dvc: bump to 2.7.4

* tests: additional callbacks test for remaining filesystems

* requirements: bump pydrive2 to 1.9.4

Co-authored-by: Ruslan Kuprieiev <[email protected]>

.github/workflows/packages.yaml

@@ -5,6 +5,7 @@ on:
     - master
   release:
     types: [released, prereleased]
+  workflow_dispatch:
 jobs:
   deb:
     runs-on: ubuntu-18.04

.github/workflows/tests.yaml

@@ -1,12 +1,8 @@
 name: Tests
 
 on:
-  pull_request:
-    branches:
-      - master
-  push:
-    branches:
-      - master
+  pull_request: {}
+  push: {}
   schedule:
     - cron: '5 1 * * *'  # every day at 01:05
 
@@ -17,6 +13,7 @@ env:
 
 jobs:
   lint:
+    timeout-minutes: 10
     runs-on: ubuntu-latest
     steps:
     - uses: actions/[email protected]
@@ -27,7 +24,7 @@ jobs:
     - name: Install requirements
       run: |
         pip install wheel
-        pip install ".[all,tests]" Pygments collective.checkdocs pre-commit
+        pip install ".[all,tests,terraform]" Pygments collective.checkdocs pre-commit
     - name: install temporary dependencies
       run: |
         pip install git+https://github.com/isidentical/gdrivefs@service-account
@@ -36,6 +33,7 @@ jobs:
       run: python setup.py checkdocs
     - uses: pre-commit/[email protected]
   tests:
+    timeout-minutes: 45
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
@@ -77,7 +75,7 @@ jobs:
       if: steps.cache.pip-cache-dir.cache-hit != 'true'
       run: |
         pip install --upgrade pip setuptools wheel
-        pip install -e ".[all,tests]"
+        pip install -e ".[all,tests,terraform]"
     - name: install temporary dependencies
       run: |
         pip install git+https://github.com/isidentical/gdrivefs@service-account
@@ -101,6 +99,7 @@ jobs:
       run: |
         echo "extra_test_args=--disable-gdrive" >> $GITHUB_ENV
     - name: run tests
+      timeout-minutes: 30
       env:
         GDRIVE_CREDENTIALS_DATA: ${{ secrets.GDRIVE_CREDENTIALS_DATA }}
         AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
@@ -113,7 +112,7 @@ jobs:
         --cov-report=xml --cov-report=term
         --all ${{ env.extra_test_args }}
     - name: upload coverage report
-      uses: codecov/codecov-action@v2.0.2
+      uses: codecov/codecov-action@v2.1.0
       with:
         file: ./coverage.xml
         fail_ci_if_error: false

.pre-commit-config.yaml

@@ -6,7 +6,7 @@ repos:
       - id: black
         language_version: python3
     repo: https://github.com/ambv/black
-    rev: 21.7b0
+    rev: 21.8b0
   - hooks:
       - id: codespell
         args: 

MANIFEST.in

@@ -1,3 +1,3 @@
 include fastentrypoints.py
 include LICENSE
-include test_requirements.txt
+recursive-include requirements *.txt

README.rst

@@ -8,7 +8,7 @@
 • `Tutorial <https://dvc.org/doc/get-started>`_
 • `Mailing List <https://sweedom.us10.list-manage.com/subscribe/post?u=a08bf93caae4063c4e6a351f6&id=24c0ecc49a>`_
 
-|Release| |CI| |Maintainability| |Coverage| |Donate| |DOI|
+|CI| |Maintainability| |Coverage| |Donate| |DOI|
 
 |PyPI| |Packages| |Brew| |Conda| |Choco| |Snap|
 
@@ -20,8 +20,9 @@ learning projects. Key features:
 #. Simple **command line** Git-like experience. Does not require installing and maintaining
    any databases. Does not depend on any proprietary online services.
 
-#. Management and versioning of **datasets** and **machine learning models**. Data is saved in
-   S3, Google cloud, Azure, Alibaba cloud, SSH server, HDFS, or even local HDD RAID.
+#. Management and versioning of **datasets** and **machine learning
+   models**. Data can be saved in S3, Google cloud, Azure, Alibaba 
+   cloud, SSH server, HDFS, or even local HDD RAID.
 
 #. Makes projects **reproducible** and **shareable**; helping to answer questions about how
    a model was built.
@@ -65,25 +66,25 @@ Quick start
 
 Please read `Get Started <https://dvc.org/doc/get-started>`_ guide for a full version. Common workflow commands include:
 
-+-----------------------------------+-------------------------------------------------------------------+
-| Step                              | Command                                                           |
-+===================================+===================================================================+
-| Track data                        | | ``$ git add train.py``                                          |
-|                                   | | ``$ dvc add images.zip``                                        |
-+-----------------------------------+-------------------------------------------------------------------+
-| Connect code and data by commands | | ``$ dvc run -d images.zip -o images/ unzip -q images.zip``      |
-|                                   | | ``$ dvc run -d images/ -d train.py -o model.p python train.py`` |
-+-----------------------------------+-------------------------------------------------------------------+
-| Make changes and reproduce        | | ``$ vi train.py``                                               |
-|                                   | | ``$ dvc repro model.p.dvc``                                     |
-+-----------------------------------+-------------------------------------------------------------------+
-| Share code                        | | ``$ git add .``                                                 |
-|                                   | | ``$ git commit -m 'The baseline model'``                        |
-|                                   | | ``$ git push``                                                  |
-+-----------------------------------+-------------------------------------------------------------------+
-| Share data and ML models          | | ``$ dvc remote add myremote -d s3://mybucket/image_cnn``        |
-|                                   | | ``$ dvc push``                                                  |
-+-----------------------------------+-------------------------------------------------------------------+
++-----------------------------------+----------------------------------------------------------------------------+
+| Step                              | Command                                                                    |
++===================================+============================================================================+
+| Track data                        | | ``$ git add train.py``                                                   |
+|                                   | | ``$ dvc add images.zip``                                                 |
++-----------------------------------+----------------------------------------------------------------------------+
+| Connect code and data by commands | | ``$ dvc run -n prepare -d images.zip -o images/ unzip -q images.zip``    |
+|                                   | | ``$ dvc run -n train -d images/ -d train.py -o model.p python train.py`` |
++-----------------------------------+----------------------------------------------------------------------------+
+| Make changes and reproduce        | | ``$ vi train.py``                                                        |
+|                                   | | ``$ dvc repro model.p.dvc``                                              |
++-----------------------------------+----------------------------------------------------------------------------+
+| Share code                        | | ``$ git add .``                                                          |
+|                                   | | ``$ git commit -m 'The baseline model'``                                 |
+|                                   | | ``$ git push``                                                           |
++-----------------------------------+----------------------------------------------------------------------------+
+| Share data and ML models          | | ``$ dvc remote add myremote -d s3://mybucket/image_cnn``                 |
+|                                   | | ``$ dvc push``                                                           |
++-----------------------------------+----------------------------------------------------------------------------+
 
 Installation
 ============
@@ -182,30 +183,37 @@ Fedora / CentOS (rpm)
 Comparison to related technologies
 ==================================
 
-#. *Data Engineering* tools such as `AirFlow <https://airflow.apache.org/>`_, `Luigi <https://github.com/spotify/luigi>`_ and
-   others - in DVC data, model and ML pipelines represent a single ML project and focuses on data scientist' experience while
-   data engineering pipeline orchestrates multiple data projects and focuses on efficient execution. DVC project can be used
-   from data pipeline as a single execution step. `AirFlow DVC <https://github.com/covid-genomics/airflow-dvc>`_ is an example
-   of such integration.
+#. Data Engineering tools such as `AirFlow <https://airflow.apache.org/>`,
+   `Luigi <https://github.com/spotify/luigi>`, and others - in DVC data,
+   model and ML pipelines represent a single ML project focused on data
+   scientists' experience.  Data engineering tools orchestrate multiple data
+   projects and focus on efficient execution. A DVC project can be used from
+   existing data pipelines as a single execution step.
 
 #. `Git-annex <https://git-annex.branchable.com/>`_ - DVC uses the idea of storing the content of large files (which should
    not be in a Git repository) in a local key-value store, and uses file hardlinks/symlinks instead of
    copying/duplicating files.
 
-#. `Git-LFS <https://git-lfs.github.com/>`_ - DVC is compatible with any remote storage (S3, Google Cloud, Azure, SSH,
-   etc). DVC also uses reflinks or hardlinks to avoid copy operations on checkouts; thus handling large data files
-   much more efficiently.
+#. `Git-LFS <https://git-lfs.github.com/>`_ - DVC is compatible with many
+   remote storage services (S3, Google Cloud, Azure, SSH, etc). DVC also
+   uses reflinks or hardlinks to avoid copy operations on checkouts; thus
+   handling large data files much more efficiently.
 
-#. *Makefile* (and analogues including ad-hoc scripts) - DVC tracks dependencies (in a directed acyclic graph).
+#. Makefile (and analogues including ad-hoc scripts) - DVC tracks
+   dependencies (in a directed acyclic graph).
 
 #. `Workflow Management Systems <https://en.wikipedia.org/wiki/Workflow_management_system>`_ - DVC is a workflow
    management system designed specifically to manage machine learning experiments. DVC is built on top of Git.
 
-#. `DAGsHub <https://dagshub.com/>`_ - online service of Git+DVC repositories with pipeline and metrics visualization and
-   DVC-specific cloud storage.
+#. `DAGsHub <https://dagshub.com/>`_ - online service to host DVC
+   projects.  It provides a useful UI around DVC repositories and integrates
+   other tools.
 
-#. `DVC Studio <https://studio.iterative.ai/>`_ - online service of DVC repositories visualisation from DVC team. Also,
-   integrated with `CML (CI/CD for ML) <https://cml.dev/>`_ for training models in clouds and Kubernetes.
+#. `DVC Studio <https://studio.iterative.ai/>`_ - official online
+   platform for DVC projects.  It can be used to manage data and models, run
+   and track experiments, and visualize and share results.  Also, it
+   integrates with `CML (CI/CD for ML) <https://cml.dev/>` for training
+   models in the cloud or Kubernetes.
 
 
 Contributing
@@ -214,7 +222,9 @@ Contributing
 |Maintainability| |Donate|
 
 Contributions are welcome! Please see our `Contributing Guide <https://dvc.org/doc/user-guide/contributing/core>`_ for more
-details.
+details. Thanks to all our contributors!
+
+|Contribs|
 
 Mailing List
 ============
@@ -244,10 +254,6 @@ Barrak, A., Eghan, E.E. and Adams, B. `On the Co-evolution of ML Pipelines and S
    :target: https://dvc.org
    :alt: DVC logo
 
-.. |Release| image:: https://img.shields.io/badge/release-ok-brightgreen
-   :target: https://travis-ci.com/iterative/dvc/branches
-   :alt: Release
-
 .. |CI| image:: https://github.com/iterative/dvc/workflows/Tests/badge.svg?branch=master
    :target: https://github.com/iterative/dvc/actions
    :alt: GHA Tests
@@ -295,3 +301,7 @@ Barrak, A., Eghan, E.E. and Adams, B. `On the Co-evolution of ML Pipelines and S
 .. |Flowchart| image:: https://dvc.org/img/flow.gif
    :target: https://dvc.org/img/flow.gif
    :alt: how_dvc_works
+
+.. |Contribs| image:: https://contrib.rocks/image?repo=iterative/dvc
+   :target: https://github.com/iterative/dvc/graphs/contributors
+   :alt: Contributors

dvc/cli.py

@@ -30,6 +30,7 @@ from .command import (
     install,
     live,
     ls,
+    machine,
     metrics,
     move,
     params,
@@ -87,6 +88,7 @@ COMMANDS = [
     experiments,
     check_ignore,
     live,
+    machine,
 ]
 
 

dvc/command/config.py

@@ -66,7 +66,8 @@ class CmdConfig(CmdBaseNoRepo):
             )
             return 1
 
-        levels = [self.args.level] if self.args.level else self.config.LEVELS
+        levels = self._get_appropriate_levels(self.args.level)
+
         for level in levels:
             conf = self.config.read(level)
             prefix = self._config_file_prefix(
@@ -81,9 +82,7 @@ class CmdConfig(CmdBaseNoRepo):
     def _get(self, remote, section, opt):
         from dvc.config import ConfigError
 
-        levels = (
-            [self.args.level] if self.args.level else self.config.LEVELS[::-1]
-        )
+        levels = self._get_appropriate_levels(self.args.level)[::-1]
 
         for level in levels:
             conf = self.config.read(level)
@@ -136,6 +135,20 @@ class CmdConfig(CmdBaseNoRepo):
                 f"option '{opt}' doesn't exist in {name} '{section}'"
             )
 
+    def _get_appropriate_levels(self, levels):
+        if levels:
+            self._validate_level_for_non_repo_operation(levels)
+            return [levels]
+        if self.config.dvc_dir is None:
+            return self.config.SYSTEM_LEVELS
+        return self.config.LEVELS
+
+    def _validate_level_for_non_repo_operation(self, level):
+        from dvc.config import ConfigError
+
+        if self.config.dvc_dir is None and level in self.config.REPO_LEVELS:
+            raise ConfigError("Not inside a DVC repo")
+
     @staticmethod
     def _format_config(config, prefix=""):
         from dvc.utils.flatten import flatten

dvc/command/experiments.py

@@ -1,5 +1,6 @@
 import argparse
 import logging
+import os
 from collections import Counter, OrderedDict, defaultdict
 from datetime import date, datetime
 from fnmatch import fnmatch
@@ -141,6 +142,8 @@ def _collect_rows(
     precision=DEFAULT_PRECISION,
     sort_by=None,
     sort_order=None,
+    fill_value=FILL_VALUE,
+    iso=False,
 ):
     from dvc.scm.git import Git
 
@@ -161,8 +164,8 @@ def _collect_rows(
         elif exp.get("queued"):
             state = "Queued"
         else:
-            state = FILL_VALUE
-        executor = exp.get("executor", FILL_VALUE)
+            state = fill_value
+        executor = exp.get("executor", fill_value)
         is_baseline = rev == "baseline"
 
         if is_baseline:
@@ -202,12 +205,12 @@ def _collect_rows(
             exp_name,
             name_rev,
             typ,
-            _format_time(exp.get("timestamp")),
+            _format_time(exp.get("timestamp"), fill_value, iso),
             parent,
             state,
             executor,
         ]
-        fill_value = FILL_VALUE_ERRORED if results.get("error") else FILL_VALUE
+        fill_value = FILL_VALUE_ERRORED if results.get("error") else fill_value
         _extend_row(
             row,
             metric_names,
@@ -274,14 +277,18 @@ def _sort_exp(experiments, sort_path, sort_name, typ, reverse):
     return ret
 
 
-def _format_time(timestamp):
-    if timestamp is None:
-        return FILL_VALUE
-    if timestamp.date() == date.today():
+def _format_time(datetime_obj, fill_value=FILL_VALUE, iso=False):
+    if datetime_obj is None:
+        return fill_value
+
+    if iso:
+        return datetime_obj.isoformat()
+
+    if datetime_obj.date() == date.today():
         fmt = "%I:%M %p"
     else:
         fmt = "%b %d, %Y"
-    return timestamp.strftime(fmt)
+    return datetime_obj.strftime(fmt)
 
 
 def _extend_row(row, names, items, precision, fill_value=FILL_VALUE):
@@ -320,6 +327,7 @@ def _parse_filter_list(param_list):
 
 def experiments_table(
     all_experiments,
+    headers,
     metric_headers,
     metric_names,
     param_headers,
@@ -327,22 +335,15 @@ def experiments_table(
     sort_by=None,
     sort_order=None,
     precision=DEFAULT_PRECISION,
+    fill_value=FILL_VALUE,
+    iso=False,
 ) -> "TabularData":
     from funcy import lconcat
 
     from dvc.compare import TabularData
 
-    headers = [
-        "Experiment",
-        "rev",
-        "typ",
-        "Created",
-        "parent",
-        "State",
-        "Executor",
-    ]
     td = TabularData(
-        lconcat(headers, metric_headers, param_headers), fill_value=FILL_VALUE
+        lconcat(headers, metric_headers, param_headers), fill_value=fill_value
     )
     for base_rev, experiments in all_experiments.items():
         rows = _collect_rows(
@@ -353,6 +354,8 @@ def experiments_table(
             sort_by=sort_by,
             sort_order=sort_order,
             precision=precision,
+            fill_value=fill_value,
+            iso=iso,
         )
         td.extend(rows)
 
@@ -385,8 +388,10 @@ def baseline_styler(typ):
 
 
 def show_experiments(
-    all_experiments, pager=True, no_timestamp=False, **kwargs
+    all_experiments, pager=True, no_timestamp=False, show_csv=False, **kwargs
 ):
+    from funcy.seqs import flatten as flatten_list
+
     include_metrics = _parse_filter_list(kwargs.pop("include_metrics", []))
     exclude_metrics = _parse_filter_list(kwargs.pop("exclude_metrics", []))
     include_params = _parse_filter_list(kwargs.pop("include_params", []))
@@ -399,11 +404,26 @@ def show_experiments(
         include_params=include_params,
         exclude_params=exclude_params,
     )
-    metric_headers = _normalize_headers(metric_names)
-    param_headers = _normalize_headers(param_names)
+
+    headers = [
+        "Experiment",
+        "rev",
+        "typ",
+        "Created",
+        "parent",
+        "State",
+        "Executor",
+    ]
+
+    names = {**metric_names, **param_names}
+    counter = Counter(flatten_list([list(a.keys()) for a in names.values()]))
+    counter.update(headers)
+    metric_headers = _normalize_headers(metric_names, counter)
+    param_headers = _normalize_headers(param_names, counter)
 
     td = experiments_table(
         all_experiments,
+        headers,
         metric_headers,
         metric_names,
         param_headers,
@@ -411,6 +431,8 @@ def show_experiments(
         kwargs.get("sort_by"),
         kwargs.get("sort_order"),
         kwargs.get("precision"),
+        kwargs.get("fill_value"),
+        kwargs.get("iso"),
     )
 
     if no_timestamp:
@@ -422,9 +444,12 @@ def show_experiments(
 
     row_styles = lmap(baseline_styler, td.column("typ"))
 
-    merge_headers = ["Experiment", "rev", "typ", "parent"]
-    td.column("Experiment")[:] = map(prepare_exp_id, td.as_dict(merge_headers))
-    td.drop(*merge_headers[1:])
+    if not show_csv:
+        merge_headers = ["Experiment", "rev", "typ", "parent"]
+        td.column("Experiment")[:] = map(
+            prepare_exp_id, td.as_dict(merge_headers)
+        )
+        td.drop(*merge_headers[1:])
 
     headers = {"metrics": metric_headers, "params": param_headers}
     styles = {
@@ -453,13 +478,11 @@ def show_experiments(
         rich_table=True,
         header_styles=styles,
         row_styles=row_styles,
+        show_csv=show_csv,
     )
 
 
-def _normalize_headers(names):
-    count = Counter(
-        name for path in names for name in names[path] for path in names
-    )
+def _normalize_headers(names, count):
     return [
         name if count[name] == 1 else f"{path}:{name}"
         for path in names
@@ -493,6 +516,14 @@ class CmdExperimentsShow(CmdBase):
 
             ui.write(json.dumps(all_experiments, default=_format_json))
         else:
+            precision = (
+                self.args.precision or None
+                if self.args.show_csv
+                else DEFAULT_PRECISION
+            )
+            fill_value = "" if self.args.show_csv else FILL_VALUE
+            iso = True if self.args.show_csv else False
+
             show_experiments(
                 all_experiments,
                 include_metrics=self.args.include_metrics,
@@ -502,8 +533,11 @@ class CmdExperimentsShow(CmdBase):
                 no_timestamp=self.args.no_timestamp,
                 sort_by=self.args.sort_by,
                 sort_order=self.args.sort_order,
-                precision=self.args.precision or DEFAULT_PRECISION,
+                precision=precision,
+                fill_value=fill_value,
+                iso=iso,
                 pager=not self.args.no_pager,
+                show_csv=self.args.show_csv,
             )
         return 0
 
@@ -747,12 +781,71 @@ class CmdExperimentsRemove(CmdBase):
     def run(self):
 
         self.repo.experiments.remove(
-            exp_names=self.args.experiment, queue=self.args.queue
+            exp_names=self.args.experiment,
+            queue=self.args.queue,
+            clear_all=self.args.all,
+            remote=self.args.git_remote,
         )
 
         return 0
 
 
+class CmdExperimentsInit(CmdBase):
+    CODE = "src"
+    DATA = "data"
+    MODELS = "models"
+    DEFAULT_METRICS = "metrics.json"
+    DEFAULT_PARAMS = "params.yaml"
+    PLOTS = "plots"
+    DVCLIVE = "dvclive"
+    DEFAULT_NAME = "default"
+
+    def run(self):
+        from dvc.command.stage import parse_cmd
+
+        cmd = parse_cmd(self.args.cmd)
+        if not cmd:
+            raise InvalidArgumentError("command is not specified")
+        if self.args.interactive:
+            raise NotImplementedError(
+                "'-i/--interactive' is not implemented yet."
+            )
+        if self.args.explicit:
+            raise NotImplementedError("'--explicit' is not implemented yet.")
+        if self.args.template:
+            raise NotImplementedError("template is not supported yet.")
+
+        from dvc.utils.serialize import LOADERS
+
+        code = self.args.code or self.CODE
+        data = self.args.data or self.DATA
+        models = self.args.models or self.MODELS
+        metrics = self.args.metrics or self.DEFAULT_METRICS
+        params_path = self.args.params or self.DEFAULT_PARAMS
+        plots = self.args.plots or self.PLOTS
+        dvclive = self.args.live or self.DVCLIVE
+
+        _, ext = os.path.splitext(params_path)
+        params = list(LOADERS[ext](params_path))
+
+        name = self.args.name or self.DEFAULT_NAME
+        stage = self.repo.stage.add(
+            name=name,
+            cmd=cmd,
+            deps=[code, data],
+            outs=[models],
+            params=[{params_path: params}],
+            metrics_no_cache=[metrics],
+            plots_no_cache=[plots],
+            live=dvclive,
+            force=True,
+        )
+
+        if self.args.run:
+            return self.repo.experiments.run(targets=[stage.addressing])
+        return 0
+
+
 def add_parser(subparsers, parent_parser):
     EXPERIMENTS_HELP = "Commands to run and compare experiments."
 
@@ -880,6 +973,12 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Print output in JSON format instead of a human-readable table.",
     )
+    experiments_show_parser.add_argument(
+        "--show-csv",
+        action="store_true",
+        default=False,
+        help="Print output in csv format instead of a human-readable table.",
+    )
     experiments_show_parser.add_argument(
         "--precision",
         type=int,
@@ -1109,9 +1208,9 @@ def add_parser(subparsers, parent_parser):
         nargs="?",
         default=None,
         help=(
-            "Optional Git remote name or Git URL. If provided, experiments "
-            "from the specified Git repository will be listed instead of "
-            "local experiments."
+            "Optional Git remote name or Git URL. "
+            "If provided, experiments from the specified Git repository "
+            " will be listed instead of local ones."
         ),
         metavar="[<git_remote>]",
     )
@@ -1229,7 +1328,7 @@ def add_parser(subparsers, parent_parser):
     )
     experiments_pull_parser.set_defaults(func=CmdExperimentsPull)
 
-    EXPERIMENTS_REMOVE_HELP = "Remove local experiments."
+    EXPERIMENTS_REMOVE_HELP = "Remove experiments."
     experiments_remove_parser = experiments_subparsers.add_parser(
         "remove",
         parents=[parent_parser],
@@ -1237,9 +1336,22 @@ def add_parser(subparsers, parent_parser):
         help=EXPERIMENTS_REMOVE_HELP,
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
-    experiments_remove_parser.add_argument(
+    remove_group = experiments_remove_parser.add_mutually_exclusive_group()
+    remove_group.add_argument(
         "--queue", action="store_true", help="Remove all queued experiments."
     )
+    remove_group.add_argument(
+        "-A",
+        "--all",
+        action="store_true",
+        help="Remove all committed experiments.",
+    )
+    remove_group.add_argument(
+        "-g",
+        "--git-remote",
+        metavar="<git_remote>",
+        help="Name or URL of the Git remote to remove the experiment from",
+    )
     experiments_remove_parser.add_argument(
         "experiment",
         nargs="*",
@@ -1248,6 +1360,68 @@ def add_parser(subparsers, parent_parser):
     )
     experiments_remove_parser.set_defaults(func=CmdExperimentsRemove)
 
+    EXPERIMENTS_INIT_HELP = "Initialize experiments."
+    experiments_init_parser = experiments_subparsers.add_parser(
+        "init",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_INIT_HELP, "exp/init"),
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_init_parser.add_argument(
+        "cmd",
+        nargs=argparse.REMAINDER,
+        help="Command to execute.",
+        metavar="command",
+    )
+    experiments_init_parser.add_argument(
+        "--run",
+        action="store_true",
+        help="Run the experiment after initializing it",
+    )
+    experiments_init_parser.add_argument(
+        "--interactive",
+        "-i",
+        action="store_true",
+        help="Prompt for values that are not provided",
+    )
+    experiments_init_parser.add_argument(
+        "--template", help="Stage template to use to fill with provided values"
+    )
+    experiments_init_parser.add_argument(
+        "--explicit", help="Only use the path values explicitly provided"
+    )
+    experiments_init_parser.add_argument(
+        "--name", "-n", help="Name of the stage to create"
+    )
+    experiments_init_parser.add_argument(
+        "--code",
+        help="Path to the source file or directory "
+        "which your experiments depend",
+    )
+    experiments_init_parser.add_argument(
+        "--data",
+        help="Path to the data file or directory "
+        "which your experiments depend",
+    )
+    experiments_init_parser.add_argument(
+        "--models",
+        help="Path to the model file or directory for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--params", help="Path to the parameters file for your experiments"
+    )
+    experiments_init_parser.add_argument(
+        "--metrics", help="Path to the metrics file for your experiments"
+    )
+    experiments_init_parser.add_argument(
+        "--plots",
+        help="Path to the plots file or directory for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--live", help="Path to log dvclive outputs for your experiments"
+    )
+    experiments_init_parser.set_defaults(func=CmdExperimentsInit)
+
 
 def _add_run_common(parser):
     """Add common args for 'exp run' and 'exp resume'."""

dvc/command/live.py

@@ -1,5 +1,5 @@
 import argparse
-import os
+from pathlib import Path
 
 from dvc.command import completion
 from dvc.command.base import CmdBase, fix_subparsers
@@ -13,10 +13,11 @@ class CmdLive(CmdBase):
         metrics, plots = self.repo.live.show(target=target, revs=revs)
 
         if plots:
-            html_path = self.args.target + ".html"
-            self.repo.plots.write_html(html_path, plots, metrics)
+            html_path = Path.cwd() / (self.args.target + "_html")
+            from dvc.render.utils import render
 
-            ui.write("\nfile://", os.path.abspath(html_path), sep="")
+            index_path = render(self.repo, plots, metrics, html_path)
+            ui.write(index_path.as_uri())
             return 0
         return 1
 

dvc/command/machine.py

@@ -0,0 +1,333 @@
+import argparse
+
+from dvc.command.base import CmdBase, append_doc_link, fix_subparsers
+from dvc.command.config import CmdConfig
+from dvc.config import ConfigError
+from dvc.ui import ui
+from dvc.utils import format_link
+
+
+class MachineDisabledError(ConfigError):
+    def __init__(self):
+        super().__init__("Machine feature is disabled")
+
+
+class CmdMachineConfig(CmdConfig):
+    def __init__(self, args):
+
+        super().__init__(args)
+        if not self.config["feature"].get("machine", False):
+            raise MachineDisabledError
+
+        if getattr(self.args, "name", None):
+            self.args.name = self.args.name.lower()
+
+    def _check_exists(self, conf):
+        if self.args.name not in conf["machine"]:
+            raise ConfigError(f"machine '{self.args.name}' doesn't exist.")
+
+
+class CmdMachineAdd(CmdMachineConfig):
+    def run(self):
+        from dvc.machine import validate_name
+
+        validate_name(self.args.name)
+
+        if self.args.default:
+            ui.write(f"Setting '{self.args.name}' as a default machine.")
+
+        with self.config.edit(self.args.level) as conf:
+            if self.args.name in conf["machine"] and not self.args.force:
+                raise ConfigError(
+                    "machine '{}' already exists. Use `-f|--force` to "
+                    "overwrite it.".format(self.args.name)
+                )
+
+            conf["machine"][self.args.name] = {"cloud": self.args.cloud}
+            if self.args.default:
+                conf["core"]["machine"] = self.args.name
+
+        return 0
+
+
+class CmdMachineRemove(CmdMachineConfig):
+    def run(self):
+        with self.config.edit(self.args.level) as conf:
+            self._check_exists(conf)
+            del conf["machine"][self.args.name]
+
+        up_to_level = self.args.level or "repo"
+        # Remove core.machine refs to this machine in any shadowing configs
+        for level in reversed(self.config.LEVELS):
+            with self.config.edit(level) as conf:
+                if conf["core"].get("machine") == self.args.name:
+                    del conf["core"]["machine"]
+
+            if level == up_to_level:
+                break
+
+        return 0
+
+
+class CmdMachineList(CmdMachineConfig):
+    def run(self):
+        levels = [self.args.level] if self.args.level else self.config.LEVELS
+        for level in levels:
+            conf = self.config.read(level)["machine"]
+            if self.args.name:
+                conf = conf.get(self.args.name, {})
+            prefix = self._config_file_prefix(
+                self.args.show_origin, self.config, level
+            )
+            configs = list(self._format_config(conf, prefix))
+            if configs:
+                ui.write("\n".join(configs))
+        return 0
+
+
+class CmdMachineModify(CmdMachineConfig):
+    def run(self):
+        from dvc.config import merge
+
+        with self.config.edit(self.args.level) as 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["machine"]:
+                conf["machine"][self.args.name] = {}
+            section = conf["machine"][self.args.name]
+            if self.args.unset:
+                section.pop(self.args.option, None)
+            else:
+                section[self.args.option] = self.args.value
+        return 0
+
+
+class CmdMachineDefault(CmdMachineConfig):
+    def run(self):
+        if self.args.name is None and not self.args.unset:
+            conf = self.config.read(self.args.level)
+            try:
+                print(conf["core"]["machine"])
+            except KeyError:
+                ui.write("No default machine set")
+                return 1
+        else:
+            with self.config.edit(self.args.level) as conf:
+                if self.args.unset:
+                    conf["core"].pop("machine", None)
+                else:
+                    merged_conf = self.config.load_config_to_level(
+                        self.args.level
+                    )
+                    if (
+                        self.args.name in conf["machine"]
+                        or self.args.name in merged_conf["machine"]
+                    ):
+                        conf["core"]["machine"] = self.args.name
+                    else:
+                        raise ConfigError(
+                            "default machine must be present in machine "
+                            "list."
+                        )
+        return 0
+
+
+class CmdMachineCreate(CmdBase):
+    def run(self):
+        if self.repo.machine is None:
+            raise MachineDisabledError
+
+        self.repo.machine.create(self.args.name)
+        return 0
+
+
+class CmdMachineDestroy(CmdBase):
+    def run(self):
+        if self.repo.machine is None:
+            raise MachineDisabledError
+
+        self.repo.machine.destroy(self.args.name)
+        return 0
+
+
+class CmdMachineSsh(CmdBase):
+    def run(self):
+        if self.repo.machine is None:
+            raise MachineDisabledError
+
+        self.repo.machine.run_shell(self.args.name)
+        return 0
+
+
+def add_parser(subparsers, parent_parser):
+    from dvc.command.config import parent_config_parser
+
+    machine_HELP = "Set up and manage cloud machines."
+    machine_parser = subparsers.add_parser(
+        "machine",
+        parents=[parent_parser],
+        description=append_doc_link(machine_HELP, "machine"),
+        # NOTE: suppress help during development to hide command
+        # help=machine_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+
+    machine_subparsers = machine_parser.add_subparsers(
+        dest="cmd",
+        help="Use `dvc machine CMD --help` for " "command-specific help.",
+    )
+
+    fix_subparsers(machine_subparsers)
+
+    machine_ADD_HELP = "Add a new data machine."
+    machine_add_parser = machine_subparsers.add_parser(
+        "add",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_ADD_HELP, "machine/add"),
+        help=machine_ADD_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_add_parser.add_argument("name", help="Name of the machine")
+    machine_add_parser.add_argument(
+        "cloud",
+        help="Machine cloud. See full list of supported clouds at {}".format(
+            format_link(
+                "https://github.com/iterative/"
+                "terraform-provider-iterative#machine"
+            )
+        ),
+    )
+    machine_add_parser.add_argument(
+        "-d",
+        "--default",
+        action="store_true",
+        default=False,
+        help="Set as default machine.",
+    )
+    machine_add_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Force overwriting existing configs",
+    )
+    machine_add_parser.set_defaults(func=CmdMachineAdd)
+
+    machine_DEFAULT_HELP = "Set/unset the default machine."
+    machine_default_parser = machine_subparsers.add_parser(
+        "default",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_DEFAULT_HELP, "machine/default"),
+        help=machine_DEFAULT_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_default_parser.add_argument(
+        "name", nargs="?", help="Name of the machine"
+    )
+    machine_default_parser.add_argument(
+        "-u",
+        "--unset",
+        action="store_true",
+        default=False,
+        help="Unset default machine.",
+    )
+    machine_default_parser.set_defaults(func=CmdMachineDefault)
+
+    machine_LIST_HELP = "List the configuration of one/all machines."
+    machine_list_parser = machine_subparsers.add_parser(
+        "list",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_LIST_HELP, "machine/list"),
+        help=machine_LIST_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_list_parser.add_argument(
+        "--show-origin",
+        default=False,
+        action="store_true",
+        help="Show the source file containing each config value.",
+    )
+    machine_list_parser.add_argument(
+        "name",
+        nargs="?",
+        type=str,
+        help="name of machine to specify",
+    )
+    machine_list_parser.set_defaults(func=CmdMachineList)
+    machine_MODIFY_HELP = "Modify the configuration of an machine."
+    machine_modify_parser = machine_subparsers.add_parser(
+        "modify",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_MODIFY_HELP, "machine/modify"),
+        help=machine_MODIFY_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_modify_parser.add_argument("name", help="Name of the machine")
+    machine_modify_parser.add_argument(
+        "option", help="Name of the option to modify."
+    )
+    machine_modify_parser.add_argument(
+        "value", nargs="?", help="(optional) Value of the option."
+    )
+    machine_modify_parser.add_argument(
+        "-u",
+        "--unset",
+        default=False,
+        action="store_true",
+        help="Unset option.",
+    )
+    machine_modify_parser.set_defaults(func=CmdMachineModify)
+
+    machine_REMOVE_HELP = "Remove an machine."
+    machine_remove_parser = machine_subparsers.add_parser(
+        "remove",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_REMOVE_HELP, "machine/remove"),
+        help=machine_REMOVE_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_remove_parser.add_argument(
+        "name", help="Name of the machine to remove."
+    )
+    machine_remove_parser.set_defaults(func=CmdMachineRemove)
+
+    machine_CREATE_HELP = "Create and start a machine instance."
+    machine_create_parser = machine_subparsers.add_parser(
+        "create",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_CREATE_HELP, "machine/create"),
+        help=machine_CREATE_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_create_parser.add_argument(
+        "name", help="Name of the machine to create."
+    )
+    machine_create_parser.set_defaults(func=CmdMachineCreate)
+
+    machine_DESTROY_HELP = "Destroy an machine instance."
+    machine_destroy_parser = machine_subparsers.add_parser(
+        "destroy",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_DESTROY_HELP, "machine/destroy"),
+        help=machine_DESTROY_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_destroy_parser.add_argument(
+        "name", help="Name of the machine instance to destroy."
+    )
+    machine_destroy_parser.set_defaults(func=CmdMachineDestroy)
+
+    machine_SSH_HELP = "Connect to a machine via SSH."
+    machine_ssh_parser = machine_subparsers.add_parser(
+        "ssh",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_SSH_HELP, "machine/ssh"),
+        help=machine_SSH_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_ssh_parser.add_argument(
+        "name", help="Name of the machine instance to connect to."
+    )
+    machine_ssh_parser.set_defaults(func=CmdMachineSsh)

dvc/command/plots.py

@@ -4,6 +4,7 @@ import logging
 from dvc.command import completion
 from dvc.command.base import CmdBase, append_doc_link, fix_subparsers
 from dvc.exceptions import DvcException
+from dvc.render.utils import find_vega, render
 from dvc.ui import ui
 from dvc.utils import format_link
 
@@ -35,43 +36,50 @@ class CmdPlots(CmdBase):
                 return 1
 
         try:
-            plots = self._func(targets=self.args.targets, props=self._props())
+            plots_data = self._func(
+                targets=self.args.targets, props=self._props()
+            )
+
+            if not plots_data:
+                ui.error_write(
+                    "No plots were loaded, "
+                    "visualization file will not be created."
+                )
 
             if self.args.show_vega:
                 target = self.args.targets[0]
-                ui.write(plots[target])
+                plot_json = find_vega(self.repo, plots_data, target)
+                if plot_json:
+                    ui.write(plot_json)
                 return 0
 
-        except DvcException:
-            logger.exception("")
-            return 1
-
-        if plots:
-            rel: str = self.args.out or "plots.html"
+            rel: str = self.args.out or "dvc_plots"
             path: Path = (Path.cwd() / rel).resolve()
-            self.repo.plots.write_html(
-                path, plots=plots, html_template_path=self.args.html_template
+            index_path = render(
+                self.repo,
+                plots_data,
+                path=path,
+                html_template_path=self.args.html_template,
             )
 
-            assert (
-                path.is_absolute()
-            )  # as_uri throws ValueError if not absolute
-            url = path.as_uri()
+            assert index_path.is_absolute()
+            url = index_path.as_uri()
             ui.write(url)
+
             if self.args.open:
                 import webbrowser
 
-                opened = webbrowser.open(rel)
+                opened = webbrowser.open(index_path)
                 if not opened:
                     ui.error_write(
                         "Failed to open. Please try opening it manually."
                     )
                     return 1
-        else:
-            ui.error_write(
-                "No plots were loaded, visualization file will not be created."
-            )
-        return 0
+            return 0
+
+        except DvcException:
+            logger.exception("")
+            return 1
 
 
 class CmdPlotsShow(CmdPlots):

dvc/compare.py

@@ -168,7 +168,10 @@ class TabularData(MutableSequence[Sequence["CellT"]]):
     def render(self, **kwargs: Any):
         from dvc.ui import ui
 
-        ui.table(self, headers=self.keys(), **kwargs)
+        if kwargs.pop("show_csv", False):
+            ui.write(self.to_csv(), end="")
+        else:
+            ui.table(self, headers=self.keys(), **kwargs)
 
     def as_dict(
         self, cols: Iterable[str] = None

dvc/config.py

@@ -32,6 +32,18 @@ class RemoteNotFoundError(RemoteConfigError):
     pass
 
 
+class MachineConfigError(ConfigError):
+    pass
+
+
+class NoMachineError(MachineConfigError):
+    pass
+
+
+class MachineNotFoundError(MachineConfigError):
+    pass
+
+
 @memoize
 def get_compiled_schema():
     from voluptuous import Schema
@@ -64,8 +76,10 @@ class Config(dict):
     APPNAME = "dvc"
     APPAUTHOR = "iterative"
 
+    SYSTEM_LEVELS = ("system", "global")
+    REPO_LEVELS = ("repo", "local")
     # In the order they shadow each other
-    LEVELS = ("system", "global", "repo", "local")
+    LEVELS = SYSTEM_LEVELS + REPO_LEVELS
 
     CONFIG = "config"
     CONFIG_LOCAL = "config.local"
@@ -167,7 +181,7 @@ class Config(dict):
                 conf_obj = ConfigObj(fobj)
         else:
             conf_obj = ConfigObj()
-        return _parse_remotes(_lower_keys(conf_obj.dict()))
+        return _parse_named(_lower_keys(conf_obj.dict()))
 
     def _save_config(self, level, conf_dict):
         from configobj import ConfigObj
@@ -179,7 +193,7 @@ class Config(dict):
 
         fs.makedirs(os.path.dirname(filename))
 
-        config = ConfigObj(_pack_remotes(conf_dict))
+        config = ConfigObj(_pack_named(conf_dict))
         with fs.open(filename, "wb") as fobj:
             config.write(fobj)
         config.filename = filename
@@ -251,6 +265,7 @@ class Config(dict):
                     "key_path": func,
                 }
             },
+            "machine": {str: {"startup_script": func}},
         }
         return Schema(dirs_schema, extra=ALLOW_EXTRA)(conf)
 
@@ -273,7 +288,7 @@ class Config(dict):
     def edit(self, level=None, validate=True):
         # NOTE: we write to repo config by default, same as git config
         level = level or "repo"
-        if level in {"repo", "local"} and self.dvc_dir is None:
+        if self.dvc_dir is None and level in self.REPO_LEVELS:
             raise ConfigError("Not inside a DVC repo")
 
         conf = self.load_one(level)
@@ -300,27 +315,29 @@ class Config(dict):
             raise ConfigError(str(exc)) from None
 
 
-def _parse_remotes(conf):
-    result = {"remote": {}}
+def _parse_named(conf):
+    result = {"remote": {}, "machine": {}}
 
     for section, val in conf.items():
-        name = re_find(r'^\s*remote\s*"(.*)"\s*$', section)
-        if name:
-            result["remote"][name] = val
+        match = re_find(r'^\s*(remote|machine)\s*"(.*)"\s*$', section)
+        if match:
+            key, name = match
+            result[key][name] = val
         else:
             result[section] = val
 
     return result
 
 
-def _pack_remotes(conf):
+def _pack_named(conf):
     # Drop empty sections
     result = compact(conf)
 
     # Transform remote.name -> 'remote "name"'
-    for name, val in conf["remote"].items():
-        result[f'remote "{name}"'] = val
-    result.pop("remote", None)
+    for key in ("remote", "machine"):
+        for name, val in conf[key].items():
+            result[f'{key} "{name}"'] = val
+        result.pop(key, None)
 
     return result
 

dvc/config_schema.py

@@ -118,6 +118,7 @@ SCHEMA = {
         Optional("autostage", default=False): Bool,
         Optional("experiments"): Bool,  # obsoleted
         Optional("check_update", default=True): Bool,
+        "machine": Lower,
     },
     "cache": {
         "local": str,
@@ -154,6 +155,7 @@ SCHEMA = {
                     "grant_read_acp": str,
                     "grant_write_acp": str,
                     "grant_full_control": str,
+                    "cache_regions": bool,
                     **REMOTE_COMMON,
                 },
                 "gs": {
@@ -224,10 +226,28 @@ SCHEMA = {
         "row_limit": All(Coerce(int), Range(1)),  # obsoleted
         "row_cleanup_quota": All(Coerce(int), Range(0, 100)),  # obsoleted
     },
+    "machine": {
+        str: {
+            "cloud": All(Lower, Choices("aws", "azure")),
+            "region": All(
+                Lower, Choices("us-west", "us-east", "eu-west", "eu-north")
+            ),
+            "image": str,
+            "name": str,
+            "spot": Bool,
+            "spot_price": Coerce(float),
+            "instance_hdd_size": Coerce(int),
+            "instance_type": Lower,
+            "instance_gpu": Lower,
+            "ssh_private": str,
+            "startup_script": str,
+        },
+    },
     # section for experimental features
     "feature": {
+        Optional("machine", default=False): Bool,
         # enabled by default. It's of no use, kept for backward compatibility.
-        Optional("parametrization", default=True): Bool
+        Optional("parametrization", default=True): Bool,
     },
     "plots": {"html_template": str},
 }

dvc/daemon.py

@@ -12,29 +12,34 @@ from dvc.utils import fix_env, is_binary
 
 logger = logging.getLogger(__name__)
 
-CREATE_NEW_PROCESS_GROUP = 0x00000200
-DETACHED_PROCESS = 0x00000008
+CREATE_NO_WINDOW = 0x08000000  # only available since Python 3.7 in subprocess.
 
 
 def _popen(cmd, **kwargs):
     prefix = [sys.executable]
     if not is_binary():
-        prefix += [sys.argv[0]]
-
+        main_entrypoint = os.path.join(
+            os.path.abspath(os.path.dirname(__file__)), "__main__.py"
+        )
+        prefix += [main_entrypoint]
     return Popen(prefix + cmd, close_fds=True, shell=False, **kwargs)
 
 
 def _spawn_windows(cmd, env):
-    from subprocess import STARTF_USESHOWWINDOW, STARTUPINFO
+    from subprocess import (
+        CREATE_NEW_PROCESS_GROUP,
+        STARTF_USESHOWWINDOW,
+        STARTUPINFO,
+    )
 
-    creationflags = CREATE_NEW_PROCESS_GROUP | DETACHED_PROCESS
+    # https://stackoverflow.com/a/7006424
+    # https://bugs.python.org/issue41619
+    creationflags = CREATE_NEW_PROCESS_GROUP | CREATE_NO_WINDOW
 
     startupinfo = STARTUPINFO()
     startupinfo.dwFlags |= STARTF_USESHOWWINDOW
 
-    _popen(
-        cmd, env=env, creationflags=creationflags, startupinfo=startupinfo
-    ).communicate()
+    _popen(cmd, env=env, creationflags=creationflags, startupinfo=startupinfo)
 
 
 def _spawn_posix(cmd, env):

dvc/fs/__init__.py

@@ -2,6 +2,7 @@ import posixpath
 from urllib.parse import urlparse
 
 from ..scheme import Schemes
+from . import utils  # noqa: F401
 from .azure import AzureFileSystem
 from .gdrive import GDriveFileSystem
 from .gs import GSFileSystem

dvc/fs/azure.py

@@ -11,7 +11,7 @@ from dvc.path_info import CloudURLInfo
 from dvc.scheme import Schemes
 from dvc.utils import format_link
 
-from .fsspec_wrapper import ObjectFSWrapper
+from .fsspec_wrapper import CallbackMixin, ObjectFSWrapper
 
 logger = logging.getLogger(__name__)
 _DEFAULT_CREDS_STEPS = (
@@ -39,7 +39,7 @@ def _az_config():
 
 
 # pylint:disable=abstract-method
-class AzureFileSystem(ObjectFSWrapper):
+class AzureFileSystem(CallbackMixin, ObjectFSWrapper):
     scheme = Schemes.AZURE
     PATH_CLS = CloudURLInfo
     PARAM_CHECKSUM = "etag"

dvc/fs/base.py

@@ -1,15 +1,20 @@
+import contextlib
 import logging
+import os
 from concurrent.futures import ThreadPoolExecutor, as_completed
-from functools import partial, partialmethod
+from functools import partialmethod
 from multiprocessing import cpu_count
 from typing import Any, ClassVar, Dict, FrozenSet, Optional
 
+from funcy import cached_property
+from tqdm.utils import CallbackIOWrapper
+
 from dvc.exceptions import DvcException
 from dvc.path_info import URLInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK, FsspecCallback
+from dvc.ui import ui
 from dvc.utils import tmp_fname
 from dvc.utils.fs import makedirs, move
-from dvc.utils.http import open_url
 
 logger = logging.getLogger(__name__)
 
@@ -125,13 +130,10 @@ class BaseFileSystem:
             f"dependencies: {missing}. {hint}"
         )
 
-    def open(self, path_info, mode: str = "r", encoding: str = None, **kwargs):
-        if hasattr(self, "_generate_download_url"):
-            # pylint:disable=no-member
-            func = self._generate_download_url  # type: ignore[attr-defined]
-            get_url = partial(func, path_info)
-            return open_url(get_url, mode=mode, encoding=encoding, **kwargs)
+    def checksum(self, path_info) -> str:
+        raise NotImplementedError
 
+    def open(self, path_info, mode: str = "r", encoding: str = None, **kwargs):
         raise RemoteActionNotImplemented("open", self.scheme)
 
     def exists(self, path_info) -> bool:
@@ -221,43 +223,64 @@ class BaseFileSystem:
             return False
         return hash_.endswith(cls.CHECKSUM_DIR_SUFFIX)
 
-    def upload(self, from_info, to_info, name=None, no_progress_bar=False):
-        if not hasattr(self, "_upload"):
-            raise RemoteActionNotImplemented("upload", self.scheme)
+    @cached_property
+    def _local_fs(self):
+        from dvc.fs import LocalFileSystem
+
+        return LocalFileSystem()
+
+    def upload(
+        self,
+        from_info,
+        to_info,
+        total=None,
+        desc=None,
+        callback=None,
+        no_progress_bar=False,
+        **pbar_args,
+    ):
+        is_file_obj = hasattr(from_info, "read")
+        method = "upload_fobj" if is_file_obj else "put_file"
+        if not hasattr(self, method):
+            raise RemoteActionNotImplemented(method, self.scheme)
 
         if to_info.scheme != self.scheme:
             raise NotImplementedError
 
-        if from_info.scheme != "local":
-            raise NotImplementedError
+        if not is_file_obj:
+            desc = desc or from_info.name
 
-        logger.debug("Uploading '%s' to '%s'", from_info, to_info)
+        stack = contextlib.ExitStack()
+        if not callback:
+            pbar = ui.progress(
+                desc=desc,
+                disable=no_progress_bar,
+                bytes=True,
+                total=total or -1,
+                **pbar_args,
+            )
+            stack.enter_context(pbar)
+            callback = pbar.as_callback(self._local_fs, from_info)
+            if total:
+                callback.set_size(total)
 
-        name = name or from_info.name
+        with stack:
+            if is_file_obj:
+                wrapped = CallbackIOWrapper(
+                    callback.relative_update, from_info
+                )
+                # `size` is used to provide hints to the WebdavFileSystem
+                # for legacy servers.
+                # pylint: disable=no-member
+                return self.upload_fobj(wrapped, to_info, size=total)
 
-        self._upload(  # noqa, pylint: disable=no-member
-            from_info.fspath,
-            to_info,
-            name=name,
-            no_progress_bar=no_progress_bar,
-        )
+            if from_info.scheme != "local":
+                raise NotImplementedError
 
-    def upload_fobj(
-        self, fobj, to_info, no_progress_bar=False, size=None, **pbar_args
-    ):
-        if not hasattr(self, "_upload_fobj"):
-            raise RemoteActionNotImplemented("upload_fobj", self.scheme)
-
-        with Tqdm.wrapattr(
-            fobj,
-            "read",
-            disable=no_progress_bar,
-            bytes=True,
-            total=size,
-            **pbar_args,
-        ) as wrapped:
-            self._upload_fobj(  # pylint: disable=no-member
-                wrapped, to_info, size=size
+            logger.debug("Uploading '%s' to '%s'", from_info, to_info)
+            # pylint: disable=no-member
+            return self.put_file(
+                os.fspath(from_info), to_info, callback=callback
             )
 
     def download(
@@ -265,84 +288,105 @@ class BaseFileSystem:
         from_info,
         to_info,
         name=None,
+        callback=None,
         no_progress_bar=False,
         jobs=None,
         _only_file=False,
         **kwargs,
     ):
-        if not hasattr(self, "_download"):
-            raise RemoteActionNotImplemented("download", self.scheme)
+        if not hasattr(self, "get_file"):
+            raise RemoteActionNotImplemented("get_file", self.scheme)
 
         if from_info.scheme != self.scheme:
             raise NotImplementedError
 
-        if to_info.scheme == self.scheme != "local":
-            self.copy(from_info, to_info)
-            return 0
-
         if to_info.scheme != "local":
             raise NotImplementedError
 
-        if not _only_file and self.isdir(from_info):
-            return self._download_dir(
-                from_info, to_info, name, no_progress_bar, jobs, **kwargs
+        download_dir = not _only_file and self.isdir(from_info)
+
+        desc = name or to_info.name
+        stack = contextlib.ExitStack()
+        if not callback:
+            pbar_kwargs = {"unit": "Files"} if download_dir else {}
+            pbar = ui.progress(
+                total=-1,
+                desc="Downloading directory" if download_dir else desc,
+                bytes=not download_dir,
+                disable=no_progress_bar,
+                **pbar_kwargs,
             )
-        return self._download_file(from_info, to_info, name, no_progress_bar)
+            stack.enter_context(pbar)
+            callback = pbar.as_callback(self, from_info)
+
+        with stack:
+            if download_dir:
+                return self._download_dir(
+                    from_info, to_info, callback=callback, jobs=jobs, **kwargs
+                )
+            return self._download_file(from_info, to_info, callback=callback)
 
     download_file = partialmethod(download, _only_file=True)
 
     def _download_dir(
-        self, from_info, to_info, name, no_progress_bar, jobs, **kwargs
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        jobs=None,
+        **kwargs,
     ):
         from_infos = list(self.walk_files(from_info, **kwargs))
         if not from_infos:
             makedirs(to_info, exist_ok=True)
             return None
+
         to_infos = (
             to_info / info.relative_to(from_info) for info in from_infos
         )
+        callback.set_size(len(from_infos))
 
-        with Tqdm(
-            total=len(from_infos),
-            desc="Downloading directory",
-            unit="Files",
-            disable=no_progress_bar,
-        ) as pbar:
-            download_files = pbar.wrap_fn(
-                partial(self._download_file, name=name, no_progress_bar=True)
-            )
-            max_workers = jobs or self.jobs
-            with ThreadPoolExecutor(max_workers=max_workers) as executor:
-                futures = [
-                    executor.submit(download_files, from_info, to_info)
-                    for from_info, to_info in zip(from_infos, to_infos)
-                ]
+        download_files = FsspecCallback.wrap_fn(callback, self._download_file)
+        max_workers = jobs or self.jobs
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            futures = [
+                executor.submit(download_files, from_info, to_info)
+                for from_info, to_info in zip(from_infos, to_infos)
+            ]
 
-                # NOTE: unlike pulling/fetching cache, where we need to
-                # download everything we can, not raising an error here might
-                # turn very ugly, as the user might think that he has
-                # downloaded a complete directory, while having a partial one,
-                # which might cause unexpected results in his pipeline.
-                for future in as_completed(futures):
-                    # NOTE: executor won't let us raise until all futures that
-                    # it has are finished, so we need to cancel them ourselves
-                    # before re-raising.
-                    exc = future.exception()
-                    if exc:
-                        for entry in futures:
-                            entry.cancel()
-                        raise exc
+            # NOTE: unlike pulling/fetching cache, where we need to
+            # download everything we can, not raising an error here might
+            # turn very ugly, as the user might think that he has
+            # downloaded a complete directory, while having a partial one,
+            # which might cause unexpected results in his pipeline.
+            for future in as_completed(futures):
+                # NOTE: executor won't let us raise until all futures that
+                # it has are finished, so we need to cancel them ourselves
+                # before re-raising.
+                exc = future.exception()
+                if exc:
+                    for entry in futures:
+                        entry.cancel()
+                    raise exc
 
-    def _download_file(self, from_info, to_info, name, no_progress_bar):
+    def _download_file(
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+    ):
         makedirs(to_info.parent, exist_ok=True)
-
-        logger.debug("Downloading '%s' to '%s'", from_info, to_info)
-        name = name or to_info.name
-
         tmp_file = tmp_fname(to_info)
 
-        self._download(  # noqa, pylint: disable=no-member
-            from_info, tmp_file, name=name, no_progress_bar=no_progress_bar
-        )
+        logger.debug("Downloading '%s' to '%s'", from_info, to_info)
+        try:
+            # noqa, pylint: disable=no-member
+            self.get_file(from_info, tmp_file, callback=callback)
+        except Exception:  # pylint: disable=broad-except
+            # do we need to rollback makedirs for previously not-existing
+            # directories?
+            with contextlib.suppress(FileNotFoundError):
+                os.unlink(tmp_file)
+            raise
 
         move(tmp_file, to_info)

dvc/fs/dvc.py

@@ -6,6 +6,7 @@ from dvc.exceptions import OutputNotFoundError
 from dvc.path_info import PathInfo
 from dvc.utils import relpath
 
+from ..progress import DEFAULT_CALLBACK
 from ._metadata import Metadata
 from .base import BaseFileSystem
 
@@ -256,8 +257,17 @@ class DvcFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
 
         return ret
 
-    def _download(self, from_info, to_file, **kwargs):
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         fs, path = self._get_fs_path(from_info)
-        fs._download(  # pylint: disable=protected-access
-            path, to_file, **kwargs
+        fs.get_file(  # pylint: disable=protected-access
+            path, to_file, callback=callback, **kwargs
         )
+
+    def checksum(self, path_info):
+        info = self.info(path_info)
+        md5 = info.get("md5")
+        if md5:
+            return md5
+        raise NotImplementedError

dvc/fs/fsspec_wrapper.py

@@ -3,8 +3,9 @@ import shutil
 from functools import lru_cache
 
 from funcy import cached_property
+from tqdm.utils import CallbackIOWrapper
 
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK
 
 from .base import BaseFileSystem
 from .local import LocalFileSystem
@@ -85,6 +86,9 @@ class FSSpecWrapper(BaseFileSystem):
     ):  # pylint: disable=arguments-differ
         return self.fs.open(self._with_bucket(path_info), mode=mode)
 
+    def checksum(self, path_info):
+        return self.fs.checksum(self._with_bucket(path_info))
+
     def copy(self, from_info, to_info):
         self.makedirs(to_info.parent)
         self.fs.copy(self._with_bucket(from_info), self._with_bucket(to_info))
@@ -127,43 +131,26 @@ class FSSpecWrapper(BaseFileSystem):
             self._with_bucket(path_info), exist_ok=kwargs.pop("exist_ok", True)
         )
 
-    def _upload_fobj(self, fobj, to_info, size=None):
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
+    ):
+        self.fs.put_file(
+            from_file, self._with_bucket(to_info), callback=callback, **kwargs
+        )
+        self.fs.invalidate_cache(self._with_bucket(to_info.parent))
+
+    def get_file(
+        self, from_info, to_info, callback=DEFAULT_CALLBACK, **kwargs
+    ):
+        self.fs.get_file(
+            self._with_bucket(from_info), to_info, callback=callback, **kwargs
+        )
+
+    def upload_fobj(self, fobj, to_info, **kwargs):
         self.makedirs(to_info.parent)
         with self.open(to_info, "wb") as fdest:
             shutil.copyfileobj(fobj, fdest, length=fdest.blocksize)
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **kwargs
-    ):
-        self.makedirs(to_info.parent)
-        size = os.path.getsize(from_file)
-        with open(from_file, "rb") as fobj:
-            self.upload_fobj(
-                fobj,
-                to_info,
-                size=size,
-                desc=name,
-                no_progress_bar=no_progress_bar,
-            )
-        self.fs.invalidate_cache(self._with_bucket(to_info.parent))
-
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
-    ):
-        total = self.getsize(self._with_bucket(from_info))
-        with self.open(from_info, "rb") as fobj:
-            with Tqdm.wrapattr(
-                fobj,
-                "read",
-                desc=name,
-                disable=no_progress_bar,
-                bytes=True,
-                total=total,
-                **pbar_args,
-            ) as wrapped:
-                with open(to_file, "wb") as fdest:
-                    shutil.copyfileobj(wrapped, fdest, length=fobj.blocksize)
-
 
 # pylint: disable=abstract-method
 class ObjectFSWrapper(FSSpecWrapper):
@@ -230,42 +217,62 @@ class ObjectFSWrapper(FSSpecWrapper):
         yield from self._strip_buckets(files, detail=detail)
 
 
+# pylint: disable=arguments-differ
+class NoDirectoriesMixin:
+    def isdir(self, *args, **kwargs):
+        return False
+
+    def isfile(self, *args, **kwargs):
+        return True
+
+    def find(self, *args, **kwargs):
+        raise NotImplementedError
+
+    def walk(self, *args, **kwargs):
+        raise NotImplementedError
+
+    def walk_files(self, *args, **kwargs):
+        raise NotImplementedError
+
+    def ls(self, *args, **kwargs):
+        raise NotImplementedError
+
+
 _LOCAL_FS = LocalFileSystem()
 
 
 class CallbackMixin:
-    """Use the native ``get_file()``/``put_file()`` APIs
-    if the target filesystem supports callbacks."""
+    """Provides callback support for the filesystem that don't support yet."""
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **pbar_args
+    def put_file(
+        self,
+        from_file,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        **kwargs,
     ):
-        with Tqdm(
-            desc=name,
-            disable=no_progress_bar,
-            bytes=True,
-            total=-1,
-            **pbar_args,
-        ) as pbar:
-            self.fs.put_file(
-                os.fspath(from_file),
-                self._with_bucket(to_info),
-                callback=pbar.as_callback(_LOCAL_FS, from_file),
-            )
-        self.fs.invalidate_cache(self._with_bucket(to_info.parent))
+        """Add compatibility support for Callback."""
+        # pylint: disable=protected-access
+        self.makedirs(to_info.parent)
+        size = os.path.getsize(from_file)
+        with open(from_file, "rb") as fobj:
+            callback.set_size(size)
+            wrapped = CallbackIOWrapper(callback.relative_update, fobj)
+            self.upload_fobj(wrapped, to_info)
+            self.fs.invalidate_cache(self._with_bucket(to_info.parent))
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
+    def get_file(
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        **kwargs,
     ):
-        with Tqdm(
-            desc=name,
-            disable=no_progress_bar,
-            bytes=True,
-            total=-1,
-            **pbar_args,
-        ) as pbar:
-            self.fs.get_file(
-                self._with_bucket(from_info),
-                os.fspath(to_file),
-                callback=pbar.as_callback(self, from_info),
-            )
+        # pylint: disable=protected-access
+        total: int = self.getsize(from_info)
+        if total:
+            callback.set_size(total)
+
+        with self.open(from_info, "rb") as fobj, open(to_info, "wb") as fdest:
+            wrapped = CallbackIOWrapper(callback.relative_update, fobj)
+            shutil.copyfileobj(wrapped, fdest, length=fobj.blocksize)

dvc/fs/gdrive.py

@@ -12,7 +12,7 @@ from dvc.path_info import CloudURLInfo
 from dvc.scheme import Schemes
 from dvc.utils import format_link, tmp_fname
 
-from .fsspec_wrapper import CallbackMixin, FSSpecWrapper
+from .fsspec_wrapper import FSSpecWrapper
 
 logger = logging.getLogger(__name__)
 FOLDER_MIME_TYPE = "application/vnd.google-apps.folder"
@@ -79,9 +79,7 @@ class GDriveURLInfo(CloudURLInfo):
         self._spath = re.sub("/{2,}", "/", self._spath.rstrip("/"))
 
 
-class GDriveFileSystem(
-    CallbackMixin, FSSpecWrapper
-):  # pylint:disable=abstract-method
+class GDriveFileSystem(FSSpecWrapper):  # pylint:disable=abstract-method
     scheme = Schemes.GDRIVE
     PATH_CLS = GDriveURLInfo
     PARAM_CHECKSUM = "checksum"
@@ -298,7 +296,16 @@ class GDriveFileSystem(
 
         return super()._with_bucket(path)
 
-    def _upload_fobj(self, fobj, to_info, size: int = None):
+    def _strip_bucket(self, entry):
+        try:
+            bucket, path = entry.split("/", 1)
+        except ValueError:
+            # If there is no path attached, only returns
+            # the bucket (top-level).
+            bucket, path = entry, None
+        return path or bucket
+
+    def upload_fobj(self, fobj, to_info, **kwargs):
         rpath = self._with_bucket(to_info)
         self.makedirs(os.path.dirname(rpath))
-        return self.fs.upload_fobj(fobj, rpath, size=size)
+        return self.fs.upload_fobj(fobj, rpath, **kwargs)

dvc/fs/git.py

@@ -1,9 +1,11 @@
 import errno
 import os
-import stat
+
+from tqdm.utils import CallbackIOWrapper
 
 from dvc.utils import is_exec, relpath
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 
 
@@ -96,30 +98,20 @@ class GitFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
         if not self.exists(path_info):
             return False
 
-        mode = self.stat(path_info).st_mode
+        mode = self.info(path_info)["mode"]
         return is_exec(mode)
 
     def info(self, path_info):
         key = self._get_key(path_info)
         try:
-            st = self.trie.stat(key)
+            return self.trie.info(key)
         except KeyError:
             raise FileNotFoundError(
                 errno.ENOENT, os.strerror(errno.ENOENT), path_info
             )
-        return {
-            "size": st.st_size,
-            "type": "directory" if stat.S_ISDIR(st.st_mode) else "file",
-        }
 
-    def stat(self, path):
-        key = self._get_key(path)
-        try:
-            return self.trie.stat(key)
-        except KeyError:
-            raise FileNotFoundError(
-                errno.ENOENT, os.strerror(errno.ENOENT), path
-            )
+    def checksum(self, path_info):
+        return self.info(path_info)["sha"]
 
     def walk_files(self, path_info, **kwargs):
         for root, _, files in self.walk(path_info, **kwargs):
@@ -127,16 +119,18 @@ class GitFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
                 # NOTE: os.path.join is ~5.5 times slower
                 yield f"{root}{os.sep}{file}"
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **kwargs
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
         import shutil
 
-        from dvc.progress import Tqdm
+        total = self.getsize(from_info)
+        if total:
+            callback.set_size(total)
 
-        with open(to_file, "wb+") as to_fobj:
-            with Tqdm.wrapattr(
-                to_fobj, "write", desc=name, disable=no_progress_bar
-            ) as wrapped:
-                with self.open(from_info, "rb", **kwargs) as from_fobj:
-                    shutil.copyfileobj(from_fobj, wrapped)
+        with self.open(from_info, "rb", **kwargs) as from_fobj:
+            with open(to_file, "wb+") as to_fobj:
+                wrapped = CallbackIOWrapper(
+                    callback.relative_update, from_fobj
+                )
+                shutil.copyfileobj(wrapped, to_fobj)

dvc/fs/gs.py

@@ -6,11 +6,11 @@ from funcy import cached_property, wrap_prop
 from dvc.path_info import CloudURLInfo
 from dvc.scheme import Schemes
 
-from .fsspec_wrapper import ObjectFSWrapper
-
-
 # pylint:disable=abstract-method
-class GSFileSystem(ObjectFSWrapper):
+from .fsspec_wrapper import CallbackMixin, ObjectFSWrapper
+
+
+class GSFileSystem(CallbackMixin, ObjectFSWrapper):
     scheme = Schemes.GS
     PATH_CLS = CloudURLInfo
     REQUIRES = {"gcsfs": "gcsfs"}

dvc/fs/hdfs.py

@@ -8,8 +8,10 @@ import subprocess
 from collections import deque
 from contextlib import closing, contextmanager
 
+from tqdm.utils import CallbackIOWrapper
+
 from dvc.hash_info import HashInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK
 from dvc.scheme import Schemes
 from dvc.utils import fix_env, tmp_fname
 
@@ -65,6 +67,8 @@ class HDFSFileSystem(BaseFileSystem):
     PARAM_CHECKSUM = "checksum"
     TRAVERSE_PREFIX_LEN = 2
 
+    BLOCK_SIZE = 2 ** 26  # 128MB
+
     def __init__(self, **config):
         super().__init__(**config)
 
@@ -242,42 +246,37 @@ class HDFSFileSystem(BaseFileSystem):
             size=self.getsize(path_info),
         )
 
-    def _upload_fobj(self, fobj, to_info, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         with self.hdfs(to_info) as hdfs:
             with hdfs.open_output_stream(to_info.path) as fdest:
-                shutil.copyfileobj(fobj, fdest)
+                shutil.copyfileobj(fobj, fdest, self.BLOCK_SIZE)
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         with self.hdfs(to_info) as hdfs:
+            hdfs.create_dir(to_info.parent.path)
+
             tmp_file = tmp_fname(to_info.path)
-            total = os.path.getsize(from_file)
+            total: int = os.path.getsize(from_file)
+            callback.set_size(total)
+
             with open(from_file, "rb") as fobj:
-                with Tqdm.wrapattr(
-                    fobj,
-                    "read",
-                    desc=name,
-                    total=total,
-                    disable=no_progress_bar,
-                ) as wrapped:
-                    with hdfs.open_output_stream(tmp_file) as sobj:
-                        sobj.write(wrapped.read())
+                wrapped = CallbackIOWrapper(callback.relative_update, fobj)
+                with hdfs.open_output_stream(tmp_file) as sobj:
+                    shutil.copyfileobj(wrapped, sobj, self.BLOCK_SIZE)
             hdfs.move(tmp_file, to_info.path)
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **_kwargs
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
         with self.hdfs(from_info) as hdfs:
             file_info = hdfs.get_file_info(from_info.path)
             total = file_info.size
-            with open(to_file, "wb+") as fobj:
-                with Tqdm.wrapattr(
-                    fobj,
-                    "write",
-                    desc=name,
-                    total=total,
-                    disable=no_progress_bar,
-                ) as wrapped:
-                    with hdfs.open_input_stream(from_info.path) as sobj:
-                        wrapped.write(sobj.read())
+            if total:
+                callback.set_size(total)
+
+            with hdfs.open_input_stream(from_info.path) as sobj:
+                with open(to_file, "wb+") as fobj:
+                    wrapped = CallbackIOWrapper(callback.relative_update, sobj)
+                    shutil.copyfileobj(wrapped, fobj, self.BLOCK_SIZE)

dvc/fs/http.py

@@ -1,20 +1,12 @@
-import logging
-import os.path
 import threading
-from typing import Optional
-from urllib.parse import urlparse
 
-from funcy import cached_property, memoize, wrap_prop, wrap_with
+from funcy import cached_property, memoize, wrap_with
 
 from dvc import prompt
-from dvc.exceptions import DvcException, HTTPError
 from dvc.path_info import HTTPURLInfo
-from dvc.progress import Tqdm
 from dvc.scheme import Schemes
 
-from .base import BaseFileSystem
-
-logger = logging.getLogger(__name__)
+from .fsspec_wrapper import FSSpecWrapper, NoDirectoriesMixin
 
 
 @wrap_with(threading.Lock())
@@ -26,171 +18,111 @@ def ask_password(host, user):
     )
 
 
-class HTTPFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
+def make_context(ssl_verify):
+    if isinstance(ssl_verify, bool) or ssl_verify is None:
+        return ssl_verify
+
+    # If this is a path, then we will create an
+    # SSL context for it, and load the given certificate.
+    import ssl
+
+    context = ssl.create_default_context()
+    context.load_verify_locations(ssl_verify)
+    return context
+
+
+# pylint: disable=abstract-method
+class HTTPFileSystem(NoDirectoriesMixin, FSSpecWrapper):
     scheme = Schemes.HTTP
     PATH_CLS = HTTPURLInfo
-    PARAM_CHECKSUM = "etag"
+    PARAM_CHECKSUM = "checksum"
+    REQUIRES = {"aiohttp": "aiohttp", "aiohttp-retry": "aiohttp_retry"}
     CAN_TRAVERSE = False
-    REQUIRES = {"requests": "requests"}
 
     SESSION_RETRIES = 5
     SESSION_BACKOFF_FACTOR = 0.1
     REQUEST_TIMEOUT = 60
-    CHUNK_SIZE = 2 ** 16
 
-    def __init__(self, **config):
-        super().__init__(**config)
+    def _prepare_credentials(self, **config):
+        import aiohttp
+        from fsspec.asyn import fsspec_loop
 
-        self.user = config.get("user", None)
+        from dvc.config import ConfigError
 
-        self.auth = config.get("auth", None)
-        self.custom_auth_header = config.get("custom_auth_header", None)
-        self.password = config.get("password", None)
-        self.ask_password = config.get("ask_password", False)
-        self.headers = {}
-        self.ssl_verify = config.get("ssl_verify", True)
-        self.method = config.get("method", "POST")
+        credentials = {}
+        client_kwargs = credentials.setdefault("client_kwargs", {})
 
-    def _auth_method(self, url):
-        from requests.auth import HTTPBasicAuth, HTTPDigestAuth
+        if config.get("auth"):
+            user = config.get("user")
+            password = config.get("password")
+            custom_auth_header = config.get("custom_auth_header")
 
-        if self.auth:
-            if self.ask_password and self.password is None:
-                self.password = ask_password(urlparse(url).hostname, self.user)
-            if self.auth == "basic":
-                return HTTPBasicAuth(self.user, self.password)
-            if self.auth == "digest":
-                return HTTPDigestAuth(self.user, self.password)
-            if self.auth == "custom" and self.custom_auth_header:
-                self.headers.update({self.custom_auth_header: self.password})
-        return None
+            if password is None and config.get("ask_password"):
+                password = ask_password(config.get("url"), user or "custom")
 
-    def _generate_download_url(self, path_info):
-        return path_info.url
+            auth_method = config["auth"]
+            if auth_method == "basic":
+                if user is None or password is None:
+                    raise ConfigError(
+                        "HTTP 'basic' authentication require both "
+                        "'user' and 'password'"
+                    )
 
-    @wrap_prop(threading.Lock())
-    @cached_property
-    def _session(self):
-        import requests
-        from requests.adapters import HTTPAdapter
-        from urllib3.util.retry import Retry
+                client_kwargs["auth"] = aiohttp.BasicAuth(user, password)
+            elif auth_method == "custom":
+                if custom_auth_header is None or password is None:
+                    raise ConfigError(
+                        "HTTP 'custom' authentication require both "
+                        "'custom_auth_header' and 'password'"
+                    )
+                credentials["headers"] = {custom_auth_header: password}
+            else:
+                raise NotImplementedError(
+                    f"Auth method {auth_method!r} is not supported."
+                )
 
-        session = requests.Session()
+        if "ssl_verify" in config:
+            with fsspec_loop():
+                client_kwargs["connector"] = aiohttp.TCPConnector(
+                    ssl=make_context(config["ssl_verify"])
+                )
 
-        session.verify = self.ssl_verify
+        credentials["get_client"] = self.get_client
+        self.upload_method = config.get("method", "POST")
+        return credentials
 
-        retries = Retry(
-            total=self.SESSION_RETRIES,
-            backoff_factor=self.SESSION_BACKOFF_FACTOR,
+    async def get_client(self, **kwargs):
+        import aiohttp
+        from aiohttp_retry import ExponentialRetry, RetryClient
+
+        kwargs["retry_options"] = ExponentialRetry(
+            attempts=self.SESSION_RETRIES,
+            factor=self.SESSION_BACKOFF_FACTOR,
+            max_timeout=self.REQUEST_TIMEOUT,
         )
 
-        session.mount("http://", HTTPAdapter(max_retries=retries))
-        session.mount("https://", HTTPAdapter(max_retries=retries))
+        # The default timeout for the aiohttp is 300 seconds
+        # which is too low for DVC's interactions (especially
+        # on the read) when dealing with large data blobs. We
+        # unlimit the total time to read, and only limit the
+        # time that is spent when connecting to the remote server.
+        kwargs["timeout"] = aiohttp.ClientTimeout(
+            total=None,
+            connect=self.REQUEST_TIMEOUT,
+            sock_connect=self.REQUEST_TIMEOUT,
+            sock_read=None,
+        )
 
-        return session
+        return RetryClient(**kwargs)
 
-    def request(self, method, url, **kwargs):
-        import requests
+    @cached_property
+    def fs(self):
+        from fsspec.implementations.http import (
+            HTTPFileSystem as _HTTPFileSystem,
+        )
 
-        kwargs.setdefault("allow_redirects", True)
-        kwargs.setdefault("timeout", self.REQUEST_TIMEOUT)
+        return _HTTPFileSystem(**self.fs_args)
 
-        try:
-            res = self._session.request(
-                method,
-                url,
-                auth=self._auth_method(url),
-                headers=self.headers,
-                **kwargs,
-            )
-
-            redirect_no_location = (
-                kwargs["allow_redirects"]
-                and res.status_code in (301, 302)
-                and "location" not in res.headers
-            )
-
-            if redirect_no_location:
-                # AWS s3 doesn't like to add a location header to its redirects
-                # from https://s3.amazonaws.com/<bucket name>/* type URLs.
-                # This should be treated as an error
-                raise requests.exceptions.RequestException
-
-            return res
-
-        except requests.exceptions.RequestException:
-            raise DvcException(f"could not perform a {method} request")
-
-    def _head(self, url):
-        response = self.request("HEAD", url)
-        if response.ok:
-            return response
-
-        # Sometimes servers are configured to forbid HEAD requests
-        # Context: https://github.com/iterative/dvc/issues/4131
-        with self.request("GET", url, stream=True) as r:
-            if r.ok:
-                return r
-
-        return response
-
-    def exists(self, path_info) -> bool:
-        res = self._head(path_info.url)
-        if res.status_code == 404:
-            return False
-        if bool(res):
-            return True
-        raise HTTPError(res.status_code, res.reason)
-
-    def info(self, path_info):
-        resp = self._head(path_info.url)
-        etag = resp.headers.get("ETag") or resp.headers.get("Content-MD5")
-        size = self._content_length(resp)
-        return {"etag": etag, "size": size, "type": "file"}
-
-    def _upload_fobj(self, fobj, to_info, **kwargs):
-        def chunks(fobj):
-            while True:
-                chunk = fobj.read(self.CHUNK_SIZE)
-                if not chunk:
-                    break
-                yield chunk
-
-        response = self.request(self.method, to_info.url, data=chunks(fobj))
-        if response.status_code not in (200, 201):
-            raise HTTPError(response.status_code, response.reason)
-
-    def _download(self, from_info, to_file, name=None, no_progress_bar=False):
-        response = self.request("GET", from_info.url, stream=True)
-        if response.status_code != 200:
-            raise HTTPError(response.status_code, response.reason)
-        with open(to_file, "wb") as fd:
-            with Tqdm.wrapattr(
-                fd,
-                "write",
-                total=None
-                if no_progress_bar
-                else self._content_length(response),
-                leave=False,
-                desc=from_info.url if name is None else name,
-                disable=no_progress_bar,
-            ) as fd_wrapped:
-                for chunk in response.iter_content(chunk_size=self.CHUNK_SIZE):
-                    fd_wrapped.write(chunk)
-
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
-    ):
-        with open(from_file, "rb") as fobj:
-            self.upload_fobj(
-                fobj,
-                to_info,
-                size=None if no_progress_bar else os.path.getsize(from_file),
-                no_progress_bar=no_progress_bar,
-                desc=name or to_info.url,
-            )
-
-    @staticmethod
-    def _content_length(response) -> Optional[int]:
-        res = response.headers.get("Content-Length")
-        return int(res) if res else None
+    def _entry_hook(self, entry):
+        entry["checksum"] = entry.get("ETag") or entry.get("Content-MD5")
+        return entry

dvc/fs/local.py

@@ -1,6 +1,5 @@
 import logging
 import os
-import stat
 
 from dvc.path_info import PathInfo
 from dvc.scheme import Schemes
@@ -8,6 +7,7 @@ from dvc.system import System
 from dvc.utils import is_exec, tmp_fname
 from dvc.utils.fs import copy_fobj_to_file, copyfile, makedirs, move, remove
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 
 logger = logging.getLogger(__name__)
@@ -34,7 +34,12 @@ class LocalFileSystem(BaseFileSystem):
 
     def exists(self, path_info) -> bool:
         assert isinstance(path_info, str) or path_info.scheme == "local"
-        return self.fs.exists(path_info)
+        # TODO: replace this with os.path.exists once the problem is fixed on
+        # the fsspec https://github.com/intake/filesystem_spec/issues/742
+        return os.path.lexists(path_info)
+
+    def checksum(self, path_info) -> str:
+        return self.fs.checksum(path_info)
 
     def isfile(self, path_info) -> bool:
         return os.path.isfile(path_info)
@@ -83,12 +88,9 @@ class LocalFileSystem(BaseFileSystem):
         makedirs(path_info, exist_ok=kwargs.pop("exist_ok", True))
 
     def isexec(self, path_info):
-        mode = self.stat(path_info).st_mode
+        mode = self.info(path_info)["mode"]
         return is_exec(mode)
 
-    def stat(self, path):
-        return os.stat(path)
-
     def move(self, from_info, to_info):
         if (
             isinstance(from_info, PathInfo) and from_info.scheme != "local"
@@ -107,7 +109,7 @@ class LocalFileSystem(BaseFileSystem):
             self.remove(tmp_info)
             raise
 
-    def _upload_fobj(self, fobj, to_info, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         self.makedirs(to_info.parent)
         tmp_info = to_info.parent / tmp_fname("")
         try:
@@ -154,29 +156,18 @@ class LocalFileSystem(BaseFileSystem):
     def reflink(self, from_info, to_info):
         System.reflink(from_info, to_info)
 
-    @staticmethod
-    def info(path_info):
-        st = os.stat(path_info)
-        return {
-            "size": st.st_size,
-            "type": "directory" if stat.S_ISDIR(st.st_mode) else "file",
-        }
+    def info(self, path_info):
+        return self.fs.info(path_info)
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         makedirs(to_info.parent, exist_ok=True)
-
         tmp_file = tmp_fname(to_info)
-        copyfile(
-            from_file, tmp_file, name=name, no_progress_bar=no_progress_bar
-        )
+        copyfile(from_file, tmp_file, callback=callback)
         os.replace(tmp_file, to_info)
 
-    @staticmethod
-    def _download(
-        from_info, to_file, name=None, no_progress_bar=False, **_kwargs
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
-        copyfile(
-            from_info, to_file, no_progress_bar=no_progress_bar, name=name
-        )
+        copyfile(from_info, to_file, callback=callback)

dvc/fs/oss.py

@@ -7,13 +7,13 @@ from funcy import cached_property, wrap_prop
 from dvc.path_info import CloudURLInfo
 from dvc.scheme import Schemes
 
-from .fsspec_wrapper import CallbackMixin, ObjectFSWrapper
+from .fsspec_wrapper import ObjectFSWrapper
 
 logger = logging.getLogger(__name__)
 
 
 # pylint:disable=abstract-method
-class OSSFileSystem(CallbackMixin, ObjectFSWrapper):
+class OSSFileSystem(ObjectFSWrapper):
     scheme = Schemes.OSS
     PATH_CLS = CloudURLInfo
     REQUIRES = {"ossfs": "ossfs"}

dvc/fs/repo.py

@@ -1,6 +1,5 @@
 import logging
 import os
-import stat
 import threading
 from contextlib import suppress
 from itertools import takewhile
@@ -10,6 +9,7 @@ from funcy import lfilter, wrap_with
 
 from dvc.path_info import PathInfo
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 from .dvc import DvcFileSystem
 
@@ -256,8 +256,8 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             return False
 
         try:
-            st = fs.stat(path)
-            return stat.S_ISDIR(st.st_mode)
+            info = fs.info(path)
+            return info["type"] == "directory"
         except (OSError, ValueError):
             # from CPython's os.path.isdir()
             pass
@@ -287,8 +287,8 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             return False
 
         try:
-            st = fs.stat(path)
-            return stat.S_ISREG(st.st_mode)
+            info = fs.info(path)
+            return info["type"] == "file"
         except (OSError, ValueError):
             # from CPython's os.path.isfile()
             pass
@@ -313,10 +313,6 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             return dvc_fs.isexec(path_info)
         return fs.isexec(path_info)
 
-    def stat(self, path):
-        fs, _ = self._get_fs_pair(path)
-        return fs.stat(path)
-
     def _dvc_walk(self, walk):
         try:
             root, dirs, files = next(walk)
@@ -453,19 +449,21 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             for fname in files:
                 yield PathInfo(root) / fname
 
-    def _download(self, from_info, to_file, **kwargs):
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         fs, dvc_fs = self._get_fs_pair(from_info)
         try:
-            fs._download(  # pylint: disable=protected-access
-                from_info, to_file, **kwargs
+            fs.get_file(  # pylint: disable=protected-access
+                from_info, to_file, callback=callback, **kwargs
             )
             return
         except FileNotFoundError:
             if not dvc_fs:
                 raise
 
-        dvc_fs._download(  # pylint: disable=protected-access
-            from_info, to_file, **kwargs
+        dvc_fs.get_file(  # pylint: disable=protected-access
+            from_info, to_file, callback=callback, **kwargs
         )
 
     def metadata(self, path):
@@ -478,11 +476,11 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             with suppress(FileNotFoundError):
                 dvc_meta = dvc_fs.metadata(path_info)
 
-        stat_result = None
+        info_result = None
         with suppress(FileNotFoundError):
-            stat_result = fs.stat(path_info)
+            info_result = fs.info(path_info)
 
-        if not stat_result and not dvc_meta:
+        if not info_result and not dvc_meta:
             raise FileNotFoundError
 
         from ._metadata import Metadata
@@ -492,13 +490,13 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             repo=self._get_repo(abspath) or self._main_repo,
         )
 
-        isdir = bool(stat_result) and stat.S_ISDIR(stat_result.st_mode)
+        isdir = bool(info_result) and info_result["type"] == "directory"
         meta.isdir = meta.isdir or isdir
 
         if not dvc_meta:
             from dvc.utils import is_exec
 
-            meta.is_exec = bool(stat_result) and is_exec(stat_result.st_mode)
+            meta.is_exec = bool(info_result) and is_exec(info_result["mode"])
         return meta
 
     def info(self, path_info):
@@ -508,3 +506,11 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             return fs.info(path_info)
         except FileNotFoundError:
             return dvc_fs.info(path_info)
+
+    def checksum(self, path_info):
+        fs, dvc_fs = self._get_fs_pair(path_info)
+
+        try:
+            return fs.checksum(path_info)
+        except FileNotFoundError:
+            return dvc_fs.checksum(path_info)

dvc/fs/s3.py

@@ -6,7 +6,7 @@ from collections import defaultdict
 from funcy import cached_property, wrap_prop
 
 from dvc.path_info import CloudURLInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK
 from dvc.scheme import Schemes
 
 from .fsspec_wrapper import ObjectFSWrapper
@@ -68,12 +68,12 @@ class BaseS3FileSystem(ObjectFSWrapper):
         self._transfer_config = None
         config_path = os.environ.get("AWS_CONFIG_FILE", _AWS_CONFIG_PATH)
         if not os.path.exists(config_path):
-            return None
+            return {}
 
         config = load_config(config_path)
         profile_config = config["profiles"].get(profile or "default")
         if not profile_config:
-            return None
+            return {}
 
         s3_config = profile_config.get("s3", {})
         return self._split_s3_config(s3_config)
@@ -123,6 +123,14 @@ class BaseS3FileSystem(ObjectFSWrapper):
         if shared_creds:
             os.environ.setdefault("AWS_SHARED_CREDENTIALS_FILE", shared_creds)
 
+        if (
+            client["region_name"] is None
+            and session_config["s3"].get("region_name") is None
+            and os.getenv("AWS_REGION") is None
+        ):
+            # Enable bucket region caching
+            login_info["cache_regions"] = config.get("cache_regions", True)
+
         config_path = config.get("configpath")
         if config_path:
             os.environ.setdefault("AWS_CONFIG_FILE", config_path)
@@ -196,36 +204,23 @@ class S3FileSystem(BaseS3FileSystem):  # pylint:disable=abstract-method
         return bucket.Object(path_info.path)
 
     @_translate_exceptions
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **pbar_args
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
-        total = os.path.getsize(from_file)
-        with Tqdm(
-            disable=no_progress_bar,
-            total=total,
-            bytes=True,
-            desc=name,
-            **pbar_args,
-        ) as pbar:
-            obj = self._get_obj(to_info)
-            obj.upload_file(
-                from_file,
-                Callback=pbar.update,
-                ExtraArgs=self.fs_args.get("s3_additional_kwargs"),
-                Config=self._transfer_config,
-            )
+        callback.set_size(os.path.getsize(from_file))
+        obj = self._get_obj(to_info)
+        obj.upload_file(
+            from_file,
+            Callback=callback.relative_update,
+            ExtraArgs=self.fs_args.get("s3_additional_kwargs"),
+            Config=self._transfer_config,
+        )
         self.fs.invalidate_cache(self._with_bucket(to_info.parent))
 
     @_translate_exceptions
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
+    def get_file(
+        self, from_info, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         obj = self._get_obj(from_info)
-        with Tqdm(
-            disable=no_progress_bar,
-            total=obj.content_length,
-            bytes=True,
-            desc=name,
-            **pbar_args,
-        ) as pbar:
-            obj.download_file(to_file, Callback=pbar.update)
+        callback.set_size(obj.content_length)
+        obj.download_file(to_info, Callback=callback.relative_update)

dvc/fs/ssh.py

@@ -2,13 +2,14 @@ import getpass
 import os.path
 import threading
 
-from funcy import cached_property, first, memoize, silent, wrap_prop, wrap_with
+from funcy import cached_property, memoize, silent, wrap_prop, wrap_with
 
 from dvc import prompt
 from dvc.scheme import Schemes
 from dvc.utils.fs import as_atomic
 
-from .fsspec_wrapper import CallbackMixin, FSSpecWrapper
+from ..progress import DEFAULT_CALLBACK
+from .fsspec_wrapper import FSSpecWrapper
 
 _SSH_TIMEOUT = 60 * 30
 _SSH_CONFIG_FILE = os.path.expanduser(os.path.join("~", ".ssh", "config"))
@@ -26,7 +27,7 @@ def ask_password(host, user, port):
 
 
 # pylint:disable=abstract-method
-class SSHFileSystem(CallbackMixin, FSSpecWrapper):
+class SSHFileSystem(FSSpecWrapper):
     scheme = Schemes.SSH
     REQUIRES = {"sshfs": "sshfs"}
 
@@ -66,16 +67,25 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
             or self.DEFAULT_PORT
         )
 
-        login_info["password"] = config.get("password")
-
-        if user_ssh_config.get("IdentityFile"):
-            config.setdefault(
-                "keyfile", first(user_ssh_config.get("IdentityFile"))
+        if config.get("ask_password") and config.get("password") is None:
+            config["password"] = ask_password(
+                login_info["host"], login_info["username"], login_info["port"]
             )
 
-        keyfile = config.get("keyfile")
-        if keyfile:
-            login_info["client_keys"] = [os.path.expanduser(keyfile)]
+        login_info["password"] = config.get("password")
+        login_info["passphrase"] = config.get("password")
+
+        raw_keys = []
+        if config.get("keyfile"):
+            raw_keys.append(config.get("keyfile"))
+        elif user_ssh_config.get("IdentityFile"):
+            raw_keys.extend(user_ssh_config.get("IdentityFile"))
+
+        if raw_keys:
+            login_info["client_keys"] = [
+                os.path.expanduser(key) for key in raw_keys
+            ]
+
         login_info["timeout"] = config.get("timeout", _SSH_TIMEOUT)
 
         # These two settings fine tune the asyncssh to use the
@@ -96,11 +106,6 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
         login_info["agent_forwarding"] = config.get("agent_forwarding", True)
         login_info["proxy_command"] = user_ssh_config.get("ProxyCommand")
 
-        if config.get("ask_password") and login_info["password"] is None:
-            login_info["password"] = ask_password(
-                login_info["host"], login_info["username"], login_info["port"]
-            )
-
         # We are going to automatically add stuff to known_hosts
         # something like paramiko's AutoAddPolicy()
         login_info["known_hosts"] = None
@@ -116,10 +121,12 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
     # Ensure that if an interrupt happens during the transfer, we don't
     # pollute the cache.
 
-    def _upload_fobj(self, fobj, to_info, *args, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         with as_atomic(self, to_info) as tmp_file:
-            super()._upload_fobj(fobj, tmp_file, *args, **kwargs)
+            super().upload_fobj(fobj, tmp_file, **kwargs)
 
-    def _upload(self, from_file, to_info, *args, **kwargs):
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         with as_atomic(self, to_info) as tmp_file:
-            super()._upload(from_file, tmp_file, *args, **kwargs)
+            super().put_file(from_file, tmp_file, callback=callback, **kwargs)

dvc/fs/utils.py

@@ -0,0 +1,56 @@
+import logging
+import os
+from typing import TYPE_CHECKING
+
+from .local import LocalFileSystem
+
+if TYPE_CHECKING:
+    from dvc.types import AnyPath, DvcPath
+
+    from .base import BaseFileSystem
+
+logger = logging.getLogger(__name__)
+
+
+def transfer(
+    from_fs: "BaseFileSystem",
+    from_info: "AnyPath",
+    to_fs: "BaseFileSystem",
+    to_info: "DvcPath",
+    move: bool = False,
+) -> None:
+    use_move = isinstance(from_fs, type(to_fs)) and move
+    try:
+        if use_move:
+            return to_fs.move(from_info, to_info)
+
+        if isinstance(from_fs, LocalFileSystem):
+            if not isinstance(from_info, from_fs.PATH_CLS):
+                from_info = from_fs.PATH_CLS(from_info)
+            return to_fs.upload(from_info, to_info)
+
+        if isinstance(to_fs, LocalFileSystem):
+            return from_fs.download_file(from_info, to_info)
+
+        with from_fs.open(from_info, mode="rb") as fobj:
+            size = from_fs.getsize(from_info)
+            return to_fs.upload(fobj, to_info, total=size)
+    except OSError as exc:
+        # If the target file already exists, we are going to simply
+        # ignore the exception (#4992).
+        #
+        # On Windows, it is not always guaranteed that you'll get
+        # FileExistsError (it might be PermissionError or a bare OSError)
+        # but all of those exceptions raised from the original
+        # FileExistsError so we have a separate check for that.
+        if isinstance(exc, FileExistsError) or (
+            os.name == "nt"
+            and exc.__context__
+            and isinstance(exc.__context__, FileExistsError)
+        ):
+            logger.debug("'%s' file already exists, skipping", to_info)
+            if use_move:
+                from_fs.remove(from_info)
+            return None
+
+        raise

dvc/fs/webdav.py

@@ -1,5 +1,4 @@
 import logging
-import os
 import threading
 from functools import lru_cache
 
@@ -71,15 +70,13 @@ class WebDAVFileSystem(FSSpecWrapper):  # pylint:disable=abstract-method
 
         return WebdavFileSystem(**self.fs_args)
 
-    def _upload_fobj(self, fobj, to_info, size: int = None):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         rpath = self.translate_path_info(to_info)
-        self.makedirs(os.path.dirname(rpath))
+        size = kwargs.get("size")
         # using upload_fileobj to directly upload fileobj rather than buffering
         # and using overwrite=True to avoid check for an extra exists call,
         # as caller should ensure that the file does not exist beforehand.
-        return self.fs.client.upload_fileobj(
-            fobj, rpath, overwrite=True, size=size
-        )
+        return self.fs.upload_fileobj(fobj, rpath, overwrite=True, size=size)
 
     @lru_cache(512)
     def translate_path_info(self, path):

dvc/fs/webhdfs.py

@@ -9,7 +9,7 @@ from funcy import cached_property, wrap_prop
 
 from dvc.hash_info import HashInfo
 from dvc.path_info import CloudURLInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK
 from dvc.scheme import Schemes
 
 from .base import BaseFileSystem
@@ -29,6 +29,15 @@ def update_pbar(pbar, total):
     return update
 
 
+def update_callback(callback, total):
+    def update(_, bytes_transfered):
+        if bytes_transfered == -1:
+            return callback.absolute_update(total)
+        return callback.relative_update(bytes_transfered)
+
+    return update
+
+
 class WebHDFSFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
     scheme = Schemes.WEBHDFS
     PATH_CLS = CloudURLInfo
@@ -141,31 +150,31 @@ class WebHDFSFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
         self.hdfs_client.makedirs(to_info.parent.path)
         self.hdfs_client.rename(from_info.path, to_info.path)
 
-    def _upload_fobj(self, fobj, to_info, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         with self.hdfs_client.write(to_info.path) as fdest:
             shutil.copyfileobj(fobj, fdest)
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         total = os.path.getsize(from_file)
-        with Tqdm(
-            desc=name, total=total, disable=no_progress_bar, bytes=True
-        ) as pbar:
-            self.hdfs_client.upload(
-                to_info.path,
-                from_file,
-                overwrite=True,
-                progress=update_pbar(pbar, total),
-            )
+        callback.set_size(total)
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **_kwargs
+        self.hdfs_client.makedirs(to_info.parent.path)
+        return self.hdfs_client.upload(
+            to_info.path,
+            from_file,
+            overwrite=True,
+            progress=update_callback(callback, total),
+        )
+
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
         total = self.getsize(from_info)
-        with Tqdm(
-            desc=name, total=total, disable=no_progress_bar, bytes=True
-        ) as pbar:
-            self.hdfs_client.download(
-                from_info.path, to_file, progress=update_pbar(pbar, total)
-            )
+        if total:
+            callback.set_size(total)
+
+        self.hdfs_client.download(
+            from_info.path, to_file, progress=update_callback(callback, total)
+        )

dvc/machine/__init__.py

@@ -0,0 +1,202 @@
+import logging
+import os
+from typing import (
+    TYPE_CHECKING,
+    Dict,
+    Iterable,
+    Mapping,
+    Optional,
+    Tuple,
+    Type,
+)
+
+from dvc.exceptions import DvcException
+from dvc.types import StrPath
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+    from .backend.base import BaseMachineBackend
+
+    BackendCls = Type[BaseMachineBackend]
+
+logger = logging.getLogger(__name__)
+
+
+RESERVED_NAMES = {"local", "localhost"}
+
+DEFAULT_STARTUP_SCRIPT = """#!/bin/bash
+sudo apt-get update
+sudo apt-get install --yes python3 python3-pip
+python3 -m pip install --upgrade pip
+
+pushd /etc/apt/sources.list.d
+sudo wget https://dvc.org/deb/dvc.list
+sudo apt-get update
+sudo apt-get install --yes dvc
+popd
+"""
+
+
+def validate_name(name: str):
+    from dvc.exceptions import InvalidArgumentError
+
+    name = name.lower()
+    if name in RESERVED_NAMES:
+        raise InvalidArgumentError(
+            f"Machine name '{name}' is reserved for internal DVC use."
+        )
+
+
+class MachineBackends(Mapping):
+    try:
+        from .backend.terraform import TerraformBackend
+    except ImportError:
+        TerraformBackend = None  # type: ignore[assignment, misc]
+
+    DEFAULT: Dict[str, Optional["BackendCls"]] = {
+        "terraform": TerraformBackend,
+    }
+
+    def __getitem__(self, key: str) -> "BaseMachineBackend":
+        """Lazily initialize backends and cache it afterwards"""
+        initialized = self.initialized.get(key)
+        if not initialized:
+            backend = self.backends[key]
+            initialized = backend(
+                os.path.join(self.tmp_dir, key), **self.kwargs
+            )
+            self.initialized[key] = initialized
+        return initialized
+
+    def __init__(
+        self,
+        selected: Optional[Iterable[str]],
+        tmp_dir: StrPath,
+        **kwargs,
+    ) -> None:
+        selected = selected or list(self.DEFAULT)
+        self.backends: Dict[str, "BackendCls"] = {}
+        for key in selected:
+            cls = self.DEFAULT.get(key)
+            if cls is None:
+                raise DvcException(
+                    f"'dvc machine' backend '{key}' is missing required "
+                    "dependencies. Install them with:\n"
+                    f"\tpip install dvc[{key}]"
+                )
+            self.backends[key] = cls
+
+        self.initialized: Dict[str, "BaseMachineBackend"] = {}
+
+        self.tmp_dir = tmp_dir
+        self.kwargs = kwargs
+
+    def __iter__(self):
+        return iter(self.backends)
+
+    def __len__(self) -> int:
+        return len(self.backends)
+
+    def close_initialized(self) -> None:
+        for backend in self.initialized.values():
+            backend.close()
+
+
+class MachineManager:
+    """Class that manages dvc cloud machines.
+
+    Args:
+        repo (dvc.repo.Repo): repo instance that belongs to the repo that
+            we are working on.
+
+    Raises:
+        config.ConfigError: thrown when config has invalid format.
+    """
+
+    CLOUD_BACKENDS = {
+        "aws": "terraform",
+        "azure": "terraform",
+    }
+
+    def __init__(
+        self, repo: "Repo", backends: Optional[Iterable[str]] = None, **kwargs
+    ):
+        self.repo = repo
+        tmp_dir = os.path.join(self.repo.tmp_dir, "machine")
+        self.backends = MachineBackends(backends, tmp_dir=tmp_dir, **kwargs)
+
+    def get_config_and_backend(
+        self,
+        name: Optional[str] = None,
+    ) -> Tuple[dict, "BaseMachineBackend"]:
+        from dvc.config import NoMachineError
+
+        if not name:
+            name = self.repo.config["core"].get("machine")
+
+        if name:
+            config = self._get_config(name=name)
+            backend = self._get_backend(config["cloud"])
+            return config, backend
+
+        if bool(self.repo.config["machine"]):
+            error_msg = (
+                "no machine specified. Setup default machine with\n"
+                "    dvc machine default <name>\n"
+            )
+        else:
+            error_msg = (
+                "no machine specified. Create a default machine with\n"
+                "    dvc machine add -d <name> <cloud>"
+            )
+
+        raise NoMachineError(error_msg)
+
+    def _get_config(self, **kwargs):
+        config = self.repo.config
+        name = kwargs.get("name")
+        if name:
+            try:
+                conf = config["machine"][name.lower()]
+                conf["name"] = name
+            except KeyError:
+                from dvc.config import MachineNotFoundError
+
+                raise MachineNotFoundError(f"machine '{name}' doesn't exist")
+        else:
+            conf = kwargs
+        return conf
+
+    def _get_backend(self, cloud: str) -> "BaseMachineBackend":
+        from dvc.config import NoMachineError
+
+        try:
+            backend = self.CLOUD_BACKENDS[cloud]
+            return self.backends[backend]
+        except KeyError:
+            raise NoMachineError(f"Machine platform '{cloud}' unsupported")
+
+    def create(self, name: Optional[str]):
+        """Create and start the specified machine instance."""
+        config, backend = self.get_config_and_backend(name)
+        if "startup_script" in config:
+            with open(config["startup_script"]) as fobj:
+                startup_script = fobj.read()
+        else:
+            startup_script = DEFAULT_STARTUP_SCRIPT
+        config["startup_script"] = startup_script
+        return backend.create(**config)
+
+    def destroy(self, name: Optional[str]):
+        """Destroy the specified machine instance."""
+        config, backend = self.get_config_and_backend(name)
+        return backend.destroy(**config)
+
+    def get_sshfs(self, name: Optional[str]):
+        config, backend = self.get_config_and_backend(name)
+        return backend.get_sshfs(**config)
+
+    def run_shell(self, name: Optional[str]):
+        config, backend = self.get_config_and_backend(name)
+        return backend.run_shell(**config)

dvc/machine/backend/base.py

@@ -0,0 +1,50 @@
+from abc import ABC, abstractmethod
+from contextlib import contextmanager
+from typing import TYPE_CHECKING, Iterator, Optional
+
+if TYPE_CHECKING:
+    from dvc.fs.ssh import SSHFileSystem
+    from dvc.repo.experiments.executor.base import BaseExecutor
+    from dvc.types import StrPath
+
+
+class BaseMachineBackend(ABC):
+    def __init__(self, tmp_dir: "StrPath", **kwargs):
+        raise NotImplementedError
+
+    @abstractmethod
+    def create(self, name: Optional[str] = None, **config):
+        """Create and start an instance of the specified machine."""
+
+    @abstractmethod
+    def destroy(self, name: Optional[str] = None, **config):
+        """Stop and destroy all instances of the specified machine."""
+
+    @abstractmethod
+    def instances(
+        self, name: Optional[str] = None, **config
+    ) -> Iterator[dict]:
+        """Iterate over status of all instances of the specified machine."""
+
+    def close(self):
+        pass
+
+    @abstractmethod
+    def run_shell(self, name: Optional[str] = None, **config):
+        """Spawn an interactive SSH shell for the specified machine."""
+
+    @abstractmethod
+    def get_executor(
+        self, name: Optional[str] = None, **config
+    ) -> "BaseExecutor":
+        """Return an executor instance which can be used for DVC
+        experiment/pipeline execution on the specified machine.
+        """
+
+    @abstractmethod
+    @contextmanager
+    def get_sshfs(
+        self, name: Optional[str] = None, **config
+    ) -> Iterator["SSHFileSystem"]:
+        """Return an sshfs instance for the default directory on the
+        specified machine."""

dvc/machine/backend/terraform.py

@@ -0,0 +1,35 @@
+from contextlib import contextmanager
+from typing import TYPE_CHECKING, Iterator, Optional
+
+from tpi import TerraformProviderIterative, terraform
+
+from dvc.fs.ssh import SSHFileSystem
+
+from .base import BaseMachineBackend
+
+if TYPE_CHECKING:
+    from dvc.repo.experiments.executor.base import BaseExecutor
+    from dvc.types import StrPath
+
+
+class TerraformBackend(terraform.TerraformBackend, BaseMachineBackend):
+    def __init__(self, tmp_dir: "StrPath", **kwargs):
+        super().__init__(tmp_dir)
+
+    def get_executor(
+        self, name: Optional[str] = None, **config
+    ) -> "BaseExecutor":
+        raise NotImplementedError
+
+    @contextmanager
+    def get_sshfs(  # pylint: disable=unused-argument
+        self, name: Optional[str] = None, **config
+    ) -> Iterator["SSHFileSystem"]:
+        resource = self._default_resource(name)
+        with TerraformProviderIterative.pemfile(resource) as pem:
+            fs = SSHFileSystem(
+                host=resource["instance_ip"],
+                user="ubuntu",
+                keyfile=pem,
+            )
+            yield fs

dvc/objects/db/base.py

@@ -1,12 +1,10 @@
 import itertools
 import logging
-import os
 from concurrent.futures import ThreadPoolExecutor
 from contextlib import suppress
 from copy import copy
 from typing import TYPE_CHECKING, Optional
 
-from dvc.fs.local import LocalFileSystem
 from dvc.objects.errors import ObjectDBPermissionError, ObjectFormatError
 from dvc.objects.file import HashFile
 from dvc.progress import Tqdm
@@ -85,38 +83,12 @@ class ObjectDB:
         _hash_info: "HashInfo",
         move: bool = False,
     ):
+        from dvc import fs
+
         self.makedirs(to_info.parent)
-        use_move = isinstance(from_fs, type(self.fs)) and move
-        try:
-            if use_move:
-                self.fs.move(from_info, to_info)
-            elif isinstance(from_fs, LocalFileSystem):
-                if not isinstance(from_info, from_fs.PATH_CLS):
-                    from_info = from_fs.PATH_CLS(from_info)
-                self.fs.upload(from_info, to_info)
-            elif isinstance(self.fs, LocalFileSystem):
-                from_fs.download_file(from_info, to_info)
-            else:
-                with from_fs.open(from_info, mode="rb") as fobj:
-                    self.fs.upload_fobj(fobj, to_info)
-        except OSError as exc:
-            # If the target file already exists, we are going to simply
-            # ignore the exception (#4992).
-            #
-            # On Windows, it is not always guaranteed that you'll get
-            # FileExistsError (it might be PermissionError or a bare OSError)
-            # but all of those exceptions raised from the original
-            # FileExistsError so we have a separate check for that.
-            if isinstance(exc, FileExistsError) or (
-                os.name == "nt"
-                and exc.__context__
-                and isinstance(exc.__context__, FileExistsError)
-            ):
-                logger.debug("'%s' file already exists, skipping", to_info)
-                if use_move:
-                    from_fs.remove(from_info)
-            else:
-                raise
+        return fs.utils.transfer(
+            from_fs, from_info, self.fs, to_info, move=move
+        )
 
     def add(
         self,
@@ -125,7 +97,6 @@ class ObjectDB:
         hash_info: "HashInfo",
         move: bool = True,
         verify: Optional[bool] = None,
-        **kwargs,
     ):
         if self.read_only:
             raise ObjectDBPermissionError("Cannot add to read-only ODB")
@@ -139,7 +110,7 @@ class ObjectDB:
             pass
 
         cache_info = self.hash_to_path_info(hash_info.value)
-        self._add_file(fs, path_info, cache_info, hash_info, move)
+        self._add_file(fs, path_info, cache_info, hash_info, move=move)
 
         try:
             if verify:
@@ -149,10 +120,6 @@ class ObjectDB:
         except (ObjectFormatError, FileNotFoundError):
             pass
 
-        callback = kwargs.get("download_callback")
-        if callback:
-            callback(1)
-
     def hash_to_path_info(self, hash_) -> "DvcPath":
         return self.path_info / hash_[0:2] / hash_[2:]
 

dvc/objects/db/reference.py

@@ -70,7 +70,7 @@ class ReferenceObjectDB(ObjectDB):
         ref_fobj = io.BytesIO(ref_file.to_bytes())
         ref_fobj.seek(0)
         try:
-            self.fs.upload_fobj(ref_fobj, to_info)
+            self.fs.upload(ref_fobj, to_info)
         except OSError as exc:
             if isinstance(exc, FileExistsError) or (
                 os.name == "nt"

dvc/objects/reference.py

@@ -2,7 +2,7 @@ import errno
 import logging
 import os
 import pickle
-from typing import TYPE_CHECKING, Optional, Tuple
+from typing import TYPE_CHECKING, Optional
 
 from .errors import ObjectFormatError
 from .file import HashFile
@@ -20,8 +20,7 @@ logger = logging.getLogger(__name__)
 class ReferenceHashFile(HashFile):
     PARAM_PATH = "path"
     PARAM_HASH = "hash"
-    PARAM_MTIME = "mtime"
-    PARAM_SIZE = "size"
+    PARAM_CHECKSUM = "checksum"
     PARAM_FS_CONFIG = "fs_config"
 
     def __init__(
@@ -29,14 +28,11 @@ class ReferenceHashFile(HashFile):
         path_info: "AnyPath",
         fs: "BaseFileSystem",
         hash_info: "HashInfo",
-        mtime: Optional[str] = None,
-        size: Optional[int] = None,
+        checksum: Optional[str] = None,
         **kwargs,
     ):
         super().__init__(path_info, fs, hash_info, **kwargs)
-        cur_mtime, cur_size = self._get_mtime_and_size()
-        self.mtime = cur_mtime if mtime is None else mtime
-        self.hash_info.size = cur_size if size is None else size
+        self.checksum = checksum or fs.checksum(path_info)
 
     def __str__(self):
         return f"ref object {self.hash_info} -> {self.path_info}"
@@ -47,21 +43,14 @@ class ReferenceHashFile(HashFile):
             raise FileNotFoundError(
                 errno.ENOENT, os.strerror(errno.ENOENT), self.path_info
             )
-        if not (self.mtime, self.size) == self._get_mtime_and_size():
+        if self.checksum != self._get_checksum():
             raise ObjectFormatError(f"{self} is changed")
         if check_hash:
             self._check_hash(odb)
 
-    def _get_mtime_and_size(self) -> Tuple[Optional[str], int]:
-        from dvc.utils.fs import get_mtime_and_size
-
+    def _get_checksum(self) -> str:
         assert self.fs
-        if hasattr(self.fs, "stat"):
-            try:
-                return get_mtime_and_size(self.path_info, self.fs)
-            except FileNotFoundError:
-                pass
-        return None, self.fs.getsize(self.path_info)
+        return self.fs.checksum(self.path_info)
 
     def to_bytes(self):
         from dvc.fs.repo import RepoFileSystem
@@ -79,8 +68,7 @@ class ReferenceHashFile(HashFile):
         dict_ = {
             self.PARAM_PATH: path_info,
             self.PARAM_HASH: self.hash_info,
-            self.PARAM_MTIME: self.mtime,
-            self.PARAM_SIZE: self.size,
+            self.PARAM_CHECKSUM: self.checksum,
             self.PARAM_FS_CONFIG: self.config_tuple(self.fs),
         }
         try:
@@ -118,8 +106,7 @@ class ReferenceHashFile(HashFile):
             path_info,
             fs,
             hash_info,
-            mtime=dict_.get(cls.PARAM_MTIME),
-            size=dict_.get(cls.PARAM_SIZE),
+            checksum=dict_.get(cls.PARAM_CHECKSUM),
         )
 
     @staticmethod

dvc/objects/stage.py

@@ -34,9 +34,8 @@ def _upload_file(path_info, fs, odb, upload_odb):
     tmp_info = upload_odb.path_info / tmp_fname()
     with fs.open(path_info, mode="rb", chunk_size=fs.CHUNK_SIZE) as stream:
         stream = HashedStreamReader(stream)
-        upload_odb.fs.upload_fobj(
-            stream, tmp_info, desc=path_info.name, size=fs.getsize(path_info)
-        )
+        size = fs.getsize(path_info)
+        upload_odb.fs.upload(stream, tmp_info, desc=path_info.name, total=size)
 
     odb.add(tmp_info, upload_odb.fs, stream.hash_info)
     return path_info, odb.get(stream.hash_info)

dvc/objects/transfer.py

@@ -1,20 +1,19 @@
 import errno
-import itertools
 import logging
-from concurrent import futures
-from concurrent.futures import ThreadPoolExecutor
 from functools import partial, wraps
-from typing import TYPE_CHECKING, Callable, Iterable, Optional
+from typing import TYPE_CHECKING, Any, Callable, Iterable, Optional
 
 from funcy import split
 
 from dvc.progress import Tqdm
+from dvc.utils.threadpool import ThreadPoolExecutor
 
 if TYPE_CHECKING:
     from dvc.hash_info import HashInfo
 
     from .db.base import ObjectDB
     from .db.index import ObjectDBIndexBase
+    from .tree import Tree
 
 logger = logging.getLogger(__name__)
 
@@ -38,109 +37,48 @@ def _log_exceptions(func):
     return wrapper
 
 
-def _transfer(
-    src, dest, dir_ids, file_ids, missing_ids, jobs, verify, move, **kwargs
-):
-    func = _log_exceptions(dest.add)
-    total = len(dir_ids) + len(file_ids)
-    if total == 0:
-        return 0
-    with Tqdm(total=total, unit="file", desc="Transferring") as pbar:
-        func = pbar.wrap_fn(func)
-        with ThreadPoolExecutor(max_workers=jobs) as executor:
-            processor = partial(
-                _create_tasks,
-                executor,
-                jobs,
-                func,
-                src,
-                verify,
-                move,
-            )
-            processor.add_func = func
-            _do_transfer(
-                src,
-                dest,
-                dir_ids,
-                file_ids,
-                missing_ids,
-                processor,
-                verify=verify,
-                move=move,
-                **kwargs,
-            )
-    return total
+def find_tree_by_obj_id(
+    odbs: Iterable[Optional["ObjectDB"]], obj_id: "HashInfo"
+) -> Optional["Tree"]:
+    from .errors import ObjectFormatError
+    from .tree import Tree
 
-
-def _create_tasks(executor, jobs, func, src, verify, move, obj_ids):
-    fails = 0
-    hash_iter = iter(obj_ids)
-
-    def submit(hash_info):
-        obj = src.get(hash_info)
-        return executor.submit(
-            func,
-            obj.path_info,
-            obj.fs,
-            obj.hash_info,
-            verify=verify,
-            move=move,
-        )
-
-    def create_taskset(amount):
-        return {
-            submit(hash_info)
-            for hash_info in itertools.islice(hash_iter, amount)
-        }
-
-    tasks = create_taskset(jobs * 5)
-    while tasks:
-        done, tasks = futures.wait(tasks, return_when=futures.FIRST_COMPLETED)
-        fails += sum(task.result() for task in done)
-        tasks.update(create_taskset(len(done)))
-    return fails
+    for odb in odbs:
+        if odb is not None:
+            try:
+                return Tree.load(odb, obj_id)
+            except (FileNotFoundError, ObjectFormatError):
+                pass
+    return None
 
 
 def _do_transfer(
     src: "ObjectDB",
     dest: "ObjectDB",
-    dir_ids: Iterable["HashInfo"],
-    file_ids: Iterable["HashInfo"],
+    obj_ids: Iterable["HashInfo"],
     missing_ids: Iterable["HashInfo"],
     processor: Callable,
     src_index: Optional["ObjectDBIndexBase"] = None,
     dest_index: Optional["ObjectDBIndexBase"] = None,
     cache_odb: Optional["ObjectDB"] = None,
-    **kwargs,
+    **kwargs: Any,
 ):
     from dvc.exceptions import FileTransferError
-    from dvc.objects.errors import ObjectFormatError
 
+    dir_ids, file_ids = split(lambda hash_info: hash_info.isdir, obj_ids)
     total_fails = 0
     succeeded_dir_objs = []
     all_file_ids = set(file_ids)
 
     for dir_hash in dir_ids:
-        from .tree import Tree
-
-        bound_file_ids = set()
-        dir_obj: Optional["Tree"] = None
-        for odb in (cache_odb, src):
-            if odb is not None:
-                try:
-                    dir_obj = Tree.load(odb, dir_hash)
-                    break
-                except (FileNotFoundError, ObjectFormatError):
-                    pass
+        dir_obj = find_tree_by_obj_id([cache_odb, src], dir_hash)
         assert dir_obj
+
         entry_ids = {entry.hash_info for _, entry in dir_obj}
+        bound_file_ids = all_file_ids & entry_ids
+        all_file_ids -= entry_ids
 
-        for file_hash in all_file_ids.copy():
-            if file_hash in entry_ids:
-                bound_file_ids.add(file_hash)
-                all_file_ids.remove(file_hash)
-
-        dir_fails = processor(bound_file_ids)
+        dir_fails = sum(processor(bound_file_ids))
         if dir_fails:
             logger.debug(
                 "failed to upload full contents of '%s', "
@@ -164,19 +102,13 @@ def _do_transfer(
                 dir_obj.name,
             )
         else:
-            raw_obj = src.get(dir_obj.hash_info)
-            is_dir_failed = processor.add_func(  # type: ignore[attr-defined]
-                raw_obj.path_info,
-                raw_obj.fs,
-                raw_obj.hash_info,
-                **kwargs,
-            )
+            is_dir_failed = sum(processor([dir_obj.hash_info]))
             total_fails += is_dir_failed
             if not is_dir_failed:
                 succeeded_dir_objs.append(dir_obj)
 
     # insert the rest
-    total_fails += processor(all_file_ids)
+    total_fails += sum(processor(all_file_ids))
     if total_fails:
         if src_index:
             src_index.clear()
@@ -222,18 +154,19 @@ def transfer(
     if not status.new:
         return 0
 
-    dir_ids, file_ids = split(lambda hash_info: hash_info.isdir, status.new)
-    if jobs is None:
-        jobs = dest.fs.jobs
+    def func(hash_info: "HashInfo") -> None:
+        obj = src.get(hash_info)
+        return dest.add(
+            obj.path_info, obj.fs, obj.hash_info, verify=verify, move=move
+        )
 
-    return _transfer(
-        src,
-        dest,
-        set(dir_ids),
-        set(file_ids),
-        status.missing,
-        jobs,
-        verify,
-        move,
-        **kwargs,
-    )
+    total = len(status.new)
+    jobs = jobs or dest.fs.jobs
+    with Tqdm(total=total, unit="file", desc="Transferring") as pbar:
+        with ThreadPoolExecutor(max_workers=jobs) as executor:
+            wrapped_func = pbar.wrap_fn(_log_exceptions(func))
+            processor = partial(executor.imap_unordered, wrapped_func)
+            _do_transfer(
+                src, dest, status.new, status.missing, processor, **kwargs
+            )
+    return total

dvc/output.py

@@ -81,6 +81,7 @@ def loadd_from(stage, d_list):
         desc = d.pop(Output.PARAM_DESC, False)
         isexec = d.pop(Output.PARAM_ISEXEC, False)
         live = d.pop(Output.PARAM_LIVE, False)
+        remote = d.pop(Output.PARAM_REMOTE, None)
         ret.append(
             _get(
                 stage,
@@ -94,6 +95,7 @@ def loadd_from(stage, d_list):
                 desc=desc,
                 isexec=isexec,
                 live=live,
+                remote=remote,
             )
         )
     return ret
@@ -109,6 +111,7 @@ def loads_from(
     checkpoint=False,
     isexec=False,
     live=False,
+    remote=None,
 ):
     return [
         _get(
@@ -122,6 +125,7 @@ def loads_from(
             checkpoint=checkpoint,
             isexec=isexec,
             live=live,
+            remote=remote,
         )
         for s in s_list
     ]
@@ -162,7 +166,6 @@ def load_from_pipeline(stage, data, typ="outs"):
     metric = typ == stage.PARAM_METRICS
     plot = typ == stage.PARAM_PLOTS
     live = typ == stage.PARAM_LIVE
-
     if live:
         # `live` is single object
         data = [data]
@@ -185,6 +188,7 @@ def load_from_pipeline(stage, data, typ="outs"):
                 Output.PARAM_CACHE,
                 Output.PARAM_PERSIST,
                 Output.PARAM_CHECKPOINT,
+                Output.PARAM_REMOTE,
             ],
         )
 
@@ -255,6 +259,7 @@ class Output:
     PARAM_LIVE = "live"
     PARAM_LIVE_SUMMARY = "summary"
     PARAM_LIVE_HTML = "html"
+    PARAM_REMOTE = "remote"
 
     METRIC_SCHEMA = Any(
         None,
@@ -283,6 +288,7 @@ class Output:
         live=False,
         desc=None,
         isexec=False,
+        remote=None,
     ):
         self.repo = stage.repo if stage else None
 
@@ -326,7 +332,7 @@ class Output:
         self.obj = None
         self.isexec = False if self.IS_DEPENDENCY else isexec
 
-        self.def_remote = None
+        self.remote = remote
 
     def _parse_path(self, fs, path_info):
         if fs.scheme != "local":
@@ -843,6 +849,8 @@ class Output:
         try:
             objects.check(self.odb, obj)
         except FileNotFoundError:
+            if self.remote:
+                kwargs["remote"] = self.remote
             self.repo.cloud.pull([obj.hash_info], **kwargs)
 
         if self.obj:
@@ -855,9 +863,9 @@ class Output:
 
         return self.obj
 
-    def collect_used_dir_cache(
+    def _collect_used_dir_cache(
         self, remote=None, force=False, jobs=None, filter_info=None
-    ) -> Dict[Optional["ObjectDB"], Set["HashInfo"]]:
+    ) -> Optional["Tree"]:
         """Fetch dir cache and return used object IDs for this out."""
 
         try:
@@ -878,13 +886,13 @@ class Output:
                     "unable to fully collect used cache"
                     " without cache for directory '{}'".format(self)
                 )
-            return {}
+            return None
 
         obj = self.get_obj()
         if filter_info and filter_info != self.path_info:
             prefix = filter_info.relative_to(self.path_info).parts
             obj = obj.filter(prefix)
-        return {None: set(self._named_obj_ids(obj))}
+        return obj
 
     def get_used_objs(
         self, **kwargs
@@ -917,22 +925,31 @@ class Output:
             return {}
 
         if self.is_dir_checksum:
-            return self.collect_used_dir_cache(**kwargs)
+            obj = self._collect_used_dir_cache(**kwargs)
+        else:
+            obj = self.get_obj(filter_info=kwargs.get("filter_info"))
+            if not obj:
+                obj = self.odb.get(self.hash_info)
 
-        obj = self.get_obj(filter_info=kwargs.get("filter_info"))
         if not obj:
-            obj = self.odb.get(self.hash_info)
+            return {}
 
-        return {None: set(self._named_obj_ids(obj))}
+        if self.remote:
+            remote = self.repo.cloud.get_remote_odb(name=self.remote)
+        else:
+            remote = None
+
+        return {remote: self._named_obj_ids(obj)}
 
     def _named_obj_ids(self, obj):
         name = str(self)
         obj.hash_info.obj_name = name
-        yield obj.hash_info
+        oids = {obj.hash_info}
         if isinstance(obj, Tree):
             for key, entry_obj in obj:
                 entry_obj.hash_info.obj_name = self.fs.sep.join([name, *key])
-                yield entry_obj.hash_info
+                oids.add(entry_obj.hash_info)
+        return oids
 
     def get_used_external(
         self, **kwargs
@@ -1033,4 +1050,5 @@ SCHEMA = {
     Output.PARAM_CACHE: bool,
     Output.PARAM_METRIC: Output.METRIC_SCHEMA,
     Output.PARAM_DESC: str,
+    Output.PARAM_REMOTE: str,
 }

dvc/progress.py

@@ -1,5 +1,4 @@
 """Manages progress bars for DVC repo."""
-
 import logging
 import sys
 from threading import RLock
@@ -182,9 +181,38 @@ class FsspecCallback(fsspec.Callback):
             size = self.fs.info(self.path_info)["size"]
         self.progress_bar.total = size
         self.progress_bar.refresh()
+        super().set_size(size)
 
     def relative_update(self, inc=1):
         self.progress_bar.update(inc)
+        super().relative_update(inc)
 
     def absolute_update(self, value):
         self.progress_bar.update_to(value)
+        super().absolute_update(value)
+
+    @staticmethod
+    def wrap_fn(cb, fn):
+        def wrapped(*args, **kwargs):
+            res = fn(*args, **kwargs)
+            cb.relative_update()
+            return res
+
+        return wrapped
+
+
+def tdqm_or_callback_wrapped(
+    fobj, method, total, callback=None, **pbar_kwargs
+):
+    if callback:
+        from funcy import nullcontext
+        from tqdm.utils import CallbackIOWrapper
+
+        callback.set_size(total)
+        wrapper = CallbackIOWrapper(callback.relative_update, fobj, method)
+        return nullcontext(wrapper)
+
+    return Tqdm.wrapattr(fobj, method, total=total, bytes=True, **pbar_kwargs)
+
+
+DEFAULT_CALLBACK = fsspec.callbacks.NoOpCallback()

dvc/render/__init__.py

@@ -0,0 +1,33 @@
+import logging
+from typing import TYPE_CHECKING, Dict
+
+if TYPE_CHECKING:
+    from dvc.types import StrPath
+
+logger = logging.getLogger(__name__)
+
+
+class Renderer:
+    def __init__(self, data: Dict):
+        self.data = data
+
+        from dvc.render.utils import get_files
+
+        files = get_files(self.data)
+
+        # we assume comparison of same file between revisions for now
+        assert len(files) == 1
+        self.filename = files.pop()
+
+    def _convert(self, path: "StrPath"):
+        raise NotImplementedError
+
+    @property
+    def DIV(self):
+        raise NotImplementedError
+
+    def generate_html(self, path: "StrPath"):
+        """this method might edit content of path"""
+        partial = self._convert(path)
+        div_id = f"plot_{self.filename.replace('.', '_').replace('/', '_')}"
+        return self.DIV.format(id=div_id, partial=partial)

dvc/render/html.py

@@ -1,11 +1,17 @@
-from typing import Dict, List, Optional
+import os
+from pathlib import Path
+from typing import TYPE_CHECKING, Dict, List, Optional
 
 from dvc.exceptions import DvcException
-from dvc.types import StrPath
+
+if TYPE_CHECKING:
+    from dvc.render import Renderer
+    from dvc.types import StrPath
 
 PAGE_HTML = """<!DOCTYPE html>
 <html>
 <head>
+    {refresh_tag}
     <title>DVC Plot</title>
     <script src="https://cdn.jsdelivr.net/npm/[email protected]"></script>
     <script src="https://cdn.jsdelivr.net/npm/[email protected]"></script>
@@ -16,12 +22,6 @@ PAGE_HTML = """<!DOCTYPE html>
 </body>
 </html>"""
 
-VEGA_DIV_HTML = """<div id = "{id}"></div>
-<script type = "text/javascript">
-    var spec = {vega_json};
-    vegaEmbed('#{id}', spec);
-</script>"""
-
 
 class MissingPlaceholderError(DvcException):
     def __init__(self, placeholder):
@@ -31,14 +31,23 @@ class MissingPlaceholderError(DvcException):
 class HTML:
     PLACEHOLDER = "plot_divs"
     PLACEHOLDER_FORMAT_STR = f"{{{PLACEHOLDER}}}"
+    REFRESH_PLACEHOLDER = "refresh_tag"
+    REFRESH_TAG = '<meta http-equiv="refresh" content="{}">'
 
-    def __init__(self, template: str = None):
+    def __init__(
+        self,
+        template: Optional[str] = None,
+        refresh_seconds: Optional[int] = None,
+    ):
         template = template or PAGE_HTML
         if self.PLACEHOLDER_FORMAT_STR not in template:
             raise MissingPlaceholderError(self.PLACEHOLDER_FORMAT_STR)
 
         self.template = template
         self.elements: List[str] = []
+        self.refresh_tag = ""
+        if refresh_seconds is not None:
+            self.refresh_tag = self.REFRESH_TAG.format(refresh_seconds)
 
     def with_metrics(self, metrics: Dict[str, Dict]) -> "HTML":
         import tabulate
@@ -56,41 +65,43 @@ class HTML:
         self.elements.append(tabulate.tabulate(rows, header, tablefmt="html"))
         return self
 
-    def with_plots(self, plots: Dict[str, Dict]) -> "HTML":
-        self.elements.extend(
-            [
-                VEGA_DIV_HTML.format(id=f"plot{i}", vega_json=plot)
-                for i, plot in enumerate(plots.values())
-            ]
-        )
-        return self
-
     def with_element(self, html: str) -> "HTML":
         self.elements.append(html)
         return self
 
     def embed(self) -> str:
-        kwargs = {self.PLACEHOLDER: "\n".join(self.elements)}
+        kwargs = {
+            self.PLACEHOLDER: "\n".join(self.elements),
+            self.REFRESH_PLACEHOLDER: self.refresh_tag,
+        }
         return self.template.format(**kwargs)
 
 
 def write(
-    path: StrPath,
-    plots: Dict[str, Dict],
+    path: "StrPath",
+    renderers: List["Renderer"],
     metrics: Optional[Dict[str, Dict]] = None,
-    template_path: Optional[StrPath] = None,
+    template_path: Optional["StrPath"] = None,
+    refresh_seconds: Optional[int] = None,
 ):
+
+    os.makedirs(path, exist_ok=True)
+
     page_html = None
     if template_path:
         with open(template_path) as fobj:
             page_html = fobj.read()
 
-    document = HTML(page_html)
+    document = HTML(page_html, refresh_seconds=refresh_seconds)
     if metrics:
         document.with_metrics(metrics)
         document.with_element("<br>")
 
-    document.with_plots(plots)
+    for renderer in renderers:
+        document.with_element(renderer.generate_html(path))
 
-    with open(path, "w") as fd:
+    index = Path(os.path.join(path, "index.html"))
+
+    with open(index, "w") as fd:
         fd.write(document.embed())
+    return index

dvc/render/image.py

@@ -0,0 +1,64 @@
+import os
+from typing import TYPE_CHECKING
+
+from dvc.render import Renderer
+from dvc.render.utils import get_files
+from dvc.utils import relpath
+
+if TYPE_CHECKING:
+    from dvc.types import StrPath
+
+
+class ImageRenderer(Renderer):
+    DIV = """
+        <div
+            id="{id}"
+            style="border: 1px solid;">
+            {partial}
+        </div>"""
+
+    def _write_image(
+        self,
+        page_dir_path: "StrPath",
+        revision: str,
+        filename: str,
+        image_data: bytes,
+    ):
+        static = os.path.join(page_dir_path, "static")
+        os.makedirs(static, exist_ok=True)
+
+        img_path = os.path.join(
+            static, f"{revision}_{filename.replace(os.sep, '_')}"
+        )
+        rel_img_path = relpath(img_path, page_dir_path)
+        with open(img_path, "wb") as fd:
+            fd.write(image_data)
+        return """
+        <div>
+            <p>{title}</p>
+            <img src="{src}">
+        </div>""".format(
+            title=revision, src=rel_img_path
+        )
+
+    def _convert(self, path: "StrPath"):
+        div_content = []
+        for rev, rev_data in self.data.items():
+            if "data" in rev_data:
+                for file, file_data in rev_data.get("data", {}).items():
+                    if "data" in file_data:
+                        div_content.append(
+                            self._write_image(
+                                path, rev, file, file_data["data"]
+                            )
+                        )
+        if div_content:
+            div_content.insert(0, f"<p>{self.filename}</p>")
+            return "\n".join(div_content)
+        return ""
+
+    @staticmethod
+    def matches(data):
+        files = get_files(data)
+        extensions = set(map(lambda f: os.path.splitext(f)[1], files))
+        return extensions.issubset({".jpg", ".jpeg", ".gif", ".png"})

dvc/render/utils.py

@@ -0,0 +1,71 @@
+import os.path
+from typing import Dict, List
+
+import dpath
+
+
+def get_files(data: Dict) -> List:
+    files = set()
+    for rev in data.keys():
+        for file in data[rev].get("data", {}).keys():
+            files.add(file)
+    sorted_files = sorted(files)
+    return sorted_files
+
+
+def group_by_filename(plots_data: Dict) -> List[Dict]:
+    files = get_files(plots_data)
+    grouped = []
+    for file in files:
+        grouped.append(dpath.util.search(plots_data, ["*", "*", file]))
+    return grouped
+
+
+def find_vega(repo, plots_data, target):
+    found = dpath.util.search(plots_data, ["*", "*", target])
+    from dvc.render.vega import VegaRenderer
+
+    if found and VegaRenderer.matches(found):
+        return VegaRenderer(found, repo.plots.templates).get_vega()
+    return ""
+
+
+def match_renderers(plots_data, templates):
+    from dvc.render.image import ImageRenderer
+    from dvc.render.vega import VegaRenderer
+
+    renderers = []
+    for g in group_by_filename(plots_data):
+        if VegaRenderer.matches(g):
+            renderers.append(VegaRenderer(g, templates))
+        if ImageRenderer.matches(g):
+            renderers.append(ImageRenderer(g))
+    return renderers
+
+
+def render(
+    repo,
+    plots_data,
+    metrics=None,
+    path=None,
+    html_template_path=None,
+    refresh_seconds=None,
+):
+    # TODO we could probably remove repo usages (here and in VegaRenderer)
+    renderers = match_renderers(plots_data, repo.plots.templates)
+    if not html_template_path:
+        html_template_path = repo.config.get("plots", {}).get(
+            "html_template", None
+        )
+        if html_template_path and not os.path.isabs(html_template_path):
+            html_template_path = os.path.join(repo.dvc_dir, html_template_path)
+
+    from dvc.render.html import write
+
+    return write(
+        path,
+        renderers,
+        metrics=metrics,
+        template_path=html_template_path,
+        refresh_seconds=refresh_seconds,
+    )

dvc/render/vega.py

@@ -0,0 +1,175 @@
+import os
+from copy import copy, deepcopy
+from typing import TYPE_CHECKING, Dict, List, Optional, Union
+
+from funcy import first
+
+from dvc.exceptions import DvcException
+from dvc.render import Renderer
+from dvc.render.utils import get_files
+
+if TYPE_CHECKING:
+    from dvc.repo.plots.template import PlotTemplates
+    from dvc.types import StrPath
+
+REVISION_FIELD = "rev"
+INDEX_FIELD = "step"
+
+
+class PlotMetricTypeError(DvcException):
+    def __init__(self, file):
+        super().__init__(
+            "'{}' - file type error\n"
+            "Only JSON, YAML, CSV and TSV formats are supported.".format(file)
+        )
+
+
+class PlotDataStructureError(DvcException):
+    def __init__(self):
+        super().__init__(
+            "Plot data extraction failed. Please see "
+            "https://man.dvc.org/plot for supported data formats."
+        )
+
+
+def _filter_fields(
+    datapoints: List[Dict], filename, revision, fields=None
+) -> List[Dict]:
+    if not fields:
+        return datapoints
+    assert isinstance(fields, set)
+
+    new_data = []
+    for data_point in datapoints:
+        new_dp = copy(data_point)
+
+        keys = set(data_point.keys())
+        if keys & fields != fields:
+            raise DvcException(
+                "Could not find fields: '{}' for '{}' at '{}'.".format(
+                    ", ".join(fields), filename, revision
+                )
+            )
+
+        to_del = keys - fields
+        for key in to_del:
+            del new_dp[key]
+        new_data.append(new_dp)
+    return new_data
+
+
+def _lists(dictionary):
+    for _, value in dictionary.items():
+        if isinstance(value, dict):
+            yield from _lists(value)
+        elif isinstance(value, list):
+            yield value
+
+
+def _find_data(data: Union[Dict, List], fields=None) -> List[Dict]:
+    if not isinstance(data, dict):
+        return data
+
+    if not fields:
+        # just look for first list of dicts
+        fields = set()
+
+    for lst in _lists(data):
+        if (
+            all(isinstance(dp, dict) for dp in lst)
+            and set(first(lst).keys()) & fields == fields
+        ):
+            return lst
+    raise PlotDataStructureError()
+
+
+def _append_index(datapoints: List[Dict], append_index=False) -> List[Dict]:
+    if not append_index or INDEX_FIELD in first(datapoints).keys():
+        return datapoints
+
+    for index, data_point in enumerate(datapoints):
+        data_point[INDEX_FIELD] = index
+    return datapoints
+
+
+def _append_revision(datapoints: List[Dict], revision) -> List[Dict]:
+    for data_point in datapoints:
+        data_point[REVISION_FIELD] = revision
+    return datapoints
+
+
+class VegaRenderer(Renderer):
+    DIV = """
+    <div id = "{id}">
+        <script type = "text/javascript">
+            var spec = {partial};
+            vegaEmbed('#{id}', spec);
+        </script>
+    </div>
+    """
+
+    def __init__(self, data: Dict, templates: "PlotTemplates"):
+        super().__init__(data)
+        self.templates = templates
+
+    def _squash_props(self) -> Dict:
+        resolved: Dict[str, str] = {}
+        for rev_data in self.data.values():
+            for file_data in rev_data.get("data", {}).values():
+                props = file_data.get("props", {})
+                resolved = {**resolved, **props}
+        return resolved
+
+    def _datapoints(self, props: Dict):
+        fields = props.get("fields", set())
+        if fields:
+            fields = {*fields, props.get("x"), props.get("y")} - {None}
+
+        datapoints = []
+        for revision, rev_data in self.data.items():
+            for filename, file_data in rev_data.get("data", {}).items():
+                if "data" in file_data:
+                    tmp = deepcopy(file_data.get("data"))
+                    tmp = _find_data(tmp, fields=fields - {INDEX_FIELD})
+                    tmp = _append_index(
+                        tmp, append_index=props.get("append_index", False)
+                    )
+                    tmp = _filter_fields(
+                        tmp,
+                        filename=filename,
+                        revision=revision,
+                        fields=fields,
+                    )
+                    tmp = _append_revision(tmp, revision=revision)
+                    datapoints.extend(tmp)
+        return datapoints
+
+    def get_vega(self) -> Optional[str]:
+        props = self._squash_props()
+
+        template = self.templates.load(props.get("template") or "default")
+
+        if not props.get("x") and template.has_anchor("x"):
+            props["append_index"] = True
+            props["x"] = INDEX_FIELD
+
+        datapoints = self._datapoints(props)
+
+        if datapoints:
+            if not props.get("y") and template.has_anchor("y"):
+                fields = list(first(datapoints))
+                skip = (REVISION_FIELD, props.get("x"))
+                props["y"] = first(
+                    f for f in reversed(fields) if f not in skip
+                )
+            return template.render(datapoints, props=props)
+        return None
+
+    def _convert(self, path: "StrPath"):
+        return self.get_vega()
+
+    @staticmethod
+    def matches(data):
+        files = get_files(data)
+        extensions = set(map(lambda f: os.path.splitext(f)[1], files))
+        return extensions.issubset({".yml", ".yaml", ".json", ".csv", ".tsv"})

dvc/repo/__init__.py

@@ -12,6 +12,7 @@ from dvc.exceptions import IsADirectoryError as DvcIsADirectoryError
 from dvc.exceptions import NotDvcRepoError, OutputNotFoundError
 from dvc.ignore import DvcIgnoreFilter
 from dvc.path_info import PathInfo
+from dvc.utils import env2bool
 from dvc.utils.fs import path_isin
 
 if TYPE_CHECKING:
@@ -134,6 +135,7 @@ class Repo:
         from dvc.data_cloud import DataCloud
         from dvc.fs.local import LocalFileSystem
         from dvc.lock import LockNoop, make_lock
+        from dvc.machine import MachineManager
         from dvc.objects.db import ODBManager
         from dvc.repo.live import Live
         from dvc.repo.metrics import Metrics
@@ -196,6 +198,14 @@ class Repo:
         self.params = Params(self)
         self.live = Live(self)
 
+        if self.tmp_dir and (
+            self.config["feature"].get("machine", False)
+            or env2bool("DVC_TEST")
+        ):
+            self.machine = MachineManager(self)
+        else:
+            self.machine = None
+
         self.stage_collection_error_handler: Optional[
             Callable[[str, Exception], None]
         ] = None

dvc/repo/collect.py

@@ -49,7 +49,6 @@ def _collect_paths(
                 )
             else:
                 logger.warning("'%s' was not found at: '%s'.", path_info, rev)
-            continue
         target_infos.append(path_info)
     return target_infos
 

dvc/repo/diff.py

@@ -4,7 +4,6 @@ from collections import defaultdict
 from typing import Dict, List
 
 from dvc.exceptions import PathMissingError
-from dvc.objects.stage import get_file_hash
 from dvc.repo import locked
 from dvc.repo.experiments.utils import fix_exp_head
 
@@ -45,7 +44,7 @@ def diff(self, a_rev="HEAD", b_rev=None, targets=None):
                 repo_fs, targets
             )
 
-        results[rev] = _paths_checksums(self, repo_fs, targets_path_infos)
+        results[rev] = _paths_checksums(self, targets_path_infos)
 
     if targets is not None:
         # check for overlapping missing targets between a_rev and b_rev
@@ -98,7 +97,7 @@ def diff(self, a_rev="HEAD", b_rev=None, targets=None):
     return ret if any(ret.values()) else {}
 
 
-def _paths_checksums(repo, repo_fs, targets):
+def _paths_checksums(repo, targets):
     """
     A dictionary of checksums addressed by relpaths collected from
     the current fs outputs.
@@ -110,10 +109,10 @@ def _paths_checksums(repo, repo_fs, targets):
         file:      "data"
     """
 
-    return dict(_output_paths(repo, repo_fs, targets))
+    return dict(_output_paths(repo, targets))
 
 
-def _output_paths(repo, repo_fs, targets):
+def _output_paths(repo, targets):
     from dvc.fs.local import LocalFileSystem
     from dvc.objects.stage import stage as ostage
 
@@ -131,45 +130,47 @@ def _output_paths(repo, repo_fs, targets):
             else os.path.join(str(output), "")
         )
 
-    def _to_checksum(output):
-        if on_working_fs:
-            return ostage(
-                repo.odb.local,
-                output.path_info,
-                repo.odb.local.fs,
-                "md5",
-                dry_run=True,
-            )[1].hash_info.value
-        return output.hash_info.value
-
     for output in repo.index.outs:
         if _exists(output):
             yield_output = targets is None or any(
                 output.path_info.isin_or_eq(target) for target in targets
             )
 
+            if on_working_fs:
+                _, obj = ostage(
+                    repo.odb.local,
+                    output.path_info,
+                    repo.odb.local.fs,
+                    "md5",
+                    dry_run=True,
+                    dvcignore=output.dvcignore,
+                )
+                hash_info = obj.hash_info
+            else:
+                hash_info = output.hash_info
+                obj = output.get_obj()
+
             if yield_output:
-                yield _to_path(output), _to_checksum(output)
+                yield _to_path(output), hash_info.value
+
+            if not obj:
+                continue
 
             if output.is_dir_checksum and (
                 yield_output
                 or any(target.isin(output.path_info) for target in targets)
             ):
-                yield from _dir_output_paths(repo_fs, output, targets)
+                yield from _dir_output_paths(output.path_info, obj, targets)
 
 
-def _dir_output_paths(repo_fs, output, targets=None):
-    from dvc.config import NoRemoteError
-
-    try:
-        for fname in repo_fs.walk_files(output.path_info):
-            if targets is None or any(
-                fname.isin_or_eq(target) for target in targets
-            ):
-                # pylint: disable=no-member
-                yield str(fname), get_file_hash(fname, repo_fs, "md5").value
-    except NoRemoteError:
-        logger.warning("dir cache entry for '%s' is missing", output)
+def _dir_output_paths(path_info, obj, targets=None):
+    for key, entry_obj in obj:
+        fname = path_info.joinpath(*key)
+        if targets is None or any(
+            fname.isin_or_eq(target) for target in targets
+        ):
+            # pylint: disable=no-member
+            yield str(fname), entry_obj.hash_info.value
 
 
 def _filter_missing(repo_fs, paths):

dvc/repo/experiments/executor/base.py

@@ -348,6 +348,7 @@ class BaseExecutor(ABC):
                     quiet=True,
                     allow_missing=True,
                     checkpoint_reset=checkpoint_reset,
+                    recursive=kwargs.get("recursive", False),
                 )
 
             checkpoint_func = partial(

dvc/repo/experiments/pull.py

@@ -4,7 +4,7 @@ from dvc.exceptions import DvcException, InvalidArgumentError
 from dvc.repo import locked
 from dvc.repo.scm_context import scm_context
 
-from .utils import exp_commits, remote_exp_refs_by_name
+from .utils import exp_commits, resolve_exp_ref
 
 logger = logging.getLogger(__name__)
 
@@ -14,7 +14,11 @@ logger = logging.getLogger(__name__)
 def pull(
     repo, git_remote, exp_name, *args, force=False, pull_cache=False, **kwargs
 ):
-    exp_ref = _get_exp_ref(repo, git_remote, exp_name)
+    exp_ref = resolve_exp_ref(repo.scm, exp_name, git_remote)
+    if not exp_ref:
+        raise InvalidArgumentError(
+            f"Experiment '{exp_name}' does not exist in '{git_remote}'"
+        )
 
     def on_diverged(refname: str, rev: str) -> bool:
         if repo.scm.get_ref(refname) == rev:
@@ -35,33 +39,6 @@ def pull(
         _pull_cache(repo, exp_ref, **kwargs)
 
 
-def _get_exp_ref(repo, git_remote, exp_name):
-    if exp_name.startswith("refs/"):
-        return exp_name
-
-    exp_refs = list(remote_exp_refs_by_name(repo.scm, git_remote, exp_name))
-    if not exp_refs:
-        raise InvalidArgumentError(
-            f"Experiment '{exp_name}' does not exist in '{git_remote}'"
-        )
-    if len(exp_refs) > 1:
-        cur_rev = repo.scm.get_rev()
-        for info in exp_refs:
-            if info.baseline_sha == cur_rev:
-                return info
-        msg = [
-            (
-                f"Ambiguous name '{exp_name}' refers to multiple "
-                "experiments in '{git_remote}'. Use full refname to pull one "
-                "of the following:"
-            ),
-            "",
-        ]
-        msg.extend([f"\t{info}" for info in exp_refs])
-        raise InvalidArgumentError("\n".join(msg))
-    return exp_refs[0]
-
-
 def _pull_cache(repo, exp_ref, dvc_remote=None, jobs=None, run_cache=False):
     revs = list(exp_commits(repo.scm, [exp_ref]))
     logger.debug("dvc fetch experiment '%s'", exp_ref)

dvc/repo/experiments/push.py

@@ -2,10 +2,9 @@ import logging
 
 from dvc.exceptions import DvcException, InvalidArgumentError
 from dvc.repo import locked
-from dvc.repo.experiments.base import ExpRefInfo
 from dvc.repo.scm_context import scm_context
 
-from .utils import exp_commits, exp_refs_by_name
+from .utils import exp_commits, resolve_exp_ref
 
 logger = logging.getLogger(__name__)
 
@@ -21,7 +20,11 @@ def push(
     push_cache=False,
     **kwargs,
 ):
-    exp_ref = _get_exp_ref(repo, exp_name)
+    exp_ref = resolve_exp_ref(repo.scm, exp_name)
+    if not exp_ref:
+        raise InvalidArgumentError(
+            f"'{exp_name}' is not a valid experiment name"
+        )
 
     def on_diverged(refname: str, rev: str) -> bool:
         if repo.scm.get_ref(refname) == rev:
@@ -42,33 +45,6 @@ def push(
         _push_cache(repo, exp_ref, **kwargs)
 
 
-def _get_exp_ref(repo, exp_name: str) -> ExpRefInfo:
-    if exp_name.startswith("refs/"):
-        return ExpRefInfo.from_ref(exp_name)
-
-    exp_refs = list(exp_refs_by_name(repo.scm, exp_name))
-    if not exp_refs:
-        raise InvalidArgumentError(
-            f"'{exp_name}' is not a valid experiment name"
-        )
-    if len(exp_refs) > 1:
-        cur_rev = repo.scm.get_rev()
-        for info in exp_refs:
-            if info.baseline_sha == cur_rev:
-                return info
-        msg = [
-            (
-                f"Ambiguous name '{exp_name}' refers to multiple "
-                "experiments. Use full refname to push one of the "
-                "following:"
-            ),
-            "",
-        ]
-        msg.extend([f"\t{info}" for info in exp_refs])
-        raise InvalidArgumentError("\n".join(msg))
-    return exp_refs[0]
-
-
 def _push_cache(repo, exp_ref, dvc_remote=None, jobs=None, run_cache=False):
     revs = list(exp_commits(repo.scm, [exp_ref]))
     logger.debug("dvc push experiment '%s'", exp_ref)

dvc/repo/experiments/remove.py

@@ -6,33 +6,47 @@ from dvc.repo import locked
 from dvc.repo.scm_context import scm_context
 from dvc.scm.base import RevError
 
-from .base import EXPS_NAMESPACE, ExpRefInfo
-from .utils import exp_refs_by_name, remove_exp_refs
+from .utils import exp_refs, remove_exp_refs, resolve_exp_ref
 
 logger = logging.getLogger(__name__)
 
 
 @locked
 @scm_context
-def remove(repo, exp_names=None, queue=False, **kwargs):
-    if not exp_names and not queue:
+def remove(
+    repo,
+    exp_names=None,
+    queue=False,
+    clear_all=False,
+    remote=None,
+    **kwargs,
+):
+    if not any([exp_names, queue, clear_all]):
         return 0
 
     removed = 0
     if queue:
-        removed += len(repo.experiments.stash)
-        repo.experiments.stash.clear()
+        removed += _clear_stash(repo)
+    if clear_all:
+        removed += _clear_all(repo)
+
     if exp_names:
-        remained = _remove_commited_exps(repo, exp_names)
-        remained = _remove_queued_exps(repo, remained)
-        if remained:
-            raise InvalidArgumentError(
-                "'{}' is not a valid experiment".format(";".join(remained))
-            )
-        removed += len(exp_names) - len(remained)
+        removed += _remove_exp_by_names(repo, remote, exp_names)
     return removed
 
 
+def _clear_stash(repo):
+    removed = len(repo.experiments.stash)
+    repo.experiments.stash.clear()
+    return removed
+
+
+def _clear_all(repo):
+    ref_infos = list(exp_refs(repo.scm))
+    remove_exp_refs(repo.scm, ref_infos)
+    return len(ref_infos)
+
+
 def _get_exp_stash_index(repo, ref_or_rev: str) -> Optional[int]:
     stash_revs = repo.experiments.stash_revs
     for _, ref_info in stash_revs.items():
@@ -47,46 +61,24 @@ def _get_exp_stash_index(repo, ref_or_rev: str) -> Optional[int]:
     return None
 
 
-def _get_exp_ref(repo, exp_name: str) -> Optional[ExpRefInfo]:
-    cur_rev = repo.scm.get_rev()
-    if exp_name.startswith(EXPS_NAMESPACE):
-        if repo.scm.get_ref(exp_name):
-            return ExpRefInfo.from_ref(exp_name)
-    else:
-        exp_refs = list(exp_refs_by_name(repo.scm, exp_name))
-        if exp_refs:
-            return _get_ref(exp_refs, exp_name, cur_rev)
-    return None
-
-
-def _get_ref(ref_infos, name, cur_rev) -> Optional[ExpRefInfo]:
-    if len(ref_infos) > 1:
-        for info in ref_infos:
-            if info.baseline_sha == cur_rev:
-                return info
-        msg = [
-            (
-                f"Ambiguous name '{name}' refers to multiple "
-                "experiments. Use full refname to remove one of "
-                "the following:"
-            )
-        ]
-        msg.extend([f"\t{info}" for info in ref_infos])
-        raise InvalidArgumentError("\n".join(msg))
-    return ref_infos[0]
-
-
-def _remove_commited_exps(repo, refs: List[str]) -> List[str]:
+def _remove_commited_exps(
+    repo, remote: Optional[str], exp_names: List[str]
+) -> List[str]:
     remain_list = []
     remove_list = []
-    for ref in refs:
-        ref_info = _get_exp_ref(repo, ref)
+    for exp_name in exp_names:
+        ref_info = resolve_exp_ref(repo.scm, exp_name, remote)
+
         if ref_info:
             remove_list.append(ref_info)
         else:
-            remain_list.append(ref)
+            remain_list.append(exp_name)
     if remove_list:
-        remove_exp_refs(repo.scm, remove_list)
+        if not remote:
+            remove_exp_refs(repo.scm, remove_list)
+        else:
+            for ref_info in remove_list:
+                repo.scm.push_refspec(remote, None, str(ref_info))
     return remain_list
 
 
@@ -99,3 +91,14 @@ def _remove_queued_exps(repo, refs_or_revs: List[str]) -> List[str]:
         else:
             repo.experiments.stash.drop(stash_index)
     return remain_list
+
+
+def _remove_exp_by_names(repo, remote, exp_names: List[str]) -> int:
+    remained = _remove_commited_exps(repo, remote, exp_names)
+    if not remote:
+        remained = _remove_queued_exps(repo, remained)
+    if remained:
+        raise InvalidArgumentError(
+            "'{}' is not a valid experiment".format(";".join(remained))
+        )
+    return len(exp_names) - len(remained)

dvc/repo/experiments/utils.py

@@ -1,5 +1,6 @@
 from typing import Generator, Iterable, Optional, Set
 
+from dvc.exceptions import InvalidArgumentError
 from dvc.scm.git import Git
 
 from .base import (
@@ -115,3 +116,44 @@ def fix_exp_head(scm: "Git", ref: Optional[str]) -> Optional[str]:
         if name == "HEAD" and scm.get_ref(EXEC_BASELINE):
             return "".join((EXEC_BASELINE, tail))
     return ref
+
+
+def resolve_exp_ref(
+    scm, exp_name: str, git_remote: Optional[str] = None
+) -> Optional[ExpRefInfo]:
+    if exp_name.startswith("refs/"):
+        return ExpRefInfo.from_ref(exp_name)
+
+    if git_remote:
+        exp_ref_list = list(remote_exp_refs_by_name(scm, git_remote, exp_name))
+    else:
+        exp_ref_list = list(exp_refs_by_name(scm, exp_name))
+
+    if not exp_ref_list:
+        return None
+    if len(exp_ref_list) > 1:
+        cur_rev = scm.get_rev()
+        for info in exp_ref_list:
+            if info.baseline_sha == cur_rev:
+                return info
+        if git_remote:
+            msg = [
+                (
+                    f"Ambiguous name '{exp_name}' refers to multiple "
+                    "experiments. Use full refname to push one of the "
+                    "following:"
+                ),
+                "",
+            ]
+        else:
+            msg = [
+                (
+                    f"Ambiguous name '{exp_name}' refers to multiple "
+                    f"experiments in '{git_remote}'. Use full refname to pull "
+                    "one of the following:"
+                ),
+                "",
+            ]
+        msg.extend([f"\t{info}" for info in exp_ref_list])
+        raise InvalidArgumentError("\n".join(msg))
+    return exp_ref_list[0]

dvc/repo/live.py

@@ -2,6 +2,8 @@ import logging
 import os
 from typing import TYPE_CHECKING, List, Optional
 
+from dvc.render.utils import render
+
 logger = logging.getLogger(__name__)
 
 if TYPE_CHECKING:
@@ -15,10 +17,12 @@ def create_summary(out):
 
     metrics, plots = out.repo.live.show(str(out.path_info))
 
-    html_path = out.path_info.with_suffix(".html")
+    html_path = out.path_info.fspath + "_dvc_plots"
 
-    out.repo.plots.write_html(html_path, plots, metrics)
-    logger.info(f"\nfile://{os.path.abspath(html_path)}")
+    index_path = render(
+        out.repo, plots, metrics=metrics, path=html_path, refresh_seconds=5
+    )
+    logger.info(f"\nfile://{os.path.abspath(index_path)}")
 
 
 def summary_path_info(out: "Output") -> Optional["PathInfo"]:

dvc/repo/plots/__init__.py

@@ -8,8 +8,7 @@ from typing import TYPE_CHECKING, Callable, Dict, List, Optional
 from funcy import cached_property, first, project
 
 from dvc.exceptions import DvcException
-from dvc.repo.plots.data import PlotMetricTypeError
-from dvc.types import StrPath
+from dvc.render.vega import PlotMetricTypeError
 from dvc.utils import (
     error_handler,
     errored_revisions,
@@ -113,58 +112,33 @@ class Plots:
             props = props or {}
 
             for path, repo_path in plot_files:
-                res[repo_path] = {"props": rev_props}
+                joined_props = {**rev_props, **props}
+                res[repo_path] = {"props": joined_props}
                 res[repo_path].update(
                     parse(
                         fs,
                         path,
-                        rev_props=rev_props,
-                        props=props,
+                        props=joined_props,
                         onerror=onerror,
                     )
                 )
         return res
 
-    @staticmethod
-    def render(data, revs=None, props=None, templates=None):
-        """Renders plots"""
-        props = props or {}
-
-        # Merge data by plot file and apply overriding props
-        plots = _prepare_plots(data, revs, props)
-
-        result = {}
-        for datafile, desc in plots.items():
-            rendered = _render(
-                datafile,
-                desc["data"],
-                desc["props"],
-                templates,
-            )
-            if rendered:
-                result[datafile] = rendered
-
-        return result
-
     def show(
         self,
         targets: List[str] = None,
         revs=None,
         props=None,
-        templates=None,
         recursive=False,
         onerror=None,
     ):
         if onerror is None:
             onerror = onerror_collect
-        data = self.collect(
+
+        return self.collect(
             targets, revs, recursive, onerror=onerror, props=props
         )
 
-        if templates is None:
-            templates = self.templates
-        return self.render(data, revs, props, templates)
-
     def diff(self, *args, **kwargs):
         from .diff import diff
 
@@ -217,26 +191,6 @@ class Plots:
 
         return PlotTemplates(self.repo.dvc_dir)
 
-    def write_html(
-        self,
-        path: StrPath,
-        plots: Dict[str, Dict],
-        metrics: Optional[Dict[str, Dict]] = None,
-        html_template_path: Optional[StrPath] = None,
-    ):
-        if not html_template_path:
-            html_template_path = self.repo.config.get("plots", {}).get(
-                "html_template", None
-            )
-            if html_template_path and not os.path.isabs(html_template_path):
-                html_template_path = os.path.join(
-                    self.repo.dvc_dir, html_template_path
-                )
-
-        from dvc.utils.html import write
-
-        write(path, plots, metrics, html_template_path)
-
 
 def _is_plot(out: "Output") -> bool:
     return bool(out.plot) or bool(out.live)
@@ -264,17 +218,19 @@ def _collect_plots(
 
 
 @error_handler
-def parse(fs, path, props=None, rev_props=None, **kwargs):
+def parse(fs, path, props=None, **kwargs):
     props = props or {}
-    rev_props = rev_props or {}
     _, extension = os.path.splitext(path)
     if extension in (".tsv", ".csv"):
-        header = {**rev_props, **props}.get("header", True)
+        header = props.get("header", True)
         if extension == ".csv":
             return _load_sv(path=path, fs=fs, delimiter=",", header=header)
         return _load_sv(path=path, fs=fs, delimiter="\t", header=header)
     if extension in LOADERS or extension in (".yml", ".yaml"):
         return LOADERS[extension](path=path, fs=fs)
+    if extension in (".jpeg", ".jpg", ".gif", ".png"):
+        with fs.open(path, "rb") as fd:
+            return fd.read()
     raise PlotMetricTypeError(path)
 
 
@@ -291,89 +247,6 @@ def _plot_props(out: "Output") -> Dict:
     return project(out.plot, PLOT_PROPS)
 
 
-def _prepare_plots(data, revs, props):
-    """Groups data by plot file.
-
-    Also resolves props conflicts between revs and applies global props.
-    """
-    # we go in order revs are supplied on props conflict first ones win.
-    revs = iter(data) if revs is None else revs
-
-    plots, props_revs = {}, {}
-    for rev in revs:
-        # Asked for revision without data
-        if rev not in data:
-            continue
-
-        for datafile, desc in data[rev].get("data", {}).items():
-            # We silently skip on an absent data file,
-            # see also try/except/pass in .collect()
-            if "data" not in desc:
-                continue
-
-            # props from command line overwrite plot props from out definition
-            full_props = {**desc["props"], **props}
-
-            if datafile in plots:
-                saved = plots[datafile]
-                if saved["props"] != full_props:
-                    logger.warning(
-                        f"Inconsistent plot props for '{datafile}' in "
-                        f"'{props_revs[datafile]}' and '{rev}'. "
-                        f"Going to use ones from '{props_revs[datafile]}'"
-                    )
-
-                saved["data"][rev] = desc["data"]
-            else:
-                plots[datafile] = {
-                    "props": full_props,
-                    "data": {rev: desc["data"]},
-                }
-                # Save rev we got props from
-                props_revs[datafile] = rev
-
-    return plots
-
-
-def _render(datafile, datas, props, templates):
-    from .data import PlotData, plot_data
-
-    # Copy it to not modify a passed value
-    props = props.copy()
-
-    # Add x and y to fields if set
-    fields = props.get("fields")
-    if fields is not None:
-        fields = {*fields, props.get("x"), props.get("y")} - {None}
-
-    template = templates.load(props.get("template") or "default")
-
-    # If x is not set add index field
-    if not props.get("x") and template.has_anchor("x"):
-        props["append_index"] = True
-        props["x"] = PlotData.INDEX_FIELD
-
-    # Parse all data, preprocess it and collect as a list of dicts
-    data = []
-    for rev, unstructured_data in datas.items():
-        rev_data = plot_data(datafile, rev, unstructured_data).to_datapoints(
-            fields=fields,
-            path=props.get("path"),
-            append_index=props.get("append_index", False),
-        )
-        data.extend(rev_data)
-
-    # If y is not set then use last field not used yet
-    if data:
-        if not props.get("y") and template.has_anchor("y"):
-            fields = list(first(data))
-            skip = (PlotData.REVISION_FIELD, props.get("x"))
-            props["y"] = first(f for f in reversed(fields) if f not in skip)
-
-        return template.render(data, props=props)
-    return None
-
-
 def _load_sv(path, fs, delimiter=",", header=True):
     with fs.open(path, "r") as fd:
         content = fd.read()

dvc/repo/plots/data.py

@@ -1,138 +0,0 @@
-import os
-from copy import copy
-
-from funcy import first
-
-from dvc.exceptions import DvcException
-from dvc.utils.serialize import ParseError
-
-
-class PlotMetricTypeError(DvcException):
-    def __init__(self, file):
-        super().__init__(
-            "'{}' - file type error\n"
-            "Only JSON, YAML, CSV and TSV formats are supported.".format(file)
-        )
-
-
-class PlotDataStructureError(DvcException):
-    def __init__(self):
-        super().__init__(
-            "Plot data extraction failed. Please see "
-            "https://man.dvc.org/plot for supported data formats."
-        )
-
-
-class PlotParsingError(ParseError):
-    def __init__(self, path, revision):
-        self.path = path
-        self.revision = revision
-
-        super().__init__(path, f"revision: '{revision}'")
-
-
-def plot_data(filename, revision, content):
-    _, extension = os.path.splitext(filename.lower())
-    if extension in (".json", ".yaml"):
-        return DictData(filename, revision, content)
-    if extension in (".csv", ".tsv"):
-        return ListData(filename, revision, content)
-    raise PlotMetricTypeError(filename)
-
-
-def _filter_fields(data_points, filename, revision, fields=None, **kwargs):
-    if not fields:
-        return data_points
-    assert isinstance(fields, set)
-
-    new_data = []
-    for data_point in data_points:
-        new_dp = copy(data_point)
-
-        keys = set(data_point.keys())
-        if keys & fields != fields:
-            raise DvcException(
-                "Could not find fields: '{}' for '{}' at '{}'.".format(
-                    ", ".join(fields), filename, revision
-                )
-            )
-
-        to_del = keys - fields
-        for key in to_del:
-            del new_dp[key]
-        new_data.append(new_dp)
-    return new_data
-
-
-def _lists(dictionary):
-    for _, value in dictionary.items():
-        if isinstance(value, dict):
-            yield from _lists(value)
-        elif isinstance(value, list):
-            yield value
-
-
-def _find_data(data, fields=None, **kwargs):
-    if not isinstance(data, dict):
-        return data
-
-    if not fields:
-        # just look for first list of dicts
-        fields = set()
-
-    for lst in _lists(data):
-        if (
-            all(isinstance(dp, dict) for dp in lst)
-            and set(first(lst).keys()) & fields == fields
-        ):
-            return lst
-    raise PlotDataStructureError()
-
-
-def _append_index(data_points, append_index=False, **kwargs):
-    if not append_index or PlotData.INDEX_FIELD in first(data_points).keys():
-        return data_points
-
-    for index, data_point in enumerate(data_points):
-        data_point[PlotData.INDEX_FIELD] = index
-    return data_points
-
-
-def _append_revision(data_points, revision, **kwargs):
-    for data_point in data_points:
-        data_point[PlotData.REVISION_FIELD] = revision
-    return data_points
-
-
-class PlotData:
-    REVISION_FIELD = "rev"
-    INDEX_FIELD = "step"
-
-    def __init__(self, filename, revision, content, **kwargs):
-        self.filename = filename
-        self.revision = revision
-        self.content = content
-
-    def _processors(self):
-        return [_filter_fields, _append_index, _append_revision]
-
-    def to_datapoints(self, **kwargs):
-        data = self.content
-
-        for data_proc in self._processors():
-            data = data_proc(
-                data, filename=self.filename, revision=self.revision, **kwargs
-            )
-        return data
-
-
-class DictData(PlotData):
-    # For files usually parsed as dicts: eg JSON, Yaml
-    def _processors(self):
-        parent_processors = super()._processors()
-        return [_find_data] + parent_processors
-
-
-class ListData(PlotData):
-    # For files parsed as list: CSV, TSV
-    pass

dvc/schema.py

@@ -49,6 +49,7 @@ OUT_PSTAGE_DETAILED_SCHEMA = {
         Output.PARAM_PERSIST: bool,
         Output.PARAM_CHECKPOINT: bool,
         Output.PARAM_DESC: str,
+        Output.PARAM_REMOTE: str,
     }
 }
 

dvc/scm/base.py

@@ -47,6 +47,15 @@ class InvalidRemoteSCMRepo(SCMError):
         super().__init__(msg)
 
 
+class GitAuthError(SCMError):
+    def __init__(self, url: str):
+        super().__init__(
+            f"HTTP Git authentication is not supported: '{url}'"
+            "\nSee https://dvc.org/doc//user-guide/"
+            "troubleshooting#git-auth"
+        )
+
+
 class Base:
     """Base class for source control management driver implementations."""
 

dvc/scm/git/backend/dulwich.py

@@ -20,7 +20,7 @@ from funcy import cached_property
 
 from dvc.path_info import PathInfo
 from dvc.progress import Tqdm
-from dvc.scm.base import InvalidRemoteSCMRepo, SCMError
+from dvc.scm.base import GitAuthError, InvalidRemoteSCMRepo, SCMError
 from dvc.utils import relpath
 
 from ..objects import GitObject
@@ -39,7 +39,7 @@ class DulwichObject(GitObject):
         self.repo = repo
         self._name = name
         self._mode = mode
-        self.sha = sha
+        self._sha = sha
 
     def open(self, mode: str = "r", encoding: str = None):
         if not encoding:
@@ -68,6 +68,14 @@ class DulwichObject(GitObject):
                 self.repo, entry.path.decode(), entry.mode, entry.sha
             )
 
+    @property
+    def size(self) -> int:
+        return len(self.repo[self.sha].as_raw_string())
+
+    @property
+    def sha(self) -> str:
+        return self._sha
+
 
 class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
     """Dulwich Git backend."""
@@ -347,7 +355,7 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 yield os.fsdecode(key)
 
     def iter_remote_refs(self, url: str, base: Optional[str] = None):
-        from dulwich.client import get_transport_and_path
+        from dulwich.client import HTTPUnauthorized, get_transport_and_path
         from dulwich.errors import NotGitRepository
         from dulwich.porcelain import get_remote_repo
 
@@ -368,6 +376,8 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 yield from (os.fsdecode(ref) for ref in client.get_refs(path))
         except NotGitRepository as exc:
             raise InvalidRemoteSCMRepo(url) from exc
+        except HTTPUnauthorized:
+            raise GitAuthError(url)
 
     def get_refs_containing(self, rev: str, pattern: Optional[str] = None):
         raise NotImplementedError
@@ -380,7 +390,7 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
         force: bool = False,
         on_diverged: Optional[Callable[[str, str], bool]] = None,
     ):
-        from dulwich.client import get_transport_and_path
+        from dulwich.client import HTTPUnauthorized, get_transport_and_path
         from dulwich.errors import NotGitRepository, SendPackError
         from dulwich.porcelain import (
             DivergedBranches,
@@ -399,9 +409,11 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
             ) from exc
 
         def update_refs(refs):
+            from dulwich.objects import ZERO_SHA
+
             new_refs = {}
             for ref, value in zip(dest_refs, values):
-                if ref in refs:
+                if ref in refs and value != ZERO_SHA:
                     local_sha = self.repo.refs[ref]
                     remote_sha = refs[ref]
                     try:
@@ -437,6 +449,8 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 )
         except (NotGitRepository, SendPackError) as exc:
             raise SCMError("Git failed to push '{src}' to '{url}'") from exc
+        except HTTPUnauthorized:
+            raise GitAuthError(url)
 
     def _push_dest_refs(
         self, src: Optional[str], dest: str

dvc/scm/git/backend/gitpython.py

@@ -92,6 +92,14 @@ class GitPythonObject(GitObject):
         for obj in self.obj:
             yield GitPythonObject(obj)
 
+    @property
+    def size(self) -> int:
+        return self.obj.size
+
+    @property
+    def sha(self) -> str:
+        return self.obj.hexsha
+
 
 class GitPythonBackend(BaseGitBackend):  # pylint:disable=abstract-method
     """git-python Git backend."""

dvc/scm/git/backend/pygit2.py

@@ -58,6 +58,14 @@ class Pygit2Object(GitObject):
             return stat.S_IFDIR
         return self.obj.filemode
 
+    @property
+    def size(self) -> int:
+        return len(self.obj.read_raw())
+
+    @property
+    def sha(self) -> str:
+        return self.obj.hex
+
     def scandir(self) -> Iterable["Pygit2Object"]:
         for entry in self.obj:  # noqa: B301
             yield Pygit2Object(entry)

dvc/scm/git/objects.py

@@ -1,4 +1,3 @@
-import os
 import stat
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
@@ -44,6 +43,16 @@ class GitObject(ABC):
     def issubmodule(self) -> bool:
         return S_ISGITLINK(self.mode)
 
+    @property
+    @abstractmethod
+    def size(self) -> int:
+        pass
+
+    @property
+    @abstractmethod
+    def sha(self) -> str:
+        pass
+
 
 class GitTrie:
     def __init__(self, tree: GitObject, rev: str):
@@ -116,9 +125,14 @@ class GitTrie:
         if not topdown:
             yield top, dirs, nondirs
 
-    def stat(self, key: tuple) -> os.stat_result:
+    def info(self, key: tuple) -> dict:
         obj = self.trie[key]
-        return os.stat_result((obj.mode, 0, 0, 0, 0, 0, 0, 0, 0, 0))
+        return {
+            "size": obj.size,
+            "type": "directory" if stat.S_ISDIR(obj.mode) else "file",
+            "sha": obj.sha,
+            "mode": obj.mode,
+        }
 
 
 @dataclass

dvc/stage/serialize.py

@@ -28,6 +28,7 @@ PARAM_PLOT = Output.PARAM_PLOT
 PARAM_PERSIST = Output.PARAM_PERSIST
 PARAM_CHECKPOINT = Output.PARAM_CHECKPOINT
 PARAM_DESC = Output.PARAM_DESC
+PARAM_REMOTE = Output.PARAM_REMOTE
 
 DEFAULT_PARAMS_FILE = ParamsDependency.DEFAULT_PARAMS_FILE
 
@@ -52,6 +53,8 @@ def _get_flags(out):
         yield from out.plot.items()
     if out.live and isinstance(out.live, dict):
         yield from out.live.items()
+    if out.remote:
+        yield PARAM_REMOTE, out.remote
 
 
 def _serialize_out(out):

dvc/state.py

@@ -53,7 +53,6 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         self.tmp_dir = tmp_dir
         self.root_dir = root_dir
         self.dvcignore = dvcignore
-        self.fs = LocalFileSystem()
 
         if not tmp_dir:
             return
@@ -85,7 +84,7 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         assert isinstance(hash_info, HashInfo)
         assert os.path.exists(path_info)
 
-        mtime, size = get_mtime_and_size(path_info, self.fs, self.dvcignore)
+        mtime, size = get_mtime_and_size(path_info, fs, self.dvcignore)
         inode = get_inode(path_info)
 
         logger.debug(
@@ -121,7 +120,7 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         if not os.path.exists(path):
             return None
 
-        mtime, size = get_mtime_and_size(path, self.fs, self.dvcignore)
+        mtime, size = get_mtime_and_size(path, fs, self.dvcignore)
         inode = get_inode(path)
 
         value = self.md5s.get(inode)
@@ -143,10 +142,10 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
 
         assert isinstance(path_info, str) or path_info.scheme == "local"
 
-        if not self.fs.exists(path_info):
+        if not fs.exists(path_info):
             return
 
-        mtime, _ = get_mtime_and_size(path_info, self.fs, self.dvcignore)
+        mtime, _ = get_mtime_and_size(path_info, fs, self.dvcignore)
         inode = get_inode(path_info)
         relative_path = relpath(path_info, self.root_dir)
 
@@ -168,11 +167,11 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
             for relative_path in ref:
                 path = os.path.join(self.root_dir, relative_path)
 
-                if path in used or not self.fs.exists(path):
+                if path in used or not fs.exists(path):
                     continue
 
                 inode = get_inode(path)
-                mtime, _ = get_mtime_and_size(path, self.fs, self.dvcignore)
+                mtime, _ = get_mtime_and_size(path, fs, self.dvcignore)
 
                 if ref[relative_path] == (inode, mtime):
                     logger.debug("Removing '%s' as unused link.", path)

dvc/utils/fs.py

@@ -47,22 +47,22 @@ def get_mtime_and_size(path, fs, dvcignore=None):
             walk_iterator = fs.walk_files(path)
         for file_path in walk_iterator:
             try:
-                stats = fs.stat(file_path)
+                stats = fs.info(file_path)
             except OSError as exc:
                 # NOTE: broken symlink case.
                 if exc.errno != errno.ENOENT:
                     raise
                 continue
-            size += stats.st_size
-            files_mtimes[os.fspath(file_path)] = stats.st_mtime
+            size += stats["size"]
+            files_mtimes[os.fspath(file_path)] = stats["mtime"]
 
         # We track file changes and moves, which cannot be detected with simply
         # max(mtime(f) for f in non_ignored_files)
         mtime = dict_md5(files_mtimes)
     else:
-        base_stat = fs.stat(path)
-        size = base_stat.st_size
-        mtime = base_stat.st_mtime
+        base_stat = fs.info(path)
+        size = base_stat["size"]
+        mtime = base_stat["mtime"]
         mtime = int(nanotime.timestamp(mtime))
 
     return str(mtime), size
@@ -193,33 +193,39 @@ def makedirs(path, exist_ok=False, mode=None):
         logger.trace("failed to chmod '%o' '%s'", mode, path, exc_info=True)
 
 
-def copyfile(src, dest, no_progress_bar=False, name=None):
+def copyfile(src, dest, callback=None, no_progress_bar=False, name=None):
     """Copy file with progress bar"""
-    from dvc.progress import Tqdm
-
     name = name if name else os.path.basename(dest)
     total = os.stat(src).st_size
 
     if os.path.isdir(dest):
         dest = os.path.join(dest, os.path.basename(src))
 
+    if callback:
+        callback.set_size(total)
+
     try:
         System.reflink(src, dest)
     except DvcException:
+        from dvc.progress import tdqm_or_callback_wrapped
+
         with open(src, "rb") as fsrc, open(dest, "wb+") as fdest:
-            with Tqdm.wrapattr(
+            with tdqm_or_callback_wrapped(
                 fdest,
                 "write",
-                desc=name,
+                total,
+                callback=callback,
                 disable=no_progress_bar,
-                total=total,
-                bytes=True,
-            ) as fdest_wrapped:
+                desc=name,
+            ) as wrapped:
                 while True:
                     buf = fsrc.read(LOCAL_CHUNK_SIZE)
                     if not buf:
                         break
-                    fdest_wrapped.write(buf)
+                    wrapped.write(buf)
+
+    if callback:
+        callback.absolute_update(total)
 
 
 def copy_fobj_to_file(fsrc, dest):

dvc/utils/http.py

@@ -24,13 +24,15 @@ def open_url(url, mode="r", encoding=None, **iter_opts):
 
 
 @contextmanager
-def iter_url(url, chunk_size=io.DEFAULT_BUFFER_SIZE):
+def iter_url(url, chunk_size=io.DEFAULT_BUFFER_SIZE, auth=None):
     """Iterate over chunks requested from url."""
     import requests
 
     def request(headers=None):
         the_url = url() if callable(url) else url
-        response = requests.get(the_url, stream=True, headers=headers)
+        response = requests.get(
+            the_url, stream=True, headers=headers, auth=auth
+        )
         if response.status_code == 404:
             raise FileNotFoundError(f"Can't open {the_url}")
         response.raise_for_status()

dvc/utils/serialize/__init__.py

@@ -14,6 +14,13 @@ LOADERS.update(
     {".toml": load_toml, ".json": load_json, ".py": load_py}  # noqa: F405
 )
 
+DUMPERS: DefaultDict[str, DumperFn] = defaultdict(  # noqa: F405
+    lambda: dump_yaml  # noqa: F405
+)
+DUMPERS.update(
+    {".toml": dump_toml, ".json": dump_json, ".py": dump_py}  # noqa: F405
+)
+
 MODIFIERS: DefaultDict[str, ModifierFn] = defaultdict(  # noqa: F405
     lambda: modify_yaml  # noqa: F405
 )

dvc/utils/serialize/_common.py

@@ -63,10 +63,16 @@ def _load_data(path: "AnyPath", parser: ParserFn, fs: "BaseFileSystem" = None):
             return parser(fd.read(), path)
 
 
-def _dump_data(path, data: Any, dumper: DumperFn, fs: "BaseFileSystem" = None):
+def _dump_data(
+    path,
+    data: Any,
+    dumper: DumperFn,
+    fs: "BaseFileSystem" = None,
+    **dumper_args,
+):
     open_fn = fs.open if fs else open
     with open_fn(path, "w+", encoding="utf-8") as fd:  # type: ignore
-        dumper(data, fd)
+        dumper(data, fd, **dumper_args)
 
 
 @contextmanager

dvc/utils/serialize/_json.py

@@ -19,8 +19,8 @@ def parse_json(text, path, **kwargs):
         return json.loads(text, **kwargs) or {}
 
 
-def dump_json(path, data, fs=None):
-    return _dump_data(path, data, dumper=json.dump, fs=fs)
+def dump_json(path, data, fs=None, **kwargs):
+    return _dump_data(path, data, dumper=json.dump, fs=fs, **kwargs)
 
 
 @contextmanager

dvc/utils/serialize/_toml.py

@@ -40,8 +40,8 @@ def _dump(data, stream):
     return toml.dump(data, stream, encoder=toml.TomlPreserveCommentEncoder())
 
 
-def dump_toml(path, data, fs=None):
-    return _dump_data(path, data, dumper=_dump, fs=fs)
+def dump_toml(path, data, fs=None, **kwargs):
+    return _dump_data(path, data, dumper=_dump, fs=fs, **kwargs)
 
 
 @contextmanager

dvc/utils/serialize/_yaml.py

@@ -23,17 +23,10 @@ def load_yaml(path, fs=None):
 def parse_yaml(text, path, typ="safe"):
     from ruamel.yaml import YAML
     from ruamel.yaml import YAMLError as _YAMLError
-    from ruamel.yaml.constructor import DuplicateKeyError
 
     yaml = YAML(typ=typ)
-    try:
-        with reraise(_YAMLError, YAMLFileCorruptedError(path)):
-            return yaml.load(text) or {}
-    except DuplicateKeyError as exc:
-        # NOTE: unfortunately this one doesn't inherit from YAMLError, so we
-        # have to catch it by-hand. See
-        # https://yaml.readthedocs.io/en/latest/api.html#duplicate-keys
-        raise YAMLError(path, exc.problem)
+    with reraise(_YAMLError, YAMLFileCorruptedError(path)):
+        return yaml.load(text) or {}
 
 
 def parse_yaml_for_update(text, path):
@@ -65,8 +58,8 @@ def _dump(data, stream):
     return yaml.dump(data, stream)
 
 
-def dump_yaml(path, data, fs=None):
-    return _dump_data(path, data, dumper=_dump, fs=fs)
+def dump_yaml(path, data, fs=None, **kwargs):
+    return _dump_data(path, data, dumper=_dump, fs=fs, **kwargs)
 
 
 def loads_yaml(s, typ="safe"):

dvc/utils/threadpool.py

@@ -0,0 +1,34 @@
+from concurrent import futures
+from itertools import islice
+from typing import Any, Callable, Iterable, Iterator, Set, TypeVar
+
+_T = TypeVar("_T")
+
+
+class ThreadPoolExecutor(futures.ThreadPoolExecutor):
+    _max_workers: int
+
+    @property
+    def max_workers(self) -> int:
+        return self._max_workers
+
+    def imap_unordered(
+        self, fn: Callable[..., _T], *iterables: Iterable[Any]
+    ) -> Iterator[_T]:
+        """Lazier version of map that does not preserve ordering of results.
+
+        It does not create all the futures at once to reduce memory usage.
+        """
+
+        def create_taskset(n: int) -> Set[futures.Future]:
+            return {self.submit(fn, *args) for args in islice(it, n)}
+
+        it = zip(*iterables)
+        tasks = create_taskset(self.max_workers * 5)
+        while tasks:
+            done, tasks = futures.wait(
+                tasks, return_when=futures.FIRST_COMPLETED
+            )
+            for fut in done:
+                yield fut.result()
+            tasks.update(create_taskset(len(done)))

dvc/version.py

@@ -6,7 +6,7 @@
 import os
 import subprocess
 
-_BASE_VERSION = "2.6.3"
+_BASE_VERSION = "2.7.4"
 
 
 def _generate_version(base_version):

requirements/azure.txt

@@ -0,0 +1,3 @@
+adlfs==2021.9.1
+azure-identity>=1.4.0
+knack

requirements/default.txt

@@ -0,0 +1,46 @@
+# See https://github.com/pyinstaller/pyinstaller/issues/1945
+ply>=3.9
+colorama>=0.3.9
+configobj>=5.0.6
+gitpython>3
+dulwich>=0.20.23
+pygit2>=1.5.0
+setuptools>=34.0.0
+nanotime>=0.5.2
+pyasn1>=0.4.1
+voluptuous>=0.11.7
+requests>=2.22.0
+# See https://github.com/bdcht/grandalf/issues/26
+grandalf==0.6
+distro>=1.3.0
+appdirs>=1.4.3
+ruamel.yaml>=0.17.11
+toml>=0.10.1
+funcy>=1.14
+pathspec>=0.6.0,<0.9.0
+shortuuid>=0.5.0
+tqdm>=4.45.0,<5
+packaging>=19.0
+zc.lockfile>=1.2.1
+flufl.lock>=3.2,<4
+win-unicode-console>=0.5; sys_platform == 'win32'
+pywin32>=225; sys_platform == 'win32' and python_version < '3.10'
+networkx>=2.5
+psutil>=5.8.0
+pydot>=1.2.4
+speedcopy>=2.0.1; python_version < '3.8' and sys_platform == 'win32'
+dataclasses==0.7; python_version < '3.7'
+importlib-metadata>=1.4; python_version < '3.8'
+flatten_dict>=0.4.1,<1
+tabulate>=0.8.7
+pygtrie>=2.3.2
+dpath>=2.0.2,<3
+shtab>=1.3.4,<2
+rich>=10.0.0
+dictdiffer>=0.8.1
+python-benedict>=0.21.1
+pyparsing==2.4.7
+typing_extensions>=3.10.0.2
+fsspec[http]>=2021.8.1
+aiohttp-retry==2.4.5
+diskcache>=5.2.1

requirements/gdrive.txt

@@ -0,0 +1 @@
+pydrive2[fsspec]>=1.9.4

requirements/gs.txt

@@ -0,0 +1 @@
+gcsfs==2021.8.1

requirements/hdfs.txt

@@ -0,0 +1 @@
+pyarrow>=2.0.0; python_version < '3.10'

requirements/oss.txt

@@ -0,0 +1 @@
+ossfs==2021.8.0

requirements/s3.txt

@@ -0,0 +1,2 @@
+s3fs==2021.8.1
+aiobotocore[boto3]>1.0.1

requirements/ssh.txt

@@ -0,0 +1 @@
+sshfs>=2021.8.1

requirements/ssh_gssapi.txt

@@ -0,0 +1 @@
+sshfs[gssapi]>=2021.8.1

requirements/terraform.txt

@@ -0,0 +1 @@
+tpi[ssh]>=0.0.0

requirements/tests.txt

@@ -1,28 +1,26 @@
 wheel==0.37.0
 
 # Test requirements
-pytest==6.2.4
+pytest==6.2.5
 pytest-cov==2.12.1
 pytest-xdist==2.3.0
 pytest-mock==3.6.1
 pytest-lazy-fixture==0.6.3
-pytest-timeout==1.4.2
 pytest-docker==0.10.3
 
 flaky==3.7.0
 mock==4.0.3
 rangehttpserver==1.2.0
-mock-ssh-server==0.9.0
-beautifulsoup4==4.9.3
+mock-ssh-server==0.9.1
 wget==3.2
 filelock==3.0.12
 wsgidav==3.1.1
 crc32c==2.2.post0
 xmltodict==0.12.0
 google-compute-engine==2.8.13
-google-cloud-storage==1.42.0
-
+google-cloud-storage==1.42.1
 urllib3==1.26.6
+dvclive>=0.2.1
 
 # required by collective.checkdocs
 Pygments==2.10.0
@@ -44,6 +42,3 @@ types-requests
 types-paramiko
 types-tabulate
 types-toml
-
-# pypi doesn't allow for direct dependencies
-# gdrivefs @ git+https://github.com/intake/gdrivefs.git

requirements/webdav.txt

@@ -0,0 +1 @@
+webdav4>=0.9.1

requirements/webhdfs.txt

@@ -0,0 +1 @@
+hdfs==2.6.0

scripts/pyinstaller/hooks/hook-dvc.py

@@ -14,6 +14,8 @@ datas += copy_metadata("boto3")
 datas += copy_metadata("ossfs")
 datas += copy_metadata("sshfs")
 datas += copy_metadata("webdav4")
+datas += copy_metadata("aiohttp")
+datas += copy_metadata("aiohttp_retry")
 
 # https://github.com/pypa/setuptools/issues/1963
 hiddenimports = ["pkg_resources.py2_warn"]

setup.cfg

@@ -17,8 +17,6 @@ known_first_party=dvc,tests
 line_length=79
 
 [tool:pytest]
-timeout = 600
-timeout_method = thread
 log_level = debug
 addopts = -ra
 markers =

setup.py

@@ -46,75 +46,25 @@ class build_py(_build_py):
         _build_py.run(self)
 
 
-install_requires = [
-    "ply>=3.9",  # See https://github.com/pyinstaller/pyinstaller/issues/1945
-    "colorama>=0.3.9",
-    "configobj>=5.0.6",
-    "gitpython>3",
-    "dulwich>=0.20.23",
-    "pygit2>=1.5.0",
-    "setuptools>=34.0.0",
-    "nanotime>=0.5.2",
-    "pyasn1>=0.4.1",
-    "voluptuous>=0.11.7",
-    "requests>=2.22.0",
-    "grandalf==0.6",
-    "distro>=1.3.0",
-    "appdirs>=1.4.3",
-    "ruamel.yaml>=0.16.1",
-    "toml>=0.10.1",
-    "funcy>=1.14",
-    "pathspec>=0.6.0,<0.9.0",
-    "shortuuid>=0.5.0",
-    "tqdm>=4.45.0,<5",
-    "packaging>=19.0",
-    "zc.lockfile>=1.2.1",
-    "flufl.lock>=3.2,<4",
-    "win-unicode-console>=0.5; sys_platform == 'win32'",
-    "pywin32>=225; sys_platform == 'win32' and python_version < '3.10'",
-    "networkx>=2.5",
-    "psutil>=5.8.0",
-    "pydot>=1.2.4",
-    "speedcopy>=2.0.1; python_version < '3.8' and sys_platform == 'win32'",
-    "dataclasses==0.7; python_version < '3.7'",
-    "importlib-metadata>=1.4; python_version < '3.8'",
-    "flatten_dict>=0.4.1,<1",
-    "tabulate>=0.8.7",
-    "pygtrie>=2.3.2",
-    "dpath>=2.0.1,<3",
-    "shtab>=1.3.4,<2",
-    "rich>=10.0.0",
-    "dictdiffer>=0.8.1",
-    "python-benedict>=0.21.1",
-    "pyparsing==2.4.7",
-    "typing_extensions>=3.7.4",
-    "fsspec>=2021.7.0",
-    "diskcache>=5.2.1",
-]
-
-
 # Extra dependencies for remote integrations
+requirements = {
+    path.stem: path.read_text().strip().splitlines()
+    for path in Path("requirements").glob("*.txt")
+}
 
-gs = ["gcsfs==2021.7.0"]
-gdrive = ["pydrive2[fsspec]>=1.9.1"]
-s3 = ["s3fs==2021.7.0", "aiobotocore[boto3]>1.0.1"]
-azure = ["adlfs==2021.7.1", "azure-identity>=1.4.0", "knack"]
-oss = ["ossfs==2021.7.5"]
-ssh = ["sshfs>=2021.7.1"]
-
-hdfs = ["pyarrow>=2.0.0; python_version < '3.10'"]
-webhdfs = ["hdfs==2.5.8"]
-webdav = ["webdav4>=0.9.0"]
 # gssapi should not be included in all_remotes, because it doesn't have wheels
 # for linux and mac, so it will fail to compile if user doesn't have all the
 # requirements, including kerberos itself. Once all the wheels are available,
 # we can start shipping it by default.
-ssh_gssapi = ["sshfs[gssapi]>=2021.7.1"]
-all_remotes = gs + s3 + azure + ssh + oss + gdrive + hdfs + webhdfs + webdav
 
-tests_requirements = (
-    Path("test_requirements.txt").read_text().strip().splitlines()
-)
+install_requires = requirements.pop("default")
+requirements["all"] = [
+    requirements
+    for key, requirements in requirements.items()
+    if key not in ("tests", "ssh_gssapi", "terraform")
+]
+requirements["tests"] += requirements["terraform"]
+requirements["dev"] = requirements["all"] + requirements["tests"]
 
 setup(
     name="dvc",
@@ -123,23 +73,12 @@ setup(
     long_description=open("README.rst", encoding="UTF-8").read(),
     author="Dmitry Petrov",
     author_email="[email protected]",
+    maintainer="Iterative",
+    maintainer_email="[email protected]",
     download_url="https://github.com/iterative/dvc",
     license="Apache License 2.0",
     install_requires=install_requires,
-    extras_require={
-        "all": all_remotes,
-        "gs": gs,
-        "gdrive": gdrive,
-        "s3": s3,
-        "azure": azure,
-        "oss": oss,
-        "ssh": ssh,
-        "ssh_gssapi": ssh_gssapi,
-        "hdfs": hdfs,
-        "webhdfs": webhdfs,
-        "webdav": webdav,
-        "tests": tests_requirements,
-    },
+    extras_require=requirements,
     keywords="data-science data-version-control machine-learning git"
     " developer-tools reproducibility collaboration ai",
     python_requires=">=3.6",

tests/conftest.py

@@ -5,6 +5,7 @@ import pytest
 
 from .dir_helpers import *  # noqa, pylint: disable=wildcard-import
 from .remotes import *  # noqa, pylint: disable=wildcard-import
+from .utils.scriptify import scriptify
 
 # Prevent updater and analytics from running their processes
 os.environ["DVC_TEST"] = "true"
@@ -155,3 +156,15 @@ def pytest_configure(config):
             enabled_remotes.discard(remote_name)
         if enabled:
             enabled_remotes.add(remote_name)
+
+
+@pytest.fixture()
+def custom_template(tmp_dir, dvc):
+    import shutil
+
+    template = tmp_dir / "custom_template.json"
+    shutil.copy(tmp_dir / ".dvc" / "plots" / "default.json", template)
+    return template
+
+
+scriptify_fixture = pytest.fixture(lambda: scriptify, name="scriptify")

tests/dir_helpers.py

@@ -49,12 +49,14 @@ import os
 import pathlib
 import sys
 from contextlib import contextmanager
+from functools import partialmethod
 from textwrap import dedent
 
 import pytest
 from funcy import lmap, retry
 
 from dvc.logger import disable_other_loggers
+from dvc.utils import serialize
 from dvc.utils.fs import makedirs
 
 __all__ = [
@@ -67,6 +69,8 @@ __all__ = [
     "erepo_dir",
     "git_dir",
     "git_init",
+    "git_upstream",
+    "git_downstream",
 ]
 
 
@@ -248,6 +252,22 @@ class TmpDir(pathlib.Path):
     def hash_to_path_info(self, hash_):
         return self / hash_[0:2] / hash_[2:]
 
+    def dump(self, *args, **kwargs):
+        return serialize.DUMPERS[self.suffix](self, *args, **kwargs)
+
+    def parse(self, *args, **kwargs):
+        return serialize.LOADERS[self.suffix](self, *args, **kwargs)
+
+    def modify(self, *args, **kwargs):
+        return serialize.MODIFIERS[self.suffix](self, *args, **kwargs)
+
+    load_yaml = partialmethod(serialize.load_yaml)
+    dump_yaml = partialmethod(serialize.dump_yaml)
+    load_json = partialmethod(serialize.load_json)
+    dump_json = partialmethod(serialize.dump_json)
+    load_toml = partialmethod(serialize.load_toml)
+    dump_toml = partialmethod(serialize.dump_toml)
+
 
 def _coerce_filenames(filenames):
     if isinstance(filenames, (str, bytes, pathlib.PurePath)):
@@ -407,3 +427,23 @@ def git_dir(make_tmp_dir):
     path = make_tmp_dir("git-erepo", scm=True)
     path.scm.commit("init repo")
     return path
+
+
+@pytest.fixture
+def git_upstream(tmp_dir, erepo_dir, git_dir, request):
+    remote = erepo_dir if "dvc" in request.fixturenames else git_dir
+    url = "file://{}".format(remote.resolve().as_posix())
+    tmp_dir.scm.gitpython.repo.create_remote("upstream", url)
+    remote.remote = "upstream"
+    remote.url = url
+    return remote
+
+
+@pytest.fixture
+def git_downstream(tmp_dir, erepo_dir, git_dir, request):
+    remote = erepo_dir if "dvc" in request.fixturenames else git_dir
+    url = "file://{}".format(tmp_dir.resolve().as_posix())
+    remote.scm.gitpython.repo.create_remote("upstream", url)
+    remote.remote = "upstream"
+    remote.url = url
+    return remote

tests/func/experiments/conftest.py

@@ -91,18 +91,14 @@ def checkpoint_stage(tmp_dir, scm, dvc, mocker):
 
 
 @pytest.fixture
-def git_upstream(tmp_dir, erepo_dir):
-    url = "file://{}".format(erepo_dir.resolve().as_posix())
-    tmp_dir.scm.gitpython.repo.create_remote("upstream", url)
-    erepo_dir.remote = "upstream"
-    erepo_dir.url = url
-    return erepo_dir
+def http_auth_patch(mocker):
+    from dulwich.client import HTTPUnauthorized
 
+    url = "https://0.0.0.0"
+    client = mocker.MagicMock()
+    client.get_refs.side_effect = HTTPUnauthorized("", url)
+    client.send_pack.side_effect = HTTPUnauthorized("", url)
 
-@pytest.fixture
-def git_downstream(tmp_dir, erepo_dir):
-    url = "file://{}".format(tmp_dir.resolve().as_posix())
-    erepo_dir.scm.gitpython.repo.create_remote("upstream", url)
-    erepo_dir.remote = "upstream"
-    erepo_dir.url = url
-    return erepo_dir
+    patch = mocker.patch("dulwich.client.get_transport_and_path")
+    patch.return_value = (client, url)
+    return url

tests/func/experiments/test_experiments.py

@@ -657,3 +657,12 @@ def test_modified_data_dep(tmp_dir, scm, dvc, workspace, params, target):
     if workspace:
         assert (tmp_dir / "metrics.yaml").read_text().strip() == params
         assert (tmp_dir / "data").read_text().strip() == "modified"
+
+
+def test_exp_run_recursive(tmp_dir, scm, dvc, run_copy_metrics):
+    tmp_dir.dvc_gen("metric_t.json", '{"foo": 1}')
+    run_copy_metrics(
+        "metric_t.json", "metric.json", metrics=["metric.json"], no_exec=True
+    )
+    assert dvc.experiments.run(".", recursive=True)
+    assert (tmp_dir / "metric.json").parse() == {"foo": 1}

tests/func/experiments/test_init.py

@@ -0,0 +1,34 @@
+import os
+
+from dvc.command.experiments import CmdExperimentsInit
+from dvc.main import main
+from dvc.utils.serialize import load_yaml
+
+
+def test_init(tmp_dir, dvc):
+    tmp_dir.gen(
+        {
+            CmdExperimentsInit.CODE: {"copy.py": ""},
+            "data": "data",
+            "params.yaml": '{"foo": 1}',
+            "dvclive": {},
+            "plots": {},
+        }
+    )
+    code_path = os.path.join(CmdExperimentsInit.CODE, "copy.py")
+    script = f"python {code_path}"
+
+    assert main(["exp", "init", script]) == 0
+    assert load_yaml(tmp_dir / "dvc.yaml") == {
+        "stages": {
+            "default": {
+                "cmd": script,
+                "deps": ["data", "src"],
+                "live": {"dvclive": {"html": True, "summary": True}},
+                "metrics": [{"metrics.json": {"cache": False}}],
+                "outs": ["models"],
+                "params": ["foo"],
+                "plots": [{"plots": {"cache": False}}],
+            }
+        }
+    }

tests/func/experiments/test_remote.py

@@ -252,3 +252,37 @@ def test_push_pull_cache(
         path = os.path.join(dvc.odb.local.cache_dir, hash_[:2], hash_[2:])
         assert os.path.exists(path)
         assert open(path).read() == str(x)
+
+
+def test_auth_error_list(tmp_dir, scm, dvc, http_auth_patch):
+    from dvc.scm.base import GitAuthError
+
+    with pytest.raises(
+        GitAuthError,
+        match=f"HTTP Git authentication is not supported: '{http_auth_patch}'",
+    ):
+        dvc.experiments.ls(git_remote=http_auth_patch)
+
+
+def test_auth_error_pull(tmp_dir, scm, dvc, http_auth_patch):
+    from dvc.scm.base import GitAuthError
+
+    with pytest.raises(
+        GitAuthError,
+        match=f"HTTP Git authentication is not supported: '{http_auth_patch}'",
+    ):
+        dvc.experiments.pull(http_auth_patch, "foo")
+
+
+def test_auth_error_push(tmp_dir, scm, dvc, exp_stage, http_auth_patch):
+    from dvc.scm.base import GitAuthError
+
+    results = dvc.experiments.run(exp_stage.addressing, params=["foo=2"])
+    exp = first(results)
+    ref_info = first(exp_refs_by_rev(scm, exp))
+
+    with pytest.raises(
+        GitAuthError,
+        match=f"HTTP Git authentication is not supported: '{http_auth_patch}'",
+    ):
+        dvc.experiments.push(http_auth_patch, ref_info.name)

tests/func/experiments/test_remove.py

@@ -72,3 +72,47 @@ def test_remove_special_queued_experiments(tmp_dir, scm, dvc, exp_stage):
     assert rev3 in dvc.experiments.stash_revs
     assert scm.get_ref(str(ref_info1)) is None
     assert scm.get_ref(str(ref_info2)) is not None
+
+
+def test_remove_all(tmp_dir, scm, dvc, exp_stage):
+    results = dvc.experiments.run(exp_stage.addressing, params=["foo=1"])
+    ref_info1 = first(exp_refs_by_rev(scm, first(results)))
+    dvc.experiments.run(exp_stage.addressing, params=["foo=2"], queue=True)
+    scm.add(["dvc.yaml", "dvc.lock", "copy.py", "params.yaml", "metrics.yaml"])
+    scm.commit("update baseline")
+
+    results = dvc.experiments.run(exp_stage.addressing, params=["foo=3"])
+    ref_info2 = first(exp_refs_by_rev(scm, first(results)))
+    dvc.experiments.run(exp_stage.addressing, params=["foo=4"], queue=True)
+
+    removed = dvc.experiments.remove(clear_all=True)
+    assert removed == 2
+    assert len(dvc.experiments.stash) == 2
+    assert scm.get_ref(str(ref_info2)) is None
+    assert scm.get_ref(str(ref_info1)) is None
+
+
+@pytest.mark.parametrize("use_url", [True, False])
+def test_remove_remote(tmp_dir, scm, dvc, exp_stage, git_upstream, use_url):
+    remote = git_upstream.url if use_url else git_upstream.remote
+
+    ref_info_list = []
+    exp_list = []
+    for i in range(3):
+        results = dvc.experiments.run(
+            exp_stage.addressing, params=[f"foo={i}"]
+        )
+        exp = first(results)
+        exp_list.append(exp)
+        ref_info = first(exp_refs_by_rev(scm, exp))
+        ref_info_list.append(ref_info)
+        dvc.experiments.push(remote, ref_info.name)
+        assert git_upstream.scm.get_ref(str(ref_info)) == exp
+
+    dvc.experiments.remove(
+        remote=remote, exp_names=[str(ref_info_list[0]), ref_info_list[1].name]
+    )
+
+    assert git_upstream.scm.get_ref(str(ref_info_list[0])) is None
+    assert git_upstream.scm.get_ref(str(ref_info_list[1])) is None
+    assert git_upstream.scm.get_ref(str(ref_info_list[2])) == exp_list[2]