dvc: bump to 1.2.2

Turns out macs are not happy with double-forks without exec.

Fixes #4294

.codeclimate.yml

@@ -1,19 +0,0 @@
-version: "2"
-checks:
-    argument-count:
-        config:
-            threshold: 8
-    complex-logic:
-        config:
-            threshold: 6
-    file-lines:
-        config:
-            threshold: 350
-    method-complexity:
-        config:
-            threshold: 15
-    method-count:
-        config:
-            threshold: 30
-exclude_patterns:
-    - tests

.deepsource.toml

@@ -1,11 +0,0 @@
-version = 1
-
-test_patterns = [
-  'tests/**'
-]
-
-[[analyzers]]
-name = 'python'
-enabled = true
-runtime_version = '3.x.x'
-

.dvc/.gitignore

@@ -9,3 +9,4 @@
 /pkg
 /repos
 /tmp
+/experiments

.github/PULL_REQUEST_TEMPLATE.md

@@ -2,6 +2,4 @@
 
 * [ ] 📖 If this PR requires [documentation](https://dvc.org/doc) updates, I have created a separate PR (or issue, at least) in [dvc.org](https://github.com/iterative/dvc.org) and linked it here.
 
-* [ ] ❌ I will check DeepSource, CodeClimate, and other sanity checks below. (We consider them recommendatory and don't expect everything to be addressed. Please fix things that actually improve code or fix bugs.)
-
 Thank you for the contribution - we'll try to review it as soon as possible. 🙏

.pylintrc

@@ -15,7 +15,11 @@ ignore-patterns=
 
 # Python code to execute, usually for sys.path manipulation such as
 # pygtk.require().
-#init-hook=
+init-hook="import sys; \
+           from pathlib import Path; \
+           from pylint.config import find_pylintrc; \
+           sys.path.append(Path(find_pylintrc()).parent / 'tests');"
+
 
 # Use multiple processes to speed up Pylint. Specifying 0 will auto-detect the
 # number of processors available to use.
@@ -28,7 +32,7 @@ limit-inference-results=100
 
 # List of plugins (as comma separated values of python modules names) to load,
 # usually to register additional checkers.
-load-plugins=
+load-plugins=pylint_pytest,pylint_plugin_disable
 
 # Pickle collected data for later comparisons.
 persistent=yes
@@ -60,41 +64,127 @@ confidence=
 # --enable=similarities". If you want to run only the classes checker, but have
 # no Warning level messages displayed, use "--disable=all --enable=classes
 # --disable=W".
-disable=all
+disable=format,
+        # ^ most of them are covered by flake8,
+        # and are conventions rather than warnings or errors
+        refactoring,  # will be noisy if it starts giving suggestions
+        spelling,
+        # design is too-few-* and too-many-* checks, too-much-* for us right now
+        design,
+        # following were generated by default
+        # most of them are redundant Py3 checks for Py2+3 codebase
+        print-statement,
+        parameter-unpacking,
+        unpacking-in-except,
+        old-raise-syntax,
+        backtick,
+        long-suffix,
+        old-ne-operator,
+        old-octal-literal,
+        import-star-module-level,
+        non-ascii-bytes-literal,
+        raw-checker-failed,
+        bad-inline-option,
+        locally-disabled,
+        file-ignored,
+        suppressed-message,
+        useless-suppression,
+        deprecated-pragma,
+        use-symbolic-message-instead,
+        apply-builtin,
+        basestring-builtin,
+        buffer-builtin,
+        cmp-builtin,
+        coerce-builtin,
+        execfile-builtin,
+        file-builtin,
+        long-builtin,
+        raw_input-builtin,
+        reduce-builtin,
+        standarderror-builtin,
+        unicode-builtin,
+        xrange-builtin,
+        coerce-method,
+        delslice-method,
+        getslice-method,
+        setslice-method,
+        no-absolute-import,
+        old-division,
+        dict-iter-method,
+        dict-view-method,
+        next-method-called,
+        metaclass-assignment,
+        indexing-exception,
+        raising-string,
+        reload-builtin,
+        oct-method,
+        hex-method,
+        nonzero-method,
+        cmp-method,
+        input-builtin,
+        round-builtin,
+        intern-builtin,
+        unichr-builtin,
+        map-builtin-not-iterating,
+        zip-builtin-not-iterating,
+        range-builtin-not-iterating,
+        filter-builtin-not-iterating,
+        using-cmp-argument,
+        eq-without-hash,
+        div-method,
+        idiv-method,
+        rdiv-method,
+        exception-message-attribute,
+        invalid-str-codec,
+        sys-max-int,
+        bad-python3-import,
+        deprecated-string-function,
+        deprecated-str-translate-call,
+        deprecated-itertools-function,
+        deprecated-types-field,
+        next-method-defined,
+        dict-items-not-iterating,
+        dict-keys-not-iterating,
+        dict-values-not-iterating,
+        deprecated-operator-function,
+        deprecated-urllib-function,
+        xreadlines-attribute,
+        deprecated-sys-function,
+        exception-escape,
+        comprehension-escape,
+        # custom disables start here
+        no-self-use,
+        invalid-name,
+        # already a check enabled for `wildcard-import`
+        unused-wildcard-import,
+        misplaced-comparison-constant,
+        # probably soon: https://github.com/iterative/dvc/issues/1843
+        logging-format-interpolation,
+        logging-fstring-interpolation,
+        # we don't have proper docstrings yet
+        # even if we do, we'll be doing it gradually
+        missing-function-docstring,
+        missing-module-docstring,
+        missing-class-docstring,
+        duplicate-code,
+        # too noisy, and detects issues even when import is inside a function
+        # see: https://github.com/PyCQA/pylint/issues/850
+        cyclic-import,
+        # isort should cover these:
+        wrong-import-order,
+        wrong-import-position,
+        ungrouped-imports,
+        multiple-imports,
+        # we use it to speedup/optimize
+        import-outside-toplevel,
+        fixme,
+
 
 # Enable the message, report, category or checker with the given id(s). You can
 # either give multiple identifier separated by comma (,) or put this option
 # multiple time (only on the command line, not in the configuration file where
 # it should appear only once). See also the "--disable" option for examples.
-enable=print-statement,
-       c-extension-no-member,
-       import-error,
-       import-self,
-       reimported,
-       deprecated-module,
-       wildcard-import,
-       useless-import-alias,
-       unpacking-in-except,
-       parameter-unpacking,
-       unpacking-non-sequence,
-       invalid-all-object,
-       no-name-in-module,
-       undefined-variable,
-       undefined-all-variable,
-       used-before-assignment,
-       cell-var-from-loop,
-       global-variable-undefined,
-       self-cls-assignment,
-       unbalanced-tuple-unpacking,
-       possibly-unused-variable,
-       redefined-builtin,
-       redefine-in-handler,
-       unused-import,
-
-# Needs Cleanup
-# unused-argument,
-# redefined-outer-name,
-
+enable=c-extension-no-member
 
 
 [REPORTS]
@@ -175,7 +265,7 @@ contextmanager-decorators=contextlib.contextmanager
 # List of members which are set dynamically and missed by pylint inference
 # system, and so shouldn't trigger E1101 when accessed. Python regular
 # expressions are accepted.
-generated-members=
+generated-members=pytest.lazy_fixture,logging.TRACE,logger.trace,sys.getwindowsversion
 
 # Tells whether missing members accessed in mixin class should be ignored. A
 # mixin class is detected if its name ends with "mixin" (case insensitive).
@@ -196,7 +286,7 @@ ignore-on-opaque-inference=yes
 # List of class names for which member attributes should not be checked (useful
 # for classes with dynamically set attributes). This supports the use of
 # qualified names.
-ignored-classes=optparse.Values,thread._local,_thread._local
+ignored-classes=optparse.Values,thread._local,_thread._local,Dvcfile
 
 # List of module names for which member attributes should not be checked
 # (useful for modules/projects where namespaces are manipulated during runtime
@@ -237,7 +327,7 @@ dummy-variables-rgx=_+$|(_[a-zA-Z0-9_]*[a-zA-Z0-9]+?$)|dummy|^ignored_|^unused_
 
 # Argument names that match this expression will be ignored. Default to name
 # with leading underscore.
-ignored-argument-names=_.*|^ignored_|^unused_
+ignored-argument-names=_.*|^ignored_|^unused_|args|kwargs
 
 # Tells whether we should check for unused import in __init__ files.
 init-import=no

.travis.yml

@@ -160,11 +160,3 @@ deploy:
       condition: "$(./scripts/ci/deploy_condition.sh dvc_*.snap) && ($TRAVIS_BRANCH = master || -n $TRAVIS_TAG) && -n $SNAP_CHANNEL"
       repo: iterative/dvc
       stage: build
-  - provider: script
-    skip_cleanup: true
-    script: "(echo $SNAP_TOKEN | snapcraft login --with -) && timeout 600 snapcraft push dvc_*.snap --release $SNAP_CHANNEL_MAJOR || echo timed out"
-    on:
-      all_branches: true
-      condition: "$(./scripts/ci/deploy_condition.sh dvc_*.snap) && ($TRAVIS_BRANCH = master || -n $TRAVIS_TAG) && -n $SNAP_CHANNEL_MAJOR"
-      repo: iterative/dvc
-      stage: build

README.rst

@@ -10,7 +10,7 @@
 
 |Release| |CI| |Maintainability| |Coverage| |Donate| |DOI|
 
-|Snap| |Choco| |Brew| |Conda| |PyPI| |Packages|
+|PyPI| |Packages| |Brew| |Conda| |Choco| |Snap|
 
 |
 
@@ -131,8 +131,6 @@ Conda (Anaconda)
 
    conda install -c conda-forge dvc
 
-Currently, this includes support for Python versions 2.7, 3.6 and 3.7.
-
 pip (PyPI)
 ----------
 

dvc/api.py

@@ -27,10 +27,10 @@ def get_url(path, repo=None, rev=None, remote=None):
     """
     with _make_repo(repo, rev=rev) as _repo:
         if not isinstance(_repo, Repo):
-            raise UrlNotDvcRepoError(_repo.url)
+            raise UrlNotDvcRepoError(_repo.url)  # pylint: disable=no-member
         out = _repo.find_out_by_relpath(path)
         remote_obj = _repo.cloud.get_remote(remote)
-        return str(remote_obj.hash_to_path_info(out.checksum))
+        return str(remote_obj.tree.hash_to_path_info(out.checksum))
 
 
 def open(  # noqa, pylint: disable=redefined-builtin
@@ -60,7 +60,9 @@ def open(  # noqa, pylint: disable=redefined-builtin
 
 
 class _OpenContextManager(GCM):
-    def __init__(self, func, args, kwds):
+    def __init__(
+        self, func, args, kwds
+    ):  # pylint: disable=super-init-not-called
         self.gen = func(*args, **kwds)
         self.func, self.args, self.kwds = func, args, kwds
 

dvc/cache/__init__.py

@@ -28,8 +28,8 @@ def _make_remote_property(name):
     """
 
     def getter(self):
-        from dvc.remote import get_cloud_tree
-        from dvc.remote.base import CloudCache
+        from ..tree import get_cloud_tree
+        from .base import CloudCache
 
         remote = self.config.get(name)
         if not remote:
@@ -52,8 +52,8 @@ class Cache:
     CACHE_DIR = "cache"
 
     def __init__(self, repo):
-        from dvc.remote import get_cloud_tree
-        from dvc.remote.local import LocalCache
+        from ..tree import get_cloud_tree
+        from .local import LocalCache
 
         self.repo = repo
         self.config = config = repo.config["cache"]
@@ -106,6 +106,7 @@ class NamedCacheItem:
 
 
 class NamedCache:
+    # pylint: disable=protected-access
     def __init__(self):
         self._items = defaultdict(lambda: defaultdict(NamedCacheItem))
         self.external = defaultdict(set)

dvc/cache/base.py

@@ -0,0 +1,554 @@
+import json
+import logging
+from copy import copy
+
+from shortuuid import uuid
+
+import dvc.prompt as prompt
+from dvc.exceptions import CheckoutError, ConfirmRemoveError, DvcException
+from dvc.path_info import WindowsPathInfo
+from dvc.progress import Tqdm
+from dvc.remote.slow_link_detection import slow_link_guard
+
+from ..tree.base import RemoteActionNotImplemented
+
+logger = logging.getLogger(__name__)
+
+STATUS_OK = 1
+STATUS_MISSING = 2
+STATUS_NEW = 3
+STATUS_DELETED = 4
+
+STATUS_MAP = {
+    # (local_exists, remote_exists)
+    (True, True): STATUS_OK,
+    (False, False): STATUS_MISSING,
+    (True, False): STATUS_NEW,
+    (False, True): STATUS_DELETED,
+}
+
+
+class DirCacheError(DvcException):
+    def __init__(self, hash_):
+        super().__init__(
+            f"Failed to load dir cache for hash value: '{hash_}'."
+        )
+
+
+class CloudCache:
+    """Cloud cache class."""
+
+    DEFAULT_CACHE_TYPES = ["copy"]
+    CACHE_MODE = None
+
+    def __init__(self, tree):
+        self.tree = tree
+        self.repo = tree.repo
+
+        self.cache_types = tree.config.get("type") or copy(
+            self.DEFAULT_CACHE_TYPES
+        )
+        self.cache_type_confirmed = False
+        self._dir_info = {}
+
+    def get_dir_cache(self, hash_):
+        assert hash_
+
+        dir_info = self._dir_info.get(hash_)
+        if dir_info:
+            return dir_info
+
+        try:
+            dir_info = self.load_dir_cache(hash_)
+        except DirCacheError:
+            dir_info = []
+
+        self._dir_info[hash_] = dir_info
+        return dir_info
+
+    def load_dir_cache(self, hash_):
+        path_info = self.tree.hash_to_path_info(hash_)
+
+        try:
+            with self.tree.open(path_info, "r") as fobj:
+                d = json.load(fobj)
+        except (ValueError, FileNotFoundError) as exc:
+            raise DirCacheError(hash_) from exc
+
+        if not isinstance(d, list):
+            logger.error(
+                "dir cache file format error '%s' [skipping the file]",
+                path_info,
+            )
+            return []
+
+        if self.tree.PATH_CLS == WindowsPathInfo:
+            # only need to convert it for Windows
+            for info in d:
+                # NOTE: here is a BUG, see comment to .as_posix() below
+                relpath = info[self.tree.PARAM_RELPATH]
+                info[self.tree.PARAM_RELPATH] = relpath.replace(
+                    "/", self.tree.PATH_CLS.sep
+                )
+
+        return d
+
+    def changed(self, path_info, hash_info):
+        """Checks if data has changed.
+
+        A file is considered changed if:
+            - It doesn't exist on the working directory (was unlinked)
+            - Hash value is not computed (saving a new file)
+            - The hash value stored is different from the given one
+            - There's no file in the cache
+
+        Args:
+            path_info: dict with path information.
+            hash: expected hash value for this data.
+
+        Returns:
+            bool: True if data has changed, False otherwise.
+        """
+
+        logger.debug(
+            "checking if '%s'('%s') has changed.", path_info, hash_info
+        )
+
+        if not self.tree.exists(path_info):
+            logger.debug("'%s' doesn't exist.", path_info)
+            return True
+
+        hash_ = hash_info.get(self.tree.PARAM_CHECKSUM)
+        if hash_ is None:
+            logger.debug("hash value for '%s' is missing.", path_info)
+            return True
+
+        if self.changed_cache(hash_):
+            logger.debug("cache for '%s'('%s') has changed.", path_info, hash_)
+            return True
+
+        actual = self.tree.get_hash(path_info)
+        if hash_ != actual:
+            logger.debug(
+                "hash value '%s' for '%s' has changed (actual '%s').",
+                hash_,
+                actual,
+                path_info,
+            )
+            return True
+
+        logger.debug("'%s' hasn't changed.", path_info)
+        return False
+
+    def link(self, from_info, to_info):
+        self._link(from_info, to_info, self.cache_types)
+
+    def _link(self, from_info, to_info, link_types):
+        assert self.tree.isfile(from_info)
+
+        self.tree.makedirs(to_info.parent)
+
+        self._try_links(from_info, to_info, link_types)
+
+    def _verify_link(self, path_info, link_type):
+        if self.cache_type_confirmed:
+            return
+
+        is_link = getattr(self.tree, f"is_{link_type}", None)
+        if is_link and not is_link(path_info):
+            self.tree.remove(path_info)
+            raise DvcException(f"failed to verify {link_type}")
+
+        self.cache_type_confirmed = True
+
+    @slow_link_guard
+    def _try_links(self, from_info, to_info, link_types):
+        while link_types:
+            link_method = getattr(self.tree, link_types[0])
+            try:
+                self._do_link(from_info, to_info, link_method)
+                self._verify_link(to_info, link_types[0])
+                return
+
+            except DvcException as exc:
+                logger.debug(
+                    "Cache type '%s' is not supported: %s", link_types[0], exc
+                )
+                del link_types[0]
+
+        raise DvcException("no possible cache types left to try out.")
+
+    def _do_link(self, from_info, to_info, link_method):
+        if self.tree.exists(to_info):
+            raise DvcException(f"Link '{to_info}' already exists!")
+
+        link_method(from_info, to_info)
+
+        logger.debug(
+            "Created '%s': %s -> %s", self.cache_types[0], from_info, to_info,
+        )
+
+    def _save_file(self, path_info, tree, hash_, save_link=True, **kwargs):
+        assert hash_
+
+        cache_info = self.tree.hash_to_path_info(hash_)
+        if tree == self.tree:
+            if self.changed_cache(hash_):
+                self.tree.move(path_info, cache_info, mode=self.CACHE_MODE)
+                self.link(cache_info, path_info)
+            elif self.tree.iscopy(path_info) and self._cache_is_copy(
+                path_info
+            ):
+                # Default relink procedure involves unneeded copy
+                self.tree.unprotect(path_info)
+            else:
+                self.tree.remove(path_info)
+                self.link(cache_info, path_info)
+
+            if save_link:
+                self.tree.state.save_link(path_info)
+            # we need to update path and cache, since in case of reflink,
+            # or copy cache type moving original file results in updates on
+            # next executed command, which causes md5 recalculation
+            self.tree.state.save(path_info, hash_)
+        else:
+            if self.changed_cache(hash_):
+                with tree.open(path_info, mode="rb") as fobj:
+                    # if tree has fetch enabled, DVC out will be fetched on
+                    # open and we do not need to read/copy any data
+                    if not (
+                        tree.isdvc(path_info, strict=False) and tree.fetch
+                    ):
+                        self.tree.copy_fobj(fobj, cache_info)
+                callback = kwargs.get("download_callback")
+                if callback:
+                    callback(1)
+
+        self.tree.state.save(cache_info, hash_)
+        return {self.tree.PARAM_CHECKSUM: hash_}
+
+    def _cache_is_copy(self, path_info):
+        """Checks whether cache uses copies."""
+        if self.cache_type_confirmed:
+            return self.cache_types[0] == "copy"
+
+        if set(self.cache_types) <= {"copy"}:
+            return True
+
+        workspace_file = path_info.with_name("." + uuid())
+        test_cache_file = self.tree.path_info / ".cache_type_test_file"
+        if not self.tree.exists(test_cache_file):
+            with self.tree.open(test_cache_file, "wb") as fobj:
+                fobj.write(bytes(1))
+        try:
+            self.link(test_cache_file, workspace_file)
+        finally:
+            self.tree.remove(workspace_file)
+            self.tree.remove(test_cache_file)
+
+        self.cache_type_confirmed = True
+        return self.cache_types[0] == "copy"
+
+    def _save_dir(self, path_info, tree, hash_, save_link=True, **kwargs):
+        dir_info = self.get_dir_cache(hash_)
+        for entry in Tqdm(
+            dir_info, desc="Saving " + path_info.name, unit="file"
+        ):
+            entry_info = path_info / entry[self.tree.PARAM_RELPATH]
+            entry_hash = entry[self.tree.PARAM_CHECKSUM]
+            self._save_file(
+                entry_info, tree, entry_hash, save_link=False, **kwargs
+            )
+
+        if save_link:
+            self.tree.state.save_link(path_info)
+        if self.tree.exists(path_info):
+            self.tree.state.save(path_info, hash_)
+
+        cache_info = self.tree.hash_to_path_info(hash_)
+        self.tree.state.save(cache_info, hash_)
+        return {self.tree.PARAM_CHECKSUM: hash_}
+
+    def save(self, path_info, tree, hash_info, save_link=True, **kwargs):
+        if path_info.scheme != self.tree.scheme:
+            raise RemoteActionNotImplemented(
+                f"save {path_info.scheme} -> {self.tree.scheme}",
+                self.tree.scheme,
+            )
+
+        if not hash_info:
+            hash_info = self.tree.save_info(path_info, tree=tree, **kwargs)
+        hash_ = hash_info[self.tree.PARAM_CHECKSUM]
+        return self._save(path_info, tree, hash_, save_link, **kwargs)
+
+    def _save(self, path_info, tree, hash_, save_link=True, **kwargs):
+        to_info = self.tree.hash_to_path_info(hash_)
+        logger.debug("Saving '%s' to '%s'.", path_info, to_info)
+
+        if tree.isdir(path_info):
+            return self._save_dir(path_info, tree, hash_, save_link, **kwargs)
+        return self._save_file(path_info, tree, hash_, save_link, **kwargs)
+
+    # Override to return path as a string instead of PathInfo for clouds
+    # which support string paths (see local)
+    def hash_to_path(self, hash_):
+        return self.tree.hash_to_path_info(hash_)
+
+    def changed_cache_file(self, hash_):
+        """Compare the given hash with the (corresponding) actual one.
+
+        - Use `State` as a cache for computed hashes
+            + The entries are invalidated by taking into account the following:
+                * mtime
+                * inode
+                * size
+                * hash
+
+        - Remove the file from cache if it doesn't match the actual hash
+        """
+        # Prefer string path over PathInfo when possible due to performance
+        cache_info = self.hash_to_path(hash_)
+        if self.tree.is_protected(cache_info):
+            logger.debug(
+                "Assuming '%s' is unchanged since it is read-only", cache_info
+            )
+            return False
+
+        actual = self.tree.get_hash(cache_info)
+
+        logger.debug(
+            "cache '%s' expected '%s' actual '%s'", cache_info, hash_, actual,
+        )
+
+        if not hash_ or not actual:
+            return True
+
+        if actual.split(".")[0] == hash_.split(".")[0]:
+            # making cache file read-only so we don't need to check it
+            # next time
+            self.tree.protect(cache_info)
+            return False
+
+        if self.tree.exists(cache_info):
+            logger.warning("corrupted cache file '%s'.", cache_info)
+            self.tree.remove(cache_info)
+
+        return True
+
+    def _changed_dir_cache(self, hash_, path_info=None, filter_info=None):
+        if self.changed_cache_file(hash_):
+            return True
+
+        for entry in self.get_dir_cache(hash_):
+            entry_hash = entry[self.tree.PARAM_CHECKSUM]
+
+            if path_info and filter_info:
+                entry_info = path_info / entry[self.tree.PARAM_RELPATH]
+                if not entry_info.isin_or_eq(filter_info):
+                    continue
+
+            if self.changed_cache_file(entry_hash):
+                return True
+
+        return False
+
+    def changed_cache(self, hash_, path_info=None, filter_info=None):
+        if self.tree.is_dir_hash(hash_):
+            return self._changed_dir_cache(
+                hash_, path_info=path_info, filter_info=filter_info
+            )
+        return self.changed_cache_file(hash_)
+
+    def already_cached(self, path_info):
+        current = self.tree.get_hash(path_info)
+
+        if not current:
+            return False
+
+        return not self.changed_cache(current)
+
+    def safe_remove(self, path_info, force=False):
+        if not self.tree.exists(path_info):
+            return
+
+        if not force and not self.already_cached(path_info):
+            msg = (
+                "file '{}' is going to be removed."
+                " Are you sure you want to proceed?".format(str(path_info))
+            )
+
+            if not prompt.confirm(msg):
+                raise ConfirmRemoveError(str(path_info))
+
+        self.tree.remove(path_info)
+
+    def _checkout_file(
+        self, path_info, hash_, force, progress_callback=None, relink=False
+    ):
+        """The file is changed we need to checkout a new copy"""
+        added, modified = True, False
+        cache_info = self.tree.hash_to_path_info(hash_)
+        if self.tree.exists(path_info):
+            logger.debug("data '%s' will be replaced.", path_info)
+            self.safe_remove(path_info, force=force)
+            added, modified = False, True
+
+        self.link(cache_info, path_info)
+        self.tree.state.save_link(path_info)
+        self.tree.state.save(path_info, hash_)
+        if progress_callback:
+            progress_callback(str(path_info))
+
+        return added, modified and not relink
+
+    def _checkout_dir(
+        self,
+        path_info,
+        hash_,
+        force,
+        progress_callback=None,
+        relink=False,
+        filter_info=None,
+    ):
+        added, modified = False, False
+        # Create dir separately so that dir is created
+        # even if there are no files in it
+        if not self.tree.exists(path_info):
+            added = True
+            self.tree.makedirs(path_info)
+
+        dir_info = self.get_dir_cache(hash_)
+
+        logger.debug("Linking directory '%s'.", path_info)
+
+        for entry in dir_info:
+            relative_path = entry[self.tree.PARAM_RELPATH]
+            entry_hash = entry[self.tree.PARAM_CHECKSUM]
+            entry_cache_info = self.tree.hash_to_path_info(entry_hash)
+            entry_info = path_info / relative_path
+
+            if filter_info and not entry_info.isin_or_eq(filter_info):
+                continue
+
+            entry_hash_info = {self.tree.PARAM_CHECKSUM: entry_hash}
+            if relink or self.changed(entry_info, entry_hash_info):
+                modified = True
+                self.safe_remove(entry_info, force=force)
+                self.link(entry_cache_info, entry_info)
+                self.tree.state.save(entry_info, entry_hash)
+            if progress_callback:
+                progress_callback(str(entry_info))
+
+        modified = (
+            self._remove_redundant_files(path_info, dir_info, force)
+            or modified
+        )
+
+        self.tree.state.save_link(path_info)
+        self.tree.state.save(path_info, hash_)
+
+        # relink is not modified, assume it as nochange
+        return added, not added and modified and not relink
+
+    def _remove_redundant_files(self, path_info, dir_info, force):
+        existing_files = set(self.tree.walk_files(path_info))
+
+        needed_files = {
+            path_info / entry[self.tree.PARAM_RELPATH] for entry in dir_info
+        }
+        redundant_files = existing_files - needed_files
+        for path in redundant_files:
+            self.safe_remove(path, force)
+
+        return bool(redundant_files)
+
+    def checkout(
+        self,
+        path_info,
+        hash_info,
+        force=False,
+        progress_callback=None,
+        relink=False,
+        filter_info=None,
+    ):
+        if path_info.scheme not in ["local", self.tree.scheme]:
+            raise NotImplementedError
+
+        hash_ = hash_info.get(self.tree.PARAM_CHECKSUM)
+        failed = None
+        skip = False
+        if not hash_:
+            logger.warning(
+                "No file hash info found for '%s'. " "It won't be created.",
+                path_info,
+            )
+            self.safe_remove(path_info, force=force)
+            failed = path_info
+
+        elif not relink and not self.changed(path_info, hash_info):
+            logger.debug("Data '%s' didn't change.", path_info)
+            skip = True
+
+        elif self.changed_cache(
+            hash_, path_info=path_info, filter_info=filter_info
+        ):
+            logger.warning(
+                "Cache '%s' not found. File '%s' won't be created.",
+                hash_,
+                path_info,
+            )
+            self.safe_remove(path_info, force=force)
+            failed = path_info
+
+        if failed or skip:
+            if progress_callback:
+                progress_callback(
+                    str(path_info),
+                    self.get_files_number(
+                        self.tree.path_info, hash_, filter_info
+                    ),
+                )
+            if failed:
+                raise CheckoutError([failed])
+            return
+
+        logger.debug("Checking out '%s' with cache '%s'.", path_info, hash_)
+
+        return self._checkout(
+            path_info, hash_, force, progress_callback, relink, filter_info,
+        )
+
+    def _checkout(
+        self,
+        path_info,
+        hash_,
+        force=False,
+        progress_callback=None,
+        relink=False,
+        filter_info=None,
+    ):
+        if not self.tree.is_dir_hash(hash_):
+            return self._checkout_file(
+                path_info, hash_, force, progress_callback, relink
+            )
+
+        return self._checkout_dir(
+            path_info, hash_, force, progress_callback, relink, filter_info
+        )
+
+    def get_files_number(self, path_info, hash_, filter_info):
+        from funcy.py3 import ilen
+
+        if not hash_:
+            return 0
+
+        if not self.tree.is_dir_hash(hash_):
+            return 1
+
+        if not filter_info:
+            return len(self.get_dir_cache(hash_))
+
+        return ilen(
+            filter_info.isin_or_eq(path_info / entry[self.tree.PARAM_CHECKSUM])
+            for entry in self.get_dir_cache(hash_)
+        )

dvc/cache/local.py

@@ -0,0 +1,431 @@
+import errno
+import logging
+import os
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from functools import partial, wraps
+
+from funcy import cached_property, concat
+
+from dvc.exceptions import DownloadError, UploadError
+from dvc.path_info import PathInfo
+from dvc.progress import Tqdm
+
+from ..remote.base import index_locked
+from ..tree.local import LocalTree
+from .base import (
+    STATUS_DELETED,
+    STATUS_MAP,
+    STATUS_MISSING,
+    STATUS_NEW,
+    CloudCache,
+)
+
+logger = logging.getLogger(__name__)
+
+
+def _log_exceptions(func, operation):
+    @wraps(func)
+    def wrapper(from_info, to_info, *args, **kwargs):
+        try:
+            func(from_info, to_info, *args, **kwargs)
+            return 0
+        except Exception as exc:  # pylint: disable=broad-except
+            # NOTE: this means we ran out of file descriptors and there is no
+            # reason to try to proceed, as we will hit this error anyways.
+            # pylint: disable=no-member
+            if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
+                raise
+
+            logger.exception(
+                "failed to %s '%s' to '%s'", operation, from_info, to_info
+            )
+            return 1
+
+    return wrapper
+
+
+class LocalCache(CloudCache):
+    DEFAULT_CACHE_TYPES = ["reflink", "copy"]
+    CACHE_MODE = LocalTree.CACHE_MODE
+
+    def __init__(self, tree):
+        super().__init__(tree)
+        self.cache_dir = tree.config.get("url")
+
+    @property
+    def cache_dir(self):
+        return self.tree.path_info.fspath if self.tree.path_info else None
+
+    @cache_dir.setter
+    def cache_dir(self, value):
+        self.tree.path_info = PathInfo(value) if value else None
+
+    @classmethod
+    def supported(cls, config):  # pylint: disable=unused-argument
+        return True
+
+    @cached_property
+    def cache_path(self):
+        return os.path.abspath(self.cache_dir)
+
+    def hash_to_path(self, hash_):
+        # NOTE: `self.cache_path` is already normalized so we can simply use
+        # `os.sep` instead of `os.path.join`. This results in this helper
+        # being ~5.5 times faster.
+        return f"{self.cache_path}{os.sep}{hash_[0:2]}{os.sep}{hash_[2:]}"
+
+    def hashes_exist(
+        self, hashes, jobs=None, name=None
+    ):  # pylint: disable=unused-argument
+        return [
+            hash_
+            for hash_ in Tqdm(
+                hashes,
+                unit="file",
+                desc="Querying "
+                + ("cache in " + name if name else "local cache"),
+            )
+            if not self.changed_cache_file(hash_)
+        ]
+
+    def already_cached(self, path_info):
+        assert path_info.scheme in ["", "local"]
+
+        current_md5 = self.tree.get_hash(path_info)
+
+        if not current_md5:
+            return False
+
+        return not self.changed_cache(current_md5)
+
+    def _verify_link(self, path_info, link_type):
+        if link_type == "hardlink" and self.tree.getsize(path_info) == 0:
+            return
+
+        super()._verify_link(path_info, link_type)
+
+    @index_locked
+    def status(
+        self,
+        named_cache,
+        remote,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+    ):
+        # Return flattened dict containing all status info
+        dir_status, file_status, _ = self._status(
+            named_cache,
+            remote,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=download,
+        )
+        return dict(dir_status, **file_status)
+
+    def _status(
+        self,
+        named_cache,
+        remote,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+    ):
+        """Return a tuple of (dir_status_info, file_status_info, dir_contents).
+
+        dir_status_info contains status for .dir files, file_status_info
+        contains status for all other files, and dir_contents is a dict of
+        {dir_hash: set(file_hash, ...)} which can be used to map
+        a .dir file to its file contents.
+        """
+        logger.debug(
+            f"Preparing to collect status from {remote.tree.path_info}"
+        )
+        md5s = set(named_cache.scheme_keys(self.tree.scheme))
+
+        logger.debug("Collecting information from local cache...")
+        local_exists = frozenset(
+            self.hashes_exist(md5s, jobs=jobs, name=self.cache_dir)
+        )
+
+        # This is a performance optimization. We can safely assume that,
+        # if the resources that we want to fetch are already cached,
+        # there's no need to check the remote storage for the existence of
+        # those files.
+        if download and local_exists == md5s:
+            remote_exists = local_exists
+        else:
+            logger.debug("Collecting information from remote cache...")
+            remote_exists = set()
+            dir_md5s = set(named_cache.dir_keys(self.tree.scheme))
+            if dir_md5s:
+                remote_exists.update(
+                    self._indexed_dir_hashes(named_cache, remote, dir_md5s)
+                )
+                md5s.difference_update(remote_exists)
+            if md5s:
+                remote_exists.update(
+                    remote.hashes_exist(
+                        md5s, jobs=jobs, name=str(remote.tree.path_info)
+                    )
+                )
+        return self._make_status(
+            named_cache, show_checksums, local_exists, remote_exists
+        )
+
+    def _make_status(
+        self, named_cache, show_checksums, local_exists, remote_exists
+    ):
+        def make_names(hash_, names):
+            return {"name": hash_ if show_checksums else " ".join(names)}
+
+        dir_status = {}
+        file_status = {}
+        dir_contents = {}
+        for hash_, item in named_cache[self.tree.scheme].items():
+            if item.children:
+                dir_status[hash_] = make_names(hash_, item.names)
+                dir_contents[hash_] = set()
+                for child_hash, child in item.children.items():
+                    file_status[child_hash] = make_names(
+                        child_hash, child.names
+                    )
+                    dir_contents[hash_].add(child_hash)
+            else:
+                file_status[hash_] = make_names(hash_, item.names)
+
+        self._fill_statuses(dir_status, local_exists, remote_exists)
+        self._fill_statuses(file_status, local_exists, remote_exists)
+
+        self._log_missing_caches(dict(dir_status, **file_status))
+
+        return dir_status, file_status, dir_contents
+
+    def _indexed_dir_hashes(self, named_cache, remote, dir_md5s):
+        # Validate our index by verifying all indexed .dir hashes
+        # still exist on the remote
+        indexed_dirs = set(remote.index.dir_hashes())
+        indexed_dir_exists = set()
+        if indexed_dirs:
+            indexed_dir_exists.update(
+                remote.tree.list_hashes_exists(indexed_dirs)
+            )
+            missing_dirs = indexed_dirs.difference(indexed_dir_exists)
+            if missing_dirs:
+                logger.debug(
+                    "Remote cache missing indexed .dir hashes '{}', "
+                    "clearing remote index".format(", ".join(missing_dirs))
+                )
+                remote.index.clear()
+
+        # Check if non-indexed (new) dir hashes exist on remote
+        dir_exists = dir_md5s.intersection(indexed_dir_exists)
+        dir_exists.update(
+            remote.tree.list_hashes_exists(dir_md5s - dir_exists)
+        )
+
+        # If .dir hash exists on the remote, assume directory contents
+        # still exists on the remote
+        for dir_hash in dir_exists:
+            file_hashes = list(
+                named_cache.child_keys(self.tree.scheme, dir_hash)
+            )
+            if dir_hash not in remote.index:
+                logger.debug(
+                    "Indexing new .dir '{}' with '{}' nested files".format(
+                        dir_hash, len(file_hashes)
+                    )
+                )
+                remote.index.update([dir_hash], file_hashes)
+            yield dir_hash
+            yield from file_hashes
+
+    @staticmethod
+    def _fill_statuses(hash_info_dir, local_exists, remote_exists):
+        # Using sets because they are way faster for lookups
+        local = set(local_exists)
+        remote = set(remote_exists)
+
+        for md5, info in hash_info_dir.items():
+            status = STATUS_MAP[(md5 in local, md5 in remote)]
+            info["status"] = status
+
+    def _get_plans(self, download, remote, status_info, status):
+        cache = []
+        path_infos = []
+        names = []
+        hashes = []
+        for md5, info in Tqdm(
+            status_info.items(), desc="Analysing status", unit="file"
+        ):
+            if info["status"] == status:
+                cache.append(self.tree.hash_to_path_info(md5))
+                path_infos.append(remote.tree.hash_to_path_info(md5))
+                names.append(info["name"])
+                hashes.append(md5)
+
+        if download:
+            to_infos = cache
+            from_infos = path_infos
+        else:
+            to_infos = path_infos
+            from_infos = cache
+
+        return from_infos, to_infos, names, hashes
+
+    def _process(
+        self,
+        named_cache,
+        remote,
+        jobs=None,
+        show_checksums=False,
+        download=False,
+    ):
+        logger.debug(
+            "Preparing to {} '{}'".format(
+                "download data from" if download else "upload data to",
+                remote.tree.path_info,
+            )
+        )
+
+        if download:
+            func = partial(
+                _log_exceptions(remote.tree.download, "download"),
+                dir_mode=self.tree.dir_mode,
+                file_mode=self.tree.file_mode,
+            )
+            status = STATUS_DELETED
+            desc = "Downloading"
+        else:
+            func = _log_exceptions(remote.tree.upload, "upload")
+            status = STATUS_NEW
+            desc = "Uploading"
+
+        if jobs is None:
+            jobs = remote.tree.JOBS
+
+        dir_status, file_status, dir_contents = self._status(
+            named_cache,
+            remote,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=download,
+        )
+
+        dir_plans = self._get_plans(download, remote, dir_status, status)
+        file_plans = self._get_plans(download, remote, file_status, status)
+
+        total = len(dir_plans[0]) + len(file_plans[0])
+        if total == 0:
+            return 0
+
+        with Tqdm(total=total, unit="file", desc=desc) as pbar:
+            func = pbar.wrap_fn(func)
+            with ThreadPoolExecutor(max_workers=jobs) as executor:
+                if download:
+                    from_infos, to_infos, names, _ = (
+                        d + f for d, f in zip(dir_plans, file_plans)
+                    )
+                    fails = sum(
+                        executor.map(func, from_infos, to_infos, names)
+                    )
+                else:
+                    # for uploads, push files first, and any .dir files last
+
+                    file_futures = {}
+                    for from_info, to_info, name, hash_ in zip(*file_plans):
+                        file_futures[hash_] = executor.submit(
+                            func, from_info, to_info, name
+                        )
+                    dir_futures = {}
+                    for from_info, to_info, name, dir_hash in zip(*dir_plans):
+                        wait_futures = {
+                            future
+                            for file_hash, future in file_futures.items()
+                            if file_hash in dir_contents[dir_hash]
+                        }
+                        dir_futures[dir_hash] = executor.submit(
+                            self._dir_upload,
+                            func,
+                            wait_futures,
+                            from_info,
+                            to_info,
+                            name,
+                        )
+                    fails = sum(
+                        future.result()
+                        for future in concat(
+                            file_futures.values(), dir_futures.values()
+                        )
+                    )
+
+        if fails:
+            if download:
+                remote.index.clear()
+                raise DownloadError(fails)
+            raise UploadError(fails)
+
+        if not download:
+            # index successfully pushed dirs
+            for dir_hash, future in dir_futures.items():
+                if future.result() == 0:
+                    file_hashes = dir_contents[dir_hash]
+                    logger.debug(
+                        "Indexing pushed dir '{}' with "
+                        "'{}' nested files".format(dir_hash, len(file_hashes))
+                    )
+                    remote.index.update([dir_hash], file_hashes)
+
+        return len(dir_plans[0]) + len(file_plans[0])
+
+    @staticmethod
+    def _dir_upload(func, futures, from_info, to_info, name):
+        for future in as_completed(futures):
+            if future.result():
+                # do not upload this .dir file if any file in this
+                # directory failed to upload
+                logger.debug(
+                    "failed to upload full contents of '{}', "
+                    "aborting .dir file upload".format(name)
+                )
+                logger.error(f"failed to upload '{from_info}' to '{to_info}'")
+                return 1
+        return func(from_info, to_info, name)
+
+    @index_locked
+    def push(self, named_cache, remote, jobs=None, show_checksums=False):
+        return self._process(
+            named_cache,
+            remote,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=False,
+        )
+
+    @index_locked
+    def pull(self, named_cache, remote, jobs=None, show_checksums=False):
+        return self._process(
+            named_cache,
+            remote,
+            jobs=jobs,
+            show_checksums=show_checksums,
+            download=True,
+        )
+
+    @staticmethod
+    def _log_missing_caches(hash_info_dict):
+        missing_caches = [
+            (md5, info)
+            for md5, info in hash_info_dict.items()
+            if info["status"] == STATUS_MISSING
+        ]
+        if missing_caches:
+            missing_desc = "\n".join(
+                "name: {}, md5: {}".format(info["name"], md5)
+                for md5, info in missing_caches
+            )
+            msg = (
+                "Some of the cache files do not exist neither locally "
+                "nor on remote. Missing cache files:\n{}".format(missing_desc)
+            )
+            logger.warning(msg)

dvc/cli.py

@@ -15,6 +15,7 @@ from .command import (
     data_sync,
     destroy,
     diff,
+    experiments,
     freeze,
     gc,
     get,
@@ -77,37 +78,40 @@ COMMANDS = [
     update,
     git_hook,
     plots,
+    experiments,
 ]
 
 
+def _find_parser(parser, cmd_cls):
+    defaults = parser._defaults  # pylint: disable=protected-access
+    if not cmd_cls or cmd_cls == defaults.get("func"):
+        parser.print_help()
+        raise DvcParserError()
+
+    actions = parser._actions  # pylint: disable=protected-access
+    for action in actions:
+        if not isinstance(action.choices, dict):
+            # NOTE: we are only interested in subparsers
+            continue
+        for subparser in action.choices.values():
+            _find_parser(subparser, cmd_cls)
+
+
 class DvcParser(argparse.ArgumentParser):
     """Custom parser class for dvc CLI."""
 
-    def error(self, message, command=None):
-        """Custom error method.
-        Args:
-            message (str): error message.
-            command (str): subcommand name for help message
-        Raises:
-            dvc.exceptions.DvcParser: dvc parser exception.
-
-        """
+    def error(self, message, cmd_cls=None):  # pylint: disable=arguments-differ
         logger.error(message)
-        if command is not None:
-            for action in self._actions:
-                if action.dest == "cmd" and command in action.choices:
-                    subparser = action.choices[command]
-                    subparser.print_help()
-                    raise DvcParserError()
-        self.print_help()
-        raise DvcParserError()
+        _find_parser(self, cmd_cls)
 
-    # override this to send subcommand name to error method
     def parse_args(self, args=None, namespace=None):
+        # NOTE: overriding to provide a more granular help message.
+        # E.g. `dvc plots diff --bad-flag` would result in a `dvc plots diff`
+        # help message instead of generic `dvc` usage.
         args, argv = self.parse_known_args(args, namespace)
         if argv:
             msg = "unrecognized arguments: %s"
-            self.error(msg % " ".join(argv), args.cmd)
+            self.error(msg % " ".join(argv), getattr(args, "func", None))
         return args
 
 

dvc/command/add.py

@@ -63,10 +63,6 @@ def add_parser(subparsers, parent_parser):
         metavar="<filename>",
     )
     parser.add_argument(
-        "targets",
-        nargs="+",
-        help="Input files/directories to add.",
-        metavar="targets",
-        choices=completion.Required.FILE,
-    )
+        "targets", nargs="+", help="Input files/directories to add.",
+    ).complete = completion.FILE
     parser.set_defaults(func=CmdAdd)

dvc/command/base.py

@@ -1,4 +1,5 @@
 import logging
+from abc import ABC, abstractmethod
 
 logger = logging.getLogger(__name__)
 
@@ -26,7 +27,7 @@ def append_doc_link(help_message, path):
     )
 
 
-class CmdBase:
+class CmdBase(ABC):
     def __init__(self, args):
         from dvc.repo import Repo
         from dvc.updater import Updater
@@ -46,11 +47,11 @@ class CmdBase:
         logger.debug(f"assuming default target '{PIPELINE_FILE}'.")
         return [PIPELINE_FILE]
 
-    # Abstract methods that have to be implemented by any inheritance class
+    @abstractmethod
     def run(self):
-        raise NotImplementedError
+        pass
 
 
 class CmdBaseNoRepo(CmdBase):
-    def __init__(self, args):
+    def __init__(self, args):  # pylint: disable=super-init-not-called
         self.args = args

dvc/command/cache.py

@@ -56,7 +56,5 @@ def add_parser(subparsers, parent_parser):
         help="Path to cache directory. Relative paths are resolved relative "
         "to the current directory and saved to config relative to the "
         "config file location.",
-        metavar="value",
-        choices=completion.Required.DIR,
-    )
+    ).complete = completion.DIR
     cache_dir_parser.set_defaults(func=CmdCacheDir)

dvc/command/checkout.py

@@ -112,9 +112,9 @@ def add_parser(subparsers, parent_parser):
     checkout_parser.add_argument(
         "targets",
         nargs="*",
-        help="DVC-files to checkout. Optional. "
-        "(Finds all DVC-files in the workspace by default.)",
-        metavar="targets",
-        choices=completion.Optional.DVC_FILE,
-    )
+        help=(
+            "Limit command scope to these tracked files/directories, "
+            ".dvc files, or stage names."
+        ),
+    ).complete = completion.DVC_FILE
     checkout_parser.set_defaults(func=CmdCheckout)

dvc/command/commit.py

@@ -67,7 +67,5 @@ def add_parser(subparsers, parent_parser):
         nargs="*",
         help="DVC-files to commit. Optional. "
         "(Finds all DVC-files in the workspace by default.)",
-        metavar="targets",
-        choices=completion.Optional.DVC_FILE,
-    )
+    ).complete = completion.DVC_FILE
     commit_parser.set_defaults(func=CmdCommit)

dvc/command/completion.py

@@ -6,9 +6,11 @@ import shtab
 from dvc.command.base import CmdBaseNoRepo, append_doc_link
 
 logger = logging.getLogger(__name__)
-CHOICE_FUNCTIONS = {
-    "bash": {"DVCFile": "_dvc_compgen_DVCFiles"},
-    "zsh": {"DVCFile": "_files -g '(*?.dvc|Dvcfile|dvc.yaml)'"},
+FILE = shtab.FILE
+DIR = shtab.DIRECTORY
+DVC_FILE = {
+    "bash": "_dvc_compgen_DVCFiles",
+    "zsh": "_files -g '(*?.dvc|Dvcfile|dvc.yaml)'",
 }
 PREAMBLE = {
     "bash": """
@@ -24,26 +26,13 @@ _dvc_compgen_DVCFiles() {
 }
 
 
-class Optional(shtab.Optional):
-    DVC_FILE = [shtab.Choice("DVCFile", required=False)]
-
-
-class Required(shtab.Required):
-    DVC_FILE = [shtab.Choice("DVCFile", required=True)]
-
-
 class CmdCompletion(CmdBaseNoRepo):
     def run(self):
         from dvc.cli import get_main_parser
 
         parser = get_main_parser()
         shell = self.args.shell
-        script = shtab.complete(
-            parser,
-            shell=shell,
-            preamble=PREAMBLE[shell],
-            choice_functions=CHOICE_FUNCTIONS[shell],
-        )
+        script = shtab.complete(parser, shell=shell, preamble=PREAMBLE)
         print(script)
         return 0
 

dvc/command/daemon.py

@@ -66,9 +66,6 @@ def add_parser(subparsers, parent_parser):
         help=DAEMON_ANALYTICS_HELP,
     )
     daemon_analytics_parser.add_argument(
-        "target",
-        help="Analytics file.",
-        metavar="target",
-        choices=completion.Required.FILE,
-    )
+        "target", help="Analytics file.",
+    ).complete = completion.FILE
     daemon_analytics_parser.set_defaults(func=CmdDaemonAnalytics)

dvc/command/data_sync.py

@@ -103,11 +103,11 @@ def shared_parent_parser():
     parent_parser.add_argument(
         "targets",
         nargs="*",
-        help="Limit command scope to these DVC-files. "
-        "Using -R, directories to search DVC-files in can also be given.",
-        metavar="targets",
-        choices=completion.Optional.DVC_FILE,
-    )
+        help=(
+            "Limit command scope to these tracked files/directories, "
+            ".dvc files, or stage names."
+        ),
+    ).complete = completion.DVC_FILE
 
     return parent_parser
 

dvc/command/destroy.py

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

dvc/command/diff.py

@@ -56,7 +56,7 @@ class CmdDiff(CmdBase):
         """
 
         def _digest(checksum):
-            if type(checksum) is str:
+            if isinstance(checksum, str):
                 return checksum[0:8]
             return "{}..{}".format(checksum["old"][0:8], checksum["new"][0:8])
 

dvc/command/experiments.py

@@ -0,0 +1,354 @@
+import argparse
+import io
+import logging
+from collections import OrderedDict
+
+from dvc.command.base import CmdBase, append_doc_link, fix_subparsers
+from dvc.command.metrics import DEFAULT_PRECISION
+from dvc.exceptions import DvcException
+
+logger = logging.getLogger(__name__)
+
+
+def _update_names(names, items):
+    from flatten_json import flatten
+
+    for name, item in items:
+        if isinstance(item, dict):
+            item = flatten(item, ".")
+            names.update(item.keys())
+        else:
+            names.add(name)
+
+
+def _collect_names(all_experiments):
+    metric_names = set()
+    param_names = set()
+
+    for _, experiments in all_experiments.items():
+        for exp in experiments.values():
+            _update_names(metric_names, exp.get("metrics", {}).items())
+            _update_names(param_names, exp.get("params", {}).items())
+
+    return sorted(metric_names), sorted(param_names)
+
+
+def _collect_rows(
+    base_rev, experiments, metric_names, param_names, precision=None
+):
+    from flatten_json import flatten
+
+    if precision is None:
+        precision = DEFAULT_PRECISION
+
+    def _round(val):
+        if isinstance(val, float):
+            return round(val, precision)
+
+        return val
+
+    def _extend(row, names, items):
+        for fname, item in items:
+            if isinstance(item, dict):
+                item = flatten(item, ".")
+            else:
+                item = {fname: item}
+            for name in names:
+                if name in item:
+                    row.append(str(_round(item[name])))
+                else:
+                    row.append("-")
+
+    for i, (rev, exp) in enumerate(experiments.items()):
+        row = []
+        style = None
+        if rev == "baseline":
+            row.append(f"{base_rev}")
+            style = "bold"
+        elif i < len(experiments) - 1:
+            row.append(f"├── {rev[:7]}")
+        else:
+            row.append(f"└── {rev[:7]}")
+
+        _extend(row, metric_names, exp.get("metrics", {}).items())
+        _extend(row, param_names, exp.get("params", {}).items())
+
+        yield row, style
+
+
+def _show_experiments(all_experiments, console, precision=None):
+    from rich.table import Table
+    from dvc.scm.git import Git
+
+    metric_names, param_names = _collect_names(all_experiments)
+
+    table = Table()
+    table.add_column("Experiment")
+    for name in metric_names:
+        table.add_column(name, justify="right")
+    for name in param_names:
+        table.add_column(name, justify="left")
+
+    for base_rev, experiments in all_experiments.items():
+        if Git.is_sha(base_rev):
+            base_rev = base_rev[:7]
+
+        for row, _, in _collect_rows(
+            base_rev,
+            experiments,
+            metric_names,
+            param_names,
+            precision=precision,
+        ):
+            table.add_row(*row)
+
+    console.print(table)
+
+
+class CmdExperimentsShow(CmdBase):
+    def run(self):
+        from rich.console import Console
+        from dvc.utils.pager import pager
+
+        if not self.repo.experiments:
+            return 0
+
+        try:
+            all_experiments = self.repo.experiments.show(
+                all_branches=self.args.all_branches,
+                all_tags=self.args.all_tags,
+                all_commits=self.args.all_commits,
+            )
+
+            # Note: rich does not currently include a native way to force
+            # infinite width for use with a pager
+            console = Console(
+                file=io.StringIO(), force_terminal=True, width=9999
+            )
+
+            _show_experiments(all_experiments, console)
+
+            pager(console.file.getvalue())
+        except DvcException:
+            logger.exception("failed to show experiments")
+            return 1
+
+        return 0
+
+
+class CmdExperimentsCheckout(CmdBase):
+    def run(self):
+        if not self.repo.experiments:
+            return 0
+
+        self.repo.experiments.checkout(
+            self.args.experiment, force=self.args.force
+        )
+
+        return 0
+
+
+def _show_diff(
+    diff, title="", markdown=False, no_path=False, old=False, precision=None
+):
+    from dvc.utils.diff import table
+
+    if precision is None:
+        precision = DEFAULT_PRECISION
+
+    def _round(val):
+        if isinstance(val, float):
+            return round(val, precision)
+
+        return val
+
+    rows = []
+    for fname, diff_ in diff.items():
+        sorted_diff = OrderedDict(sorted(diff_.items()))
+        for item, change in sorted_diff.items():
+            row = [] if no_path else [fname]
+            row.append(item)
+            if old:
+                row.append(_round(change.get("old")))
+            row.append(_round(change["new"]))
+            row.append(_round(change.get("diff", "diff not supported")))
+            rows.append(row)
+
+    header = [] if no_path else ["Path"]
+    header.append(title)
+    if old:
+        header.extend(["Old", "New"])
+    else:
+        header.append("Value")
+    header.append("Change")
+
+    return table(header, rows, markdown)
+
+
+class CmdExperimentsDiff(CmdBase):
+    def run(self):
+        if not self.repo.experiments:
+            return 0
+
+        try:
+            diff = self.repo.experiments.diff(
+                a_rev=self.args.a_rev,
+                b_rev=self.args.b_rev,
+                all=self.args.all,
+            )
+
+            if self.args.show_json:
+                import json
+
+                logger.info(json.dumps(diff))
+            else:
+                diffs = [("metrics", "Metric"), ("params", "Param")]
+                for key, title in diffs:
+                    table = _show_diff(
+                        diff[key],
+                        title=title,
+                        markdown=self.args.show_md,
+                        no_path=self.args.no_path,
+                        old=self.args.old,
+                        precision=self.args.precision,
+                    )
+                    if table:
+                        logger.info(table)
+                        logger.info("")
+
+        except DvcException:
+            logger.exception("failed to show experiments diff")
+            return 1
+
+        return 0
+
+
+def add_parser(subparsers, parent_parser):
+    EXPERIMENTS_HELP = "Commands to display and compare experiments."
+
+    experiments_parser = subparsers.add_parser(
+        "experiments",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_HELP, "experiments"),
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+
+    experiments_subparsers = experiments_parser.add_subparsers(
+        dest="cmd",
+        help="Use `dvc experiments CMD --help` to display "
+        "command-specific help.",
+    )
+
+    fix_subparsers(experiments_subparsers)
+
+    EXPERIMENTS_SHOW_HELP = "Print experiments."
+    experiments_show_parser = experiments_subparsers.add_parser(
+        "show",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_SHOW_HELP, "experiments/show"),
+        help=EXPERIMENTS_SHOW_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_show_parser.add_argument(
+        "-a",
+        "--all-branches",
+        action="store_true",
+        default=False,
+        help="Show metrics for all branches.",
+    )
+    experiments_show_parser.add_argument(
+        "-T",
+        "--all-tags",
+        action="store_true",
+        default=False,
+        help="Show metrics for all tags.",
+    )
+    experiments_show_parser.add_argument(
+        "--all-commits",
+        action="store_true",
+        default=False,
+        help="Show metrics for all commits.",
+    )
+    experiments_show_parser.set_defaults(func=CmdExperimentsShow)
+
+    EXPERIMENTS_CHECKOUT_HELP = "Checkout experiments."
+    experiments_checkout_parser = experiments_subparsers.add_parser(
+        "checkout",
+        parents=[parent_parser],
+        description=append_doc_link(
+            EXPERIMENTS_CHECKOUT_HELP, "experiments/checkout"
+        ),
+        help=EXPERIMENTS_CHECKOUT_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_checkout_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Overwrite your current workspace with changes from the "
+        "experiment.",
+    )
+    experiments_checkout_parser.add_argument(
+        "experiment", help="Checkout this experiment.",
+    )
+    experiments_checkout_parser.set_defaults(func=CmdExperimentsCheckout)
+
+    EXPERIMENTS_DIFF_HELP = (
+        "Show changes between experiments in the DVC repository."
+    )
+    experiments_diff_parser = experiments_subparsers.add_parser(
+        "diff",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_DIFF_HELP, "experiments/diff"),
+        help=EXPERIMENTS_DIFF_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_diff_parser.add_argument(
+        "a_rev", nargs="?", help="Old experiment to compare (defaults to HEAD)"
+    )
+    experiments_diff_parser.add_argument(
+        "b_rev",
+        nargs="?",
+        help="New experiment to compare (defaults to the current workspace)",
+    )
+    experiments_diff_parser.add_argument(
+        "--all",
+        action="store_true",
+        default=False,
+        help="Show unchanged metrics/params as well.",
+    )
+    experiments_diff_parser.add_argument(
+        "--show-json",
+        action="store_true",
+        default=False,
+        help="Show output in JSON format.",
+    )
+    experiments_diff_parser.add_argument(
+        "--show-md",
+        action="store_true",
+        default=False,
+        help="Show tabulated output in the Markdown format (GFM).",
+    )
+    experiments_diff_parser.add_argument(
+        "--old",
+        action="store_true",
+        default=False,
+        help="Show old metric/param value.",
+    )
+    experiments_diff_parser.add_argument(
+        "--no-path",
+        action="store_true",
+        default=False,
+        help="Don't show metric/param path.",
+    )
+    experiments_diff_parser.add_argument(
+        "--precision",
+        type=int,
+        help=(
+            "Round metrics/params to `n` digits precision after the decimal "
+            f"point. Rounds to {DEFAULT_PRECISION} digits by default."
+        ),
+        metavar="<n>",
+    )
+    experiments_diff_parser.set_defaults(func=CmdExperimentsDiff)

dvc/command/freeze.py

@@ -31,7 +31,7 @@ class CmdUnfreeze(CmdFreezeBase):
 
 
 def add_parser(subparsers, parent_parser):
-    FREEZE_HELP = "Freeze DVC-files."
+    FREEZE_HELP = "Freeze stages or .dvc files."
     freeze_parser = subparsers.add_parser(
         "freeze",
         parents=[parent_parser],
@@ -40,15 +40,11 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     freeze_parser.add_argument(
-        "targets",
-        nargs="+",
-        help="DVC-files to freeze.",
-        metavar="targets",
-        choices=completion.Required.DVC_FILE,
-    )
+        "targets", nargs="+", help="Stages or .dvc files to freeze",
+    ).complete = completion.DVC_FILE
     freeze_parser.set_defaults(func=CmdFreeze)
 
-    UNFREEZE_HELP = "Unfreeze DVC-files."
+    UNFREEZE_HELP = "Unfreeze stages or .dvc files."
     unfreeze_parser = subparsers.add_parser(
         "unfreeze",
         parents=[parent_parser],
@@ -57,10 +53,6 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     unfreeze_parser.add_argument(
-        "targets",
-        nargs="+",
-        help="DVC-files to unfreeze.",
-        metavar="targets",
-        choices=completion.Required.DVC_FILE,
-    )
+        "targets", nargs="+", help="Stages or .dvc files to unfreeze",
+    ).complete = completion.DVC_FILE
     unfreeze_parser.set_defaults(func=CmdUnfreeze)

dvc/command/get.py

@@ -63,19 +63,15 @@ def add_parser(subparsers, parent_parser):
         "url", help="Location of DVC or Git repository to download from"
     )
     get_parser.add_argument(
-        "path",
-        help="Path to a file or directory within the repository",
-        metavar="path",
-        choices=completion.Required.FILE,
-    )
+        "path", help="Path to a file or directory within the repository",
+    ).complete = completion.FILE
     get_parser.add_argument(
         "-o",
         "--out",
         nargs="?",
         help="Destination path to download files to",
         metavar="<path>",
-        choices=completion.Optional.DIR,
-    )
+    ).complete = completion.DIR
     get_parser.add_argument(
         "--rev",
         nargs="?",

dvc/command/get_url.py

@@ -34,10 +34,6 @@ def add_parser(subparsers, parent_parser):
         "url", help="See `dvc import-url -h` for full list of supported URLs."
     )
     get_parser.add_argument(
-        "out",
-        nargs="?",
-        help="Destination path to put data to.",
-        metavar="out",
-        choices=completion.Optional.DIR,
-    )
+        "out", nargs="?", help="Destination path to put data to.",
+    ).complete = completion.DIR
     get_parser.set_defaults(func=CmdGetUrl)

dvc/command/git_hook.py

@@ -19,6 +19,9 @@ class CmdHookBase(CmdBaseNoRepo):
 
         return self._run()
 
+    def _run(self):
+        raise NotImplementedError
+
 
 class CmdPreCommit(CmdHookBase):
     def _run(self):

dvc/command/imp.py

@@ -44,19 +44,15 @@ def add_parser(subparsers, parent_parser):
         "url", help="Location of DVC or Git repository to download from"
     )
     import_parser.add_argument(
-        "path",
-        help="Path to a file or directory within the repository",
-        metavar="path",
-        choices=completion.Required.FILE,
-    )
+        "path", help="Path to a file or directory within the repository",
+    ).complete = completion.FILE
     import_parser.add_argument(
         "-o",
         "--out",
         nargs="?",
         help="Destination path to download files to",
         metavar="<path>",
-        choices=completion.Optional.DIR,
-    )
+    ).complete = completion.DIR
     import_parser.add_argument(
         "--rev",
         nargs="?",

dvc/command/imp_url.py

@@ -58,18 +58,13 @@ def add_parser(subparsers, parent_parser):
         "remote://myremote/path/to/file (see `dvc remote`)",
     )
     import_parser.add_argument(
-        "out",
-        nargs="?",
-        help="Destination path to put files to.",
-        metavar="out",
-        choices=completion.Optional.DIR,
-    )
+        "out", nargs="?", help="Destination path to put files to.",
+    ).complete = completion.DIR
     import_parser.add_argument(
         "--file",
         help="Specify name of the DVC-file this command will generate.",
         metavar="<filename>",
-        choices=completion.Optional.DIR,
-    )
+    ).complete = completion.DIR
     import_parser.add_argument(
         "--no-exec",
         action="store_true",

dvc/command/install.py

@@ -10,7 +10,7 @@ class CmdInstall(CmdBase):
     def run(self):
         try:
             self.repo.install(self.args.use_pre_commit_tool)
-        except Exception:
+        except Exception:  # noqa, pylint:disable=broad-except
             logger.exception("failed to install DVC Git hooks")
             return 1
         return 0

dvc/command/ls/__init__.py

@@ -75,7 +75,5 @@ def add_parser(subparsers, parent_parser):
         "path",
         nargs="?",
         help="Path to directory within the repository to list outputs for",
-        metavar="path",
-        choices=completion.Optional.DIR,
-    )
+    ).complete = completion.DIR
     list_parser.set_defaults(func=CmdList)

dvc/command/metrics.py

@@ -170,9 +170,7 @@ def add_parser(subparsers, parent_parser):
             "Limit command scope to these metric files. Using -R, "
             "directories to search metric files in can also be given."
         ),
-        metavar="targets",
-        choices=completion.Optional.FILE,
-    )
+    ).complete = completion.FILE
     metrics_show_parser.add_argument(
         "-a",
         "--all-branches",
@@ -238,8 +236,7 @@ def add_parser(subparsers, parent_parser):
             "directories to search metric files in can also be given."
         ),
         metavar="<paths>",
-        choices=completion.Optional.FILE,
-    )
+    ).complete = completion.FILE
     metrics_diff_parser.add_argument(
         "-R",
         "--recursive",

dvc/command/move.py

@@ -37,15 +37,9 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     move_parser.add_argument(
-        "src",
-        help="Source path to a data file or directory.",
-        metavar="src",
-        choices=completion.Required.FILE,
-    )
+        "src", help="Source path to a data file or directory.",
+    ).complete = completion.FILE
     move_parser.add_argument(
-        "dst",
-        help="Destination path.",
-        metavar="dst",
-        choices=completion.Required.FILE,
-    )
+        "dst", help="Destination path.",
+    ).complete = completion.FILE
     move_parser.set_defaults(func=CmdMove)

dvc/command/plots.py

@@ -129,9 +129,7 @@ def add_parser(subparsers, parent_parser):
         "targets",
         nargs="*",
         help="Plots files to visualize. Shows all plots by default.",
-        metavar="targets",
-        choices=completion.Optional.FILE,
-    )
+    ).complete = completion.FILE
     _add_props_arguments(plots_show_parser)
     _add_output_arguments(plots_show_parser)
     plots_show_parser.set_defaults(func=CmdPlotsShow)
@@ -152,8 +150,7 @@ def add_parser(subparsers, parent_parser):
         nargs="*",
         help="Plots file to visualize. Shows all plots by default.",
         metavar="<path>",
-        choices=completion.Optional.FILE,
-    )
+    ).complete = completion.FILE
     plots_diff_parser.add_argument(
         "revisions", nargs="*", default=None, help="Git commits to plot from",
     )
@@ -170,11 +167,8 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     plots_modify_parser.add_argument(
-        "target",
-        help="Metric file to set properties to",
-        metavar="target",
-        choices=completion.Required.FILE,
-    )
+        "target", help="Metric file to set properties to",
+    ).complete = completion.FILE
     _add_props_arguments(plots_modify_parser)
     plots_modify_parser.add_argument(
         "--unset",
@@ -198,8 +192,7 @@ def _add_props_arguments(parser):
             )
         ),
         metavar="<path>",
-        choices=completion.Optional.FILE,
-    )
+    ).complete = completion.FILE
     parser.add_argument(
         "-x", default=None, help="Field name for X axis.", metavar="<field>"
     )
@@ -231,8 +224,7 @@ def _add_output_arguments(parser):
         default=None,
         help="Destination path to save plots to",
         metavar="<path>",
-        choices=completion.Optional.DIR,
-    )
+    ).complete = completion.DIR
     parser.add_argument(
         "--show-vega",
         action="store_true",

dvc/command/remove.py

@@ -37,10 +37,6 @@ def add_parser(subparsers, parent_parser):
         help="Remove outputs as well.",
     )
     remove_parser.add_argument(
-        "targets",
-        nargs="+",
-        help="DVC-files to remove.",
-        metavar="targets",
-        choices=completion.Required.DVC_FILE,
-    )
+        "targets", nargs="+", help="DVC-files to remove.",
+    ).complete = completion.DVC_FILE
     remove_parser.set_defaults(func=CmdRemove)

dvc/command/repro.py

@@ -40,6 +40,7 @@ class CmdRepro(CmdBase):
                     downstream=self.args.downstream,
                     recursive=self.args.recursive,
                     force_downstream=self.args.force_downstream,
+                    experiment=self.args.experiment,
                 )
 
                 if len(stages) == 0:
@@ -72,9 +73,7 @@ def add_parser(subparsers, parent_parser):
         "targets",
         nargs="*",
         help=f"Stages to reproduce. '{PIPELINE_FILE}' by default.",
-        metavar="targets",
-        choices=completion.Optional.DVC_FILE,
-    )
+    ).complete = completion.DVC_FILE
     repro_parser.add_argument(
         "-f",
         "--force",
@@ -168,4 +167,11 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Start from the specified stages when reproducing pipelines.",
     )
+    repro_parser.add_argument(
+        "-e",
+        "--experiment",
+        action="store_true",
+        default=False,
+        help=argparse.SUPPRESS,
+    )
     repro_parser.set_defaults(func=CmdRepro)

dvc/command/status.py

@@ -63,7 +63,7 @@ class CmdDataStatus(CmdDataBase):
             else:
                 logger.info(self.UP_TO_DATE_MSG)
 
-        except Exception:
+        except Exception:  # noqa, pylint: disable=broad-except
             logger.exception("failed to obtain data status")
             return 1
         return 0

dvc/command/unprotect.py

@@ -33,10 +33,6 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     unprotect_parser.add_argument(
-        "targets",
-        nargs="+",
-        help="Data files/directories to unprotect.",
-        metavar="targets",
-        choices=completion.Required.FILE,
-    )
+        "targets", nargs="+", help="Data files/directories to unprotect.",
+    ).complete = completion.FILE
     unprotect_parser.set_defaults(func=CmdUnprotect)

dvc/command/update.py

@@ -33,12 +33,8 @@ def add_parser(subparsers, parent_parser):
         formatter_class=argparse.RawDescriptionHelpFormatter,
     )
     update_parser.add_argument(
-        "targets",
-        nargs="+",
-        help="DVC-files to update.",
-        metavar="targets",
-        choices=completion.Required.DVC_FILE,
-    )
+        "targets", nargs="+", help="DVC-files to update.",
+    ).complete = completion.DVC_FILE
     update_parser.add_argument(
         "--rev",
         nargs="?",

dvc/command/version.py

@@ -122,7 +122,7 @@ class CmdVersion(CmdBaseNoRepo):
 
     @staticmethod
     def get_supported_remotes():
-        from dvc.remote import TREES
+        from dvc.tree import TREES
 
         supported_remotes = []
         for tree_cls in TREES:

dvc/config.py

@@ -127,6 +127,7 @@ SCHEMA = {
         Optional("analytics", default=True): Bool,
         Optional("hardlink_lock", default=False): Bool,
         Optional("no_scm", default=False): Bool,
+        Optional("experiments", default=False): Bool,
     },
     "cache": {
         "local": str,
@@ -148,6 +149,8 @@ SCHEMA = {
                     "profile": str,
                     "credentialpath": str,
                     "endpointurl": str,
+                    "access_key_id": str,
+                    "secret_access_key": str,
                     Optional("listobjects", default=False): Bool,
                     Optional("use_ssl", default=True): Bool,
                     "sse": str,
@@ -231,7 +234,11 @@ class Config(dict):
     CONFIG = "config"
     CONFIG_LOCAL = "config.local"
 
-    def __init__(self, dvc_dir=None, validate=True):
+    def __init__(
+        self, dvc_dir=None, validate=True, tree=None,
+    ):  # pylint: disable=super-init-not-called
+        from dvc.tree.local import LocalTree
+
         self.dvc_dir = dvc_dir
 
         if not dvc_dir:
@@ -244,6 +251,9 @@ class Config(dict):
         else:
             self.dvc_dir = os.path.abspath(os.path.realpath(dvc_dir))
 
+        self.wtree = LocalTree(None, {"url": self.dvc_dir})
+        self.tree = tree or self.wtree
+
         self.load(validate=validate)
 
     @classmethod
@@ -302,8 +312,30 @@ class Config(dict):
         if not self["cache"].get("dir") and self.dvc_dir:
             self["cache"]["dir"] = os.path.join(self.dvc_dir, "cache")
 
+    def _load_config(self, level):
+        filename = self.files[level]
+
+        if self.tree.exists(filename, use_dvcignore=False):
+            with self.tree.open(filename) as fobj:
+                conf_obj = configobj.ConfigObj(fobj)
+        else:
+            conf_obj = configobj.ConfigObj()
+        return _parse_remotes(_lower_keys(conf_obj.dict()))
+
+    def _save_config(self, level, conf_dict):
+        filename = self.files[level]
+
+        logger.debug(f"Writing '{filename}'.")
+
+        self.tree.makedirs(os.path.dirname(filename))
+
+        config = configobj.ConfigObj(_pack_remotes(conf_dict))
+        with self.tree.open(filename, "wb") as fobj:
+            config.write(fobj)
+        config.filename = filename
+
     def load_one(self, level):
-        conf = _load_config(self.files[level])
+        conf = self._load_config(level)
         conf = self._load_paths(conf, self.files[level])
 
         # Auto-verify sections
@@ -373,7 +405,7 @@ class Config(dict):
         _merge(merged_conf, conf)
         self.validate(merged_conf)
 
-        _save_config(self.files[level], conf)
+        self._save_config(level, conf)
         self.load()
 
     @staticmethod
@@ -384,20 +416,6 @@ class Config(dict):
             raise ConfigError(str(exc)) from None
 
 
-def _load_config(filename):
-    conf_obj = configobj.ConfigObj(filename)
-    return _parse_remotes(_lower_keys(conf_obj.dict()))
-
-
-def _save_config(filename, conf_dict):
-    logger.debug(f"Writing '{filename}'.")
-    os.makedirs(os.path.dirname(filename), exist_ok=True)
-
-    config = configobj.ConfigObj(_pack_remotes(conf_dict))
-    config.filename = filename
-    config.write()
-
-
 def _parse_remotes(conf):
     result = {"remote": {}}
 

dvc/daemon.py

@@ -3,6 +3,7 @@
 import inspect
 import logging
 import os
+import platform
 import sys
 from subprocess import Popen
 
@@ -15,6 +16,14 @@ CREATE_NEW_PROCESS_GROUP = 0x00000200
 DETACHED_PROCESS = 0x00000008
 
 
+def _popen(cmd, **kwargs):
+    prefix = [sys.executable]
+    if not is_binary():
+        prefix += [sys.argv[0]]
+
+    return Popen(prefix + cmd, close_fds=True, shell=False, **kwargs)
+
+
 def _spawn_windows(cmd, env):
     from subprocess import STARTUPINFO, STARTF_USESHOWWINDOW
 
@@ -23,17 +32,14 @@ def _spawn_windows(cmd, env):
     startupinfo = STARTUPINFO()
     startupinfo.dwFlags |= STARTF_USESHOWWINDOW
 
-    Popen(
-        cmd,
-        env=env,
-        close_fds=True,
-        shell=False,
-        creationflags=creationflags,
-        startupinfo=startupinfo,
+    _popen(
+        cmd, env=env, creationflags=creationflags, startupinfo=startupinfo,
     ).communicate()
 
 
 def _spawn_posix(cmd, env):
+    from dvc.main import main
+
     # NOTE: using os._exit instead of sys.exit, because dvc built
     # with PyInstaller has trouble with SystemExit exception and throws
     # errors such as "[26338] Failed to execute script __main__"
@@ -59,7 +65,13 @@ def _spawn_posix(cmd, env):
     sys.stdout.close()
     sys.stderr.close()
 
-    Popen(cmd, env=env, close_fds=True, shell=False).communicate()
+    if platform.system() == "Darwin":
+        # workaround for MacOS bug
+        # https://github.com/iterative/dvc/issues/4294
+        _popen(cmd, env=env).communicate()
+    else:
+        os.environ.update(env)
+        main(cmd)
 
     os._exit(0)  # pylint: disable=protected-access
 
@@ -87,10 +99,7 @@ def daemon(args):
         logger.debug("skipping launching a new daemon.")
         return
 
-    cmd = [sys.executable]
-    if not is_binary():
-        cmd += [sys.argv[0]]
-    cmd += ["daemon", "-q"] + args
+    cmd = ["daemon", "-q"] + args
 
     env = fix_env()
     file_path = os.path.abspath(inspect.stack()[0][1])

dvc/data_cloud.py

@@ -92,7 +92,7 @@ class DataCloud:
 
     def _save_pulled_checksums(self, cache):
         for checksum in cache.scheme_keys("local"):
-            cache_file = self.repo.cache.local.hash_to_path_info(checksum)
+            cache_file = self.repo.cache.local.tree.hash_to_path_info(checksum)
             if self.repo.cache.local.tree.exists(cache_file):
                 # We can safely save here, as existing corrupted files will
                 # be removed upon status, while files corrupted during

dvc/dependency/__init__.py

@@ -12,9 +12,9 @@ from dvc.dependency.param import ParamsDependency
 from dvc.dependency.s3 import S3Dependency
 from dvc.dependency.ssh import SSHDependency
 from dvc.output.base import BaseOutput
-from dvc.remote import get_remote
 from dvc.scheme import Schemes
 
+from ..tree import get_cloud_tree
 from .repo import RepoDependency
 
 DEPS = [
@@ -54,8 +54,8 @@ SCHEMA.update(ParamsDependency.PARAM_SCHEMA)
 def _get(stage, p, info):
     parsed = urlparse(p) if p else None
     if parsed and parsed.scheme == "remote":
-        remote = get_remote(stage.repo, name=parsed.netloc)
-        return DEP_MAP[remote.scheme](stage, p, info, remote=remote)
+        tree = get_cloud_tree(stage.repo, name=parsed.netloc)
+        return DEP_MAP[tree.scheme](stage, p, info, tree=tree)
 
     if info and info.get(RepoDependency.PARAM_REPO):
         repo = info.pop(RepoDependency.PARAM_REPO)

dvc/dependency/azure.py

@@ -1,7 +1,8 @@
 from dvc.dependency.base import BaseDependency
 from dvc.output.base import BaseOutput
-from dvc.remote.azure import AzureRemoteTree
+
+from ..tree.azure import AzureTree
 
 
 class AzureDependency(BaseDependency, BaseOutput):
-    TREE_CLS = AzureRemoteTree
+    TREE_CLS = AzureTree

dvc/dependency/http.py

@@ -1,7 +1,8 @@
 from dvc.dependency.base import BaseDependency
 from dvc.output.base import BaseOutput
-from dvc.remote.http import HTTPRemoteTree
+
+from ..tree.http import HTTPTree
 
 
 class HTTPDependency(BaseDependency, BaseOutput):
-    TREE_CLS = HTTPRemoteTree
+    TREE_CLS = HTTPTree

dvc/dependency/https.py

@@ -1,7 +1,6 @@
-from dvc.remote.https import HTTPSRemoteTree
-
+from ..tree.https import HTTPSTree
 from .http import HTTPDependency
 
 
 class HTTPSDependency(HTTPDependency):
-    TREE_CLS = HTTPSRemoteTree
+    TREE_CLS = HTTPSTree

dvc/dependency/param.py

@@ -2,6 +2,7 @@ import os
 from collections import defaultdict
 
 import dpath.util
+import toml
 import yaml
 from voluptuous import Any
 
@@ -21,6 +22,8 @@ class ParamsDependency(LocalDependency):
     PARAM_PARAMS = "params"
     PARAM_SCHEMA = {PARAM_PARAMS: Any(dict, list, None)}
     DEFAULT_PARAMS_FILE = "params.yaml"
+    PARAMS_FILE_LOADERS = defaultdict(lambda: yaml.safe_load)
+    PARAMS_FILE_LOADERS.update({".toml": toml.load})
 
     def __init__(self, stage, path, params):
         info = {}
@@ -45,9 +48,8 @@ class ParamsDependency(LocalDependency):
         if not values:
             return
         for param in self.params:
-            value = values.get(param)
-            if value:
-                self.info[param] = value
+            if param in values:
+                self.info[param] = values[param]
 
     def save(self):
         super().save()
@@ -88,8 +90,10 @@ class ParamsDependency(LocalDependency):
 
         with self.repo.tree.open(self.path_info, "r") as fobj:
             try:
-                config = yaml.safe_load(fobj)
-            except yaml.YAMLError as exc:
+                config = self.PARAMS_FILE_LOADERS[
+                    self.path_info.suffix.lower()
+                ](fobj)
+            except (yaml.YAMLError, toml.TomlDecodeError) as exc:
                 raise BadParamFileError(
                     f"Unable to read parameters from '{self}'"
                 ) from exc

dvc/dependency/repo.py

@@ -26,7 +26,7 @@ class RepoDependency(LocalDependency):
         self.def_repo = def_repo
         super().__init__(stage, *args, **kwargs)
 
-    def _parse_path(self, remote, path):
+    def _parse_path(self, tree, path):
         return None
 
     @property
@@ -64,7 +64,7 @@ class RepoDependency(LocalDependency):
 
                 # We are polluting our repo cache with some dir listing here
                 if tree.isdir(path):
-                    return self.repo.cache.local.get_hash(path, tree)
+                    return self.repo.cache.local.tree.get_hash(path, tree=tree)
                 return tree.get_file_hash(path)
 
     def status(self):
@@ -98,3 +98,9 @@ class RepoDependency(LocalDependency):
 
         with self._make_repo(locked=False) as repo:
             self.def_repo[self.PARAM_REV_LOCK] = repo.get_rev()
+
+    def changed_checksum(self):
+        # From current repo point of view what describes RepoDependency is its
+        # origin project url and rev_lock, and it makes RepoDependency
+        # immutable, hence its impossible for checksum to change.
+        return False

dvc/dvcfile.py

@@ -1,3 +1,4 @@
+import collections
 import contextlib
 import logging
 import os
@@ -101,13 +102,13 @@ class FileMixin:
 
     @classmethod
     def validate(cls, d, fname=None):
-        assert cls.SCHEMA
+        assert isinstance(cls.SCHEMA, collections.abc.Callable)
         try:
-            cls.SCHEMA(d)
+            cls.SCHEMA(d)  # pylint: disable=not-callable
         except MultipleInvalid as exc:
             raise StageFileFormatError(f"'{fname}' format error: {exc}")
 
-    def remove(self, force=False):
+    def remove(self, force=False):  # pylint: disable=unused-argument
         with contextlib.suppress(FileNotFoundError):
             os.unlink(self.path)
 
@@ -118,9 +119,6 @@ class FileMixin:
 class SingleStageFile(FileMixin):
     from dvc.schema import COMPILED_SINGLE_STAGE_SCHEMA as SCHEMA
 
-    def __init__(self, repo, path):
-        super().__init__(repo, path)
-
     @property
     def stage(self):
         data, raw = self._load()
@@ -143,7 +141,7 @@ class SingleStageFile(FileMixin):
         dump_yaml(self.path, serialize.to_single_stage_file(stage))
         self.repo.scm.track_file(self.relpath)
 
-    def remove_stage(self, stage):
+    def remove_stage(self, stage):  # pylint: disable=unused-argument
         self.remove()
 
 
@@ -156,7 +154,9 @@ class PipelineFile(FileMixin):
     def _lockfile(self):
         return Lockfile(self.repo, os.path.splitext(self.path)[0] + ".lock")
 
-    def dump(self, stage, update_pipeline=False, no_lock=False):
+    def dump(
+        self, stage, update_pipeline=False, no_lock=False, **kwargs
+    ):  # pylint: disable=arguments-differ
         """Dumps given stage appropriately in the dvcfile."""
         from dvc.stage import PipelineStage
 
@@ -233,7 +233,11 @@ class PipelineFile(FileMixin):
 
         logger.debug("Removing '%s' from '%s'", stage.name, self.path)
         del d["stages"][stage.name]
-        dump_yaml(self.path, d)
+
+        if d["stages"]:
+            dump_yaml(self.path, d)
+        else:
+            super().remove()
 
 
 class Lockfile(FileMixin):
@@ -286,7 +290,10 @@ class Lockfile(FileMixin):
         logger.debug("Removing '%s' from '%s'", stage.name, self.path)
         del d[stage.name]
 
-        dump_yaml(self.path, d)
+        if d:
+            dump_yaml(self.path, d)
+        else:
+            self.remove()
 
 
 class Dvcfile:

dvc/external_repo.py

@@ -19,8 +19,9 @@ from dvc.exceptions import (
 from dvc.path_info import PathInfo
 from dvc.repo import Repo
 from dvc.repo.tree import RepoTree
+from dvc.scm.base import CloneError
 from dvc.scm.git import Git
-from dvc.scm.tree import is_working_tree
+from dvc.tree.local import LocalTree
 from dvc.utils.fs import remove
 
 logger = logging.getLogger(__name__)
@@ -31,7 +32,10 @@ def external_repo(url, rev=None, for_write=False):
     logger.debug("Creating external repo %s@%s", url, rev)
     path = _cached_clone(url, rev, for_write=for_write)
     if not rev:
-        rev = "HEAD"
+        # Local HEAD points to the tip of whatever branch we first cloned from
+        # (which may not be the default branch), use origin/HEAD here to get
+        # the tip of the default branch
+        rev = "refs/remotes/origin/HEAD"
     try:
         repo = ExternalRepo(path, url, rev, for_write=for_write)
     except NotDvcRepoError:
@@ -59,7 +63,7 @@ CACHE_DIRS = {}
 
 def clean_repos():
     # Outside code should not see cache while we are removing
-    paths = list(CLONES.values()) + list(CACHE_DIRS.values())
+    paths = [path for path, _ in CLONES.values()] + list(CACHE_DIRS.values())
     CLONES.clear()
     CACHE_DIRS.clear()
 
@@ -68,9 +72,13 @@ def clean_repos():
 
 
 class BaseExternalRepo:
+    # pylint: disable=no-member
 
     _local_cache = None
 
+    def __str__(self):
+        return self.url
+
     @property
     def local_cache(self):
         if hasattr(self, "cache"):
@@ -98,10 +106,8 @@ class BaseExternalRepo:
         return RepoTree(self, fetch=True)
 
     def get_rev(self):
-        if is_working_tree(self.tree):
+        if isinstance(self.tree, LocalTree):
             return self.scm.get_rev()
-        if hasattr(self.tree, "tree"):
-            return self.tree.tree.rev
         return self.tree.rev
 
     def fetch_external(self, paths: Iterable, **kwargs):
@@ -249,10 +255,10 @@ def _cached_clone(url, rev, for_write=False):
     """
     # even if we have already cloned this repo, we may need to
     # fetch/fast-forward to get specified rev
-    clone_path = _clone_default_branch(url, rev)
+    clone_path, shallow = _clone_default_branch(url, rev, for_write=for_write)
 
     if not for_write and (url) in CLONES:
-        return CLONES[url]
+        return CLONES[url][0]
 
     # Copy to a new dir to keep the clone clean
     repo_path = tempfile.mkdtemp("dvc-erepo")
@@ -263,17 +269,17 @@ def _cached_clone(url, rev, for_write=False):
     if for_write:
         _git_checkout(repo_path, rev)
     else:
-        CLONES[url] = repo_path
+        CLONES[url] = (repo_path, shallow)
     return repo_path
 
 
 @wrap_with(threading.Lock())
-def _clone_default_branch(url, rev):
+def _clone_default_branch(url, rev, for_write=False):
     """Get or create a clean clone of the url.
 
     The cloned is reactualized with git pull unless rev is a known sha.
     """
-    clone_path = CLONES.get(url)
+    clone_path, shallow = CLONES.get(url, (None, False))
 
     git = None
     try:
@@ -281,18 +287,55 @@ def _clone_default_branch(url, rev):
             git = Git(clone_path)
             # Do not pull for known shas, branches and tags might move
             if not Git.is_sha(rev) or not git.has_rev(rev):
-                logger.debug("erepo: git pull %s", url)
-                git.pull()
+                if shallow:
+                    # If we are missing a rev in a shallow clone, fallback to
+                    # a full (unshallowed) clone. Since fetching specific rev
+                    # SHAs is only available in certain git versions, if we
+                    # have need to reference multiple specific revs for a
+                    # given repo URL it is easier/safer for us to work with
+                    # full clones in this case.
+                    logger.debug("erepo: unshallowing clone for '%s'", url)
+                    _unshallow(git)
+                    shallow = False
+                    CLONES[url] = (clone_path, shallow)
+                else:
+                    logger.debug("erepo: git pull '%s'", url)
+                    git.pull()
         else:
-            logger.debug("erepo: git clone %s to a temporary dir", url)
+            logger.debug("erepo: git clone '%s' to a temporary dir", url)
             clone_path = tempfile.mkdtemp("dvc-clone")
-            git = Git.clone(url, clone_path)
-            CLONES[url] = clone_path
+            if not for_write and rev and not Git.is_sha(rev):
+                # If rev is a tag or branch name try shallow clone first
+                try:
+                    git = Git.clone(url, clone_path, shallow_branch=rev)
+                    shallow = True
+                    logger.debug(
+                        "erepo: using shallow clone for branch '%s'", rev
+                    )
+                except CloneError:
+                    pass
+            if not git:
+                git = Git.clone(url, clone_path)
+                shallow = False
+            CLONES[url] = (clone_path, shallow)
     finally:
         if git:
             git.close()
 
-    return clone_path
+    return clone_path, shallow
+
+
+def _unshallow(git):
+    if git.repo.head.is_detached:
+        # If this is a detached head (i.e. we shallow cloned a tag) switch to
+        # the default branch
+        origin_refs = git.repo.remotes["origin"].refs
+        ref = origin_refs["HEAD"].reference
+        branch_name = ref.name.split("/")[-1]
+        branch = git.repo.create_head(branch_name, ref)
+        branch.set_tracking_branch(ref)
+        branch.checkout()
+    git.pull(unshallow=True)
 
 
 def _git_checkout(repo_path, rev):

dvc/ignore.py

@@ -1,14 +1,14 @@
 import logging
 import os
 import re
-from itertools import groupby
+from itertools import groupby, takewhile
 
-from funcy import cached_property
 from pathspec.patterns import GitWildMatchPattern
 from pathspec.util import normalize_file
+from pygtrie import StringTrie
 
 from dvc.path_info import PathInfo
-from dvc.scm.tree import BaseTree
+from dvc.pathspec_math import merge_patterns
 from dvc.system import System
 from dvc.utils import relpath
 
@@ -23,39 +23,46 @@ class DvcIgnore:
 
 
 class DvcIgnorePatterns(DvcIgnore):
-    def __init__(self, ignore_file_path, tree):
+    def __init__(self, pattern_list, dirname):
+
+        self.pattern_list = pattern_list
+        self.dirname = dirname
+        self.prefix = self.dirname + os.sep
+
+        regex_pattern_list = map(
+            GitWildMatchPattern.pattern_to_regex, pattern_list
+        )
+
+        self.ignore_spec = [
+            (ignore, re.compile("|".join(item[0] for item in group)))
+            for ignore, group in groupby(regex_pattern_list, lambda x: x[1])
+            if ignore is not None
+        ]
+
+    @classmethod
+    def from_files(cls, ignore_file_path, tree):
         assert os.path.isabs(ignore_file_path)
-
-        self.ignore_file_path = ignore_file_path
-        self.dirname = os.path.normpath(os.path.dirname(ignore_file_path))
-
+        dirname = os.path.normpath(os.path.dirname(ignore_file_path))
         with tree.open(ignore_file_path, encoding="utf-8") as fobj:
-            path_spec_lines = fobj.readlines()
-            regex_pattern_list = map(
-                GitWildMatchPattern.pattern_to_regex, path_spec_lines
-            )
-            self.ignore_spec = [
-                (ignore, re.compile("|".join(item[0] for item in group)))
-                for ignore, group in groupby(
-                    regex_pattern_list, lambda x: x[1]
-                )
-                if ignore is not None
+            path_spec_lines = [
+                line for line in map(str.strip, fobj.readlines()) if line
             ]
 
+        return cls(path_spec_lines, dirname)
+
     def __call__(self, root, dirs, files):
         files = [f for f in files if not self.matches(root, f)]
-        dirs = [d for d in dirs if not self.matches(root, d)]
+        dirs = [d for d in dirs if not self.matches(root, d, True)]
 
         return dirs, files
 
-    def matches(self, dirname, basename):
+    def matches(self, dirname, basename, is_dir=False):
         # NOTE: `relpath` is too slow, so we have to assume that both
         # `dirname` and `self.dirname` are relative or absolute together.
-        prefix = self.dirname + os.sep
         if dirname == self.dirname:
             path = basename
-        elif dirname.startswith(prefix):
-            rel = dirname[len(prefix) :]
+        elif dirname.startswith(self.prefix):
+            rel = dirname[len(self.prefix) :]
             # NOTE: `os.path.join` is ~x5.5 slower
             path = f"{rel}{os.sep}{basename}"
         else:
@@ -63,188 +70,178 @@ class DvcIgnorePatterns(DvcIgnore):
 
         if not System.is_unix():
             path = normalize_file(path)
-        return self.ignore(path)
+        return self.ignore(path, is_dir)
 
-    def ignore(self, path):
+    def ignore(self, path, is_dir):
         result = False
-        for ignore, pattern in self.ignore_spec:
-            if pattern.match(path):
-                result = ignore
+        if is_dir:
+            path_dir = f"{path}/"
+            for ignore, pattern in self.ignore_spec:
+                if pattern.match(path) or pattern.match(path_dir):
+                    result = ignore
+        else:
+            for ignore, pattern in self.ignore_spec:
+                if pattern.match(path):
+                    result = ignore
         return result
 
     def __hash__(self):
-        return hash(self.ignore_file_path)
+        return hash(self.dirname + ":" + "\n".join(self.pattern_list))
 
     def __eq__(self, other):
         if not isinstance(other, DvcIgnorePatterns):
             return NotImplemented
+        return (self.dirname == other.dirname) & (
+            self.pattern_list == other.pattern_list
+        )
 
-        return self.ignore_file_path == other.ignore_file_path
+    def __bool__(self):
+        return bool(self.pattern_list)
 
 
-class DvcIgnoreDirs(DvcIgnore):
-    def __init__(self, basenames):
-        self.basenames = set(basenames)
+class DvcIgnoreFilterNoop:
+    def __init__(self, tree, root_dir):
+        pass
 
     def __call__(self, root, dirs, files):
-        dirs = [d for d in dirs if d not in self.basenames]
-
         return dirs, files
 
-    def __hash__(self):
-        return hash(tuple(self.basenames))
+    def is_ignored_dir(self, _):
+        return False
 
-    def __eq__(self, other):
-        if not isinstance(other, DvcIgnoreDirs):
-            return NotImplemented
-
-        return self.basenames == other.basenames
-
-
-class DvcIgnoreRepo(DvcIgnore):
-    def __call__(self, root, dirs, files):
-        def is_dvc_repo(directory):
-            from dvc.repo import Repo
-
-            return os.path.isdir(os.path.join(root, directory, Repo.DVC_DIR))
-
-        dirs = [d for d in dirs if not is_dvc_repo(d)]
-
-        return dirs, files
+    def is_ignored_file(self, _):
+        return False
 
 
 class DvcIgnoreFilter:
-    def __init__(self, tree, root_dir):
-        self.tree = tree
-        self.root_dir = root_dir
-        self.ignores = {
-            DvcIgnoreDirs([".git", ".hg", ".dvc"]),
-            DvcIgnoreRepo(),
-        }
-        for root, dirs, files in self.tree.walk(self.root_dir):
-            self._update(root)
-            dirs[:], files[:] = self(root, dirs, files)
-
-    def _update(self, dirname):
-        ignore_file_path = os.path.join(dirname, DvcIgnore.DVCIGNORE_FILE)
-        if self.tree.exists(ignore_file_path):
-            self.ignores.add(DvcIgnorePatterns(ignore_file_path, self.tree))
-
-    def __call__(self, root, dirs, files):
-        for ignore in self.ignores:
-            dirs, files = ignore(root, dirs, files)
-
-        return dirs, files
-
-
-class CleanTree(BaseTree):
-    def __init__(self, tree, tree_root=None):
-        self.tree = tree
-        if tree_root:
-            self._tree_root = tree_root
-        else:
-            self._tree_root = self.tree.tree_root
-
-    @cached_property
-    def dvcignore(self):
-        return DvcIgnoreFilter(self.tree, self.tree_root)
-
-    @property
-    def tree_root(self):
-        return self._tree_root
-
-    def open(self, path, mode="r", encoding="utf-8"):
-        if self.isfile(path):
-            return self.tree.open(path, mode, encoding)
-        raise FileNotFoundError
-
-    def exists(self, path):
-        if self.tree.exists(path) and self._parents_exist(path):
-            if self.tree.isdir(path):
-                return self._valid_dirname(path)
-            return self._valid_filename(path)
-        return False
-
-    def isdir(self, path):
-        return (
-            self.tree.isdir(path)
-            and self._parents_exist(path)
-            and self._valid_dirname(path)
-        )
-
-    def _valid_dirname(self, path):
-        path = os.path.abspath(path)
-        if path == self.tree_root:
-            return True
-        dirname, basename = os.path.split(path)
-        dirs, _ = self.dvcignore(dirname, [basename], [])
-        if dirs:
-            return True
-        return False
-
-    def isfile(self, path):
-        return (
-            self.tree.isfile(path)
-            and self._parents_exist(path)
-            and self._valid_filename(path)
-        )
-
-    def _valid_filename(self, path):
-        dirname, basename = os.path.split(os.path.normpath(path))
-        _, files = self.dvcignore(os.path.abspath(dirname), [], [basename])
-        if files:
-            return True
-        return False
-
-    def isexec(self, path):
-        return self.exists(path) and self.tree.isexec(path)
-
-    def _parents_exist(self, path):
+    @staticmethod
+    def _is_dvc_repo(root, directory):
         from dvc.repo import Repo
 
+        return os.path.isdir(os.path.join(root, directory, Repo.DVC_DIR))
+
+    def __init__(self, tree, root_dir):
+        from dvc.repo import Repo
+
+        default_ignore_patterns = [".hg/", ".git/", "{}/".format(Repo.DVC_DIR)]
+
+        self.tree = tree
+        self.root_dir = root_dir
+        self.ignores_trie_tree = StringTrie(separator=os.sep)
+        self.ignores_trie_tree[root_dir] = DvcIgnorePatterns(
+            default_ignore_patterns, root_dir
+        )
+        self._update(self.root_dir)
+
+    def _update(self, dirname):
+        old_pattern = self.ignores_trie_tree.longest_prefix(dirname).value
+        matches = old_pattern.matches(dirname, DvcIgnore.DVCIGNORE_FILE, False)
+
+        ignore_file_path = os.path.join(dirname, DvcIgnore.DVCIGNORE_FILE)
+        if not matches and self.tree.exists(
+            ignore_file_path, use_dvcignore=False
+        ):
+            new_pattern = DvcIgnorePatterns.from_files(
+                ignore_file_path, self.tree
+            )
+            if old_pattern:
+                self.ignores_trie_tree[dirname] = DvcIgnorePatterns(
+                    *merge_patterns(
+                        old_pattern.pattern_list,
+                        old_pattern.dirname,
+                        new_pattern.pattern_list,
+                        new_pattern.dirname,
+                    )
+                )
+            else:
+                self.ignores_trie_tree[dirname] = new_pattern
+        elif old_pattern:
+            self.ignores_trie_tree[dirname] = old_pattern
+
+        # NOTE: using `walk` + `break` because tree doesn't have `listdir()`
+        for root, dirs, _ in self.tree.walk(dirname, use_dvcignore=False):
+            self._update_sub_repo(root, dirs)
+            break
+
+    def _update_sub_repo(self, root, dirs):
+        for d in dirs:
+            if self._is_dvc_repo(root, d):
+                old_pattern = self.ignores_trie_tree.longest_prefix(root).value
+                if old_pattern:
+                    self.ignores_trie_tree[root] = DvcIgnorePatterns(
+                        *merge_patterns(
+                            old_pattern.pattern_list,
+                            old_pattern.dirname,
+                            ["/{}/".format(d)],
+                            root,
+                        )
+                    )
+                else:
+                    self.ignores_trie_tree[root] = DvcIgnorePatterns(
+                        ["/{}/".format(d)], root
+                    )
+
+    def __call__(self, root, dirs, files):
+        ignore_pattern = self._get_trie_pattern(root)
+        if ignore_pattern:
+            return ignore_pattern(root, dirs, files)
+        else:
+            return dirs, files
+
+    def _get_trie_pattern(self, dirname):
+        ignore_pattern = self.ignores_trie_tree.get(dirname)
+        if ignore_pattern:
+            return ignore_pattern
+
+        prefix = self.ignores_trie_tree.longest_prefix(dirname).key
+        if not prefix:
+            # outside of the repo
+            return None
+
+        dirs = list(
+            takewhile(
+                lambda path: path != prefix,
+                (parent.fspath for parent in PathInfo(dirname).parents),
+            )
+        )
+        dirs.reverse()
+        dirs.append(dirname)
+
+        for parent in dirs:
+            self._update(parent)
+
+        return self.ignores_trie_tree.get(dirname)
+
+    def _is_ignored(self, path, is_dir=False):
+        if self._outside_repo(path):
+            return True
+        dirname, basename = os.path.split(os.path.normpath(path))
+        ignore_pattern = self._get_trie_pattern(dirname)
+        if ignore_pattern:
+            return ignore_pattern.matches(dirname, basename, is_dir)
+        else:
+            return False
+
+    def is_ignored_dir(self, path):
+        path = os.path.abspath(path)
+        if path == self.root_dir:
+            return False
+
+        return self._is_ignored(path, True)
+
+    def is_ignored_file(self, path):
+        return self._is_ignored(path, False)
+
+    def _outside_repo(self, path):
         path = PathInfo(path)
 
-        # if parent is tree_root or inside a .dvc dir we can skip this check
-        if path.parent == self.tree_root or Repo.DVC_DIR in path.parts:
-            return True
-
-        # paths outside of the CleanTree root should be ignored
-        path = relpath(path, self.tree_root)
+        # paths outside of the repo should be ignored
+        path = relpath(path, self.root_dir)
         if path.startswith("..") or (
             os.name == "nt"
             and not os.path.commonprefix(
-                [os.path.abspath(path), self.tree_root]
+                [os.path.abspath(path), self.root_dir]
             )
         ):
-            return False
-
-        # check if parent directories are in our ignores, starting from
-        # tree_root
-        for parent_dir in reversed(PathInfo(path).parents):
-            dirname, basename = os.path.split(parent_dir)
-            if basename == ".":
-                # parent_dir == tree_root
-                continue
-            dirs, _ = self.dvcignore(os.path.abspath(dirname), [basename], [])
-            if not dirs:
-                return False
-        return True
-
-    def walk(self, top, topdown=True, onerror=None):
-        for root, dirs, files in self.tree.walk(
-            top, topdown=topdown, onerror=onerror
-        ):
-            dirs[:], files[:] = self.dvcignore(
-                os.path.abspath(root), dirs, files
-            )
-
-            yield root, dirs, files
-
-    def stat(self, path):
-        if self.exists(path):
-            return self.tree.stat(path)
-        raise FileNotFoundError
-
-    @property
-    def hash_jobs(self):
-        return self.tree.hash_jobs
+            return True
+        return False

dvc/lock.py

@@ -88,7 +88,9 @@ class HardlinkLock(flufl.lock.Lock):
         tmp_dir (str): a directory to store claim files.
     """
 
-    def __init__(self, lockfile, tmp_dir=None, **kwargs):
+    def __init__(
+        self, lockfile, tmp_dir=None, **kwargs
+    ):  # pylint: disable=super-init-not-called
         import socket
 
         self._tmp_dir = tmp_dir
@@ -117,7 +119,7 @@ class HardlinkLock(flufl.lock.Lock):
     def lockfile(self):
         return self._lockfile
 
-    def lock(self):
+    def lock(self):  # pylint: disable=arguments-differ
         try:
             super().lock(timedelta(seconds=DEFAULT_TIMEOUT))
         except flufl.lock.TimeOutError:

dvc/logger.py

@@ -34,6 +34,7 @@ def addLoggingLevel(levelName, levelNum, methodName=None):
 
     def logForLevel(self, message, *args, **kwargs):
         if self.isEnabledFor(levelNum):
+            # pylint: disable=protected-access
             self._log(levelNum, message, args, **kwargs)
 
     def logToRoot(message, *args, **kwargs):
@@ -137,7 +138,7 @@ class LoggerHandler(logging.StreamHandler):
             self.flush()
         except RecursionError:
             raise
-        except Exception:
+        except Exception:  # noqa, pylint: disable=broad-except
             self.handleError(record)
 
 

dvc/main.py

@@ -9,7 +9,7 @@ from dvc.config import ConfigError
 from dvc.exceptions import DvcException, DvcParserError, NotDvcRepoError
 from dvc.external_repo import clean_repos
 from dvc.logger import FOOTER, disable_other_loggers
-from dvc.remote.pool import close_pools
+from dvc.tree.pool import close_pools
 from dvc.utils import format_link
 
 # Workaround for CPython bug. See [1] and [2] for more info.
@@ -65,8 +65,8 @@ def main(argv=None):
     except DvcException:
         ret = 255
         logger.exception("")
-    except Exception as exc:  # pylint: disable=broad-except
-        if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
+    except OSError as exc:
+        if exc.errno == errno.EMFILE:
             logger.exception(
                 "too many open files, please visit "
                 "{} to see how to handle this "
@@ -78,10 +78,13 @@ def main(argv=None):
         else:
             logger.exception("unexpected error")
         ret = 255
+    except Exception:  # noqa, pylint: disable=broad-except
+        logger.exception("unexpected error")
+        ret = 255
     finally:
         logger.setLevel(outerLogLevel)
 
-        # Closing pools by-hand to prevent wierd messages when closing SSH
+        # Closing pools by-hand to prevent weird messages when closing SSH
         # connections. See https://github.com/iterative/dvc/issues/3248 for
         # more info.
         close_pools()

dvc/output/__init__.py

@@ -10,12 +10,13 @@ from dvc.output.hdfs import HDFSOutput
 from dvc.output.local import LocalOutput
 from dvc.output.s3 import S3Output
 from dvc.output.ssh import SSHOutput
-from dvc.remote import get_remote
-from dvc.remote.hdfs import HDFSRemoteTree
-from dvc.remote.local import LocalRemoteTree
-from dvc.remote.s3 import S3RemoteTree
 from dvc.scheme import Schemes
 
+from ..tree import get_cloud_tree
+from ..tree.hdfs import HDFSTree
+from ..tree.local import LocalTree
+from ..tree.s3 import S3Tree
+
 OUTS = [
     HDFSOutput,
     S3Output,
@@ -47,9 +48,9 @@ CHECKSUM_SCHEMA = Any(
 # so when a few types of outputs share the same name, we only need
 # specify it once.
 CHECKSUMS_SCHEMA = {
-    LocalRemoteTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
-    S3RemoteTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
-    HDFSRemoteTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
+    LocalTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
+    S3Tree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
+    HDFSTree.PARAM_CHECKSUM: CHECKSUM_SCHEMA,
 }
 
 SCHEMA = CHECKSUMS_SCHEMA.copy()
@@ -66,13 +67,13 @@ def _get(
     parsed = urlparse(p)
 
     if parsed.scheme == "remote":
-        remote = get_remote(stage.repo, name=parsed.netloc)
-        return OUTS_MAP[remote.scheme](
+        tree = get_cloud_tree(stage.repo, name=parsed.netloc)
+        return OUTS_MAP[tree.scheme](
             stage,
             p,
             info,
             cache=cache,
-            remote=remote,
+            tree=tree,
             metric=metric,
             plot=plot,
             persist=persist,
@@ -85,7 +86,7 @@ def _get(
                 p,
                 info,
                 cache=cache,
-                remote=None,
+                tree=None,
                 metric=metric,
                 plot=plot,
                 persist=persist,
@@ -95,7 +96,7 @@ def _get(
         p,
         info,
         cache=cache,
-        remote=None,
+        tree=None,
         metric=metric,
         plot=plot,
         persist=persist,

dvc/output/base.py

@@ -12,7 +12,8 @@ from dvc.exceptions import (
     DvcException,
     RemoteCacheRequiredError,
 )
-from dvc.remote.base import BaseRemoteTree, Remote
+
+from ..tree.base import BaseTree
 
 logger = logging.getLogger(__name__)
 
@@ -47,8 +48,7 @@ class OutputIsStageFileError(DvcException):
 class BaseOutput:
     IS_DEPENDENCY = False
 
-    REMOTE_CLS = Remote
-    TREE_CLS = BaseRemoteTree
+    TREE_CLS = BaseTree
 
     PARAM_PATH = "path"
     PARAM_CACHE = "cache"
@@ -85,7 +85,7 @@ class BaseOutput:
         stage,
         path,
         info=None,
-        remote=None,
+        tree=None,
         cache=True,
         metric=False,
         plot=False,
@@ -106,24 +106,23 @@ class BaseOutput:
         self.repo = stage.repo if stage else None
         self.def_path = path
         self.info = info
-        if remote:
-            self.remote = remote
+        if tree:
+            self.tree = tree
         else:
-            tree = self.TREE_CLS(self.repo, {})
-            self.remote = self.REMOTE_CLS(tree)
+            self.tree = self.TREE_CLS(self.repo, {})
         self.use_cache = False if self.IS_DEPENDENCY else cache
         self.metric = False if self.IS_DEPENDENCY else metric
         self.plot = False if self.IS_DEPENDENCY else plot
         self.persist = persist
 
-        self.path_info = self._parse_path(remote, path)
+        self.path_info = self._parse_path(tree, path)
         if self.use_cache and self.cache is None:
             raise RemoteCacheRequiredError(self.path_info)
 
-    def _parse_path(self, remote, path):
-        if remote:
+    def _parse_path(self, tree, path):
+        if tree:
             parsed = urlparse(path)
-            return remote.path_info / parsed.path.lstrip("/")
+            return tree.path_info / parsed.path.lstrip("/")
         return self.TREE_CLS.PATH_CLS(path)
 
     def __repr__(self):
@@ -163,33 +162,33 @@ class BaseOutput:
 
     @property
     def cache_path(self):
-        return self.cache.hash_to_path_info(self.checksum).url
+        return self.cache.tree.hash_to_path_info(self.checksum).url
 
     @property
     def checksum_type(self):
-        return self.remote.tree.PARAM_CHECKSUM
+        return self.tree.PARAM_CHECKSUM
 
     @property
     def checksum(self):
-        return self.info.get(self.remote.tree.PARAM_CHECKSUM)
+        return self.info.get(self.tree.PARAM_CHECKSUM)
 
     @checksum.setter
     def checksum(self, checksum):
-        self.info[self.remote.tree.PARAM_CHECKSUM] = checksum
+        self.info[self.tree.PARAM_CHECKSUM] = checksum
 
     def get_checksum(self):
-        return self.remote.get_hash(self.path_info)
+        return self.tree.get_hash(self.path_info)
 
     @property
     def is_dir_checksum(self):
-        return self.remote.is_dir_hash(self.checksum)
+        return self.tree.is_dir_hash(self.checksum)
 
     @property
     def exists(self):
-        return self.remote.tree.exists(self.path_info)
+        return self.tree.exists(self.path_info)
 
     def save_info(self):
-        return self.remote.save_info(self.path_info)
+        return self.tree.save_info(self.path_info)
 
     def changed_checksum(self):
         return self.checksum != self.get_checksum()
@@ -222,13 +221,15 @@ class BaseOutput:
 
     @property
     def is_empty(self):
-        return self.remote.tree.is_empty(self.path_info)
+        return self.tree.is_empty(self.path_info)
 
     def isdir(self):
-        return self.remote.tree.isdir(self.path_info)
+        return self.tree.isdir(self.path_info)
 
     def isfile(self):
-        return self.remote.tree.isfile(self.path_info)
+        return self.tree.isfile(self.path_info)
+
+    # pylint: disable=no-member
 
     def ignore(self):
         if not self.use_scm_ignore:
@@ -245,6 +246,8 @@ class BaseOutput:
 
         self.repo.scm.ignore_remove(self.fspath)
 
+    # pylint: enable=no-member
+
     def save(self):
         if not self.exists:
             raise self.DoesNotExistError(self)
@@ -312,7 +315,7 @@ class BaseOutput:
         raise DvcException(f"verify metric is not supported for {self.scheme}")
 
     def download(self, to):
-        self.remote.tree.download(self.path_info, to.path_info)
+        self.tree.download(self.path_info, to.path_info)
 
     def checkout(
         self,
@@ -338,7 +341,7 @@ class BaseOutput:
         )
 
     def remove(self, ignore_remove=False):
-        self.remote.tree.remove(self.path_info)
+        self.tree.remove(self.path_info)
         if self.scheme != "local":
             return
 
@@ -346,10 +349,11 @@ class BaseOutput:
             self.ignore_remove()
 
     def move(self, out):
+        # pylint: disable=no-member
         if self.scheme == "local" and self.use_scm_ignore:
             self.repo.scm.ignore_remove(self.fspath)
 
-        self.remote.tree.move(self.path_info, out.path_info)
+        self.tree.move(self.path_info, out.path_info)
         self.def_path = out.def_path
         self.path_info = out.path_info
         self.save()
@@ -368,7 +372,7 @@ class BaseOutput:
 
     def unprotect(self):
         if self.exists:
-            self.remote.tree.unprotect(self.path_info)
+            self.tree.unprotect(self.path_info)
 
     def get_dir_cache(self, **kwargs):
         if not self.is_dir_checksum:
@@ -428,8 +432,8 @@ class BaseOutput:
         filter_path = str(filter_info) if filter_info else None
         is_win = os.name == "nt"
         for entry in self.dir_cache:
-            checksum = entry[self.remote.tree.PARAM_CHECKSUM]
-            entry_relpath = entry[self.remote.tree.PARAM_RELPATH]
+            checksum = entry[self.tree.PARAM_CHECKSUM]
+            entry_relpath = entry[self.tree.PARAM_RELPATH]
             if is_win:
                 entry_relpath = entry_relpath.replace("/", os.sep)
             entry_path = os.path.join(path, entry_relpath)

dvc/output/gs.py

@@ -1,6 +1,7 @@
 from dvc.output.s3 import S3Output
-from dvc.remote.gs import GSRemoteTree
+
+from ..tree.gs import GSTree
 
 
 class GSOutput(S3Output):
-    TREE_CLS = GSRemoteTree
+    TREE_CLS = GSTree

dvc/output/hdfs.py

@@ -1,6 +1,7 @@
 from dvc.output.base import BaseOutput
-from dvc.remote.hdfs import HDFSRemoteTree
+
+from ..tree.hdfs import HDFSTree
 
 
 class HDFSOutput(BaseOutput):
-    TREE_CLS = HDFSRemoteTree
+    TREE_CLS = HDFSTree

dvc/output/local.py

@@ -5,16 +5,16 @@ from urllib.parse import urlparse
 from dvc.exceptions import DvcException
 from dvc.istextfile import istextfile
 from dvc.output.base import BaseOutput
-from dvc.remote.local import LocalRemote, LocalRemoteTree
 from dvc.utils import relpath
 from dvc.utils.fs import path_isin
 
+from ..tree.local import LocalTree
+
 logger = logging.getLogger(__name__)
 
 
 class LocalOutput(BaseOutput):
-    REMOTE_CLS = LocalRemote
-    TREE_CLS = LocalRemoteTree
+    TREE_CLS = LocalTree
     sep = os.sep
 
     def __init__(self, stage, path, *args, **kwargs):
@@ -22,11 +22,17 @@ class LocalOutput(BaseOutput):
             path = relpath(path, stage.wdir)
 
         super().__init__(stage, path, *args, **kwargs)
+        if (
+            self.is_in_repo
+            and self.repo
+            and isinstance(self.repo.tree, LocalTree)
+        ):
+            self.tree = self.repo.tree
 
-    def _parse_path(self, remote, path):
+    def _parse_path(self, tree, path):
         parsed = urlparse(path)
         if parsed.scheme == "remote":
-            p = remote.path_info / parsed.path.lstrip("/")
+            p = tree.path_info / parsed.path.lstrip("/")
         else:
             # NOTE: we can path either from command line or .dvc file,
             # so we should expect both posix and windows style paths.

dvc/output/s3.py

@@ -1,6 +1,7 @@
 from dvc.output.base import BaseOutput
-from dvc.remote.s3 import S3RemoteTree
+
+from ..tree.s3 import S3Tree
 
 
 class S3Output(BaseOutput):
-    TREE_CLS = S3RemoteTree
+    TREE_CLS = S3Tree

dvc/output/ssh.py

@@ -1,7 +1,7 @@
 from dvc.output.base import BaseOutput
-from dvc.remote.ssh import SSHRemote, SSHRemoteTree
+
+from ..tree.ssh import SSHTree
 
 
 class SSHOutput(BaseOutput):
-    REMOTE_CLS = SSHRemote
-    TREE_CLS = SSHRemoteTree
+    TREE_CLS = SSHTree

dvc/path_info.py

@@ -1,3 +1,4 @@
+# pylint: disable=protected-access
 import os
 import pathlib
 import posixpath
@@ -17,7 +18,7 @@ class _BasePath:
         return self.isin_or_eq(other) or other.isin(self)
 
     def isin_or_eq(self, other):
-        return self == other or self.isin(other)
+        return self == other or self.isin(other)  # pylint: disable=no-member
 
 
 class PathInfo(pathlib.PurePath, _BasePath):
@@ -37,7 +38,7 @@ class PathInfo(pathlib.PurePath, _BasePath):
         return cls._from_parts(args)
 
     def as_posix(self):
-        f = self._flavour
+        f = self._flavour  # pylint: disable=no-member
         # Unlike original implementation [1] that uses `str()` we actually need
         # to use `fspath`, because we've overridden `__str__` method to return
         # relative paths, which will break original `as_posix`.
@@ -264,7 +265,7 @@ class HTTPURLInfo(URLInfo):
         params=None,
         query=None,
         fragment=None,
-    ):
+    ):  # pylint: disable=arguments-differ
         assert bool(host) ^ bool(netloc)
 
         if netloc is not None:

dvc/pathspec_math.py

@@ -0,0 +1,87 @@
+# Path Specification Pattern Math
+# Including changing base dir of path specification patterns and merging
+# of two path specification patterns with different base
+# All the operations follow the documents of `gitignore`
+import os
+
+from pathspec.util import normalize_file
+
+from dvc.utils import relpath
+
+
+def _not_ignore(rule):
+    return (True, rule[1:]) if rule.startswith("!") else (False, rule)
+
+
+def _is_comment(rule):
+    return rule.startswith("#")
+
+
+def _remove_slash(rule):
+    if rule.startswith("\\"):
+        return rule[1:]
+    return rule
+
+
+def _match_all_level(rule):
+    if rule[:-1].find("/") >= 0 and not rule.startswith("**/"):
+        if rule.startswith("/"):
+            rule = rule[1:]
+        return False, rule
+    if rule.startswith("**/"):
+        rule = rule[3:]
+    return True, rule
+
+
+def change_rule(rule, rel):
+    rule = rule.strip()
+    if _is_comment(rule):
+        return rule
+    not_ignore, rule = _not_ignore(rule)
+    match_all, rule = _match_all_level(rule)
+    rule = _remove_slash(rule)
+    if not match_all:
+        rule = f"/{rule}"
+    else:
+        rule = f"/**/{rule}"
+    if not_ignore:
+        rule = f"!/{rel}{rule}"
+    else:
+        rule = f"/{rel}{rule}"
+    rule = normalize_file(rule)
+    return rule
+
+
+def _change_dirname(dirname, pattern_list, new_dirname):
+    if new_dirname == dirname:
+        return pattern_list
+    rel = relpath(dirname, new_dirname)
+    if rel.startswith(".."):
+        raise ValueError("change dirname can only change to parent path")
+
+    return [change_rule(rule, rel) for rule in pattern_list]
+
+
+def merge_patterns(pattern_a, prefix_a, pattern_b, prefix_b):
+    """
+    Merge two path specification patterns.
+
+    This implementation merge two path specification patterns on different
+    bases. It returns the longest common parent directory, and the patterns
+    based on this new base directory.
+    """
+    if not pattern_a:
+        return pattern_b, prefix_b
+    elif not pattern_b:
+        return pattern_a, prefix_a
+
+    longest_common_dir = os.path.commonpath([prefix_a, prefix_b])
+    new_pattern_a = _change_dirname(prefix_a, pattern_a, longest_common_dir)
+    new_pattern_b = _change_dirname(prefix_b, pattern_b, longest_common_dir)
+
+    if len(prefix_a) <= len(prefix_b):
+        merged_pattern = new_pattern_a + new_pattern_b
+    else:
+        merged_pattern = new_pattern_b + new_pattern_a
+
+    return merged_pattern, longest_common_dir

dvc/remote/__init__.py

@@ -1,82 +1,7 @@
-import posixpath
-from urllib.parse import urlparse
-
-from dvc.remote.azure import AzureRemoteTree
-from dvc.remote.base import Remote
-from dvc.remote.gdrive import GDriveRemoteTree
-from dvc.remote.gs import GSRemoteTree
-from dvc.remote.hdfs import HDFSRemoteTree
-from dvc.remote.http import HTTPRemoteTree
-from dvc.remote.https import HTTPSRemoteTree
-from dvc.remote.local import LocalRemote, LocalRemoteTree
-from dvc.remote.oss import OSSRemoteTree
-from dvc.remote.s3 import S3RemoteTree
-from dvc.remote.ssh import SSHRemote, SSHRemoteTree
-
-TREES = [
-    AzureRemoteTree,
-    GDriveRemoteTree,
-    GSRemoteTree,
-    HDFSRemoteTree,
-    HTTPRemoteTree,
-    HTTPSRemoteTree,
-    S3RemoteTree,
-    SSHRemoteTree,
-    OSSRemoteTree,
-    # NOTE: LocalRemoteTree is the default
-]
-
-
-def _get_tree(remote_conf):
-    for tree_cls in TREES:
-        if tree_cls.supported(remote_conf):
-            return tree_cls
-    return LocalRemoteTree
-
-
-def _get_conf(repo, **kwargs):
-    name = kwargs.get("name")
-    if name:
-        remote_conf = repo.config["remote"][name.lower()]
-    else:
-        remote_conf = kwargs
-    return _resolve_remote_refs(repo.config, remote_conf)
-
-
-def _resolve_remote_refs(config, remote_conf):
-    # Support for cross referenced remotes.
-    # This will merge the settings, shadowing base ref with remote_conf.
-    # For example, having:
-    #
-    #       dvc remote add server ssh://localhost
-    #       dvc remote modify server user root
-    #       dvc remote modify server ask_password true
-    #
-    #       dvc remote add images remote://server/tmp/pictures
-    #       dvc remote modify images user alice
-    #       dvc remote modify images ask_password false
-    #       dvc remote modify images password asdf1234
-    #
-    # Results on a config dictionary like:
-    #
-    #       {
-    #           "url": "ssh://localhost/tmp/pictures",
-    #           "user": "alice",
-    #           "password": "asdf1234",
-    #           "ask_password": False,
-    #       }
-    parsed = urlparse(remote_conf["url"])
-    if parsed.scheme != "remote":
-        return remote_conf
-
-    base = config["remote"][parsed.netloc]
-    url = posixpath.join(base["url"], parsed.path.lstrip("/"))
-    return {**base, **remote_conf, "url": url}
-
-
-def get_cloud_tree(repo, **kwargs):
-    remote_conf = _get_conf(repo, **kwargs)
-    return _get_tree(remote_conf)(repo, remote_conf)
+from ..tree import get_cloud_tree
+from .base import Remote
+from .local import LocalRemote
+from .ssh import SSHRemote
 
 
 def get_remote(repo, **kwargs):

dvc/remote/https.py

@@ -1,7 +0,0 @@
-from dvc.scheme import Schemes
-
-from .http import HTTPRemoteTree
-
-
-class HTTPSRemoteTree(HTTPRemoteTree):
-    scheme = Schemes.HTTPS

dvc/remote/index.py

@@ -66,7 +66,7 @@ class RemoteIndex:
     INDEX_SUFFIX = ".idx"
     VERSION = 1
     INDEX_TABLE = "remote_index"
-    INDEX_TABLE_LAYOUT = "checksum TEXT PRIMARY KEY, " "dir INTEGER NOT NULL"
+    INDEX_TABLE_LAYOUT = "checksum TEXT PRIMARY KEY, dir INTEGER NOT NULL"
 
     def __init__(self, repo, name, dir_suffix=".dir"):
         self.path = os.path.join(repo.index_dir, f"{name}{self.INDEX_SUFFIX}")

dvc/remote/local.py

@@ -1,726 +1,6 @@
-import errno
-import logging
-import os
-import stat
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from functools import partial, wraps
-
-from funcy import cached_property, concat
-from shortuuid import uuid
-
-from dvc.exceptions import DownloadError, DvcException, UploadError
-from dvc.path_info import PathInfo
-from dvc.progress import Tqdm
-from dvc.remote.base import (
-    STATUS_DELETED,
-    STATUS_MAP,
-    STATUS_MISSING,
-    STATUS_NEW,
-    BaseRemoteTree,
-    CloudCache,
-    Remote,
-    index_locked,
-)
-from dvc.remote.index import RemoteIndexNoop
-from dvc.scheme import Schemes
-from dvc.scm.tree import WorkingTree, is_working_tree
-from dvc.system import System
-from dvc.utils import file_md5, relpath, tmp_fname
-from dvc.utils.fs import (
-    copy_fobj_to_file,
-    copyfile,
-    makedirs,
-    move,
-    remove,
-    walk_files,
-)
-
-logger = logging.getLogger(__name__)
-
-
-class LocalRemoteTree(BaseRemoteTree):
-    scheme = Schemes.LOCAL
-    PATH_CLS = PathInfo
-    PARAM_CHECKSUM = "md5"
-    PARAM_PATH = "path"
-    TRAVERSE_PREFIX_LEN = 2
-    UNPACKED_DIR_SUFFIX = ".unpacked"
-
-    CACHE_MODE = 0o444
-    SHARED_MODE_MAP = {None: (0o644, 0o755), "group": (0o664, 0o775)}
-
-    def __init__(self, repo, config):
-        super().__init__(repo, config)
-        url = config.get("url")
-        self.path_info = self.PATH_CLS(url) if url else None
-
-    @property
-    def state(self):
-        from dvc.state import StateNoop
-
-        return self.repo.state if self.repo else StateNoop()
-
-    @cached_property
-    def work_tree(self):
-        # When using repo.brancher, repo.tree may change to/from WorkingTree to
-        # GitTree arbitarily. When repo.tree is GitTree, local cache needs to
-        # use its own WorkingTree instance.
-        if self.repo:
-            return WorkingTree(self.repo.root_dir)
-        return None
-
-    @staticmethod
-    def open(path_info, mode="r", encoding=None):
-        return open(path_info, mode=mode, encoding=encoding)
-
-    def exists(self, path_info):
-        assert isinstance(path_info, str) or path_info.scheme == "local"
-        if not self.repo:
-            return os.path.exists(path_info)
-        return self.work_tree.exists(path_info)
-
-    def isfile(self, path_info):
-        if not self.repo:
-            return os.path.isfile(path_info)
-        return self.work_tree.isfile(path_info)
-
-    def isdir(self, path_info):
-        if not self.repo:
-            return os.path.isdir(path_info)
-        return self.work_tree.isdir(path_info)
-
-    def iscopy(self, path_info):
-        return not (
-            System.is_symlink(path_info) or System.is_hardlink(path_info)
-        )
-
-    def walk_files(self, path_info, **kwargs):
-        for fname in self.work_tree.walk_files(path_info):
-            yield PathInfo(fname)
-
-    def is_empty(self, path_info):
-        path = path_info.fspath
-
-        if self.isfile(path_info) and os.path.getsize(path) == 0:
-            return True
-
-        if self.isdir(path_info) and len(os.listdir(path)) == 0:
-            return True
-
-        return False
-
-    def remove(self, path_info):
-        if isinstance(path_info, PathInfo):
-            if path_info.scheme != "local":
-                raise NotImplementedError
-            path = path_info.fspath
-        else:
-            path = path_info
-
-        if self.exists(path):
-            remove(path)
-
-    def makedirs(self, path_info):
-        makedirs(path_info, exist_ok=True, mode=self.dir_mode)
-
-    def move(self, from_info, to_info, mode=None):
-        if from_info.scheme != "local" or to_info.scheme != "local":
-            raise NotImplementedError
-
-        self.makedirs(to_info.parent)
-
-        if mode is None:
-            if self.isfile(from_info):
-                mode = self.file_mode
-            else:
-                mode = self.dir_mode
-
-        move(from_info, to_info, mode=mode)
-
-    def copy(self, from_info, to_info):
-        tmp_info = to_info.parent / tmp_fname(to_info.name)
-        try:
-            System.copy(from_info, tmp_info)
-            os.chmod(tmp_info, self.file_mode)
-            os.rename(tmp_info, to_info)
-        except Exception:
-            self.remove(tmp_info)
-            raise
-
-    def copy_fobj(self, fobj, to_info):
-        self.makedirs(to_info.parent)
-        tmp_info = to_info.parent / tmp_fname(to_info.name)
-        try:
-            copy_fobj_to_file(fobj, tmp_info)
-            os.chmod(tmp_info, self.file_mode)
-            os.rename(tmp_info, to_info)
-        except Exception:
-            self.remove(tmp_info)
-            raise
-
-    @staticmethod
-    def symlink(from_info, to_info):
-        System.symlink(from_info, to_info)
-
-    @staticmethod
-    def is_symlink(path_info):
-        return System.is_symlink(path_info)
-
-    def hardlink(self, from_info, to_info):
-        # If there are a lot of empty files (which happens a lot in datasets),
-        # and the cache type is `hardlink`, we might reach link limits and
-        # will get something like: `too many links error`
-        #
-        # This is because all those empty files will have the same hash
-        # (i.e. 68b329da9893e34099c7d8ad5cb9c940), therefore, they will be
-        # linked to the same file in the cache.
-        #
-        # From https://en.wikipedia.org/wiki/Hard_link
-        #   * ext4 limits the number of hard links on a file to 65,000
-        #   * Windows with NTFS has a limit of 1024 hard links on a file
-        #
-        # That's why we simply create an empty file rather than a link.
-        if self.getsize(from_info) == 0:
-            self.open(to_info, "w").close()
-
-            logger.debug(
-                "Created empty file: {src} -> {dest}".format(
-                    src=str(from_info), dest=str(to_info)
-                )
-            )
-            return
-
-        System.hardlink(from_info, to_info)
-
-    @staticmethod
-    def is_hardlink(path_info):
-        return System.is_hardlink(path_info)
-
-    def reflink(self, from_info, to_info):
-        tmp_info = to_info.parent / tmp_fname(to_info.name)
-        System.reflink(from_info, tmp_info)
-        # NOTE: reflink has its own separate inode, so you can set permissions
-        # that are different from the source.
-        os.chmod(tmp_info, self.file_mode)
-        os.rename(tmp_info, to_info)
-
-    def _unprotect_file(self, path):
-        if System.is_symlink(path) or System.is_hardlink(path):
-            logger.debug(f"Unprotecting '{path}'")
-            tmp = os.path.join(os.path.dirname(path), "." + uuid())
-
-            # The operations order is important here - if some application
-            # would access the file during the process of copyfile then it
-            # would get only the part of file. So, at first, the file should be
-            # copied with the temporary name, and then original file should be
-            # replaced by new.
-            copyfile(path, tmp, name="Unprotecting '{}'".format(relpath(path)))
-            remove(path)
-            os.rename(tmp, path)
-
-        else:
-            logger.debug(
-                "Skipping copying for '{}', since it is not "
-                "a symlink or a hardlink.".format(path)
-            )
-
-        os.chmod(path, self.file_mode)
-
-    def _unprotect_dir(self, path):
-        assert is_working_tree(self.repo.tree)
-
-        for fname in self.repo.tree.walk_files(path):
-            self._unprotect_file(fname)
-
-    def unprotect(self, path_info):
-        path = path_info.fspath
-        if not os.path.exists(path):
-            raise DvcException(f"can't unprotect non-existing data '{path}'")
-
-        if os.path.isdir(path):
-            self._unprotect_dir(path)
-        else:
-            self._unprotect_file(path)
-
-    def protect(self, path_info):
-        path = os.fspath(path_info)
-        mode = self.CACHE_MODE
-
-        try:
-            os.chmod(path, mode)
-        except OSError as exc:
-            # There is nothing we need to do in case of a read-only file system
-            if exc.errno == errno.EROFS:
-                return
-
-            # In shared cache scenario, we might not own the cache file, so we
-            # need to check if cache file is already protected.
-            if exc.errno not in [errno.EPERM, errno.EACCES]:
-                raise
-
-            actual = stat.S_IMODE(os.stat(path).st_mode)
-            if actual != mode:
-                raise
-
-    def is_protected(self, path_info):
-        try:
-            mode = os.stat(path_info).st_mode
-        except FileNotFoundError:
-            return False
-
-        return stat.S_IMODE(mode) == self.CACHE_MODE
-
-    def get_file_hash(self, path_info):
-        return file_md5(path_info)[0]
-
-    @staticmethod
-    def getsize(path_info):
-        return os.path.getsize(path_info)
-
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_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
-        )
-
-        self.protect(tmp_file)
-        os.rename(tmp_file, to_info)
-
-    @staticmethod
-    def _download(
-        from_info, to_file, name=None, no_progress_bar=False, **_kwargs
-    ):
-        copyfile(
-            from_info, to_file, no_progress_bar=no_progress_bar, name=name
-        )
-
-    def list_paths(self, prefix=None, progress_callback=None):
-        assert self.path_info is not None
-        if prefix:
-            path_info = self.path_info / prefix[:2]
-            if not self.exists(path_info):
-                return
-        else:
-            path_info = self.path_info
-        # NOTE: use utils.fs walk_files since tree.walk_files will not follow
-        # symlinks
-        if progress_callback:
-            for path in walk_files(path_info):
-                progress_callback()
-                yield path
-        else:
-            yield from walk_files(path_info)
-
-    def _remove_unpacked_dir(self, hash_):
-        info = self.hash_to_path_info(hash_)
-        path_info = info.with_name(info.name + self.UNPACKED_DIR_SUFFIX)
-        self.remove(path_info)
-
-
-def _log_exceptions(func, operation):
-    @wraps(func)
-    def wrapper(from_info, to_info, *args, **kwargs):
-        try:
-            func(from_info, to_info, *args, **kwargs)
-            return 0
-        except Exception as exc:
-            # NOTE: this means we ran out of file descriptors and there is no
-            # reason to try to proceed, as we will hit this error anyways.
-            if isinstance(exc, OSError) and exc.errno == errno.EMFILE:
-                raise
-
-            logger.exception(
-                "failed to %s '%s' to '%s'", operation, from_info, to_info
-            )
-            return 1
-
-    return wrapper
+from .base import Remote
+from .index import RemoteIndexNoop
 
 
 class LocalRemote(Remote):
     INDEX_CLS = RemoteIndexNoop
-
-
-class LocalCache(CloudCache):
-    DEFAULT_CACHE_TYPES = ["reflink", "copy"]
-    CACHE_MODE = LocalRemoteTree.CACHE_MODE
-
-    def __init__(self, tree):
-        super().__init__(tree)
-        self.cache_dir = tree.config.get("url")
-
-    @property
-    def cache_dir(self):
-        return self.tree.path_info.fspath if self.tree.path_info else None
-
-    @cache_dir.setter
-    def cache_dir(self, value):
-        self.tree.path_info = PathInfo(value) if value else None
-
-    @classmethod
-    def supported(cls, config):
-        return True
-
-    @cached_property
-    def cache_path(self):
-        return os.path.abspath(self.cache_dir)
-
-    def hash_to_path(self, hash_):
-        # NOTE: `self.cache_path` is already normalized so we can simply use
-        # `os.sep` instead of `os.path.join`. This results in this helper
-        # being ~5.5 times faster.
-        return f"{self.cache_path}{os.sep}{hash_[0:2]}{os.sep}{hash_[2:]}"
-
-    def hashes_exist(self, hashes, jobs=None, name=None):
-        return [
-            hash_
-            for hash_ in Tqdm(
-                hashes,
-                unit="file",
-                desc="Querying "
-                + ("cache in " + name if name else "local cache"),
-            )
-            if not self.changed_cache_file(hash_)
-        ]
-
-    def already_cached(self, path_info):
-        assert path_info.scheme in ["", "local"]
-
-        current_md5 = self.get_hash(path_info)
-
-        if not current_md5:
-            return False
-
-        return not self.changed_cache(current_md5)
-
-    def _verify_link(self, path_info, link_type):
-        if link_type == "hardlink" and self.tree.getsize(path_info) == 0:
-            return
-
-        super()._verify_link(path_info, link_type)
-
-    @index_locked
-    def status(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        # Return flattened dict containing all status info
-        dir_status, file_status, _ = self._status(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=download,
-        )
-        return dict(dir_status, **file_status)
-
-    def _status(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        """Return a tuple of (dir_status_info, file_status_info, dir_contents).
-
-        dir_status_info contains status for .dir files, file_status_info
-        contains status for all other files, and dir_contents is a dict of
-        {dir_hash: set(file_hash, ...)} which can be used to map
-        a .dir file to its file contents.
-        """
-        logger.debug(f"Preparing to collect status from {remote.path_info}")
-        md5s = set(named_cache.scheme_keys(self.scheme))
-
-        logger.debug("Collecting information from local cache...")
-        local_exists = frozenset(
-            self.hashes_exist(md5s, jobs=jobs, name=self.cache_dir)
-        )
-
-        # This is a performance optimization. We can safely assume that,
-        # if the resources that we want to fetch are already cached,
-        # there's no need to check the remote storage for the existence of
-        # those files.
-        if download and local_exists == md5s:
-            remote_exists = local_exists
-        else:
-            logger.debug("Collecting information from remote cache...")
-            remote_exists = set()
-            dir_md5s = set(named_cache.dir_keys(self.scheme))
-            if dir_md5s:
-                remote_exists.update(
-                    self._indexed_dir_hashes(named_cache, remote, dir_md5s)
-                )
-                md5s.difference_update(remote_exists)
-            if md5s:
-                remote_exists.update(
-                    remote.hashes_exist(
-                        md5s, jobs=jobs, name=str(remote.path_info)
-                    )
-                )
-        return self._make_status(
-            named_cache, show_checksums, local_exists, remote_exists
-        )
-
-    def _make_status(
-        self, named_cache, show_checksums, local_exists, remote_exists
-    ):
-        def make_names(hash_, names):
-            return {"name": hash_ if show_checksums else " ".join(names)}
-
-        dir_status = {}
-        file_status = {}
-        dir_contents = {}
-        for hash_, item in named_cache[self.scheme].items():
-            if item.children:
-                dir_status[hash_] = make_names(hash_, item.names)
-                dir_contents[hash_] = set()
-                for child_hash, child in item.children.items():
-                    file_status[child_hash] = make_names(
-                        child_hash, child.names
-                    )
-                    dir_contents[hash_].add(child_hash)
-            else:
-                file_status[hash_] = make_names(hash_, item.names)
-
-        self._fill_statuses(dir_status, local_exists, remote_exists)
-        self._fill_statuses(file_status, local_exists, remote_exists)
-
-        self._log_missing_caches(dict(dir_status, **file_status))
-
-        return dir_status, file_status, dir_contents
-
-    def _indexed_dir_hashes(self, named_cache, remote, dir_md5s):
-        # Validate our index by verifying all indexed .dir hashes
-        # still exist on the remote
-        indexed_dirs = set(remote.index.dir_hashes())
-        indexed_dir_exists = set()
-        if indexed_dirs:
-            indexed_dir_exists.update(
-                remote.tree.list_hashes_exists(indexed_dirs)
-            )
-            missing_dirs = indexed_dirs.difference(indexed_dir_exists)
-            if missing_dirs:
-                logger.debug(
-                    "Remote cache missing indexed .dir hashes '{}', "
-                    "clearing remote index".format(", ".join(missing_dirs))
-                )
-                remote.index.clear()
-
-        # Check if non-indexed (new) dir hashes exist on remote
-        dir_exists = dir_md5s.intersection(indexed_dir_exists)
-        dir_exists.update(
-            remote.tree.list_hashes_exists(dir_md5s - dir_exists)
-        )
-
-        # If .dir hash exists on the remote, assume directory contents
-        # still exists on the remote
-        for dir_hash in dir_exists:
-            file_hashes = list(named_cache.child_keys(self.scheme, dir_hash))
-            if dir_hash not in remote.index:
-                logger.debug(
-                    "Indexing new .dir '{}' with '{}' nested files".format(
-                        dir_hash, len(file_hashes)
-                    )
-                )
-                remote.index.update([dir_hash], file_hashes)
-            yield dir_hash
-            yield from file_hashes
-
-    @staticmethod
-    def _fill_statuses(hash_info_dir, local_exists, remote_exists):
-        # Using sets because they are way faster for lookups
-        local = set(local_exists)
-        remote = set(remote_exists)
-
-        for md5, info in hash_info_dir.items():
-            status = STATUS_MAP[(md5 in local, md5 in remote)]
-            info["status"] = status
-
-    def _get_plans(self, download, remote, status_info, status):
-        cache = []
-        path_infos = []
-        names = []
-        hashes = []
-        for md5, info in Tqdm(
-            status_info.items(), desc="Analysing status", unit="file"
-        ):
-            if info["status"] == status:
-                cache.append(self.hash_to_path_info(md5))
-                path_infos.append(remote.hash_to_path_info(md5))
-                names.append(info["name"])
-                hashes.append(md5)
-
-        if download:
-            to_infos = cache
-            from_infos = path_infos
-        else:
-            to_infos = path_infos
-            from_infos = cache
-
-        return from_infos, to_infos, names, hashes
-
-    def _process(
-        self,
-        named_cache,
-        remote,
-        jobs=None,
-        show_checksums=False,
-        download=False,
-    ):
-        logger.debug(
-            "Preparing to {} '{}'".format(
-                "download data from" if download else "upload data to",
-                remote.path_info,
-            )
-        )
-
-        if download:
-            func = partial(
-                _log_exceptions(remote.tree.download, "download"),
-                dir_mode=self.tree.dir_mode,
-                file_mode=self.tree.file_mode,
-            )
-            status = STATUS_DELETED
-            desc = "Downloading"
-        else:
-            func = _log_exceptions(remote.tree.upload, "upload")
-            status = STATUS_NEW
-            desc = "Uploading"
-
-        if jobs is None:
-            jobs = remote.tree.JOBS
-
-        dir_status, file_status, dir_contents = self._status(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=download,
-        )
-
-        dir_plans = self._get_plans(download, remote, dir_status, status)
-        file_plans = self._get_plans(download, remote, file_status, status)
-
-        total = len(dir_plans[0]) + len(file_plans[0])
-        if total == 0:
-            return 0
-
-        with Tqdm(total=total, unit="file", desc=desc) as pbar:
-            func = pbar.wrap_fn(func)
-            with ThreadPoolExecutor(max_workers=jobs) as executor:
-                if download:
-                    from_infos, to_infos, names, _ = (
-                        d + f for d, f in zip(dir_plans, file_plans)
-                    )
-                    fails = sum(
-                        executor.map(func, from_infos, to_infos, names)
-                    )
-                else:
-                    # for uploads, push files first, and any .dir files last
-
-                    file_futures = {}
-                    for from_info, to_info, name, hash_ in zip(*file_plans):
-                        file_futures[hash_] = executor.submit(
-                            func, from_info, to_info, name
-                        )
-                    dir_futures = {}
-                    for from_info, to_info, name, dir_hash in zip(*dir_plans):
-                        wait_futures = {
-                            future
-                            for file_hash, future in file_futures.items()
-                            if file_hash in dir_contents[dir_hash]
-                        }
-                        dir_futures[dir_hash] = executor.submit(
-                            self._dir_upload,
-                            func,
-                            wait_futures,
-                            from_info,
-                            to_info,
-                            name,
-                        )
-                    fails = sum(
-                        future.result()
-                        for future in concat(
-                            file_futures.values(), dir_futures.values()
-                        )
-                    )
-
-        if fails:
-            if download:
-                remote.index.clear()
-                raise DownloadError(fails)
-            raise UploadError(fails)
-
-        if not download:
-            # index successfully pushed dirs
-            for dir_hash, future in dir_futures.items():
-                if future.result() == 0:
-                    file_hashes = dir_contents[dir_hash]
-                    logger.debug(
-                        "Indexing pushed dir '{}' with "
-                        "'{}' nested files".format(dir_hash, len(file_hashes))
-                    )
-                    remote.index.update([dir_hash], file_hashes)
-
-        return len(dir_plans[0]) + len(file_plans[0])
-
-    @staticmethod
-    def _dir_upload(func, futures, from_info, to_info, name):
-        for future in as_completed(futures):
-            if future.result():
-                # do not upload this .dir file if any file in this
-                # directory failed to upload
-                logger.debug(
-                    "failed to upload full contents of '{}', "
-                    "aborting .dir file upload".format(name)
-                )
-                logger.error(f"failed to upload '{from_info}' to '{to_info}'")
-                return 1
-        return func(from_info, to_info, name)
-
-    @index_locked
-    def push(self, named_cache, remote, jobs=None, show_checksums=False):
-        return self._process(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=False,
-        )
-
-    @index_locked
-    def pull(self, named_cache, remote, jobs=None, show_checksums=False):
-        return self._process(
-            named_cache,
-            remote,
-            jobs=jobs,
-            show_checksums=show_checksums,
-            download=True,
-        )
-
-    @staticmethod
-    def _log_missing_caches(hash_info_dict):
-        missing_caches = [
-            (md5, info)
-            for md5, info in hash_info_dict.items()
-            if info["status"] == STATUS_MISSING
-        ]
-        if missing_caches:
-            missing_desc = "".join(
-                "\nname: {}, md5: {}".format(info["name"], md5)
-                for md5, info in missing_caches
-            )
-            msg = (
-                "Some of the cache files do not exist neither locally "
-                "nor on remote. Missing cache files: {}".format(missing_desc)
-            )
-            logger.warning(msg)

dvc/remote/ssh.py

@@ -0,0 +1,73 @@
+import errno
+import itertools
+import logging
+from concurrent.futures import ThreadPoolExecutor
+
+from dvc.progress import Tqdm
+from dvc.utils import to_chunks
+
+from .base import Remote
+
+logger = logging.getLogger(__name__)
+
+
+class SSHRemote(Remote):
+    def batch_exists(self, path_infos, callback):
+        def _exists(chunk_and_channel):
+            chunk, channel = chunk_and_channel
+            ret = []
+            for path in chunk:
+                try:
+                    channel.stat(path)
+                    ret.append(True)
+                except OSError as exc:
+                    if exc.errno != errno.ENOENT:
+                        raise
+                    ret.append(False)
+                callback(path)
+            return ret
+
+        with self.tree.ssh(path_infos[0]) as ssh:
+            channels = ssh.open_max_sftp_channels()
+            max_workers = len(channels)
+
+            with ThreadPoolExecutor(max_workers=max_workers) as executor:
+                paths = [path_info.path for path_info in path_infos]
+                chunks = to_chunks(paths, num_chunks=max_workers)
+                chunks_and_channels = zip(chunks, channels)
+                outcome = executor.map(_exists, chunks_and_channels)
+                results = list(itertools.chain.from_iterable(outcome))
+
+            return results
+
+    def hashes_exist(self, hashes, jobs=None, name=None):
+        """This is older implementation used in remote/base.py
+        We are reusing it in RemoteSSH, because SSH's batch_exists proved to be
+        faster than current approach (relying on exists(path_info)) applied in
+        remote/base.
+        """
+        if not self.tree.CAN_TRAVERSE:
+            return list(set(hashes) & set(self.tree.all()))
+
+        # possibly prompt for credentials before "Querying" progress output
+        self.tree.ensure_credentials()
+
+        with Tqdm(
+            desc="Querying "
+            + ("cache in " + name if name else "remote cache"),
+            total=len(hashes),
+            unit="file",
+        ) as pbar:
+
+            def exists_with_progress(chunks):
+                return self.batch_exists(chunks, callback=pbar.update_msg)
+
+            with ThreadPoolExecutor(
+                max_workers=jobs or self.tree.JOBS
+            ) as executor:
+                path_infos = [self.tree.hash_to_path_info(x) for x in hashes]
+                chunks = to_chunks(path_infos, num_chunks=self.tree.JOBS)
+                results = executor.map(exists_with_progress, chunks)
+                in_remote = itertools.chain.from_iterable(results)
+                ret = list(itertools.compress(hashes, in_remote))
+                return ret

dvc/repo/__init__.py

@@ -14,10 +14,8 @@ from dvc.exceptions import (
     NotDvcRepoError,
     OutputNotFoundError,
 )
-from dvc.ignore import CleanTree
 from dvc.path_info import PathInfo
 from dvc.repo.tree import RepoTree
-from dvc.scm.tree import is_working_tree
 from dvc.utils.fs import path_isin
 
 from ..stage.exceptions import StageFileDoesNotExistError, StageNotFound
@@ -31,6 +29,7 @@ def locked(f):
     @wraps(f)
     def wrapper(repo, *args, **kwargs):
         with repo.lock, repo.state:
+            # pylint: disable=protected-access
             repo._reset()
             ret = f(repo, *args, **kwargs)
             # Our graph cache is no longer valid after we release the repo.lock
@@ -73,31 +72,38 @@ class Repo:
         from dvc.scm import SCM
         from dvc.cache import Cache
         from dvc.data_cloud import DataCloud
+        from dvc.repo.experiments import Experiments
         from dvc.repo.metrics import Metrics
         from dvc.repo.plots import Plots
         from dvc.repo.params import Params
-        from dvc.scm.tree import WorkingTree
+        from dvc.tree.local import LocalTree
         from dvc.utils.fs import makedirs
         from dvc.stage.cache import StageCache
 
         if scm:
-            # use GitTree instead of WorkingTree as default repo tree instance
             tree = scm.get_tree(rev)
             self.root_dir = self.find_root(root_dir, tree)
             self.scm = scm
-            self.tree = tree
+            self.tree = scm.get_tree(
+                rev, use_dvcignore=True, dvcignore_root=self.root_dir
+            )
             self.state = StateNoop()
         else:
             root_dir = self.find_root(root_dir)
             self.root_dir = os.path.abspath(os.path.realpath(root_dir))
+            self.tree = LocalTree(
+                self,
+                {"url": self.root_dir},
+                use_dvcignore=True,
+                dvcignore_root=self.root_dir,
+            )
 
         self.dvc_dir = os.path.join(self.root_dir, self.DVC_DIR)
-        self.config = Config(self.dvc_dir)
+        self.config = Config(self.dvc_dir, tree=self.tree)
 
         if not scm:
             no_scm = self.config["core"].get("no_scm", False)
             self.scm = SCM(self.root_dir, no_scm=no_scm)
-            self.tree = WorkingTree(self.root_dir)
 
         self.tmp_dir = os.path.join(self.dvc_dir, "tmp")
         self.index_dir = os.path.join(self.tmp_dir, "index")
@@ -126,6 +132,11 @@ class Repo:
         self.plots = Plots(self)
         self.params = Params(self)
 
+        try:
+            self.experiments = Experiments(self)
+        except NotImplementedError:
+            self.experiments = None
+
         self._ignore()
 
     @property
@@ -134,13 +145,7 @@ class Repo:
 
     @tree.setter
     def tree(self, tree):
-        if is_working_tree(tree) or tree.tree_root == self.root_dir:
-            root = None
-        else:
-            root = self.root_dir
-        self._tree = (
-            tree if isinstance(tree, CleanTree) else CleanTree(tree, root)
-        )
+        self._tree = tree
         # Our graph cache is no longer valid, as it was based on the previous
         # tree.
         self._reset()
@@ -190,7 +195,12 @@ class Repo:
         return self.cache.local.tree.unprotect(PathInfo(target))
 
     def _ignore(self):
-        flist = [self.config.files["local"], self.tmp_dir]
+        flist = [
+            self.config.files["local"],
+            self.tmp_dir,
+        ]
+        if self.experiments:
+            flist.append(self.experiments.exp_dir)
 
         if path_isin(self.cache.local.cache_dir, self.root_dir):
             flist += [self.cache.local.cache_dir]
@@ -605,4 +615,4 @@ class Repo:
         self.__dict__.pop("graph", None)
         self.__dict__.pop("stages", None)
         self.__dict__.pop("pipelines", None)
-        self.__dict__.pop("dvcignore", None)
+        self.tree.__dict__.pop("dvcignore", None)

dvc/repo/add.py

@@ -109,17 +109,17 @@ def add(
 def _find_all_targets(repo, target, recursive):
     if os.path.isdir(target) and recursive:
         return [
-            fname
-            for fname in Tqdm(
+            os.fspath(path)
+            for path in Tqdm(
                 repo.tree.walk_files(target),
                 desc="Searching " + target,
                 bar_format=Tqdm.BAR_FMT_NOTOTAL,
                 unit="file",
             )
-            if not repo.is_dvc_internal(fname)
-            if not is_dvc_file(fname)
-            if not repo.scm.belongs_to_scm(fname)
-            if not repo.scm.is_tracked(fname)
+            if not repo.is_dvc_internal(os.fspath(path))
+            if not is_dvc_file(os.fspath(path))
+            if not repo.scm.belongs_to_scm(os.fspath(path))
+            if not repo.scm.is_tracked(os.fspath(path))
         ]
     return [target]
 
@@ -147,7 +147,7 @@ def _create_stages(repo, targets, fname, pbar=None, external=False):
         if stage:
             Dvcfile(repo, stage.path).remove()
 
-        repo._reset()
+        repo._reset()  # pylint: disable=protected-access
 
         if not stage:
             if pbar is not None:

dvc/repo/brancher.py

@@ -1,6 +1,6 @@
 from funcy import group_by
 
-from dvc.scm.tree import WorkingTree
+from dvc.tree.local import LocalTree
 
 
 def brancher(  # noqa: E302
@@ -29,7 +29,7 @@ def brancher(  # noqa: E302
 
     scm = self.scm
 
-    self.tree = WorkingTree(self.root_dir)
+    self.tree = LocalTree(self, {"url": self.root_dir}, use_dvcignore=True)
     yield "workspace"
 
     if revs and "workspace" in revs:
@@ -47,7 +47,9 @@ def brancher(  # noqa: E302
     try:
         if revs:
             for sha, names in group_by(scm.resolve_rev, revs).items():
-                self.tree = scm.get_tree(sha)
+                self.tree = scm.get_tree(
+                    sha, use_dvcignore=True, dvcignore_root=self.root_dir
+                )
                 # ignore revs that don't contain repo root
                 # (i.e. revs from before a subdir=True repo was init'ed)
                 if self.tree.exists(self.root_dir):

dvc/repo/diff.py

@@ -1,7 +1,7 @@
 import os
 
 from dvc.repo import locked
-from dvc.scm.tree import is_working_tree
+from dvc.tree.local import LocalTree
 
 
 @locked
@@ -33,11 +33,11 @@ def diff(self, a_rev="HEAD", b_rev=None):
                 else os.path.join(str(output), "")
             )
 
-        on_working_tree = is_working_tree(self.tree)
+        on_working_tree = isinstance(self.tree, LocalTree)
 
         def _to_checksum(output):
             if on_working_tree:
-                return self.cache.local.get_hash(output.path_info)
+                return self.cache.local.tree.get_hash(output.path_info)
             return output.checksum
 
         def _exists(output):

dvc/repo/experiments/__init__.py

@@ -0,0 +1,197 @@
+import logging
+import os
+import tempfile
+from contextlib import contextmanager
+
+from funcy import cached_property
+
+from dvc.exceptions import DvcException
+from dvc.scm.git import Git
+from dvc.stage.serialize import to_lockfile
+from dvc.utils import dict_sha256, env2bool, relpath
+from dvc.utils.fs import remove
+
+logger = logging.getLogger(__name__)
+
+
+class UnchangedExperimentError(DvcException):
+    pass
+
+
+class Experiments:
+    """Class that manages experiments in a DVC repo.
+
+    Args:
+        repo (dvc.repo.Repo): repo instance that these experiments belong to.
+    """
+
+    EXPERIMENTS_DIR = "experiments"
+
+    def __init__(self, repo):
+        if not (
+            env2bool("DVC_TEST")
+            or repo.config["core"].get("experiments", False)
+        ):
+            raise NotImplementedError
+
+        self.repo = repo
+
+    @cached_property
+    def exp_dir(self):
+        return os.path.join(self.repo.dvc_dir, self.EXPERIMENTS_DIR)
+
+    @cached_property
+    def scm(self):
+        """Experiments clone scm instance."""
+        if os.path.exists(self.exp_dir):
+            return Git(self.exp_dir)
+        return self._init_clone()
+
+    @cached_property
+    def exp_dvc_dir(self):
+        dvc_dir = relpath(self.repo.dvc_dir, self.repo.scm.root_dir)
+        return os.path.join(self.exp_dir, dvc_dir)
+
+    @cached_property
+    def exp_dvc(self):
+        """Return clone dvc Repo instance."""
+        from dvc.repo import Repo
+
+        return Repo(self.exp_dvc_dir)
+
+    @staticmethod
+    def exp_hash(stages):
+        exp_data = {}
+        for stage in stages:
+            exp_data.update(to_lockfile(stage))
+        return dict_sha256(exp_data)
+
+    @contextmanager
+    def chdir(self):
+        cwd = os.getcwd()
+        os.chdir(self.exp_dvc.root_dir)
+        yield
+        os.chdir(cwd)
+
+    def _init_clone(self):
+        src_dir = self.repo.scm.root_dir
+        logger.debug("Initializing experiments clone")
+        git = Git.clone(src_dir, self.exp_dir)
+        self._config_clone()
+        return git
+
+    def _config_clone(self):
+        dvc_dir = relpath(self.repo.dvc_dir, self.repo.scm.root_dir)
+        local_config = os.path.join(self.exp_dir, dvc_dir, "config.local")
+        cache_dir = self.repo.cache.local.cache_dir
+        logger.debug("Writing experiments local config '%s'", local_config)
+        with open(local_config, "w") as fobj:
+            fobj.write(f"[cache]\n    dir = {cache_dir}")
+
+    def _scm_checkout(self, rev):
+        self.scm.repo.git.reset(hard=True)
+        if not Git.is_sha(rev) or not self.scm.has_rev(rev):
+            self.scm.fetch(all=True)
+        logger.debug("Checking out base experiment commit '%s'", rev)
+        self.scm.checkout(rev)
+
+    def _patch_exp(self):
+        """Create a patch based on the current (parent) workspace and apply it
+        to the experiment workspace.
+        """
+        tmp = tempfile.NamedTemporaryFile(delete=False).name
+        try:
+            self.repo.scm.repo.git.diff(patch=True, output=tmp)
+            if os.path.getsize(tmp):
+                logger.debug("Patching experiment workspace")
+                self.scm.repo.git.apply(tmp)
+            else:
+                raise UnchangedExperimentError(
+                    "Experiment identical to baseline commit."
+                )
+        finally:
+            remove(tmp)
+
+    def _commit(self, stages, check_exists=True, branch=True, rev=None):
+        """Commit stages as an experiment and return the commit SHA."""
+        hash_ = self.exp_hash(stages)
+        exp_name = f"{rev[:7]}-{hash_}"
+        if branch:
+            if check_exists and exp_name in self.scm.list_branches():
+                logger.debug("Using existing experiment branch '%s'", exp_name)
+                return self.scm.resolve_rev(exp_name)
+            self.scm.checkout(exp_name, create_new=True)
+        logger.debug("Commit new experiment branch '%s'", exp_name)
+        self.scm.repo.git.add(A=True)
+        self.scm.commit(f"Add experiment {exp_name}")
+        return self.scm.get_rev()
+
+    def _reproduce(self, *args, **kwargs):
+        """Run `dvc repro` inside the experiments workspace."""
+        with self.chdir():
+            return self.exp_dvc.reproduce(*args, **kwargs)
+
+    def new(self, *args, workspace=True, **kwargs):
+        """Create a new experiment.
+
+        Experiment will be reproduced and checked out into the user's
+        workspace.
+        """
+        rev = self.repo.scm.get_rev()
+        self._scm_checkout(rev)
+        if workspace:
+            try:
+                self._patch_exp()
+            except UnchangedExperimentError as exc:
+                logger.info("Reproducing existing experiment '%s'.", rev[:7])
+                raise exc
+        else:
+            # configure params via command line here
+            pass
+        self.exp_dvc.checkout()
+        stages = self._reproduce(*args, **kwargs)
+        exp_rev = self._commit(stages, rev=rev)
+        self.checkout_exp(exp_rev, force=True)
+        logger.info("Generated experiment '%s'.", exp_rev[:7])
+        return stages
+
+    def checkout_exp(self, rev, force=False):
+        """Checkout an experiment to the user's workspace."""
+        from git.exc import GitCommandError
+        from dvc.repo.checkout import _checkout as dvc_checkout
+
+        if force:
+            self.repo.scm.repo.git.reset(hard=True)
+        logger.debug(f"checkout {rev}")
+        self._scm_checkout(rev)
+
+        tmp = tempfile.NamedTemporaryFile(delete=False).name
+        self.scm.repo.head.commit.diff("HEAD~1", patch=True, output=tmp)
+        try:
+            if os.path.getsize(tmp):
+                logger.debug("Patching local workspace")
+                self.repo.scm.repo.git.apply(tmp, reverse=True)
+            dvc_checkout(self.repo)
+        except GitCommandError:
+            raise DvcException(
+                "Checkout failed, experiment contains changes which "
+                "conflict with your current workspace. To overwrite "
+                "your workspace, use `dvc experiments checkout --force`."
+            )
+        finally:
+            remove(tmp)
+
+    def checkout(self, *args, **kwargs):
+        from dvc.repo.experiments.checkout import checkout
+
+        return checkout(self.repo, *args, **kwargs)
+
+    def diff(self, *args, **kwargs):
+        from dvc.repo.experiments.diff import diff
+
+        return diff(self.repo, *args, **kwargs)
+
+    def show(self, *args, **kwargs):
+        from dvc.repo.experiments.show import show
+
+        return show(self.repo, *args, **kwargs)

dvc/repo/experiments/checkout.py

@@ -0,0 +1,17 @@
+import logging
+
+from dvc.repo import locked
+from dvc.repo.scm_context import scm_context
+
+logger = logging.getLogger(__name__)
+
+
+@locked
+@scm_context
+def checkout(repo, rev, *args, **kwargs):
+    repo.experiments.checkout_exp(rev, *args, **kwargs)
+    logger.info(
+        "Changes for experiment '%s' have been applied to your current "
+        "workspace.",
+        rev,
+    )

dvc/repo/experiments/diff.py

@@ -0,0 +1,36 @@
+import logging
+
+from dvc.utils.diff import diff as _diff
+from dvc.utils.diff import format_dict
+
+logger = logging.getLogger(__name__)
+
+
+def diff(repo, *args, a_rev=None, b_rev=None, **kwargs):
+    from dvc.repo.experiments.show import _collect_experiment
+
+    if repo.scm.no_commits:
+        return {}
+
+    if a_rev:
+        with repo.experiments.chdir():
+            old = _collect_experiment(repo.experiments.exp_dvc, a_rev)
+    else:
+        old = _collect_experiment(repo, "HEAD")
+
+    if b_rev:
+        with repo.experiments.chdir():
+            new = _collect_experiment(repo.experiments.exp_dvc, b_rev)
+    else:
+        new = _collect_experiment(repo, "workspace")
+
+    with_unchanged = kwargs.pop("all", False)
+
+    return {
+        key: _diff(
+            format_dict(old[key]),
+            format_dict(new[key]),
+            with_unchanged=with_unchanged,
+        )
+        for key in ["metrics", "params"]
+    }

dvc/repo/experiments/show.py

@@ -0,0 +1,64 @@
+import logging
+import re
+from collections import OrderedDict, defaultdict
+
+from dvc.repo import locked
+from dvc.repo.metrics.show import _collect_metrics, _read_metrics
+from dvc.repo.params.show import _collect_configs, _read_params
+
+logger = logging.getLogger(__name__)
+
+
+EXP_RE = re.compile(r"(?P<rev_sha>[a-f0-9]{7})-(?P<exp_sha>[a-f0-9]+)")
+
+
+def _collect_experiment(repo, branch):
+    res = defaultdict(dict)
+    for rev in repo.brancher(revs=[branch]):
+        configs = _collect_configs(repo)
+        params = _read_params(repo, configs, rev)
+        if params:
+            res["params"] = params
+
+        metrics = _collect_metrics(repo, None, False)
+        vals = _read_metrics(repo, metrics, rev)
+        if vals:
+            res["metrics"] = vals
+
+    return res
+
+
+@locked
+def show(
+    repo, all_branches=False, all_tags=False, revs=None, all_commits=False
+):
+    res = defaultdict(OrderedDict)
+
+    if revs is None:
+        revs = [repo.scm.get_rev()]
+
+    revs = set(
+        repo.brancher(
+            revs=revs,
+            all_branches=all_branches,
+            all_tags=all_tags,
+            all_commits=all_commits,
+        )
+    )
+
+    for rev in revs:
+        res[rev]["baseline"] = _collect_experiment(repo, rev)
+
+    for exp_branch in repo.experiments.scm.list_branches():
+        m = re.match(EXP_RE, exp_branch)
+        if m:
+            rev = repo.scm.resolve_rev(m.group("rev_sha"))
+            if rev in revs:
+                exp_rev = repo.experiments.scm.resolve_rev(exp_branch)
+                with repo.experiments.chdir():
+                    experiment = _collect_experiment(
+                        repo.experiments.exp_dvc, exp_branch
+                    )
+                res[rev][exp_rev] = experiment
+
+    return res

dvc/repo/gc.py

@@ -52,10 +52,9 @@ def gc(
     from contextlib import ExitStack
     from dvc.repo import Repo
 
-    all_repos = []
-
-    if repos:
-        all_repos = [Repo(path) for path in repos]
+    if not repos:
+        repos = []
+    all_repos = [Repo(path) for path in repos]
 
     with ExitStack() as stack:
         for repo in all_repos:

dvc/repo/ls.py

@@ -53,9 +53,9 @@ def _ls(repo, path_info, recursive=None, dvc_only=False):
     def onerror(exc):
         raise exc
 
-    # use our own RepoTree instance instead of repo.repo_tree since we do not
-    # want fetch/stream enabled for ls
-    tree = RepoTree(repo)
+    # use our own RepoTree instance instead of repo.repo_tree since we want to
+    # fetch directory listings, but don't want to fetch file contents.
+    tree = RepoTree(repo, stream=True)
 
     ret = {}
     try:

dvc/repo/metrics/show.py

@@ -60,19 +60,19 @@ def _read_metrics(repo, metrics, rev):
         if not tree.exists(metric):
             continue
 
-        with tree.open(metric, "r") as fobj:
-            try:
+        try:
+            with tree.open(metric, "r") as fobj:
                 # NOTE this also supports JSON
                 val = yaml.safe_load(fobj)
-            except yaml.YAMLError:
-                logger.debug(
-                    "failed to read '%s' on '%s'", metric, rev, exc_info=True
-                )
-                continue
+        except (FileNotFoundError, yaml.YAMLError):
+            logger.debug(
+                "failed to read '%s' on '%s'", metric, rev, exc_info=True
+            )
+            continue
 
-            val = _extract_metrics(val)
-            if val:
-                res[str(metric)] = val
+        val = _extract_metrics(val)
+        if val:
+            res[str(metric)] = val
 
     return res
 

dvc/repo/params/show.py

@@ -1,5 +1,6 @@
 import logging
 
+import toml
 import yaml
 
 from dvc.dependency.param import ParamsDependency
@@ -34,8 +35,10 @@ def _read_params(repo, configs, rev):
 
         with repo.tree.open(config, "r") as fobj:
             try:
-                res[str(config)] = yaml.safe_load(fobj)
-            except yaml.YAMLError:
+                res[str(config)] = ParamsDependency.PARAMS_FILE_LOADERS[
+                    config.suffix.lower()
+                ](fobj)
+            except (yaml.YAMLError, toml.TomlDecodeError):
                 logger.debug(
                     "failed to read '%s' on '%s'", config, rev, exc_info=True
                 )

dvc/repo/plots/__init__.py

@@ -13,7 +13,8 @@ logger = logging.getLogger(__name__)
 class NotAPlotError(DvcException):
     def __init__(self, out):
         super().__init__(
-            f"'{out}' is not a plot. Use `dvc plots modify` to change that."
+            f"'{out}' is not a known plot. Use `dvc plots modify` to turn it "
+            "into one."
         )
 
 

dvc/repo/plots/data.py

@@ -181,7 +181,7 @@ class CSVPlotData(PlotData):
         super().__init__(filename, revision, content)
         self.delimiter = delimiter
 
-    def raw(self, header=True, **kwargs):
+    def raw(self, header=True, **kwargs):  # pylint: disable=arguments-differ
         first_row = first(csv.reader(io.StringIO(self.content)))
 
         if header:

dvc/repo/plots/template.py

@@ -29,9 +29,13 @@ class Template:
     EXTENSION = ".json"
     ANCHOR = "<DVC_METRIC_{}>"
 
+    DEFAULT_CONTENT = None
+    DEFAULT_NAME = None
+
     def __init__(self, content=None, name=None):
         self.content = self.DEFAULT_CONTENT if content is None else content
         self.name = name or self.DEFAULT_NAME
+        assert self.content and self.name
         self.filename = self.name + self.EXTENSION
 
     def render(self, data, props=None):

dvc/repo/pull.py

@@ -22,7 +22,7 @@ def pull(
     if isinstance(targets, str):
         targets = [targets]
 
-    processed_files_count = self._fetch(
+    processed_files_count = self._fetch(  # pylint: disable=protected-access
         targets,
         jobs,
         remote=remote,
@@ -33,7 +33,7 @@ def pull(
         recursive=recursive,
         run_cache=run_cache,
     )
-    stats = self._checkout(
+    stats = self._checkout(  # pylint: disable=protected-access
         targets=targets, with_deps=with_deps, force=force, recursive=recursive
     )
 

dvc/repo/reproduce.py

@@ -1,6 +1,7 @@
 import logging
 
 from dvc.exceptions import InvalidArgumentError, ReproductionError
+from dvc.repo.experiments import UnchangedExperimentError
 from dvc.repo.scm_context import scm_context
 
 from . import locked
@@ -69,6 +70,19 @@ def reproduce(
             "Neither `target` nor `--all-pipelines` are specified."
         )
 
+    experiment = kwargs.pop("experiment", False)
+    if experiment and self.experiments:
+        try:
+            return self.experiments.new(
+                target=target,
+                recursive=recursive,
+                all_pipelines=all_pipelines,
+                **kwargs
+            )
+        except UnchangedExperimentError:
+            # If experiment contains no changes, just run regular repro
+            pass
+
     interactive = kwargs.get("interactive", False)
     if not interactive:
         kwargs["interactive"] = self.config["core"].get("interactive", False)

dvc/repo/status.py

@@ -11,7 +11,7 @@ logger = logging.getLogger(__name__)
 
 
 def _joint_status(stages):
-    status = {}
+    status_info = {}
 
     for stage in stages:
         if stage.frozen and not stage.is_repo_import:
@@ -20,9 +20,9 @@ def _joint_status(stages):
                 " not going to be shown in the status output.".format(stage)
             )
 
-        status.update(stage.status(check_updates=True))
+        status_info.update(stage.status(check_updates=True))
 
-    return status
+    return status_info
 
 
 def _local_status(self, targets=None, with_deps=False, recursive=False):
@@ -74,7 +74,7 @@ def _cloud_status(
 
             { "bar": "deleted" }
     """
-    import dvc.remote.base as cloud
+    import dvc.cache.base as cloud
 
     used = self.used_cache(
         targets,
@@ -92,13 +92,13 @@ def _cloud_status(
     status_info = self.cloud.status(used, jobs, remote=remote)
     for info in status_info.values():
         name = info["name"]
-        status = info["status"]
-        if status in [cloud.STATUS_OK, cloud.STATUS_MISSING]:
+        status_ = info["status"]
+        if status_ in [cloud.STATUS_OK, cloud.STATUS_MISSING]:
             continue
 
         prefix_map = {cloud.STATUS_DELETED: "deleted", cloud.STATUS_NEW: "new"}
 
-        ret[name] = prefix_map[status]
+        ret[name] = prefix_map[status_]
 
     return ret
 

dvc/repo/tree.py

@@ -4,15 +4,14 @@ import os
 from dvc.dvcfile import is_valid_filename
 from dvc.exceptions import OutputNotFoundError
 from dvc.path_info import PathInfo
-from dvc.remote.base import RemoteActionNotImplemented
-from dvc.scm.tree import BaseTree
+from dvc.tree.base import BaseTree, RemoteActionNotImplemented
 from dvc.utils import file_md5
 from dvc.utils.fs import copy_fobj_to_file, makedirs
 
 logger = logging.getLogger(__name__)
 
 
-class DvcTree(BaseTree):
+class DvcTree(BaseTree):  # pylint:disable=abstract-method
     """DVC repo tree.
 
     Args:
@@ -27,7 +26,7 @@ class DvcTree(BaseTree):
     """
 
     def __init__(self, repo, fetch=False, stream=False):
-        self.repo = repo
+        super().__init__(repo, {"url": repo.root_dir})
         self.fetch = fetch
         self.stream = stream
 
@@ -44,18 +43,21 @@ class DvcTree(BaseTree):
         return outs
 
     def _get_granular_checksum(self, path, out, remote=None):
+        assert isinstance(path, PathInfo)
         if not self.fetch and not self.stream:
             raise FileNotFoundError
         dir_cache = out.get_dir_cache(remote=remote)
         for entry in dir_cache:
-            entry_relpath = entry[out.remote.tree.PARAM_RELPATH]
+            entry_relpath = entry[out.tree.PARAM_RELPATH]
             if os.name == "nt":
                 entry_relpath = entry_relpath.replace("/", os.sep)
             if path == out.path_info / entry_relpath:
-                return entry[out.remote.tree.PARAM_CHECKSUM]
+                return entry[out.tree.PARAM_CHECKSUM]
         raise FileNotFoundError
 
-    def open(self, path, mode="r", encoding="utf-8", remote=None):
+    def open(
+        self, path, mode="r", encoding="utf-8", remote=None
+    ):  # pylint: disable=arguments-differ
         try:
             outs = self._find_outs(path, strict=False)
         except OutputNotFoundError as exc:
@@ -82,8 +84,8 @@ class DvcTree(BaseTree):
                 else:
                     checksum = out.checksum
                 try:
-                    remote_info = remote_obj.hash_to_path_info(checksum)
-                    return remote_obj.open(
+                    remote_info = remote_obj.tree.hash_to_path_info(checksum)
+                    return remote_obj.tree.open(
                         remote_info, mode=mode, encoding=encoding
                     )
                 except RemoteActionNotImplemented:
@@ -93,19 +95,19 @@ class DvcTree(BaseTree):
 
         if out.is_dir_checksum:
             checksum = self._get_granular_checksum(path, out)
-            cache_path = out.cache.hash_to_path_info(checksum).url
+            cache_path = out.cache.tree.hash_to_path_info(checksum).url
         else:
             cache_path = out.cache_path
         return open(cache_path, mode=mode, encoding=encoding)
 
-    def exists(self, path):
+    def exists(self, path):  # pylint: disable=arguments-differ
         try:
             self._find_outs(path, strict=False, recursive=True)
             return True
         except OutputNotFoundError:
             return False
 
-    def isdir(self, path):
+    def isdir(self, path):  # pylint: disable=arguments-differ
         if not self.exists(path):
             return False
 
@@ -126,21 +128,47 @@ class DvcTree(BaseTree):
         # for dir checksum, we need to check if this is a file inside the
         # directory
         try:
-            self._get_granular_checksum(path, out)
+            self._get_granular_checksum(path_info, out)
             return False
         except FileNotFoundError:
             return True
 
-    def isfile(self, path):
+    def isfile(self, path):  # pylint: disable=arguments-differ
         if not self.exists(path):
             return False
 
         return not self.isdir(path)
 
-    def _walk(self, root, trie, topdown=True):
+    def _add_dir(self, top, trie, out, download_callback=None, **kwargs):
+        if not self.fetch and not self.stream:
+            return
+
+        # pull dir cache if needed
+        dir_cache = out.get_dir_cache(**kwargs)
+
+        # pull dir contents if needed
+        if self.fetch:
+            if out.changed_cache(filter_info=top):
+                used_cache = out.get_used_cache(filter_info=top)
+                downloaded = self.repo.cloud.pull(used_cache, **kwargs)
+                if download_callback:
+                    download_callback(downloaded)
+
+        for entry in dir_cache:
+            entry_relpath = entry[out.tree.PARAM_RELPATH]
+            if os.name == "nt":
+                entry_relpath = entry_relpath.replace("/", os.sep)
+            path_info = out.path_info / entry_relpath
+            trie[path_info.parts] = None
+
+    def _walk(self, root, trie, topdown=True, **kwargs):
         dirs = set()
         files = []
 
+        out = trie.get(root.parts)
+        if out and out.is_dir_checksum:
+            self._add_dir(root, trie, out, **kwargs)
+
         root_len = len(root.parts)
         for key, out in trie.iteritems(prefix=root.parts):  # noqa: B301
             if key == root.parts:
@@ -160,9 +188,7 @@ class DvcTree(BaseTree):
         for dname in dirs:
             yield from self._walk(root / dname, trie)
 
-    def walk(
-        self, top, topdown=True, onerror=None, download_callback=None, **kwargs
-    ):
+    def walk(self, top, topdown=True, onerror=None, **kwargs):
         from pygtrie import Trie
 
         assert topdown
@@ -185,26 +211,10 @@ class DvcTree(BaseTree):
         for out in outs:
             trie[out.path_info.parts] = out
 
-            if out.is_dir_checksum and (self.fetch or self.stream):
-                # pull dir cache if needed
-                dir_cache = out.get_dir_cache(**kwargs)
+            if out.is_dir_checksum and root.isin_or_eq(out.path_info):
+                self._add_dir(top, trie, out, **kwargs)
 
-                # pull dir contents if needed
-                if self.fetch:
-                    if out.changed_cache(filter_info=top):
-                        used_cache = out.get_used_cache(filter_info=top)
-                        downloaded = self.repo.cloud.pull(used_cache, **kwargs)
-                        if download_callback:
-                            download_callback(downloaded)
-
-                for entry in dir_cache:
-                    entry_relpath = entry[out.remote.tree.PARAM_RELPATH]
-                    if os.name == "nt":
-                        entry_relpath = entry_relpath.replace("/", os.sep)
-                    path_info = out.path_info / entry_relpath
-                    trie[path_info.parts] = None
-
-        yield from self._walk(root, trie, topdown=topdown)
+        yield from self._walk(root, trie, topdown=topdown, **kwargs)
 
     def isdvc(self, path, **kwargs):
         try:
@@ -213,7 +223,7 @@ class DvcTree(BaseTree):
             pass
         return False
 
-    def isexec(self, path):
+    def isexec(self, path):  # pylint: disable=unused-argument
         return False
 
     def get_file_hash(self, path_info):
@@ -226,7 +236,7 @@ class DvcTree(BaseTree):
         return out.checksum
 
 
-class RepoTree(BaseTree):
+class RepoTree(BaseTree):  # pylint:disable=abstract-method
     """DVC + git-tracked files tree.
 
     Args:
@@ -236,7 +246,7 @@ class RepoTree(BaseTree):
     """
 
     def __init__(self, repo, **kwargs):
-        self.repo = repo
+        super().__init__(repo, {"url": repo.root_dir})
         if hasattr(repo, "dvc_dir"):
             self.dvctree = DvcTree(repo, **kwargs)
         else:
@@ -255,7 +265,9 @@ class RepoTree(BaseTree):
             return self.dvctree.stream
         return False
 
-    def open(self, path, mode="r", encoding="utf-8", **kwargs):
+    def open(
+        self, path, mode="r", encoding="utf-8", **kwargs
+    ):  # pylint: disable=arguments-differ
         if "b" in mode:
             encoding = None
 
@@ -265,12 +277,14 @@ class RepoTree(BaseTree):
             )
         return self.repo.tree.open(path, mode=mode, encoding=encoding)
 
-    def exists(self, path):
+    def exists(
+        self, path, use_dvcignore=True
+    ):  # pylint: disable=arguments-differ
         return self.repo.tree.exists(path) or (
             self.dvctree and self.dvctree.exists(path)
         )
 
-    def isdir(self, path):
+    def isdir(self, path):  # pylint: disable=arguments-differ
         return self.repo.tree.isdir(path) or (
             self.dvctree and self.dvctree.isdir(path)
         )
@@ -278,7 +292,7 @@ class RepoTree(BaseTree):
     def isdvc(self, path, **kwargs):
         return self.dvctree is not None and self.dvctree.isdvc(path, **kwargs)
 
-    def isfile(self, path):
+    def isfile(self, path):  # pylint: disable=arguments-differ
         return self.repo.tree.isfile(path) or (
             self.dvctree and self.dvctree.isfile(path)
         )
@@ -337,7 +351,9 @@ class RepoTree(BaseTree):
             elif dirname in repo_set:
                 yield from self._walk_one(repo_walk)
 
-    def walk(self, top, topdown=True, onerror=None, dvcfiles=False, **kwargs):
+    def walk(
+        self, top, topdown=True, onerror=None, dvcfiles=False, **kwargs
+    ):  # pylint: disable=arguments-differ
         """Walk and merge both DVC and repo trees.
 
         Args:
@@ -380,7 +396,7 @@ class RepoTree(BaseTree):
         repo_walk = self.repo.tree.walk(top, topdown=topdown)
         yield from self._walk(dvc_walk, repo_walk, dvcfiles=dvcfiles)
 
-    def walk_files(self, top, **kwargs):
+    def walk_files(self, top, **kwargs):  # pylint: disable=arguments-differ
         for root, _, files in self.walk(top, **kwargs):
             for fname in files:
                 yield PathInfo(root) / fname
@@ -416,12 +432,13 @@ class RepoTree(BaseTree):
 
         for root, _, files in self.walk(top):
             root = PathInfo(root)
-            makedirs(dest, exist_ok=True)
+            dest_dir = dest / root.relative_to(top)
+            makedirs(dest_dir, exist_ok=True)
             for fname in files:
                 src = root / fname
                 with self.open(src, mode="rb") as fobj:
-                    copy_fobj_to_file(fobj, dest / fname)
+                    copy_fobj_to_file(fobj, dest_dir / fname)
 
     @property
-    def hash_jobs(self):
+    def hash_jobs(self):  # pylint: disable=invalid-overridden-method
         return self.repo.tree.hash_jobs

dvc/scm/git.py

@@ -3,6 +3,7 @@
 import logging
 import os
 import shlex
+from functools import partial
 
 import yaml
 from funcy import cached_property
@@ -17,7 +18,6 @@ from dvc.scm.base import (
     RevError,
     SCMError,
 )
-from dvc.scm.git.tree import GitTree
 from dvc.utils import fix_env, is_binary, relpath
 from dvc.utils.fs import path_isin
 
@@ -93,7 +93,7 @@ class Git(Base):
         return self.repo.working_tree_dir
 
     @staticmethod
-    def clone(url, to_path, rev=None):
+    def clone(url, to_path, rev=None, shallow_branch=None):
         import git
 
         ld_key = "LD_LIBRARY_PATH"
@@ -110,16 +110,25 @@ class Git(Base):
             env[ld_key] = ""
 
         try:
+            if shallow_branch is not None and os.path.exists(url):
+                # git disables --depth for local clones unless file:// url
+                # scheme is used
+                url = f"file://{url}"
             with TqdmGit(desc="Cloning", unit="obj") as pbar:
-                tmp_repo = git.Repo.clone_from(
+                clone_from = partial(
+                    git.Repo.clone_from,
                     url,
                     to_path,
                     env=env,  # needed before we can fix it in __init__
                     no_single_branch=True,
                     progress=pbar.update_git,
                 )
+                if shallow_branch is None:
+                    tmp_repo = clone_from()
+                else:
+                    tmp_repo = clone_from(branch=shallow_branch, depth=1)
             tmp_repo.close()
-        except git.exc.GitCommandError as exc:
+        except git.exc.GitCommandError as exc:  # pylint: disable=no-member
             raise CloneError(url, to_path) from exc
 
         # NOTE: using our wrapper to make sure that env is fixed in __init__
@@ -128,7 +137,7 @@ class Git(Base):
         if rev:
             try:
                 repo.checkout(rev)
-            except git.exc.GitCommandError as exc:
+            except git.exc.GitCommandError as exc:  # pylint: disable=no-member
                 raise RevError(
                     "failed to access revision '{}' for repo '{}'".format(
                         rev, url
@@ -251,8 +260,8 @@ class Git(Base):
         else:
             self.repo.git.checkout(branch)
 
-    def pull(self):
-        infos = self.repo.remote().pull()
+    def pull(self, **kwargs):
+        infos = self.repo.remote().pull(**kwargs)
         for info in infos:
             if info.flags & info.ERROR:
                 raise SCMError(f"pull failed: {info.note}")
@@ -375,8 +384,10 @@ class Git(Base):
         path_parts = os.path.normpath(path).split(os.path.sep)
         return basename == self.ignore_file or Git.GIT_DIR in path_parts
 
-    def get_tree(self, rev):
-        return GitTree(self.repo, self.resolve_rev(rev))
+    def get_tree(self, rev, **kwargs):
+        from dvc.tree.git import GitTree
+
+        return GitTree(self.repo, self.resolve_rev(rev), **kwargs)
 
     def get_rev(self):
         return self.repo.rev_parse("HEAD").hexsha

dvc/scm/tree.py

@@ -1,97 +0,0 @@
-import os
-import stat
-from multiprocessing import cpu_count
-
-from funcy import cached_property
-
-
-class BaseTree:
-    """Abstract class to represent access to files"""
-
-    @property
-    def tree_root(self):
-        pass
-
-    def open(self, path, mode="r", encoding="utf-8"):
-        """Open file and return a stream."""
-
-    def exists(self, path):
-        """Test whether a path exists."""
-
-    def isdir(self, path):
-        """Return true if the pathname refers to an existing directory."""
-
-    def isfile(self, path):
-        """Test whether a path is a regular file"""
-
-    def walk(self, top, topdown=True, onerror=None):
-        """Directory tree generator.
-
-        See `os.walk` for the docs. Differences:
-        - no support for symlinks
-        """
-
-    def walk_files(self, top):
-        for root, _, files in self.walk(top):
-            for file in files:
-                # NOTE: os.path.join is ~5.5 times slower
-                yield f"{root}{os.sep}{file}"
-
-
-class WorkingTree(BaseTree):
-    """Proxies the repo file access methods to working tree files"""
-
-    def __init__(self, repo_root=None):
-        repo_root = repo_root or os.getcwd()
-        self.repo_root = repo_root
-
-    @property
-    def tree_root(self):
-        return self.repo_root
-
-    def open(self, path, mode="r", encoding="utf-8"):
-        """Open file and return a stream."""
-        if "b" in mode:
-            encoding = None
-        return open(path, mode=mode, encoding=encoding)
-
-    def exists(self, path):
-        """Test whether a path exists."""
-        return os.path.lexists(path)
-
-    def isdir(self, path):
-        """Return true if the pathname refers to an existing directory."""
-        return os.path.isdir(path)
-
-    def isfile(self, path):
-        """Test whether a path is a regular file"""
-        return os.path.isfile(path)
-
-    def walk(self, top, topdown=True, onerror=None):
-        """Directory tree generator.
-
-        See `os.walk` for the docs. Differences:
-        - no support for symlinks
-        """
-        for root, dirs, files in os.walk(
-            top, topdown=topdown, onerror=onerror
-        ):
-            yield os.path.normpath(root), dirs, files
-
-    def isexec(self, path):
-        mode = os.stat(path).st_mode
-        return mode & (stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH)
-
-    @staticmethod
-    def stat(path):
-        return os.stat(path)
-
-    @cached_property
-    def hash_jobs(self):
-        return max(1, min(4, cpu_count() // 2))
-
-
-def is_working_tree(tree):
-    return isinstance(tree, WorkingTree) or isinstance(
-        getattr(tree, "tree", None), WorkingTree
-    )

dvc/stage/__init__.py

@@ -88,6 +88,9 @@ def create_stage(cls, repo, path, external=False, **kwargs):
 
 
 class Stage(params.StageParams):
+    # pylint:disable=no-value-for-parameter
+    # rwlocked() confuses pylint
+
     def __init__(
         self,
         repo,

dvc/stage/cache.py

@@ -6,7 +6,7 @@ import yaml
 from funcy import first
 from voluptuous import Invalid
 
-from dvc.remote.local import _log_exceptions
+from dvc.cache.local import _log_exceptions
 from dvc.schema import COMPILED_LOCK_FILE_STAGE_SCHEMA
 from dvc.utils import dict_sha256, relpath
 from dvc.utils.fs import makedirs

dvc/stage/decorators.py

@@ -15,7 +15,7 @@ def rwlocked(call, read=None, write=None):
     if write is None:
         write = []
 
-    stage = call._args[0]
+    stage = call._args[0]  # pylint: disable=protected-access
 
     assert stage.repo.lock.is_locked
 
@@ -40,7 +40,7 @@ def unlocked_repo(f):
     def wrapper(stage, *args, **kwargs):
         stage.repo.state.dump()
         stage.repo.lock.unlock()
-        stage.repo._reset()
+        stage.repo._reset()  # pylint: disable=protected-access
         try:
             ret = f(stage, *args, **kwargs)
         finally:

dvc/stage/loader.py

@@ -125,7 +125,7 @@ class SingleStageLoader(Mapping):
     def load_stage(cls, dvcfile, d, stage_text):
         path, wdir = resolve_paths(dvcfile.path, d.get(Stage.PARAM_WDIR))
         stage = loads_from(Stage, dvcfile.repo, path, wdir, d)
-        stage._stage_text = stage_text
+        stage._stage_text = stage_text  # noqa, pylint:disable=protected-access
         stage.deps = dependency.loadd_from(
             stage, d.get(Stage.PARAM_DEPS) or []
         )

dvc/stage/run.py

@@ -60,7 +60,10 @@ def cmd_run(stage, *args, **kwargs):
         warn_if_fish(executable)
         cmd = _nix_cmd(executable, stage.cmd)
 
-    main_thread = isinstance(threading.current_thread(), threading._MainThread)
+    main_thread = isinstance(
+        threading.current_thread(),
+        threading._MainThread,  # pylint: disable=protected-access
+    )
     old_handler = None
     p = None
 

dvc/stage/serialize.py

@@ -173,8 +173,9 @@ def to_single_stage_file(stage: "Stage"):
     # - reparse the same yaml text with a slow but smart ruamel yaml parser
     # - apply changes to a returned structure
     # - serialize it
-    if stage._stage_text is not None:
-        saved_state = parse_yaml_for_update(stage._stage_text, stage.path)
+    text = stage._stage_text  # noqa, pylint: disable=protected-access
+    if text is not None:
+        saved_state = parse_yaml_for_update(text, stage.path)
         # Stage doesn't work with meta in any way, so .dumpd() doesn't
         # have it. We simply copy it over.
         if "meta" in saved_state:

dvc/stage/utils.py

@@ -8,8 +8,8 @@ from dvc import dependency, output
 from dvc.utils.fs import path_isin
 
 from ..dependency import ParamsDependency
-from ..remote.local import LocalRemoteTree
-from ..remote.s3 import S3RemoteTree
+from ..tree.local import LocalTree
+from ..tree.s3 import S3Tree
 from ..utils import dict_md5, format_link, relpath
 from .exceptions import (
     MissingDataSource,
@@ -133,8 +133,8 @@ def stage_dump_eq(stage_cls, old_d, new_d):
     new_d.pop(stage_cls.PARAM_MD5, None)
     outs = old_d.get(stage_cls.PARAM_OUTS, [])
     for out in outs:
-        out.pop(LocalRemoteTree.PARAM_CHECKSUM, None)
-        out.pop(S3RemoteTree.PARAM_CHECKSUM, None)
+        out.pop(LocalTree.PARAM_CHECKSUM, None)
+        out.pop(S3Tree.PARAM_CHECKSUM, None)
 
     # outs and deps are lists of dicts. To check equality, we need to make
     # them independent of the order, so, we convert them to dicts.

dvc/state.py

@@ -35,7 +35,7 @@ class StateNoop:
     def save(self, path_info, checksum):
         pass
 
-    def get(self, path_info):
+    def get(self, path_info):  # pylint: disable=unused-argument
         return None
 
     def save_link(self, path_info):
@@ -89,10 +89,12 @@ class State:  # pylint: disable=too-many-instance-attributes
     MAX_UINT = 2 ** 64 - 2
 
     def __init__(self, local_cache):
+        from dvc.tree.local import LocalTree
+
         repo = local_cache.repo
         self.repo = repo
         self.root_dir = repo.root_dir
-        self.tree = local_cache.tree.work_tree
+        self.tree = LocalTree(None, {"url": self.root_dir})
 
         state_config = repo.config.get("state", {})
         self.row_limit = state_config.get("row_limit", self.STATE_ROW_LIMIT)
@@ -394,8 +396,8 @@ class State:  # pylint: disable=too-many-instance-attributes
         assert isinstance(path_info, str) or path_info.scheme == "local"
         path = os.fspath(path_info)
 
-        # NOTE: use os.path.exists instead of WorkingTree.exists
-        # WorkingTree.exists uses lexists() and will return True for broken
+        # NOTE: use os.path.exists instead of LocalTree.exists
+        # because it uses lexists() and will return True for broken
         # symlinks that we cannot stat() in get_mtime_and_size
         if not os.path.exists(path):
             return None
@@ -445,9 +447,9 @@ class State:  # pylint: disable=too-many-instance-attributes
 
         self._execute(f"SELECT * FROM {self.LINK_STATE_TABLE}")
         for row in self.cursor:
-            relpath, inode, mtime = row
+            relative_path, inode, mtime = row
             inode = self._from_sqlite(inode)
-            path = os.path.join(self.root_dir, relpath)
+            path = os.path.join(self.root_dir, relative_path)
 
             if path in used or not self.tree.exists(path):
                 continue
@@ -457,7 +459,7 @@ class State:  # pylint: disable=too-many-instance-attributes
 
             if (inode, mtime) == (actual_inode, actual_mtime):
                 logger.debug("Removing '%s' as unused link.", path)
-                unused.append(relpath)
+                unused.append(relative_path)
 
         return unused
 

dvc/system.py

@@ -77,7 +77,7 @@ class System:
         )
 
     @staticmethod
-    def _reflink_windows(src, dst):
+    def _reflink_windows(_src, _dst):
         return -1
 
     @staticmethod