dvc: bump to 1.10.2

* parametrization: print pretty interpolation errors

* Fix col/loc in exception msg

* Fix tests

.codecov.yml

@@ -1 +1,10 @@
-
+comment: false
+coverage:
+  status:
+    project:
+      default:
+        target: 85
+        threshold: 10%
+    patch:
+      default:
+        threshold: 25%

.coveragerc

@@ -1,10 +1,14 @@
 [run]
+branch = True
+source = dvc
 omit =
     *__main__.py
 
 [report]
 exclude_lines =
     if __name__ == .__main__.:
+    if typing.TYPE_CHECKING:
+    if TYPE_CHECKING:
     pass
     raise NotImplementedError
     # pragma: no cover

.github/release-drafter.yml

@@ -1,5 +1,49 @@
 # Config for https://github.com/apps/release-drafter
-branches:
-  - master
+name-template: "$RESOLVED_VERSION 🦉 "
+tag-template: "$RESOLVED_VERSION"
+categories:
+  - title: "🚀 New Features and Enhancements "
+    labels:
+      - "feature"
+      - "enhancement"
+      - "ui"
+  - title: "💣 Breaking Changes "
+    labels:
+      - "breaking-change"
+  - title: "🏇 Optimizations "
+    labels:
+      - "optimize"
+      - "performance"
+  - title: "🐛 Bug Fixes "
+    labels:
+      - "bugfix"
+      - "bug"
+  - title: "🔨 Maintenance "
+    labels:
+      - "maintenance"
+      - "refactoring"
+      - "chore"
+      - "build"
+      - "ci"
+change-template: "- $TITLE (#$NUMBER) @$AUTHOR"
+version-resolver:
+  major:
+    labels:
+      - "major"
+  minor:
+    labels:
+      - "minor"
+  patch:
+    labels:
+      - "patch"
+  default: patch
+exclude-labels:
+  - "skip-changelog"
 template: |
+  > Refer to https://dvc.org/doc/install for installation instructions.
+
+  ## Changes
+
   $CHANGES
+
+  Thanks again to $CONTRIBUTORS for the contributions! 🎉

.github/workflows/check-patch.yaml

@@ -1,19 +0,0 @@
-name: Lint Python
-
-on: [push, pull_request]
-
-jobs:
-  build:
-
-    runs-on: ubuntu-latest
-
-    steps:
-    - uses: actions/checkout@v1
-    - name: Set up Python 3.8
-      uses: actions/setup-python@v1
-      with:
-        python-version: 3.8
-    - name: Install requirements
-      run: bash ./scripts/ci/install.sh
-    - name: Check Patch
-      run: ./scripts/ci/check_patch.sh

.github/workflows/packages.yaml

@@ -0,0 +1,131 @@
+name: Build packages
+on:
+  push:
+    branches:
+    - master
+  release:
+    types: [released]
+jobs:
+  linux:
+    runs-on: ubuntu-16.04
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python 3.7
+      uses: actions/[email protected]
+      with:
+        python-version: 3.7
+    - name: Set up Ruby 2.4
+      uses: actions/setup-ruby@v1
+      with:
+        ruby-version: 2.4
+    - name: Build
+      run: ./scripts/build_posix.sh
+    - name: Publish deb
+      if: github.event_name == 'release'
+      uses: actions/[email protected]
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      with:
+        upload_url: ${{ github.event.release.upload_url }}
+        asset_path: dvc_${{ github.event.release.tag_name }}_amd64.deb
+        asset_name: dvc_${{ github.event.release.tag_name }}_amd64.deb
+        asset_content_type: binary/octet-stream
+    - name: Publish rpm
+      if: github.event_name == 'release'
+      uses: actions/[email protected]
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      with:
+        upload_url: ${{ github.event.release.upload_url }}
+        asset_path: dvc-${{ github.event.release.tag_name }}-1.x86_64.rpm
+        asset_name: dvc-${{ github.event.release.tag_name }}-1.x86_64.rpm
+        asset_content_type: binary/octet-stream
+  mac:
+    runs-on: macos-10.15
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python 3.7
+      uses: actions/[email protected]
+      with:
+        python-version: 3.7
+    - name: Set up Ruby 2.4
+      uses: actions/setup-ruby@v1
+      with:
+        ruby-version: 2.4
+    - name: Build
+      run: ./scripts/build_posix.sh
+    - name: Publish
+      if: github.event_name == 'release'
+      uses: actions/[email protected]
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      with:
+        upload_url: ${{ github.event.release.upload_url }}
+        asset_path: dvc-${{ github.event.release.tag_name }}.pkg
+        asset_name: dvc-${{ github.event.release.tag_name }}.pkg
+        asset_content_type: binary/octet-stream
+  windows:
+    runs-on: windows-2019
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python 3.7
+      uses: actions/[email protected]
+      with:
+        python-version: 3.7
+    - name: Build
+      run: scripts\build_windows.cmd
+      shell: cmd
+    - name: Publish
+      if: github.event_name == 'release'
+      env:
+        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      uses: actions/[email protected]
+      with:
+        upload_url: ${{ github.event.release.upload_url }}
+        asset_path: dvc-${{ github.event.release.tag_name }}.exe
+        asset_name: dvc-${{ github.event.release.tag_name }}.exe
+        asset_content_type: binary/octet-stream
+  python:
+    runs-on: ubuntu-16.04
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python 3.7
+      uses: actions/[email protected]
+      with:
+        python-version: 3.7
+    - name: Install
+      run: pip install ".[all,tests]"
+    - name: Build packages
+      run: ./scripts/build_package.sh
+    - name: Publish packages to PyPI
+      if: github.event_name == 'release'
+      uses: pypa/gh-action-pypi-publish@master
+      with:
+        user: __token__
+        password: ${{ secrets.PYPI_TOKEN }}
+  snap:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/[email protected]
+    - name: Build the package
+      uses: snapcore/action-build@v1
+      id: snapcraft
+    - name: Test the package
+      run: |
+        sudo snap install --dangerous --classic ${{ steps.snapcraft.outputs.snap }}
+        dvc version
+        sudo snap remove dvc
+    - name: Publish to the edge channel
+      uses: snapcore/action-publish@v1
+      if: github.event_name == 'push'
+      with:
+        store_login: ${{ secrets.STORE_LOGIN }}
+        snap: ${{ steps.snapcraft.outputs.snap }}
+        release: edge
+    - name: Publish to the stable channel
+      uses: snapcore/action-publish@v1
+      if: github.event_name == 'release'
+      with:
+        store_login: ${{ secrets.STORE_LOGIN }}
+        snap: ${{ steps.snapcraft.outputs.snap }}
+        release: stable

.github/workflows/tests.yaml

@@ -0,0 +1,86 @@
+name: Tests
+
+on:
+  pull_request:
+    branches:
+      - master
+  push:
+    branches:
+      - master
+
+env:
+  DVC_TEST: "true"
+  HOMEBREW_NO_AUTO_UPDATE: 1
+  SHELL: /bin/bash
+  GDRIVE_CREDENTIALS_DATA: ${{ secrets.GDRIVE_CREDENTIALS_DATA }}
+  AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
+  AWS_SECRET_ACCESS_KEY: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
+  GS_CREDS_JSON: ${{ secrets.GS_CREDS_JSON }}
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python 3.8
+      uses: actions/[email protected]
+      with:
+        python-version: 3.8
+    - name: Install requirements
+      run: pip install ".[all,tests]" Pygments collective.checkdocs pre-commit
+    - name: Check README
+      run: python setup.py checkdocs
+    - name: Check formatting
+      run: pre-commit run --all-files
+  tests:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-18.04, windows-2019, macos-10.15]
+        pyv: ["3.6", "3.7", "3.8", "3.9"]
+        include:
+          - os: "macos-11.0"
+            pyv: "3.9"
+        exclude:
+          - os: macos-10.15
+            pyv: "3.9"
+
+    steps:
+    - uses: actions/[email protected]
+    - name: Set up Python
+      uses: actions/[email protected]
+      with:
+        python-version: ${{ matrix.pyv }}
+    - name: install
+      run: |
+        pip install wheel
+        pip install -e ".[all,tests]"
+    - name: setup git
+      run: |
+        git config --global user.email "[email protected]"
+        git config --global user.name "DVC Tester"
+    - name: setup gs creds
+      shell: bash
+      if: env.GS_CREDS_JSON != ''
+      run: |
+        mkdir -p scripts/ci
+        echo "$GS_CREDS_JSON" > scripts/ci/gcp-creds.json
+    - name: run tests
+      run: python -m tests --all --cov-report=xml --cov-report=term --tap-combined
+    - name: upload coverage report
+      uses: codecov/[email protected]
+      with:
+        file: ./coverage.xml
+        fail_ci_if_error: false
+    - name: Post TAP
+      # Running custom fork with an option, that does not fail CI on error
+      uses: skshetry/flaky-service/packages/action@master
+      if: always() && github.ref == 'refs/heads/master'
+      with:
+        file-path: ${{github.workspace}}/testresults.tap
+        repo-token: ${{ secrets.GITHUB_TOKEN }}
+        os: ${{runner.os}}
+        matrix: ${{toJson(matrix)}}
+        repo-description: ${{github.event.repository.description}}
+        fail_ci_if_error: false

.gitignore

@@ -1,12 +1,11 @@
 neatlynx.conf
 neatlynx/__pycache__
 .idea
-cache
+/cache
 *.pyc
 .env/
 .env2.7/
 
-cache
 .dvc.conf.lock
 .DS_Store
 build/

.pre-commit-config.yaml

@@ -4,15 +4,11 @@ repos:
         language_version: python3
     repo: https://github.com/ambv/black
     rev: 19.10b0
-  - repo: https://github.com/asottile/seed-isort-config
-    rev: v2.1.1
-    hooks:
-      - id: seed-isort-config
   - hooks:
       - id: isort
         language_version: python3
     repo: https://github.com/timothycrosley/isort
-    rev: 4.3.21
+    rev: 5.4.2
   - hooks:
       - id: flake8
         language_version: python3
@@ -22,7 +18,7 @@ repos:
           - flake8-debugger
           - flake8-string-format
     repo: https://gitlab.com/pycqa/flake8
-    rev: 3.7.9
+    rev: 3.8.3
   - repo: local
     hooks:
     - id: pylint

.pylintrc

@@ -265,7 +265,7 @@ contextmanager-decorators=contextlib.contextmanager
 # List of members which are set dynamically and missed by pylint inference
 # system, and so shouldn't trigger E1101 when accessed. Python regular
 # expressions are accepted.
-generated-members=pytest.lazy_fixture,logging.TRACE,logger.trace,sys.getwindowsversion
+generated-members=hdfs.*,pytest.lazy_fixture,logging.TRACE,logger.trace,sys.getwindowsversion
 
 # Tells whether missing members accessed in mixin class should be ignored. A
 # mixin class is detected if its name ends with "mixin" (case insensitive).

.restyled.yaml

@@ -1,20 +0,0 @@
----
-restylers:
-  - name: black
-    image: restyled/restyler-black:v19.10b0
-    command:
-    - black
-    arguments: []
-    include:
-    - "**/*.py"
-    interpreters:
-    - python
-  - name: isort
-    image: restyled/restyler-isort:v4.3.21
-    command:
-    - isort
-    arguments: []
-    include:
-    - "**/*.py"
-    interpreters:
-    - python

.travis.yml

@@ -1,162 +0,0 @@
-cache:
-  directories:
-    - $HOME/.cache/pip
-    - $HOME/.pyenv
-dist: xenial
-env:
-  global:
-    - HOMEBREW_NO_AUTO_UPDATE=1
-    - PYENV_ROOT="$HOME/.pyenv"
-    - PATH="$PYENV_ROOT/bin:$PATH"
-    - PIP_CACHE_DIR="$HOME/.cache/pip"  # unify pip cache location for all platforms
-stages:
-  - check
-  - test
-  - build
-# Travis doesn't support testing python on Mac yet, so we need
-# to workaround it by installing it directly with brew.
-#
-# Note that we specify language as ruby-2.4.3 to help us satisfy
-# fpm dependencies, as travis runs el capitan with an old ruby
-# version, that causes fpm to crash.
-#
-# Each job will use the previous job's `stage` tag if not specified.
-jobs:
-  include:
-  # patch formatting checks
-  - name: "Check patch formatting"
-    stage: check
-    os: linux
-    language: python
-    python: 3.7
-    before_install: bash ./scripts/ci/install.sh
-    install:
-    script: ./scripts/ci/check_patch.sh
-  # test jobs
-  - name: "3.7 on Windows"
-    stage: test
-    os: windows           # Windows 10.0.17134 N/A Build 17134
-    language: shell       # 'language: python' is an error on Travis CI Windows
-    env:
-      - PYTHON_VERSION=3.7.5
-      - PATH=/c/Python37:/c/Python37/Scripts:$PATH
-  - name: "3.8 on Windows"
-    os: windows           # Windows 10.0.17134 N/A Build 17134
-    language: shell       # 'language: python' is an error on Travis CI Windows
-    env:
-      - PYTHON_VERSION=3.8.0
-      - PATH=/c/Python38:/c/Python38/Scripts:$PATH
-  - name: "3.6 on Linux"
-    language: python
-    python: 3.6
-  - name: "3.7 on Linux"
-    language: python
-    python: 3.7
-  - name: "3.8 on Linux"
-    language: python
-    python: 3.8
-  - name: "3.7 on Mac"
-    os: osx
-    osx_image: xcode11.3
-  # build jobs
-  - name: "Mac pkg"
-    stage: build
-    os: osx
-    osx_image: xcode9.4
-    language: ruby
-    rvm: 2.4.3
-    install:
-    script: ./scripts/build_posix.sh
-  - name: "Windows pkg"
-    os: windows
-    language: shell
-    install:
-    script: ./scripts/build_windows.cmd
-    env:
-      - PYTHON_VERSION=3.7.5
-      - PATH=/c/Python37:/c/Python37/Scripts:$PATH
-  - name: "Linux pkgs"
-    language: python
-    python: 3.7
-    before_install:
-    install:
-    script: ./scripts/build_posix.sh
-  - name: "PyPI pkgs"
-    language: python
-    python: 3.8
-    script: ./scripts/build_package.sh
-  - name: Snapcraft snap
-    language: python
-    python: 3.8
-    addons:
-     snaps:
-     - name: snapcraft
-       channel: stable
-       confinement: classic
-     - name: review-tools
-       channel: stable
-     - name: lxd
-       channel: stable
-    env:
-    - SNAPCRAFT_IMAGE_INFO: |
-        '{"build_url": "$TRAVIS_JOB_WEB_URL"}'
-    - SNAPCRAFT_BUILD_ENVIRONMENT: lxd
-    - SNAPCRAFT_BUILD_INFO: 1  # https://snapcraft.io/blog/introducing-developer-notifications-for-snap-security-updates
-    install:
-    - sudo usermod --append --groups lxd $USER
-    - sudo /snap/bin/lxd.migrate -yes
-    - sudo /snap/bin/lxd waitready
-    - sudo /snap/bin/lxd init --auto
-    script:
-    - ./scripts/build_snap.sh
-    after_failure:
-    - sudo journalctl -u snapd
-
-before_install:
-  - bash ./scripts/ci/before_install.sh
-  - source env.sh
-  - bash ./scripts/ci/check_python.sh
-install:
-  - bash ./scripts/ci/install.sh
-script:
-  - bash ./scripts/ci/script.sh
-deploy:
-  - provider: releases
-    api_key:
-      secure: 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
-    file_glob: true
-    file:
-      - dvc*.rpm
-      - dvc*.deb
-      - dvc*.pkg
-      - dvc*.exe
-    skip_cleanup: true
-    on:
-      condition: "$(./scripts/ci/deploy_condition.sh dvc*.rpm dvc*.deb dvc*.pkg dvc*.exe)"
-      tags: true
-      repo: iterative/dvc
-      stage: build
-  - provider: pypi
-    user:
-      secure: "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"
-    password:
-      secure: "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"
-    skip_existing: true
-    file_glob: true
-    file:
-      - dist/dvc*.whl
-      - dist/dvc*.tar.gz
-    skip_cleanup: true
-    on:
-      condition: "$(./scripts/ci/deploy_condition.sh dist/dvc*.whl dist/dvc*.tar.gz)"
-      tags: true
-      repo: iterative/dvc
-      stage: build
-  - provider: script
-    skip_cleanup: true
-    script: "(echo $SNAP_TOKEN | snapcraft login --with -) && timeout 600 snapcraft push dvc_*.snap --release $SNAP_CHANNEL || echo timed out"
-    on:
-      all_branches: true
-      condition: "$(./scripts/ci/deploy_condition.sh dvc_*.snap) && ($TRAVIS_BRANCH = master || -n $TRAVIS_TAG) && -n $SNAP_CHANNEL"
-      repo: iterative/dvc
-      stage: build

README.rst

@@ -262,9 +262,9 @@ Iterative, *DVC: Data Version Control - Git for Data & Models* (2020)
    :target: https://travis-ci.com/iterative/dvc/branches
    :alt: Release
 
-.. |CI| image:: https://img.shields.io/travis/com/iterative/dvc/master?label=dev&logo=travis
-   :target: https://travis-ci.com/iterative/dvc/builds
-   :alt: Travis dev branch
+.. |CI| image:: https://github.com/iterative/dvc/workflows/Tests/badge.svg?branch=master
+   :target: https://github.com/iterative/dvc/actions
+   :alt: GHA Tests
 
 .. |Maintainability| image:: https://codeclimate.com/github/iterative/dvc/badges/gpa.svg
    :target: https://codeclimate.com/github/iterative/dvc

dvc/api.py

@@ -2,35 +2,40 @@ import os
 from contextlib import _GeneratorContextManager as GCM
 from contextlib import contextmanager
 
-from dvc.exceptions import DvcException, NotDvcRepoError
+from funcy import reraise
+
+from dvc.exceptions import (
+    NotDvcRepoError,
+    OutputNotFoundError,
+    PathMissingError,
+)
 from dvc.external_repo import external_repo
+from dvc.path_info import PathInfo
 from dvc.repo import Repo
 
 
-class UrlNotDvcRepoError(DvcException):
-    """Thrown if the given URL is not a DVC repository."""
-
-    def __init__(self, url):
-        super().__init__(f"'{url}' is not a DVC repository.")
-
-
 def get_url(path, repo=None, rev=None, remote=None):
     """
     Returns the URL to the storage location of a data file or directory tracked
     in a DVC repo. For Git repos, HEAD is used unless a rev argument is
     supplied. The default remote is tried unless a remote argument is supplied.
 
-    Raises UrlNotDvcRepoError if repo is not a DVC project.
+    Raises OutputNotFoundError if the file is not tracked by DVC.
 
     NOTE: This function does not check for the actual existence of the file or
     directory in the remote storage.
     """
     with _make_repo(repo, rev=rev) as _repo:
-        if not isinstance(_repo, Repo):
-            raise UrlNotDvcRepoError(_repo.url)  # pylint: disable=no-member
-        out = _repo.find_out_by_relpath(path)
-        remote_obj = _repo.cloud.get_remote(remote)
-        return str(remote_obj.tree.hash_to_path_info(out.checksum))
+        path_info = PathInfo(_repo.root_dir) / path
+        with reraise(FileNotFoundError, PathMissingError(path, repo)):
+            metadata = _repo.repo_tree.metadata(path_info)
+
+        if not metadata.is_dvc:
+            raise OutputNotFoundError(path, repo)
+
+        cloud = metadata.repo.cloud
+        hash_info = _repo.repo_tree.get_hash(path_info)
+        return cloud.get_url_for(remote, checksum=hash_info.value)
 
 
 def open(  # noqa, pylint: disable=redefined-builtin
@@ -97,9 +102,41 @@ def _make_repo(repo_url=None, rev=None):
     repo_url = repo_url or os.getcwd()
     if rev is None and os.path.exists(repo_url):
         try:
-            yield Repo(repo_url)
+            yield Repo(repo_url, subrepos=True)
             return
         except NotDvcRepoError:
             pass  # fallthrough to external_repo
     with external_repo(url=repo_url, rev=rev) as repo:
         yield repo
+
+
+def make_checkpoint():
+    """
+    Signal DVC to create a checkpoint experiment.
+
+    If the current process is being run from DVC, this function will block
+    until DVC has finished creating the checkpoint. Otherwise, this function
+    will return immediately.
+    """
+    import builtins
+    from time import sleep
+
+    from dvc.env import DVC_CHECKPOINT, DVC_ROOT
+    from dvc.stage.run import CHECKPOINT_SIGNAL_FILE
+
+    if os.getenv(DVC_CHECKPOINT) is None:
+        return
+
+    root_dir = os.getenv(DVC_ROOT, Repo.find_root())
+    signal_file = os.path.join(
+        root_dir, Repo.DVC_DIR, "tmp", CHECKPOINT_SIGNAL_FILE
+    )
+
+    with builtins.open(signal_file, "w") as fobj:
+        # NOTE: force flushing/writing empty file to disk, otherwise when
+        # run in certain contexts (pytest) file may not actually be written
+        fobj.write("")
+        fobj.flush()
+        os.fsync(fobj.fileno())
+    while os.path.exists(signal_file):
+        sleep(1)

dvc/cache/__init__.py

@@ -1,45 +1,31 @@
 """Manages cache of a DVC repo."""
 from collections import defaultdict
 
-from funcy import cached_property
+from ..scheme import Schemes
 
 
-def _make_remote_property(name):
-    """
-    The config file is stored in a way that allows you to have a
-    cache for each remote.
+def get_cloud_cache(tree):
+    from .base import CloudCache
+    from .local import LocalCache
+    from .ssh import SSHCache
 
-    This is needed when specifying external outputs
-    (as they require you to have an external cache location).
+    if tree.scheme == Schemes.LOCAL:
+        return LocalCache(tree)
 
-    Imagine a config file like the following:
+    if tree.scheme == Schemes.SSH:
+        return SSHCache(tree)
 
-            ['remote "dvc-storage"']
-            url = ssh://localhost/tmp
-            ask_password = true
+    return CloudCache(tree)
 
-            [cache]
-            ssh = dvc-storage
 
-    This method creates a cached property, containing cache named `name`:
+def _get_cache(repo, settings):
+    from ..tree import get_cloud_tree
 
-        self.config == {'ssh': 'dvc-storage'}
-        self.ssh  # a RemoteSSH instance
-    """
+    if not settings:
+        return None
 
-    def getter(self):
-        from ..tree import get_cloud_tree
-        from .base import CloudCache
-
-        remote = self.config.get(name)
-        if not remote:
-            return None
-
-        tree = get_cloud_tree(self.repo, name=remote)
-        return CloudCache(tree)
-
-    getter.__name__ = name
-    return cached_property(getter)
+    tree = get_cloud_tree(repo, **settings)
+    return get_cloud_cache(tree)
 
 
 class Cache:
@@ -50,29 +36,48 @@ class Cache:
     """
 
     CACHE_DIR = "cache"
+    CLOUD_SCHEMES = [
+        Schemes.S3,
+        Schemes.GS,
+        Schemes.SSH,
+        Schemes.HDFS,
+        Schemes.WEBHDFS,
+    ]
 
     def __init__(self, repo):
-        from ..tree import get_cloud_tree
-        from .local import LocalCache
-
         self.repo = repo
         self.config = config = repo.config["cache"]
+        self._cache = {}
 
         local = config.get("local")
 
         if local:
             settings = {"name": local}
+        elif "dir" not in config:
+            settings = None
         else:
-            settings = {**config, "url": config["dir"]}
+            from ..config import LOCAL_COMMON
 
-        tree = get_cloud_tree(repo, **settings)
-        self.local = LocalCache(tree)
+            settings = {"url": config["dir"]}
+            for opt in LOCAL_COMMON.keys():
+                if opt in config:
+                    settings[str(opt)] = config.get(opt)
 
-    s3 = _make_remote_property("s3")
-    gs = _make_remote_property("gs")
-    ssh = _make_remote_property("ssh")
-    hdfs = _make_remote_property("hdfs")
-    azure = _make_remote_property("azure")
+        self._cache[Schemes.LOCAL] = _get_cache(repo, settings)
+
+        for scheme in self.CLOUD_SCHEMES:
+            remote = self.config.get(scheme)
+            settings = {"name": remote} if remote else None
+            self._cache[scheme] = _get_cache(repo, settings)
+
+    def __getattr__(self, name):
+        try:
+            return self._cache[name]
+        except KeyError as exc:
+            raise AttributeError from exc
+
+    def by_scheme(self):
+        yield from self._cache.items()
 
 
 class NamedCacheItem:

dvc/cache/base.py

@@ -1,32 +1,30 @@
+import itertools
 import json
 import logging
+from concurrent.futures import ThreadPoolExecutor
 from copy import copy
+from typing import Optional
 
+from funcy import decorator
 from shortuuid import uuid
 
 import dvc.prompt as prompt
-from dvc.exceptions import CheckoutError, ConfirmRemoveError, DvcException
-from dvc.path_info import WindowsPathInfo
+from dvc.dir_info import DirInfo
+from dvc.exceptions import (
+    CacheLinkError,
+    CheckoutError,
+    ConfirmRemoveError,
+    DvcException,
+)
 from dvc.progress import Tqdm
-from dvc.remote.slow_link_detection import slow_link_guard
+from dvc.remote.slow_link_detection import (  # type: ignore[attr-defined]
+    slow_link_guard,
+)
 
 from ..tree.base import RemoteActionNotImplemented
 
 logger = logging.getLogger(__name__)
 
-STATUS_OK = 1
-STATUS_MISSING = 2
-STATUS_NEW = 3
-STATUS_DELETED = 4
-
-STATUS_MAP = {
-    # (local_exists, remote_exists)
-    (True, True): STATUS_OK,
-    (False, False): STATUS_MISSING,
-    (True, False): STATUS_NEW,
-    (False, True): STATUS_DELETED,
-}
-
 
 class DirCacheError(DvcException):
     def __init__(self, hash_):
@@ -35,11 +33,18 @@ class DirCacheError(DvcException):
         )
 
 
+@decorator
+def use_state(call):
+    tree = call._args[0].tree  # pylint: disable=protected-access
+    with tree.state:
+        return call()
+
+
 class CloudCache:
     """Cloud cache class."""
 
     DEFAULT_CACHE_TYPES = ["copy"]
-    CACHE_MODE = None
+    CACHE_MODE: Optional[int] = None
 
     def __init__(self, tree):
         self.tree = tree
@@ -51,47 +56,38 @@ class CloudCache:
         self.cache_type_confirmed = False
         self._dir_info = {}
 
-    def get_dir_cache(self, hash_):
-        assert hash_
+    def get_dir_cache(self, hash_info):
+        assert hash_info
 
-        dir_info = self._dir_info.get(hash_)
+        dir_info = self._dir_info.get(hash_info.value)
         if dir_info:
             return dir_info
 
         try:
-            dir_info = self.load_dir_cache(hash_)
+            dir_info = self.load_dir_cache(hash_info)
         except DirCacheError:
-            dir_info = []
+            dir_info = DirInfo()
 
-        self._dir_info[hash_] = dir_info
+        self._dir_info[hash_info.value] = dir_info
         return dir_info
 
-    def load_dir_cache(self, hash_):
-        path_info = self.tree.hash_to_path_info(hash_)
+    def load_dir_cache(self, hash_info):
+        path_info = self.tree.hash_to_path_info(hash_info.value)
 
         try:
             with self.tree.open(path_info, "r") as fobj:
                 d = json.load(fobj)
         except (ValueError, FileNotFoundError) as exc:
-            raise DirCacheError(hash_) from exc
+            raise DirCacheError(hash_info) from exc
 
         if not isinstance(d, list):
             logger.error(
                 "dir cache file format error '%s' [skipping the file]",
                 path_info,
             )
-            return []
+            d = []
 
-        if self.tree.PATH_CLS == WindowsPathInfo:
-            # only need to convert it for Windows
-            for info in d:
-                # NOTE: here is a BUG, see comment to .as_posix() below
-                relpath = info[self.tree.PARAM_RELPATH]
-                info[self.tree.PARAM_RELPATH] = relpath.replace(
-                    "/", self.tree.PATH_CLS.sep
-                )
-
-        return d
+        return DirInfo.from_list(d)
 
     def changed(self, path_info, hash_info):
         """Checks if data has changed.
@@ -118,20 +114,21 @@ class CloudCache:
             logger.debug("'%s' doesn't exist.", path_info)
             return True
 
-        hash_ = hash_info.get(self.tree.PARAM_CHECKSUM)
-        if hash_ is None:
+        if not hash_info:
             logger.debug("hash value for '%s' is missing.", path_info)
             return True
 
-        if self.changed_cache(hash_):
-            logger.debug("cache for '%s'('%s') has changed.", path_info, hash_)
+        if self.changed_cache(hash_info):
+            logger.debug(
+                "cache for '%s'('%s') has changed.", path_info, hash_info
+            )
             return True
 
         actual = self.tree.get_hash(path_info)
-        if hash_ != actual:
+        if hash_info != actual:
             logger.debug(
                 "hash value '%s' for '%s' has changed (actual '%s').",
-                hash_,
+                hash_info,
                 actual,
                 path_info,
             )
@@ -176,7 +173,7 @@ class CloudCache:
                 )
                 del link_types[0]
 
-        raise DvcException("no possible cache types left to try out.")
+        raise CacheLinkError([to_info])
 
     def _do_link(self, from_info, to_info, link_method):
         if self.tree.exists(to_info):
@@ -188,12 +185,12 @@ class CloudCache:
             "Created '%s': %s -> %s", self.cache_types[0], from_info, to_info,
         )
 
-    def _save_file(self, path_info, tree, hash_, save_link=True, **kwargs):
-        assert hash_
+    def _save_file(self, path_info, tree, hash_info, save_link=True, **kwargs):
+        assert hash_info
 
-        cache_info = self.tree.hash_to_path_info(hash_)
+        cache_info = self.tree.hash_to_path_info(hash_info.value)
         if tree == self.tree:
-            if self.changed_cache(hash_):
+            if self.changed_cache(hash_info):
                 self.tree.move(path_info, cache_info, mode=self.CACHE_MODE)
                 self.link(cache_info, path_info)
             elif self.tree.iscopy(path_info) and self._cache_is_copy(
@@ -210,9 +207,9 @@ class CloudCache:
             # we need to update path and cache, since in case of reflink,
             # or copy cache type moving original file results in updates on
             # next executed command, which causes md5 recalculation
-            self.tree.state.save(path_info, hash_)
+            self.tree.state.save(path_info, hash_info)
         else:
-            if self.changed_cache(hash_):
+            if self.changed_cache(hash_info):
                 with tree.open(path_info, mode="rb") as fobj:
                     # if tree has fetch enabled, DVC out will be fetched on
                     # open and we do not need to read/copy any data
@@ -224,8 +221,7 @@ class CloudCache:
                 if callback:
                     callback(1)
 
-        self.tree.state.save(cache_info, hash_)
-        return {self.tree.PARAM_CHECKSUM: hash_}
+        self.tree.state.save(cache_info, hash_info)
 
     def _cache_is_copy(self, path_info):
         """Checks whether cache uses copies."""
@@ -249,13 +245,55 @@ class CloudCache:
         self.cache_type_confirmed = True
         return self.cache_types[0] == "copy"
 
-    def _save_dir(self, path_info, tree, hash_, save_link=True, **kwargs):
-        dir_info = self.get_dir_cache(hash_)
-        for entry in Tqdm(
-            dir_info, desc="Saving " + path_info.name, unit="file"
+    def _get_dir_info_hash(self, dir_info):
+        import tempfile
+
+        from dvc.path_info import PathInfo
+        from dvc.utils import tmp_fname
+
+        tmp = tempfile.NamedTemporaryFile(delete=False).name
+        with open(tmp, "w+") as fobj:
+            json.dump(dir_info.to_list(), fobj, sort_keys=True)
+
+        from_info = PathInfo(tmp)
+        to_info = self.tree.path_info / tmp_fname("")
+        self.tree.upload(from_info, to_info, no_progress_bar=True)
+
+        hash_info = self.tree.get_file_hash(to_info)
+        hash_info.value += self.tree.CHECKSUM_DIR_SUFFIX
+        hash_info.dir_info = dir_info
+        hash_info.nfiles = dir_info.nfiles
+
+        return hash_info, to_info
+
+    @use_state
+    def save_dir_info(self, dir_info, hash_info=None):
+        if (
+            hash_info
+            and hash_info.name == self.tree.PARAM_CHECKSUM
+            and not self.changed_cache_file(hash_info)
+        ):
+            return hash_info
+
+        hi, tmp_info = self._get_dir_info_hash(dir_info)
+        new_info = self.tree.hash_to_path_info(hi.value)
+        if self.changed_cache_file(hi):
+            self.tree.makedirs(new_info.parent)
+            self.tree.move(tmp_info, new_info, mode=self.CACHE_MODE)
+
+        self.tree.state.save(new_info, hi)
+
+        return hi
+
+    def _save_dir(self, path_info, tree, hash_info, save_link=True, **kwargs):
+        if not hash_info.dir_info:
+            hash_info.dir_info = tree.cache.get_dir_cache(hash_info)
+        hi = self.save_dir_info(hash_info.dir_info, hash_info)
+        for entry_info, entry_hash in Tqdm(
+            hi.dir_info.items(path_info),
+            desc="Saving " + path_info.name,
+            unit="file",
         ):
-            entry_info = path_info / entry[self.tree.PARAM_RELPATH]
-            entry_hash = entry[self.tree.PARAM_CHECKSUM]
             self._save_file(
                 entry_info, tree, entry_hash, save_link=False, **kwargs
             )
@@ -263,12 +301,12 @@ class CloudCache:
         if save_link:
             self.tree.state.save_link(path_info)
         if self.tree.exists(path_info):
-            self.tree.state.save(path_info, hash_)
+            self.tree.state.save(path_info, hi)
 
-        cache_info = self.tree.hash_to_path_info(hash_)
-        self.tree.state.save(cache_info, hash_)
-        return {self.tree.PARAM_CHECKSUM: hash_}
+        cache_info = self.tree.hash_to_path_info(hi.value)
+        self.tree.state.save(cache_info, hi)
 
+    @use_state
     def save(self, path_info, tree, hash_info, save_link=True, **kwargs):
         if path_info.scheme != self.tree.scheme:
             raise RemoteActionNotImplemented(
@@ -276,25 +314,23 @@ class CloudCache:
                 self.tree.scheme,
             )
 
-        if not hash_info:
-            hash_info = self.tree.save_info(path_info, tree=tree, **kwargs)
-        hash_ = hash_info[self.tree.PARAM_CHECKSUM]
-        return self._save(path_info, tree, hash_, save_link, **kwargs)
+        self._save(path_info, tree, hash_info, save_link, **kwargs)
 
-    def _save(self, path_info, tree, hash_, save_link=True, **kwargs):
-        to_info = self.tree.hash_to_path_info(hash_)
+    def _save(self, path_info, tree, hash_info, save_link=True, **kwargs):
+        to_info = self.tree.hash_to_path_info(hash_info.value)
         logger.debug("Saving '%s' to '%s'.", path_info, to_info)
 
         if tree.isdir(path_info):
-            return self._save_dir(path_info, tree, hash_, save_link, **kwargs)
-        return self._save_file(path_info, tree, hash_, save_link, **kwargs)
+            self._save_dir(path_info, tree, hash_info, save_link, **kwargs)
+        else:
+            self._save_file(path_info, tree, hash_info, save_link, **kwargs)
 
     # Override to return path as a string instead of PathInfo for clouds
     # which support string paths (see local)
     def hash_to_path(self, hash_):
         return self.tree.hash_to_path_info(hash_)
 
-    def changed_cache_file(self, hash_):
+    def changed_cache_file(self, hash_info):
         """Compare the given hash with the (corresponding) actual one.
 
         - Use `State` as a cache for computed hashes
@@ -307,7 +343,7 @@ class CloudCache:
         - Remove the file from cache if it doesn't match the actual hash
         """
         # Prefer string path over PathInfo when possible due to performance
-        cache_info = self.hash_to_path(hash_)
+        cache_info = self.hash_to_path(hash_info.value)
         if self.tree.is_protected(cache_info):
             logger.debug(
                 "Assuming '%s' is unchanged since it is read-only", cache_info
@@ -317,13 +353,16 @@ class CloudCache:
         actual = self.tree.get_hash(cache_info)
 
         logger.debug(
-            "cache '%s' expected '%s' actual '%s'", cache_info, hash_, actual,
+            "cache '%s' expected '%s' actual '%s'",
+            cache_info,
+            hash_info,
+            actual,
         )
 
-        if not hash_ or not actual:
+        if not hash_info or not actual:
             return True
 
-        if actual.split(".")[0] == hash_.split(".")[0]:
+        if actual.value.split(".")[0] == hash_info.value.split(".")[0]:
             # making cache file read-only so we don't need to check it
             # next time
             self.tree.protect(cache_info)
@@ -335,15 +374,13 @@ class CloudCache:
 
         return True
 
-    def _changed_dir_cache(self, hash_, path_info=None, filter_info=None):
-        if self.changed_cache_file(hash_):
+    def _changed_dir_cache(self, hash_info, path_info=None, filter_info=None):
+        if self.changed_cache_file(hash_info):
             return True
 
-        for entry in self.get_dir_cache(hash_):
-            entry_hash = entry[self.tree.PARAM_CHECKSUM]
-
+        dir_info = self.get_dir_cache(hash_info)
+        for entry_info, entry_hash in dir_info.items(path_info):
             if path_info and filter_info:
-                entry_info = path_info / entry[self.tree.PARAM_RELPATH]
                 if not entry_info.isin_or_eq(filter_info):
                     continue
 
@@ -352,12 +389,12 @@ class CloudCache:
 
         return False
 
-    def changed_cache(self, hash_, path_info=None, filter_info=None):
-        if self.tree.is_dir_hash(hash_):
+    def changed_cache(self, hash_info, path_info=None, filter_info=None):
+        if hash_info.isdir:
             return self._changed_dir_cache(
-                hash_, path_info=path_info, filter_info=filter_info
+                hash_info, path_info=path_info, filter_info=filter_info
             )
-        return self.changed_cache_file(hash_)
+        return self.changed_cache_file(hash_info)
 
     def already_cached(self, path_info):
         current = self.tree.get_hash(path_info)
@@ -383,11 +420,11 @@ class CloudCache:
         self.tree.remove(path_info)
 
     def _checkout_file(
-        self, path_info, hash_, force, progress_callback=None, relink=False
+        self, path_info, hash_info, force, progress_callback=None, relink=False
     ):
         """The file is changed we need to checkout a new copy"""
         added, modified = True, False
-        cache_info = self.tree.hash_to_path_info(hash_)
+        cache_info = self.tree.hash_to_path_info(hash_info.value)
         if self.tree.exists(path_info):
             logger.debug("data '%s' will be replaced.", path_info)
             self.safe_remove(path_info, force=force)
@@ -395,7 +432,7 @@ class CloudCache:
 
         self.link(cache_info, path_info)
         self.tree.state.save_link(path_info)
-        self.tree.state.save(path_info, hash_)
+        self.tree.state.save(path_info, hash_info)
         if progress_callback:
             progress_callback(str(path_info))
 
@@ -404,7 +441,7 @@ class CloudCache:
     def _checkout_dir(
         self,
         path_info,
-        hash_,
+        hash_info,
         force,
         progress_callback=None,
         relink=False,
@@ -417,25 +454,23 @@ class CloudCache:
             added = True
             self.tree.makedirs(path_info)
 
-        dir_info = self.get_dir_cache(hash_)
+        dir_info = self.get_dir_cache(hash_info)
 
         logger.debug("Linking directory '%s'.", path_info)
 
-        for entry in dir_info:
-            relative_path = entry[self.tree.PARAM_RELPATH]
-            entry_hash = entry[self.tree.PARAM_CHECKSUM]
-            entry_cache_info = self.tree.hash_to_path_info(entry_hash)
-            entry_info = path_info / relative_path
+        for entry_info, entry_hash_info in dir_info.items(path_info):
+            entry_cache_info = self.tree.hash_to_path_info(
+                entry_hash_info.value
+            )
 
             if filter_info and not entry_info.isin_or_eq(filter_info):
                 continue
 
-            entry_hash_info = {self.tree.PARAM_CHECKSUM: entry_hash}
             if relink or self.changed(entry_info, entry_hash_info):
                 modified = True
                 self.safe_remove(entry_info, force=force)
                 self.link(entry_cache_info, entry_info)
-                self.tree.state.save(entry_info, entry_hash)
+                self.tree.state.save(entry_info, entry_hash_info)
             if progress_callback:
                 progress_callback(str(entry_info))
 
@@ -445,7 +480,7 @@ class CloudCache:
         )
 
         self.tree.state.save_link(path_info)
-        self.tree.state.save(path_info, hash_)
+        self.tree.state.save(path_info, hash_info)
 
         # relink is not modified, assume it as nochange
         return added, not added and modified and not relink
@@ -453,15 +488,14 @@ class CloudCache:
     def _remove_redundant_files(self, path_info, dir_info, force):
         existing_files = set(self.tree.walk_files(path_info))
 
-        needed_files = {
-            path_info / entry[self.tree.PARAM_RELPATH] for entry in dir_info
-        }
+        needed_files = {info for info, _ in dir_info.items(path_info)}
         redundant_files = existing_files - needed_files
         for path in redundant_files:
             self.safe_remove(path, force)
 
         return bool(redundant_files)
 
+    @use_state
     def checkout(
         self,
         path_info,
@@ -470,18 +504,19 @@ class CloudCache:
         progress_callback=None,
         relink=False,
         filter_info=None,
+        quiet=False,
     ):
         if path_info.scheme not in ["local", self.tree.scheme]:
             raise NotImplementedError
 
-        hash_ = hash_info.get(self.tree.PARAM_CHECKSUM)
         failed = None
         skip = False
-        if not hash_:
-            logger.warning(
-                "No file hash info found for '%s'. " "It won't be created.",
-                path_info,
-            )
+        if not hash_info:
+            if not quiet:
+                logger.warning(
+                    "No file hash info found for '%s'. It won't be created.",
+                    path_info,
+                )
             self.safe_remove(path_info, force=force)
             failed = path_info
 
@@ -490,13 +525,14 @@ class CloudCache:
             skip = True
 
         elif self.changed_cache(
-            hash_, path_info=path_info, filter_info=filter_info
+            hash_info, path_info=path_info, filter_info=filter_info
         ):
-            logger.warning(
-                "Cache '%s' not found. File '%s' won't be created.",
-                hash_,
-                path_info,
-            )
+            if not quiet:
+                logger.warning(
+                    "Cache '%s' not found. File '%s' won't be created.",
+                    hash_info,
+                    path_info,
+                )
             self.safe_remove(path_info, force=force)
             failed = path_info
 
@@ -505,50 +541,397 @@ class CloudCache:
                 progress_callback(
                     str(path_info),
                     self.get_files_number(
-                        self.tree.path_info, hash_, filter_info
+                        self.tree.path_info, hash_info, filter_info
                     ),
                 )
             if failed:
                 raise CheckoutError([failed])
             return
 
-        logger.debug("Checking out '%s' with cache '%s'.", path_info, hash_)
+        logger.debug(
+            "Checking out '%s' with cache '%s'.", path_info, hash_info
+        )
 
         return self._checkout(
-            path_info, hash_, force, progress_callback, relink, filter_info,
+            path_info,
+            hash_info,
+            force,
+            progress_callback,
+            relink,
+            filter_info,
         )
 
     def _checkout(
         self,
         path_info,
-        hash_,
+        hash_info,
         force=False,
         progress_callback=None,
         relink=False,
         filter_info=None,
     ):
-        if not self.tree.is_dir_hash(hash_):
+        if not hash_info.isdir:
             return self._checkout_file(
-                path_info, hash_, force, progress_callback, relink
+                path_info, hash_info, force, progress_callback, relink
             )
 
         return self._checkout_dir(
-            path_info, hash_, force, progress_callback, relink, filter_info
+            path_info, hash_info, force, progress_callback, relink, filter_info
         )
 
-    def get_files_number(self, path_info, hash_, filter_info):
+    def get_files_number(self, path_info, hash_info, filter_info):
         from funcy.py3 import ilen
 
-        if not hash_:
+        if not hash_info:
             return 0
 
-        if not self.tree.is_dir_hash(hash_):
+        if not hash_info.isdir:
             return 1
 
         if not filter_info:
-            return len(self.get_dir_cache(hash_))
+            return self.get_dir_cache(hash_info).nfiles
 
         return ilen(
-            filter_info.isin_or_eq(path_info / entry[self.tree.PARAM_CHECKSUM])
-            for entry in self.get_dir_cache(hash_)
+            filter_info.isin_or_eq(path_info / relpath)
+            for relpath, _ in self.get_dir_cache(hash_info).items()
         )
+
+    def _get_dir_size(self, dir_info):
+        try:
+            return sum(
+                self.tree.getsize(self.tree.hash_to_path_info(hi.value))
+                for _, hi in dir_info.items()
+            )
+        except FileNotFoundError:
+            return None
+
+    def merge(self, ancestor_info, our_info, their_info):
+        assert our_info
+        assert their_info
+
+        if ancestor_info:
+            ancestor = self.get_dir_cache(ancestor_info)
+        else:
+            ancestor = DirInfo()
+
+        our = self.get_dir_cache(our_info)
+        their = self.get_dir_cache(their_info)
+
+        merged = our.merge(ancestor, their)
+        hash_info = self.save_dir_info(merged)
+        hash_info.size = self._get_dir_size(merged)
+        return hash_info
+
+    @use_state
+    def get_hash(self, tree, path_info):
+        hash_info = tree.get_hash(path_info)
+        if not hash_info.isdir:
+            assert hash_info.name == self.tree.PARAM_CHECKSUM
+            return hash_info
+
+        hi = self.save_dir_info(hash_info.dir_info, hash_info)
+        hi.size = hash_info.size
+        return hi
+
+    def set_dir_info(self, hash_info):
+        assert hash_info.isdir
+
+        hash_info.dir_info = self.get_dir_cache(hash_info)
+        hash_info.nfiles = hash_info.dir_info.nfiles
+
+    def _list_paths(self, prefix=None, progress_callback=None):
+        if prefix:
+            if len(prefix) > 2:
+                path_info = self.tree.path_info / prefix[:2] / prefix[2:]
+            else:
+                path_info = self.tree.path_info / prefix[:2]
+            prefix = True
+        else:
+            path_info = self.tree.path_info
+            prefix = False
+        if progress_callback:
+            for file_info in self.tree.walk_files(path_info, prefix=prefix):
+                progress_callback()
+                yield file_info.path
+        else:
+            yield from self.tree.walk_files(path_info, prefix=prefix)
+
+    def _path_to_hash(self, path):
+        parts = self.tree.PATH_CLS(path).parts[-2:]
+
+        if not (len(parts) == 2 and parts[0] and len(parts[0]) == 2):
+            raise ValueError(f"Bad cache file path '{path}'")
+
+        return "".join(parts)
+
+    def list_hashes(self, prefix=None, progress_callback=None):
+        """Iterate over hashes in this tree.
+
+        If `prefix` is specified, only hashes which begin with `prefix`
+        will be returned.
+        """
+        for path in self._list_paths(prefix, progress_callback):
+            try:
+                yield self._path_to_hash(path)
+            except ValueError:
+                logger.debug(
+                    "'%s' doesn't look like a cache file, skipping", path
+                )
+
+    def _hashes_with_limit(self, limit, prefix=None, progress_callback=None):
+        count = 0
+        for hash_ in self.list_hashes(prefix, progress_callback):
+            yield hash_
+            count += 1
+            if count > limit:
+                logger.debug(
+                    "`list_hashes()` returned max '{}' hashes, "
+                    "skipping remaining results".format(limit)
+                )
+                return
+
+    def _max_estimation_size(self, hashes):
+        # Max remote size allowed for us to use traverse method
+        return max(
+            self.tree.TRAVERSE_THRESHOLD_SIZE,
+            len(hashes)
+            / self.tree.TRAVERSE_WEIGHT_MULTIPLIER
+            * self.tree.LIST_OBJECT_PAGE_SIZE,
+        )
+
+    def _estimate_remote_size(self, hashes=None, name=None):
+        """Estimate tree size based on number of entries beginning with
+        "00..." prefix.
+        """
+        prefix = "0" * self.tree.TRAVERSE_PREFIX_LEN
+        total_prefixes = pow(16, self.tree.TRAVERSE_PREFIX_LEN)
+        if hashes:
+            max_hashes = self._max_estimation_size(hashes)
+        else:
+            max_hashes = None
+
+        with Tqdm(
+            desc="Estimating size of "
+            + (f"cache in '{name}'" if name else "remote cache"),
+            unit="file",
+        ) as pbar:
+
+            def update(n=1):
+                pbar.update(n * total_prefixes)
+
+            if max_hashes:
+                hashes = self._hashes_with_limit(
+                    max_hashes / total_prefixes, prefix, update
+                )
+            else:
+                hashes = self.list_hashes(prefix, update)
+
+            remote_hashes = set(hashes)
+            if remote_hashes:
+                remote_size = total_prefixes * len(remote_hashes)
+            else:
+                remote_size = total_prefixes
+            logger.debug(f"Estimated remote size: {remote_size} files")
+        return remote_size, remote_hashes
+
+    def list_hashes_traverse(
+        self, remote_size, remote_hashes, jobs=None, name=None
+    ):
+        """Iterate over all hashes found in this tree.
+        Hashes are fetched in parallel according to prefix, except in
+        cases where the remote size is very small.
+
+        All hashes from the remote (including any from the size
+        estimation step passed via the `remote_hashes` argument) will be
+        returned.
+
+        NOTE: For large remotes the list of hashes will be very
+        big(e.g. 100M entries, md5 for each is 32 bytes, so ~3200Mb list)
+        and we don't really need all of it at the same time, so it makes
+        sense to use a generator to gradually iterate over it, without
+        keeping all of it in memory.
+        """
+        num_pages = remote_size / self.tree.LIST_OBJECT_PAGE_SIZE
+        if num_pages < 256 / self.tree.JOBS:
+            # Fetching prefixes in parallel requires at least 255 more
+            # requests, for small enough remotes it will be faster to fetch
+            # entire cache without splitting it into prefixes.
+            #
+            # NOTE: this ends up re-fetching hashes that were already
+            # fetched during remote size estimation
+            traverse_prefixes = [None]
+            initial = 0
+        else:
+            yield from remote_hashes
+            initial = len(remote_hashes)
+            traverse_prefixes = [f"{i:02x}" for i in range(1, 256)]
+            if self.tree.TRAVERSE_PREFIX_LEN > 2:
+                traverse_prefixes += [
+                    "{0:0{1}x}".format(i, self.tree.TRAVERSE_PREFIX_LEN)
+                    for i in range(
+                        1, pow(16, self.tree.TRAVERSE_PREFIX_LEN - 2)
+                    )
+                ]
+        with Tqdm(
+            desc="Querying "
+            + (f"cache in '{name}'" if name else "remote cache"),
+            total=remote_size,
+            initial=initial,
+            unit="file",
+        ) as pbar:
+
+            def list_with_update(prefix):
+                return list(
+                    self.list_hashes(
+                        prefix=prefix, progress_callback=pbar.update
+                    )
+                )
+
+            with ThreadPoolExecutor(
+                max_workers=jobs or self.tree.JOBS
+            ) as executor:
+                in_remote = executor.map(list_with_update, traverse_prefixes,)
+                yield from itertools.chain.from_iterable(in_remote)
+
+    def all(self, jobs=None, name=None):
+        """Iterate over all hashes in this tree.
+
+        Hashes will be fetched in parallel threads according to prefix
+        (except for small remotes) and a progress bar will be displayed.
+        """
+        logger.debug(
+            "Fetching all hashes from '{}'".format(
+                name if name else "remote cache"
+            )
+        )
+
+        if not self.tree.CAN_TRAVERSE:
+            return self.list_hashes()
+
+        remote_size, remote_hashes = self._estimate_remote_size(name=name)
+        return self.list_hashes_traverse(
+            remote_size, remote_hashes, jobs, name
+        )
+
+    def _remove_unpacked_dir(self, hash_):
+        pass
+
+    def gc(self, used, jobs=None):
+        removed = False
+        # hashes must be sorted to ensure we always remove .dir files first
+        for hash_ in sorted(
+            self.all(jobs, str(self.tree.path_info)),
+            key=self.tree.is_dir_hash,
+            reverse=True,
+        ):
+            if hash_ in used:
+                continue
+            path_info = self.tree.hash_to_path_info(hash_)
+            if self.tree.is_dir_hash(hash_):
+                # backward compatibility
+                # pylint: disable=protected-access
+                self._remove_unpacked_dir(hash_)
+            self.tree.remove(path_info)
+            removed = True
+
+        return removed
+
+    def list_hashes_exists(self, hashes, jobs=None, name=None):
+        """Return list of the specified hashes which exist in this tree.
+        Hashes will be queried individually.
+        """
+        logger.debug(
+            "Querying {} hashes via object_exists".format(len(hashes))
+        )
+        with Tqdm(
+            desc="Querying "
+            + ("cache in " + name if name else "remote cache"),
+            total=len(hashes),
+            unit="file",
+        ) as pbar:
+
+            def exists_with_progress(path_info):
+                ret = self.tree.exists(path_info)
+                pbar.update_msg(str(path_info))
+                return ret
+
+            with ThreadPoolExecutor(
+                max_workers=jobs or self.tree.JOBS
+            ) as executor:
+                path_infos = map(self.tree.hash_to_path_info, hashes)
+                in_remote = executor.map(exists_with_progress, path_infos)
+                ret = list(itertools.compress(hashes, in_remote))
+                return ret
+
+    def hashes_exist(self, hashes, jobs=None, name=None):
+        """Check if the given hashes are stored in the remote.
+
+        There are two ways of performing this check:
+
+        - Traverse method: Get a list of all the files in the remote
+            (traversing the cache directory) and compare it with
+            the given hashes. Cache entries will be retrieved in parallel
+            threads according to prefix (i.e. entries starting with, "00...",
+            "01...", and so on) and a progress bar will be displayed.
+
+        - Exists method: For each given hash, run the `exists`
+            method and filter the hashes that aren't on the remote.
+            This is done in parallel threads.
+            It also shows a progress bar when performing the check.
+
+        The reason for such an odd logic is that most of the remotes
+        take much shorter time to just retrieve everything they have under
+        a certain prefix (e.g. s3, gs, ssh, hdfs). Other remotes that can
+        check if particular file exists much quicker, use their own
+        implementation of hashes_exist (see ssh, local).
+
+        Which method to use will be automatically determined after estimating
+        the size of the remote cache, and comparing the estimated size with
+        len(hashes). To estimate the size of the remote cache, we fetch
+        a small subset of cache entries (i.e. entries starting with "00...").
+        Based on the number of entries in that subset, the size of the full
+        cache can be estimated, since the cache is evenly distributed according
+        to hash.
+
+        Returns:
+            A list with hashes that were found in the remote
+        """
+        # Remotes which do not use traverse prefix should override
+        # hashes_exist() (see ssh, local)
+        assert self.tree.TRAVERSE_PREFIX_LEN >= 2
+
+        hashes = set(hashes)
+        if len(hashes) == 1 or not self.tree.CAN_TRAVERSE:
+            remote_hashes = self.list_hashes_exists(hashes, jobs, name)
+            return remote_hashes
+
+        # Max remote size allowed for us to use traverse method
+        remote_size, remote_hashes = self._estimate_remote_size(hashes, name)
+
+        traverse_pages = remote_size / self.tree.LIST_OBJECT_PAGE_SIZE
+        # For sufficiently large remotes, traverse must be weighted to account
+        # for performance overhead from large lists/sets.
+        # From testing with S3, for remotes with 1M+ files, object_exists is
+        # faster until len(hashes) is at least 10k~100k
+        if remote_size > self.tree.TRAVERSE_THRESHOLD_SIZE:
+            traverse_weight = (
+                traverse_pages * self.tree.TRAVERSE_WEIGHT_MULTIPLIER
+            )
+        else:
+            traverse_weight = traverse_pages
+        if len(hashes) < traverse_weight:
+            logger.debug(
+                "Large remote ('{}' hashes < '{}' traverse weight), "
+                "using object_exists for remaining hashes".format(
+                    len(hashes), traverse_weight
+                )
+            )
+            return list(hashes & remote_hashes) + self.list_hashes_exists(
+                hashes - remote_hashes, jobs, name
+            )
+
+        logger.debug("Querying '{}' hashes via traverse".format(len(hashes)))
+        remote_hashes = set(
+            self.list_hashes_traverse(remote_size, remote_hashes, jobs, name)
+        )
+        return list(hashes & set(remote_hashes))

dvc/cache/local.py

@@ -1,52 +1,23 @@
-import errno
 import logging
 import os
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from functools import partial, wraps
 
-from funcy import cached_property, concat
+from funcy import cached_property
 
-from dvc.exceptions import DownloadError, UploadError
+from dvc.hash_info import HashInfo
 from dvc.path_info import PathInfo
 from dvc.progress import Tqdm
 
-from ..remote.base import index_locked
 from ..tree.local import LocalTree
-from .base import (
-    STATUS_DELETED,
-    STATUS_MAP,
-    STATUS_MISSING,
-    STATUS_NEW,
-    CloudCache,
-)
+from ..utils.fs import walk_files
+from .base import CloudCache
 
 logger = logging.getLogger(__name__)
 
 
-def _log_exceptions(func, operation):
-    @wraps(func)
-    def wrapper(from_info, to_info, *args, **kwargs):
-        try:
-            func(from_info, to_info, *args, **kwargs)
-            return 0
-        except Exception as exc:  # pylint: disable=broad-except
-            # NOTE: this means we ran out of file descriptors and there is no
-            # reason to try to proceed, as we will hit this error anyways.
-            # pylint: disable=no-member
-            if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
-                raise
-
-            logger.exception(
-                "failed to %s '%s' to '%s'", operation, from_info, to_info
-            )
-            return 1
-
-    return wrapper
-
-
 class LocalCache(CloudCache):
     DEFAULT_CACHE_TYPES = ["reflink", "copy"]
     CACHE_MODE = LocalTree.CACHE_MODE
+    UNPACKED_DIR_SUFFIX = ".unpacked"
 
     def __init__(self, tree):
         super().__init__(tree)
@@ -60,10 +31,6 @@ class LocalCache(CloudCache):
     def cache_dir(self, value):
         self.tree.path_info = PathInfo(value) if value else None
 
-    @classmethod
-    def supported(cls, config):  # pylint: disable=unused-argument
-        return True
-
     @cached_property
     def cache_path(self):
         return os.path.abspath(self.cache_dir)
@@ -85,18 +52,15 @@ class LocalCache(CloudCache):
                 desc="Querying "
                 + ("cache in " + name if name else "local cache"),
             )
-            if not self.changed_cache_file(hash_)
+            if not self.changed_cache_file(
+                HashInfo(self.tree.PARAM_CHECKSUM, hash_)
+            )
         ]
 
     def already_cached(self, path_info):
         assert path_info.scheme in ["", "local"]
 
-        current_md5 = self.tree.get_hash(path_info)
-
-        if not current_md5:
-            return False
-
-        return not self.changed_cache(current_md5)
+        return super().already_cached(path_info)
 
     def _verify_link(self, path_info, link_type):
         if link_type == "hardlink" and self.tree.getsize(path_info) == 0:
@@ -104,328 +68,24 @@ class LocalCache(CloudCache):
 
         super()._verify_link(path_info, link_type)
 
-    @index_locked
-    def status(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        # Return flattened dict containing all status info
-        dir_status, file_status, _ = self._status(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=download,
-        )
-        return dict(dir_status, **file_status)
-
-    def _status(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        """Return a tuple of (dir_status_info, file_status_info, dir_contents).
-
-        dir_status_info contains status for .dir files, file_status_info
-        contains status for all other files, and dir_contents is a dict of
-        {dir_hash: set(file_hash, ...)} which can be used to map
-        a .dir file to its file contents.
-        """
-        logger.debug(
-            f"Preparing to collect status from {remote.tree.path_info}"
-        )
-        md5s = set(named_cache.scheme_keys(self.tree.scheme))
-
-        logger.debug("Collecting information from local cache...")
-        local_exists = frozenset(
-            self.hashes_exist(md5s, jobs=jobs, name=self.cache_dir)
-        )
-
-        # This is a performance optimization. We can safely assume that,
-        # if the resources that we want to fetch are already cached,
-        # there's no need to check the remote storage for the existence of
-        # those files.
-        if download and local_exists == md5s:
-            remote_exists = local_exists
+    def _list_paths(self, prefix=None, progress_callback=None):
+        assert self.tree.path_info is not None
+        if prefix:
+            path_info = self.tree.path_info / prefix[:2]
+            if not self.tree.exists(path_info):
+                return
         else:
-            logger.debug("Collecting information from remote cache...")
-            remote_exists = set()
-            dir_md5s = set(named_cache.dir_keys(self.tree.scheme))
-            if dir_md5s:
-                remote_exists.update(
-                    self._indexed_dir_hashes(named_cache, remote, dir_md5s)
-                )
-                md5s.difference_update(remote_exists)
-            if md5s:
-                remote_exists.update(
-                    remote.hashes_exist(
-                        md5s, jobs=jobs, name=str(remote.tree.path_info)
-                    )
-                )
-        return self._make_status(
-            named_cache, show_checksums, local_exists, remote_exists
-        )
-
-    def _make_status(
-        self, named_cache, show_checksums, local_exists, remote_exists
-    ):
-        def make_names(hash_, names):
-            return {"name": hash_ if show_checksums else " ".join(names)}
-
-        dir_status = {}
-        file_status = {}
-        dir_contents = {}
-        for hash_, item in named_cache[self.tree.scheme].items():
-            if item.children:
-                dir_status[hash_] = make_names(hash_, item.names)
-                dir_contents[hash_] = set()
-                for child_hash, child in item.children.items():
-                    file_status[child_hash] = make_names(
-                        child_hash, child.names
-                    )
-                    dir_contents[hash_].add(child_hash)
-            else:
-                file_status[hash_] = make_names(hash_, item.names)
-
-        self._fill_statuses(dir_status, local_exists, remote_exists)
-        self._fill_statuses(file_status, local_exists, remote_exists)
-
-        self._log_missing_caches(dict(dir_status, **file_status))
-
-        return dir_status, file_status, dir_contents
-
-    def _indexed_dir_hashes(self, named_cache, remote, dir_md5s):
-        # Validate our index by verifying all indexed .dir hashes
-        # still exist on the remote
-        indexed_dirs = set(remote.index.dir_hashes())
-        indexed_dir_exists = set()
-        if indexed_dirs:
-            indexed_dir_exists.update(
-                remote.tree.list_hashes_exists(indexed_dirs)
-            )
-            missing_dirs = indexed_dirs.difference(indexed_dir_exists)
-            if missing_dirs:
-                logger.debug(
-                    "Remote cache missing indexed .dir hashes '{}', "
-                    "clearing remote index".format(", ".join(missing_dirs))
-                )
-                remote.index.clear()
-
-        # Check if non-indexed (new) dir hashes exist on remote
-        dir_exists = dir_md5s.intersection(indexed_dir_exists)
-        dir_exists.update(
-            remote.tree.list_hashes_exists(dir_md5s - dir_exists)
-        )
-
-        # If .dir hash exists on the remote, assume directory contents
-        # still exists on the remote
-        for dir_hash in dir_exists:
-            file_hashes = list(
-                named_cache.child_keys(self.tree.scheme, dir_hash)
-            )
-            if dir_hash not in remote.index:
-                logger.debug(
-                    "Indexing new .dir '{}' with '{}' nested files".format(
-                        dir_hash, len(file_hashes)
-                    )
-                )
-                remote.index.update([dir_hash], file_hashes)
-            yield dir_hash
-            yield from file_hashes
-
-    @staticmethod
-    def _fill_statuses(hash_info_dir, local_exists, remote_exists):
-        # Using sets because they are way faster for lookups
-        local = set(local_exists)
-        remote = set(remote_exists)
-
-        for md5, info in hash_info_dir.items():
-            status = STATUS_MAP[(md5 in local, md5 in remote)]
-            info["status"] = status
-
-    def _get_plans(self, download, remote, status_info, status):
-        cache = []
-        path_infos = []
-        names = []
-        hashes = []
-        for md5, info in Tqdm(
-            status_info.items(), desc="Analysing status", unit="file"
-        ):
-            if info["status"] == status:
-                cache.append(self.tree.hash_to_path_info(md5))
-                path_infos.append(remote.tree.hash_to_path_info(md5))
-                names.append(info["name"])
-                hashes.append(md5)
-
-        if download:
-            to_infos = cache
-            from_infos = path_infos
+            path_info = self.tree.path_info
+        # NOTE: use utils.fs walk_files since tree.walk_files will not follow
+        # symlinks
+        if progress_callback:
+            for path in walk_files(path_info):
+                progress_callback()
+                yield path
         else:
-            to_infos = path_infos
-            from_infos = cache
+            yield from walk_files(path_info)
 
-        return from_infos, to_infos, names, hashes
-
-    def _process(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        logger.debug(
-            "Preparing to {} '{}'".format(
-                "download data from" if download else "upload data to",
-                remote.tree.path_info,
-            )
-        )
-
-        if download:
-            func = partial(
-                _log_exceptions(remote.tree.download, "download"),
-                dir_mode=self.tree.dir_mode,
-                file_mode=self.tree.file_mode,
-            )
-            status = STATUS_DELETED
-            desc = "Downloading"
-        else:
-            func = _log_exceptions(remote.tree.upload, "upload")
-            status = STATUS_NEW
-            desc = "Uploading"
-
-        if jobs is None:
-            jobs = remote.tree.JOBS
-
-        dir_status, file_status, dir_contents = self._status(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=download,
-        )
-
-        dir_plans = self._get_plans(download, remote, dir_status, status)
-        file_plans = self._get_plans(download, remote, file_status, status)
-
-        total = len(dir_plans[0]) + len(file_plans[0])
-        if total == 0:
-            return 0
-
-        with Tqdm(total=total, unit="file", desc=desc) as pbar:
-            func = pbar.wrap_fn(func)
-            with ThreadPoolExecutor(max_workers=jobs) as executor:
-                if download:
-                    from_infos, to_infos, names, _ = (
-                        d + f for d, f in zip(dir_plans, file_plans)
-                    )
-                    fails = sum(
-                        executor.map(func, from_infos, to_infos, names)
-                    )
-                else:
-                    # for uploads, push files first, and any .dir files last
-
-                    file_futures = {}
-                    for from_info, to_info, name, hash_ in zip(*file_plans):
-                        file_futures[hash_] = executor.submit(
-                            func, from_info, to_info, name
-                        )
-                    dir_futures = {}
-                    for from_info, to_info, name, dir_hash in zip(*dir_plans):
-                        wait_futures = {
-                            future
-                            for file_hash, future in file_futures.items()
-                            if file_hash in dir_contents[dir_hash]
-                        }
-                        dir_futures[dir_hash] = executor.submit(
-                            self._dir_upload,
-                            func,
-                            wait_futures,
-                            from_info,
-                            to_info,
-                            name,
-                        )
-                    fails = sum(
-                        future.result()
-                        for future in concat(
-                            file_futures.values(), dir_futures.values()
-                        )
-                    )
-
-        if fails:
-            if download:
-                remote.index.clear()
-                raise DownloadError(fails)
-            raise UploadError(fails)
-
-        if not download:
-            # index successfully pushed dirs
-            for dir_hash, future in dir_futures.items():
-                if future.result() == 0:
-                    file_hashes = dir_contents[dir_hash]
-                    logger.debug(
-                        "Indexing pushed dir '{}' with "
-                        "'{}' nested files".format(dir_hash, len(file_hashes))
-                    )
-                    remote.index.update([dir_hash], file_hashes)
-
-        return len(dir_plans[0]) + len(file_plans[0])
-
-    @staticmethod
-    def _dir_upload(func, futures, from_info, to_info, name):
-        for future in as_completed(futures):
-            if future.result():
-                # do not upload this .dir file if any file in this
-                # directory failed to upload
-                logger.debug(
-                    "failed to upload full contents of '{}', "
-                    "aborting .dir file upload".format(name)
-                )
-                logger.error(f"failed to upload '{from_info}' to '{to_info}'")
-                return 1
-        return func(from_info, to_info, name)
-
-    @index_locked
-    def push(self, named_cache, remote, jobs=None, show_checksums=False):
-        return self._process(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=False,
-        )
-
-    @index_locked
-    def pull(self, named_cache, remote, jobs=None, show_checksums=False):
-        return self._process(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=True,
-        )
-
-    @staticmethod
-    def _log_missing_caches(hash_info_dict):
-        missing_caches = [
-            (md5, info)
-            for md5, info in hash_info_dict.items()
-            if info["status"] == STATUS_MISSING
-        ]
-        if missing_caches:
-            missing_desc = "\n".join(
-                "name: {}, md5: {}".format(info["name"], md5)
-                for md5, info in missing_caches
-            )
-            msg = (
-                "Some of the cache files do not exist neither locally "
-                "nor on remote. Missing cache files:\n{}".format(missing_desc)
-            )
-            logger.warning(msg)
+    def _remove_unpacked_dir(self, hash_):
+        info = self.tree.hash_to_path_info(hash_)
+        path_info = info.with_name(info.name + self.UNPACKED_DIR_SUFFIX)
+        self.tree.remove(path_info)

dvc/cache/ssh.py

@@ -1,17 +1,18 @@
 import errno
 import itertools
 import logging
+import posixpath
 from concurrent.futures import ThreadPoolExecutor
 
 from dvc.progress import Tqdm
 from dvc.utils import to_chunks
 
-from .base import Remote
+from .base import CloudCache
 
 logger = logging.getLogger(__name__)
 
 
-class SSHRemote(Remote):
+class SSHCache(CloudCache):
     def batch_exists(self, path_infos, callback):
         def _exists(chunk_and_channel):
             chunk, channel = chunk_and_channel
@@ -47,7 +48,7 @@ class SSHRemote(Remote):
         remote/base.
         """
         if not self.tree.CAN_TRAVERSE:
-            return list(set(hashes) & set(self.tree.all()))
+            return list(set(hashes) & set(self.all()))
 
         # possibly prompt for credentials before "Querying" progress output
         self.tree.ensure_credentials()
@@ -71,3 +72,19 @@ class SSHRemote(Remote):
                 in_remote = itertools.chain.from_iterable(results)
                 ret = list(itertools.compress(hashes, in_remote))
                 return ret
+
+    def _list_paths(self, prefix=None, progress_callback=None):
+        if prefix:
+            root = posixpath.join(self.tree.path_info.path, prefix[:2])
+        else:
+            root = self.tree.path_info.path
+        with self.tree.ssh(self.tree.path_info) as ssh:
+            if prefix and not ssh.exists(root):
+                return
+            # If we simply return an iterator then with above closes instantly
+            if progress_callback:
+                for path in ssh.walk_files(root):
+                    progress_callback()
+                    yield path
+            else:
+                yield from ssh.walk_files(root)

dvc/cli.py

@@ -1,11 +1,13 @@
 """DVC command line interface"""
 import argparse
 import logging
+import os
 import sys
 
 from .command import (
     add,
     cache,
+    check_ignore,
     checkout,
     commit,
     completion,
@@ -79,6 +81,7 @@ COMMANDS = [
     git_hook,
     plots,
     experiments,
+    check_ignore,
 ]
 
 
@@ -135,6 +138,18 @@ def get_parent_parser():
     """
     parent_parser = argparse.ArgumentParser(add_help=False)
 
+    parent_parser.add_argument(
+        "--cprofile",
+        action="store_true",
+        default=False,
+        help=argparse.SUPPRESS,
+    )
+    parent_parser.add_argument("--cprofile-dump", help=argparse.SUPPRESS)
+
+    parent_parser.add_argument(
+        "--pdb", action="store_true", default=False, help=argparse.SUPPRESS,
+    )
+
     log_level_group = parent_parser.add_mutually_exclusive_group()
     log_level_group.add_argument(
         "-q", "--quiet", action="count", default=0, help="Be quiet."
@@ -181,6 +196,14 @@ def get_main_parser():
         help="Show program's version.",
     )
 
+    parser.add_argument(
+        "--cd",
+        default=os.path.curdir,
+        metavar="<path>",
+        help="Change to directory before executing.",
+        type=str,
+    )
+
     # Sub commands
     subparsers = parser.add_subparsers(
         title="Available Commands",

dvc/command/add.py

@@ -20,6 +20,8 @@ class CmdAdd(CmdBase):
                 no_commit=self.args.no_commit,
                 fname=self.args.file,
                 external=self.args.external,
+                glob=self.args.glob,
+                desc=self.args.desc,
             )
 
         except DvcException:
@@ -57,11 +59,26 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Allow targets that are outside of the DVC repository.",
     )
+    parser.add_argument(
+        "--glob",
+        action="store_true",
+        default=False,
+        help="Allows targets containing shell-style wildcards.",
+    )
     parser.add_argument(
         "--file",
         help="Specify name of the DVC-file this command will generate.",
         metavar="<filename>",
     )
+    parser.add_argument(
+        "--desc",
+        type=str,
+        metavar="<text>",
+        help=(
+            "User description of the data (optional). "
+            "This doesn't affect any DVC operations."
+        ),
+    )
     parser.add_argument(
         "targets", nargs="+", help="Input files/directories to add.",
     ).complete = completion.FILE

dvc/command/base.py

@@ -1,4 +1,5 @@
 import logging
+import os
 from abc import ABC, abstractmethod
 
 logger = logging.getLogger(__name__)
@@ -28,16 +29,21 @@ def append_doc_link(help_message, path):
 
 
 class CmdBase(ABC):
+    UNINITIALIZED = False
+
     def __init__(self, args):
         from dvc.repo import Repo
         from dvc.updater import Updater
 
-        self.repo = Repo()
+        os.chdir(args.cd)
+
+        self.repo = Repo(uninitialized=self.UNINITIALIZED)
         self.config = self.repo.config
         self.args = args
-        hardlink_lock = self.config["core"].get("hardlink_lock", False)
-        updater = Updater(self.repo.tmp_dir, hardlink_lock=hardlink_lock)
-        updater.check()
+        if self.repo.tmp_dir:
+            hardlink_lock = self.config["core"].get("hardlink_lock", False)
+            updater = Updater(self.repo.tmp_dir, hardlink_lock=hardlink_lock)
+            updater.check()
 
     @property
     def default_targets(self):
@@ -55,3 +61,5 @@ class CmdBase(ABC):
 class CmdBaseNoRepo(CmdBase):
     def __init__(self, args):  # pylint: disable=super-init-not-called
         self.args = args
+
+        os.chdir(args.cd)

dvc/command/cache.py

@@ -1,12 +1,18 @@
 import argparse
+import logging
 
 from dvc.command import completion
 from dvc.command.base import append_doc_link, fix_subparsers
 from dvc.command.config import CmdConfig
 
+logger = logging.getLogger(__name__)
+
 
 class CmdCacheDir(CmdConfig):
     def run(self):
+        if self.args.value is None and not self.args.unset:
+            logger.info(self.config["cache"]["dir"])
+            return 0
         with self.config.edit(level=self.args.level) as edit:
             edit["cache"]["dir"] = self.args.value
         return 0
@@ -55,6 +61,8 @@ def add_parser(subparsers, parent_parser):
         "value",
         help="Path to cache directory. Relative paths are resolved relative "
         "to the current directory and saved to config relative to the "
-        "config file location.",
+        "config file location. If no path is provided, it returns the "
+        "current cache directory.",
+        nargs="?",
     ).complete = completion.DIR
     cache_dir_parser.set_defaults(func=CmdCacheDir)

dvc/command/check_ignore.py

@@ -0,0 +1,126 @@
+import argparse
+import logging
+
+from dvc.command import completion
+from dvc.command.base import CmdBase, append_doc_link
+from dvc.exceptions import DvcException
+
+logger = logging.getLogger(__name__)
+
+
+class CmdCheckIgnore(CmdBase):
+    def __init__(self, args):
+        super().__init__(args)
+        self.ignore_filter = self.repo.tree.dvcignore
+
+    def _show_results(self, result):
+        if not result.match and not self.args.non_matching:
+            return
+
+        if self.args.details:
+            patterns = result.patterns
+            if not self.args.all:
+                patterns = patterns[-1:]
+
+            for pattern in patterns:
+                logger.info("{}\t{}".format(pattern, result.file))
+        else:
+            logger.info(result.file)
+
+    def _check_one_file(self, target):
+        result = self.ignore_filter.check_ignore(target)
+        self._show_results(result)
+        if result.match:
+            return 0
+        return 1
+
+    def _interactive_mode(self):
+        ret = 1
+        while True:
+            try:
+                target = input()
+            except (KeyboardInterrupt, EOFError):
+                break
+            if target == "":
+                break
+            if not self._check_one_file(target):
+                ret = 0
+        return ret
+
+    def _normal_mode(self):
+        ret = 1
+        for target in self.args.targets:
+            if not self._check_one_file(target):
+                ret = 0
+        return ret
+
+    def _check_args(self):
+        if not self.args.stdin and not self.args.targets:
+            raise DvcException("`targets` or `--stdin` needed")
+
+        if self.args.stdin and self.args.targets:
+            raise DvcException("cannot have both `targets` and `--stdin`")
+
+        if self.args.non_matching and not self.args.details:
+            raise DvcException(
+                "`--non-matching` is only valid with `--details`"
+            )
+
+        if self.args.all and not self.args.details:
+            raise DvcException("`--all` is only valid with `--details`")
+
+        if self.args.quiet and self.args.details:
+            raise DvcException("cannot use both `--details` and `--quiet`")
+
+    def run(self):
+        self._check_args()
+        if self.args.stdin:
+            return self._interactive_mode()
+        return self._normal_mode()
+
+
+def add_parser(subparsers, parent_parser):
+    ADD_HELP = (
+        "Check whether files or directories are excluded due to `.dvcignore`."
+    )
+
+    parser = subparsers.add_parser(
+        "check-ignore",
+        parents=[parent_parser],
+        description=append_doc_link(ADD_HELP, "check-ignore"),
+        help=ADD_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    parser.add_argument(
+        "-d",
+        "--details",
+        action="store_true",
+        default=False,
+        help="Show the exclude patterns along with each target path.",
+    )
+    parser.add_argument(
+        "-a",
+        "--all",
+        action="store_true",
+        default=False,
+        help="Include the target paths which don’t match any pattern "
+        "in the `--details` list.",
+    )
+    parser.add_argument(
+        "-n",
+        "--non-matching",
+        action="store_true",
+        default=False,
+        help="Include the target paths which don’t match any pattern "
+        "in the `--details` list.",
+    )
+    parser.add_argument(
+        "--stdin",
+        action="store_true",
+        default=False,
+        help="Read paths from standard input instead of providing `targets`.",
+    )
+    parser.add_argument(
+        "targets", nargs="*", help="File or directory paths to check",
+    ).complete = completion.FILE
+    parser.set_defaults(func=CmdCheckIgnore)

dvc/command/daemon.py

@@ -9,8 +9,9 @@ class CmdDaemonBase(CmdBaseNoRepo):
 class CmdDaemonUpdater(CmdDaemonBase):
     def run(self):
         import os
-        from dvc.repo import Repo
+
         from dvc.config import Config
+        from dvc.repo import Repo
         from dvc.updater import Updater
 
         root_dir = Repo.find_root()

dvc/command/dag.py

@@ -2,14 +2,14 @@ import argparse
 import logging
 
 from dvc.command.base import CmdBase, append_doc_link
-from dvc.exceptions import DvcException
+from dvc.stage import Stage
 
 logger = logging.getLogger(__name__)
 
 
 def _show_ascii(G):
-    from dvc.repo.graph import get_pipelines
     from dvc.dagascii import draw
+    from dvc.repo.graph import get_pipelines
 
     pipelines = get_pipelines(G)
 
@@ -22,6 +22,7 @@ def _show_ascii(G):
 
 def _show_dot(G):
     import io
+
     from networkx.drawing.nx_pydot import write_dot
 
     dot_file = io.StringIO()
@@ -29,57 +30,83 @@ def _show_dot(G):
     return dot_file.getvalue()
 
 
-def _build(G, target=None, full=False):
+def _collect_targets(repo, target, outs):
+    if not target:
+        return []
+
+    pairs = repo.collect_granular(target)
+    if not outs:
+        return [stage.addressing for stage, _ in pairs]
+
+    targets = []
+    for stage, info in pairs:
+        if not info:
+            targets.extend([str(out) for out in stage.outs])
+            continue
+
+        for out in repo.outs_trie.itervalues(prefix=info.parts):  # noqa: B301
+            targets.extend(str(out))
+
+    return targets
+
+
+def _transform(repo, outs):
     import networkx as nx
-    from dvc.repo.graph import get_pipeline, get_pipelines
 
-    if target:
-        H = get_pipeline(get_pipelines(G), target)
-        if not full:
-            descendants = nx.descendants(G, target)
+    def _relabel(node) -> str:
+        return node.addressing if isinstance(node, Stage) else str(node)
+
+    G = repo.outs_graph if outs else repo.graph
+    return nx.relabel_nodes(G, _relabel, copy=True)
+
+
+def _filter(G, targets, full):
+    import networkx as nx
+
+    if not targets:
+        return G
+
+    H = G.copy()
+    if not full:
+        descendants = set()
+        for target in targets:
+            descendants.update(nx.descendants(G, target))
             descendants.add(target)
-            H.remove_nodes_from(set(G.nodes()) - descendants)
-    else:
-        H = G
+        H.remove_nodes_from(set(G.nodes()) - descendants)
 
-    def _relabel(stage):
-        return stage.addressing
+    undirected = H.to_undirected()
+    connected = set()
+    for target in targets:
+        connected.update(nx.node_connected_component(undirected, target))
 
-    return nx.relabel_nodes(H, _relabel, copy=False)
+    H.remove_nodes_from(set(H.nodes()) - connected)
+
+    return H
+
+
+def _build(repo, target=None, full=False, outs=False):
+    targets = _collect_targets(repo, target, outs)
+    G = _transform(repo, outs)
+    return _filter(G, targets, full)
 
 
 class CmdDAG(CmdBase):
     def run(self):
-        try:
-            target = None
-            if self.args.target:
-                stages = self.repo.collect(self.args.target)
-                if len(stages) > 1:
-                    logger.error(
-                        f"'{self.args.target}' contains more than one stage "
-                        "{stages}, please specify one stage"
-                    )
-                    return 1
-                target = stages[0]
+        G = _build(
+            self.repo,
+            target=self.args.target,
+            full=self.args.full,
+            outs=self.args.outs,
+        )
 
-            G = _build(self.repo.graph, target=target, full=self.args.full,)
+        if self.args.dot:
+            logger.info(_show_dot(G))
+        else:
+            from dvc.utils.pager import pager
 
-            if self.args.dot:
-                logger.info(_show_dot(G))
-            else:
-                from dvc.utils.pager import pager
+            pager(_show_ascii(G))
 
-                pager(_show_ascii(G))
-
-            return 0
-        except DvcException:
-            msg = "failed to show "
-            if self.args.target:
-                msg += f"a pipeline for '{target}'"
-            else:
-                msg += "pipelines"
-            logger.exception(msg)
-            return 1
+        return 0
 
 
 def add_parser(subparsers, parent_parser):
@@ -106,6 +133,13 @@ def add_parser(subparsers, parent_parser):
             "showing DAG consisting only of ancestors."
         ),
     )
+    dag_parser.add_argument(
+        "-o",
+        "--outs",
+        action="store_true",
+        default=False,
+        help="Print output files instead of stages.",
+    )
     dag_parser.add_argument(
         "target",
         nargs="?",

dvc/command/data_sync.py

@@ -97,7 +97,11 @@ def shared_parent_parser():
         "-j",
         "--jobs",
         type=int,
-        help="Number of jobs to run simultaneously.",
+        help=(
+            "Number of jobs to run simultaneously. "
+            "The default value is 4 * cpu_count(). "
+            "For SSH remotes, the default is 4. "
+        ),
         metavar="<number>",
     )
     parent_parser.add_argument(

dvc/command/destroy.py

@@ -25,7 +25,7 @@ class CmdDestroy(CmdBase):
                 )
 
             self.repo.destroy()
-        except Exception:  # noqa, pylint: disable=broad-except
+        except DvcException:
             logger.exception("failed to destroy DVC")
             return 1
         return 0

dvc/command/diff.py

@@ -11,30 +11,43 @@ from dvc.exceptions import DvcException
 logger = logging.getLogger(__name__)
 
 
-def _show_md(diff):
+def _digest(checksum):
+    if isinstance(checksum, str):
+        return checksum[0:8]
+    return "{}..{}".format(checksum["old"][0:8], checksum["new"][0:8])
+
+
+def _show_md(diff, show_hash=False, hide_missing=False):
     from dvc.utils.diff import table
 
+    header = ["Status", "Hash", "Path"] if show_hash else ["Status", "Path"]
     rows = []
-    for status in ["added", "deleted", "modified"]:
+    statuses = ["added", "deleted", "modified"]
+    if not hide_missing:
+        statuses.append("not in cache")
+    for status in statuses:
         entries = diff.get(status, [])
         if not entries:
             continue
-        paths = sorted(entry["path"] for entry in entries)
-        for path in paths:
-            rows.append([status, path])
+        for entry in entries:
+            path = entry["path"]
+            if show_hash:
+                check_sum = _digest(entry.get("hash", ""))
+                rows.append([status, check_sum, path])
+            else:
+                rows.append([status, path])
 
-    return table(["Status", "Path"], rows, True)
+    return table(header, rows, True)
 
 
 class CmdDiff(CmdBase):
     @staticmethod
-    def _format(diff):
+    def _format(diff, hide_missing=False):
         """
         Given a diff structure, generate a string of paths separated
         by new lines and grouped together by their state.
 
-        A group's header is colored and its entries are sorted to enhance
-        readability, for example:
+        A group's header is colored to enhance readability, for example:
 
             Added:
                 another_file.txt
@@ -55,21 +68,20 @@ class CmdDiff(CmdBase):
         At the bottom, include a summary with the number of files per state.
         """
 
-        def _digest(checksum):
-            if isinstance(checksum, str):
-                return checksum[0:8]
-            return "{}..{}".format(checksum["old"][0:8], checksum["new"][0:8])
-
         colors = {
             "added": colorama.Fore.GREEN,
             "modified": colorama.Fore.YELLOW,
             "deleted": colorama.Fore.RED,
+            "not in cache": colorama.Fore.YELLOW,
         }
 
         summary = {}
         groups = []
 
-        for state in ["added", "deleted", "modified"]:
+        states = ["added", "deleted", "modified"]
+        if not hide_missing:
+            states.append("not in cache")
+        for state in states:
             summary[state] = 0
             entries = diff[state]
 
@@ -100,28 +112,42 @@ class CmdDiff(CmdBase):
                 )
             )
 
-        groups.append(
+        if not sum(summary.values()):
+            return None
+
+        fmt = (
             "files summary: {added} added, {deleted} deleted,"
-            " {modified} modified".format_map(summary)
+            " {modified} modified"
         )
+        if not hide_missing:
+            fmt += ", {not in cache} not in cache"
+        groups.append(fmt.format_map(summary))
 
         return "\n\n".join(groups)
 
     def run(self):
         try:
             diff = self.repo.diff(self.args.a_rev, self.args.b_rev)
+            show_hash = self.args.show_hash
+            hide_missing = self.args.b_rev or self.args.hide_missing
+            if hide_missing:
+                del diff["not in cache"]
 
-            if not self.args.show_hash:
-                for _, entries in diff.items():
+            for key, entries in diff.items():
+                entries = sorted(entries, key=lambda entry: entry["path"])
+                if not show_hash:
                     for entry in entries:
                         del entry["hash"]
+                diff[key] = entries
 
             if self.args.show_json:
                 logger.info(json.dumps(diff))
             elif self.args.show_md:
-                logger.info(_show_md(diff))
+                logger.info(_show_md(diff, show_hash, hide_missing))
             elif diff:
-                logger.info(self._format(diff))
+                output = self._format(diff, hide_missing)
+                if output:
+                    logger.info(output)
 
         except DvcException:
             logger.exception("failed to get diff")
@@ -170,4 +196,9 @@ def add_parser(subparsers, parent_parser):
         action="store_true",
         default=False,
     )
+    diff_parser.add_argument(
+        "--hide-missing",
+        help="Hide missing cache file status.",
+        action="store_true",
+    )
     diff_parser.set_defaults(func=CmdDiff)

dvc/command/experiments.py

@@ -1,27 +1,77 @@
 import argparse
 import io
 import logging
+import os
 from collections import OrderedDict
+from collections.abc import Mapping
+from datetime import date, datetime
+from itertools import groupby
+from typing import Iterable, Optional
 
+import dvc.prompt as prompt
 from dvc.command.base import CmdBase, append_doc_link, fix_subparsers
 from dvc.command.metrics import DEFAULT_PRECISION
-from dvc.exceptions import DvcException
+from dvc.command.repro import CmdRepro
+from dvc.command.repro import add_arguments as add_repro_arguments
+from dvc.exceptions import DvcException, InvalidArgumentError
+from dvc.repo.experiments import Experiments
+from dvc.scm.git import Git
+from dvc.utils.flatten import flatten
 
 logger = logging.getLogger(__name__)
 
 
-def _update_names(names, items):
-    from flatten_json import flatten
+def _filter_names(
+    names: Iterable,
+    label: str,
+    include: Optional[Iterable],
+    exclude: Optional[Iterable],
+):
+    if include and exclude:
+        intersection = set(include) & set(exclude)
+        if intersection:
+            values = ", ".join(intersection)
+            raise InvalidArgumentError(
+                f"'{values}' specified in both --include-{label} and"
+                f" --exclude-{label}"
+            )
 
+    names = [tuple(name.split(".")) for name in names]
+
+    def _filter(filters, update_func):
+        filters = [tuple(name.split(".")) for name in filters]
+        for length, groups in groupby(filters, len):
+            for group in groups:
+                matches = [name for name in names if name[:length] == group]
+                if not matches:
+                    name = ".".join(group)
+                    raise InvalidArgumentError(
+                        f"'{name}' does not match any known {label}"
+                    )
+                update_func({match: None for match in matches})
+
+    if include:
+        ret: OrderedDict = OrderedDict()
+        _filter(include, ret.update)
+    else:
+        ret = OrderedDict({name: None for name in names})
+
+    if exclude:
+        _filter(exclude, ret.difference_update)  # type: ignore[attr-defined]
+
+    return [".".join(name) for name in ret]
+
+
+def _update_names(names, items):
     for name, item in items:
         if isinstance(item, dict):
-            item = flatten(item, ".")
+            item = flatten(item)
             names.update(item.keys())
         else:
-            names.add(name)
+            names[name] = None
 
 
-def _collect_names(all_experiments):
+def _collect_names(all_experiments, **kwargs):
     metric_names = set()
     param_names = set()
 
@@ -30,84 +80,223 @@ def _collect_names(all_experiments):
             _update_names(metric_names, exp.get("metrics", {}).items())
             _update_names(param_names, exp.get("params", {}).items())
 
-    return sorted(metric_names), sorted(param_names)
+    metric_names = _filter_names(
+        sorted(metric_names),
+        "metrics",
+        kwargs.get("include_metrics"),
+        kwargs.get("exclude_metrics"),
+    )
+    param_names = _filter_names(
+        sorted(param_names),
+        "params",
+        kwargs.get("include_params"),
+        kwargs.get("exclude_params"),
+    )
+
+    return metric_names, param_names
 
 
 def _collect_rows(
-    base_rev, experiments, metric_names, param_names, precision=None
+    base_rev,
+    experiments,
+    metric_names,
+    param_names,
+    precision=DEFAULT_PRECISION,
+    no_timestamp=False,
+    sort_by=None,
+    sort_order=None,
 ):
-    from flatten_json import flatten
-
-    if precision is None:
-        precision = DEFAULT_PRECISION
-
-    def _round(val):
-        if isinstance(val, float):
-            return round(val, precision)
-
-        return val
-
-    def _extend(row, names, items):
-        for fname, item in items:
-            if isinstance(item, dict):
-                item = flatten(item, ".")
-            else:
-                item = {fname: item}
-            for name in names:
-                if name in item:
-                    row.append(str(_round(item[name])))
-                else:
-                    row.append("-")
+    if sort_by:
+        if sort_by in metric_names:
+            sort_type = "metrics"
+        elif sort_by in param_names:
+            sort_type = "params"
+        else:
+            raise InvalidArgumentError(f"Unknown sort column '{sort_by}'")
+        reverse = sort_order == "desc"
+        experiments = _sort_exp(experiments, sort_by, sort_type, reverse)
 
+    new_checkpoint = True
     for i, (rev, exp) in enumerate(experiments.items()):
         row = []
         style = None
-        if rev == "baseline":
-            row.append(f"{base_rev}")
-            style = "bold"
-        elif i < len(experiments) - 1:
-            row.append(f"├── {rev[:7]}")
-        else:
-            row.append(f"└── {rev[:7]}")
+        queued = "*" if exp.get("queued", False) else ""
 
-        _extend(row, metric_names, exp.get("metrics", {}).items())
-        _extend(row, param_names, exp.get("params", {}).items())
+        tip = exp.get("checkpoint_tip")
+        parent = ""
+        if rev == "baseline":
+            if Git.is_sha(base_rev):
+                name_rev = base_rev[:7]
+            else:
+                name_rev = base_rev
+            name = exp.get("name", name_rev)
+            row.append(f"{name}")
+            style = "bold"
+        else:
+            if tip:
+                parent_rev = exp.get("checkpoint_parent", "")
+                parent_exp = experiments.get(parent_rev, {})
+                parent_tip = parent_exp.get("checkpoint_tip")
+                if tip == parent_tip:
+                    if new_checkpoint:
+                        tree = "│ ╓"
+                    else:
+                        tree = "│ ╟"
+                    new_checkpoint = False
+                else:
+                    if parent_rev == base_rev:
+                        tree = "├─╨"
+                    else:
+                        tree = "│ ╟"
+                        parent = f" ({parent_rev[:7]})"
+                    new_checkpoint = True
+            else:
+                if i < len(experiments) - 1:
+                    tree = "├──"
+                else:
+                    tree = "└──"
+                new_checkpoint = True
+            row.append(f"{tree} {queued}{rev[:7]}{parent}")
+
+        if not no_timestamp:
+            row.append(_format_time(exp.get("timestamp")))
+
+        _extend_row(
+            row, metric_names, exp.get("metrics", {}).items(), precision
+        )
+        _extend_row(row, param_names, exp.get("params", {}).items(), precision)
 
         yield row, style
 
 
-def _show_experiments(all_experiments, console, precision=None):
-    from rich.table import Table
-    from dvc.scm.git import Git
+def _sort_exp(experiments, sort_by, typ, reverse):
+    def _sort(item):
+        rev, exp = item
+        tip = exp.get("checkpoint_tip")
+        if tip and tip != rev:
+            # Sort checkpoint experiments by tip commit
+            return _sort((tip, experiments[tip]))
+        for fname, item in exp.get(typ, {}).items():
+            if isinstance(item, dict):
+                item = flatten(item)
+            else:
+                item = {fname: item}
+            if sort_by in item:
+                val = item[sort_by]
+                return (val is None, val)
+        return (True, None)
 
-    metric_names, param_names = _collect_names(all_experiments)
+    ret = OrderedDict()
+    if "baseline" in experiments:
+        ret["baseline"] = experiments.pop("baseline")
+
+    ret.update(sorted(experiments.items(), key=_sort, reverse=reverse))
+    return ret
+
+
+def _format_time(timestamp):
+    if timestamp is None:
+        return "-"
+    if timestamp.date() == date.today():
+        fmt = "%I:%M %p"
+    else:
+        fmt = "%b %d, %Y"
+    return timestamp.strftime(fmt)
+
+
+def _format_field(val, precision=DEFAULT_PRECISION):
+    if isinstance(val, float):
+        fmt = f"{{:.{precision}g}}"
+        return fmt.format(val)
+    elif isinstance(val, Mapping):
+        return {k: _format_field(v) for k, v in val.items()}
+    elif isinstance(val, list):
+        return [_format_field(x) for x in val]
+    return str(val)
+
+
+def _extend_row(row, names, items, precision):
+    from rich.text import Text
+
+    if not items:
+        row.extend(["-"] * len(names))
+        return
+
+    for fname, item in items:
+        if isinstance(item, dict):
+            item = flatten(item)
+        else:
+            item = {fname: item}
+        for name in names:
+            if name in item:
+                value = item[name]
+                if value is None:
+                    text = "-"
+                else:
+                    # wrap field data in rich.Text, otherwise rich may
+                    # interpret unescaped braces from list/dict types as rich
+                    # markup tags
+                    text = Text(str(_format_field(value, precision)))
+                row.append(text)
+            else:
+                row.append("-")
+
+
+def _parse_list(param_list):
+    ret = []
+    for param_str in param_list:
+        # we don't care about filename prefixes for show, silently
+        # ignore it if provided to keep usage consistent with other
+        # metric/param list command options
+        _, _, param_str = param_str.rpartition(":")
+        ret.extend(param_str.split(","))
+    return ret
+
+
+def _show_experiments(all_experiments, console, **kwargs):
+    from rich.table import Table
+
+    include_metrics = _parse_list(kwargs.pop("include_metrics", []))
+    exclude_metrics = _parse_list(kwargs.pop("exclude_metrics", []))
+    include_params = _parse_list(kwargs.pop("include_params", []))
+    exclude_params = _parse_list(kwargs.pop("exclude_params", []))
+
+    metric_names, param_names = _collect_names(
+        all_experiments,
+        include_metrics=include_metrics,
+        exclude_metrics=exclude_metrics,
+        include_params=include_params,
+        exclude_params=exclude_params,
+    )
 
     table = Table()
-    table.add_column("Experiment")
+    table.add_column("Experiment", no_wrap=True)
+    if not kwargs.get("no_timestamp", False):
+        table.add_column("Created")
     for name in metric_names:
-        table.add_column(name, justify="right")
+        table.add_column(name, justify="right", no_wrap=True)
     for name in param_names:
         table.add_column(name, justify="left")
 
     for base_rev, experiments in all_experiments.items():
-        if Git.is_sha(base_rev):
-            base_rev = base_rev[:7]
-
         for row, _, in _collect_rows(
-            base_rev,
-            experiments,
-            metric_names,
-            param_names,
-            precision=precision,
+            base_rev, experiments, metric_names, param_names, **kwargs,
         ):
             table.add_row(*row)
 
     console.print(table)
 
 
+def _format_json(item):
+    if isinstance(item, (date, datetime)):
+        return item.isoformat()
+    raise TypeError
+
+
 class CmdExperimentsShow(CmdBase):
     def run(self):
         from rich.console import Console
+
         from dvc.utils.pager import pager
 
         if not self.repo.experiments:
@@ -118,17 +307,44 @@ class CmdExperimentsShow(CmdBase):
                 all_branches=self.args.all_branches,
                 all_tags=self.args.all_tags,
                 all_commits=self.args.all_commits,
+                sha_only=self.args.sha,
             )
 
-            # Note: rich does not currently include a native way to force
-            # infinite width for use with a pager
-            console = Console(
-                file=io.StringIO(), force_terminal=True, width=9999
+            if self.args.show_json:
+                import json
+
+                logger.info(json.dumps(all_experiments, default=_format_json))
+                return 0
+
+            if self.args.no_pager:
+                console = Console()
+            else:
+                # Note: rich does not currently include a native way to force
+                # infinite width for use with a pager
+                console = Console(
+                    file=io.StringIO(), force_terminal=True, width=9999
+                )
+
+            if self.args.precision is None:
+                precision = DEFAULT_PRECISION
+            else:
+                precision = self.args.precision
+
+            _show_experiments(
+                all_experiments,
+                console,
+                include_metrics=self.args.include_metrics,
+                exclude_metrics=self.args.exclude_metrics,
+                include_params=self.args.include_params,
+                exclude_params=self.args.exclude_params,
+                no_timestamp=self.args.no_timestamp,
+                sort_by=self.args.sort_by,
+                sort_order=self.args.sort_order,
+                precision=precision,
             )
 
-            _show_experiments(all_experiments, console)
-
-            pager(console.file.getvalue())
+            if not self.args.no_pager:
+                pager(console.file.getvalue())
         except DvcException:
             logger.exception("failed to show experiments")
             return 1
@@ -141,27 +357,21 @@ class CmdExperimentsCheckout(CmdBase):
         if not self.repo.experiments:
             return 0
 
-        self.repo.experiments.checkout(
-            self.args.experiment, force=self.args.force
-        )
+        self.repo.experiments.checkout(self.args.experiment)
 
         return 0
 
 
 def _show_diff(
-    diff, title="", markdown=False, no_path=False, old=False, precision=None
+    diff,
+    title="",
+    markdown=False,
+    no_path=False,
+    old=False,
+    precision=DEFAULT_PRECISION,
 ):
     from dvc.utils.diff import table
 
-    if precision is None:
-        precision = DEFAULT_PRECISION
-
-    def _round(val):
-        if isinstance(val, float):
-            return round(val, precision)
-
-        return val
-
     rows = []
     for fname, diff_ in diff.items():
         sorted_diff = OrderedDict(sorted(diff_.items()))
@@ -169,9 +379,13 @@ def _show_diff(
             row = [] if no_path else [fname]
             row.append(item)
             if old:
-                row.append(_round(change.get("old")))
-            row.append(_round(change["new"]))
-            row.append(_round(change.get("diff", "diff not supported")))
+                row.append(_format_field(change.get("old"), precision))
+            row.append(_format_field(change["new"], precision))
+            row.append(
+                _format_field(
+                    change.get("diff", "diff not supported"), precision
+                )
+            )
             rows.append(row)
 
     header = [] if no_path else ["Path"]
@@ -202,6 +416,11 @@ class CmdExperimentsDiff(CmdBase):
 
                 logger.info(json.dumps(diff))
             else:
+                if self.args.precision is None:
+                    precision = DEFAULT_PRECISION
+                else:
+                    precision = self.args.precision
+
                 diffs = [("metrics", "Metric"), ("params", "Param")]
                 for key, title in diffs:
                     table = _show_diff(
@@ -210,7 +429,7 @@ class CmdExperimentsDiff(CmdBase):
                         markdown=self.args.show_md,
                         no_path=self.args.no_path,
                         old=self.args.old,
-                        precision=self.args.precision,
+                        precision=precision,
                     )
                     if table:
                         logger.info(table)
@@ -223,12 +442,103 @@ class CmdExperimentsDiff(CmdBase):
         return 0
 
 
+class CmdExperimentsRun(CmdRepro):
+    def run(self):
+        if not self.repo.experiments:
+            return 0
+
+        saved_dir = os.path.realpath(os.curdir)
+        os.chdir(self.args.cwd)
+
+        # Dirty hack so the for loop below can at least enter once
+        if self.args.all_pipelines:
+            self.args.targets = [None]
+        elif not self.args.targets:
+            self.args.targets = self.default_targets
+
+        ret = 0
+        for target in self.args.targets:
+            try:
+                self.repo.experiments.run(
+                    target,
+                    queue=self.args.queue,
+                    run_all=self.args.run_all,
+                    jobs=self.args.jobs,
+                    params=self.args.params,
+                    checkpoint_resume=self.args.checkpoint_resume,
+                    **self._repro_kwargs,
+                )
+            except DvcException:
+                logger.exception("")
+                ret = 1
+                break
+
+        os.chdir(saved_dir)
+        return ret
+
+
+class CmdExperimentsGC(CmdRepro):
+    def run(self):
+        from dvc.repo.gc import _raise_error_if_all_disabled
+
+        if not self.repo.experiments:
+            return 0
+
+        _raise_error_if_all_disabled(
+            all_branches=self.args.all_branches,
+            all_tags=self.args.all_tags,
+            all_commits=self.args.all_commits,
+            workspace=self.args.workspace,
+        )
+
+        msg = "This will remove all experiments except those derived from "
+
+        msg += "the workspace"
+        if self.args.all_commits:
+            msg += " and all git commits"
+        elif self.args.all_branches and self.args.all_tags:
+            msg += " and all git branches and tags"
+        elif self.args.all_branches:
+            msg += " and all git branches"
+        elif self.args.all_tags:
+            msg += " and all git tags"
+        msg += " of the current repo."
+        if self.args.queued:
+            msg += " Run queued experiments will be preserved."
+        if self.args.queued:
+            msg += " Run queued experiments will be removed."
+
+        logger.warning(msg)
+
+        msg = "Are you sure you want to proceed?"
+        if not self.args.force and not prompt.confirm(msg):
+            return 1
+
+        removed = self.repo.experiments.gc(
+            all_branches=self.args.all_branches,
+            all_tags=self.args.all_tags,
+            all_commits=self.args.all_commits,
+            workspace=self.args.workspace,
+            queued=self.args.queued,
+        )
+
+        if removed:
+            logger.info(
+                f"Removed {removed} experiments. To remove unused cache files "
+                "use 'dvc gc'."
+            )
+        else:
+            logger.info("No experiments to remove.")
+        return 0
+
+
 def add_parser(subparsers, parent_parser):
     EXPERIMENTS_HELP = "Commands to display and compare experiments."
 
     experiments_parser = subparsers.add_parser(
         "experiments",
         parents=[parent_parser],
+        aliases=["exp"],
         description=append_doc_link(EXPERIMENTS_HELP, "experiments"),
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
@@ -269,6 +579,78 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Show metrics for all commits.",
     )
+    experiments_show_parser.add_argument(
+        "--no-pager",
+        action="store_true",
+        default=False,
+        help="Do not pipe output into a pager.",
+    )
+    experiments_show_parser.add_argument(
+        "--include-metrics",
+        action="append",
+        default=[],
+        help="Include the specified metrics in output table.",
+        metavar="<metrics_list>",
+    )
+    experiments_show_parser.add_argument(
+        "--exclude-metrics",
+        action="append",
+        default=[],
+        help="Exclude the specified metrics from output table.",
+        metavar="<metrics_list>",
+    )
+    experiments_show_parser.add_argument(
+        "--include-params",
+        action="append",
+        default=[],
+        help="Include the specified params in output table.",
+        metavar="<params_list>",
+    )
+    experiments_show_parser.add_argument(
+        "--exclude-params",
+        action="append",
+        default=[],
+        help="Exclude the specified params from output table.",
+        metavar="<params_list>",
+    )
+    experiments_show_parser.add_argument(
+        "--sort-by",
+        help="Sort related experiments by the specified metric or param.",
+        metavar="<metric/param>",
+    )
+    experiments_show_parser.add_argument(
+        "--sort-order",
+        help="Sort order to use with --sort-by.",
+        choices=("asc", "desc"),
+        default="asc",
+    )
+    experiments_show_parser.add_argument(
+        "--no-timestamp",
+        action="store_true",
+        default=False,
+        help="Do not show experiment timestamps.",
+    )
+    experiments_show_parser.add_argument(
+        "--sha",
+        action="store_true",
+        default=False,
+        help="Always show git commit SHAs instead of branch/tag names.",
+    )
+    experiments_show_parser.add_argument(
+        "--show-json",
+        action="store_true",
+        default=False,
+        help="Print output in JSON format instead of a human-readable table.",
+    )
+    experiments_show_parser.add_argument(
+        "--precision",
+        type=int,
+        help=(
+            "Round metrics/params to `n` digits precision after the decimal "
+            f"point. Rounds to {DEFAULT_PRECISION} digits by default."
+        ),
+        metavar="<n>",
+    )
     experiments_show_parser.set_defaults(func=CmdExperimentsShow)
 
     EXPERIMENTS_CHECKOUT_HELP = "Checkout experiments."
@@ -281,14 +663,6 @@ def add_parser(subparsers, parent_parser):
         help=EXPERIMENTS_CHECKOUT_HELP,
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
-    experiments_checkout_parser.add_argument(
-        "-f",
-        "--force",
-        action="store_true",
-        default=False,
-        help="Overwrite your current workspace with changes from the "
-        "experiment.",
-    )
     experiments_checkout_parser.add_argument(
         "experiment", help="Checkout this experiment.",
     )
@@ -352,3 +726,137 @@ def add_parser(subparsers, parent_parser):
         metavar="<n>",
     )
     experiments_diff_parser.set_defaults(func=CmdExperimentsDiff)
+
+    EXPERIMENTS_RUN_HELP = (
+        "Reproduce complete or partial experiment pipelines."
+    )
+    experiments_run_parser = experiments_subparsers.add_parser(
+        "run",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_RUN_HELP, "experiments/run"),
+        help=EXPERIMENTS_RUN_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    _add_run_common(experiments_run_parser)
+    experiments_run_parser.add_argument(
+        "--checkpoint-resume", type=str, default=None, help=argparse.SUPPRESS,
+    )
+    experiments_run_parser.set_defaults(func=CmdExperimentsRun)
+
+    EXPERIMENTS_RESUME_HELP = "Resume checkpoint experiments."
+    experiments_resume_parser = experiments_subparsers.add_parser(
+        "resume",
+        parents=[parent_parser],
+        aliases=["res"],
+        description=append_doc_link(
+            EXPERIMENTS_RESUME_HELP, "experiments/resume"
+        ),
+        help=EXPERIMENTS_RESUME_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    _add_run_common(experiments_resume_parser)
+    experiments_resume_parser.add_argument(
+        "-r",
+        "--rev",
+        type=str,
+        default=Experiments.LAST_CHECKPOINT,
+        dest="checkpoint_resume",
+        help=(
+            "Continue the specified checkpoint experiment. "
+            "If no experiment revision is provided, "
+            "the most recently run checkpoint experiment will be used."
+        ),
+        metavar="<experiment_rev>",
+    )
+    experiments_resume_parser.set_defaults(func=CmdExperimentsRun)
+
+    EXPERIMENTS_GC_HELP = "Garbage collect unneeded experiments."
+    EXPERIMENTS_GC_DESCRIPTION = (
+        "Removes all experiments which are not derived from the specified"
+        "Git revisions."
+    )
+    experiments_gc_parser = experiments_subparsers.add_parser(
+        "gc",
+        parents=[parent_parser],
+        description=append_doc_link(
+            EXPERIMENTS_GC_DESCRIPTION, "experiments/gc"
+        ),
+        help=EXPERIMENTS_GC_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_gc_parser.add_argument(
+        "-w",
+        "--workspace",
+        action="store_true",
+        default=False,
+        help="Keep experiments derived from the current workspace.",
+    )
+    experiments_gc_parser.add_argument(
+        "-a",
+        "--all-branches",
+        action="store_true",
+        default=False,
+        help="Keep experiments derived from the tips of all Git branches.",
+    )
+    experiments_gc_parser.add_argument(
+        "-T",
+        "--all-tags",
+        action="store_true",
+        default=False,
+        help="Keep experiments derived from all Git tags.",
+    )
+    experiments_gc_parser.add_argument(
+        "--all-commits",
+        action="store_true",
+        default=False,
+        help="Keep experiments derived from all Git commits.",
+    )
+    experiments_gc_parser.add_argument(
+        "--queued",
+        action="store_true",
+        default=False,
+        help=(
+            "Keep queued experiments (experiments run queue will be cleared "
+            "by default)."
+        ),
+    )
+    experiments_gc_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Force garbage collection - automatically agree to all prompts.",
+    )
+    experiments_gc_parser.set_defaults(func=CmdExperimentsGC)
+
+
+def _add_run_common(parser):
+    """Add common args for 'exp run' and 'exp resume'."""
+    # inherit arguments from `dvc repro`
+    add_repro_arguments(parser)
+    parser.add_argument(
+        "--params",
+        action="append",
+        default=[],
+        help="Use the specified param values when reproducing pipelines.",
+        metavar="[<filename>:]<params_list>",
+    )
+    parser.add_argument(
+        "--queue",
+        action="store_true",
+        default=False,
+        help="Stage this experiment in the run queue for future execution.",
+    )
+    parser.add_argument(
+        "--run-all",
+        action="store_true",
+        default=False,
+        help="Execute all experiments in the run queue.",
+    )
+    parser.add_argument(
+        "-j",
+        "--jobs",
+        type=int,
+        help="Run the specified number of experiments at a time in parallel.",
+        metavar="<number>",
+    )

dvc/command/gc.py

@@ -123,7 +123,11 @@ def add_parser(subparsers, parent_parser):
         "-j",
         "--jobs",
         type=int,
-        help="Number of jobs to run simultaneously.",
+        help=(
+            "Number of jobs to run simultaneously. "
+            "The default value is 4 * cpu_count(). "
+            "For SSH remotes, the default is 4. "
+        ),
         metavar="<number>",
     )
     gc_parser.add_argument(

dvc/command/git_hook.py

@@ -61,6 +61,26 @@ class CmdPrePush(CmdHookBase):
         return main(["push"])
 
 
+class CmdMergeDriver(CmdHookBase):
+    def _run(self):
+        from dvc.dvcfile import Dvcfile
+        from dvc.repo import Repo
+
+        dvc = Repo()
+
+        try:
+            with dvc.state:
+                ancestor = Dvcfile(dvc, self.args.ancestor, verify=False)
+                our = Dvcfile(dvc, self.args.our, verify=False)
+                their = Dvcfile(dvc, self.args.their, verify=False)
+
+                our.merge(ancestor, their)
+
+                return 0
+        finally:
+            dvc.close()
+
+
 def add_parser(subparsers, parent_parser):
     GIT_HOOK_HELP = "Run GIT hook."
 
@@ -113,3 +133,27 @@ def add_parser(subparsers, parent_parser):
         "args", nargs="*", help="Arguments passed by GIT or pre-commit tool.",
     )
     pre_push_parser.set_defaults(func=CmdPrePush)
+
+    MERGE_DRIVER_HELP = "Run GIT merge driver."
+    merge_driver_parser = git_hook_subparsers.add_parser(
+        "merge-driver",
+        parents=[parent_parser],
+        description=MERGE_DRIVER_HELP,
+        help=MERGE_DRIVER_HELP,
+    )
+    merge_driver_parser.add_argument(
+        "--ancestor",
+        required=True,
+        help="Ancestor's version of the conflicting file.",
+    )
+    merge_driver_parser.add_argument(
+        "--our",
+        required=True,
+        help="Current version of the conflicting file.",
+    )
+    merge_driver_parser.add_argument(
+        "--their",
+        required=True,
+        help="Other branch's version of the conflicting file.",
+    )
+    merge_driver_parser.set_defaults(func=CmdMergeDriver)

dvc/command/imp.py

@@ -15,7 +15,10 @@ class CmdImport(CmdBase):
                 self.args.url,
                 self.args.path,
                 out=self.args.out,
+                fname=self.args.file,
                 rev=self.args.rev,
+                no_exec=self.args.no_exec,
+                desc=self.args.desc,
             )
         except DvcException:
             logger.exception(
@@ -59,4 +62,24 @@ def add_parser(subparsers, parent_parser):
         help="Git revision (e.g. SHA, branch, tag)",
         metavar="<commit>",
     )
+    import_parser.add_argument(
+        "--file",
+        help="Specify name of the DVC-file this command will generate.",
+        metavar="<filename>",
+    )
+    import_parser.add_argument(
+        "--no-exec",
+        action="store_true",
+        default=False,
+        help="Only create DVC-file without actually downloading it.",
+    )
+    import_parser.add_argument(
+        "--desc",
+        type=str,
+        metavar="<text>",
+        help=(
+            "User description of the data (optional). "
+            "This doesn't affect any DVC operations."
+        ),
+    )
     import_parser.set_defaults(func=CmdImport)

dvc/command/imp_url.py

@@ -16,6 +16,7 @@ class CmdImportUrl(CmdBase):
                 out=self.args.out,
                 fname=self.args.file,
                 no_exec=self.args.no_exec,
+                desc=self.args.desc,
             )
         except DvcException:
             logger.exception(
@@ -42,20 +43,16 @@ def add_parser(subparsers, parent_parser):
     )
     import_parser.add_argument(
         "url",
-        help="Supported urls:\n"
-        "/path/to/file\n"
-        "/path/to/directory\n"
-        "C:\\\\path\\to\\file\n"
-        "C:\\\\path\\to\\directory\n"
+        help="Location of the data to download. Supported URLs:\n"
+        "/absolute/path/to/file/or/dir\n"
+        "relative/path/to/file/or/dir\n"
+        "C:\\\\path\\to\\file\\or\\dir\n"
         "https://example.com/path/to/file\n"
-        "s3://bucket/path/to/file\n"
-        "s3://bucket/path/to/directory\n"
-        "gs://bucket/path/to/file\n"
-        "gs://bucket/path/to/directory\n"
+        "s3://bucket/key/path\n"
+        "gs://bucket/path/to/file/or/dir\n"
         "hdfs://example.com/path/to/file\n"
-        "ssh://example.com:/path/to/file\n"
-        "ssh://example.com:/path/to/directory\n"
-        "remote://myremote/path/to/file (see `dvc remote`)",
+        "ssh://example.com/absolute/path/to/file/or/dir\n"
+        "remote://remote_name/path/to/file/or/dir (see `dvc remote`)",
     )
     import_parser.add_argument(
         "out", nargs="?", help="Destination path to put files to.",
@@ -71,4 +68,13 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Only create DVC-file without actually downloading it.",
     )
+    import_parser.add_argument(
+        "--desc",
+        type=str,
+        metavar="<text>",
+        help=(
+            "User description of the data (optional). "
+            "This doesn't affect any DVC operations."
+        ),
+    )
     import_parser.set_defaults(func=CmdImportUrl)

dvc/command/init.py

@@ -1,15 +1,42 @@
 import argparse
 import logging
 
+import colorama
+
+from dvc import analytics
 from dvc.command.base import CmdBaseNoRepo, append_doc_link
+from dvc.utils import boxify
+from dvc.utils import format_link as fmt_link
 
 logger = logging.getLogger(__name__)
 
 
+def _welcome_message():
+    if analytics.is_enabled():
+        logger.info(
+            boxify(
+                "DVC has enabled anonymous aggregate usage analytics.\n"
+                "Read the analytics documentation (and how to opt-out) here:\n"
+                + fmt_link("https://dvc.org/doc/user-guide/analytics"),
+                border_color="red",
+            )
+        )
+
+    msg = (
+        "{yellow}What's next?{nc}\n"
+        "{yellow}------------{nc}\n"
+        f"- Check out the documentation: {fmt_link('https://dvc.org/doc')}\n"
+        f"- Get help and share ideas: {fmt_link('https://dvc.org/chat')}\n"
+        f"- Star us on GitHub: {fmt_link('https://github.com/iterative/dvc')}"
+    ).format(yellow=colorama.Fore.YELLOW, nc=colorama.Fore.RESET)
+
+    logger.info(msg)
+
+
 class CmdInit(CmdBaseNoRepo):
     def run(self):
-        from dvc.repo import Repo
         from dvc.exceptions import InitError
+        from dvc.repo import Repo
 
         try:
             self.repo = Repo.init(
@@ -19,6 +46,7 @@ class CmdInit(CmdBaseNoRepo):
                 subdir=self.args.subdir,
             )
             self.config = self.repo.config
+            _welcome_message()
         except InitError:
             logger.exception("failed to initiate DVC")
             return 1

dvc/command/install.py

@@ -2,6 +2,7 @@ import argparse
 import logging
 
 from dvc.command.base import CmdBase, append_doc_link
+from dvc.exceptions import DvcException
 
 logger = logging.getLogger(__name__)
 
@@ -10,7 +11,7 @@ class CmdInstall(CmdBase):
     def run(self):
         try:
             self.repo.install(self.args.use_pre_commit_tool)
-        except Exception:  # noqa, pylint:disable=broad-except
+        except DvcException:
             logger.exception("failed to install DVC Git hooks")
             return 1
         return 0

dvc/command/metrics.py

@@ -11,34 +11,41 @@ logger = logging.getLogger(__name__)
 DEFAULT_PRECISION = 5
 
 
-def show_metrics(
+def _show_metrics(
     metrics, all_branches=False, all_tags=False, all_commits=False
 ):
-    from flatten_json import flatten
     from dvc.utils.diff import format_dict
+    from dvc.utils.flatten import flatten
 
     # When `metrics` contains a `None` key, it means that some files
     # specified as `targets` in `repo.metrics.show` didn't contain any metrics.
     missing = metrics.pop(None, None)
 
+    lines = []
     for branch, val in metrics.items():
         if all_branches or all_tags or all_commits:
-            logger.info(f"{branch}:")
+            lines.append(f"{branch}:")
 
         for fname, metric in val.items():
             if not isinstance(metric, dict):
-                logger.info("\t{}: {}".format(fname, str(metric)))
+                lines.append("\t{}: {}".format(fname, str(metric)))
                 continue
 
-            logger.info(f"\t{fname}:")
-            for key, value in flatten(format_dict(metric), ".").items():
-                logger.info(f"\t\t{key}: {value}")
+            lines.append(f"\t{fname}:")
+            for key, value in flatten(format_dict(metric)).items():
+                lines.append(f"\t\t{key}: {value}")
 
     if missing:
         raise BadMetricError(missing)
 
+    return "\n".join(lines)
 
-class CmdMetricsShow(CmdBase):
+
+class CmdMetricsBase(CmdBase):
+    UNINITIALIZED = True
+
+
+class CmdMetricsShow(CmdMetricsBase):
     def run(self):
         try:
             metrics = self.repo.metrics.show(
@@ -54,20 +61,22 @@ class CmdMetricsShow(CmdBase):
 
                 logger.info(json.dumps(metrics))
             else:
-                show_metrics(
+                table = _show_metrics(
                     metrics,
                     self.args.all_branches,
                     self.args.all_tags,
                     self.args.all_commits,
                 )
+                if table:
+                    logger.info(table)
         except DvcException:
-            logger.exception("failed to show metrics")
+            logger.exception("")
             return 1
 
         return 0
 
 
-def _show_diff(diff, markdown=False, no_path=False, old=False, precision=None):
+def _show_diff(diff, markdown=False, no_path=False, precision=None):
     from collections import OrderedDict
 
     from dvc.utils.diff import table
@@ -87,24 +96,20 @@ def _show_diff(diff, markdown=False, no_path=False, old=False, precision=None):
         for metric, change in sorted_mdiff.items():
             row = [] if no_path else [fname]
             row.append(metric)
-            if old:
-                row.append(_round(change.get("old")))
+            row.append(_round(change.get("old")))
             row.append(_round(change["new"]))
-            row.append(_round(change.get("diff", "diff not supported")))
+            row.append(_round(change.get("diff")))
             rows.append(row)
 
     header = [] if no_path else ["Path"]
     header.append("Metric")
-    if old:
-        header.extend(["Old", "New"])
-    else:
-        header.append("Value")
+    header.extend(["Old", "New"])
     header.append("Change")
 
     return table(header, rows, markdown)
 
 
-class CmdMetricsDiff(CmdBase):
+class CmdMetricsDiff(CmdMetricsBase):
     def run(self):
         try:
             diff = self.repo.metrics.diff(
@@ -124,7 +129,6 @@ class CmdMetricsDiff(CmdBase):
                     diff,
                     self.args.show_md,
                     self.args.no_path,
-                    self.args.old,
                     precision=self.args.precision,
                 )
                 if table:
@@ -167,8 +171,8 @@ def add_parser(subparsers, parent_parser):
         "targets",
         nargs="*",
         help=(
-            "Limit command scope to these metric files. Using -R, "
-            "directories to search metric files in can also be given."
+            "Limit command scope to these metrics files. Using -R, "
+            "directories to search metrics files in can also be given."
         ),
     ).complete = completion.FILE
     metrics_show_parser.add_argument(
@@ -204,7 +208,7 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help=(
             "If any target is a directory, recursively search and process "
-            "metric files."
+            "metrics files."
         ),
     )
     metrics_show_parser.set_defaults(func=CmdMetricsShow)
@@ -232,8 +236,8 @@ def add_parser(subparsers, parent_parser):
         "--targets",
         nargs="*",
         help=(
-            "Limit command scope to these metric files. Using -R, "
-            "directories to search metric files in can also be given."
+            "Limit command scope to these metrics files. Using -R, "
+            "directories to search metrics files in can also be given."
         ),
         metavar="<paths>",
     ).complete = completion.FILE
@@ -244,7 +248,7 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help=(
             "If any target is a directory, recursively search and process "
-            "metric files."
+            "metrics files."
         ),
     )
     metrics_diff_parser.add_argument(
@@ -265,12 +269,6 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Show tabulated output in the Markdown format (GFM).",
     )
-    metrics_diff_parser.add_argument(
-        "--old",
-        action="store_true",
-        default=False,
-        help="Show old metric value.",
-    )
     metrics_diff_parser.add_argument(
         "--no-path",
         action="store_true",

dvc/command/plots.py

@@ -65,12 +65,11 @@ class CmdPlots(CmdBase):
             html = PAGE_HTML.format(divs="\n".join(divs))
             path = self.args.out or "plots.html"
 
+            path = os.path.join(os.getcwd(), path)
             with open(path, "w") as fobj:
                 fobj.write(html)
 
-            logger.info(
-                "file://{}".format(os.path.join(self.repo.root_dir, path))
-            )
+            logger.info(f"file://{path}")
 
         except DvcException:
             logger.exception("")
@@ -80,13 +79,22 @@ class CmdPlots(CmdBase):
 
 
 class CmdPlotsShow(CmdPlots):
+    UNINITIALIZED = True
+
     def _func(self, *args, **kwargs):
         return self.repo.plots.show(*args, **kwargs)
 
 
 class CmdPlotsDiff(CmdPlots):
+    UNINITIALIZED = True
+
     def _func(self, *args, **kwargs):
-        return self.repo.plots.diff(*args, revs=self.args.revisions, **kwargs)
+        return self.repo.plots.diff(
+            *args,
+            revs=self.args.revisions,
+            experiment=self.args.experiment,
+            **kwargs,
+        )
 
 
 class CmdPlotsModify(CmdPlots):
@@ -117,7 +125,7 @@ def add_parser(subparsers, parent_parser):
 
     fix_subparsers(plots_subparsers)
 
-    SHOW_HELP = "Generate plots from metric files."
+    SHOW_HELP = "Generate plots from metrics files."
     plots_show_parser = plots_subparsers.add_parser(
         "show",
         parents=[parent_parser],
@@ -128,7 +136,9 @@ def add_parser(subparsers, parent_parser):
     plots_show_parser.add_argument(
         "targets",
         nargs="*",
-        help="Plots files to visualize. Shows all plots by default.",
+        help="Files to visualize (supports any file, "
+        "even when not found as `plots` in `dvc.yaml`). "
+        "Shows all plots by default.",
     ).complete = completion.FILE
     _add_props_arguments(plots_show_parser)
     _add_output_arguments(plots_show_parser)
@@ -148,9 +158,18 @@ def add_parser(subparsers, parent_parser):
     plots_diff_parser.add_argument(
         "--targets",
         nargs="*",
-        help="Plots file to visualize. Shows all plots by default.",
+        help="Files to visualize (supports any file, "
+        "even when not found as `plots` in `dvc.yaml`). "
+        "Shows all plots by default.",
         metavar="<path>",
     ).complete = completion.FILE
+    plots_diff_parser.add_argument(
+        "-e",
+        "--experiment",
+        action="store_true",
+        default=False,
+        help=argparse.SUPPRESS,
+    )
     plots_diff_parser.add_argument(
         "revisions", nargs="*", default=None, help="Git commits to plot from",
     )
@@ -158,7 +177,7 @@ def add_parser(subparsers, parent_parser):
     _add_output_arguments(plots_diff_parser)
     plots_diff_parser.set_defaults(func=CmdPlotsDiff)
 
-    PLOTS_MODIFY_HELP = "Modify display properties of plot metric files."
+    PLOTS_MODIFY_HELP = "Modify display properties of plot metrics files."
     plots_modify_parser = plots_subparsers.add_parser(
         "modify",
         parents=[parent_parser],

dvc/command/remote.py

@@ -3,7 +3,7 @@ import logging
 
 from dvc.command.base import append_doc_link, fix_subparsers
 from dvc.command.config import CmdConfig
-from dvc.config import ConfigError
+from dvc.config import ConfigError, merge
 from dvc.utils import format_link
 
 logger = logging.getLogger(__name__)
@@ -18,7 +18,7 @@ class CmdRemote(CmdConfig):
 
     def _check_exists(self, conf):
         if self.args.name not in conf["remote"]:
-            raise ConfigError(f"remote '{self.args.name}' doesn't exists.")
+            raise ConfigError(f"remote '{self.args.name}' doesn't exist.")
 
 
 class CmdRemoteAdd(CmdRemote):
@@ -61,7 +61,12 @@ class CmdRemoteRemove(CmdRemote):
 class CmdRemoteModify(CmdRemote):
     def run(self):
         with self.config.edit(self.args.level) as conf:
-            self._check_exists(conf)
+            merged = self.config.load_config_to_level(self.args.level)
+            merge(merged, conf)
+            self._check_exists(merged)
+
+            if self.args.name not in conf["remote"]:
+                conf["remote"][self.args.name] = {}
             section = conf["remote"][self.args.name]
             if self.args.unset:
                 section.pop(self.args.option, None)
@@ -167,7 +172,7 @@ def add_parser(subparsers, parent_parser):
     remote_add_parser.add_argument("name", help="Name of the remote")
     remote_add_parser.add_argument(
         "url",
-        help="URL. See full list of supported urls at {}".format(
+        help="Remote location. See full list of supported URLs at {}".format(
             format_link("https://man.dvc.org/remote")
         ),
     )
@@ -196,7 +201,7 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     remote_default_parser.add_argument(
-        "name", nargs="?", help="Name of the remote."
+        "name", nargs="?", help="Name of the remote"
     )
     remote_default_parser.add_argument(
         "-u",
@@ -215,7 +220,7 @@ def add_parser(subparsers, parent_parser):
         help=REMOTE_MODIFY_HELP,
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
-    remote_modify_parser.add_argument("name", help="Name of the remote.")
+    remote_modify_parser.add_argument("name", help="Name of the remote")
     remote_modify_parser.add_argument(
         "option", help="Name of the option to modify."
     )

dvc/command/repro.py

@@ -4,7 +4,7 @@ import os
 
 from dvc.command import completion
 from dvc.command.base import CmdBase, append_doc_link
-from dvc.command.metrics import show_metrics
+from dvc.command.metrics import _show_metrics
 from dvc.command.status import CmdDataStatus
 from dvc.dvcfile import PIPELINE_FILE
 from dvc.exceptions import DvcException
@@ -15,8 +15,7 @@ logger = logging.getLogger(__name__)
 class CmdRepro(CmdBase):
     def run(self):
         saved_dir = os.path.realpath(os.curdir)
-        if self.args.cwd:
-            os.chdir(self.args.cwd)
+        os.chdir(self.args.cwd)
 
         # Dirty hack so the for loop below can at least enter once
         if self.args.all_pipelines:
@@ -27,28 +26,20 @@ class CmdRepro(CmdBase):
         ret = 0
         for target in self.args.targets:
             try:
-                stages = self.repo.reproduce(
-                    target,
-                    single_item=self.args.single_item,
-                    force=self.args.force,
-                    dry=self.args.dry,
-                    interactive=self.args.interactive,
-                    pipeline=self.args.pipeline,
-                    all_pipelines=self.args.all_pipelines,
-                    run_cache=not self.args.no_run_cache,
-                    no_commit=self.args.no_commit,
-                    downstream=self.args.downstream,
-                    recursive=self.args.recursive,
-                    force_downstream=self.args.force_downstream,
-                    experiment=self.args.experiment,
-                )
+                stages = self.repo.reproduce(target, **self._repro_kwargs)
 
                 if len(stages) == 0:
                     logger.info(CmdDataStatus.UP_TO_DATE_MSG)
+                else:
+                    logger.info(
+                        "Use `dvc push` to send your updates to "
+                        "remote storage."
+                    )
 
                 if self.args.metrics:
                     metrics = self.repo.metrics.show()
-                    show_metrics(metrics)
+                    logger.info(_show_metrics(metrics))
+
             except DvcException:
                 logger.exception("")
                 ret = 1
@@ -57,18 +48,25 @@ class CmdRepro(CmdBase):
         os.chdir(saved_dir)
         return ret
 
+    @property
+    def _repro_kwargs(self):
+        return {
+            "single_item": self.args.single_item,
+            "force": self.args.force,
+            "dry": self.args.dry,
+            "interactive": self.args.interactive,
+            "pipeline": self.args.pipeline,
+            "all_pipelines": self.args.all_pipelines,
+            "run_cache": not self.args.no_run_cache,
+            "no_commit": self.args.no_commit,
+            "downstream": self.args.downstream,
+            "recursive": self.args.recursive,
+            "force_downstream": self.args.force_downstream,
+            "pull": self.args.pull,
+        }
 
-def add_parser(subparsers, parent_parser):
-    REPRO_HELP = (
-        "Reproduce complete or partial pipelines by executing their stages."
-    )
-    repro_parser = subparsers.add_parser(
-        "repro",
-        parents=[parent_parser],
-        description=append_doc_link(REPRO_HELP, "repro"),
-        help=REPRO_HELP,
-        formatter_class=argparse.RawDescriptionHelpFormatter,
-    )
+
+def add_arguments(repro_parser):
     repro_parser.add_argument(
         "targets",
         nargs="*",
@@ -93,7 +91,8 @@ def add_parser(subparsers, parent_parser):
         "-c",
         "--cwd",
         default=os.path.curdir,
-        help="Directory within your repo to reproduce from.",
+        help="Directory within your repo to reproduce from. Note: deprecated "
+        "by `dvc --cd <path>`.",
         metavar="<path>",
     )
     repro_parser.add_argument(
@@ -168,10 +167,26 @@ def add_parser(subparsers, parent_parser):
         help="Start from the specified stages when reproducing pipelines.",
     )
     repro_parser.add_argument(
-        "-e",
-        "--experiment",
+        "--pull",
         action="store_true",
         default=False,
-        help=argparse.SUPPRESS,
+        help=(
+            "Try automatically pulling missing cache for outputs restored "
+            "from the run-cache."
+        ),
     )
+
+
+def add_parser(subparsers, parent_parser):
+    REPRO_HELP = (
+        "Reproduce complete or partial pipelines by executing their stages."
+    )
+    repro_parser = subparsers.add_parser(
+        "repro",
+        parents=[parent_parser],
+        description=append_doc_link(REPRO_HELP, "repro"),
+        help=REPRO_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    add_arguments(repro_parser)
     repro_parser.set_defaults(func=CmdRepro)

dvc/command/run.py

@@ -1,6 +1,7 @@
 import argparse
 import logging
 
+from dvc.command import completion
 from dvc.command.base import CmdBase, append_doc_link
 from dvc.exceptions import DvcException
 
@@ -20,6 +21,7 @@ class CmdRun(CmdBase):
                 self.args.plots_no_cache,
                 self.args.outs_persist,
                 self.args.outs_persist_no_cache,
+                self.args.checkpoints,
                 self.args.params,
                 self.args.command,
             ]
@@ -44,16 +46,18 @@ class CmdRun(CmdBase):
                 params=self.args.params,
                 fname=self.args.file,
                 wdir=self.args.wdir,
-                no_exec=self.args.no_exec,
+                no_exec=(self.args.no_exec or bool(self.args.checkpoints)),
                 force=self.args.force,
                 run_cache=not self.args.no_run_cache,
                 no_commit=self.args.no_commit,
                 outs_persist=self.args.outs_persist,
                 outs_persist_no_cache=self.args.outs_persist_no_cache,
+                checkpoints=self.args.checkpoints,
                 always_changed=self.args.always_changed,
                 name=self.args.name,
                 single_stage=self.args.single_stage,
                 external=self.args.external,
+                desc=self.args.desc,
             )
         except DvcException:
             logger.exception("")
@@ -99,7 +103,7 @@ def add_parser(subparsers, parent_parser):
         default=[],
         help="Declare dependencies for reproducible cmd.",
         metavar="<path>",
-    )
+    ).complete = completion.FILE
     run_parser.add_argument(
         "-n", "--name", help="Stage name.",
     )
@@ -110,7 +114,7 @@ def add_parser(subparsers, parent_parser):
         default=[],
         help="Declare output file or directory.",
         metavar="<filename>",
-    )
+    ).complete = completion.FILE
     run_parser.add_argument(
         "-O",
         "--outs-no-cache",
@@ -119,7 +123,7 @@ def add_parser(subparsers, parent_parser):
         help="Declare output file or directory "
         "(do not put into DVC cache).",
         metavar="<filename>",
-    )
+    ).complete = completion.FILE
     run_parser.add_argument(
         "-p",
         "--params",
@@ -127,13 +131,13 @@ def add_parser(subparsers, parent_parser):
         default=[],
         help="Declare parameter to use as additional dependency.",
         metavar="[<filename>:]<params_list>",
-    )
+    ).complete = completion.FILE
     run_parser.add_argument(
         "-m",
         "--metrics",
         action="append",
         default=[],
-        help="Declare output metric file.",
+        help="Declare output metrics file.",
         metavar="<path>",
     )
     run_parser.add_argument(
@@ -141,7 +145,7 @@ def add_parser(subparsers, parent_parser):
         "--metrics-no-cache",
         action="append",
         default=[],
-        help="Declare output metric file (do not put into DVC cache).",
+        help="Declare output metrics file (do not put into DVC cache).",
         metavar="<path>",
     )
     run_parser.add_argument(
@@ -211,6 +215,14 @@ def add_parser(subparsers, parent_parser):
         "removed upon repro (do not put into DVC cache).",
         metavar="<filename>",
     )
+    run_parser.add_argument(
+        "-c",
+        "--checkpoints",
+        action="append",
+        default=[],
+        help=argparse.SUPPRESS,
+        metavar="<filename>",
+    ).complete = completion.FILE
     run_parser.add_argument(
         "--always-changed",
         action="store_true",
@@ -229,6 +241,15 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Allow outputs that are outside of the DVC repository.",
     )
+    run_parser.add_argument(
+        "--desc",
+        type=str,
+        metavar="<text>",
+        help=(
+            "User description of the stage (optional). "
+            "This doesn't affect any DVC operations."
+        ),
+    )
     run_parser.add_argument(
         "command", nargs=argparse.REMAINDER, help="Command to execute."
     )

dvc/command/status.py

@@ -1,6 +1,8 @@
 import logging
 
 from dvc.command.data_sync import CmdDataBase
+from dvc.exceptions import DvcException
+from dvc.utils import format_link
 
 logger = logging.getLogger(__name__)
 
@@ -9,6 +11,11 @@ class CmdDataStatus(CmdDataBase):
     STATUS_LEN = 20
     STATUS_INDENT = "\t"
     UP_TO_DATE_MSG = "Data and pipelines are up to date."
+    IN_SYNC_MSG = "Cache and remote '{remote}' are in sync."
+    EMPTY_PROJECT_MSG = (
+        "There are no data or pipelines tracked in this project yet.\n"
+        "See {link} to get started!"
+    ).format(link=format_link("https://dvc.org/doc/start"))
 
     def _normalize(self, s):
         s += ":"
@@ -60,10 +67,17 @@ class CmdDataStatus(CmdDataBase):
                 logger.info(json.dumps(st))
             elif st:
                 self._show(st, indent)
+            elif not self.repo.stages:
+                logger.info(self.EMPTY_PROJECT_MSG)
+            elif self.args.cloud or self.args.remote:
+                remote = self.args.remote or self.repo.config["core"].get(
+                    "remote"
+                )
+                logger.info(self.IN_SYNC_MSG.format(remote=remote))
             else:
                 logger.info(self.UP_TO_DATE_MSG)
 
-        except Exception:  # noqa, pylint: disable=broad-except
-            logger.exception("failed to obtain data status")
+        except DvcException:
+            logger.exception("")
             return 1
         return 0

dvc/command/version.py

@@ -1,159 +1,18 @@
 import argparse
-import itertools
 import logging
-import os
-import pathlib
-import platform
-import uuid
 
 from dvc.command.base import CmdBaseNoRepo, append_doc_link
-from dvc.exceptions import DvcException, NotDvcRepoError
-from dvc.scm.base import SCMError
-from dvc.system import System
-from dvc.utils import relpath
-from dvc.utils.pkg import PKG
-from dvc.version import __version__
-
-try:
-    import psutil
-except ImportError:
-    psutil = None
-
+from dvc.info import get_dvc_info
 
 logger = logging.getLogger(__name__)
 
 
 class CmdVersion(CmdBaseNoRepo):
     def run(self):
-        from dvc.repo import Repo
-
-        package = PKG
-        if PKG is None:
-            package = ""
-        else:
-            package = f"({PKG})"
-
-        info = [
-            f"DVC version: {__version__} {package}",
-            "---------------------------------",
-            f"Platform: Python {platform.python_version()} on "
-            f"{platform.platform()}",
-            f"Supports: {self.get_supported_remotes()}",
-        ]
-
-        try:
-            repo = Repo()
-            root_directory = repo.root_dir
-
-            # cache_dir might not exist yet (e.g. after `dvc init`), and we
-            # can't auto-create it, as it might cause issues if the user
-            # later decides to enable shared cache mode with
-            # `dvc config cache.shared group`.
-            if os.path.exists(repo.cache.local.cache_dir):
-                info.append(
-                    "Cache types: {}".format(
-                        self.get_linktype_support_info(repo)
-                    )
-                )
-                if psutil:
-                    fs_type = self.get_fs_type(repo.cache.local.cache_dir)
-                    info.append(f"Cache directory: {fs_type}")
-            else:
-                logger.warning(
-                    "Unable to detect supported link types, as cache "
-                    "directory '{}' doesn't exist. It is usually auto-created "
-                    "by commands such as `dvc add/fetch/pull/run/import`, "
-                    "but you could create it manually to enable this "
-                    "check.".format(relpath(repo.cache.local.cache_dir))
-                )
-
-        except NotDvcRepoError:
-            root_directory = os.getcwd()
-        except SCMError:
-            root_directory = os.getcwd()
-            info.append("Repo: dvc, git (broken)")
-        else:
-            if psutil:
-                fs_root = self.get_fs_type(os.path.abspath(root_directory))
-                info.append(f"Workspace directory: {fs_root}")
-
-            info.append("Repo: {}".format(_get_dvc_repo_info(repo)))
-
-        logger.info("\n".join(info))
+        dvc_info = get_dvc_info()
+        logger.info(dvc_info)
         return 0
 
-    @staticmethod
-    def get_fs_type(path):
-        partition = {
-            pathlib.Path(part.mountpoint): (part.fstype + " on " + part.device)
-            for part in psutil.disk_partitions(all=True)
-        }
-
-        path = pathlib.Path(path)
-
-        for parent in itertools.chain([path], path.parents):
-            if parent in partition:
-                return partition[parent]
-        return ("unknown", "none")
-
-    @staticmethod
-    def get_linktype_support_info(repo):
-        links = {
-            "reflink": (System.reflink, None),
-            "hardlink": (System.hardlink, System.is_hardlink),
-            "symlink": (System.symlink, System.is_symlink),
-        }
-
-        fname = "." + str(uuid.uuid4())
-        src = os.path.join(repo.cache.local.cache_dir, fname)
-        open(src, "w").close()
-        dst = os.path.join(repo.root_dir, fname)
-
-        cache = []
-
-        for name, (link, is_link) in links.items():
-            try:
-                link(src, dst)
-                status = "supported"
-                if is_link and not is_link(dst):
-                    status = "broken"
-                os.unlink(dst)
-            except DvcException:
-                status = "not supported"
-
-            if status == "supported":
-                cache.append(name)
-        os.remove(src)
-
-        return ", ".join(cache)
-
-    @staticmethod
-    def get_supported_remotes():
-        from dvc.tree import TREES
-
-        supported_remotes = []
-        for tree_cls in TREES:
-            if not tree_cls.get_missing_deps():
-                supported_remotes.append(tree_cls.scheme)
-
-        if len(supported_remotes) == len(TREES):
-            return "All remotes"
-
-        if len(supported_remotes) == 1:
-            return supported_remotes
-
-        return ", ".join(supported_remotes)
-
-
-def _get_dvc_repo_info(repo):
-    if repo.config.get("core", {}).get("no_scm", False):
-        return "dvc (no_scm)"
-
-    if repo.root_dir != repo.scm.root_dir:
-        return "dvc (subdir), git"
-
-    return "dvc, git"
-
 
 def add_parser(subparsers, parent_parser):
     VERSION_HELP = (
@@ -165,5 +24,6 @@ def add_parser(subparsers, parent_parser):
         description=append_doc_link(VERSION_HELP, "version"),
         help=VERSION_HELP,
         formatter_class=argparse.RawDescriptionHelpFormatter,
+        aliases=["doctor"],
     )
     version_parser.set_defaults(func=CmdVersion)

dvc/config.py

@@ -103,6 +103,7 @@ class RelPath(str):
 REMOTE_COMMON = {
     "url": str,
     "checksum_jobs": All(Coerce(int), Range(1)),
+    "jobs": All(Coerce(int), Range(1)),
     Optional("no_traverse"): Bool,  # obsoleted
     "verify": Bool,
 }
@@ -118,7 +119,19 @@ HTTP_COMMON = {
     "user": str,
     "password": str,
     "ask_password": Bool,
+    "ssl_verify": Bool,
+    "method": str,
 }
+WEBDAV_COMMON = {
+    "user": str,
+    "password": str,
+    "ask_password": Bool,
+    "token": str,
+    "cert_path": str,
+    "key_path": str,
+    "timeout": Coerce(int),
+}
+
 SCHEMA = {
     "core": {
         "remote": Lower,
@@ -127,13 +140,16 @@ SCHEMA = {
         Optional("analytics", default=True): Bool,
         Optional("hardlink_lock", default=False): Bool,
         Optional("no_scm", default=False): Bool,
+        Optional("autostage", default=False): Bool,
         Optional("experiments", default=False): Bool,
+        Optional("check_update", default=True): Bool,
     },
     "cache": {
         "local": str,
         "s3": str,
         "gs": str,
         "hdfs": str,
+        "webhdfs": str,
         "ssh": str,
         "azure": str,
         # This is for default local cache
@@ -151,7 +167,7 @@ SCHEMA = {
                     "endpointurl": str,
                     "access_key_id": str,
                     "secret_access_key": str,
-                    Optional("listobjects", default=False): Bool,
+                    Optional("listobjects", default=False): Bool,  # obsoleted
                     Optional("use_ssl", default=True): Bool,
                     "sse": str,
                     "sse_kms_key_id": str,
@@ -176,9 +192,17 @@ SCHEMA = {
                     "keyfile": str,
                     "timeout": Coerce(int),
                     "gss_auth": Bool,
+                    "allow_agent": Bool,
                     **REMOTE_COMMON,
                 },
                 "hdfs": {"user": str, **REMOTE_COMMON},
+                "webhdfs": {
+                    "hdfscli_config": str,
+                    "webhdfs_token": str,
+                    "user": str,
+                    "webhdfs_alias": str,
+                    **REMOTE_COMMON,
+                },
                 "azure": {"connection_string": str, **REMOTE_COMMON},
                 "oss": {
                     "oss_key_id": str,
@@ -199,6 +223,8 @@ SCHEMA = {
                 },
                 "http": {**HTTP_COMMON, **REMOTE_COMMON},
                 "https": {**HTTP_COMMON, **REMOTE_COMMON},
+                "webdav": {**WEBDAV_COMMON, **REMOTE_COMMON},
+                "webdavs": {**WEBDAV_COMMON, **REMOTE_COMMON},
                 "remote": {str: object},  # Any of the above options are valid
             }
         )
@@ -207,6 +233,8 @@ SCHEMA = {
         "row_limit": All(Coerce(int), Range(1)),
         "row_cleanup_quota": All(Coerce(int), Range(0, 100)),
     },
+    # section for experimental features
+    "feature": {Optional("parametrization", default=False): Bool},
 }
 COMPILED_SCHEMA = Schema(SCHEMA)
 
@@ -258,7 +286,7 @@ class Config(dict):
 
     @classmethod
     def get_dir(cls, level):
-        from appdirs import user_config_dir, site_config_dir
+        from appdirs import site_config_dir, user_config_dir
 
         assert level in ("global", "system")
 
@@ -312,11 +340,17 @@ class Config(dict):
         if not self["cache"].get("dir") and self.dvc_dir:
             self["cache"]["dir"] = os.path.join(self.dvc_dir, "cache")
 
+    def _get_tree(self, level):
+        # NOTE: this might be a GitTree, which doesn't see things outside of
+        # the repo.
+        return self.tree if level == "repo" else self.wtree
+
     def _load_config(self, level):
         filename = self.files[level]
+        tree = self._get_tree(level)
 
-        if self.tree.exists(filename, use_dvcignore=False):
-            with self.tree.open(filename) as fobj:
+        if tree.exists(filename, use_dvcignore=False):
+            with tree.open(filename) as fobj:
                 conf_obj = configobj.ConfigObj(fobj)
         else:
             conf_obj = configobj.ConfigObj()
@@ -324,13 +358,14 @@ class Config(dict):
 
     def _save_config(self, level, conf_dict):
         filename = self.files[level]
+        tree = self._get_tree(level)
 
         logger.debug(f"Writing '{filename}'.")
 
-        self.tree.makedirs(os.path.dirname(filename))
+        tree.makedirs(os.path.dirname(filename))
 
         config = configobj.ConfigObj(_pack_remotes(conf_dict))
-        with self.tree.open(filename, "wb") as fobj:
+        with tree.open(filename, "wb") as fobj:
             config.write(fobj)
         config.filename = filename
 
@@ -351,6 +386,12 @@ class Config(dict):
         def resolve(path):
             if os.path.isabs(path) or re.match(r"\w+://", path):
                 return path
+
+            # on windows convert slashes to backslashes
+            # to have path compatible with abs_conf_dir
+            if os.path.sep == "\\" and "/" in path:
+                path = path.replace("/", "\\")
+
             return RelPath(os.path.join(abs_conf_dir, path))
 
         return Config._map_dirs(conf, resolve)
@@ -377,7 +418,15 @@ class Config(dict):
         dirs_schema = {
             "cache": {"dir": func},
             "remote": {
-                str: {"url": func, "gdrive_user_credentials_file": func}
+                str: {
+                    "url": func,
+                    "gdrive_user_credentials_file": func,
+                    "gdrive_service_account_p12_file_path": func,
+                    "credentialpath": func,
+                    "keyfile": func,
+                    "cert_path": func,
+                    "key_path": func,
+                }
             },
         }
         return Schema(dirs_schema, extra=ALLOW_EXTRA)(conf)
@@ -388,7 +437,7 @@ class Config(dict):
             if merge_level == level:
                 break
             if merge_level in self.files:
-                _merge(merged_conf, self.load_one(merge_level))
+                merge(merged_conf, self.load_one(merge_level))
         return merged_conf
 
     @contextmanager
@@ -402,7 +451,7 @@ class Config(dict):
         conf = self._save_paths(conf, self.files[level])
 
         merged_conf = self.load_config_to_level(level)
-        _merge(merged_conf, conf)
+        merge(merged_conf, conf)
         self.validate(merged_conf)
 
         self._save_config(level, conf)
@@ -441,11 +490,11 @@ def _pack_remotes(conf):
     return result
 
 
-def _merge(into, update):
+def merge(into, update):
     """Merges second dict into first recursively"""
     for key, val in update.items():
         if isinstance(into.get(key), dict) and isinstance(val, dict):
-            _merge(into[key], val)
+            merge(into[key], val)
         else:
             into[key] = val
 

dvc/daemon.py

@@ -25,7 +25,7 @@ def _popen(cmd, **kwargs):
 
 
 def _spawn_windows(cmd, env):
-    from subprocess import STARTUPINFO, STARTF_USESHOWWINDOW
+    from subprocess import STARTF_USESHOWWINDOW, STARTUPINFO
 
     creationflags = CREATE_NEW_PROCESS_GROUP | DETACHED_PROCESS
 

dvc/data_cloud.py

@@ -62,8 +62,11 @@ class DataCloud:
         """
         remote = self.get_remote(remote, "push")
 
-        return self.repo.cache.local.push(
-            cache, jobs=jobs, remote=remote, show_checksums=show_checksums,
+        return remote.push(
+            self.repo.cache.local,
+            cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
         )
 
     def pull(
@@ -81,26 +84,21 @@ class DataCloud:
         """
         remote = self.get_remote(remote, "pull")
 
-        downloaded_items_num = self.repo.cache.local.pull(
-            cache, jobs=jobs, remote=remote, show_checksums=show_checksums
+        return remote.pull(
+            self.repo.cache.local,
+            cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
         )
 
-        if not remote.tree.verify:
-            self._save_pulled_checksums(cache)
-
-        return downloaded_items_num
-
-    def _save_pulled_checksums(self, cache):
-        for checksum in cache.scheme_keys("local"):
-            cache_file = self.repo.cache.local.tree.hash_to_path_info(checksum)
-            if self.repo.cache.local.tree.exists(cache_file):
-                # We can safely save here, as existing corrupted files will
-                # be removed upon status, while files corrupted during
-                # download will not be moved from tmp_file
-                # (see `RemoteBASE.download()`)
-                self.repo.state.save(cache_file, checksum)
-
-    def status(self, cache, jobs=None, remote=None, show_checksums=False):
+    def status(
+        self,
+        cache,
+        jobs=None,
+        remote=None,
+        show_checksums=False,
+        log_missing=True,
+    ):
         """Check status of data items in a cloud-agnostic way.
 
         Args:
@@ -111,8 +109,18 @@ class DataCloud:
                 is used.
             show_checksums (bool): show checksums instead of file names in
                 information messages.
+            log_missing (bool): log warning messages if file doesn't exist
+                neither in cache, neither in cloud.
         """
         remote = self.get_remote(remote, "status")
-        return self.repo.cache.local.status(
-            cache, jobs=jobs, remote=remote, show_checksums=show_checksums
+        return remote.status(
+            self.repo.cache.local,
+            cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            log_missing=log_missing,
         )
+
+    def get_url_for(self, remote, checksum):
+        remote = self.get_remote(remote)
+        return str(remote.tree.hash_to_path_info(checksum))

dvc/dependency/__init__.py

@@ -11,6 +11,9 @@ from dvc.dependency.local import LocalDependency
 from dvc.dependency.param import ParamsDependency
 from dvc.dependency.s3 import S3Dependency
 from dvc.dependency.ssh import SSHDependency
+from dvc.dependency.webdav import WebDAVDependency
+from dvc.dependency.webdavs import WebDAVSDependency
+from dvc.dependency.webhdfs import WebHDFSDependency
 from dvc.output.base import BaseOutput
 from dvc.scheme import Schemes
 
@@ -25,6 +28,9 @@ DEPS = [
     HTTPSDependency,
     S3Dependency,
     SSHDependency,
+    WebDAVDependency,
+    WebDAVSDependency,
+    WebHDFSDependency,
     # NOTE: LocalDependency is the default choice
 ]
 
@@ -37,16 +43,20 @@ DEP_MAP = {
     Schemes.HDFS: HDFSDependency,
     Schemes.HTTP: HTTPDependency,
     Schemes.HTTPS: HTTPSDependency,
+    Schemes.WEBDAV: WebDAVDependency,
+    Schemes.WEBDAVS: WebDAVSDependency,
+    Schemes.WEBHDFS: WebHDFSDependency,
 }
 
 
 # NOTE: schema for dependencies is basically the same as for outputs, but
 # without output-specific entries like 'cache' (whether or not output is
 # cached, see -o and -O flags for `dvc run`) and 'metric' (whether or not
-# output is a metric file and how to parse it, see `-M` flag for `dvc run`).
+# output is a metrics file and how to parse it, see `-M` flag for `dvc run`).
 SCHEMA = output.SCHEMA.copy()
 del SCHEMA[BaseOutput.PARAM_CACHE]
 del SCHEMA[BaseOutput.PARAM_METRIC]
+del SCHEMA[BaseOutput.PARAM_DESC]
 SCHEMA.update(RepoDependency.REPO_SCHEMA)
 SCHEMA.update(ParamsDependency.PARAM_SCHEMA)
 

dvc/dependency/base.py

@@ -1,3 +1,5 @@
+from typing import Type
+
 from dvc.exceptions import DvcException
 
 
@@ -21,9 +23,11 @@ class DependencyIsStageFileError(DvcException):
 class BaseDependency:
     IS_DEPENDENCY = True
 
-    DoesNotExistError = DependencyDoesNotExistError
-    IsNotFileOrDirError = DependencyIsNotFileOrDirError
-    IsStageFileError = DependencyIsStageFileError
+    DoesNotExistError = DependencyDoesNotExistError  # type: Type[DvcException]
+    IsNotFileOrDirError = (
+        DependencyIsNotFileOrDirError
+    )  # type: Type[DvcException]
+    IsStageFileError = DependencyIsStageFileError  # type: Type[DvcException]
 
     def update(self, rev=None):
         pass

dvc/dependency/param.py

@@ -2,12 +2,12 @@ import os
 from collections import defaultdict
 
 import dpath.util
-import toml
-import yaml
 from voluptuous import Any
 
 from dvc.dependency.local import LocalDependency
 from dvc.exceptions import DvcException
+from dvc.hash_info import HashInfo
+from dvc.utils.serialize import LOADERS, ParseError
 
 
 class MissingParamsError(DvcException):
@@ -22,8 +22,6 @@ class ParamsDependency(LocalDependency):
     PARAM_PARAMS = "params"
     PARAM_SCHEMA = {PARAM_PARAMS: Any(dict, list, None)}
     DEFAULT_PARAMS_FILE = "params.yaml"
-    PARAMS_FILE_LOADERS = defaultdict(lambda: yaml.safe_load)
-    PARAMS_FILE_LOADERS.update({".toml": toml.load})
 
     def __init__(self, stage, path, params):
         info = {}
@@ -34,7 +32,7 @@ class ParamsDependency(LocalDependency):
             else:
                 assert isinstance(params, dict)
                 self.params = list(params.keys())
-                info = params
+                info = {self.PARAM_PARAMS: params}
 
         super().__init__(
             stage,
@@ -43,60 +41,61 @@ class ParamsDependency(LocalDependency):
             info=info,
         )
 
+    def dumpd(self):
+        ret = super().dumpd()
+        if not self.hash_info:
+            ret[self.PARAM_PARAMS] = self.params
+        return ret
+
     def fill_values(self, values=None):
         """Load params values dynamically."""
         if not values:
             return
+        info = {}
         for param in self.params:
             if param in values:
-                self.info[param] = values[param]
+                info[param] = values[param]
+        self.hash_info = HashInfo(self.PARAM_PARAMS, info)
 
-    def save(self):
-        super().save()
-        self.info = self.save_info()
+    def workspace_status(self):
+        status = super().workspace_status()
 
-    def status(self):
-        status = super().status()
-
-        if status[str(self)] == "deleted":
+        if status.get(str(self)) == "deleted":
             return status
 
         status = defaultdict(dict)
-        info = self.read_params()
+        info = self.hash_info.value if self.hash_info else {}
+        actual = self.read_params()
         for param in self.params:
-            if param not in info.keys():
+            if param not in actual.keys():
                 st = "deleted"
-            elif param not in self.info:
+            elif param not in info:
                 st = "new"
-            elif info[param] != self.info[param]:
+            elif actual[param] != info[param]:
                 st = "modified"
             else:
-                assert info[param] == self.info[param]
+                assert actual[param] == info[param]
                 continue
 
             status[str(self)][param] = st
 
         return status
 
-    def dumpd(self):
-        return {
-            self.PARAM_PATH: self.def_path,
-            self.PARAM_PARAMS: self.info or self.params,
-        }
+    def status(self):
+        return self.workspace_status()
 
     def read_params(self):
         if not self.exists:
             return {}
 
-        with self.repo.tree.open(self.path_info, "r") as fobj:
-            try:
-                config = self.PARAMS_FILE_LOADERS[
-                    self.path_info.suffix.lower()
-                ](fobj)
-            except (yaml.YAMLError, toml.TomlDecodeError) as exc:
-                raise BadParamFileError(
-                    f"Unable to read parameters from '{self}'"
-                ) from exc
+        suffix = self.path_info.suffix.lower()
+        loader = LOADERS[suffix]
+        try:
+            config = loader(self.path_info, tree=self.repo.tree)
+        except ParseError as exc:
+            raise BadParamFileError(
+                f"Unable to read parameters from '{self}'"
+            ) from exc
 
         ret = {}
         for param in self.params:
@@ -106,7 +105,7 @@ class ParamsDependency(LocalDependency):
                 pass
         return ret
 
-    def save_info(self):
+    def get_hash(self):
         info = self.read_params()
 
         missing_params = set(self.params) - set(info.keys())
@@ -117,4 +116,4 @@ class ParamsDependency(LocalDependency):
                 )
             )
 
-        return info
+        return HashInfo(self.PARAM_PARAMS, info)

dvc/dependency/repo.py

@@ -1,10 +1,5 @@
-import os
-
 from voluptuous import Required
 
-from dvc.exceptions import OutputNotFoundError
-from dvc.path_info import PathInfo
-
 from .local import LocalDependency
 
 
@@ -42,40 +37,31 @@ class RepoDependency(LocalDependency):
     def __str__(self):
         return "{} ({})".format(self.def_path, self.def_repo[self.PARAM_URL])
 
-    def _make_repo(self, *, locked=True):
+    def _make_repo(self, *, locked=True, **kwargs):
         from dvc.external_repo import external_repo
 
         d = self.def_repo
         rev = (d.get("rev_lock") if locked else None) or d.get("rev")
-        return external_repo(d["url"], rev=rev)
+        return external_repo(d["url"], rev=rev, **kwargs)
 
-    def _get_checksum(self, locked=True):
-        from dvc.repo.tree import RepoTree
+    def _get_hash(self, locked=True):
+        # we want stream but not fetch, so DVC out directories are
+        # walked, but dir contents is not fetched
+        with self._make_repo(locked=locked, fetch=False, stream=True) as repo:
+            return repo.get_checksum(self.def_path)
 
-        with self._make_repo(locked=locked) as repo:
-            try:
-                return repo.find_out_by_relpath(self.def_path).info["md5"]
-            except OutputNotFoundError:
-                path = PathInfo(os.path.join(repo.root_dir, self.def_path))
+    def workspace_status(self):
+        current = self._get_hash(locked=True)
+        updated = self._get_hash(locked=False)
 
-                # we want stream but not fetch, so DVC out directories are
-                # walked, but dir contents is not fetched
-                tree = RepoTree(repo, stream=True)
-
-                # We are polluting our repo cache with some dir listing here
-                if tree.isdir(path):
-                    return self.repo.cache.local.tree.get_hash(path, tree=tree)
-                return tree.get_file_hash(path)
-
-    def status(self):
-        current_checksum = self._get_checksum(locked=True)
-        updated_checksum = self._get_checksum(locked=False)
-
-        if current_checksum != updated_checksum:
+        if current != updated:
             return {str(self): "update available"}
 
         return {}
 
+    def status(self):
+        return self.workspace_status()
+
     def save(self):
         pass
 
@@ -83,14 +69,15 @@ class RepoDependency(LocalDependency):
         return {self.PARAM_PATH: self.def_path, self.PARAM_REPO: self.def_repo}
 
     def download(self, to):
-        with self._make_repo() as repo:
+        cache = self.repo.cache.local
+
+        with self._make_repo(cache_dir=cache.cache_dir) as repo:
             if self.def_repo.get(self.PARAM_REV_LOCK) is None:
                 self.def_repo[self.PARAM_REV_LOCK] = repo.get_rev()
 
-            cache = self.repo.cache.local
-            with repo.use_cache(cache):
-                _, _, cache_infos = repo.fetch_external([self.def_path])
-            cache.checkout(to.path_info, cache_infos[0])
+            _, _, cache_infos = repo.fetch_external([self.def_path])
+
+        cache.checkout(to.path_info, cache_infos[0])
 
     def update(self, rev=None):
         if rev:

dvc/dependency/webdav.py

@@ -0,0 +1,8 @@
+from dvc.dependency.base import BaseDependency
+from dvc.output.base import BaseOutput
+
+from ..tree.webdav import WebDAVTree
+
+
+class WebDAVDependency(BaseDependency, BaseOutput):
+    TREE_CLS = WebDAVTree

dvc/dependency/webdavs.py

@@ -0,0 +1,6 @@
+from ..tree.webdavs import WebDAVSTree
+from .webdav import WebDAVDependency
+
+
+class WebDAVSDependency(WebDAVDependency):
+    TREE_CLS = WebDAVSTree

dvc/dependency/webhdfs.py

@@ -0,0 +1,6 @@
+from dvc.dependency.base import BaseDependency
+from dvc.output.webhdfs import WebHDFSOutput
+
+
+class WebHDFSDependency(BaseDependency, WebHDFSOutput):
+    pass

dvc/dir_info.py

@@ -0,0 +1,102 @@
+import posixpath
+from operator import itemgetter
+
+from pygtrie import Trie
+
+from .hash_info import HashInfo
+
+
+def _diff(ancestor, other, allow_removed=False):
+    from dictdiffer import diff
+
+    from dvc.exceptions import MergeError
+
+    allowed = ["add"]
+    if allow_removed:
+        allowed.append("remove")
+
+    result = list(diff(ancestor, other))
+    for typ, _, _ in result:
+        if typ not in allowed:
+            raise MergeError(
+                "unable to auto-merge directories with diff that contains "
+                f"'{typ}'ed files"
+            )
+    return result
+
+
+def _merge(ancestor, our, their):
+    import copy
+
+    from dictdiffer import patch
+
+    our_diff = _diff(ancestor, our)
+    if not our_diff:
+        return copy.deepcopy(their)
+
+    their_diff = _diff(ancestor, their)
+    if not their_diff:
+        return copy.deepcopy(our)
+
+    # make sure there are no conflicting files
+    _diff(our, their, allow_removed=True)
+
+    return patch(our_diff + their_diff, ancestor)
+
+
+class DirInfo:
+    PARAM_RELPATH = "relpath"
+
+    def __init__(self):
+        self.trie = Trie()
+
+    @property
+    def size(self):
+        try:
+            return sum(
+                hash_info.size
+                for _, hash_info in self.trie.iteritems()  # noqa: B301
+            )
+        except TypeError:
+            return None
+
+    @property
+    def nfiles(self):
+        return len(self.trie)
+
+    def items(self, path_info=None):
+        for key, hash_info in self.trie.iteritems():  # noqa: B301
+            path = posixpath.sep.join(key)
+            if path_info is not None:
+                path = path_info / path
+            yield path, hash_info
+
+    @classmethod
+    def from_list(cls, lst):
+        ret = DirInfo()
+        for _entry in lst:
+            entry = _entry.copy()
+            relpath = entry.pop(cls.PARAM_RELPATH)
+            parts = tuple(relpath.split(posixpath.sep))
+            ret.trie[parts] = HashInfo.from_dict(entry)
+        return ret
+
+    def to_list(self):
+        # Sorting the list by path to ensure reproducibility
+        return sorted(
+            (
+                {
+                    # NOTE: not using hash_info.to_dict() because we don't want
+                    # size/nfiles fields at this point.
+                    hash_info.name: hash_info.value,
+                    self.PARAM_RELPATH: posixpath.sep.join(parts),
+                }
+                for parts, hash_info in self.trie.iteritems()  # noqa: B301
+            ),
+            key=itemgetter(self.PARAM_RELPATH),
+        )
+
+    def merge(self, ancestor, their):
+        merged = DirInfo()
+        merged.trie = _merge(ancestor.trie, self.trie, their.trie)
+        return merged

dvc/dvcfile.py

@@ -16,7 +16,13 @@ from dvc.stage.exceptions import (
 from dvc.stage.loader import SingleStageLoader, StageLoader
 from dvc.utils import relpath
 from dvc.utils.collections import apply_diff
-from dvc.utils.yaml import dump_yaml, parse_yaml, parse_yaml_for_update
+from dvc.utils.serialize import (
+    dump_yaml,
+    load_yaml,
+    modify_yaml,
+    parse_yaml,
+    parse_yaml_for_update,
+)
 
 logger = logging.getLogger(__name__)
 
@@ -30,6 +36,10 @@ class LockfileCorruptedError(DvcException):
     pass
 
 
+class ParametrizedDumpError(DvcException):
+    pass
+
+
 def is_valid_filename(path):
     return path.endswith(DVC_FILE_SUFFIX) or os.path.basename(path) in [
         DVC_FILE,
@@ -39,10 +49,14 @@ def is_valid_filename(path):
 
 def is_dvc_file(path):
     return os.path.isfile(path) and (
-        is_valid_filename(path) or os.path.basename(path) == PIPELINE_LOCK
+        is_valid_filename(path) or is_lock_file(path)
     )
 
 
+def is_lock_file(path):
+    return os.path.basename(path) == PIPELINE_LOCK
+
+
 def check_dvc_filename(path):
     if not is_valid_filename(path):
         raise StageFileBadNameError(
@@ -56,9 +70,10 @@ def check_dvc_filename(path):
 class FileMixin:
     SCHEMA = None
 
-    def __init__(self, repo, path, **kwargs):
+    def __init__(self, repo, path, verify=True, **kwargs):
         self.repo = repo
         self.path = path
+        self.verify = verify
 
     def __repr__(self):
         return "{}: {}".format(
@@ -90,7 +105,8 @@ class FileMixin:
         # 3. path doesn't represent a regular file
         if not self.exists():
             raise StageFileDoesNotExistError(self.path)
-        check_dvc_filename(self.path)
+        if self.verify:
+            check_dvc_filename(self.path)
         if not self.repo.tree.isfile(self.path):
             raise StageFileIsNotDvcFileError(self.path)
 
@@ -115,6 +131,9 @@ class FileMixin:
     def dump(self, stage, **kwargs):
         raise NotImplementedError
 
+    def merge(self, ancestor, other):
+        raise NotImplementedError
+
 
 class SingleStageFile(FileMixin):
     from dvc.schema import COMPILED_SINGLE_STAGE_SCHEMA as SCHEMA
@@ -134,7 +153,8 @@ class SingleStageFile(FileMixin):
         from dvc.stage import PipelineStage
 
         assert not isinstance(stage, PipelineStage)
-        check_dvc_filename(self.path)
+        if self.verify:
+            check_dvc_filename(self.path)
         logger.debug(
             "Saving information to '{file}'.".format(file=relpath(self.path))
         )
@@ -144,6 +164,14 @@ class SingleStageFile(FileMixin):
     def remove_stage(self, stage):  # pylint: disable=unused-argument
         self.remove()
 
+    def merge(self, ancestor, other):
+        assert isinstance(ancestor, SingleStageFile)
+        assert isinstance(other, SingleStageFile)
+
+        stage = self.stage
+        stage.merge(ancestor.stage, other.stage)
+        self.dump(stage)
+
 
 class PipelineFile(FileMixin):
     """Abstraction for pipelines file, .yaml + .lock combined."""
@@ -155,48 +183,52 @@ class PipelineFile(FileMixin):
         return Lockfile(self.repo, os.path.splitext(self.path)[0] + ".lock")
 
     def dump(
-        self, stage, update_pipeline=False, no_lock=False, **kwargs
+        self, stage, update_pipeline=True, update_lock=True, **kwargs
     ):  # pylint: disable=arguments-differ
         """Dumps given stage appropriately in the dvcfile."""
         from dvc.stage import PipelineStage
 
         assert isinstance(stage, PipelineStage)
-        check_dvc_filename(self.path)
+        if self.verify:
+            check_dvc_filename(self.path)
 
         if update_pipeline and not stage.is_data_source:
             self._dump_pipeline_file(stage)
 
-        if not no_lock:
+        if update_lock:
             self._dump_lockfile(stage)
 
     def _dump_lockfile(self, stage):
         self._lockfile.dump(stage)
 
+    @staticmethod
+    def _check_if_parametrized(stage):
+        if stage.raw_data.parametrized:
+            raise ParametrizedDumpError(f"cannot dump a parametrized {stage}")
+
     def _dump_pipeline_file(self, stage):
-        data = {}
-        if self.exists():
-            with open(self.path) as fd:
-                data = parse_yaml_for_update(fd.read(), self.path)
-        else:
-            logger.info("Creating '%s'", self.relpath)
-            open(self.path, "w+").close()
-
-        data["stages"] = data.get("stages", {})
+        self._check_if_parametrized(stage)
         stage_data = serialize.to_pipeline_file(stage)
-        existing_entry = stage.name in data["stages"]
 
-        action = "Modifying" if existing_entry else "Adding"
-        logger.info("%s stage '%s' in '%s'", action, stage.name, self.relpath)
+        with modify_yaml(self.path, tree=self.repo.tree) as data:
+            if not data:
+                logger.info("Creating '%s'", self.relpath)
 
-        if existing_entry:
-            orig_stage_data = data["stages"][stage.name]
-            if "meta" in orig_stage_data:
-                stage_data[stage.name]["meta"] = orig_stage_data["meta"]
-            apply_diff(stage_data[stage.name], orig_stage_data)
-        else:
-            data["stages"].update(stage_data)
+            data["stages"] = data.get("stages", {})
+            existing_entry = stage.name in data["stages"]
+            action = "Modifying" if existing_entry else "Adding"
+            logger.info(
+                "%s stage '%s' in '%s'", action, stage.name, self.relpath
+            )
+
+            if existing_entry:
+                orig_stage_data = data["stages"][stage.name]
+                if "meta" in orig_stage_data:
+                    stage_data[stage.name]["meta"] = orig_stage_data["meta"]
+                apply_diff(stage_data[stage.name], orig_stage_data)
+            else:
+                data["stages"].update(stage_data)
 
-        dump_yaml(self.path, data)
         self.repo.scm.track_file(self.relpath)
 
     @property
@@ -209,7 +241,7 @@ class PipelineFile(FileMixin):
     def stages(self):
         data, _ = self._load()
         lockfile_data = self._lockfile.load()
-        return StageLoader(self, data.get("stages", {}), lockfile_data)
+        return StageLoader(self, data, lockfile_data)
 
     def remove(self, force=False):
         if not force:
@@ -239,6 +271,9 @@ class PipelineFile(FileMixin):
         else:
             super().remove()
 
+    def merge(self, ancestor, other):
+        raise NotImplementedError
+
 
 class Lockfile(FileMixin):
     from dvc.schema import COMPILED_LOCKFILE_SCHEMA as SCHEMA
@@ -246,8 +281,8 @@ class Lockfile(FileMixin):
     def load(self):
         if not self.exists():
             return {}
-        with self.repo.tree.open(self.path) as fd:
-            data = parse_yaml(fd.read(), self.path)
+
+        data = load_yaml(self.path, tree=self.repo.tree)
         try:
             self.validate(data, fname=self.relpath)
         except StageFileFormatError:
@@ -258,21 +293,18 @@ class Lockfile(FileMixin):
 
     def dump(self, stage, **kwargs):
         stage_data = serialize.to_lockfile(stage)
-        if not self.exists():
-            modified = True
-            logger.info("Generating lock file '%s'", self.relpath)
-            data = stage_data
-            open(self.path, "w+").close()
-        else:
-            with self.repo.tree.open(self.path, "r") as fd:
-                data = parse_yaml_for_update(fd.read(), self.path)
+
+        with modify_yaml(self.path, tree=self.repo.tree) as data:
+            if not data:
+                logger.info("Generating lock file '%s'", self.relpath)
+
             modified = data.get(stage.name, {}) != stage_data.get(
                 stage.name, {}
             )
             if modified:
                 logger.info("Updating lock file '%s'", self.relpath)
             data.update(stage_data)
-        dump_yaml(self.path, data)
+
         if modified:
             self.repo.scm.track_file(self.relpath)
 
@@ -295,6 +327,9 @@ class Lockfile(FileMixin):
         else:
             self.remove()
 
+    def merge(self, ancestor, other):
+        raise NotImplementedError
+
 
 class Dvcfile:
     def __new__(cls, repo, path, **kwargs):

dvc/env.py

@@ -1,2 +1,4 @@
+DVC_CHECKPOINT = "DVC_CHECKPOINT"
 DVC_DAEMON = "DVC_DAEMON"
 DVC_PAGER = "DVC_PAGER"
+DVC_ROOT = "DVC_ROOT"

dvc/exceptions.py

@@ -1,7 +1,7 @@
 """Exceptions raised by the dvc."""
 from funcy import first
 
-from dvc.utils import format_link, relpath
+from dvc.utils import error_link, format_link, relpath
 
 
 class DvcException(Exception):
@@ -42,7 +42,7 @@ class OutputDuplicationError(DvcException):
 
 
 class OutputNotFoundError(DvcException):
-    """Thrown if a file/directory not found in repository pipelines.
+    """Thrown if a file/directory is not found as an output in any pipeline.
 
     Args:
         output (unicode): path to the file/directory.
@@ -105,7 +105,7 @@ class ArgumentDuplicationError(DvcException):
 
 
 class MoveNotDataSourceError(DvcException):
-    """Thrown if attempted to move a file/directory that is not an output
+    """Thrown when trying to move a file/directory that is not an output
     in a data source stage.
 
     Args:
@@ -164,33 +164,30 @@ class BadMetricError(DvcException):
     def __init__(self, paths):
         super().__init__(
             "the following metrics do not exist, "
-            "are not metric files or are malformed: {paths}".format(
+            "are not metrics files or are malformed: {paths}".format(
                 paths=", ".join(f"'{path}'" for path in paths)
             )
         )
 
 
 class NoMetricsError(DvcException):
-    def __init__(self):
+    pass
+
+
+class NoMetricsParsedError(NoMetricsError):
+    def __init__(self, command):
         super().__init__(
-            "no metric files in this repository. Use `-m/-M` options for "
-            "`dvc run` to mark stage outputs as  metrics."
+            f"Could not parse {command} files. Use `-v` option to see more "
+            "details."
         )
 
 
-class NoPlotsError(DvcException):
-    def __init__(self):
+class NoMetricsFoundError(NoMetricsError):
+    def __init__(self, command, run_options):
         super().__init__(
-            "no plots in this repository. Use `--plots/--plots-no-cache` "
-            "options for `dvc run` to mark stage outputs as plots."
-        )
-
-
-class YAMLFileCorruptedError(DvcException):
-    def __init__(self, path):
-        path = relpath(path)
-        super().__init__(
-            f"unable to read: '{path}', YAML file structure is corrupted"
+            f"No {command} files in this repository. "
+            f"Use `{run_options}` options for "
+            f"`dvc run` to mark stage outputs as {command}."
         )
 
 
@@ -230,6 +227,11 @@ class FileMissingError(DvcException):
         )
 
 
+class FileOwnershipError(DvcException):
+    def __init__(self, path):
+        super().__init__(f"file '{path}' not owned by user! ")
+
+
 class DvcIgnoreInCollectedDirError(DvcException):
     def __init__(self, ignore_dirname):
         super().__init__(
@@ -268,8 +270,10 @@ class CheckoutError(DvcException):
         self.stats = stats
         targets = [str(t) for t in target_infos]
         m = (
-            "Checkout failed for following targets:\n{}\nDid you "
-            "forget to fetch?".format("\n".join(targets))
+            "Checkout failed for following targets:\n{}\nIs your "
+            "cache up to date?\n{}".format(
+                "\n".join(targets), error_link("missing-files"),
+            )
         )
         super().__init__(m)
 
@@ -344,3 +348,22 @@ class NoOutputOrStageError(DvcException):
             f"'{target}' "
             f"does not exist as an output or a stage name in '{file}'"
         )
+
+
+class MergeError(DvcException):
+    pass
+
+
+class CacheLinkError(DvcException):
+    SUPPORT_LINK = "See {} for more information.".format(
+        format_link(
+            "https://dvc.org/doc/user-guide/troubleshooting#cache-types"
+        )
+    )
+
+    def __init__(self, path_infos):
+        msg = "No possible cache link types for '{}'. {}".format(
+            ", ".join([str(path) for path in path_infos]), self.SUPPORT_LINK,
+        )
+        super().__init__(msg)
+        self.path_infos = path_infos

dvc/external_repo.py

@@ -4,12 +4,14 @@ import tempfile
 import threading
 from contextlib import contextmanager
 from distutils.dir_util import copy_tree
-from typing import Iterable
+from typing import Dict, Iterable
 
-from funcy import cached_property, retry, wrap_with
+from funcy import cached_property, reraise, retry, wrap_with
 
+from dvc.cache import Cache
 from dvc.config import NoRemoteError, NotDvcRepoError
 from dvc.exceptions import (
+    DvcException,
     FileMissingError,
     NoOutputInExternalRepoError,
     NoRemoteInExternalRepoError,
@@ -18,47 +20,61 @@ from dvc.exceptions import (
 )
 from dvc.path_info import PathInfo
 from dvc.repo import Repo
-from dvc.repo.tree import RepoTree
 from dvc.scm.base import CloneError
 from dvc.scm.git import Git
 from dvc.tree.local import LocalTree
+from dvc.tree.repo import RepoTree
+from dvc.utils import relpath
 from dvc.utils.fs import remove
 
 logger = logging.getLogger(__name__)
 
 
+class IsADVCRepoError(DvcException):
+    """Raised when it is not expected to be a dvc repo."""
+
+
 @contextmanager
-def external_repo(url, rev=None, for_write=False):
+def external_repo(url, rev=None, for_write=False, **kwargs):
     logger.debug("Creating external repo %s@%s", url, rev)
     path = _cached_clone(url, rev, for_write=for_write)
-    if not rev:
-        # Local HEAD points to the tip of whatever branch we first cloned from
-        # (which may not be the default branch), use origin/HEAD here to get
-        # the tip of the default branch
-        rev = "refs/remotes/origin/HEAD"
-    try:
-        repo = ExternalRepo(path, url, rev, for_write=for_write)
-    except NotDvcRepoError:
-        repo = ExternalGitRepo(path, url, rev)
+    # Local HEAD points to the tip of whatever branch we first cloned from
+    # (which may not be the default branch), use origin/HEAD here to get
+    # the tip of the default branch
+    rev = rev or "refs/remotes/origin/HEAD"
+
+    root_dir = path if for_write else os.path.realpath(path)
+    conf = dict(
+        root_dir=root_dir,
+        url=url,
+        scm=None if for_write else Git(root_dir),
+        rev=None if for_write else rev,
+        for_write=for_write,
+        uninitialized=True,
+        **kwargs,
+    )
+    repo = ExternalRepo(**conf)
 
     try:
         yield repo
-    except NoRemoteError:
-        raise NoRemoteInExternalRepoError(url)
+    except NoRemoteError as exc:
+        raise NoRemoteInExternalRepoError(url) from exc
     except OutputNotFoundError as exc:
         if exc.repo is repo:
-            raise NoOutputInExternalRepoError(exc.output, repo.root_dir, url)
+            raise NoOutputInExternalRepoError(
+                exc.output, repo.root_dir, url
+            ) from exc
         raise
     except FileMissingError as exc:
-        raise PathMissingError(exc.path, url)
+        raise PathMissingError(exc.path, url) from exc
     finally:
         repo.close()
         if for_write:
             _remove(path)
 
 
-CLONES = {}
-CACHE_DIRS = {}
+CLONES: Dict[str, str] = {}
+CACHE_DIRS: Dict[str, str] = {}
 
 
 def clean_repos():
@@ -71,45 +87,67 @@ def clean_repos():
         _remove(path)
 
 
-class BaseExternalRepo:
+class ExternalRepo(Repo):
     # pylint: disable=no-member
 
-    _local_cache = None
+    def __init__(
+        self,
+        root_dir,
+        url,
+        scm=None,
+        rev=None,
+        for_write=False,
+        cache_dir=None,
+        cache_types=None,
+        uninitialized=False,
+        **kwargs,
+    ):
+        super().__init__(
+            root_dir, scm=scm, rev=rev, uninitialized=uninitialized
+        )
+
+        self.url = url
+        self.for_write = for_write
+        self.cache_dir = cache_dir or self._get_cache_dir()
+        self.cache_types = cache_types
+
+        self._setup_cache(self)
+        self._fix_upstream(self)
+        self.tree_confs = kwargs
 
     def __str__(self):
         return self.url
 
-    @property
-    def local_cache(self):
-        if hasattr(self, "cache"):
-            return self.cache.local
-        return self._local_cache
-
-    @contextmanager
-    def use_cache(self, cache):
-        """Use the specified cache in place of default tmpdir cache for
-        download operations.
-        """
-        if hasattr(self, "cache"):
-            save_cache = self.cache.local
-            self.cache.local = cache
-        self._local_cache = cache
-
-        yield
-
-        if hasattr(self, "cache"):
-            self.cache.local = save_cache
-        self._local_cache = None
-
     @cached_property
     def repo_tree(self):
-        return RepoTree(self, fetch=True)
+        return self._get_tree_for(
+            self, subrepos=not self.for_write, repo_factory=self.make_repo
+        )
 
     def get_rev(self):
+        assert self.scm
         if isinstance(self.tree, LocalTree):
             return self.scm.get_rev()
         return self.tree.rev
 
+    def _fetch_to_cache(self, path_info, repo, callback, **kwargs):
+        # don't support subrepo traversal as it might fail due to difference
+        # in remotes
+        tree = self._get_tree_for(repo)
+        cache = repo.cache.local
+
+        hash_info = tree.get_hash(
+            path_info, download_callback=callback, **kwargs
+        )
+        cache.save(
+            path_info,
+            tree,
+            hash_info,
+            save_link=False,
+            download_callback=callback,
+        )
+        return hash_info
+
     def fetch_external(self, paths: Iterable, **kwargs):
         """Fetch specified external repo paths into cache.
 
@@ -120,130 +158,135 @@ class BaseExternalRepo:
         """
         download_results = []
         failed = 0
+        root = PathInfo(self.root_dir)
 
-        paths = [PathInfo(self.root_dir) / path for path in paths]
+        paths = [root / path for path in paths]
 
         def download_update(result):
             download_results.append(result)
 
-        save_infos = []
+        hash_infos = []
         for path in paths:
-            if not self.repo_tree.exists(path):
-                raise PathMissingError(path, self.url)
-            save_info = self.local_cache.save(
-                path,
-                self.repo_tree,
-                None,
-                save_link=False,
-                download_callback=download_update,
-            )
-            save_infos.append(save_info)
+            with reraise(FileNotFoundError, PathMissingError(path, self.url)):
+                metadata = self.repo_tree.metadata(path)
 
-        return sum(download_results), failed, save_infos
+            self._check_repo(path, metadata.repo)
+            repo = metadata.repo
+            hash_info = self._fetch_to_cache(
+                path, repo, download_update, **kwargs
+            )
+            hash_infos.append(hash_info)
+
+        return sum(download_results), failed, hash_infos
+
+    def _check_repo(self, path_info, repo):
+        if not repo:
+            return
+
+        repo_path = PathInfo(repo.root_dir)
+        if path_info == repo_path and isinstance(repo, Repo):
+            message = "Cannot fetch a complete DVC repository"
+            if repo.root_dir != self.root_dir:
+                rel = relpath(repo.root_dir, self.root_dir)
+                message += f" '{rel}'"
+            raise IsADVCRepoError(message)
 
     def get_external(self, path, dest):
         """Convenience wrapper for fetch_external and checkout."""
-        if self.local_cache:
+        path_info = PathInfo(self.root_dir) / path
+        with reraise(FileNotFoundError, PathMissingError(path, self.url)):
+            metadata = self.repo_tree.metadata(path_info)
+
+        self._check_repo(path_info, metadata.repo)
+        if metadata.output_exists:
+            repo = metadata.repo
+            cache = repo.cache.local
             # fetch DVC and git files to tmpdir cache, then checkout
-            _, _, save_infos = self.fetch_external([path])
-            self.local_cache.checkout(PathInfo(dest), save_infos[0])
+            save_info = self._fetch_to_cache(path_info, repo, None)
+            cache.checkout(PathInfo(dest), save_info)
         else:
-            # git-only erepo with no cache, just copy files directly
-            # to dest
-            path = PathInfo(self.root_dir) / path
-            if not self.repo_tree.exists(path):
-                raise PathMissingError(path, self.url)
-            self.repo_tree.copytree(path, dest)
+            # git-only folder, just copy files directly to dest
+            tree = self._get_tree_for(metadata.repo)  # ignore subrepos
+            tree.copytree(path_info, dest)
 
+    def _get_tree_for(self, repo, **kwargs):
+        """
+        Provides a combined tree of a single repo with dvc + git/local tree.
+        """
+        kw = {**self.tree_confs, **kwargs}
+        if "fetch" not in kw:
+            kw["fetch"] = True
+        return RepoTree(repo, **kw)
 
-class ExternalRepo(Repo, BaseExternalRepo):
-    def __init__(self, root_dir, url, rev, for_write=False):
-        if for_write:
-            super().__init__(root_dir)
-        else:
-            root_dir = os.path.realpath(root_dir)
-            super().__init__(root_dir, scm=Git(root_dir), rev=rev)
-        self.url = url
-        self._set_cache_dir()
-        self._fix_upstream()
+    def get_checksum(self, path):
+        path_info = PathInfo(self.root_dir) / path
+        with reraise(FileNotFoundError, PathMissingError(path, self.url)):
+            metadata = self.repo_tree.metadata(path_info)
 
-    @wrap_with(threading.Lock())
-    def _set_cache_dir(self):
-        try:
-            cache_dir = CACHE_DIRS[self.url]
-        except KeyError:
-            cache_dir = CACHE_DIRS[self.url] = tempfile.mkdtemp("dvc-cache")
+        # skip subrepos to check for
+        tree = self._get_tree_for(metadata.repo)
+        return tree.get_hash(path_info)
 
-        self.cache.local.cache_dir = cache_dir
-        self._local_cache = self.cache.local
-
-    def _fix_upstream(self):
-        if not os.path.isdir(self.url):
-            return
-
-        try:
-            src_repo = Repo(self.url)
-        except NotDvcRepoError:
-            # If ExternalRepo does not throw NotDvcRepoError and Repo does,
-            # the self.url might be a bare git repo.
-            # NOTE: This will fail to resolve remote with relative path,
-            # same as if it was a remote DVC repo.
-            return
-
-        try:
-            remote_name = self.config["core"].get("remote")
-            if remote_name:
-                self._fix_local_remote(src_repo, remote_name)
-            else:
-                self._add_upstream(src_repo)
-        finally:
-            src_repo.close()
-
-    def _fix_local_remote(self, src_repo, remote_name):
+    @staticmethod
+    def _fix_local_remote(orig_repo, src_repo, remote_name):
         # If a remote URL is relative to the source repo,
         # it will have changed upon config load and made
         # relative to this new repo. Restore the old one here.
-        new_remote = self.config["remote"][remote_name]
+        new_remote = orig_repo.config["remote"][remote_name]
         old_remote = src_repo.config["remote"][remote_name]
         if new_remote["url"] != old_remote["url"]:
             new_remote["url"] = old_remote["url"]
 
-    def _add_upstream(self, src_repo):
+    @staticmethod
+    def _add_upstream(orig_repo, src_repo):
         # Fill the empty upstream entry with a new remote pointing to the
         # original repo's cache location.
         cache_dir = src_repo.cache.local.cache_dir
-        self.config["remote"]["auto-generated-upstream"] = {"url": cache_dir}
-        self.config["core"]["remote"] = "auto-generated-upstream"
+        orig_repo.config["remote"]["auto-generated-upstream"] = {
+            "url": cache_dir
+        }
+        orig_repo.config["core"]["remote"] = "auto-generated-upstream"
 
+    def make_repo(self, path):
+        repo = Repo(path, scm=self.scm, rev=self.get_rev())
 
-class ExternalGitRepo(BaseExternalRepo):
-    def __init__(self, root_dir, url, rev):
-        self.root_dir = os.path.realpath(root_dir)
-        self.url = url
-        self.tree = self.scm.get_tree(rev)
+        self._setup_cache(repo)
+        self._fix_upstream(repo)
 
-    @cached_property
-    def scm(self):
-        return Git(self.root_dir)
+        return repo
 
-    def close(self):
-        if "scm" in self.__dict__:
-            self.scm.close()
+    def _setup_cache(self, repo):
+        repo.config["cache"]["dir"] = self.cache_dir
+        repo.cache = Cache(repo)
+        if self.cache_types:
+            repo.cache.local.cache_types = self.cache_types
 
-    def find_out_by_relpath(self, path):
-        raise OutputNotFoundError(path, self)
+    def _fix_upstream(self, repo):
+        if not os.path.isdir(self.url):
+            return
 
-    @contextmanager
-    def open_by_relpath(self, path, mode="r", encoding=None, **kwargs):
-        """Opens a specified resource as a file object."""
-        tree = RepoTree(self)
         try:
-            with tree.open(
-                path, mode=mode, encoding=encoding, **kwargs
-            ) as fobj:
-                yield fobj
-        except FileNotFoundError:
-            raise PathMissingError(path, self.url)
+            rel_path = os.path.relpath(repo.root_dir, self.root_dir)
+            src_repo = Repo(PathInfo(self.url) / rel_path)
+        except NotDvcRepoError:
+            return
+
+        try:
+            remote_name = repo.config["core"].get("remote")
+            if remote_name:
+                self._fix_local_remote(repo, src_repo, remote_name)
+            else:
+                self._add_upstream(repo, src_repo)
+        finally:
+            src_repo.close()
+
+    @wrap_with(threading.Lock())
+    def _get_cache_dir(self):
+        try:
+            cache_dir = CACHE_DIRS[self.url]
+        except KeyError:
+            cache_dir = CACHE_DIRS[self.url] = tempfile.mkdtemp("dvc-cache")
+        return cache_dir
 
 
 def _cached_clone(url, rev, for_write=False):
@@ -351,6 +394,11 @@ def _remove(path):
     if os.name == "nt":
         # git.exe may hang for a while not permitting to remove temp dir
         os_retry = retry(5, errors=OSError, timeout=0.1)
-        os_retry(remove)(path)
+        try:
+            os_retry(remove)(path)
+        except PermissionError:
+            logger.warning(
+                "Failed to remove '%s'", relpath(path), exc_info=True
+            )
     else:
         remove(path)

dvc/hash_info.py

@@ -0,0 +1,52 @@
+from collections import OrderedDict
+from dataclasses import dataclass, field
+from typing import Dict, Optional
+
+HASH_DIR_SUFFIX = ".dir"
+
+DirInfo = Dict[str, str]
+
+
+@dataclass
+class HashInfo:
+    PARAM_SIZE = "size"
+    PARAM_NFILES = "nfiles"
+
+    name: Optional[str]
+    value: Optional[str]
+    dir_info: Optional[DirInfo] = field(default=None, compare=False)
+    size: Optional[int] = field(default=None, compare=False)
+    nfiles: Optional[int] = field(default=None, compare=False)
+
+    def __bool__(self):
+        return bool(self.value)
+
+    @classmethod
+    def from_dict(cls, d):
+        _d = d.copy() if d else {}
+        size = _d.pop(cls.PARAM_SIZE, None)
+        nfiles = _d.pop(cls.PARAM_NFILES, None)
+
+        if not _d:
+            return cls(None, None)
+
+        ((name, value),) = _d.items()
+        return cls(name, value, size=size, nfiles=nfiles)
+
+    def to_dict(self):
+        ret = OrderedDict()
+        if not self:
+            return ret
+
+        ret[self.name] = self.value
+        if self.size is not None:
+            ret[self.PARAM_SIZE] = self.size
+        if self.nfiles is not None:
+            ret[self.PARAM_NFILES] = self.nfiles
+        return ret
+
+    @property
+    def isdir(self):
+        if not self:
+            return False
+        return self.value.endswith(HASH_DIR_SUFFIX)

dvc/ignore.py

@@ -1,16 +1,17 @@
 import logging
 import os
 import re
+from collections import namedtuple
 from itertools import groupby, takewhile
 
 from pathspec.patterns import GitWildMatchPattern
 from pathspec.util import normalize_file
-from pygtrie import StringTrie
 
 from dvc.path_info import PathInfo
-from dvc.pathspec_math import merge_patterns
+from dvc.pathspec_math import PatternInfo, merge_patterns
 from dvc.system import System
 from dvc.utils import relpath
+from dvc.utils.collections import PathStringTrie
 
 logger = logging.getLogger(__name__)
 
@@ -24,18 +25,26 @@ class DvcIgnore:
 
 class DvcIgnorePatterns(DvcIgnore):
     def __init__(self, pattern_list, dirname):
+        if pattern_list:
+            if isinstance(pattern_list[0], str):
+                pattern_list = [
+                    PatternInfo(pattern, "") for pattern in pattern_list
+                ]
 
         self.pattern_list = pattern_list
         self.dirname = dirname
         self.prefix = self.dirname + os.sep
 
-        regex_pattern_list = map(
-            GitWildMatchPattern.pattern_to_regex, pattern_list
-        )
+        self.regex_pattern_list = [
+            GitWildMatchPattern.pattern_to_regex(pattern_info.patterns)
+            for pattern_info in pattern_list
+        ]
 
         self.ignore_spec = [
             (ignore, re.compile("|".join(item[0] for item in group)))
-            for ignore, group in groupby(regex_pattern_list, lambda x: x[1])
+            for ignore, group in groupby(
+                self.regex_pattern_list, lambda x: x[1]
+            )
             if ignore is not None
         ]
 
@@ -43,9 +52,19 @@ class DvcIgnorePatterns(DvcIgnore):
     def from_files(cls, ignore_file_path, tree):
         assert os.path.isabs(ignore_file_path)
         dirname = os.path.normpath(os.path.dirname(ignore_file_path))
+        ignore_file_rel_path = os.path.relpath(
+            ignore_file_path, tree.tree_root
+        )
         with tree.open(ignore_file_path, encoding="utf-8") as fobj:
             path_spec_lines = [
-                line for line in map(str.strip, fobj.readlines()) if line
+                PatternInfo(
+                    line,
+                    "{}:{}:{}".format(ignore_file_rel_path, line_no + 1, line),
+                )
+                for line_no, line in enumerate(
+                    map(str.strip, fobj.readlines())
+                )
+                if line and not (line.strip().startswith("#"))
             ]
 
         return cls(path_spec_lines, dirname)
@@ -56,7 +75,7 @@ class DvcIgnorePatterns(DvcIgnore):
 
         return dirs, files
 
-    def matches(self, dirname, basename, is_dir=False):
+    def _get_normalize_path(self, dirname, basename):
         # NOTE: `relpath` is too slow, so we have to assume that both
         # `dirname` and `self.dirname` are relative or absolute together.
         if dirname == self.dirname:
@@ -70,6 +89,12 @@ class DvcIgnorePatterns(DvcIgnore):
 
         if not System.is_unix():
             path = normalize_file(path)
+        return path
+
+    def matches(self, dirname, basename, is_dir=False):
+        path = self._get_normalize_path(dirname, basename)
+        if not path:
+            return False
         return self.ignore(path, is_dir)
 
     def ignore(self, path, is_dir):
@@ -85,25 +110,57 @@ class DvcIgnorePatterns(DvcIgnore):
                     result = ignore
         return result
 
+    def match_details(self, dirname, basename, is_dir=False):
+        path = self._get_normalize_path(dirname, basename)
+        if not path:
+            return False
+        return self._ignore_details(path, is_dir)
+
+    def _ignore_details(self, path, is_dir):
+        result = []
+        for ignore, pattern in zip(self.regex_pattern_list, self.pattern_list):
+            regex = re.compile(ignore[0])
+            # skip system pattern
+            if not pattern.file_info:
+                continue
+            if is_dir:
+                path_dir = f"{path}/"
+                if regex.match(path) or regex.match(path_dir):
+                    result.append(pattern.file_info)
+            else:
+                if regex.match(path):
+                    result.append(pattern.file_info)
+        return result
+
     def __hash__(self):
-        return hash(self.dirname + ":" + "\n".join(self.pattern_list))
+        return hash(self.dirname + ":" + str(self.pattern_list))
 
     def __eq__(self, other):
         if not isinstance(other, DvcIgnorePatterns):
             return NotImplemented
         return (self.dirname == other.dirname) & (
-            self.pattern_list == other.pattern_list
+            [pattern.patterns for pattern in self.pattern_list]
+            == [pattern.patterns for pattern in other.pattern_list]
         )
 
     def __bool__(self):
         return bool(self.pattern_list)
 
 
+CheckIgnoreResult = namedtuple(
+    "CheckIgnoreResult", ["file", "match", "patterns"]
+)
+
+
+def _no_match(path):
+    return CheckIgnoreResult(path, False, ["::"])
+
+
 class DvcIgnoreFilterNoop:
     def __init__(self, tree, root_dir):
         pass
 
-    def __call__(self, root, dirs, files):
+    def __call__(self, root, dirs, files, **kwargs):
         return dirs, files
 
     def is_ignored_dir(self, _):
@@ -112,6 +169,12 @@ class DvcIgnoreFilterNoop:
     def is_ignored_file(self, _):
         return False
 
+    def check_ignore(self, path):
+        return _no_match(path)
+
+    def is_ignored(self, _):
+        return False
+
 
 class DvcIgnoreFilter:
     @staticmethod
@@ -127,10 +190,11 @@ class DvcIgnoreFilter:
 
         self.tree = tree
         self.root_dir = root_dir
-        self.ignores_trie_tree = StringTrie(separator=os.sep)
+        self.ignores_trie_tree = PathStringTrie()
         self.ignores_trie_tree[root_dir] = DvcIgnorePatterns(
             default_ignore_patterns, root_dir
         )
+        self._ignored_subrepos = PathStringTrie()
         self._update(self.root_dir)
 
     def _update(self, dirname):
@@ -166,27 +230,33 @@ class DvcIgnoreFilter:
     def _update_sub_repo(self, root, dirs):
         for d in dirs:
             if self._is_dvc_repo(root, d):
+                self._ignored_subrepos[root] = self._ignored_subrepos.get(
+                    root, set()
+                ) | {d}
+                pattern_info = PatternInfo(
+                    f"/{d}/", "in sub_repo:{}".format(d)
+                )
+                new_pattern = DvcIgnorePatterns([pattern_info], root)
                 old_pattern = self.ignores_trie_tree.longest_prefix(root).value
                 if old_pattern:
                     self.ignores_trie_tree[root] = DvcIgnorePatterns(
                         *merge_patterns(
                             old_pattern.pattern_list,
                             old_pattern.dirname,
-                            ["/{}/".format(d)],
-                            root,
+                            new_pattern.pattern_list,
+                            new_pattern.dirname,
                         )
                     )
                 else:
-                    self.ignores_trie_tree[root] = DvcIgnorePatterns(
-                        ["/{}/".format(d)], root
-                    )
+                    self.ignores_trie_tree[root] = new_pattern
 
-    def __call__(self, root, dirs, files):
+    def __call__(self, root, dirs, files, ignore_subrepos=True):
         ignore_pattern = self._get_trie_pattern(root)
         if ignore_pattern:
-            return ignore_pattern(root, dirs, files)
-        else:
-            return dirs, files
+            dirs, files = ignore_pattern(root, dirs, files)
+            if not ignore_subrepos:
+                dirs.extend(self._ignored_subrepos.get(root, []))
+        return dirs, files
 
     def _get_trie_pattern(self, dirname):
         ignore_pattern = self.ignores_trie_tree.get(dirname)
@@ -214,7 +284,7 @@ class DvcIgnoreFilter:
 
     def _is_ignored(self, path, is_dir=False):
         if self._outside_repo(path):
-            return True
+            return False
         dirname, basename = os.path.split(os.path.normpath(path))
         ignore_pattern = self._get_trie_pattern(dirname)
         if ignore_pattern:
@@ -222,14 +292,21 @@ class DvcIgnoreFilter:
         else:
             return False
 
-    def is_ignored_dir(self, path):
+    def _is_subrepo(self, path):
+        dirname, basename = os.path.split(os.path.normpath(path))
+        return basename in self._ignored_subrepos.get(dirname, set())
+
+    def is_ignored_dir(self, path, ignore_subrepos=True):
         path = os.path.abspath(path)
+        if not ignore_subrepos:
+            return not self._is_subrepo(path)
         if path == self.root_dir:
             return False
 
         return self._is_ignored(path, True)
 
     def is_ignored_file(self, path):
+        path = os.path.abspath(path)
         return self._is_ignored(path, False)
 
     def _outside_repo(self, path):
@@ -245,3 +322,37 @@ class DvcIgnoreFilter:
         ):
             return True
         return False
+
+    def check_ignore(self, target):
+        full_target = os.path.abspath(target)
+        if not self._outside_repo(full_target):
+            dirname, basename = os.path.split(os.path.normpath(full_target))
+            pattern = self._get_trie_pattern(dirname)
+            if pattern:
+                matches = pattern.match_details(
+                    dirname, basename, os.path.isdir(full_target)
+                )
+
+                if matches:
+                    return CheckIgnoreResult(target, True, matches)
+        return _no_match(target)
+
+    def is_ignored(self, path):
+        # NOTE: can't use self.check_ignore(path).match for now, see
+        # https://github.com/iterative/dvc/issues/4555
+        return self.is_ignored_dir(path) or self.is_ignored_file(path)
+
+
+def init(path):
+    dvcignore = os.path.join(path, DvcIgnore.DVCIGNORE_FILE)
+    if os.path.exists(dvcignore):
+        return dvcignore
+
+    with open(dvcignore, "w") as fobj:
+        fobj.write(
+            "# Add patterns of files dvc should ignore, which could improve\n"
+            "# the performance. Learn more at\n"
+            "# https://dvc.org/doc/user-guide/dvcignore\n"
+        )
+
+    return dvcignore

dvc/info.py

@@ -0,0 +1,160 @@
+import itertools
+import os
+import pathlib
+import platform
+import uuid
+
+from dvc.exceptions import DvcException, NotDvcRepoError
+from dvc.repo import Repo
+from dvc.scm.base import SCMError
+from dvc.system import System
+from dvc.tree import TREES, get_tree_cls, get_tree_config
+from dvc.utils import error_link
+from dvc.utils.pkg import PKG
+from dvc.version import __version__
+
+try:
+    import psutil
+except ImportError:
+    psutil = None
+
+if PKG is None:
+    package = ""
+else:
+    package = f"({PKG})"
+
+
+def get_dvc_info():
+    info = [
+        f"DVC version: {__version__} {package}",
+        "---------------------------------",
+        f"Platform: Python {platform.python_version()} on "
+        f"{platform.platform()}",
+        f"Supports: {_get_supported_remotes()}",
+    ]
+
+    try:
+        repo = Repo()
+
+        # cache_dir might not exist yet (e.g. after `dvc init`), and we
+        # can't auto-create it, as it might cause issues if the user
+        # later decides to enable shared cache mode with
+        # `dvc config cache.shared group`.
+        if os.path.exists(repo.cache.local.cache_dir):
+            info.append(
+                "Cache types: {}".format(_get_linktype_support_info(repo))
+            )
+            if psutil:
+                fs_type = get_fs_type(repo.cache.local.cache_dir)
+                info.append(f"Cache directory: {fs_type}")
+        else:
+            info.append("Cache types: " + error_link("no-dvc-cache"))
+
+        info.append(f"Caches: {_get_caches(repo.cache)}")
+
+        info.append(f"Remotes: {_get_remotes(repo.config)}")
+
+    except NotDvcRepoError:
+        pass
+    except SCMError:
+        info.append("Repo: dvc, git (broken)")
+    else:
+        root_directory = repo.root_dir
+        if psutil:
+            fs_root = get_fs_type(os.path.abspath(root_directory))
+            info.append(f"Workspace directory: {fs_root}")
+        info.append("Repo: {}".format(_get_dvc_repo_info(repo)))
+    return "\n".join(info)
+
+
+def _get_caches(cache):
+    caches = (
+        cache_type
+        for cache_type, cache_instance in cache.by_scheme()
+        if cache_instance
+    )
+
+    # Caches will be always non-empty including the local cache
+    return ", ".join(caches)
+
+
+def _get_remotes(config):
+    schemes = (
+        get_tree_cls(get_tree_config(config, name=remote)).scheme
+        for remote in config["remote"]
+    )
+
+    return ", ".join(schemes) or "None"
+
+
+def _get_linktype_support_info(repo):
+
+    links = {
+        "reflink": (System.reflink, None),
+        "hardlink": (System.hardlink, System.is_hardlink),
+        "symlink": (System.symlink, System.is_symlink),
+    }
+
+    fname = "." + str(uuid.uuid4())
+    src = os.path.join(repo.cache.local.cache_dir, fname)
+    open(src, "w").close()
+    dst = os.path.join(repo.root_dir, fname)
+
+    cache = []
+
+    for name, (link, is_link) in links.items():
+        try:
+            link(src, dst)
+            status = "supported"
+            if is_link and not is_link(dst):
+                status = "broken"
+            os.unlink(dst)
+        except DvcException:
+            status = "not supported"
+
+        if status == "supported":
+            cache.append(name)
+    os.remove(src)
+
+    return ", ".join(cache)
+
+
+def _get_supported_remotes():
+
+    supported_remotes = []
+    for tree_cls in TREES:
+        if not tree_cls.get_missing_deps():
+            supported_remotes.append(tree_cls.scheme)
+
+    if len(supported_remotes) == len(TREES):
+        return "All remotes"
+
+    if len(supported_remotes) == 1:
+        return supported_remotes
+
+    return ", ".join(supported_remotes)
+
+
+def get_fs_type(path):
+
+    partition = {
+        pathlib.Path(part.mountpoint): (part.fstype + " on " + part.device)
+        for part in psutil.disk_partitions(all=True)
+    }
+
+    path = pathlib.Path(path)
+
+    for parent in itertools.chain([path], path.parents):
+        if parent in partition:
+            return partition[parent]
+    return ("unknown", "none")
+
+
+def _get_dvc_repo_info(self):
+    if self.config.get("core", {}).get("no_scm", False):
+        return "dvc (no_scm)"
+
+    if self.root_dir != self.scm.root_dir:
+        return "dvc (subdir), git"
+
+    return "dvc, git"

dvc/lock.py

@@ -2,22 +2,26 @@
 
 import hashlib
 import os
-import time
+from abc import ABC, abstractmethod
 from datetime import timedelta
 
 import flufl.lock
 import zc.lockfile
+from funcy import retry
 
 from dvc.exceptions import DvcException
 from dvc.progress import Tqdm
 from dvc.utils import format_link
 
-DEFAULT_TIMEOUT = 5
+DEFAULT_TIMEOUT = 3
+
 
 FAILED_TO_LOCK_MESSAGE = (
-    "cannot perform the command because another DVC process seems to be "
-    "running on this project. If that is not the case, manually remove "
-    "`.dvc/tmp/lock` and try again."
+    "Unable to acquire lock. Most likely another DVC process is running or "
+    "was terminated abruptly. Check the page {} for other possible reasons "
+    "and to learn how to resolve this."
+).format(
+    format_link("https://dvc.org/doc/user-guide/troubleshooting#lock-issue")
 )
 
 
@@ -25,20 +29,74 @@ class LockError(DvcException):
     """Thrown when unable to acquire the lock for DVC repo."""
 
 
-class Lock:
+class LockBase(ABC):
+    @abstractmethod
+    def __init__(self, lockfile):
+        self._lockfile = lockfile
+
+    @property
+    def lockfile(self):
+        return self._lockfile
+
+    @abstractmethod
+    def lock(self):
+        pass
+
+    @abstractmethod
+    def unlock(self):
+        pass
+
+    @property
+    @abstractmethod
+    def is_locked(self):
+        pass
+
+    @abstractmethod
+    def __enter__(self):
+        pass
+
+    @abstractmethod
+    def __exit__(self, typ, value, tbck):
+        pass
+
+
+class LockNoop(LockBase):
+    def __init__(
+        self, *args, **kwargs
+    ):  # pylint: disable=super-init-not-called
+        self._lock = False
+
+    def lock(self):
+        self._lock = True
+
+    def unlock(self):
+        self._lock = False
+
+    @property
+    def is_locked(self):
+        return self._lock
+
+    def __enter__(self):
+        self.lock()
+
+    def __exit__(self, typ, value, tbck):
+        self.unlock()
+
+
+class Lock(LockBase):
     """Class for DVC repo lock.
 
     Uses zc.lockfile as backend.
     """
 
     def __init__(self, lockfile, friendly=False, **kwargs):
+        super().__init__(lockfile)
         self._friendly = friendly
-        self.lockfile = lockfile
         self._lock = None
 
     @property
     def files(self):
-        return [self.lockfile]
+        return [self._lockfile]
 
     def _do_lock(self):
         try:
@@ -51,18 +109,15 @@ class Lock:
                     )
                 ),
             ):
-                self._lock = zc.lockfile.LockFile(self.lockfile)
+                self._lock = zc.lockfile.LockFile(self._lockfile)
         except zc.lockfile.LockError:
             raise LockError(FAILED_TO_LOCK_MESSAGE)
 
     def lock(self):
-        try:
-            self._do_lock()
-            return
-        except LockError:
-            time.sleep(DEFAULT_TIMEOUT)
-
-        self._do_lock()
+        retries = 6
+        delay = DEFAULT_TIMEOUT / retries
+        lock_retry = retry(retries, LockError, timeout=delay)(self._do_lock)
+        lock_retry()
 
     def unlock(self):
         self._lock.close()
@@ -79,7 +134,7 @@ class Lock:
         self.unlock()
 
 
-class HardlinkLock(flufl.lock.Lock):
+class HardlinkLock(flufl.lock.Lock, LockBase):
     """Class for DVC repo lock.
 
     Args:
@@ -93,6 +148,8 @@ class HardlinkLock(flufl.lock.Lock):
     ):  # pylint: disable=super-init-not-called
         import socket
 
+        super().__init__(lockfile)
+
         self._tmp_dir = tmp_dir
 
         # NOTE: this is basically Lock.__init__ copy-paste, except that
@@ -108,17 +165,12 @@ class HardlinkLock(flufl.lock.Lock):
         # [1] https://github.com/iterative/dvc/issues/2582
         self._hostname = socket.gethostname()
 
-        self._lockfile = lockfile
         self._lifetime = timedelta(days=365)  # Lock for good by default
         self._separator = flufl.lock.SEP
         self._set_claimfile()
         self._owned = True
         self._retry_errnos = []
 
-    @property
-    def lockfile(self):
-        return self._lockfile
-
     def lock(self):  # pylint: disable=arguments-differ
         try:
             super().lock(timedelta(seconds=DEFAULT_TIMEOUT))

dvc/main.py

@@ -2,6 +2,7 @@
 
 import errno
 import logging
+from contextlib import contextmanager
 
 from dvc import analytics
 from dvc.cli import parse_args
@@ -10,7 +11,7 @@ from dvc.exceptions import DvcException, DvcParserError, NotDvcRepoError
 from dvc.external_repo import clean_repos
 from dvc.logger import FOOTER, disable_other_loggers
 from dvc.tree.pool import close_pools
-from dvc.utils import format_link
+from dvc.utils import error_link
 
 # Workaround for CPython bug. See [1] and [2] for more info.
 # [1] https://github.com/aws/aws-cli/blob/1.16.277/awscli/clidriver.py#L55
@@ -21,7 +22,40 @@ from dvc.utils import format_link
 logger = logging.getLogger("dvc")
 
 
-def main(argv=None):
+@contextmanager
+def profile(enable, dump):
+    if not enable:
+        yield
+        return
+
+    import cProfile
+
+    prof = cProfile.Profile()
+    prof.enable()
+
+    yield
+
+    prof.disable()
+    if not dump:
+        prof.print_stats(sort="cumtime")
+        return
+    prof.dump_stats(dump)
+
+
+@contextmanager
+def debug(enable):
+    try:
+        yield
+        return
+    except Exception:
+        if enable:
+            import pdb  # noqa: T100
+
+            pdb.post_mortem()
+        raise
+
+
+def main(argv=None):  # noqa: C901
     """Run dvc CLI command.
 
     Args:
@@ -37,20 +71,23 @@ def main(argv=None):
     try:
         args = parse_args(argv)
 
-        verbosity = args.verbose - args.quiet
-        if verbosity:
-            logger.setLevel(
-                {
-                    -2: logging.CRITICAL,
-                    -1: logging.ERROR,
-                    1: logging.DEBUG,
-                    2: logging.TRACE,
-                }[max(-2, min(verbosity, 2))]
-            )
+        level = None
+        if args.quiet:
+            level = logging.CRITICAL
+        elif args.verbose == 1:
+            level = logging.DEBUG
+        elif args.verbose > 1:
+            level = logging.TRACE
+
+        if level is not None:
+            logger.setLevel(level)
+
         logger.trace(args)
 
-        cmd = args.func(args)
-        ret = cmd.run()
+        with profile(enable=args.cprofile, dump=args.cprofile_dump):
+            with debug(args.pdb):
+                cmd = args.func(args)
+                ret = cmd.run()
     except ConfigError:
         logger.exception("configuration error")
         ret = 251
@@ -65,22 +102,31 @@ def main(argv=None):
     except DvcException:
         ret = 255
         logger.exception("")
-    except OSError as exc:
-        if exc.errno == errno.EMFILE:
+    except Exception as exc:  # noqa, pylint: disable=broad-except
+        # pylint: disable=no-member
+        if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
             logger.exception(
                 "too many open files, please visit "
                 "{} to see how to handle this "
-                "problem".format(
-                    format_link("https://error.dvc.org/many-files")
-                ),
+                "problem".format(error_link("many-files")),
                 extra={"tb_only": True},
             )
         else:
+            from dvc.info import get_dvc_info
+
             logger.exception("unexpected error")
+
+            dvc_info = get_dvc_info()
+            logger.debug("Version info for developers:\n%s", dvc_info)
+
+            logger.info(FOOTER)
         ret = 255
-    except Exception:  # noqa, pylint: disable=broad-except
-        logger.exception("unexpected error")
-        ret = 255
+
+    try:
+        if analytics.is_enabled():
+            analytics.collect_and_send_report(args, ret)
+
+        return ret
     finally:
         logger.setLevel(outerLogLevel)
 
@@ -92,11 +138,3 @@ def main(argv=None):
         # Remove cached repos in the end of the call, these are anonymous
         # so won't be reused by any other subsequent run anyway.
         clean_repos()
-
-    if ret != 0:
-        logger.info(FOOTER)
-
-    if analytics.is_enabled():
-        analytics.collect_and_send_report(args, ret)
-
-    return ret

dvc/output/__init__.py

@@ -4,24 +4,28 @@ from urllib.parse import urlparse
 from funcy import collecting, project
 from voluptuous import And, Any, Coerce, Length, Lower, Required, SetTo
 
+from dvc.hash_info import HashInfo
 from dvc.output.base import BaseOutput
 from dvc.output.gs import GSOutput
 from dvc.output.hdfs import HDFSOutput
 from dvc.output.local import LocalOutput
 from dvc.output.s3 import S3Output
 from dvc.output.ssh import SSHOutput
+from dvc.output.webhdfs import WebHDFSOutput
 from dvc.scheme import Schemes
 
 from ..tree import get_cloud_tree
 from ..tree.hdfs import HDFSTree
 from ..tree.local import LocalTree
 from ..tree.s3 import S3Tree
+from ..tree.webhdfs import WebHDFSTree
 
 OUTS = [
     HDFSOutput,
     S3Output,
     GSOutput,
     SSHOutput,
+    WebHDFSOutput,
     # NOTE: LocalOutput is the default choice
 ]
 
@@ -31,6 +35,7 @@ OUTS_MAP = {
     Schemes.GS: GSOutput,
     Schemes.SSH: SSHOutput,
     Schemes.LOCAL: LocalOutput,
+    Schemes.WEBHDFS: WebHDFSOutput,
 }
 
 CHECKSUM_SCHEMA = Any(
@@ -51,6 +56,7 @@ CHECKSUMS_SCHEMA = {
     LocalTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
     S3Tree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
     HDFSTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
+    WebHDFSTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
 }
 
 SCHEMA = CHECKSUMS_SCHEMA.copy()
@@ -59,10 +65,22 @@ SCHEMA[BaseOutput.PARAM_CACHE] = bool
 SCHEMA[BaseOutput.PARAM_METRIC] = BaseOutput.METRIC_SCHEMA
 SCHEMA[BaseOutput.PARAM_PLOT] = bool
 SCHEMA[BaseOutput.PARAM_PERSIST] = bool
+SCHEMA[BaseOutput.PARAM_CHECKPOINT] = bool
+SCHEMA[HashInfo.PARAM_SIZE] = int
+SCHEMA[HashInfo.PARAM_NFILES] = int
+SCHEMA[BaseOutput.PARAM_DESC] = str
 
 
 def _get(
-    stage, p, info=None, cache=True, metric=False, plot=False, persist=False
+    stage,
+    p,
+    info=None,
+    cache=True,
+    metric=False,
+    plot=False,
+    persist=False,
+    checkpoint=False,
+    desc=None,
 ):
     parsed = urlparse(p)
 
@@ -77,6 +95,8 @@ def _get(
             metric=metric,
             plot=plot,
             persist=persist,
+            checkpoint=checkpoint,
+            desc=desc,
         )
 
     for o in OUTS:
@@ -90,6 +110,8 @@ def _get(
                 metric=metric,
                 plot=plot,
                 persist=persist,
+                checkpoint=checkpoint,
+                desc=desc,
             )
     return LocalOutput(
         stage,
@@ -100,6 +122,8 @@ def _get(
         metric=metric,
         plot=plot,
         persist=persist,
+        checkpoint=checkpoint,
+        desc=desc,
     )
 
 
@@ -111,6 +135,8 @@ def loadd_from(stage, d_list):
         metric = d.pop(BaseOutput.PARAM_METRIC, False)
         plot = d.pop(BaseOutput.PARAM_PLOT, False)
         persist = d.pop(BaseOutput.PARAM_PERSIST, False)
+        checkpoint = d.pop(BaseOutput.PARAM_CHECKPOINT, False)
+        desc = d.pop(BaseOutput.PARAM_DESC, False)
         ret.append(
             _get(
                 stage,
@@ -120,13 +146,21 @@ def loadd_from(stage, d_list):
                 metric=metric,
                 plot=plot,
                 persist=persist,
+                checkpoint=checkpoint,
+                desc=desc,
             )
         )
     return ret
 
 
 def loads_from(
-    stage, s_list, use_cache=True, metric=False, plot=False, persist=False
+    stage,
+    s_list,
+    use_cache=True,
+    metric=False,
+    plot=False,
+    persist=False,
+    checkpoint=False,
 ):
     return [
         _get(
@@ -137,6 +171,7 @@ def loads_from(
             metric=metric,
             plot=plot,
             persist=persist,
+            checkpoint=checkpoint,
         )
         for s in s_list
     ]
@@ -180,5 +215,12 @@ def load_from_pipeline(stage, s_list, typ="outs"):
             from dvc.schema import PLOT_PROPS
 
             plt_d, flags = _split_dict(flags, keys=PLOT_PROPS.keys())
-        extra = project(flags, ["cache", "persist"])
+        extra = project(
+            flags,
+            [
+                BaseOutput.PARAM_CACHE,
+                BaseOutput.PARAM_PERSIST,
+                BaseOutput.PARAM_CHECKPOINT,
+            ],
+        )
         yield _get(stage, path, {}, plot=plt_d or plot, metric=metric, **extra)

dvc/output/base.py

@@ -1,6 +1,7 @@
 import logging
 import os
 from copy import copy
+from typing import Type
 from urllib.parse import urlparse
 
 from voluptuous import Any
@@ -8,10 +9,13 @@ from voluptuous import Any
 import dvc.prompt as prompt
 from dvc.cache import NamedCache
 from dvc.exceptions import (
+    CheckoutError,
     CollectCacheError,
     DvcException,
+    MergeError,
     RemoteCacheRequiredError,
 )
+from dvc.hash_info import HashInfo
 
 from ..tree.base import BaseTree
 
@@ -45,6 +49,12 @@ class OutputIsStageFileError(DvcException):
         super().__init__(f"Stage file '{path}' cannot be an output.")
 
 
+class OutputIsIgnoredError(DvcException):
+    def __init__(self, match):
+        lines = "\n".join(match.patterns)
+        super().__init__(f"Path '{match.file}' is ignored by\n{lines}")
+
+
 class BaseOutput:
     IS_DEPENDENCY = False
 
@@ -52,6 +62,7 @@ class BaseOutput:
 
     PARAM_PATH = "path"
     PARAM_CACHE = "cache"
+    PARAM_CHECKPOINT = "checkpoint"
     PARAM_METRIC = "metric"
     PARAM_METRIC_TYPE = "type"
     PARAM_METRIC_XPATH = "xpath"
@@ -64,6 +75,7 @@ class BaseOutput:
     PARAM_PLOT_TITLE = "title"
     PARAM_PLOT_HEADER = "header"
     PARAM_PERSIST = "persist"
+    PARAM_DESC = "desc"
 
     METRIC_SCHEMA = Any(
         None,
@@ -74,9 +86,10 @@ class BaseOutput:
         },
     )
 
-    DoesNotExistError = OutputDoesNotExistError
-    IsNotFileOrDirError = OutputIsNotFileOrDirError
-    IsStageFileError = OutputIsStageFileError
+    DoesNotExistError = OutputDoesNotExistError  # type: Type[DvcException]
+    IsNotFileOrDirError = OutputIsNotFileOrDirError  # type: Type[DvcException]
+    IsStageFileError = OutputIsStageFileError  # type: Type[DvcException]
+    IsIgnoredError = OutputIsIgnoredError  # type: Type[DvcException]
 
     sep = "/"
 
@@ -90,8 +103,10 @@ class BaseOutput:
         metric=False,
         plot=False,
         persist=False,
+        checkpoint=False,
+        desc=None,
     ):
-        self._validate_output_path(path)
+        self._validate_output_path(path, stage)
         # This output (and dependency) objects have too many paths/urls
         # here is a list and comments:
         #
@@ -105,7 +120,7 @@ class BaseOutput:
         self.stage = stage
         self.repo = stage.repo if stage else None
         self.def_path = path
-        self.info = info
+        self.hash_info = HashInfo.from_dict(info)
         if tree:
             self.tree = tree
         else:
@@ -114,6 +129,8 @@ class BaseOutput:
         self.metric = False if self.IS_DEPENDENCY else metric
         self.plot = False if self.IS_DEPENDENCY else plot
         self.persist = persist
+        self.checkpoint = checkpoint
+        self.desc = desc
 
         self.path_info = self._parse_path(tree, path)
         if self.use_cache and self.cache is None:
@@ -154,7 +171,7 @@ class BaseOutput:
 
     @property
     def dir_cache(self):
-        return self.cache.get_dir_cache(self.checksum)
+        return self.cache.get_dir_cache(self.hash_info)
 
     @classmethod
     def supported(cls, url):
@@ -162,58 +179,50 @@ class BaseOutput:
 
     @property
     def cache_path(self):
-        return self.cache.tree.hash_to_path_info(self.checksum).url
+        return self.cache.tree.hash_to_path_info(self.hash_info.value).url
 
-    @property
-    def checksum_type(self):
-        return self.tree.PARAM_CHECKSUM
-
-    @property
-    def checksum(self):
-        return self.info.get(self.tree.PARAM_CHECKSUM)
-
-    @checksum.setter
-    def checksum(self, checksum):
-        self.info[self.tree.PARAM_CHECKSUM] = checksum
-
-    def get_checksum(self):
-        return self.tree.get_hash(self.path_info)
+    def get_hash(self):
+        if not self.use_cache:
+            return self.tree.get_hash(self.path_info)
+        return self.cache.get_hash(self.tree, self.path_info)
 
     @property
     def is_dir_checksum(self):
-        return self.tree.is_dir_hash(self.checksum)
+        return self.hash_info.isdir
 
     @property
     def exists(self):
         return self.tree.exists(self.path_info)
 
-    def save_info(self):
-        return self.tree.save_info(self.path_info)
-
     def changed_checksum(self):
-        return self.checksum != self.get_checksum()
+        return self.hash_info != self.get_hash()
 
     def changed_cache(self, filter_info=None):
-        if not self.use_cache or not self.checksum:
+        if not self.use_cache or not self.hash_info:
             return True
 
-        return self.cache.changed_cache(self.checksum, filter_info=filter_info)
-
-    def status(self):
-        if self.checksum and self.use_cache and self.changed_cache():
-            return {str(self): "not in cache"}
+        return self.cache.changed_cache(
+            self.hash_info, filter_info=filter_info
+        )
 
+    def workspace_status(self):
         if not self.exists:
             return {str(self): "deleted"}
 
         if self.changed_checksum():
             return {str(self): "modified"}
 
-        if not self.checksum:
+        if not self.hash_info:
             return {str(self): "new"}
 
         return {}
 
+    def status(self):
+        if self.hash_info and self.use_cache and self.changed_cache():
+            return {str(self): "not in cache"}
+
+        return self.workspace_status()
+
     def changed(self):
         status = self.status()
         logger.debug(str(status))
@@ -264,7 +273,7 @@ class BaseOutput:
             self.verify_metric()
 
         if not self.use_cache:
-            self.info = self.save_info()
+            self.hash_info = self.get_hash()
             if not self.IS_DEPENDENCY:
                 logger.debug(
                     "Output '%s' doesn't use cache. Skipping saving.", self
@@ -277,19 +286,26 @@ class BaseOutput:
             logger.debug("Output '%s' didn't change. Skipping saving.", self)
             return
 
-        self.info = self.save_info()
+        self.hash_info = self.get_hash()
 
     def commit(self):
+        if not self.exists:
+            raise self.DoesNotExistError(self)
+
+        assert self.hash_info
         if self.use_cache:
-            self.cache.save(self.path_info, self.cache.tree, self.info)
+            self.cache.save(self.path_info, self.cache.tree, self.hash_info)
 
     def dumpd(self):
-        ret = copy(self.info)
+        ret = copy(self.hash_info.to_dict())
         ret[self.PARAM_PATH] = self.def_path
 
         if self.IS_DEPENDENCY:
             return ret
 
+        if self.desc:
+            ret[self.PARAM_DESC] = self.desc
+
         if not self.use_cache:
             ret[self.PARAM_CACHE] = self.use_cache
 
@@ -309,6 +325,9 @@ class BaseOutput:
         if self.persist:
             ret[self.PARAM_PERSIST] = self.persist
 
+        if self.checkpoint:
+            ret[self.PARAM_CHECKPOINT] = self.checkpoint
+
         return ret
 
     def verify_metric(self):
@@ -323,6 +342,8 @@ class BaseOutput:
         progress_callback=None,
         relink=False,
         filter_info=None,
+        allow_missing=False,
+        **kwargs,
     ):
         if not self.use_cache:
             if progress_callback:
@@ -331,14 +352,20 @@ class BaseOutput:
                 )
             return None
 
-        return self.cache.checkout(
-            self.path_info,
-            self.info,
-            force=force,
-            progress_callback=progress_callback,
-            relink=relink,
-            filter_info=filter_info,
-        )
+        try:
+            return self.cache.checkout(
+                self.path_info,
+                self.hash_info,
+                force=force,
+                progress_callback=progress_callback,
+                relink=relink,
+                filter_info=filter_info,
+                **kwargs,
+            )
+        except CheckoutError:
+            if allow_missing or self.checkpoint:
+                return None
+            raise
 
     def remove(self, ignore_remove=False):
         self.tree.remove(self.path_info)
@@ -367,7 +394,7 @@ class BaseOutput:
             return 0
 
         return self.cache.get_files_number(
-            self.path_info, self.checksum, filter_info
+            self.path_info, self.hash_info, filter_info
         )
 
     def unprotect(self):
@@ -377,12 +404,15 @@ class BaseOutput:
     def get_dir_cache(self, **kwargs):
         if not self.is_dir_checksum:
             raise DvcException("cannot get dir cache for file checksum")
-        if self.cache.changed_cache_file(self.checksum):
+        if self.cache.changed_cache_file(self.hash_info):
             self.repo.cloud.pull(
-                NamedCache.make("local", self.checksum, str(self)),
+                NamedCache.make("local", self.hash_info.value, str(self)),
                 show_checksums=False,
                 **kwargs,
             )
+            self.hash_info.dir_info = None
+        if not self.hash_info.dir_info:
+            self.cache.set_dir_info(self.hash_info)
         return self.dir_cache
 
     def collect_used_dir_cache(
@@ -414,7 +444,7 @@ class BaseOutput:
         except DvcException:
             logger.debug(f"failed to pull cache for '{self}'")
 
-        if self.cache.changed_cache_file(self.checksum):
+        if self.cache.changed_cache_file(self.hash_info):
             msg = (
                 "Missing cache for directory '{}'. "
                 "Cache for files inside will be lost. "
@@ -431,9 +461,7 @@ class BaseOutput:
         path = str(self.path_info)
         filter_path = str(filter_info) if filter_info else None
         is_win = os.name == "nt"
-        for entry in self.dir_cache:
-            checksum = entry[self.tree.PARAM_CHECKSUM]
-            entry_relpath = entry[self.tree.PARAM_RELPATH]
+        for entry_relpath, entry_hash_info in self.dir_cache.items():
             if is_win:
                 entry_relpath = entry_relpath.replace("/", os.sep)
             entry_path = os.path.join(path, entry_relpath)
@@ -442,7 +470,7 @@ class BaseOutput:
                 or entry_path == filter_path
                 or entry_path.startswith(filter_path + os.sep)
             ):
-                cache.add(self.scheme, checksum, entry_path)
+                cache.add(self.scheme, entry_hash_info.value, entry_path)
 
         return cache
 
@@ -464,7 +492,7 @@ class BaseOutput:
             cache.external[dep.repo_pair].add(dep.def_path)
             return cache
 
-        if not self.checksum:
+        if not self.hash_info:
             msg = (
                 "Output '{}'({}) is missing version info. "
                 "Cache for it will not be collected. "
@@ -483,20 +511,68 @@ class BaseOutput:
             logger.warning(msg)
             return NamedCache()
 
-        ret = NamedCache.make(self.scheme, self.checksum, str(self))
+        ret = NamedCache.make(self.scheme, self.hash_info.value, str(self))
 
         if not self.is_dir_checksum:
             return ret
 
         ret.add_child_cache(
-            self.checksum, self.collect_used_dir_cache(**kwargs),
+            self.hash_info.value, self.collect_used_dir_cache(**kwargs),
         )
 
         return ret
 
     @classmethod
-    def _validate_output_path(cls, path):
+    def _validate_output_path(cls, path, stage=None):
         from dvc.dvcfile import is_valid_filename
 
         if is_valid_filename(path):
             raise cls.IsStageFileError(path)
+
+        if stage:
+            check = stage.repo.tree.dvcignore.check_ignore(path)
+            if check.match:
+                raise cls.IsIgnoredError(check)
+
+    def _check_can_merge(self, out):
+        if self.scheme != out.scheme:
+            raise MergeError("unable to auto-merge outputs of different types")
+
+        my = self.dumpd()
+        other = out.dumpd()
+
+        ignored = [
+            self.tree.PARAM_CHECKSUM,
+            HashInfo.PARAM_SIZE,
+            HashInfo.PARAM_NFILES,
+        ]
+
+        for opt in ignored:
+            my.pop(opt, None)
+            other.pop(opt, None)
+
+        if my != other:
+            raise MergeError(
+                "unable to auto-merge outputs with different options"
+            )
+
+        if not out.is_dir_checksum:
+            raise MergeError(
+                "unable to auto-merge outputs that are not directories"
+            )
+
+    def merge(self, ancestor, other):
+        assert other
+
+        if ancestor:
+            self._check_can_merge(ancestor)
+            ancestor_info = ancestor.hash_info
+        else:
+            ancestor_info = None
+
+        self._check_can_merge(self)
+        self._check_can_merge(other)
+
+        self.hash_info = self.cache.merge(
+            ancestor_info, self.hash_info, other.hash_info
+        )

dvc/output/gs.py

@@ -1,7 +1,7 @@
-from dvc.output.s3 import S3Output
+from dvc.output.base import BaseOutput
 
 from ..tree.gs import GSTree
 
 
-class GSOutput(S3Output):
+class GSOutput(BaseOutput):
     TREE_CLS = GSTree

dvc/output/local.py

@@ -87,8 +87,9 @@ class LocalOutput(BaseOutput):
 
         name = "metrics" if self.metric else "plot"
         if os.path.isdir(path):
-            msg = "directory '{}' cannot be used as {}."
-            raise IsADirectoryError(msg.format(self.path_info, name))
+            msg = "directory '%s' cannot be used as %s."
+            logger.debug(msg, str(self.path_info), name)
+            return
 
         if not istextfile(path):
             msg = "binary file '{}' cannot be used as {}."

dvc/output/webhdfs.py

@@ -0,0 +1,7 @@
+from dvc.output.base import BaseOutput
+
+from ..tree.webhdfs import WebHDFSTree
+
+
+class WebHDFSOutput(BaseOutput):
+    TREE_CLS = WebHDFSTree

dvc/parsing/__init__.py

@@ -0,0 +1,287 @@
+import logging
+import os
+from collections import defaultdict
+from collections.abc import Mapping, Sequence
+from copy import deepcopy
+from itertools import starmap
+from typing import TYPE_CHECKING, List, Set
+
+from funcy import join
+
+from dvc.dependency.param import ParamsDependency
+from dvc.exceptions import DvcException
+from dvc.parsing.interpolate import ParseError
+from dvc.path_info import PathInfo
+from dvc.utils import relpath
+
+from .context import (
+    Context,
+    ContextError,
+    KeyNotInContext,
+    MergeError,
+    Meta,
+    Node,
+    ParamsFileNotFound,
+    SetError,
+)
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+logger = logging.getLogger(__name__)
+
+STAGES_KWD = "stages"
+VARS_KWD = "vars"
+WDIR_KWD = "wdir"
+DEFAULT_PARAMS_FILE = ParamsDependency.DEFAULT_PARAMS_FILE
+PARAMS_KWD = "params"
+FOREACH_KWD = "foreach"
+DO_KWD = "do"
+SET_KWD = "set"
+
+DEFAULT_SENTINEL = object()
+
+JOIN = "@"
+
+
+class ResolveError(DvcException):
+    pass
+
+
+def format_and_raise(exc, msg, path):
+    spacing = "\n" if isinstance(exc, (ParseError, MergeError)) else " "
+    message = f"failed to parse {msg} in '{path}':{spacing}{str(exc)}"
+
+    # FIXME: cannot reraise because of how we log "cause" of the exception
+    # the error message is verbose, hence need control over the spacing
+    _reraise_err(ResolveError, message, from_exc=exc)
+
+
+def _reraise_err(exc_cls, *args, from_exc=None):
+    err = exc_cls(*args)
+    if from_exc and logger.isEnabledFor(logging.DEBUG):
+        raise err from from_exc
+    raise err
+
+
+class DataResolver:
+    def __init__(self, repo: "Repo", wdir: PathInfo, d: dict):
+        self.data: dict = d
+        self.wdir = wdir
+        self.repo = repo
+        self.tree = self.repo.tree
+        self.imported_files: Set[str] = set()
+        self.relpath = relpath(self.wdir / "dvc.yaml")
+
+        to_import: PathInfo = wdir / DEFAULT_PARAMS_FILE
+        if self.tree.exists(to_import):
+            self.imported_files = {os.path.abspath(to_import)}
+            self.global_ctx = Context.load_from(self.tree, to_import)
+        else:
+            self.global_ctx = Context()
+            logger.debug(
+                "%s does not exist, it won't be used in parametrization",
+                to_import,
+            )
+
+        vars_ = d.get(VARS_KWD, [])
+        try:
+            self.load_from_vars(
+                self.global_ctx, vars_, wdir, skip_imports=self.imported_files
+            )
+        except (ParamsFileNotFound, MergeError) as exc:
+            format_and_raise(exc, "'vars'", self.relpath)
+
+    def load_from_vars(
+        self,
+        context: "Context",
+        vars_: List,
+        wdir: PathInfo,
+        skip_imports: Set[str],
+        stage_name: str = None,
+    ):
+        stage_name = stage_name or ""
+        for index, item in enumerate(vars_):
+            assert isinstance(item, (str, dict))
+            if isinstance(item, str):
+                path_info = wdir / item
+                path = os.path.abspath(path_info)
+                if path in skip_imports:
+                    continue
+
+                context.merge_from(self.tree, path_info)
+                skip_imports.add(path)
+            else:
+                joiner = "." if stage_name else ""
+                meta = Meta(source=f"{stage_name}{joiner}vars[{index}]")
+                context.merge_update(Context(item, meta=meta))
+
+    def _resolve_entry(self, name: str, definition):
+        context = Context.clone(self.global_ctx)
+        if FOREACH_KWD in definition:
+            assert DO_KWD in definition
+            self.set_context_from(
+                context, definition.get(SET_KWD, {}), source=[name, "set"]
+            )
+            return self._foreach(
+                context, name, definition[FOREACH_KWD], definition[DO_KWD]
+            )
+
+        try:
+            return self._resolve_stage(context, name, definition)
+        except ContextError as exc:
+            format_and_raise(exc, f"stage '{name}'", self.relpath)
+
+    def resolve(self):
+        stages = self.data.get(STAGES_KWD, {})
+        data = join(starmap(self._resolve_entry, stages.items())) or {}
+        logger.trace("Resolved dvc.yaml:\n%s", data)
+        return {STAGES_KWD: data}
+
+    def _resolve_stage(self, context: Context, name: str, definition) -> dict:
+        definition = deepcopy(definition)
+        self.set_context_from(
+            context, definition.pop(SET_KWD, {}), source=[name, "set"]
+        )
+        wdir = self._resolve_wdir(context, name, definition.get(WDIR_KWD))
+        if self.wdir != wdir:
+            logger.debug(
+                "Stage %s has different wdir than dvc.yaml file", name
+            )
+
+        vars_ = definition.pop(VARS_KWD, [])
+        # FIXME: Should `vars` be templatized?
+        self.load_from_vars(
+            context,
+            vars_,
+            wdir,
+            skip_imports=deepcopy(self.imported_files),
+            stage_name=name,
+        )
+
+        logger.trace(  # type: ignore[attr-defined]
+            "Context during resolution of stage %s:\n%s", name, context
+        )
+
+        with context.track():
+            resolved = {}
+            for key, value in definition.items():
+                # NOTE: we do not pop "wdir", and resolve it again
+                # this does not affect anything and is done to try to
+                # track the source of `wdir` interpolation.
+                # This works because of the side-effect that we do not
+                # allow overwriting and/or str interpolating complex objects.
+                # Fix if/when those assumptions are no longer valid.
+                try:
+                    resolved[key] = context.resolve(value)
+                except (ParseError, KeyNotInContext) as exc:
+                    format_and_raise(
+                        exc, f"'stages.{name}.{key}'", self.relpath
+                    )
+
+        # FIXME: Decide if we should track them or not (it does right now)
+        params = resolved.get(PARAMS_KWD, []) + self._resolve_params(
+            context, wdir
+        )
+        if params:
+            resolved[PARAMS_KWD] = params
+        return {name: resolved}
+
+    def _resolve_params(self, context: Context, wdir):
+        tracked = defaultdict(set)
+        for src, keys in context.tracked.items():
+            tracked[str(PathInfo(src).relative_to(wdir))].update(keys)
+
+        return [{file: list(keys)} for file, keys in tracked.items()]
+
+    def _resolve_wdir(
+        self, context: Context, name: str, wdir: str = None
+    ) -> PathInfo:
+        if not wdir:
+            return self.wdir
+
+        try:
+            wdir = str(context.resolve_str(wdir, unwrap=True))
+        except (ContextError, ParseError) as exc:
+            format_and_raise(exc, f"'stages.{name}.wdir'", self.relpath)
+        return self.wdir / str(wdir)
+
+    def _foreach(self, context: Context, name: str, foreach_data, do_data):
+        iterable = self._resolve_foreach_data(context, name, foreach_data)
+        args = (context, name, do_data, iterable)
+        it = (
+            range(len(iterable))
+            if not isinstance(iterable, Mapping)
+            else iterable
+        )
+        gen = (self._each_iter(*args, i) for i in it)
+        return join(gen)
+
+    def _each_iter(self, context: Context, name: str, do_data, iterable, key):
+        value = iterable[key]
+        c = Context.clone(context)
+        suffix = c["item"] = value
+        if isinstance(iterable, Mapping):
+            suffix = c["key"] = key
+
+        generated = f"{name}{JOIN}{suffix}"
+        try:
+            return self._resolve_stage(c, generated, do_data)
+        except ContextError as exc:
+            # pylint: disable=no-member
+            if isinstance(exc, MergeError) and exc.key in self._inserted_keys(
+                iterable
+            ):
+                raise ResolveError(
+                    f"attempted to redefine '{exc.key}' in stage '{generated}'"
+                    " generated through 'foreach'"
+                )
+            format_and_raise(
+                exc, f"stage '{generated}' (gen. from '{name}')", self.relpath
+            )
+
+    def _resolve_foreach_data(
+        self, context: "Context", name: str, foreach_data
+    ):
+        try:
+            iterable = context.resolve(foreach_data, unwrap=False)
+        except (ContextError, ParseError) as exc:
+            format_and_raise(exc, f"'stages.{name}.foreach'", self.relpath)
+        if isinstance(iterable, str) or not isinstance(
+            iterable, (Sequence, Mapping)
+        ):
+            raise ResolveError(
+                f"failed to resolve 'stages.{name}.foreach'"
+                f" in '{self.relpath}': expected list/dictionary, got "
+                + type(
+                    iterable.value if isinstance(iterable, Node) else iterable
+                ).__name__
+            )
+
+        warn_for = [k for k in self._inserted_keys(iterable) if k in context]
+        if warn_for:
+            logger.warning(
+                "%s %s already specified, "
+                "will be overwritten for stages generated from '%s'",
+                " and ".join(warn_for),
+                "is" if len(warn_for) == 1 else "are",
+                name,
+            )
+
+        return iterable
+
+    @staticmethod
+    def _inserted_keys(iterable):
+        keys = ["item"]
+        if isinstance(iterable, Mapping):
+            keys.append("key")
+        return keys
+
+    @classmethod
+    def set_context_from(cls, context: Context, to_set, source=None):
+        try:
+            for key, value in to_set.items():
+                src_set = [*(source or []), key]
+                context.set(key, value, source=".".join(src_set))
+        except SetError as exc:
+            _reraise_err(ResolveError, str(exc), from_exc=exc)

dvc/parsing/context.py

@@ -0,0 +1,436 @@
+import os
+from abc import ABC, abstractmethod
+from collections import defaultdict
+from collections.abc import Mapping, MutableMapping, MutableSequence, Sequence
+from contextlib import contextmanager
+from copy import deepcopy
+from dataclasses import dataclass, field, replace
+from typing import Any, List, Optional, Union
+
+from funcy import identity, rpartial
+
+from dvc.exceptions import DvcException
+from dvc.parsing.interpolate import (
+    ParseError,
+    get_expression,
+    get_matches,
+    is_exact_string,
+    is_interpolated_string,
+    str_interpolate,
+)
+from dvc.path_info import PathInfo
+from dvc.utils import relpath
+from dvc.utils.serialize import LOADERS
+
+SeqOrMap = Union[Sequence, Mapping]
+
+
+class ContextError(DvcException):
+    pass
+
+
+class SetError(ContextError):
+    pass
+
+
+class MergeError(ContextError):
+    def __init__(self, key, new, into):
+        self.key = key
+        if not isinstance(into[key], Node) or not isinstance(new, Node):
+            super().__init__(
+                f"cannot merge '{key}' as it already exists in {into}"
+            )
+            return
+
+        preexisting = into[key].meta.source
+        new_src = new.meta.source
+        path = new.meta.path()
+        super().__init__(
+            f"cannot redefine '{path}' from '{new_src}'"
+            f" as it already exists in '{preexisting}'"
+        )
+
+
+class ParamsFileNotFound(ContextError):
+    pass
+
+
+class KeyNotInContext(ContextError):
+    def __init__(self, key) -> None:
+        self.key = key
+        super().__init__(f"Could not find '{key}'")
+
+
+def _merge(into, update, overwrite):
+    for key, val in update.items():
+        if isinstance(into.get(key), Mapping) and isinstance(val, Mapping):
+            _merge(into[key], val, overwrite)
+        else:
+            if key in into and not overwrite:
+                raise MergeError(key, val, into)
+            into[key] = val
+            assert isinstance(into[key], Node)
+
+
+@dataclass
+class Meta:
+    source: Optional[str] = None
+    dpaths: List[str] = field(default_factory=list)
+    local: bool = True
+
+    @staticmethod
+    def update_path(meta: "Meta", path: Union[str, int]):
+        dpaths = meta.dpaths[:] + [str(path)]
+        return replace(meta, dpaths=dpaths)
+
+    def __str__(self):
+        string = self.source or "<local>"
+        string += ":" + self.path()
+        return string
+
+    def path(self):
+        return ".".join(self.dpaths)
+
+
+def _default_meta():
+    return Meta(source=None)
+
+
+class Node:
+    def get_sources(self):
+        raise NotImplementedError
+
+    @property
+    @abstractmethod
+    def value(self):
+        pass
+
+
+@dataclass
+class Value(Node):
+    _value: Any
+    meta: Meta = field(
+        compare=False, default_factory=_default_meta, repr=False
+    )
+
+    def __repr__(self):
+        return repr(self._value)
+
+    def __str__(self) -> str:
+        return str(self._value)
+
+    def get_sources(self):
+        return {self.meta.source: self.meta.path()}
+
+    @property
+    def value(self):
+        return self._value
+
+
+PRIMITIVES = (int, float, str, bytes, bool)
+
+
+class Container(Node, ABC):
+    meta: Meta
+    data: Union[list, dict]
+    _key_transform = staticmethod(identity)
+
+    def __init__(self, meta=None) -> None:
+        self.meta = meta or _default_meta()
+
+    def _convert(self, key, value):
+        meta = Meta.update_path(self.meta, key)
+        return self._convert_with_meta(value, meta)
+
+    @staticmethod
+    def _convert_with_meta(value, meta: Meta = None):
+        if value is None or isinstance(value, PRIMITIVES):
+            assert meta
+            return Value(value, meta=meta)
+        elif isinstance(value, Node):
+            return value
+        elif isinstance(value, (list, dict)):
+            assert meta
+            container = CtxDict if isinstance(value, dict) else CtxList
+            return container(value, meta=meta)
+        else:
+            msg = (
+                "Unsupported value of type "
+                f"'{type(value).__name__}' in '{meta}'"
+            )
+            raise TypeError(msg)
+
+    def __repr__(self):
+        return repr(self.data)
+
+    def __getitem__(self, key):
+        return self.data[key]
+
+    def __setitem__(self, key, value):
+        self.data[key] = self._convert(key, value)
+
+    def __delitem__(self, key):
+        del self.data[key]
+
+    def __len__(self):
+        return len(self.data)
+
+    def __iter__(self):
+        return iter(self.data)
+
+    def __eq__(self, o):
+        container = type(self)
+        if isinstance(o, container):
+            return o.data == self.data
+        return container(o) == self
+
+    def select(self, key: str):
+        index, *rems = key.split(sep=".", maxsplit=1)
+        index = index.strip()
+        index = self._key_transform(index)
+        try:
+            d = self[index]
+        except LookupError as exc:
+            raise ValueError(
+                f"Could not find '{index}' in {self.data}"
+            ) from exc
+
+        if not rems:
+            return d
+
+        rem = rems[0]
+        if not isinstance(d, Container):
+            raise ValueError(
+                f"{index} is a primitive value, cannot get '{rem}'"
+            )
+        return d.select(rem)
+
+    def get_sources(self):
+        return {}
+
+
+class CtxList(Container, MutableSequence):
+    _key_transform = staticmethod(int)
+
+    def __init__(self, values: Sequence, meta: Meta = None):
+        super().__init__(meta=meta)
+        self.data: list = []
+        self.extend(values)
+
+    def insert(self, index: int, value):
+        self.data.insert(index, self._convert(index, value))
+
+    def get_sources(self):
+        return {self.meta.source: self.meta.path()}
+
+    @property
+    def value(self):
+        return [node.value for node in self]
+
+
+class CtxDict(Container, MutableMapping):
+    def __init__(self, mapping: Mapping = None, meta: Meta = None, **kwargs):
+        super().__init__(meta=meta)
+
+        self.data: dict = {}
+        if mapping:
+            self.update(mapping)
+        self.update(kwargs)
+
+    def __setitem__(self, key, value):
+        if not isinstance(key, str):
+            # limitation for the interpolation
+            # ignore other kinds of keys
+            return
+        return super().__setitem__(key, value)
+
+    def merge_update(self, *args, overwrite=False):
+        for d in args:
+            _merge(self, d, overwrite=overwrite)
+
+    @property
+    def value(self):
+        return {key: node.value for key, node in self.items()}
+
+
+class Context(CtxDict):
+    def __init__(self, *args, **kwargs):
+        """
+        Top level mutable dict, with some helpers to create context and track
+        """
+        super().__init__(*args, **kwargs)
+        self._track = False
+        self._tracked_data = defaultdict(set)
+
+    @contextmanager
+    def track(self):
+        self._track = True
+        yield
+        self._track = False
+
+    def _track_data(self, node):
+        if not self._track:
+            return
+
+        if node.meta and node.meta.local:
+            return
+
+        for source, keys in node.get_sources().items():
+            if not source:
+                continue
+            params_file = self._tracked_data[source]
+            keys = [keys] if isinstance(keys, str) else keys
+            params_file.update(keys)
+
+    @property
+    def tracked(self):
+        return self._tracked_data
+
+    def select(
+        self, key: str, unwrap=False
+    ):  # pylint: disable=arguments-differ
+        """Select the item using key, similar to `__getitem__`
+           but can track the usage of the data on interpolation
+           as well and can get from nested data structure by using
+           "." separated key (eg: "key1.key2.key3")
+
+        Args:
+            key: key to select value from
+            unwrap: Convert CtxList/CtxDict/Value items to it's original data
+                    Defaults to False. Note that the default is different from
+                    `resolve`.
+        """
+        try:
+            node = super().select(key)
+        except ValueError as exc:
+            raise KeyNotInContext(key) from exc
+
+        self._track_data(node)
+        assert isinstance(node, Node)
+        return node.value if unwrap else node
+
+    @classmethod
+    def load_from(cls, tree, path: PathInfo) -> "Context":
+        file = relpath(path)
+        if not tree.exists(path):
+            raise ParamsFileNotFound(f"'{file}' does not exist")
+
+        _, ext = os.path.splitext(file)
+        loader = LOADERS[ext]
+
+        meta = Meta(source=file, local=False)
+        return cls(loader(path, tree=tree), meta=meta)
+
+    def merge_from(self, tree, path: PathInfo, overwrite=False):
+        self.merge_update(Context.load_from(tree, path), overwrite=overwrite)
+
+    @classmethod
+    def clone(cls, ctx: "Context") -> "Context":
+        """Clones given context."""
+        return cls(deepcopy(ctx.data))
+
+    def set(self, key, value, source=None):
+        """
+        Sets a value, either non-interpolated values to a key,
+        or an interpolated string after resolving it.
+
+        >>> c = Context({"foo": "foo", "bar": [1, 2], "lorem": {"a": "z"}})
+        >>> c.set("foobar", "${bar}")
+        >>> c
+        {'foo': 'foo', 'bar': [1, 2], 'lorem': {'a': 'z'}, 'foobar': [1, 2]}
+        """
+        try:
+            self._set(key, value, source)
+        except (ParseError, ValueError, ContextError) as exc:
+            sp = "\n" if isinstance(exc, ParseError) else " "
+            raise SetError(f"Failed to set '{source}':{sp}{str(exc)}") from exc
+
+    def _set(self, key, value, source):
+        if key in self:
+            raise ValueError(f"Cannot set '{key}', key already exists")
+        if isinstance(value, str):
+            self._check_joined_with_interpolation(key, value)
+            value = self.resolve_str(value, unwrap=False)
+        elif isinstance(value, (Sequence, Mapping)):
+            self._check_not_nested_collection(key, value)
+            self._check_interpolation_collection(key, value)
+        self[key] = self._convert_with_meta(value, Meta(source=source))
+
+    def resolve(self, src, unwrap=True):
+        """Recursively resolves interpolation and returns resolved data.
+
+        Args:
+            src: Data (str/list/dict etc.) to resolve
+            unwrap: Unwrap CtxDict/CtxList/Value to it's original data if
+                    inside `src`. Defaults to True.
+
+        >>> c = Context({"three": 3})
+        >>> c.resolve({"lst": [1, 2, "${three}"]})
+        {'lst': [1, 2, 3]}
+        """
+        Seq = (list, tuple, set)
+
+        resolve = rpartial(self.resolve, unwrap)
+        if isinstance(src, Mapping):
+            return dict(map(resolve, kv) for kv in src.items())
+        elif isinstance(src, Seq):
+            return type(src)(map(resolve, src))
+        elif isinstance(src, str):
+            return self.resolve_str(src, unwrap=unwrap)
+        return src
+
+    def resolve_str(self, src: str, unwrap=True):
+        """Resolves interpolated string to it's original value,
+        or in case of multiple interpolations, a combined string.
+
+        >>> c = Context({"enabled": True})
+        >>> c.resolve_str("${enabled}")
+        True
+        >>> c.resolve_str("enabled? ${enabled}")
+        'enabled? True'
+        """
+        matches = get_matches(src)
+        if is_exact_string(src, matches):
+            # replace "${enabled}", if `enabled` is a boolean, with it's actual
+            # value rather than it's string counterparts.
+            expr = get_expression(matches[0])
+            return self.select(expr, unwrap=unwrap)
+        # but not "${num} days"
+        return str_interpolate(src, matches, self)
+
+    @staticmethod
+    def _check_not_nested_collection(key: str, value: SeqOrMap):
+        values = value.values() if isinstance(value, Mapping) else value
+        has_nested = any(
+            not isinstance(item, str) and isinstance(item, (Mapping, Sequence))
+            for item in values
+        )
+        if has_nested:
+            raise ValueError(f"Cannot set '{key}', has nested dict/list")
+
+    @staticmethod
+    def _check_interpolation_collection(key: str, value: SeqOrMap):
+        values = value.values() if isinstance(value, Mapping) else value
+        interpolated = any(is_interpolated_string(item) for item in values)
+        if interpolated:
+            raise ValueError(
+                f"Cannot set '{key}', "
+                "having interpolation inside "
+                f"'{type(value).__name__}' is not supported."
+            )
+
+    @staticmethod
+    def _check_joined_with_interpolation(key: str, value: str):
+        matches = get_matches(value)
+        if matches and not is_exact_string(value, matches):
+            raise ValueError(
+                f"Cannot set '{key}', "
+                "joining string with interpolated string "
+                "is not supported"
+            )
+
+
+if __name__ == "__main__":
+    import doctest
+
+    doctest.testmod()

dvc/parsing/interpolate.py

@@ -0,0 +1,115 @@
+import re
+import typing
+
+from pyparsing import (
+    CharsNotIn,
+    ParseException,
+    ParserElement,
+    Suppress,
+    ZeroOrMore,
+)
+
+from dvc.exceptions import DvcException
+from dvc.utils import colorize
+
+if typing.TYPE_CHECKING:
+    from typing import List, Match
+
+    from .context import Context
+
+LBRACK = "["
+RBRACK = "]"
+PERIOD = "."
+KEYCRE = re.compile(
+    r"""
+    (?<!\\)                   # escape \${} or ${{}}
+    \$                        # starts with $
+    (?:({{)|({))              # either starts with double braces or single
+    (.*?)                     # match every char inside
+    (?(1)}})(?(2)})           # end with same kinds of braces it opened with
+""",
+    re.VERBOSE,
+)
+
+ParserElement.enablePackrat()
+
+
+word = CharsNotIn(f"{PERIOD}{LBRACK}{RBRACK}")
+idx = Suppress(LBRACK) + word + Suppress(RBRACK)
+attr = Suppress(PERIOD) + word
+parser = word + ZeroOrMore(attr ^ idx)
+parser.setParseAction(PERIOD.join)
+
+
+class ParseError(DvcException):
+    pass
+
+
+def get_matches(template: str):
+    return list(KEYCRE.finditer(template))
+
+
+def is_interpolated_string(val):
+    return bool(get_matches(val)) if isinstance(val, str) else False
+
+
+def format_and_raise_parse_error(exc):
+    raise ParseError(_format_exc_msg(exc))
+
+
+def _format_exc_msg(exc: ParseException):
+    exc.loc += 2  # 2 because we append `${` at the start of expr below
+
+    expr = exc.pstr
+    exc.pstr = "${" + exc.pstr + "}"
+    error = ParseException.explain(exc, depth=0)
+
+    _, pointer, *explains = error.splitlines()
+    pstr = "{brace_open}{expr}{brace_close}".format(
+        brace_open=colorize("${", color="blue"),
+        expr=colorize(expr, color="magenta"),
+        brace_close=colorize("}", color="blue"),
+    )
+    msg = "\n".join(explains)
+    pointer = colorize(pointer, color="red")
+    return "\n".join([pstr, pointer, colorize(msg, color="red", style="bold")])
+
+
+def parse_expr(s: str):
+    try:
+        result = parser.parseString(s, parseAll=True)
+    except ParseException as exc:
+        format_and_raise_parse_error(exc)
+
+    joined = result.asList()
+    assert len(joined) == 1
+    return joined[0]
+
+
+def get_expression(match: "Match"):
+    _, _, inner = match.groups()
+    return parse_expr(inner)
+
+
+def str_interpolate(template: str, matches: "List[Match]", context: "Context"):
+    from .context import PRIMITIVES
+
+    index, buf = 0, ""
+    for match in matches:
+        start, end = match.span(0)
+        expr = get_expression(match)
+        value = context.select(expr, unwrap=True)
+        if value is not None and not isinstance(value, PRIMITIVES):
+            raise ParseError(
+                f"Cannot interpolate data of type '{type(value).__name__}'"
+            )
+        buf += template[index:start] + str(value)
+        index = end
+    buf += template[index:]
+    # regex already backtracks and avoids any `${` starting with
+    # backslashes(`\`). We just need to replace those by `${`.
+    return buf.replace(r"\${", "${")
+
+
+def is_exact_string(src: str, matches: "List[Match]"):
+    return len(matches) == 1 and src == matches[0].group(0)

dvc/path_info.py

@@ -2,6 +2,7 @@
 import os
 import pathlib
 import posixpath
+from typing import Callable
 from urllib.parse import urlparse
 
 from funcy import cached_property
@@ -75,6 +76,17 @@ class PathInfo(pathlib.PurePath, _BasePath):
         n = len(other._cparts)
         return len(self._cparts) > n and self._cparts[:n] == other._cparts
 
+    def relative_to(self, other):  # pylint: disable=arguments-differ
+        # pathlib relative_to raises exception when one path is not a direct
+        # descendant of the other when os.path.relpath would return abspath.
+        # For DVC PathInfo we only need the relpath behavior.
+        # See: https://bugs.python.org/issue40358
+        try:
+            path = super().relative_to(other)
+        except ValueError:
+            path = relpath(self, other)
+        return self.__class__(path)
+
 
 class WindowsPathInfo(PathInfo, pathlib.PureWindowsPath):
     pass
@@ -113,7 +125,6 @@ class URLInfo(_BasePath):
         p = urlparse(url)
         assert not p.query and not p.params and not p.fragment
         assert p.password is None
-
         self._fill_parts(p.scheme, p.hostname, p.username, p.port, p.path)
 
     @classmethod
@@ -243,7 +254,7 @@ class CloudURLInfo(URLInfo):
 
 
 class HTTPURLInfo(URLInfo):
-    __hash__ = URLInfo.__hash__
+    __hash__: Callable[["HTTPURLInfo"], int] = URLInfo.__hash__
 
     def __init__(self, url):
         p = urlparse(url)
@@ -253,6 +264,15 @@ class HTTPURLInfo(URLInfo):
         self.query = p.query
         self.fragment = p.fragment
 
+    def replace(self, path=None):
+        return self.from_parts(
+            *self._base_parts,
+            params=self.params,
+            query=self.query,
+            fragment=self.fragment,
+            path=path,
+        )
+
     @classmethod
     def from_parts(
         cls,
@@ -315,3 +335,11 @@ class HTTPURLInfo(URLInfo):
             and self._path == other._path
             and self._extra_parts == other._extra_parts
         )
+
+
+class WebDAVURLInfo(URLInfo):
+    @cached_property
+    def url(self):
+        return "{}://{}{}".format(
+            self.scheme.replace("webdav", "http"), self.netloc, self._spath
+        )

dvc/pathspec_math.py

@@ -3,11 +3,14 @@
 # of two path specification patterns with different base
 # All the operations follow the documents of `gitignore`
 import os
+from collections import namedtuple
 
 from pathspec.util import normalize_file
 
 from dvc.utils import relpath
 
+PatternInfo = namedtuple("PatternInfo", ["patterns", "file_info"])
+
 
 def _not_ignore(rule):
     return (True, rule[1:]) if rule.startswith("!") else (False, rule)
@@ -59,7 +62,10 @@ def _change_dirname(dirname, pattern_list, new_dirname):
     if rel.startswith(".."):
         raise ValueError("change dirname can only change to parent path")
 
-    return [change_rule(rule, rel) for rule in pattern_list]
+    return [
+        PatternInfo(change_rule(rule.patterns, rel), rule.file_info)
+        for rule in pattern_list
+    ]
 
 
 def merge_patterns(pattern_a, prefix_a, pattern_b, prefix_b):

dvc/progress.py

@@ -31,9 +31,12 @@ class Tqdm(tqdm):
     BAR_FMT_NOTOTAL = (
         "{desc}{bar:b}|{postfix[info]}{n_fmt} [{elapsed}, {rate_fmt:>11}]"
     )
-    BYTES_DEFAULTS = dict(
-        unit="B", unit_scale=True, unit_divisor=1024, miniters=1
-    )
+    BYTES_DEFAULTS = {
+        "unit": "B",
+        "unit_scale": True,
+        "unit_divisor": 1024,
+        "miniters": 1,
+    }
 
     def __init__(
         self,

dvc/prompt.py

@@ -7,7 +7,7 @@ from getpass import getpass
 logger = logging.getLogger(__name__)
 
 
-def _ask(prompt, limited_to=None):
+def ask(prompt, limited_to=None):
     if not sys.stdout.isatty():
         return None
 
@@ -39,7 +39,7 @@ def confirm(statement):
         bool: whether or not specified statement was confirmed.
     """
     prompt = f"{statement} [y/n]"
-    answer = _ask(prompt, limited_to=["yes", "no", "y", "n"])
+    answer = ask(prompt, limited_to=["yes", "no", "y", "n"])
     return answer and answer.startswith("y")
 
 

dvc/remote/__init__.py

@@ -1,13 +1,10 @@
 from ..tree import get_cloud_tree
 from .base import Remote
 from .local import LocalRemote
-from .ssh import SSHRemote
 
 
 def get_remote(repo, **kwargs):
     tree = get_cloud_tree(repo, **kwargs)
     if tree.scheme == "local":
         return LocalRemote(tree)
-    if tree.scheme == "ssh":
-        return SSHRemote(tree)
     return Remote(tree)

dvc/remote/base.py

@@ -1,19 +1,59 @@
+import errno
 import hashlib
 import logging
-from functools import wraps
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from functools import partial, wraps
 
+from funcy import concat
+
+from dvc.exceptions import DownloadError, UploadError
+from dvc.hash_info import HashInfo
+
+from ..progress import Tqdm
 from .index import RemoteIndex, RemoteIndexNoop
 
 logger = logging.getLogger(__name__)
 
+STATUS_OK = 1
+STATUS_MISSING = 2
+STATUS_NEW = 3
+STATUS_DELETED = 4
+
+STATUS_MAP = {
+    # (local_exists, remote_exists)
+    (True, True): STATUS_OK,
+    (False, False): STATUS_MISSING,
+    (True, False): STATUS_NEW,
+    (False, True): STATUS_DELETED,
+}
+
+
+def _log_exceptions(func, operation):
+    @wraps(func)
+    def wrapper(from_info, to_info, *args, **kwargs):
+        try:
+            func(from_info, to_info, *args, **kwargs)
+            return 0
+        except Exception as exc:  # pylint: disable=broad-except
+            # NOTE: this means we ran out of file descriptors and there is no
+            # reason to try to proceed, as we will hit this error anyways.
+            # pylint: disable=no-member
+            if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
+                raise
+
+            logger.exception(
+                "failed to %s '%s' to '%s'", operation, from_info, to_info
+            )
+            return 1
+
+    return wrapper
+
 
 def index_locked(f):
     @wraps(f)
-    def wrapper(obj, named_cache, remote, *args, **kwargs):
-        if hasattr(remote, "index"):
-            with remote.index:
-                return f(obj, named_cache, remote, *args, **kwargs)
-        return f(obj, named_cache, remote, *args, **kwargs)
+    def wrapper(obj, *args, **kwargs):
+        with obj.index:
+            return f(obj, *args, **kwargs)
 
     return wrapper
 
@@ -28,8 +68,11 @@ class Remote:
     INDEX_CLS = RemoteIndex
 
     def __init__(self, tree):
+        from dvc.cache import get_cloud_cache
+
         self.tree = tree
         self.repo = tree.repo
+        self.cache = get_cloud_cache(self.tree)
 
         config = tree.config
         url = config.get("url")
@@ -47,123 +90,393 @@ class Remote:
             path_info=self.tree.path_info or "No path",
         )
 
-    @property
-    def cache(self):
-        return getattr(self.repo.cache, self.tree.scheme)
+    @index_locked
+    def gc(self, *args, **kwargs):
+        removed = self.cache.gc(*args, **kwargs)
 
-    def hashes_exist(self, hashes, jobs=None, name=None):
-        """Check if the given hashes are stored in the remote.
+        if removed:
+            self.index.clear()
 
-        There are two ways of performing this check:
+        return removed
 
-        - Traverse method: Get a list of all the files in the remote
-            (traversing the cache directory) and compare it with
-            the given hashes. Cache entries will be retrieved in parallel
-            threads according to prefix (i.e. entries starting with, "00...",
-            "01...", and so on) and a progress bar will be displayed.
-
-        - Exists method: For each given hash, run the `exists`
-            method and filter the hashes that aren't on the remote.
-            This is done in parallel threads.
-            It also shows a progress bar when performing the check.
-
-        The reason for such an odd logic is that most of the remotes
-        take much shorter time to just retrieve everything they have under
-        a certain prefix (e.g. s3, gs, ssh, hdfs). Other remotes that can
-        check if particular file exists much quicker, use their own
-        implementation of hashes_exist (see ssh, local).
-
-        Which method to use will be automatically determined after estimating
-        the size of the remote cache, and comparing the estimated size with
-        len(hashes). To estimate the size of the remote cache, we fetch
-        a small subset of cache entries (i.e. entries starting with "00...").
-        Based on the number of entries in that subset, the size of the full
-        cache can be estimated, since the cache is evenly distributed according
-        to hash.
-
-        Returns:
-            A list with hashes that were found in the remote
-        """
-        # Remotes which do not use traverse prefix should override
-        # hashes_exist() (see ssh, local)
-        assert self.tree.TRAVERSE_PREFIX_LEN >= 2
+    @index_locked
+    def status(
+        self,
+        cache,
+        named_cache,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+        log_missing=True,
+    ):
+        # Return flattened dict containing all status info
+        dir_status, file_status, _ = self._status(
+            cache,
+            named_cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=download,
+            log_missing=log_missing,
+        )
+        return dict(dir_status, **file_status)
 
+    def hashes_exist(self, hashes, **kwargs):
         hashes = set(hashes)
         indexed_hashes = set(self.index.intersection(hashes))
         hashes -= indexed_hashes
+        indexed_hashes = list(indexed_hashes)
         logger.debug("Matched '{}' indexed hashes".format(len(indexed_hashes)))
         if not hashes:
             return indexed_hashes
 
-        if len(hashes) == 1 or not self.tree.CAN_TRAVERSE:
-            remote_hashes = self.tree.list_hashes_exists(hashes, jobs, name)
-            return list(indexed_hashes) + remote_hashes
+        return indexed_hashes + self.cache.hashes_exist(list(hashes), **kwargs)
 
-        # Max remote size allowed for us to use traverse method
-        remote_size, remote_hashes = self.tree.estimate_remote_size(
-            hashes, name
+    def _status(
+        self,
+        cache,
+        named_cache,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+        log_missing=True,
+    ):
+        """Return a tuple of (dir_status_info, file_status_info, dir_contents).
+
+        dir_status_info contains status for .dir files, file_status_info
+        contains status for all other files, and dir_contents is a dict of
+        {dir_hash: set(file_hash, ...)} which can be used to map
+        a .dir file to its file contents.
+        """
+        logger.debug(f"Preparing to collect status from {self.tree.path_info}")
+        md5s = set(named_cache.scheme_keys(cache.tree.scheme))
+
+        logger.debug("Collecting information from local cache...")
+        local_exists = frozenset(
+            cache.hashes_exist(md5s, jobs=jobs, name=cache.cache_dir)
         )
 
-        traverse_pages = remote_size / self.tree.LIST_OBJECT_PAGE_SIZE
-        # For sufficiently large remotes, traverse must be weighted to account
-        # for performance overhead from large lists/sets.
-        # From testing with S3, for remotes with 1M+ files, object_exists is
-        # faster until len(hashes) is at least 10k~100k
-        if remote_size > self.tree.TRAVERSE_THRESHOLD_SIZE:
-            traverse_weight = (
-                traverse_pages * self.tree.TRAVERSE_WEIGHT_MULTIPLIER
-            )
+        # This is a performance optimization. We can safely assume that,
+        # if the resources that we want to fetch are already cached,
+        # there's no need to check the remote storage for the existence of
+        # those files.
+        if download and local_exists == md5s:
+            remote_exists = local_exists
         else:
-            traverse_weight = traverse_pages
-        if len(hashes) < traverse_weight:
-            logger.debug(
-                "Large remote ('{}' hashes < '{}' traverse weight), "
-                "using object_exists for remaining hashes".format(
-                    len(hashes), traverse_weight
+            logger.debug("Collecting information from remote cache...")
+            remote_exists = set()
+            dir_md5s = set(named_cache.dir_keys(cache.tree.scheme))
+            if dir_md5s:
+                remote_exists.update(
+                    self._indexed_dir_hashes(cache, named_cache, dir_md5s)
                 )
-            )
-            return (
-                list(indexed_hashes)
-                + list(hashes & remote_hashes)
-                + self.tree.list_hashes_exists(
-                    hashes - remote_hashes, jobs, name
+                md5s.difference_update(remote_exists)
+            if md5s:
+                remote_exists.update(
+                    self.hashes_exist(
+                        md5s, jobs=jobs, name=str(self.tree.path_info)
+                    )
                 )
-            )
+        return self._make_status(
+            cache,
+            named_cache,
+            show_checksums,
+            local_exists,
+            remote_exists,
+            log_missing,
+        )
 
-        logger.debug("Querying '{}' hashes via traverse".format(len(hashes)))
-        remote_hashes = set(
-            self.tree.list_hashes_traverse(
-                remote_size, remote_hashes, jobs, name
+    def _make_status(
+        self,
+        cache,
+        named_cache,
+        show_checksums,
+        local_exists,
+        remote_exists,
+        log_missing,
+    ):
+        def make_names(hash_, names):
+            return {"name": hash_ if show_checksums else " ".join(names)}
+
+        dir_status = {}
+        file_status = {}
+        dir_contents = {}
+        for hash_, item in named_cache[cache.tree.scheme].items():
+            if item.children:
+                dir_status[hash_] = make_names(hash_, item.names)
+                dir_contents[hash_] = set()
+                for child_hash, child in item.children.items():
+                    file_status[child_hash] = make_names(
+                        child_hash, child.names
+                    )
+                    dir_contents[hash_].add(child_hash)
+            else:
+                file_status[hash_] = make_names(hash_, item.names)
+
+        self._fill_statuses(dir_status, local_exists, remote_exists)
+        self._fill_statuses(file_status, local_exists, remote_exists)
+
+        if log_missing:
+            self._log_missing_caches(dict(dir_status, **file_status))
+
+        return dir_status, file_status, dir_contents
+
+    def _indexed_dir_hashes(self, cache, named_cache, dir_md5s):
+        # Validate our index by verifying all indexed .dir hashes
+        # still exist on the remote
+        indexed_dirs = set(self.index.dir_hashes())
+        indexed_dir_exists = set()
+        if indexed_dirs:
+            indexed_dir_exists.update(
+                self.cache.list_hashes_exists(indexed_dirs)
+            )
+            missing_dirs = indexed_dirs.difference(indexed_dir_exists)
+            if missing_dirs:
+                logger.debug(
+                    "Remote cache missing indexed .dir hashes '{}', "
+                    "clearing remote index".format(", ".join(missing_dirs))
+                )
+                self.index.clear()
+
+        # Check if non-indexed (new) dir hashes exist on remote
+        dir_exists = dir_md5s.intersection(indexed_dir_exists)
+        dir_exists.update(self.cache.list_hashes_exists(dir_md5s - dir_exists))
+
+        # If .dir hash exists on the remote, assume directory contents
+        # still exists on the remote
+        for dir_hash in dir_exists:
+            file_hashes = list(
+                named_cache.child_keys(cache.tree.scheme, dir_hash)
+            )
+            if dir_hash not in self.index:
+                logger.debug(
+                    "Indexing new .dir '{}' with '{}' nested files".format(
+                        dir_hash, len(file_hashes)
+                    )
+                )
+                self.index.update([dir_hash], file_hashes)
+            yield dir_hash
+            yield from file_hashes
+
+    @staticmethod
+    def _fill_statuses(hash_info_dir, local_exists, remote_exists):
+        # Using sets because they are way faster for lookups
+        local = set(local_exists)
+        remote = set(remote_exists)
+
+        for md5, info in hash_info_dir.items():
+            status = STATUS_MAP[(md5 in local, md5 in remote)]
+            info["status"] = status
+
+    def _get_plans(self, cache_obj, download, status_info, status):
+        cache = []
+        path_infos = []
+        names = []
+        hashes = []
+        missing = []
+        for md5, info in Tqdm(
+            status_info.items(), desc="Analysing status", unit="file"
+        ):
+            if info["status"] == status:
+                cache.append(cache_obj.tree.hash_to_path_info(md5))
+                path_infos.append(self.tree.hash_to_path_info(md5))
+                names.append(info["name"])
+                hashes.append(md5)
+            elif info["status"] == STATUS_MISSING:
+                missing.append(md5)
+
+        if download:
+            to_infos = cache
+            from_infos = path_infos
+        else:
+            to_infos = path_infos
+            from_infos = cache
+
+        return (from_infos, to_infos, names, hashes), missing
+
+    def _process(
+        self,
+        cache,
+        named_cache,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+    ):
+        logger.debug(
+            "Preparing to {} '{}'".format(
+                "download data from" if download else "upload data to",
+                self.tree.path_info,
             )
         )
-        return list(indexed_hashes) + list(hashes & set(remote_hashes))
 
-    @classmethod
+        if download:
+            func = partial(
+                _log_exceptions(self.tree.download, "download"),
+                dir_mode=cache.tree.dir_mode,
+                file_mode=cache.tree.file_mode,
+            )
+            status = STATUS_DELETED
+            desc = "Downloading"
+        else:
+            func = _log_exceptions(self.tree.upload, "upload")
+            status = STATUS_NEW
+            desc = "Uploading"
+
+        if jobs is None:
+            jobs = self.tree.JOBS
+
+        dir_status, file_status, dir_contents = self._status(
+            cache,
+            named_cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=download,
+        )
+
+        dir_plans, _ = self._get_plans(cache, download, dir_status, status)
+        file_plans, missing_files = self._get_plans(
+            cache, download, file_status, status
+        )
+
+        total = len(dir_plans[0]) + len(file_plans[0])
+        if total == 0:
+            return 0
+
+        with Tqdm(total=total, unit="file", desc=desc) as pbar:
+            func = pbar.wrap_fn(func)
+            with ThreadPoolExecutor(max_workers=jobs) as executor:
+                if download:
+                    from_infos, to_infos, names, _ = (
+                        d + f for d, f in zip(dir_plans, file_plans)
+                    )
+                    fails = sum(
+                        executor.map(func, from_infos, to_infos, names)
+                    )
+                else:
+                    # for uploads, push files first, and any .dir files last
+
+                    file_futures = {}
+                    for from_info, to_info, name, hash_ in zip(*file_plans):
+                        file_futures[hash_] = executor.submit(
+                            func, from_info, to_info, name
+                        )
+                    dir_futures = {}
+                    for from_info, to_info, name, dir_hash in zip(*dir_plans):
+                        # if for some reason a file contained in this dir is
+                        # missing both locally and in the remote, we want to
+                        # push whatever file content we have, but should not
+                        # push .dir file
+                        for file_hash in missing_files:
+                            if file_hash in dir_contents[dir_hash]:
+                                logger.debug(
+                                    "directory '%s' contains missing files,"
+                                    "skipping .dir file upload",
+                                    name,
+                                )
+                                break
+                        else:
+                            wait_futures = {
+                                future
+                                for file_hash, future in file_futures.items()
+                                if file_hash in dir_contents[dir_hash]
+                            }
+                            dir_futures[dir_hash] = executor.submit(
+                                self._dir_upload,
+                                func,
+                                wait_futures,
+                                from_info,
+                                to_info,
+                                name,
+                            )
+                    fails = sum(
+                        future.result()
+                        for future in concat(
+                            file_futures.values(), dir_futures.values()
+                        )
+                    )
+
+        if fails:
+            if download:
+                self.index.clear()
+                raise DownloadError(fails)
+            raise UploadError(fails)
+
+        if not download:
+            # index successfully pushed dirs
+            for dir_hash, future in dir_futures.items():
+                if future.result() == 0:
+                    file_hashes = dir_contents[dir_hash]
+                    logger.debug(
+                        "Indexing pushed dir '{}' with "
+                        "'{}' nested files".format(dir_hash, len(file_hashes))
+                    )
+                    self.index.update([dir_hash], file_hashes)
+
+        return len(dir_plans[0]) + len(file_plans[0])
+
+    @staticmethod
+    def _dir_upload(func, futures, from_info, to_info, name):
+        for future in as_completed(futures):
+            if future.result():
+                # do not upload this .dir file if any file in this
+                # directory failed to upload
+                logger.debug(
+                    "failed to upload full contents of '{}', "
+                    "aborting .dir file upload".format(name)
+                )
+                logger.error(f"failed to upload '{from_info}' to '{to_info}'")
+                return 1
+        return func(from_info, to_info, name)
+
     @index_locked
-    def gc(cls, named_cache, remote, jobs=None):
-        tree = remote.tree
-        used = set(named_cache.scheme_keys("local"))
+    def push(self, cache, named_cache, jobs=None, show_checksums=False):
+        return self._process(
+            cache,
+            named_cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=False,
+        )
 
-        if tree.scheme != "":
-            used.update(named_cache.scheme_keys(tree.scheme))
+    @index_locked
+    def pull(self, cache, named_cache, jobs=None, show_checksums=False):
+        ret = self._process(
+            cache,
+            named_cache,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=True,
+        )
 
-        removed = False
-        # hashes must be sorted to ensure we always remove .dir files first
-        for hash_ in sorted(
-            tree.all(jobs, str(tree.path_info)),
-            key=tree.is_dir_hash,
-            reverse=True,
-        ):
-            if hash_ in used:
-                continue
-            path_info = tree.hash_to_path_info(hash_)
-            if tree.is_dir_hash(hash_):
-                # backward compatibility
-                # pylint: disable=protected-access
-                tree._remove_unpacked_dir(hash_)
-            tree.remove(path_info)
-            removed = True
+        if not self.tree.verify:
+            with cache.tree.state:
+                for checksum in named_cache.scheme_keys("local"):
+                    cache_file = cache.tree.hash_to_path_info(checksum)
+                    if cache.tree.exists(cache_file):
+                        # We can safely save here, as existing corrupted files
+                        # will be removed upon status, while files corrupted
+                        # during download will not be moved from tmp_file
+                        # (see `BaseTree.download()`)
+                        hash_info = HashInfo(
+                            cache.tree.PARAM_CHECKSUM, checksum
+                        )
+                        cache.tree.state.save(cache_file, hash_info)
 
-        if removed and hasattr(remote, "index"):
-            remote.index.clear()
-        return removed
+        return ret
+
+    @staticmethod
+    def _log_missing_caches(hash_info_dict):
+        missing_caches = [
+            (md5, info)
+            for md5, info in hash_info_dict.items()
+            if info["status"] == STATUS_MISSING
+        ]
+        if missing_caches:
+            missing_desc = "\n".join(
+                "name: {}, md5: {}".format(info["name"], md5)
+                for md5, info in missing_caches
+            )
+            msg = (
+                "Some of the cache files do not exist neither locally "
+                "nor on remote. Missing cache files:\n{}".format(missing_desc)
+            )
+            logger.warning(msg)

dvc/remote/index.py

@@ -6,6 +6,7 @@ import threading
 from funcy import lchunks
 
 from dvc.state import _connect_sqlite
+from dvc.utils.fs import makedirs
 
 logger = logging.getLogger(__name__)
 
@@ -67,9 +68,12 @@ class RemoteIndex:
     VERSION = 1
     INDEX_TABLE = "remote_index"
     INDEX_TABLE_LAYOUT = "checksum TEXT PRIMARY KEY, dir INTEGER NOT NULL"
+    INDEX_DIR = "index"
 
     def __init__(self, repo, name, dir_suffix=".dir"):
-        self.path = os.path.join(repo.index_dir, f"{name}{self.INDEX_SUFFIX}")
+        self.path = os.path.join(
+            repo.tmp_dir, self.INDEX_DIR, f"{name}{self.INDEX_SUFFIX}"
+        )
 
         self.dir_suffix = dir_suffix
         self.database = None
@@ -150,6 +154,7 @@ class RemoteIndex:
             assert self.cursor is None
 
             empty = not os.path.isfile(self.path)
+            makedirs(os.path.dirname(self.path), exist_ok=True)
             self.database = _connect_sqlite(self.path, {"nolock": 1})
             self.cursor = self.database.cursor()
 

dvc/remote/local.py

@@ -3,4 +3,4 @@ from .index import RemoteIndexNoop
 
 
 class LocalRemote(Remote):
-    INDEX_CLS = RemoteIndexNoop
+    INDEX_CLS = RemoteIndexNoop  # type: ignore[assignment]

dvc/remote/slow_link_detection.py

@@ -1,3 +1,4 @@
+# type: ignore
 import logging
 import sys
 import time

dvc/repo/__init__.py

@@ -2,8 +2,10 @@ import logging
 import os
 from contextlib import contextmanager
 from functools import wraps
+from typing import TYPE_CHECKING
 
-from funcy import cached_property, cat, first
+from funcy import cached_property, cat
+from git import InvalidGitRepositoryError
 
 from dvc.config import Config
 from dvc.dvcfile import PIPELINE_FILE, Dvcfile, is_valid_filename
@@ -15,26 +17,47 @@ from dvc.exceptions import (
     OutputNotFoundError,
 )
 from dvc.path_info import PathInfo
-from dvc.repo.tree import RepoTree
+from dvc.scm import Base
+from dvc.scm.base import SCMError
+from dvc.tree.repo import RepoTree
 from dvc.utils.fs import path_isin
 
 from ..stage.exceptions import StageFileDoesNotExistError, StageNotFound
 from ..utils import parse_target
-from .graph import check_acyclic, get_pipeline, get_pipelines
+from .graph import build_graph, build_outs_graph, get_pipeline, get_pipelines
+from .trie import build_outs_trie
+
+if TYPE_CHECKING:
+    from dvc.tree.base import BaseTree
+
 
 logger = logging.getLogger(__name__)
 
 
+@contextmanager
+def lock_repo(repo: "Repo"):
+    # pylint: disable=protected-access
+    depth = repo._lock_depth
+    repo._lock_depth += 1
+
+    try:
+        if depth > 0:
+            yield
+        else:
+            with repo.lock, repo.state:
+                repo._reset()
+                yield
+                # Graph cache is no longer valid after we release the repo.lock
+                repo._reset()
+    finally:
+        repo._lock_depth = depth
+
+
 def locked(f):
     @wraps(f)
     def wrapper(repo, *args, **kwargs):
-        with repo.lock, repo.state:
-            # pylint: disable=protected-access
-            repo._reset()
-            ret = f(repo, *args, **kwargs)
-            # Our graph cache is no longer valid after we release the repo.lock
-            repo._reset()
-            return ret
+        with lock_repo(repo):
+            return f(repo, *args, **kwargs)
 
     return wrapper
 
@@ -42,109 +65,149 @@ def locked(f):
 class Repo:
     DVC_DIR = ".dvc"
 
-    from dvc.repo.destroy import destroy
-    from dvc.repo.install import install
     from dvc.repo.add import add
-    from dvc.repo.remove import remove
-    from dvc.repo.ls import ls
+    from dvc.repo.brancher import brancher
+    from dvc.repo.checkout import checkout
+    from dvc.repo.commit import commit
+    from dvc.repo.destroy import destroy
+    from dvc.repo.diff import diff
+    from dvc.repo.fetch import fetch
     from dvc.repo.freeze import freeze, unfreeze
-    from dvc.repo.move import move
-    from dvc.repo.run import run
+    from dvc.repo.gc import gc
+    from dvc.repo.get import get as _get
+    from dvc.repo.get_url import get_url as _get_url
     from dvc.repo.imp import imp
     from dvc.repo.imp_url import imp_url
-    from dvc.repo.reproduce import reproduce
-    from dvc.repo.checkout import _checkout
-    from dvc.repo.push import push
-    from dvc.repo.fetch import _fetch
+    from dvc.repo.install import install
+    from dvc.repo.ls import ls as _ls
+    from dvc.repo.move import move
     from dvc.repo.pull import pull
+    from dvc.repo.push import push
+    from dvc.repo.remove import remove
+    from dvc.repo.reproduce import reproduce
+    from dvc.repo.run import run
     from dvc.repo.status import status
-    from dvc.repo.gc import gc
-    from dvc.repo.commit import commit
-    from dvc.repo.diff import diff
-    from dvc.repo.brancher import brancher
-    from dvc.repo.get import get
-    from dvc.repo.get_url import get_url
     from dvc.repo.update import update
 
-    def __init__(self, root_dir=None, scm=None, rev=None):
-        from dvc.state import State, StateNoop
-        from dvc.lock import make_lock
+    ls = staticmethod(_ls)
+    get = staticmethod(_get)
+    get_url = staticmethod(_get_url)
+
+    def _get_repo_dirs(
+        self,
+        root_dir: str = None,
+        scm: Base = None,
+        rev: str = None,
+        uninitialized: bool = False,
+    ):
+        assert bool(scm) == bool(rev)
+
         from dvc.scm import SCM
+        from dvc.scm.git import Git
+        from dvc.utils.fs import makedirs
+
+        dvc_dir = None
+        tmp_dir = None
+        try:
+            tree = scm.get_tree(rev) if isinstance(scm, Git) and rev else None
+            root_dir = self.find_root(root_dir, tree)
+            dvc_dir = os.path.join(root_dir, self.DVC_DIR)
+            tmp_dir = os.path.join(dvc_dir, "tmp")
+            makedirs(tmp_dir, exist_ok=True)
+        except NotDvcRepoError:
+            if not uninitialized:
+                raise
+
+            try:
+                scm = SCM(root_dir or os.curdir)
+            except (SCMError, InvalidGitRepositoryError):
+                scm = SCM(os.curdir, no_scm=True)
+
+            assert isinstance(scm, Base)
+            root_dir = scm.root_dir
+
+        return root_dir, dvc_dir, tmp_dir
+
+    def __init__(
+        self,
+        root_dir=None,
+        scm=None,
+        rev=None,
+        subrepos=False,
+        uninitialized=False,
+    ):
         from dvc.cache import Cache
         from dvc.data_cloud import DataCloud
+        from dvc.lock import LockNoop, make_lock
         from dvc.repo.experiments import Experiments
         from dvc.repo.metrics import Metrics
-        from dvc.repo.plots import Plots
         from dvc.repo.params import Params
-        from dvc.tree.local import LocalTree
-        from dvc.utils.fs import makedirs
+        from dvc.repo.plots import Plots
         from dvc.stage.cache import StageCache
+        from dvc.state import State, StateNoop
+        from dvc.tree.local import LocalTree
 
-        if scm:
-            tree = scm.get_tree(rev)
-            self.root_dir = self.find_root(root_dir, tree)
-            self.scm = scm
-            self.tree = scm.get_tree(
-                rev, use_dvcignore=True, dvcignore_root=self.root_dir
-            )
-            self.state = StateNoop()
-        else:
-            root_dir = self.find_root(root_dir)
-            self.root_dir = os.path.abspath(os.path.realpath(root_dir))
-            self.tree = LocalTree(
-                self,
-                {"url": self.root_dir},
-                use_dvcignore=True,
-                dvcignore_root=self.root_dir,
-            )
-
-        self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
-        self.config = Config(self.dvc_dir, tree=self.tree)
-
-        if not scm:
-            no_scm = self.config["core"].get("no_scm", False)
-            self.scm = SCM(self.root_dir, no_scm=no_scm)
-
-        self.tmp_dir = os.path.join(self.dvc_dir, "tmp")
-        self.index_dir = os.path.join(self.tmp_dir, "index")
-        makedirs(self.index_dir, exist_ok=True)
-
-        hardlink_lock = self.config["core"].get("hardlink_lock", False)
-        self.lock = make_lock(
-            os.path.join(self.tmp_dir, "lock"),
-            tmp_dir=self.tmp_dir,
-            hardlink_lock=hardlink_lock,
-            friendly=True,
+        self.root_dir, self.dvc_dir, self.tmp_dir = self._get_repo_dirs(
+            root_dir=root_dir, scm=scm, rev=rev, uninitialized=uninitialized
         )
 
+        tree_kwargs = {"use_dvcignore": True, "dvcignore_root": self.root_dir}
+        if scm:
+            self.tree = scm.get_tree(rev, **tree_kwargs)
+        else:
+            self.tree = LocalTree(self, {"url": self.root_dir}, **tree_kwargs)
+
+        self.config = Config(self.dvc_dir, tree=self.tree)
+        self._scm = scm
+
+        # used by RepoTree to determine if it should traverse subrepos
+        self.subrepos = subrepos
+
         self.cache = Cache(self)
         self.cloud = DataCloud(self)
 
-        if not scm:
+        if scm or not self.dvc_dir:
+            self.lock = LockNoop()
+            self.state = StateNoop()
+        else:
+            self.lock = make_lock(
+                os.path.join(self.tmp_dir, "lock"),
+                tmp_dir=self.tmp_dir,
+                hardlink_lock=self.config["core"].get("hardlink_lock", False),
+                friendly=True,
+            )
+
             # NOTE: storing state and link_state in the repository itself to
             # avoid any possible state corruption in 'shared cache dir'
             # scenario.
-            self.state = State(self.cache.local)
+            self.state = State(self)
+            self.stage_cache = StageCache(self)
 
-        self.stage_cache = StageCache(self)
+            try:
+                self.experiments = Experiments(self)
+            except NotImplementedError:
+                self.experiments = None
+
+            self._ignore()
 
         self.metrics = Metrics(self)
         self.plots = Plots(self)
         self.params = Params(self)
+        self._lock_depth = 0
 
-        try:
-            self.experiments = Experiments(self)
-        except NotImplementedError:
-            self.experiments = None
+    @cached_property
+    def scm(self):
+        from dvc.scm import SCM
 
-        self._ignore()
+        no_scm = self.config["core"].get("no_scm", False)
+        return self._scm if self._scm else SCM(self.root_dir, no_scm=no_scm)
 
     @property
-    def tree(self):
+    def tree(self) -> "BaseTree":
         return self._tree
 
     @tree.setter
-    def tree(self, tree):
+    def tree(self, tree: "BaseTree"):
         self._tree = tree
         # Our graph cache is no longer valid, as it was based on the previous
         # tree.
@@ -154,7 +217,7 @@ class Repo:
         return f"{self.__class__.__name__}: '{self.root_dir}'"
 
     @classmethod
-    def find_root(cls, root=None, tree=None):
+    def find_root(cls, root=None, tree=None) -> str:
         root_dir = os.path.realpath(root or os.curdir)
 
         if tree:
@@ -241,7 +304,7 @@ class Repo:
         #
         # [1] https://github.com/iterative/dvc/issues/2671
         if not getattr(self, "_skip_graph_checks", False):
-            self._collect_graph(self.stages + new_stages)
+            build_graph(self.stages + new_stages)
 
     def _collect_inside(self, path, graph):
         import networkx as nx
@@ -400,113 +463,17 @@ class Repo:
 
         return cache
 
-    def _collect_graph(self, stages):
-        """Generate a graph by using the given stages on the given directory
-
-        The nodes of the graph are the stage's path relative to the root.
-
-        Edges are created when the output of one stage is used as a
-        dependency in other stage.
-
-        The direction of the edges goes from the stage to its dependency:
-
-        For example, running the following:
-
-            $ dvc run -o A "echo A > A"
-            $ dvc run -d A -o B "echo B > B"
-            $ dvc run -d B -o C "echo C > C"
-
-        Will create the following graph:
-
-               ancestors <--
-                           |
-                C.dvc -> B.dvc -> A.dvc
-                |          |
-                |          --> descendants
-                |
-                ------- pipeline ------>
-                           |
-                           v
-              (weakly connected components)
-
-        Args:
-            stages (list): used to build a graph, if None given, collect stages
-                in the repository.
-
-        Raises:
-            OutputDuplicationError: two outputs with the same path
-            StagePathAsOutputError: stage inside an output directory
-            OverlappingOutputPathsError: output inside output directory
-            CyclicGraphError: resulting graph has cycles
-        """
-        import networkx as nx
-        from pygtrie import Trie
-        from dvc.exceptions import (
-            OutputDuplicationError,
-            StagePathAsOutputError,
-            OverlappingOutputPathsError,
-        )
-
-        G = nx.DiGraph()
-        stages = stages or self.stages
-        outs = Trie()  # Use trie to efficiently find overlapping outs and deps
-
-        for stage in filter(bool, stages):  # bug? not using it later
-            for out in stage.outs:
-                out_key = out.path_info.parts
-
-                # Check for dup outs
-                if out_key in outs:
-                    dup_stages = [stage, outs[out_key].stage]
-                    raise OutputDuplicationError(str(out), dup_stages)
-
-                # Check for overlapping outs
-                if outs.has_subtrie(out_key):
-                    parent = out
-                    overlapping = first(outs.values(prefix=out_key))
-                else:
-                    parent = outs.shortest_prefix(out_key).value
-                    overlapping = out
-                if parent and overlapping:
-                    msg = (
-                        "Paths for outs:\n'{}'('{}')\n'{}'('{}')\n"
-                        "overlap. To avoid unpredictable behaviour, "
-                        "rerun command with non overlapping outs paths."
-                    ).format(
-                        str(parent),
-                        parent.stage.addressing,
-                        str(overlapping),
-                        overlapping.stage.addressing,
-                    )
-                    raise OverlappingOutputPathsError(parent, overlapping, msg)
-
-                outs[out_key] = out
-
-        for stage in stages:
-            out = outs.shortest_prefix(PathInfo(stage.path).parts).value
-            if out:
-                raise StagePathAsOutputError(stage, str(out))
-
-        # Building graph
-        G.add_nodes_from(stages)
-        for stage in stages:
-            for dep in stage.deps:
-                if dep.path_info is None:
-                    continue
-
-                dep_key = dep.path_info.parts
-                overlapping = [n.value for n in outs.prefixes(dep_key)]
-                if outs.has_subtrie(dep_key):
-                    overlapping.extend(outs.values(prefix=dep_key))
-
-                G.add_edges_from((stage, out.stage) for out in overlapping)
-        check_acyclic(G)
-
-        return G
+    @cached_property
+    def outs_trie(self):
+        return build_outs_trie(self.stages)
 
     @cached_property
     def graph(self):
-        return self._collect_graph(self.stages)
+        return build_graph(self.stages, self.outs_trie)
+
+    @cached_property
+    def outs_graph(self):
+        return build_outs_graph(self.graph, self.outs_trie)
 
     @cached_property
     def pipelines(self):
@@ -525,12 +492,6 @@ class Repo:
         """
         return self._collect_stages()
 
-    @cached_property
-    def plot_templates(self):
-        from .plots.template import PlotTemplates
-
-        return PlotTemplates(self.dvc_dir)
-
     def _collect_stages(self):
         stages = []
         outs = set()
@@ -580,19 +541,20 @@ class Repo:
         path_parts = os.path.normpath(path).split(os.path.sep)
         return self.DVC_DIR in path_parts
 
+    @cached_property
+    def repo_tree(self):
+        return RepoTree(self, subrepos=self.subrepos, fetch=True)
+
     @contextmanager
     def open_by_relpath(self, path, remote=None, mode="r", encoding=None):
         """Opens a specified resource as a file descriptor"""
 
-        tree = RepoTree(self, stream=True)
-        path = os.path.join(self.root_dir, path)
+        tree = RepoTree(self, stream=True, subrepos=True)
+        path = PathInfo(self.root_dir) / path
         try:
             with self.state:
                 with tree.open(
-                    os.path.join(self.root_dir, path),
-                    mode=mode,
-                    encoding=encoding,
-                    remote=remote,
+                    path, mode=mode, encoding=encoding, remote=remote,
                 ) as fobj:
                     yield fobj
         except FileNotFoundError as exc:
@@ -603,16 +565,9 @@ class Repo:
     def close(self):
         self.scm.close()
 
-    @locked
-    def checkout(self, *args, **kwargs):
-        return self._checkout(*args, **kwargs)
-
-    @locked
-    def fetch(self, *args, **kwargs):
-        return self._fetch(*args, **kwargs)
-
     def _reset(self):
+        self.__dict__.pop("outs_trie", None)
+        self.__dict__.pop("outs_graph", None)
         self.__dict__.pop("graph", None)
         self.__dict__.pop("stages", None)
         self.__dict__.pop("pipelines", None)
-        self.tree.__dict__.pop("dvcignore", None)

dvc/repo/add.py

@@ -6,6 +6,7 @@ import colorama
 from dvc.dvcfile import Dvcfile, is_dvc_file
 
 from ..exceptions import (
+    CacheLinkError,
     OutputDuplicationError,
     OverlappingOutputPathsError,
     RecursiveAddingWhileUsingFilename,
@@ -22,7 +23,7 @@ logger = logging.getLogger(__name__)
 @locked
 @scm_context
 def add(
-    repo, targets, recursive=False, no_commit=False, fname=None, external=False
+    repo, targets, recursive=False, no_commit=False, fname=None, **kwargs,
 ):
     if recursive and fname:
         raise RecursiveAddingWhileUsingFilename()
@@ -30,6 +31,7 @@ def add(
     if isinstance(targets, str):
         targets = [targets]
 
+    link_failures = []
     stages_list = []
     num_targets = len(targets)
     with Tqdm(total=num_targets, desc="Add", unit="file", leave=True) as pbar:
@@ -55,7 +57,7 @@ def add(
                 )
 
             stages = _create_stages(
-                repo, sub_targets, fname, pbar=pbar, external=external
+                repo, sub_targets, fname, pbar=pbar, **kwargs,
             )
 
             try:
@@ -79,33 +81,56 @@ def add(
                     exc.output, set(exc.stages) - set(stages)
                 )
 
-            with Tqdm(
-                total=len(stages),
-                desc="Processing",
-                unit="file",
-                disable=len(stages) == 1,
-            ) as pbar_stages:
-                for stage in stages:
-                    try:
-                        stage.save()
-                    except OutputDoesNotExistError:
-                        pbar.n -= 1
-                        raise
-
-                    if not no_commit:
-                        stage.commit()
-
-                    Dvcfile(repo, stage.path).dump(stage)
-                    pbar_stages.update()
-
+            link_failures.extend(
+                _process_stages(repo, stages, no_commit, pbar)
+            )
             stages_list += stages
 
         if num_targets == 1:  # restore bar format for stats
             pbar.bar_format = pbar.BAR_FMT_DEFAULT
 
+    if link_failures:
+        msg = (
+            "Some targets could not be linked from cache to workspace.\n{}\n"
+            "To re-link these targets, reconfigure cache types and then run:\n"
+            "\n\tdvc checkout {}"
+        ).format(
+            CacheLinkError.SUPPORT_LINK,
+            " ".join([str(stage.relpath) for stage in link_failures]),
+        )
+        logger.warning(msg)
+
     return stages_list
 
 
+def _process_stages(repo, stages, no_commit, pbar):
+    link_failures = []
+
+    with Tqdm(
+        total=len(stages),
+        desc="Processing",
+        unit="file",
+        disable=len(stages) == 1,
+    ) as pbar_stages:
+        for stage in stages:
+            try:
+                stage.save()
+            except OutputDoesNotExistError:
+                pbar.n -= 1
+                raise
+
+            try:
+                if not no_commit:
+                    stage.commit()
+            except CacheLinkError:
+                link_failures.append(stage)
+
+            Dvcfile(repo, stage.path).dump(stage)
+            pbar_stages.update()
+
+    return link_failures
+
+
 def _find_all_targets(repo, target, recursive):
     if os.path.isdir(target) and recursive:
         return [
@@ -124,15 +149,27 @@ def _find_all_targets(repo, target, recursive):
     return [target]
 
 
-def _create_stages(repo, targets, fname, pbar=None, external=False):
+def _create_stages(
+    repo, targets, fname, pbar=None, external=False, glob=False, desc=None,
+):
+    from glob import iglob
+
     from dvc.stage import Stage, create_stage
 
-    stages = []
+    if glob:
+        expanded_targets = [
+            exp_target
+            for target in targets
+            for exp_target in iglob(target, recursive=True)
+        ]
+    else:
+        expanded_targets = targets
 
+    stages = []
     for out in Tqdm(
-        targets,
+        expanded_targets,
         desc="Creating DVC-files",
-        disable=len(targets) < LARGE_DIR_SIZE,
+        disable=len(expanded_targets) < LARGE_DIR_SIZE,
         unit="file",
     ):
         path, wdir, out = resolve_paths(repo, out)
@@ -146,6 +183,8 @@ def _create_stages(repo, targets, fname, pbar=None, external=False):
         )
         if stage:
             Dvcfile(repo, stage.path).remove()
+            if desc:
+                stage.outs[0].desc = desc
 
         repo._reset()  # pylint: disable=protected-access
 

dvc/repo/brancher.py

@@ -4,7 +4,12 @@ from dvc.tree.local import LocalTree
 
 
 def brancher(  # noqa: E302
-    self, revs=None, all_branches=False, all_tags=False, all_commits=False
+    self,
+    revs=None,
+    all_branches=False,
+    all_tags=False,
+    all_commits=False,
+    sha_only=False,
 ):
     """Generator that iterates over specified revisions.
 
@@ -13,6 +18,7 @@ def brancher(  # noqa: E302
         all_branches (bool): iterate over all available branches.
         all_commits (bool): iterate over all commits.
         all_tags (bool): iterate over all available tags.
+        sha_only (bool): only return git SHA for a revision.
 
     Yields:
         str: the display name for the currently selected tree, it could be:
@@ -53,6 +59,9 @@ def brancher(  # noqa: E302
                 # ignore revs that don't contain repo root
                 # (i.e. revs from before a subdir=True repo was init'ed)
                 if self.tree.exists(self.root_dir):
-                    yield ", ".join(names)
+                    if sha_only:
+                        yield sha
+                    else:
+                        yield ", ".join(names)
     finally:
         self.tree = saved_tree

dvc/repo/checkout.py

@@ -9,6 +9,8 @@ from dvc.exceptions import (
 from dvc.progress import Tqdm
 from dvc.utils import relpath
 
+from . import locked
+
 logger = logging.getLogger(__name__)
 
 
@@ -36,13 +38,16 @@ def get_all_files_numbers(pairs):
     )
 
 
-def _checkout(
+@locked
+def checkout(
     self,
     targets=None,
     with_deps=False,
     force=False,
     relink=False,
     recursive=False,
+    allow_missing=False,
+    **kwargs,
 ):
     from dvc.stage.exceptions import (
         StageFileBadNameError,
@@ -93,6 +98,8 @@ def _checkout(
                 progress_callback=pbar.update_msg,
                 relink=relink,
                 filter_info=filter_info,
+                allow_missing=allow_missing,
+                **kwargs,
             )
             for key, items in result.items():
                 stats[key].extend(_fspath_dir(path) for path in items)

dvc/repo/collect.py

@@ -0,0 +1,87 @@
+import logging
+import os
+from typing import Callable, Iterable, Set, Tuple
+
+from dvc.output.base import BaseOutput
+from dvc.path_info import PathInfo
+from dvc.repo import Repo
+from dvc.tree.repo import RepoTree
+from dvc.types import DvcPath
+
+logger = logging.getLogger(__name__)
+
+
+FilterFn = Callable[[BaseOutput], bool]
+Outputs = Set[BaseOutput]
+DvcPaths = Set[DvcPath]
+
+
+def _collect_outs(
+    repo: Repo, output_filter: FilterFn = None, deps: bool = False,
+) -> Outputs:
+    outs = {
+        out
+        for stage in repo.stages
+        for out in (stage.deps if deps else stage.outs)
+    }
+    return set(filter(output_filter, outs)) if output_filter else outs
+
+
+def _collect_paths(
+    repo: Repo,
+    targets: Iterable[str],
+    recursive: bool = False,
+    rev: str = None,
+):
+    path_infos = {PathInfo(os.path.abspath(target)) for target in targets}
+    tree = RepoTree(repo)
+
+    target_infos = set()
+    for path_info in path_infos:
+
+        if recursive and tree.isdir(path_info):
+            target_infos.update(set(tree.walk_files(path_info)))
+
+        if not tree.isfile(path_info):
+            if not recursive:
+                logger.warning(
+                    "'%s' was not found at: '%s'.", path_info, rev,
+                )
+            continue
+        target_infos.add(path_info)
+    return target_infos
+
+
+def _filter_duplicates(
+    outs: Outputs, path_infos: DvcPaths
+) -> Tuple[Outputs, DvcPaths]:
+    res_outs: Outputs = set()
+    res_infos = set(path_infos)
+
+    for out in outs:
+        if out.path_info in path_infos:
+            res_outs.add(out)
+            res_infos.remove(out.path_info)
+
+    return res_outs, res_infos
+
+
+def collect(
+    repo: Repo,
+    deps: bool = False,
+    targets: Iterable[str] = None,
+    output_filter: FilterFn = None,
+    rev: str = None,
+    recursive: bool = False,
+) -> Tuple[Outputs, DvcPaths]:
+    assert targets or output_filter
+
+    outs: Outputs = _collect_outs(repo, output_filter=output_filter, deps=deps)
+
+    if not targets:
+        path_infos: DvcPaths = set()
+        return outs, path_infos
+
+    target_infos = _collect_paths(repo, targets, recursive=recursive, rev=rev)
+
+    return _filter_duplicates(outs, target_infos)

dvc/repo/commit.py

@@ -45,5 +45,5 @@ def commit(self, target, with_deps=False, recursive=False, force=False):
             stage.save()
         stage.commit()
 
-        Dvcfile(self, stage.path).dump(stage)
+        Dvcfile(self, stage.path).dump(stage, update_pipeline=False)
     return stages

dvc/repo/diff.py

@@ -1,7 +1,11 @@
+import logging
 import os
 
 from dvc.repo import locked
 from dvc.tree.local import LocalTree
+from dvc.tree.repo import RepoTree
+
+logger = logging.getLogger(__name__)
 
 
 @locked
@@ -14,64 +18,32 @@ def diff(self, a_rev="HEAD", b_rev=None):
     `dvc diff` would be the same as `dvc diff HEAD`.
     """
 
-    def _paths_checksums():
-        """
-        A dictionary of checksums addressed by relpaths collected from
-        the current tree outputs.
-
-        To help distinguish between a directory and a file output,
-        the former one will come with a trailing slash in the path:
-
-            directory: "data/"
-            file:      "data"
-        """
-
-        def _to_path(output):
-            return (
-                str(output)
-                if not output.is_dir_checksum
-                else os.path.join(str(output), "")
-            )
-
-        on_working_tree = isinstance(self.tree, LocalTree)
-
-        def _to_checksum(output):
-            if on_working_tree:
-                return self.cache.local.tree.get_hash(output.path_info)
-            return output.checksum
-
-        def _exists(output):
-            if on_working_tree:
-                return output.exists
-            return True
-
-        return {
-            _to_path(output): _to_checksum(output)
-            for stage in self.stages
-            for output in stage.outs
-            if _exists(output)
-        }
-
     if self.scm.no_commits:
         return {}
 
-    working_tree = self.tree
-    a_tree = self.scm.get_tree(a_rev)
-    b_tree = self.scm.get_tree(b_rev) if b_rev else working_tree
+    b_rev = b_rev if b_rev else "workspace"
+    results = {}
+    for rev in self.brancher(revs=[a_rev, b_rev]):
+        if rev == "workspace" and rev != b_rev:
+            # brancher always returns workspace, but we only need to compute
+            # workspace paths/checksums if b_rev was None
+            continue
+        results[rev] = _paths_checksums(self)
 
-    try:
-        self.tree = a_tree
-        old = _paths_checksums()
-
-        self.tree = b_tree
-        new = _paths_checksums()
-    finally:
-        self.tree = working_tree
+    old = results[a_rev]
+    new = results[b_rev]
 
     # Compare paths between the old and new tree.
     # set() efficiently converts dict keys to a set
     added = sorted(set(new) - set(old))
-    deleted = sorted(set(old) - set(new))
+    deleted_or_missing = set(old) - set(new)
+    if b_rev == "workspace":
+        # missing status is only applicable when diffing local workspace
+        # against a commit
+        missing = sorted(_filter_missing(self, deleted_or_missing))
+    else:
+        missing = []
+    deleted = sorted(deleted_or_missing - set(missing))
     modified = sorted(set(old) & set(new))
 
     ret = {
@@ -82,6 +54,73 @@ def diff(self, a_rev="HEAD", b_rev=None):
             for path in modified
             if old[path] != new[path]
         ],
+        "not in cache": [
+            {"path": path, "hash": old[path]} for path in missing
+        ],
     }
 
     return ret if any(ret.values()) else {}
+
+
+def _paths_checksums(repo):
+    """
+    A dictionary of checksums addressed by relpaths collected from
+    the current tree outputs.
+
+    To help distinguish between a directory and a file output,
+    the former one will come with a trailing slash in the path:
+
+        directory: "data/"
+        file:      "data"
+    """
+
+    return dict(_output_paths(repo))
+
+
+def _output_paths(repo):
+    repo_tree = RepoTree(repo, stream=True)
+    on_working_tree = isinstance(repo.tree, LocalTree)
+
+    def _exists(output):
+        if on_working_tree:
+            return output.exists
+        return True
+
+    def _to_path(output):
+        return (
+            str(output)
+            if not output.is_dir_checksum
+            else os.path.join(str(output), "")
+        )
+
+    def _to_checksum(output):
+        if on_working_tree:
+            return repo.cache.local.tree.get_hash(output.path_info).value
+        return output.hash_info.value
+
+    for stage in repo.stages:
+        for output in stage.outs:
+            if _exists(output):
+                yield _to_path(output), _to_checksum(output)
+                if output.is_dir_checksum:
+                    yield from _dir_output_paths(repo_tree, output)
+
+
+def _dir_output_paths(repo_tree, output):
+    from dvc.config import NoRemoteError
+
+    try:
+        for fname in repo_tree.walk_files(output.path_info):
+            yield str(fname), repo_tree.get_file_hash(fname).value
+    except NoRemoteError:
+        logger.warning("dir cache entry for '%s' is missing", output)
+
+
+def _filter_missing(repo, paths):
+    repo_tree = RepoTree(repo, stream=True)
+    for path in paths:
+        metadata = repo_tree.metadata(path)
+        if metadata.is_dvc:
+            out = metadata.outs[0]
+            if out.status().get(str(out)) == "not in cache":
+                yield path

dvc/repo/experiments/executor.py

@@ -0,0 +1,207 @@
+import logging
+import os
+import pickle
+from tempfile import TemporaryDirectory
+from typing import Iterable, Optional
+
+from funcy import cached_property
+
+from dvc.path_info import PathInfo
+from dvc.stage import PipelineStage
+from dvc.tree.base import BaseTree
+from dvc.tree.local import LocalTree
+from dvc.tree.repo import RepoTree
+
+logger = logging.getLogger(__name__)
+
+
+class ExperimentExecutor:
+    """Base class for executing experiments in parallel.
+
+    Args:
+        baseline_rev: Baseline revision that this experiment is derived from.
+
+    Optional keyword args:
+        branch: Existing git branch for this experiment.
+        rev: Git revision to be checked out for this experiment, defaults to
+            branch, baseline_rev in that order.
+        repro_args: Args to be passed into reproduce.
+        repro_kwargs: Keyword args to be passed into reproduce.
+    """
+
+    def __init__(
+        self,
+        baseline_rev: str,
+        branch: Optional[str] = None,
+        rev: Optional[str] = None,
+        **kwargs,
+    ):
+        self.baseline_rev = baseline_rev
+        self._rev = rev
+        self.branch = branch
+        self.repro_args = kwargs.pop("repro_args", [])
+        self.repro_kwargs = kwargs.pop("repro_kwargs", {})
+        self.tmp_dir: Optional[TemporaryDirectory] = None
+
+    @property
+    def tree(self) -> BaseTree:
+        raise NotImplementedError
+
+    @staticmethod
+    def reproduce(dvc_dir, cwd=None, **kwargs):
+        raise NotImplementedError
+
+    def collect_output(self) -> Iterable["PathInfo"]:
+        """Iterate over output pathnames for this executor.
+
+        For DVC outs, only the .dvc file path will be yielded. DVC outs
+        themselves should be fetched from remote executor cache in the normal
+        fetch/pull way. For local executors outs will already be available via
+        shared local cache.
+        """
+        raise NotImplementedError
+
+    def cleanup(self):
+        pass
+
+    # TODO: come up with better way to stash repro arguments
+    @staticmethod
+    def pack_repro_args(path, *args, tree=None, **kwargs):
+        dpath = os.path.dirname(path)
+        if tree:
+            open_func = tree.open
+            tree.makedirs(dpath)
+        else:
+            from dvc.utils.fs import makedirs
+
+            open_func = open
+            makedirs(dpath, exist_ok=True)
+        data = {"args": args, "kwargs": kwargs}
+        with open_func(path, "wb") as fobj:
+            pickle.dump(data, fobj)
+
+    @staticmethod
+    def unpack_repro_args(path, tree=None):
+        open_func = tree.open if tree else open
+        with open_func(path, "rb") as fobj:
+            data = pickle.load(fobj)
+        return data["args"], data["kwargs"]
+
+    @staticmethod
+    def collect_files(tree: BaseTree, repo_tree: RepoTree):
+        raise NotImplementedError
+
+
+class LocalExecutor(ExperimentExecutor):
+    """Local machine experiment executor."""
+
+    def __init__(
+        self, checkpoint_reset: Optional[bool] = False, **kwargs,
+    ):
+        from dvc.repo import Repo
+
+        dvc_dir = kwargs.pop("dvc_dir")
+        cache_dir = kwargs.pop("cache_dir")
+        super().__init__(**kwargs)
+        self.tmp_dir = TemporaryDirectory()
+
+        # init empty DVC repo (will be overwritten when input is uploaded)
+        Repo.init(root_dir=self.tmp_dir.name, no_scm=True)
+        logger.debug(
+            "Init local executor in dir '%s' with baseline '%s'.",
+            self.tmp_dir,
+            self.baseline_rev[:7],
+        )
+        self.dvc_dir = os.path.join(self.tmp_dir.name, dvc_dir)
+        self._config(cache_dir)
+        self._tree = LocalTree(self.dvc, {"url": self.dvc.root_dir})
+        # override default CACHE_MODE since files must be writable in order
+        # to run repro
+        self._tree.CACHE_MODE = 0o644
+        self.checkpoint_reset = checkpoint_reset
+        self.checkpoint = False
+
+    def _config(self, cache_dir):
+        local_config = os.path.join(self.dvc_dir, "config.local")
+        logger.debug("Writing experiments local config '%s'", local_config)
+        with open(local_config, "w") as fobj:
+            fobj.write("[core]\n    no_scm = true\n")
+            fobj.write(f"[cache]\n    dir = {cache_dir}")
+
+    @cached_property
+    def dvc(self):
+        from dvc.repo import Repo
+
+        return Repo(self.dvc_dir)
+
+    @cached_property
+    def path_info(self):
+        return PathInfo(self.tmp_dir.name)
+
+    @property
+    def tree(self):
+        return self._tree
+
+    @property
+    def rev(self):
+        return self._rev if self._rev else self.baseline_rev
+
+    @staticmethod
+    def reproduce(dvc_dir, cwd=None, **kwargs):
+        """Run dvc repro and return the result."""
+        from dvc.repo import Repo
+
+        unchanged = []
+
+        def filter_pipeline(stages):
+            unchanged.extend(
+                [stage for stage in stages if isinstance(stage, PipelineStage)]
+            )
+
+        if cwd:
+            old_cwd = os.getcwd()
+            os.chdir(cwd)
+        else:
+            old_cwd = None
+            cwd = os.getcwd()
+
+        try:
+            logger.debug("Running repro in '%s'", cwd)
+            dvc = Repo(dvc_dir)
+
+            # NOTE: for checkpoint experiments we handle persist outs slightly
+            # differently than normal:
+            #
+            # - checkpoint out may not yet exist if this is the first time this
+            #   experiment has been run, this is not an error condition for
+            #   experiments
+            # - at the start of a repro run, we need to remove the persist out
+            #   and restore it to its last known (committed) state (which may
+            #   be removed/does not yet exist) so that our executor workspace
+            #   is not polluted with the (persistent) out from an unrelated
+            #   experiment run
+            dvc.checkout(force=True, quiet=True)
+            stages = dvc.reproduce(on_unchanged=filter_pipeline, **kwargs)
+        finally:
+            if old_cwd is not None:
+                os.chdir(old_cwd)
+
+        # ideally we would return stages here like a normal repro() call, but
+        # stages is not currently picklable and cannot be returned across
+        # multiprocessing calls
+        return stages + unchanged
+
+    def collect_output(self) -> Iterable["PathInfo"]:
+        repo_tree = RepoTree(self.dvc)
+        yield from self.collect_files(self.tree, repo_tree)
+
+    @staticmethod
+    def collect_files(tree: BaseTree, repo_tree: RepoTree):
+        for fname in repo_tree.walk_files(repo_tree.root_dir, dvcfiles=True):
+            if not repo_tree.isdvc(fname):
+                yield tree.path_info / fname.relative_to(repo_tree.root_dir)
+
+    def cleanup(self):
+        logger.debug("Removing tmpdir '%s'", self.tmp_dir)
+        self.tmp_dir.cleanup()
+        super().cleanup()

dvc/repo/experiments/gc.py

@@ -0,0 +1,51 @@
+import logging
+from typing import Optional
+
+from dvc.repo import locked
+
+logger = logging.getLogger(__name__)
+
+
+@locked
+def gc(
+    repo,
+    all_branches: Optional[bool] = False,
+    all_tags: Optional[bool] = False,
+    all_commits: Optional[bool] = False,
+    workspace: Optional[bool] = False,
+    queued: Optional[bool] = False,
+):
+    keep_revs = set(
+        repo.brancher(
+            all_branches=all_branches,
+            all_tags=all_tags,
+            all_commits=all_commits,
+            sha_only=True,
+        )
+    )
+    if workspace:
+        keep_revs.add(repo.scm.get_rev())
+
+    if not keep_revs:
+        return 0
+
+    delete_branches = []
+    for exp_branch in repo.experiments.scm.list_branches():
+        m = repo.experiments.BRANCH_RE.match(exp_branch)
+        if m:
+            rev = repo.scm.resolve_rev(m.group("baseline_rev"))
+            if rev not in keep_revs:
+                delete_branches.append(exp_branch)
+    if delete_branches:
+        repo.experiments.scm.repo.delete_head(*delete_branches, force=True)
+    removed = len(delete_branches)
+
+    delete_stashes = []
+    for _, entry in repo.experiments.stash_revs.items():
+        if not queued or entry.baseline_rev not in keep_revs:
+            delete_stashes.append(entry.index)
+    for index in sorted(delete_stashes, reverse=True):
+        repo.experiments.scm.repo.git.stash("drop", index)
+    removed += len(delete_stashes)
+
+    return removed

dvc/repo/experiments/run.py

@@ -0,0 +1,68 @@
+import logging
+from typing import Iterable, Optional
+
+from dvc.exceptions import InvalidArgumentError
+from dvc.repo import locked
+from dvc.repo.experiments import UnchangedExperimentError
+
+logger = logging.getLogger(__name__)
+
+
+def _parse_params(path_params: Iterable):
+    from ruamel.yaml import YAMLError
+
+    from dvc.dependency.param import ParamsDependency
+    from dvc.utils.serialize import loads_yaml
+
+    ret = {}
+    for path_param in path_params:
+        path, _, params_str = path_param.rpartition(":")
+        # remove empty strings from params, on condition such as `-p "file1:"`
+        params = {}
+        for param_str in filter(bool, params_str.split(",")):
+            try:
+                # interpret value strings using YAML rules
+                key, value = param_str.split("=")
+                params[key] = loads_yaml(value)
+            except (ValueError, YAMLError):
+                raise InvalidArgumentError(
+                    f"Invalid param/value pair '{param_str}'"
+                )
+        if not path:
+            path = ParamsDependency.DEFAULT_PARAMS_FILE
+        ret[path] = params
+    return ret
+
+
+@locked
+def run(
+    repo,
+    target: Optional[str] = None,
+    params: Optional[Iterable] = None,
+    run_all: Optional[bool] = False,
+    jobs: Optional[int] = 1,
+    **kwargs,
+) -> dict:
+    """Reproduce the specified target as an experiment.
+
+    Accepts the same additional kwargs as Repo.reproduce.
+
+    Returns a dict mapping new experiment SHAs to the results
+    of `repro` for that experiment.
+    """
+    if run_all:
+        return repo.experiments.reproduce_queued(jobs=jobs)
+
+    if params:
+        params = _parse_params(params)
+    else:
+        params = []
+    try:
+        return repo.experiments.reproduce_one(
+            target=target, params=params, **kwargs
+        )
+    except UnchangedExperimentError:
+        # If experiment contains no changes, just run regular repro
+        kwargs.pop("queue", None)
+        kwargs.pop("checkpoint_resume", None)
+        return {None: repo.reproduce(target=target, **kwargs)}

dvc/repo/experiments/show.py

@@ -1,6 +1,6 @@
 import logging
-import re
 from collections import OrderedDict, defaultdict
+from datetime import datetime
 
 from dvc.repo import locked
 from dvc.repo.metrics.show import _collect_metrics, _read_metrics
@@ -9,56 +9,119 @@ from dvc.repo.params.show import _collect_configs, _read_params
 logger = logging.getLogger(__name__)
 
 
-EXP_RE = re.compile(r"(?P<rev_sha>[a-f0-9]{7})-(?P<exp_sha>[a-f0-9]+)")
+def _collect_experiment(repo, rev, stash=False, sha_only=True):
+    from git.exc import GitCommandError
 
-
-def _collect_experiment(repo, branch):
     res = defaultdict(dict)
-    for rev in repo.brancher(revs=[branch]):
-        configs = _collect_configs(repo)
+    for rev in repo.brancher(revs=[rev]):
+        if rev == "workspace":
+            res["timestamp"] = None
+        else:
+            commit = repo.scm.resolve_commit(rev)
+            res["timestamp"] = datetime.fromtimestamp(commit.committed_date)
+
+        configs = _collect_configs(repo, rev=rev)
         params = _read_params(repo, configs, rev)
         if params:
             res["params"] = params
 
-        metrics = _collect_metrics(repo, None, False)
-        vals = _read_metrics(repo, metrics, rev)
-        if vals:
+        res["queued"] = stash
+        if not stash:
+            metrics = _collect_metrics(repo, None, rev, False)
+            vals = _read_metrics(repo, metrics, rev)
             res["metrics"] = vals
 
+        if not sha_only and rev != "workspace":
+            try:
+                name = repo.scm.repo.git.describe(
+                    rev, all=True, exact_match=True
+                )
+                name = name.rsplit("/")[-1]
+                res["name"] = name
+            except GitCommandError:
+                pass
+
+    return res
+
+
+def _collect_checkpoint_experiment(res, repo, branch, baseline, **kwargs):
+    exp_rev = repo.scm.resolve_rev(branch)
+    prev = None
+    for rev in repo.scm.branch_revs(exp_rev, baseline):
+        exp = {"checkpoint_tip": exp_rev}
+        if prev:
+            res[prev]["checkpoint_parent"] = rev
+        if rev in res:
+            res[rev].update(exp)
+            res.move_to_end(rev)
+        else:
+            exp.update(_collect_experiment(repo, rev, **kwargs))
+            res[rev] = exp
+        prev = rev
+    res[prev]["checkpoint_parent"] = baseline
     return res
 
 
 @locked
 def show(
-    repo, all_branches=False, all_tags=False, revs=None, all_commits=False
+    repo,
+    all_branches=False,
+    all_tags=False,
+    revs=None,
+    all_commits=False,
+    sha_only=False,
 ):
     res = defaultdict(OrderedDict)
 
     if revs is None:
         revs = [repo.scm.get_rev()]
 
-    revs = set(
-        repo.brancher(
+    revs = OrderedDict(
+        (rev, None)
+        for rev in repo.brancher(
             revs=revs,
             all_branches=all_branches,
             all_tags=all_tags,
             all_commits=all_commits,
+            sha_only=True,
         )
     )
 
     for rev in revs:
-        res[rev]["baseline"] = _collect_experiment(repo, rev)
+        res[rev]["baseline"] = _collect_experiment(
+            repo, rev, sha_only=sha_only
+        )
 
-    for exp_branch in repo.experiments.scm.list_branches():
-        m = re.match(EXP_RE, exp_branch)
+    # collect reproduced experiments
+    for head in sorted(
+        repo.experiments.scm.repo.heads,
+        key=lambda h: h.commit.committed_date,
+        reverse=True,
+    ):
+        exp_branch = head.name
+        m = repo.experiments.BRANCH_RE.match(exp_branch)
         if m:
-            rev = repo.scm.resolve_rev(m.group("rev_sha"))
+            rev = repo.scm.resolve_rev(m.group("baseline_rev"))
             if rev in revs:
-                exp_rev = repo.experiments.scm.resolve_rev(exp_branch)
                 with repo.experiments.chdir():
-                    experiment = _collect_experiment(
-                        repo.experiments.exp_dvc, exp_branch
-                    )
-                res[rev][exp_rev] = experiment
+                    if m.group("checkpoint"):
+                        _collect_checkpoint_experiment(
+                            res[rev], repo.experiments.exp_dvc, exp_branch, rev
+                        )
+                    else:
+                        exp_rev = repo.experiments.scm.resolve_rev(exp_branch)
+                        experiment = _collect_experiment(
+                            repo.experiments.exp_dvc, exp_branch
+                        )
+                        res[rev][exp_rev] = experiment
+
+    # collect queued (not yet reproduced) experiments
+    for stash_rev, entry in repo.experiments.stash_revs.items():
+        if entry.baseline_rev in revs:
+            with repo.experiments.chdir():
+                experiment = _collect_experiment(
+                    repo.experiments.exp_dvc, stash_rev, stash=True
+                )
+            res[entry.baseline_rev][stash_rev] = experiment
 
     return res

dvc/repo/fetch.py

@@ -4,10 +4,13 @@ from dvc.config import NoRemoteError
 from dvc.exceptions import DownloadError
 from dvc.scm.base import CloneError
 
+from . import locked
+
 logger = logging.getLogger(__name__)
 
 
-def _fetch(
+@locked
+def fetch(
     self,
     targets=None,
     jobs=None,
@@ -33,8 +36,6 @@ def _fetch(
             remote is configured
     """
 
-    used_run_cache = self.stage_cache.pull(remote) if run_cache else []
-
     if isinstance(targets, str):
         targets = [targets]
 
@@ -48,13 +49,14 @@ def _fetch(
         remote=remote,
         jobs=jobs,
         recursive=recursive,
-        used_run_cache=used_run_cache,
     )
 
     downloaded = 0
     failed = 0
 
     try:
+        if run_cache:
+            self.stage_cache.pull(remote)
         downloaded += self.cloud.pull(
             used, jobs, remote=remote, show_checksums=show_checksums,
         )
@@ -72,19 +74,21 @@ def _fetch(
     if failed:
         raise DownloadError(failed)
 
-    return downloaded + len(used_run_cache)
+    return downloaded
 
 
 def _fetch_external(self, repo_url, repo_rev, files, jobs):
     from dvc.external_repo import external_repo
 
     failed, downloaded = 0, 0
+    cache = self.cache.local
     try:
-        with external_repo(repo_url, repo_rev) as repo:
-            with repo.use_cache(self.cache.local):
-                d, f, _ = repo.fetch_external(files, jobs=jobs)
-                downloaded += d
-                failed += f
+        with external_repo(
+            repo_url, repo_rev, cache_dir=cache.cache_dir
+        ) as repo:
+            d, f, _ = repo.fetch_external(files, jobs=jobs)
+            downloaded += d
+            failed += f
     except CloneError:
         failed += 1
         logger.exception(

dvc/repo/freeze.py

@@ -8,7 +8,7 @@ def _set(repo, target, frozen):
     path, name = parse_target(target)
     stage = repo.get_stage(path, name)
     stage.frozen = frozen
-    stage.dvcfile.dump(stage, update_pipeline=True)
+    stage.dvcfile.dump(stage, update_lock=False)
 
     return stage
 

dvc/repo/gc.py

@@ -3,19 +3,12 @@ import logging
 from dvc.cache import NamedCache
 from dvc.exceptions import InvalidArgumentError
 
+from ..scheme import Schemes
 from . import locked
 
 logger = logging.getLogger(__name__)
 
 
-def _do_gc(typ, remote, clist, jobs=None):
-    from dvc.remote.base import Remote
-
-    removed = Remote.gc(clist, remote, jobs=jobs)
-    if not removed:
-        logger.info(f"No unused '{typ}' cache to remove.")
-
-
 def _raise_error_if_all_disabled(**kwargs):
     if not any(kwargs.values()):
         raise InvalidArgumentError(
@@ -50,6 +43,7 @@ def gc(
     )
 
     from contextlib import ExitStack
+
     from dvc.repo import Repo
 
     if not repos:
@@ -75,23 +69,18 @@ def gc(
                 )
             )
 
-    # treat caches as remotes for garbage collection
-    _do_gc("local", self.cache.local, used, jobs)
+    for scheme, cache in self.cache.by_scheme():
+        if not cache:
+            continue
 
-    if self.cache.s3:
-        _do_gc("s3", self.cache.s3, used, jobs)
+        removed = cache.gc(set(used.scheme_keys(scheme)), jobs=jobs)
+        if not removed:
+            logger.info(f"No unused '{scheme}' cache to remove.")
 
-    if self.cache.gs:
-        _do_gc("gs", self.cache.gs, used, jobs)
+    if not cloud:
+        return
 
-    if self.cache.ssh:
-        _do_gc("ssh", self.cache.ssh, used, jobs)
-
-    if self.cache.hdfs:
-        _do_gc("hdfs", self.cache.hdfs, used, jobs)
-
-    if self.cache.azure:
-        _do_gc("azure", self.cache.azure, used, jobs)
-
-    if cloud:
-        _do_gc("remote", self.cloud.get_remote(remote, "gc -c"), used, jobs)
+    remote = self.cloud.get_remote(remote, "gc -c")
+    removed = remote.gc(set(used.scheme_keys(Schemes.LOCAL)), jobs=jobs)
+    if not removed:
+        logger.info("No unused cache to remove from remote.")

dvc/repo/get.py

@@ -18,10 +18,9 @@ class GetDVCFileError(DvcException):
         )
 
 
-@staticmethod
 def get(url, path, out=None, rev=None):
-    from dvc.external_repo import external_repo
     from dvc.dvcfile import is_valid_filename
+    from dvc.external_repo import external_repo
 
     out = resolve_output(path, out)
 
@@ -35,21 +34,21 @@ def get(url, path, out=None, rev=None):
     # and won't work with reflink/hardlink.
     dpath = os.path.dirname(os.path.abspath(out))
     tmp_dir = os.path.join(dpath, "." + str(shortuuid.uuid()))
-    try:
-        with external_repo(url=url, rev=rev) as repo:
-            if hasattr(repo, "cache"):
-                repo.cache.local.cache_dir = tmp_dir
 
-                # Try any links possible to avoid data duplication.
-                #
-                # Not using symlink, because we need to remove cache after we
-                # are done, and to make that work we would have to copy data
-                # over anyway before removing the cache, so we might just copy
-                # it right away.
-                #
-                # Also, we can't use theoretical "move" link type here, because
-                # the same cache file might be used a few times in a directory.
-                repo.cache.local.cache_types = ["reflink", "hardlink", "copy"]
+    # Try any links possible to avoid data duplication.
+    #
+    # Not using symlink, because we need to remove cache after we
+    # are done, and to make that work we would have to copy data
+    # over anyway before removing the cache, so we might just copy
+    # it right away.
+    #
+    # Also, we can't use theoretical "move" link type here, because
+    # the same cache file might be used a few times in a directory.
+    cache_types = ["reflink", "hardlink", "copy"]
+    try:
+        with external_repo(
+            url=url, rev=rev, cache_dir=tmp_dir, cache_types=cache_types
+        ) as repo:
             repo.get_external(path, out)
     finally:
         remove(tmp_dir)

dvc/repo/get_url.py

@@ -5,7 +5,6 @@ import dvc.output as output
 from dvc.utils import resolve_output
 
 
-@staticmethod
 def get_url(url, out=None):
     out = resolve_output(url, out)
 
@@ -16,6 +15,7 @@ def get_url(url, out=None):
 
     (dep,) = dependency.loads_from(None, [url])
     (out,) = output.loads_from(None, [out], use_cache=False)
-    dep.save()
+    if not dep.exists:
+        raise dep.DoesNotExistError(dep)
     dep.download(out)
     out.save()

dvc/repo/graph.py

@@ -1,5 +1,6 @@
 def check_acyclic(graph):
     import networkx as nx
+
     from dvc.exceptions import CyclicGraphError
 
     try:
@@ -25,3 +26,97 @@ def get_pipelines(G):
     import networkx as nx
 
     return [G.subgraph(c).copy() for c in nx.weakly_connected_components(G)]
+
+
+def build_graph(stages, outs_trie=None):
+    """Generate a graph by using the given stages on the given directory
+
+    The nodes of the graph are the stage's path relative to the root.
+
+    Edges are created when the output of one stage is used as a
+    dependency in other stage.
+
+    The direction of the edges goes from the stage to its dependency:
+
+    For example, running the following:
+
+        $ dvc run -o A "echo A > A"
+        $ dvc run -d A -o B "echo B > B"
+        $ dvc run -d B -o C "echo C > C"
+
+    Will create the following graph:
+
+           ancestors <--
+                       |
+            C.dvc -> B.dvc -> A.dvc
+            |          |
+            |          --> descendants
+            |
+            ------- pipeline ------>
+                       |
+                       v
+          (weakly connected components)
+
+    Args:
+        stages (list): used to build a graph from
+
+    Raises:
+        OutputDuplicationError: two outputs with the same path
+        StagePathAsOutputError: stage inside an output directory
+        OverlappingOutputPathsError: output inside output directory
+        CyclicGraphError: resulting graph has cycles
+    """
+    import networkx as nx
+
+    from dvc.exceptions import StagePathAsOutputError
+
+    from ..path_info import PathInfo
+    from .trie import build_outs_trie
+
+    G = nx.DiGraph()
+
+    # Use trie to efficiently find overlapping outs and deps
+    outs_trie = outs_trie or build_outs_trie(stages)
+
+    for stage in stages:
+        out = outs_trie.shortest_prefix(PathInfo(stage.path).parts).value
+        if out:
+            raise StagePathAsOutputError(stage, str(out))
+
+    # Building graph
+    G.add_nodes_from(stages)
+    for stage in stages:
+        for dep in stage.deps:
+            if dep.path_info is None:
+                continue
+
+            dep_key = dep.path_info.parts
+            overlapping = [n.value for n in outs_trie.prefixes(dep_key)]
+            if outs_trie.has_subtrie(dep_key):
+                overlapping.extend(outs_trie.values(prefix=dep_key))
+
+            G.add_edges_from((stage, out.stage) for out in overlapping)
+    check_acyclic(G)
+
+    return G
+
+
+# NOTE: using stage graph instead of just list of stages to make sure that it
+# has already passed all the sanity checks like cycles/overlapping outputs and
+# so on.
+def build_outs_graph(graph, outs_trie):
+    import networkx as nx
+
+    G = nx.DiGraph()
+
+    G.add_nodes_from(outs_trie.values())
+    for stage in graph.nodes():
+        for dep in stage.deps:
+            dep_key = dep.path_info.parts
+            overlapping = [n.value for n in outs_trie.prefixes(dep_key)]
+            if outs_trie.has_subtrie(dep_key):
+                overlapping.extend(outs_trie.values(prefix=dep_key))
+
+            for from_out in stage.outs:
+                G.add_edges_from((from_out, out) for out in overlapping)
+    return G

dvc/repo/imp.py

@@ -1,6 +1,8 @@
-def imp(self, url, path, out=None, rev=None):
+def imp(self, url, path, out=None, fname=None, rev=None, **kwargs):
     erepo = {"url": url}
     if rev is not None:
         erepo["rev"] = rev
 
-    return self.imp_url(path, out=out, erepo=erepo, frozen=True)
+    return self.imp_url(
+        path, out=out, fname=fname, erepo=erepo, frozen=True, **kwargs
+    )

dvc/repo/imp_url.py

@@ -11,7 +11,14 @@ from . import locked
 @locked
 @scm_context
 def imp_url(
-    self, url, out=None, fname=None, erepo=None, frozen=True, no_exec=False
+    self,
+    url,
+    out=None,
+    fname=None,
+    erepo=None,
+    frozen=True,
+    no_exec=False,
+    desc=None,
 ):
     from dvc.dvcfile import Dvcfile
     from dvc.stage import Stage, create_stage
@@ -40,6 +47,9 @@ def imp_url(
     if stage is None:
         return None
 
+    if desc:
+        stage.outs[0].desc = desc
+
     dvcfile = Dvcfile(self, stage.path)
     dvcfile.remove()
 

dvc/repo/init.py

@@ -1,48 +1,18 @@
 import logging
 import os
 
-import colorama
-
-from dvc import analytics
 from dvc.config import Config
 from dvc.exceptions import InitError, InvalidArgumentError
+from dvc.ignore import init as init_dvcignore
 from dvc.repo import Repo
 from dvc.scm import SCM
 from dvc.scm.base import SCMError
-from dvc.utils import boxify, relpath
+from dvc.utils import relpath
 from dvc.utils.fs import remove
 
 logger = logging.getLogger(__name__)
 
 
-def _welcome_message():
-    if analytics.is_enabled():
-        logger.info(
-            boxify(
-                "DVC has enabled anonymous aggregate usage analytics.\n"
-                "Read the analytics documentation (and how to opt-out) here:\n"
-                "{blue}https://dvc.org/doc/user-guide/analytics{nc}".format(
-                    blue=colorama.Fore.BLUE, nc=colorama.Fore.RESET
-                ),
-                border_color="red",
-            )
-        )
-
-    msg = (
-        "{yellow}What's next?{nc}\n"
-        "{yellow}------------{nc}\n"
-        "- Check out the documentation: {blue}https://dvc.org/doc{nc}\n"
-        "- Get help and share ideas: {blue}https://dvc.org/chat{nc}\n"
-        "- Star us on GitHub: {blue}https://github.com/iterative/dvc{nc}"
-    ).format(
-        yellow=colorama.Fore.YELLOW,
-        blue=colorama.Fore.BLUE,
-        nc=colorama.Fore.RESET,
-    )
-
-    logger.info(msg)
-
-
 def init(root_dir=os.curdir, no_scm=False, force=False, subdir=False):
     """
     Creates an empty repo on the given directory -- basically a
@@ -99,14 +69,18 @@ def init(root_dir=os.curdir, no_scm=False, force=False, subdir=False):
         with config.edit() as conf:
             conf["core"]["no_scm"] = True
 
+    dvcignore = init_dvcignore(root_dir)
+
     proj = Repo(root_dir)
 
-    scm.add([config.files["repo"], proj.plot_templates.templates_dir])
+    proj.plots.templates.init()
+
+    scm.add(
+        [config.files["repo"], dvcignore, proj.plots.templates.templates_dir]
+    )
 
     if scm.ignore_file:
         scm.add([os.path.join(dvc_dir, scm.ignore_file)])
         logger.info("\nYou can now commit the changes to git.\n")
 
-    _welcome_message()
-
     return proj

dvc/repo/ls.py

@@ -1,8 +1,9 @@
+from itertools import chain
+
 from dvc.exceptions import PathMissingError
 from dvc.path_info import PathInfo
 
 
-@staticmethod
 def ls(
     url, path=None, rev=None, recursive=None, dvc_only=False,
 ):
@@ -29,12 +30,14 @@ def ls(
     """
     from dvc.external_repo import external_repo
 
-    with external_repo(url, rev) as repo:
+    # use our own RepoTree instance instead of repo.repo_tree since we want to
+    # fetch directory listings, but don't want to fetch file contents.
+    with external_repo(url, rev, fetch=False, stream=True) as repo:
         path_info = PathInfo(repo.root_dir)
         if path:
             path_info /= path
 
-        ret = _ls(repo, path_info, recursive, dvc_only)
+        ret = _ls(repo.repo_tree, path_info, recursive, dvc_only)
 
         if path and not ret:
             raise PathMissingError(path, repo, dvc_only=dvc_only)
@@ -47,58 +50,36 @@ def ls(
         return ret_list
 
 
-def _ls(repo, path_info, recursive=None, dvc_only=False):
-    from dvc.repo.tree import RepoTree
-
+def _ls(tree, path_info, recursive=None, dvc_only=False):
     def onerror(exc):
         raise exc
 
-    # use our own RepoTree instance instead of repo.repo_tree since we want to
-    # fetch directory listings, but don't want to fetch file contents.
-    tree = RepoTree(repo, stream=True)
-
-    ret = {}
+    infos = []
     try:
         for root, dirs, files in tree.walk(
             path_info.fspath, onerror=onerror, dvcfiles=True
         ):
-            for fname in files:
-                info = PathInfo(root) / fname
-                dvc = tree.isdvc(info)
-                if dvc or not dvc_only:
-                    path = str(info.relative_to(path_info))
-                    ret[path] = {
-                        "isout": dvc,
-                        "isdir": False,
-                        "isexec": False if dvc else tree.isexec(info),
-                    }
-
+            entries = chain(files, dirs) if not recursive else files
+            infos.extend(PathInfo(root) / entry for entry in entries)
             if not recursive:
-                for dname in dirs:
-                    info = PathInfo(root) / dname
-                    if not dvc_only or (
-                        tree.dvctree and tree.dvctree.exists(info)
-                    ):
-                        dvc = tree.isdvc(info)
-                        path = str(info.relative_to(path_info))
-                        ret[path] = {
-                            "isout": dvc,
-                            "isdir": True,
-                            "isexec": False if dvc else tree.isexec(info),
-                        }
                 break
     except NotADirectoryError:
-        dvc = tree.isdvc(path_info)
-        if dvc or not dvc_only:
-            return {
-                path_info.name: {
-                    "isout": dvc,
-                    "isdir": False,
-                    "isexec": False if dvc else tree.isexec(path_info),
-                }
-            }
-        return {}
+        infos.append(path_info)
     except FileNotFoundError:
         return {}
 
+    ret = {}
+    for info in infos:
+        metadata = tree.metadata(info)
+        if metadata.output_exists or not dvc_only:
+            path = (
+                path_info.name
+                if path_info == info
+                else str(info.relative_to(path_info))
+            )
+            ret[path] = {
+                "isout": metadata.is_output,
+                "isdir": metadata.isdir,
+                "isexec": metadata.is_exec,
+            }
     return ret