deps: bump dvc-data to >=0.42.3

Makes for a much simpler resource management elsewhere.

.github/workflows/packages.yaml

@@ -17,9 +17,9 @@ jobs:
     strategy:
       matrix:
         include:
-         - {pkg: "deb",  os: "ubuntu-18.04", asset: "dvc_${{ github.event.release.tag_name }}_amd64.deb"}
-         - {pkg: "rpm", os: "ubuntu-18.04", asset: "dvc-${{ github.event.release.tag_name }}-1.x86_64.rpm"}
-         - {pkg: "osxpkg", os: "macos-10.15", asset: "dvc-${{ github.event.release.tag_name }}.pkg"}
+         - {pkg: "deb",  os: "ubuntu-20.04", asset: "dvc_${{ github.event.release.tag_name }}_amd64.deb"}
+         - {pkg: "rpm", os: "ubuntu-20.04", asset: "dvc-${{ github.event.release.tag_name }}-1.x86_64.rpm"}
+         - {pkg: "osxpkg", os: "macos-11", asset: "dvc-${{ github.event.release.tag_name }}.pkg"}
          - {pkg: "exe",  os: "windows-2019", asset: "dvc-${{ github.event.release.tag_name }}.exe"}
 
     name: ${{ matrix.pkg }}

.github/workflows/tests.yaml

@@ -17,7 +17,7 @@ permissions:
 
 jobs:
   tests:
-    timeout-minutes: 45
+    timeout-minutes: 50
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
@@ -41,9 +41,9 @@ jobs:
         pip install -e ".[dev]"
     - uses: pre-commit/[email protected]
     - name: run tests
-      timeout-minutes: 30
+      timeout-minutes: 40
       run: >
-        pytest -nauto --timeout=300 --durations=100
+        pytest -n=logical --timeout=300 --durations=100
         --cov --cov-report=xml --cov-report=term
     - name: upload coverage report
       uses: codecov/codecov-action@v3

.pre-commit-config.yaml

@@ -43,7 +43,7 @@ repos:
     hooks:
       - id: isort
   - repo: https://github.com/charliermarsh/ruff-pre-commit
-    rev: 'v0.0.238'
+    rev: 'v0.0.254'
     hooks:
       - id: ruff
   - repo: https://github.com/codespell-project/codespell

dvc/api/__init__.py

@@ -3,13 +3,14 @@ from dvc.fs.dvc import _DVCFileSystem as DVCFileSystem
 from .data import open  # pylint: disable=redefined-builtin
 from .data import get_url, read
 from .experiments import make_checkpoint
-from .params import params_show
+from .show import metrics_show, params_show
 
 __all__ = [
     "get_url",
     "make_checkpoint",
     "open",
     "params_show",
+    "metrics_show",
     "read",
     "DVCFileSystem",
 ]

dvc/api/show.py

@@ -4,10 +4,153 @@ from typing import Dict, Iterable, Optional, Union
 
 from funcy import first
 
-from dvc.exceptions import DvcException
 from dvc.repo import Repo
 
 
+def _onerror_raise(exception: Exception, *args, **kwargs):
+    raise exception
+
+
+def _postprocess(results):
+    processed: Dict[str, Dict] = {}
+    for rev, rev_data in results.items():
+        if not rev_data:
+            continue
+
+        processed[rev] = {}
+
+        counts: typing.Counter[str] = Counter()
+        for file_data in rev_data["data"].values():
+            for k in file_data["data"]:
+                counts[k] += 1
+
+        for file_name, file_data in rev_data["data"].items():
+            to_merge = {
+                (k if counts[k] == 1 else f"{file_name}:{k}"): v
+                for k, v in file_data["data"].items()
+            }
+            processed[rev] = {**processed[rev], **to_merge}
+
+    if "workspace" in processed:
+        del processed["workspace"]
+
+    return processed
+
+
+def metrics_show(
+    *targets: str,
+    repo: Optional[str] = None,
+    rev: Optional[str] = None,
+) -> Dict:
+    """Get metrics tracked in `repo`.
+
+    Without arguments, this function will retrieve all metrics from all tracked
+    metric files, for the current working tree.
+
+    See the options below to restrict the metrics retrieved.
+
+    Args:
+        *targets (str, optional): Names of the metric files to retrieve
+        metrics from. For example, "classifier_eval.json,
+        clustering_eval.json".
+        If no `targets` are provided, all metric files tracked in `dvc.yaml`
+        will be used.
+        Note that targets don't necessarily have to be defined in `dvc.yaml`.
+        repo (str, optional): Location of the DVC repository.
+            Defaults to the current project (found by walking up from the
+            current working directory tree).
+            It can be a URL or a file system path.
+            Both HTTP and SSH protocols are supported for online Git repos
+            (e.g. [user@]server:project.git).
+        rev (str, optional): Name of the `Git revision`_ to retrieve metrics
+            from.
+            Defaults to `None`.
+            An example of git revision can be a branch or tag name, a commit
+            hash or a dvc experiment name.
+            If `repo` is not a Git repo, this option is ignored.
+            If `None`, the current working tree will be used.
+
+    Returns:
+        Dict: See Examples below.
+
+    Examples:
+
+        - No arguments.
+
+        Working on https://github.com/iterative/example-get-started
+
+        >>> import dvc.api
+        >>> metrics = dvc.api.metrics_show()
+        >>> print(json.dumps(metrics, indent=4))
+        {
+            "avg_prec": 0.9249974999612706,
+            "roc_auc": 0.9460213440787918
+        }
+
+        ---
+
+        - Using `rev`.
+
+        Working on https://github.com/iterative/example-get-started
+
+        >>> import json
+        >>> import dvc.api
+        >>> metrics = dvc.api.metrics_show(rev="tune-hyperparams")
+        >>> print(json.dumps(metrics, indent=4))
+        {
+            "avg_prec": 0.9268792615819422,
+            "roc_auc": 0.945093365854111
+        }
+
+        ---
+
+        - Using `targets`.
+
+        Working on https://github.com/iterative/example-get-started
+
+        >>> import json
+        >>> import dvc.api
+        >>> metrics = dvc.api.metrics_show("evaluation.json")
+        >>> print(json.dumps(metrics, indent=4))
+        {
+            "avg_prec": 0.9249974999612706,
+            "roc_auc": 0.9460213440787918
+        }
+
+        ---
+
+        - Git URL as `repo`.
+
+        >>> import json
+        >>> import dvc.api
+        >>> metrics = dvc.api.metrics_show(
+        ...     repo="https://github.com/iterative/demo-fashion-mnist")
+        >>> print(json.dumps(metrics, indent=4))
+        {
+            "loss": 0.25284987688064575,
+            "accuracy": 0.9071000218391418
+        }
+
+
+    .. _Git revision:
+        https://git-scm.com/docs/revisions
+    """
+
+    with Repo.open(repo) as _repo:
+        metrics = _repo.metrics.show(
+            targets=targets,
+            revs=rev if rev is None else [rev],
+            onerror=_onerror_raise,
+        )
+
+    metrics = _postprocess(metrics)
+
+    if not metrics:
+        return {}
+
+    return metrics[first(metrics)]
+
+
 def params_show(
     *targets: str,
     repo: Optional[str] = None,
@@ -56,9 +199,6 @@ def params_show(
     Returns:
         Dict: See Examples below.
 
-    Raises:
-        DvcException: If no params are found in `repo`.
-
     Examples:
 
         - No arguments.
@@ -239,34 +379,6 @@ def params_show(
     if isinstance(stages, str):
         stages = [stages]
 
-    def _onerror_raise(_result: Dict, exception: Exception, *args, **kwargs):
-        raise exception
-
-    def _postprocess(params):
-        processed: Dict[str, Dict] = {}
-        for rev, rev_data in params.items():
-            processed[rev] = {}
-
-            counts: typing.Counter[str] = Counter()
-            for file_data in rev_data["data"].values():
-                for k in file_data["data"]:
-                    counts[k] += 1
-
-            for file_name, file_data in rev_data["data"].items():
-                to_merge = {
-                    (k if counts[k] == 1 else f"{file_name}:{k}"): v
-                    for k, v in file_data["data"].items()
-                }
-                processed[rev] = {**processed[rev], **to_merge}
-
-        if "workspace" in processed:
-            del processed["workspace"]
-
-        if not processed:
-            raise DvcException("No params found")
-
-        return processed[first(processed)]
-
     with Repo.open(repo) as _repo:
         params = _repo.params.show(
             revs=rev if rev is None else [rev],
@@ -276,4 +388,9 @@ def params_show(
             stages=stages,
         )
 
-    return _postprocess(params)
+    params = _postprocess(params)
+
+    if not params:
+        return {}
+
+    return params[first(params)]

dvc/cachemgr.py

@@ -1,3 +1,5 @@
+import os
+
 from dvc.fs import GitFileSystem, Schemes
 from dvc_data.hashfile.db import get_odb
 
@@ -9,7 +11,6 @@ def _get_odb(repo, settings, fs=None):
         return None
 
     cls, config, fs_path = get_cloud_fs(repo, **settings)
-    config["tmp_dir"] = repo.tmp_dir
     fs = fs or cls(**config)
     return get_odb(fs, fs_path, state=repo.state, **config)
 
@@ -29,16 +30,21 @@ class CacheManager:
         self.config = config = repo.config["cache"]
         self._odb = {}
 
+        default = None
+        if repo and repo.local_dvc_dir:
+            default = os.path.join(repo.local_dvc_dir, self.CACHE_DIR)
+
         local = config.get("local")
 
         if local:
             settings = {"name": local}
-        elif "dir" not in config:
+        elif "dir" not in config and not default:
             settings = None
         else:
             from dvc.config_schema import LOCAL_COMMON
 
-            settings = {"url": config["dir"]}
+            url = config.get("dir") or default
+            settings = {"url": url}
             for opt in LOCAL_COMMON:
                 if opt in config:
                     settings[str(opt)] = config.get(opt)

dvc/cli/__init__.py

@@ -39,12 +39,17 @@ def parse_args(argv=None):
 
 def _log_unknown_exceptions() -> None:
     from dvc.info import get_dvc_info
-    from dvc.logger import FOOTER
+    from dvc.ui import ui
+    from dvc.utils import colorize
 
     logger.exception("unexpected error")
     if logger.isEnabledFor(logging.DEBUG):
         logger.debug("Version info for developers:\n%s", get_dvc_info())
-    logger.info(FOOTER)
+
+    q = colorize("Having any troubles?", "yellow")
+    link = colorize("https://dvc.org/support", "blue")
+    footer = f"\n{q} Hit us up at {link}, we are always happy to help!"
+    ui.error_write(footer)
 
 
 def _log_exceptions(exc: Exception) -> Optional[int]:
@@ -141,7 +146,7 @@ def _log_exceptions(exc: Exception) -> Optional[int]:
     return None
 
 
-def main(argv=None):  # noqa: C901
+def main(argv=None):  # noqa: C901, PLR0912, PLR0915
     """Main entry point for dvc CLI.
 
     Args:
@@ -167,7 +172,7 @@ def main(argv=None):  # noqa: C901
 
     args = None
 
-    outerLogLevel = logger.level
+    outer_log_level = logger.level
     try:
         args = parse_args(argv)
 
@@ -238,7 +243,7 @@ def main(argv=None):  # noqa: C901
 
         return ret
     finally:
-        logger.setLevel(outerLogLevel)
+        logger.setLevel(outer_log_level)
 
         from dvc.external_repo import clean_repos
 

dvc/cli/parser.py

@@ -2,6 +2,7 @@
 import argparse
 import logging
 import os
+from functools import lru_cache
 
 from dvc import __version__
 from dvc.commands import (
@@ -150,6 +151,7 @@ def get_parent_parser():
     return parent_parser
 
 
+@lru_cache(maxsize=1)
 def get_main_parser():
     parent_parser = get_parent_parser()
 

dvc/commands/cache.py

@@ -1,4 +1,5 @@
 import argparse
+import os
 
 from dvc.cli import completion
 from dvc.cli.utils import append_doc_link, fix_subparsers
@@ -9,13 +10,21 @@ from dvc.ui import ui
 class CmdCacheDir(CmdConfig):
     def run(self):
         if self.args.value is None and not self.args.unset:
+            from dvc.config import ConfigError
+
             if self.args.level:
                 conf = self.config.read(level=self.args.level)
             else:
                 # Use merged config with default values
                 conf = self.config
-            self._check(conf, False, "cache", "dir")
-            ui.write(conf["cache"]["dir"])
+            try:
+                self._check(conf, False, "cache", "dir")
+                path = conf["cache"]["dir"]
+            except ConfigError:
+                if not self.config.dvc_dir or self.args.level:
+                    raise
+                path = os.path.join(self.config.dvc_dir, "cache")
+            ui.write(path)
             return 0
         with self.config.edit(level=self.args.level) as conf:
             if self.args.unset:

dvc/commands/dag.py

@@ -9,11 +9,11 @@ if TYPE_CHECKING:
     from networkx import DiGraph
 
 
-def _show_ascii(G: "DiGraph"):
+def _show_ascii(graph: "DiGraph"):
     from dvc.dagascii import draw
     from dvc.repo.graph import get_pipelines
 
-    pipelines = get_pipelines(G)
+    pipelines = get_pipelines(graph)
 
     ret = []
     for pipeline in pipelines:
@@ -31,7 +31,7 @@ def _quote_label(node):
     return label
 
 
-def _show_dot(G: "DiGraph"):
+def _show_dot(graph: "DiGraph"):
     import io
 
     import networkx as nx
@@ -39,15 +39,15 @@ def _show_dot(G: "DiGraph"):
 
     dot_file = io.StringIO()
 
-    nx.relabel_nodes(G, _quote_label, copy=False)
-    write_dot(G.reverse(), dot_file)
+    nx.relabel_nodes(graph, _quote_label, copy=False)
+    write_dot(graph.reverse(), dot_file)
     return dot_file.getvalue()
 
 
-def _show_mermaid(G, markdown: bool = False):
+def _show_mermaid(graph, markdown: bool = False):
     from dvc.repo.graph import get_pipelines
 
-    pipelines = get_pipelines(G)
+    pipelines = get_pipelines(graph)
 
     graph = "flowchart TD"
 
@@ -100,43 +100,42 @@ def _transform(index, outs):
     def _relabel(node) -> str:
         return node.addressing if isinstance(node, Stage) else str(node)
 
-    G = index.outs_graph if outs else index.graph
-    return nx.relabel_nodes(G, _relabel, copy=True)
+    graph = index.outs_graph if outs else index.graph
+    return nx.relabel_nodes(graph, _relabel, copy=True)
 
 
-def _filter(G, targets, full):
+def _filter(graph, targets, full):
     import networkx as nx
 
     if not targets:
-        return G
+        return graph
 
-    H = G.copy()
+    new_graph = graph.copy()
     if not full:
         descendants = set()
         for target in targets:
-            descendants.update(nx.descendants(G, target))
+            descendants.update(nx.descendants(graph, target))
             descendants.add(target)
-        H.remove_nodes_from(set(G.nodes()) - descendants)
+        new_graph.remove_nodes_from(set(graph.nodes()) - descendants)
 
-    undirected = H.to_undirected()
+    undirected = new_graph.to_undirected()
     connected = set()
     for target in targets:
         connected.update(nx.node_connected_component(undirected, target))
 
-    H.remove_nodes_from(set(H.nodes()) - connected)
-
-    return H
+    new_graph.remove_nodes_from(set(new_graph.nodes()) - connected)
+    return new_graph
 
 
 def _build(repo, target=None, full=False, outs=False):
     targets = _collect_targets(repo, target, outs)
-    G = _transform(repo.index, outs)
-    return _filter(G, targets, full)
+    graph = _transform(repo.index, outs)
+    return _filter(graph, targets, full)
 
 
 class CmdDAG(CmdBase):
     def run(self):
-        G = _build(
+        graph = _build(
             self.repo,
             target=self.args.target,
             full=self.args.full,
@@ -144,12 +143,12 @@ class CmdDAG(CmdBase):
         )
 
         if self.args.dot:
-            ui.write(_show_dot(G))
+            ui.write(_show_dot(graph))
         elif self.args.mermaid or self.args.markdown:
-            ui.write(_show_mermaid(G, self.args.markdown))
+            ui.write(_show_mermaid(graph, self.args.markdown))
         else:
             with ui.pager():
-                ui.write(_show_ascii(G))
+                ui.write(_show_ascii(graph))
 
         return 0
 

dvc/commands/data.py

@@ -21,12 +21,14 @@ logger = logging.getLogger(__name__)
 
 class CmdDataStatus(CmdBase):
     COLORS = {
+        "not_in_remote": "red",
         "not_in_cache": "red",
         "committed": "green",
         "uncommitted": "yellow",
         "untracked": "cyan",
     }
     LABELS = {
+        "not_in_remote": "Not in remote",
         "not_in_cache": "Not in cache",
         "committed": "DVC committed changes",
         "uncommitted": "DVC uncommitted changes",
@@ -34,6 +36,7 @@ class CmdDataStatus(CmdBase):
         "unchanged": "DVC unchanged files",
     }
     HINTS = {
+        "not_in_remote": ('use "dvc push <file>..." to upload files',),
         "not_in_cache": ('use "dvc fetch <file>..." to download files',),
         "committed": ("git commit the corresponding dvc files to update the repo",),
         "uncommitted": (
@@ -114,6 +117,7 @@ class CmdDataStatus(CmdBase):
             status = self.repo.data_status(
                 granular=self.args.granular,
                 untracked_files=self.args.untracked_files,
+                remote_refresh=self.args.remote_refresh,
             )
 
         if not self.args.unchanged:
@@ -239,6 +243,12 @@ def add_parser(subparsers, parent_parser):
         nargs="?",
         help="Show untracked files.",
     )
+    data_status_parser.add_argument(
+        "--remote-refresh",
+        action="store_true",
+        default=False,
+        help="Refresh remote index.",
+    )
     data_status_parser.set_defaults(func=CmdDataStatus)
 
     DATA_LS_HELP = "List data tracked by DVC with its metadata."

dvc/commands/data_sync.py

@@ -40,6 +40,7 @@ class CmdDataPull(CmdDataBase):
                 recursive=self.args.recursive,
                 run_cache=self.args.run_cache,
                 glob=self.args.glob,
+                allow_missing=self.args.allow_missing,
             )
             self.log_summary(stats)
         except (CheckoutError, DvcException) as exc:
@@ -197,6 +198,12 @@ def add_parser(subparsers, _parent_parser):
         default=False,
         help=argparse.SUPPRESS,
     )
+    pull_parser.add_argument(
+        "--allow-missing",
+        action="store_true",
+        default=False,
+        help="Ignore errors if some of the files or directories are missing.",
+    )
     pull_parser.set_defaults(func=CmdDataPull)
 
     # Push

dvc/commands/experiments/__init__.py

@@ -4,6 +4,7 @@ from dvc.cli.utils import append_doc_link, fix_subparsers, hide_subparsers_from_
 from dvc.commands.experiments import (
     apply,
     branch,
+    clean,
     diff,
     exec_run,
     gc,
@@ -21,6 +22,7 @@ from dvc.commands.experiments import (
 SUB_COMMANDS = [
     apply,
     branch,
+    clean,
     diff,
     exec_run,
     gc,

dvc/commands/experiments/apply.py

@@ -4,13 +4,21 @@ import logging
 from dvc.cli import completion
 from dvc.cli.command import CmdBase
 from dvc.cli.utils import append_doc_link
+from dvc.ui import ui
 
 logger = logging.getLogger(__name__)
 
 
 class CmdExperimentsApply(CmdBase):
     def run(self):
-        self.repo.experiments.apply(self.args.experiment, force=self.args.force)
+        if not self.args.force:
+            ui.write(
+                "The --no-force option is deprecated and will be removed in a future"
+                " DVC release. To revert the result of 'exp apply', run:\n"
+                "\n\tgit reset --hard\n"
+                "\tgit stash apply refs/exps/apply/stash\n"
+            )
+        self.repo.experiments.apply(self.args.experiment)
 
         return 0
 

dvc/commands/experiments/clean.py

@@ -0,0 +1,25 @@
+import argparse
+import logging
+
+from dvc.cli.command import CmdBase
+from dvc.cli.utils import append_doc_link
+
+logger = logging.getLogger(__name__)
+
+
+class CmdExperimentsClean(CmdBase):
+    def run(self):
+        self.repo.experiments.clean()
+        return 0
+
+
+def add_parser(experiments_subparsers, parent_parser):
+    EXPERIMENTS_CLEAN_HELP = "Cleanup experiments temporary internal files."
+    experiments_clean_parser = experiments_subparsers.add_parser(
+        "clean",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_CLEAN_HELP, "exp/clean"),
+        help=EXPERIMENTS_CLEAN_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    experiments_clean_parser.set_defaults(func=CmdExperimentsClean)

dvc/commands/experiments/diff.py

@@ -52,7 +52,7 @@ class CmdExperimentsDiff(CmdBase):
 
 
 def add_parser(experiments_subparsers, parent_parser):
-    EXPERIMENTS_DIFF_HELP = "Show changes between experiments in the DVC repository."
+    EXPERIMENTS_DIFF_HELP = "Show changes between experiments."
 
     experiments_diff_parser = experiments_subparsers.add_parser(
         "diff",

dvc/commands/experiments/gc.py

@@ -78,7 +78,8 @@ class CmdExperimentsGC(CmdBase):
 def add_parser(experiments_subparsers, parent_parser):
     EXPERIMENTS_GC_HELP = "Garbage collect unneeded experiments."
     EXPERIMENTS_GC_DESCRIPTION = (
-        "Removes all experiments which are not derived from the specifiedGit revisions."
+        "Removes all experiments which are not derived"
+        " from the specified Git revisions."
     )
     experiments_gc_parser = experiments_subparsers.add_parser(
         "gc",

dvc/commands/experiments/queue_worker.py

@@ -9,7 +9,9 @@ class CmdQueueWorker(CmdBase):
     """Run the exp queue worker."""
 
     def run(self):
-        self.repo.experiments.celery_queue.worker.start(self.args.name)
+        self.repo.experiments.celery_queue.worker.start(
+            self.args.name, fsapp_clean=self.args.clean
+        )
         return 0
 
 
@@ -22,4 +24,9 @@ def add_parser(experiments_subparsers, parent_parser):
         add_help=False,
     )
     parser.add_argument("name", help="Celery worker name.")
+    parser.add_argument(
+        "--clean",
+        action="store_true",
+        help="Automatically cleanup celery broker on shutdown.",
+    )
     parser.set_defaults(func=CmdQueueWorker)

dvc/commands/experiments/save.py

@@ -35,7 +35,7 @@ class CmdExperimentsSave(CmdBase):
 
 
 def add_parser(experiments_subparsers, parent_parser):
-    EXPERIMENTS_SAVE_HELP = "Save current workspace as a dvc experiment."
+    EXPERIMENTS_SAVE_HELP = "Save current workspace as an experiment."
     save_parser = experiments_subparsers.add_parser(
         "save",
         parents=[parent_parser],

dvc/commands/experiments/show.py

@@ -62,7 +62,7 @@ experiment_types = {
 }
 
 
-def _collect_rows(  # noqa: C901
+def _collect_rows(  # noqa: C901, PLR0912, PLR0913, PLR0915
     base_rev,
     experiments,
     all_headers,
@@ -244,7 +244,7 @@ def _extend_row(row, names, headers, items, precision, fill_value=FILL_VALUE):
                 row[f"{fname}:{name}"] = value
 
 
-def experiments_table(
+def experiments_table(  # noqa: PLR0913
     all_experiments,
     headers,
     metric_headers,
@@ -308,7 +308,7 @@ def baseline_styler(typ):
     return {"style": "bold"} if typ == "baseline" else {}
 
 
-def show_experiments(  # noqa: C901
+def show_experiments(  # noqa: C901, PLR0912
     all_experiments,
     keep=None,
     drop=None,
@@ -453,8 +453,12 @@ def _normalize_headers(names, count):
 
 
 def _format_json(item):
+    from dvc.repo.experiments.show import _CachedError
+
     if isinstance(item, (date, datetime)):
         return item.isoformat()
+    if isinstance(item, _CachedError):
+        return {"type": getattr(item, "typ", "_CachedError"), "msg": str(item)}
     return encode_exception(item)
 
 
@@ -472,6 +476,7 @@ class CmdExperimentsShow(CmdBase):
                 sha_only=self.args.sha,
                 param_deps=self.args.param_deps,
                 fetch_running=self.args.fetch_running,
+                force=self.args.force,
             )
         except DvcException:
             logger.exception("failed to show experiments")
@@ -650,4 +655,10 @@ def add_parser(experiments_subparsers, parent_parser):
         action="store_false",
         help=argparse.SUPPRESS,
     )
+    experiments_show_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        help="Force re-collection of experiments instead of loading from exp cache.",
+    )
     experiments_show_parser.set_defaults(func=CmdExperimentsShow)

dvc/commands/machine.py

@@ -302,7 +302,7 @@ class CmdMachineSsh(CmdBase):
         return 0
 
 
-def add_parser(subparsers, parent_parser):
+def add_parser(subparsers, parent_parser):  # noqa: PLR0915
     from dvc.commands.config import parent_config_parser
 
     machine_HELP = "Set up and manage cloud machines."

dvc/commands/plots.py

@@ -19,7 +19,7 @@ def _show_json(renderers, split=False):
     from dvc.render.convert import to_json
 
     result = {renderer.name: to_json(renderer, split) for renderer in renderers}
-    ui.write_json(result)
+    ui.write_json(result, highlight=False)
 
 
 def _adjust_vega_renderers(renderers):
@@ -76,10 +76,6 @@ class CmdPlots(CmdBase):
         props = {p: getattr(self.args, p) for p in PLOT_PROPS}
         return {k: v for k, v in props.items() if v is not None}
 
-    def _config_files(self):
-        if self.args.from_config:
-            return {self.args.from_config}
-
     def _html_template_path(self):
         html_template_path = self.args.html_template
         if not html_template_path:
@@ -91,7 +87,7 @@ class CmdPlots(CmdBase):
                 html_template_path = os.path.join(self.repo.dvc_dir, html_template_path)
         return html_template_path  # noqa: RET504
 
-    def run(self) -> int:  # noqa: C901
+    def run(self) -> int:  # noqa: C901, PLR0911, PLR0912
         from pathlib import Path
 
         from dvc.render.match import match_defs_renderers
@@ -114,10 +110,9 @@ class CmdPlots(CmdBase):
             plots_data = self._func(
                 targets=self.args.targets,
                 props=self._props(),
-                config_files=self._config_files(),
             )
 
-            if not plots_data:
+            if not plots_data and not self.args.json:
                 ui.error_write(
                     "No plots were loaded, visualization file will not be created."
                 )
@@ -150,7 +145,7 @@ class CmdPlots(CmdBase):
                 render_html(
                     renderers=renderers,
                     output_file=output_file,
-                    template_path=self._html_template_path(),
+                    html_template=self._html_template_path(),
                 )
 
                 ui.write(output_file.as_uri())
@@ -412,9 +407,3 @@ def _add_ui_arguments(parser):
         help="Custom HTML template for VEGA visualization.",
         metavar="<path>",
     )
-    parser.add_argument(
-        "--from-config",
-        default=None,
-        metavar="<path>",
-        help=argparse.SUPPRESS,
-    )

dvc/commands/update.py

@@ -31,8 +31,7 @@ class CmdUpdate(CmdBase):
 def add_parser(subparsers, parent_parser):
     UPDATE_HELP = (
         "Update data artifact imported (via dvc import or dvc import-url) "
-        "from an external DVC repository or URL, or update data from a "
-        "DVC worktree remote."
+        "from an external DVC repository or URL."
     )
     update_parser = subparsers.add_parser(
         "update",

dvc/compare.py

@@ -209,7 +209,7 @@ class TabularData(MutableSequence[Sequence["CellT"]]):
         keys = self.keys() if cols is None else set(cols)
         return [{k: self._columns[k][i] for k in keys} for i in range(len(self))]
 
-    def dropna(  # noqa: C901
+    def dropna(  # noqa: C901, PLR0912
         self,
         axis: str = "rows",
         how="any",
@@ -371,7 +371,7 @@ def diff_table(
     return td
 
 
-def show_diff(
+def show_diff(  # noqa: PLR0913
     diff,
     title: str,
     old: bool = True,

dvc/config.py

@@ -117,7 +117,7 @@ class Config(dict):
 
     @classmethod
     def get_dir(cls, level):
-        from appdirs import site_config_dir, user_config_dir
+        from platformdirs import site_config_dir, user_config_dir
 
         assert level in ("global", "system")
 
@@ -170,10 +170,6 @@ class Config(dict):
         self.clear()
         self.update(conf)
 
-        # Add resolved default cache.dir
-        if not self["cache"].get("dir") and self.dvc_dir:
-            self["cache"]["dir"] = os.path.join(self.dvc_dir, "cache")
-
     def _get_fs(self, level):
         # NOTE: this might be a Gitfs, which doesn't see things outside of
         # the repo.

dvc/config_schema.py

@@ -1,8 +1,21 @@
+import logging
 import os
 from urllib.parse import urlparse
 
-from funcy import walk_values
-from voluptuous import All, Any, Coerce, Invalid, Lower, Optional, Range, Schema
+from funcy import once, walk_values
+from voluptuous import (
+    REMOVE_EXTRA,
+    All,
+    Any,
+    Coerce,
+    Invalid,
+    Lower,
+    Optional,
+    Range,
+    Schema,
+)
+
+logger = logging.getLogger(__name__)
 
 Bool = All(
     Lower,
@@ -30,7 +43,7 @@ def supported_cache_type(types):
     return types
 
 
-def Choices(*choices):
+def Choices(*choices):  # noqa: N802
     """Checks that value belongs to the specified set of values
 
     Args:
@@ -40,7 +53,7 @@ def Choices(*choices):
     return Any(*choices, msg="expected one of {}".format(", ".join(choices)))
 
 
-def ByUrl(mapping):
+def ByUrl(mapping):  # noqa: N802
     schemas = walk_values(Schema, mapping)
 
     def validate(data):
@@ -51,6 +64,8 @@ def ByUrl(mapping):
         # Windows absolute paths should really have scheme == "" (local)
         if os.name == "nt" and len(parsed.scheme) == 1 and parsed.netloc == "":
             return schemas[""](data)
+        if parsed.netloc == "":
+            return schemas[""](data)
         if parsed.scheme not in schemas:
             raise Invalid(f"Unsupported URL type {parsed.scheme}://")
 
@@ -63,6 +78,27 @@ class RelPath(str):
     pass
 
 
+class FeatureSchema(Schema):
+    def __init__(self, schema, required=False):
+        super().__init__(schema, required=required, extra=REMOVE_EXTRA)
+
+    @staticmethod
+    @once
+    def _log_deprecated(keys):
+        # only run this once per session
+        message = "%s config option%s unsupported"
+        paths = ", ".join(f"'feature.{key}'" for key in keys)
+        pluralize = " is" if len(keys) == 1 else "s are"
+        logger.warning(message, paths, pluralize)
+
+    def __call__(self, data):
+        ret = super().__call__(data)
+        extra_keys = data.keys() - ret.keys()
+        if extra_keys:
+            self._log_deprecated(sorted(extra_keys))
+        return ret
+
+
 REMOTE_COMMON = {
     "url": str,
     "checksum_jobs": All(Coerce(int), Range(1)),
@@ -260,11 +296,12 @@ SCHEMA = {
         },
     },
     # section for experimental features
-    "feature": {
-        Optional("machine", default=False): Bool,
-        # enabled by default. It's of no use, kept for backward compatibility.
-        Optional("parametrization", default=True): Bool,
-    },
+    # only specified keys are validated, others get logged and then ignored/removed
+    "feature": FeatureSchema(
+        {
+            Optional("machine", default=False): Bool,
+        },
+    ),
     "plots": {
         "html_template": str,
         Optional("auto_open", default=False): Bool,

dvc/daemon.py

@@ -14,7 +14,14 @@ from dvc.utils import fix_env, is_binary
 logger = logging.getLogger(__name__)
 
 
-def _popen(cmd, **kwargs):
+def _suppress_resource_warning(popen: Popen):
+    """Sets the returncode to avoid ResourceWarning when popen is garbage collected."""
+    # only use for daemon processes.
+    # See https://bugs.python.org/issue38890.
+    popen.returncode = 0
+
+
+def _popen(cmd, **kwargs) -> Popen:
     prefix = [sys.executable]
     if not is_binary():
         main_entrypoint = os.path.join(
@@ -40,7 +47,10 @@ def _spawn_windows(cmd, env):
         startupinfo = STARTUPINFO()
         startupinfo.dwFlags |= STARTF_USESHOWWINDOW
 
-        _popen(cmd, env=env, creationflags=creationflags, startupinfo=startupinfo)
+        popen = _popen(
+            cmd, env=env, creationflags=creationflags, startupinfo=startupinfo
+        )
+        _suppress_resource_warning(popen)
 
 
 def _spawn_posix(cmd, env):

dvc/dagascii.py

@@ -80,7 +80,7 @@ class AsciiCanvas:
 
         self.canvas[y][x] = char
 
-    def line(self, x0, y0, x1, y1, char):  # noqa: C901
+    def line(self, x0, y0, x1, y1, char):  # noqa: C901, PLR0912
         """Create a line on ASCII canvas.
 
         Args:
@@ -213,8 +213,8 @@ def draw(vertices, edges):
     # pylint: disable=too-many-locals
     # NOTE: coordinates might me negative, so we need to shift
     # everything to the positive plane before we actually draw it.
-    Xs = []  # pylint: disable=invalid-name
-    Ys = []  # pylint: disable=invalid-name
+    Xs = []  # noqa: N806, pylint: disable=invalid-name
+    Ys = []  # noqa: N806, pylint: disable=invalid-name
 
     sug = _build_sugiyama_layout(vertices, edges)
 

dvc/data_cloud.py

@@ -18,10 +18,11 @@ logger = logging.getLogger(__name__)
 
 
 class Remote:
-    def __init__(self, name: str, path: str, fs: "FileSystem", **config):
+    def __init__(self, name: str, path: str, fs: "FileSystem", *, index=None, **config):
         self.path = path
         self.fs = fs
         self.name = name
+        self.index = index
 
         self.worktree: bool = config.pop("worktree", False)
         self.config = config
@@ -34,7 +35,7 @@ class Remote:
         if self.worktree:
             path = self.fs.path.join(path, ".dvc", "cache")
 
-        return get_odb(self.fs, path, **self.config)
+        return get_odb(self.fs, path, hash_name="md5", **self.config)
 
 
 class DataCloud:
@@ -75,7 +76,11 @@ class DataCloud:
 
             fs = cls(**config)
             config["tmp_dir"] = self.repo.index_db_dir
-            return Remote(name, fs_path, fs, **config)
+            if self.repo.data_index is not None:
+                index = self.repo.data_index.view(("remote", name))
+            else:
+                index = None
+            return Remote(name, fs_path, fs, index=index, **config)
 
         if bool(self.repo.config["remote"]):
             error_msg = (

dvc/dependency/param.py

@@ -2,9 +2,9 @@ import logging
 import os
 import typing
 from collections import defaultdict
-from typing import Dict, cast
+from typing import Dict
 
-import dpath.util
+import dpath
 from voluptuous import Any
 
 from dvc.exceptions import DvcException
@@ -87,17 +87,17 @@ class ParamsDependency(Dependency):
         if flatten:
             for param in self.params:
                 try:
-                    ret[param] = dpath.util.get(config, param, separator=".")
+                    ret[param] = dpath.get(config, param, separator=".")
                 except KeyError:
                     continue
             return ret
 
-        from dpath.util import merge
+        from dpath import merge
 
         for param in self.params:
             merge(
                 ret,
-                dpath.util.search(config, param, separator="."),
+                dpath.search(config, param, separator="."),
                 separator=".",
             )
         return ret
@@ -111,7 +111,8 @@ class ParamsDependency(Dependency):
         from funcy import ldistinct
 
         status: Dict[str, Any] = defaultdict(dict)
-        info = cast(dict, self.hash_info.value) if self.hash_info else {}
+        assert isinstance(self.hash_info.value, dict)
+        info = self.hash_info.value if self.hash_info else {}
         actual = self.read_params()
 
         # NOTE: we want to preserve the order of params as specified in the

dvc/dependency/repo.py

@@ -133,7 +133,7 @@ class RepoDependency(Dependency):
                         self._check_circular_import(odb, obj_ids)
                         used_obj_ids[odb].update(obj_ids)
                 except (NoRemoteError, NoOutputOrStageError):
-                    pass  # noqa: S110
+                    pass
 
             try:
                 object_store, meta, obj = build(

dvc/dvcfile.py

@@ -187,7 +187,7 @@ class FileMixin:
 
 class SingleStageFile(FileMixin):
     from dvc.schema import COMPILED_SINGLE_STAGE_SCHEMA as SCHEMA
-    from dvc.stage.loader import SingleStageLoader as LOADER
+    from dvc.stage.loader import SingleStageLoader as LOADER  # noqa: N814
 
     metrics: List[str] = []
     plots: Any = {}
@@ -231,7 +231,7 @@ class ProjectFile(FileMixin):
     """Abstraction for pipelines file, .yaml + .lock combined."""
 
     from dvc.schema import COMPILED_MULTI_STAGE_SCHEMA as SCHEMA
-    from dvc.stage.loader import StageLoader as LOADER
+    from dvc.stage.loader import StageLoader as LOADER  # noqa: N814
 
     @property
     def _lockfile(self):
@@ -409,11 +409,10 @@ class Lockfile(FileMixin):
             if version == LOCKFILE_VERSION.V1:
                 logger.info("Migrating lock file '%s' from v1 to v2", self.relpath)
                 migrate_lock_v1_to_v2(data, self.latest_version_info)
-            else:
-                if not data:
-                    data.update(self.latest_version_info)
-                    # order is important, meta should always be at the top
-                    logger.info("Generating lock file '%s'", self.relpath)
+            elif not data:
+                data.update(self.latest_version_info)
+                # order is important, meta should always be at the top
+                logger.info("Generating lock file '%s'", self.relpath)
 
             data["stages"] = data.get("stages", {})
             modified = data["stages"].get(stage.name, {}) != stage_data.get(

dvc/external_repo.py

@@ -189,7 +189,7 @@ def _cached_clone(url, rev, for_write=False):
 
 
 @wrap_with(threading.Lock())
-def _clone_default_branch(url, rev, for_write=False):  # noqa: C901
+def _clone_default_branch(url, rev, for_write=False):  # noqa: C901, PLR0912
     """Get or create a clean clone of the url.
 
     The cloned is reactualized with git pull unless rev is a known sha.

dvc/fs/__init__.py

@@ -6,7 +6,6 @@ from dvc.config import ConfigError as RepoConfigError
 from dvc.config_schema import SCHEMA, Invalid
 
 # pylint: disable=unused-import
-from dvc_objects.fs import utils  # noqa: F401
 from dvc_objects.fs import (  # noqa: F401
     LocalFileSystem,
     MemoryFileSystem,
@@ -17,6 +16,7 @@ from dvc_objects.fs import (  # noqa: F401
     localfs,
     registry,
     system,
+    utils,
 )
 from dvc_objects.fs.base import AnyFSPath, FileSystem  # noqa: F401
 from dvc_objects.fs.errors import (  # noqa: F401
@@ -121,7 +121,7 @@ def _get_cloud_fs(repo_config, **kwargs):
 
     remote_conf = get_fs_config(repo_config, **kwargs)
     try:
-        remote_conf = SCHEMA["remote"][str](remote_conf)
+        remote_conf = SCHEMA["remote"][str](remote_conf)  # type: ignore[index]
     except Invalid as exc:
         raise RepoConfigError(str(exc)) from None
 

dvc/fs/callbacks.py

@@ -1,6 +1,6 @@
 # pylint: disable=unused-import
 from contextlib import ExitStack
-from typing import TYPE_CHECKING, Any, BinaryIO, Dict, Optional, Union, cast
+from typing import TYPE_CHECKING, Any, BinaryIO, Dict, Optional, Union
 
 from dvc.utils.objects import cached_property
 from dvc_objects.fs.callbacks import (  # noqa: F401
@@ -54,7 +54,8 @@ class RichCallback(Callback):
             disable=self.disable,
             console=ui.error_console,
         )
-        return cast(RichTransferProgress, self._stack.enter_context(progress))
+        self._stack.enter_context(progress)
+        return progress
 
     @cached_property
     def task(self) -> "TaskID":

dvc/fs/dvc.py

@@ -289,7 +289,7 @@ class _DVCFileSystem(AbstractFileSystem):  # pylint:disable=abstract-method
                 ):
                     fs_infos[fs.path.name(info["name"])] = info
             except (FileNotFoundError, NotADirectoryError):
-                pass  # noqa: S110
+                pass
 
         dvcfiles = kwargs.get("dvcfiles", False)
 
@@ -316,7 +316,9 @@ class _DVCFileSystem(AbstractFileSystem):  # pylint:disable=abstract-method
         ignore_subrepos = kwargs.get("ignore_subrepos", True)
         return self._info(key, path, ignore_subrepos=ignore_subrepos)
 
-    def _info(self, key, path, ignore_subrepos=True, check_ignored=True):  # noqa: C901
+    def _info(  # noqa: C901, PLR0912
+        self, key, path, ignore_subrepos=True, check_ignored=True
+    ):
         repo, dvc_fs, subkey = self._get_subrepo_info(key)
 
         dvc_info = None
@@ -359,6 +361,19 @@ class _DVCFileSystem(AbstractFileSystem):  # pylint:disable=abstract-method
         info["name"] = path
         return info
 
+    def get_file(self, rpath, lpath, **kwargs):  # pylint: disable=arguments-differ
+        key = self._get_key_from_relative(rpath)
+        fs_path = self._from_key(key)
+        try:
+            return self.repo.fs.get_file(fs_path, lpath, **kwargs)
+        except FileNotFoundError:
+            _, dvc_fs, subkey = self._get_subrepo_info(key)
+            if not dvc_fs:
+                raise
+
+        dvc_path = _get_dvc_path(dvc_fs, subkey)
+        return dvc_fs.get_file(dvc_path, lpath, **kwargs)
+
 
 class DVCFileSystem(FileSystem):
     protocol = "local"

dvc/info.py

@@ -19,7 +19,6 @@ SUBPROJECTS = (
     "dvc_objects",
     "dvc_render",
     "dvc_task",
-    "dvclive",
     "scmrepo",
 )
 package = "" if PKG is None else f" ({PKG})"

dvc/logger.py

@@ -9,57 +9,47 @@ import colorama
 
 from dvc.progress import Tqdm
 
-FOOTER = (
-    "\n{yellow}Having any troubles?{nc}"
-    " Hit us up at {blue}https://dvc.org/support{nc},"
-    " we are always happy to help!"
-).format(
-    blue=colorama.Fore.BLUE,
-    nc=colorama.Fore.RESET,
-    yellow=colorama.Fore.YELLOW,
-)
 
-
-def addLoggingLevel(levelName, levelNum, methodName=None):
+def add_logging_level(level_name, level_num, method_name=None):
     """
     Adds a new logging level to the `logging` module and the
     currently configured logging class.
 
-    Uses the existing numeric levelNum if already defined.
+    Uses the existing numeric level_num if already defined.
 
     Based on https://stackoverflow.com/questions/2183233
     """
-    if methodName is None:
-        methodName = levelName.lower()
+    if method_name is None:
+        method_name = level_name.lower()
 
     # If the level name is already defined as a top-level `logging`
     # constant, then adopt the existing numeric level.
-    if hasattr(logging, levelName):
-        existingLevelNum = getattr(logging, levelName)
-        assert isinstance(existingLevelNum, int)
-        levelNum = existingLevelNum
+    if hasattr(logging, level_name):
+        existing_level_num = getattr(logging, level_name)
+        assert isinstance(existing_level_num, int)
+        level_num = existing_level_num
 
-    def logForLevel(self, message, *args, **kwargs):
-        if self.isEnabledFor(levelNum):
+    def log_for_level(self, message, *args, **kwargs):
+        if self.isEnabledFor(level_num):
             # pylint: disable=protected-access
-            self._log(levelNum, message, args, **kwargs)
+            self._log(level_num, message, args, **kwargs)
 
-    def logToRoot(message, *args, **kwargs):
-        logging.log(levelNum, message, *args, **kwargs)
+    def log_to_root(message, *args, **kwargs):
+        logging.log(level_num, message, *args, **kwargs)
 
     # getLevelName resolves the numeric log level if already defined,
     # otherwise returns a string
-    if not isinstance(logging.getLevelName(levelName), int):
-        logging.addLevelName(levelNum, levelName)
+    if not isinstance(logging.getLevelName(level_name), int):
+        logging.addLevelName(level_num, level_name)
 
-    if not hasattr(logging, levelName):
-        setattr(logging, levelName, levelNum)
+    if not hasattr(logging, level_name):
+        setattr(logging, level_name, level_num)
 
-    if not hasattr(logging.getLoggerClass(), methodName):
-        setattr(logging.getLoggerClass(), methodName, logForLevel)
+    if not hasattr(logging.getLoggerClass(), method_name):
+        setattr(logging.getLoggerClass(), method_name, log_for_level)
 
-    if not hasattr(logging, methodName):
-        setattr(logging, methodName, logToRoot)
+    if not hasattr(logging, method_name):
+        setattr(logging, method_name, log_to_root)
 
 
 class LoggingException(Exception):
@@ -139,7 +129,7 @@ class ColorFormatter(logging.Formatter):
 
 
 class LoggerHandler(logging.StreamHandler):
-    def handleError(self, record):
+    def handleError(self, record):  # noqa: N802
         super().handleError(record)
         raise LoggingException(record)
 
@@ -157,7 +147,7 @@ class LoggerHandler(logging.StreamHandler):
                         if not _is_verbose():
                             return
                     except Exception:  # noqa, pylint: disable=broad-except
-                        pass  # noqa: S110
+                        pass
 
             msg = self.format(record)
             Tqdm.write(msg, file=self.stream, end=getattr(self, "terminator", "\n"))
@@ -200,7 +190,7 @@ def setup(level: int = logging.INFO, log_colors: bool = True) -> None:
     console_debug.setFormatter(formatter)
     console_debug.addFilter(exclude_filter(logging.INFO))
 
-    addLoggingLevel("TRACE", logging.DEBUG - 5)
+    add_logging_level("TRACE", logging.DEBUG - 5)
 
     console_trace = LoggerHandler(sys.stdout)
     console_trace.setLevel(logging.TRACE)  # type: ignore[attr-defined]

dvc/output.py

@@ -2,8 +2,9 @@ import logging
 import os
 import posixpath
 from collections import defaultdict
+from contextlib import suppress
 from operator import itemgetter
-from typing import TYPE_CHECKING, Dict, List, Optional, Set, Tuple, Type, cast
+from typing import TYPE_CHECKING, Dict, List, Optional, Set, Tuple, Type
 from urllib.parse import urlparse
 
 from funcy import collecting, first, project
@@ -20,7 +21,6 @@ from dvc.exceptions import (
     RemoteCacheRequiredError,
 )
 from dvc.utils.objects import cached_property
-from dvc_data.hashfile import Tree
 from dvc_data.hashfile import check as ocheck
 from dvc_data.hashfile import load as oload
 from dvc_data.hashfile.build import build
@@ -29,6 +29,7 @@ from dvc_data.hashfile.hash_info import HashInfo
 from dvc_data.hashfile.istextfile import istextfile
 from dvc_data.hashfile.meta import Meta
 from dvc_data.hashfile.transfer import transfer as otransfer
+from dvc_data.hashfile.tree import Tree
 from dvc_objects.errors import ObjectFormatError
 
 from .annotations import ANNOTATION_FIELDS, ANNOTATION_SCHEMA, Annotation
@@ -312,7 +313,7 @@ class Output:
     IsStageFileError: Type[DvcException] = OutputIsStageFileError
     IsIgnoredError: Type[DvcException] = OutputIsIgnoredError
 
-    def __init__(
+    def __init__(  # noqa: PLR0913
         self,
         stage,
         path,
@@ -709,7 +710,7 @@ class Output:
         except _CheckoutError as exc:
             raise CheckoutError(exc.paths, {})  # noqa: B904
 
-    def commit(self, filter_info=None) -> None:
+    def commit(self, filter_info=None, relink=True) -> None:
         if not self.exists:
             raise self.DoesNotExistError(self)
 
@@ -719,8 +720,11 @@ class Output:
             granular = (
                 self.is_dir_checksum and filter_info and filter_info != self.fs_path
             )
+            # NOTE: trying to use hardlink during transfer only if we will be
+            # relinking later
+            hardlink = relink
             if granular:
-                obj = self._commit_granular_dir(filter_info)
+                obj = self._commit_granular_dir(filter_info, hardlink)
             else:
                 staging, _, obj = build(
                     self.cache,
@@ -734,41 +738,43 @@ class Output:
                     self.cache,
                     {obj.hash_info},
                     shallow=False,
-                    hardlink=True,
+                    hardlink=hardlink,
                 )
-            self._checkout(
-                filter_info or self.fs_path,
-                self.fs,
-                obj,
-                self.cache,
-                relink=True,
-                state=self.repo.state,
-                prompt=prompt.confirm,
-            )
-            self.set_exec()
+            if relink:
+                self._checkout(
+                    filter_info or self.fs_path,
+                    self.fs,
+                    obj,
+                    self.cache,
+                    relink=True,
+                    state=self.repo.state,
+                    prompt=prompt.confirm,
+                )
+                self.set_exec()
 
-    def _commit_granular_dir(self, filter_info) -> Optional["HashFile"]:
+    def _commit_granular_dir(self, filter_info, hardlink) -> Optional["HashFile"]:
         prefix = self.fs.path.parts(self.fs.path.relpath(filter_info, self.fs_path))
-        staging, _, save_obj = build(
+        staging, _, obj = build(
             self.cache,
             self.fs_path,
             self.fs,
             self.hash_name,
             ignore=self.dvcignore,
         )
-        save_obj = cast(Tree, save_obj)
-        save_obj = cast(Tree, save_obj.filter(prefix))
+        assert isinstance(obj, Tree)
+        save_obj = obj.filter(prefix)
+        assert isinstance(save_obj, Tree)
         checkout_obj = save_obj.get_obj(self.cache, prefix)
         otransfer(
             staging,
             self.cache,
             {save_obj.hash_info} | {oid for _, _, oid in save_obj},
             shallow=True,
-            hardlink=True,
+            hardlink=hardlink,
         )
         return checkout_obj
 
-    def dumpd(self, **kwargs):  # noqa: C901
+    def dumpd(self, **kwargs):  # noqa: C901, PLR0912
         ret: Dict[str, Any] = {}
         with_files = (
             (not self.IS_DEPENDENCY or self.stage.is_import)
@@ -822,7 +828,7 @@ class Output:
         if with_files:
             obj = self.obj or self.get_obj()
             if obj:
-                obj = cast("Tree", obj)
+                assert isinstance(obj, Tree)
                 ret[self.PARAM_FILES] = [
                     split_file_meta_from_cloud(f)
                     for f in _serialize_tree_obj_to_files(obj)
@@ -870,7 +876,7 @@ class Output:
         fs_path = self.fs.path
         if filter_info and filter_info != self.fs_path:
             prefix = fs_path.relparts(filter_info, self.fs_path)
-            obj = cast(Tree, obj)
+            assert isinstance(obj, Tree)
             obj = obj.get_obj(self.cache, prefix)
 
         return obj
@@ -1012,7 +1018,8 @@ class Output:
         except FileNotFoundError:
             if self.remote:
                 kwargs["remote"] = self.remote
-            self.repo.cloud.pull([obj.hash_info], **kwargs)
+            with suppress(Exception):
+                self.repo.cloud.pull([obj.hash_info], **kwargs)
 
         if self.obj:
             return self.obj
@@ -1052,14 +1059,14 @@ class Output:
             return None
 
         obj = self.get_obj()
+        assert obj is None or isinstance(obj, Tree)
         if filter_info and filter_info != self.fs_path:
             assert obj
             prefix = self.fs.path.parts(self.fs.path.relpath(filter_info, self.fs_path))
-            obj = cast(Tree, obj)
             return obj.filter(prefix)
-        return cast(Tree, obj)
+        return obj
 
-    def get_used_objs(  # noqa: C901
+    def get_used_objs(  # noqa: C901, PLR0911
         self, **kwargs
     ) -> Dict[Optional["ObjectDB"], Set["HashInfo"]]:
         """Return filtered set of used object IDs for this out."""

dvc/parsing/context.py

@@ -155,7 +155,7 @@ class Value(Node):
 PRIMITIVES = (int, float, str, bytes, bool)
 
 
-class Container(Node, ABC):  # noqa: B024
+class Container(Node, ABC):
     meta: Meta
     data: Union[list, dict]
     _key_transform = staticmethod(identity)

dvc/parsing/interpolate.py

@@ -102,9 +102,8 @@ def _(obj: dict):  # noqa: C901
         if isinstance(v, bool):
             if v:
                 result += f"--{k} "
-            else:
-                if config.get("bool", "store_true") == "boolean_optional":
-                    result += f"--no-{k} "
+            elif config.get("bool", "store_true") == "boolean_optional":
+                result += f"--no-{k} "
 
         elif isinstance(v, str):
             result += f"--{k} {escape_str(v)} "

dvc/parsing/versions.py

@@ -19,7 +19,7 @@ class VersionEnum(str, enum.Enum):
         return [v.value for v in cls]
 
 
-class LOCKFILE_VERSION(VersionEnum):
+class LOCKFILE_VERSION(VersionEnum):  # noqa: N801
     V1 = "1.0"
     V2 = "2.0"
 

dvc/proc/exceptions.py

@@ -1,23 +0,0 @@
-from dvc.exceptions import DvcException
-
-
-class ProcessNotTerminatedError(DvcException):
-    def __init__(self, name):
-        super().__init__(f"Managed process '{name}' has not been terminated.")
-
-
-class ProcessNotFoundError(DvcException):
-    def __init__(self, name):
-        super().__init__(f"Managed process '{name}' does not exist.")
-
-
-class TimeoutExpired(DvcException):
-    def __init__(self, cmd, timeout):
-        super().__init__(f"'{cmd}' did not complete before timeout '{timeout}'")
-        self.cmd = cmd
-        self.timeout = timeout
-
-
-class UnsupportedSignalError(DvcException):
-    def __init__(self, sig):
-        super().__init__(f"Unsupported signal: {sig}")

dvc/proc/manager.py

@@ -1,149 +0,0 @@
-"""Serverless process manager."""
-
-import json
-import logging
-import os
-import signal
-import sys
-from typing import Generator, List, Optional, Tuple, Union
-
-from funcy.flow import reraise
-from shortuuid import uuid
-
-from .exceptions import ProcessNotTerminatedError, UnsupportedSignalError
-from .process import ManagedProcess, ProcessInfo
-
-logger = logging.getLogger(__name__)
-
-
-class ProcessManager:
-    """Manager for controlling background ManagedProcess(es).
-
-    Spawned process entries are kept in the manager directory until they
-    are explicitly removed (with remove() or cleanup()) so that return
-    value and log information can be accessed after a process has completed.
-    """
-
-    def __init__(self, wdir: Optional[str] = None):
-        self.wdir = wdir or "."
-
-    def __iter__(self) -> Generator[str, None, None]:
-        if not os.path.exists(self.wdir):
-            return
-        yield from os.listdir(self.wdir)
-
-    def __getitem__(self, key: str) -> "ProcessInfo":
-        info_path = os.path.join(self.wdir, key, f"{key}.json")
-        try:
-            with open(info_path, encoding="utf-8") as fobj:
-                return ProcessInfo.from_dict(json.load(fobj))
-        except FileNotFoundError:
-            raise KeyError  # noqa: B904
-
-    @reraise(FileNotFoundError, KeyError)
-    def __setitem__(self, key: str, value: "ProcessInfo"):
-        info_path = os.path.join(self.wdir, key, f"{key}.json")
-        with open(info_path, "w", encoding="utf-8") as fobj:
-            return json.dump(value.asdict(), fobj)
-
-    def __delitem__(self, key: str) -> None:
-        from dvc.utils.fs import remove
-
-        path = os.path.join(self.wdir, key)
-        if os.path.exists(path):
-            remove(path)
-
-    def get(self, key: str, default=None):
-        try:
-            return self[key]
-        except KeyError:
-            return default
-
-    def processes(self) -> Generator[Tuple[str, "ProcessInfo"], None, None]:
-        for name in self:
-            try:
-                yield name, self[name]
-            except KeyError:
-                continue
-
-    def spawn(self, args: Union[str, List[str]], name: Optional[str] = None):
-        """Run the given command in the background."""
-        name = name or uuid()
-        pid = ManagedProcess.spawn(
-            args,
-            wdir=os.path.join(self.wdir, name),
-            name=name,
-        )
-        logger.debug(
-            "Spawned managed process '%s' (PID: '%d')",
-            name,
-            pid,
-        )
-
-    def send_signal(self, name: str, sig: int):
-        """Send `signal` to the specified named process."""
-        process_info = self[name]
-        if sys.platform == "win32" and sig not in (
-            signal.SIGTERM,
-            signal.CTRL_C_EVENT,  # pylint: disable=no-member
-            signal.CTRL_BREAK_EVENT,  # pylint: disable=no-member
-        ):
-            raise UnsupportedSignalError(sig)
-
-        def handle_closed_process():
-            logging.warning("Process %s had already aborted unexpectedly.", name)
-            process_info.returncode = -1
-            self[name] = process_info
-
-        if process_info.returncode is None:
-            try:
-                os.kill(process_info.pid, sig)
-            except ProcessLookupError:
-                handle_closed_process()
-                raise
-            except OSError as exc:
-                if getattr(exc, "winerror", None) == 87:
-                    handle_closed_process()
-                    raise ProcessLookupError from exc
-                raise
-
-    def terminate(self, name: str):
-        """Terminate the specified named process."""
-        self.send_signal(name, signal.SIGTERM)
-
-    def kill(self, name: str):
-        """Kill the specified named process."""
-        if sys.platform == "win32":
-            self.send_signal(name, signal.SIGTERM)
-        else:
-            self.send_signal(name, signal.SIGKILL)  # pylint: disable=no-member
-
-    def remove(self, name: str, force: bool = False):
-        """Remove the specified named process from this manager.
-
-        If the specified process is still running, it will be forcefully killed
-        if `force` is True`, otherwise an exception will be raised.
-
-        Raises:
-            ProcessNotTerminatedError if the specified process is still
-            running and was not forcefully killed.
-        """
-        try:
-            process_info = self[name]
-        except KeyError:
-            return
-        if process_info.returncode is None and not force:
-            raise ProcessNotTerminatedError(name)
-        try:
-            self.kill(name)
-        except ProcessLookupError:
-            pass
-        del self[name]
-
-    def cleanup(self, force: bool = False):
-        """Remove stale (terminated) processes from this manager."""
-        for name in self:
-            try:
-                self.remove(name, force)
-            except ProcessNotTerminatedError:
-                continue

dvc/proc/process.py

@@ -1,179 +0,0 @@
-import json
-import logging
-import os
-import shlex
-import subprocess
-from contextlib import AbstractContextManager
-from dataclasses import asdict, dataclass
-from typing import BinaryIO, List, Optional, Union
-
-from shortuuid import uuid
-
-from dvc.utils.objects import cached_property
-
-from .exceptions import TimeoutExpired
-
-logger = logging.getLogger(__name__)
-
-
-@dataclass
-class ProcessInfo:
-    pid: int
-    stdin: Optional[str]
-    stdout: Optional[str]
-    stderr: Optional[str]
-    returncode: Optional[int]
-
-    @classmethod
-    def from_dict(cls, d):
-        return cls(**d)
-
-    def asdict(self):
-        return asdict(self)
-
-
-class ManagedProcess(AbstractContextManager):
-    """Run the specified command with redirected output.
-
-    stdout and stderr will both be redirected to <name>.out.
-    Interactive processes (requiring stdin input) are currently unsupported.
-
-    Parameters:
-        args: Command to be run.
-        wdir: If specified, redirected output files will be placed in `wdir`.
-        name: Name to use for this process, if not specified a UUID will be
-            generated instead.
-    """
-
-    def __init__(
-        self,
-        args: Union[str, List[str]],
-        wdir: Optional[str] = None,
-        name: Optional[str] = None,
-    ):
-        self.args: List[str] = (
-            shlex.split(args, posix=os.name == "posix")
-            if isinstance(args, str)
-            else list(args)
-        )
-        self.wdir = wdir
-        self.name = name or uuid()
-        self.returncode: Optional[int] = None
-        self._stdout: Optional[BinaryIO] = None
-        self._stderr: Optional[BinaryIO] = None
-
-        self._run()
-
-    def __exit__(self, exc_type, exc_value, traceback):
-        self.wait()
-
-    def _close_fds(self):
-        if self._stdout:
-            self._stdout.close()
-            self._stdout = None
-        if self._stderr:
-            self._stderr.close()
-            self._stderr = None
-
-    def _make_path(self, path: str) -> str:
-        return os.path.join(self.wdir, path) if self.wdir else path
-
-    @cached_property
-    def stdout_path(self) -> str:
-        return self._make_path(f"{self.name}.out")
-
-    @cached_property
-    def info_path(self) -> str:
-        return self._make_path(f"{self.name}.json")
-
-    @cached_property
-    def pidfile_path(self) -> str:
-        return self._make_path(f"{self.name}.pid")
-
-    @property
-    def info(self) -> "ProcessInfo":
-        return ProcessInfo(
-            pid=self.pid,
-            stdin=None,
-            stdout=self.stdout_path,
-            stderr=None,
-            returncode=self.returncode,
-        )
-
-    def _make_wdir(self):
-        if self.wdir:
-            os.makedirs(self.wdir, exist_ok=True)
-
-    def _dump(self):
-        self._make_wdir()
-        with open(self.info_path, "w", encoding="utf-8") as fobj:
-            json.dump(self.info.asdict(), fobj)
-        with open(self.pidfile_path, "w", encoding="utf-8") as fobj:
-            fobj.write(str(self.pid))
-
-    def _run(self) -> None:
-        self._make_wdir()
-        logger.debug(
-            "Appending output to '%s'",
-            self.stdout_path,
-        )
-        self._stdout = open(self.stdout_path, "ab")  # noqa: SIM115
-        try:
-            self._proc = subprocess.Popen(
-                self.args,
-                stdin=subprocess.DEVNULL,
-                stdout=self._stdout,
-                stderr=subprocess.STDOUT,
-                close_fds=True,
-                shell=False,
-            )
-            self.pid: int = self._proc.pid
-            self._dump()
-        except Exception:
-            if self._proc is not None:
-                self._proc.kill()
-            self._close_fds()
-            raise
-
-    def wait(self, timeout: Optional[int] = None) -> Optional[int]:
-        """Block until a process started with `run` has completed.
-
-        Raises:
-            TimeoutExpired if `timeout` was set and the process
-            did not terminate after `timeout` seconds.
-        """
-        if self.returncode is not None:
-            return self.returncode
-        try:
-            self._proc.wait(timeout=timeout)
-        except subprocess.TimeoutExpired as exc:
-            raise TimeoutExpired(exc.cmd, exc.timeout) from exc
-        self.returncode = self._proc.returncode
-        self._close_fds()
-        self._dump()
-        return self.returncode
-
-    @classmethod
-    def spawn(cls, *args, **kwargs) -> Optional[int]:
-        """Spawn a ManagedProcess command in the background.
-
-        Returns: The spawned process PID.
-        """
-        import multiprocessing as mp
-
-        proc = mp.Process(
-            target=cls._spawn,
-            args=args,
-            kwargs=kwargs,
-            daemon=True,
-        )
-        proc.start()
-        # Do not terminate the child daemon when the main process exits
-        # pylint: disable=protected-access
-        mp.process._children.discard(proc)  # type: ignore[attr-defined]
-        return proc.pid
-
-    @classmethod
-    def _spawn(cls, *args, **kwargs):
-        with cls(*args, **kwargs):
-            pass

dvc/progress.py

@@ -36,7 +36,7 @@ class Tqdm(tqdm):
         "miniters": 1,
     }
 
-    def __init__(
+    def __init__(  # noqa: PLR0913
         self,
         iterable=None,
         disable=None,

dvc/render/convert.py

@@ -29,10 +29,8 @@ def _group_by_rev(datapoints):
 
 
 def to_json(renderer, split: bool = False) -> List[Dict]:
-    from copy import deepcopy
-
     if renderer.TYPE == "vega":
-        grouped = _group_by_rev(deepcopy(renderer.datapoints))
+        grouped = _group_by_rev(renderer.datapoints)
         if split:
             content = renderer.get_filled_template(skip_anchors=["data"])
         else:

dvc/render/converter/vega.py

@@ -1,4 +1,4 @@
-from copy import deepcopy
+import os
 from typing import Any, Dict, Iterable, List, Optional, Tuple, Union
 
 from funcy import first, last
@@ -194,7 +194,7 @@ class VegaConverter(Converter):
             return first(fields)
         return "x"
 
-    def flat_datapoints(self, revision):  # noqa: C901
+    def flat_datapoints(self, revision):  # noqa: C901, PLR0912
         file2datapoints, properties = self.convert()
 
         props_update = {}
@@ -222,7 +222,13 @@ class VegaConverter(Converter):
             )
 
         all_datapoints = []
-        all_y_fields = {y_field for _, y_field in ys}
+        if ys:
+            all_y_files, all_y_fields = list(zip(*ys))
+            all_y_fields = set(all_y_fields)
+            all_y_files = set(all_y_files)
+        else:
+            all_y_files = set()
+            all_y_fields = set()
 
         # override to unified y field name if there are different y fields
         if len(all_y_fields) > 1:
@@ -230,10 +236,16 @@ class VegaConverter(Converter):
         else:
             props_update["y"] = first(all_y_fields)
 
+        # get common prefix to drop from file names
+        if len(all_y_files) > 1:
+            common_prefix_len = len(os.path.commonpath(all_y_files))
+        else:
+            common_prefix_len = 0
+
         for i, (y_file, y_field) in enumerate(ys):
             if num_xs > 1:
                 x_file, x_field = xs[i]
-            datapoints = deepcopy(file2datapoints.get(y_file, []))
+            datapoints = [d.copy() for d in file2datapoints.get(y_file, [])]
 
             if props_update.get("y", None) == "dvc_inferred_y_value":
                 _update_from_field(
@@ -259,12 +271,13 @@ class VegaConverter(Converter):
                         "They have to have same length."
                     )
 
+            y_file_short = y_file[common_prefix_len:].strip("/\\")
             _update_all(
                 datapoints,
                 update_dict={
                     VERSION_FIELD: {
                         "revision": revision,
-                        FILENAME_FIELD: y_file,
+                        FILENAME_FIELD: y_file_short,
                         "field": y_field,
                     }
                 },

dvc/render/match.py

@@ -1,11 +1,12 @@
+import os
 from collections import defaultdict
 from typing import TYPE_CHECKING, Dict, List, Optional
 
+import dpath
 import dpath.options
-import dpath.util
 from funcy import last
 
-from dvc.repo.plots import infer_data_sources
+from dvc.repo.plots import _normpath, infer_data_sources
 from dvc.utils.plots import get_plot_id
 
 from .convert import _get_converter
@@ -43,17 +44,20 @@ class PlotsData:
 
     def get_definition_data(self, target_files, rev):
         result = {}
-        for file in target_files:
+        for definition_file in target_files:
+            if os.name == "nt":
+                source_file = _normpath(definition_file).replace("\\", "/")
+            else:
+                source_file = definition_file
             file_content = (
                 self.data.get(rev, {})
                 .get("sources", {})
                 .get("data", {})
-                .get(file, {})
+                .get(source_file, {})
                 .get("data", {})
             )
             if file_content:
-                result[file] = file_content
-
+                result[definition_file] = file_content
         return result
 
 

dvc/repo/__init__.py

@@ -3,7 +3,16 @@ import os
 from collections import defaultdict
 from contextlib import contextmanager
 from functools import wraps
-from typing import TYPE_CHECKING, Callable, Iterable, List, Optional, Tuple, Union, cast
+from typing import (
+    TYPE_CHECKING,
+    Callable,
+    ContextManager,
+    Iterable,
+    List,
+    Optional,
+    Tuple,
+    Union,
+)
 
 from dvc.exceptions import FileMissingError
 from dvc.exceptions import IsADirectoryError as DvcIsADirectoryError
@@ -14,7 +23,7 @@ from dvc.utils.fs import path_isin
 from dvc.utils.objects import cached_property
 
 if TYPE_CHECKING:
-    from dvc.fs import FileSystem, GitFileSystem
+    from dvc.fs import FileSystem
     from dvc.fs.data import DataFileSystem
     from dvc.fs.dvc import DVCFileSystem
     from dvc.lock import LockBase
@@ -23,6 +32,7 @@ if TYPE_CHECKING:
     from dvc.stage import Stage
     from dvc.types import DictStrAny
     from dvc_data.hashfile.state import StateBase
+    from dvc_data.index import DataIndex
 
     from .experiments import Experiments
     from .index import Index
@@ -99,17 +109,15 @@ class Repo:
         fs: Optional["FileSystem"] = None,
         uninitialized: bool = False,
         scm: Optional[Union["Git", "NoSCM"]] = None,
-    ) -> Tuple[str, Optional[str], Optional[str]]:
+    ) -> Tuple[str, Optional[str]]:
         from dvc.fs import localfs
         from dvc.scm import SCM, SCMError
 
         dvc_dir: Optional[str] = None
-        tmp_dir: Optional[str] = None
         try:
             root_dir = self.find_root(root_dir, fs)
             fs = fs or localfs
             dvc_dir = fs.path.join(root_dir, self.DVC_DIR)
-            tmp_dir = fs.path.join(dvc_dir, "tmp")
         except NotDvcRepoError:
             if not uninitialized:
                 raise
@@ -124,9 +132,11 @@ class Repo:
                 root_dir = scm.root_dir
 
         assert root_dir
-        return root_dir, dvc_dir, tmp_dir
+        return root_dir, dvc_dir
 
     def _get_database_dir(self, db_name: str) -> Optional[str]:
+        from dvc.fs import localfs
+
         # NOTE: by default, store SQLite-based remote indexes and state's
         # `links` and `md5s` caches in the repository itself to avoid any
         # possible state corruption in 'shared cache dir' scenario, but allow
@@ -139,18 +149,24 @@ class Repo:
 
         import hashlib
 
-        root_dir_hash = hashlib.sha224(self.root_dir.encode("utf-8")).hexdigest()
+        if self.local_dvc_dir:
+            fs: "FileSystem" = localfs
+            local_root = fs.path.parent(self.local_dvc_dir)
+        else:
+            fs = self.fs
+            local_root = self.root_dir
+        root_dir_hash = hashlib.sha224(local_root.encode("utf-8")).hexdigest()
 
-        db_dir = self.fs.path.join(
+        db_dir = fs.path.join(
             base_db_dir,
             self.DVC_DIR,
-            f"{self.fs.path.name(self.root_dir)}-{root_dir_hash[0:7]}",
+            f"{fs.path.name(local_root)}-{root_dir_hash[0:7]}",
         )
 
-        self.fs.makedirs(db_dir, exist_ok=True)
+        fs.makedirs(db_dir, exist_ok=True)
         return db_dir
 
-    def __init__(
+    def __init__(  # noqa: PLR0915
         self,
         root_dir: Optional[str] = None,
         fs: Optional["FileSystem"] = None,
@@ -179,6 +195,7 @@ class Repo:
         self._fs_conf = {"repo_factory": repo_factory}
         self._fs = fs or localfs
         self._scm = scm
+        self._data_index = None
 
         if rev and not fs:
             self._scm = scm = SCM(root_dir or os.curdir)
@@ -187,8 +204,10 @@ class Repo:
 
         self.root_dir: str
         self.dvc_dir: Optional[str]
-        self.tmp_dir: Optional[str]
-        self.root_dir, self.dvc_dir, self.tmp_dir = self._get_repo_dirs(
+        (
+            self.root_dir,
+            self.dvc_dir,
+        ) = self._get_repo_dirs(
             root_dir=root_dir,
             fs=self.fs,
             uninitialized=uninitialized,
@@ -211,11 +230,11 @@ class Repo:
             self.lock = LockNoop()
             self.state = StateNoop()
             self.cache = CacheManager(self)
-            self.tmp_dir = None
         else:
-            self.fs.makedirs(cast(str, self.tmp_dir), exist_ok=True)
-
             if isinstance(self.fs, LocalFileSystem):
+                assert self.tmp_dir
+                self.fs.makedirs(self.tmp_dir, exist_ok=True)
+
                 self.lock = make_lock(
                     self.fs.path.join(self.tmp_dir, "lock"),
                     tmp_dir=self.tmp_dir,
@@ -246,6 +265,41 @@ class Repo:
     def __str__(self):
         return self.url or self.root_dir
 
+    @cached_property
+    def local_dvc_dir(self):
+        from dvc.fs import GitFileSystem, LocalFileSystem
+
+        if not self.dvc_dir:
+            return None
+
+        if isinstance(self.fs, LocalFileSystem):
+            return self.dvc_dir
+
+        if not isinstance(self.fs, GitFileSystem):
+            return None
+
+        relparts = ()
+        if self.root_dir != "/":
+            # subrepo
+            relparts = self.fs.path.relparts(self.root_dir, "/")
+
+        dvc_dir = os.path.join(
+            self.scm.root_dir,
+            *relparts,
+            self.DVC_DIR,
+        )
+        if os.path.exists(dvc_dir):
+            return dvc_dir
+
+        return None
+
+    @cached_property
+    def tmp_dir(self):
+        if self.local_dvc_dir is None:
+            return None
+
+        return os.path.join(self.local_dvc_dir, "tmp")
+
     @cached_property
     def index(self) -> "Index":
         from dvc.repo.index import Index
@@ -304,7 +358,7 @@ class Repo:
         return DvcIgnoreFilter(self.fs, self.root_dir)
 
     def get_rev(self):
-        from dvc.fs import LocalFileSystem
+        from dvc.fs import GitFileSystem, LocalFileSystem
 
         assert self.scm
         if isinstance(self.fs, LocalFileSystem):
@@ -312,7 +366,8 @@ class Repo:
 
             with map_scm_exception():
                 return self.scm.get_rev()
-        return cast("GitFileSystem", self.fs).rev
+        assert isinstance(self.fs, GitFileSystem)
+        return self.fs.rev
 
     @cached_property
     def experiments(self) -> "Experiments":
@@ -341,6 +396,21 @@ class Repo:
         # fs.
         self._reset()
 
+    @property
+    def data_index(self) -> "DataIndex":
+        from dvc_data.index import DataIndex
+
+        if self._data_index is None:
+            if self.index_db_dir:
+                index_dir = os.path.join(self.index_db_dir, "index", "data")
+                os.makedirs(index_dir, exist_ok=True)
+
+                self._data_index = DataIndex.open(os.path.join(index_dir, "db.db"))
+            else:
+                self._data_index = DataIndex()
+
+        return self._data_index
+
     def __repr__(self):
         return f"{self.__class__.__name__}: '{self.root_dir}'"
 
@@ -402,7 +472,12 @@ class Repo:
 
         return brancher(self, *args, **kwargs)
 
-    def used_objs(
+    def switch(self, rev: str) -> ContextManager[str]:
+        from dvc.repo.brancher import switch
+
+        return switch(self, rev)
+
+    def used_objs(  # noqa: PLR0913
         self,
         targets=None,
         all_branches=False,
@@ -553,21 +628,19 @@ class Repo:
     def close(self):
         self.scm.close()
         self.state.close()
+        if self._data_index is not None:
+            self._data_index.close()
 
-    def _reset_cached_indecies(self):
+    def _reset(self):
+        self.scm._reset()  # pylint: disable=protected-access
+        self.state.close()
         self.__dict__.pop("index", None)
         self.__dict__.pop("dvcignore", None)
         self.__dict__.pop("dvcfs", None)
         self.__dict__.pop("datafs", None)
 
-    def _reset(self):
-        self.state.close()
-        self.scm._reset()  # pylint: disable=protected-access
-        self._reset_cached_indecies()
-
     def __enter__(self):
         return self
 
     def __exit__(self, exc_type, exc_val, exc_tb):
-        self._reset()
-        self.scm.close()
+        self.close()

dvc/repo/add.py

@@ -148,7 +148,7 @@ VALIDATORS = (
 @validate(*VALIDATORS)
 @locked
 @scm_context
-def add(  # noqa: C901
+def add(
     repo: "Repo",
     targets: "TargetType",
     recursive: bool = False,

dvc/repo/brancher.py

@@ -1,7 +1,17 @@
-from typing import Optional
+import logging
+from contextlib import contextmanager
+from typing import TYPE_CHECKING, Iterator, Optional, Tuple
 
+from scmrepo.git import Git
+
+from dvc.exceptions import NotDvcRepoError
 from dvc.scm import iter_revs
 
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+logger = logging.getLogger(__name__)
+
 
 def brancher(  # noqa: E302
     self,
@@ -60,6 +70,7 @@ def brancher(  # noqa: E302
 
     scm = self.scm
 
+    logger.trace("switching fs to workspace")  # type: ignore[attr-defined]
     self.fs = LocalFileSystem(url=self.root_dir)
     yield "workspace"
 
@@ -79,21 +90,68 @@ def brancher(  # noqa: E302
     )
 
     try:
-        from dvc.fs import GitFileSystem
-
         for sha, names in found_revs.items():
-            self._reset_cached_indecies()  # pylint: disable=W0212
-            self.fs = GitFileSystem(scm=scm, rev=sha)
-            self.root_dir = self.fs.path.join("/", *repo_root_parts)
-
-            if cwd_parts:
-                cwd = self.fs.path.join("/", *cwd_parts)  # type: ignore[unreachable]
-                self.fs.path.chdir(cwd)
-
-            # ignore revs that don't contain repo root
-            # (i.e. revs from before a subdir=True repo was init'ed)
-            if self.fs.exists(self.root_dir):
+            try:
+                _switch_fs(self, sha, repo_root_parts, cwd_parts)
                 yield sha if sha_only else ",".join(names)
+            except NotDvcRepoError:
+                # ignore revs that don't contain repo root
+                # (i.e. revs from before a subdir=True repo was init'ed)
+                pass
     finally:
         self.fs = saved_fs
         self.root_dir = saved_root
+
+
+def _switch_fs(
+    repo: "Repo",
+    rev: str,
+    repo_root_parts: Tuple[str, ...],
+    cwd_parts: Tuple[str, ...],
+):
+    from dvc.fs import GitFileSystem, LocalFileSystem
+
+    if rev == "workspace":
+        logger.trace("switching fs to workspace")  # type: ignore[attr-defined]
+        repo.fs = LocalFileSystem(url=repo.root_dir)
+        return
+
+    logger.trace("switching fs to revision %s", rev[:7])  # type: ignore[attr-defined]
+    assert isinstance(repo.scm, Git)
+    fs = GitFileSystem(scm=repo.scm, rev=rev)
+    root_dir = repo.fs.path.join("/", *repo_root_parts)
+    if not fs.exists(root_dir):
+        raise NotDvcRepoError(f"Commit '{rev[:7]}' does not contain a DVC repo")
+
+    repo.fs = fs
+    repo.root_dir = root_dir
+
+    if cwd_parts:
+        cwd = repo.fs.path.join("/", *cwd_parts)
+        repo.fs.path.chdir(cwd)
+
+
+@contextmanager
+def switch(repo: "Repo", rev: str) -> Iterator[str]:
+    """Switch to a specific revision."""
+    from dvc.scm import resolve_rev
+
+    if rev != "workspace":
+        rev = resolve_rev(repo.scm, rev)
+
+    repo_root_parts = ()
+    if repo.fs.path.isin(repo.root_dir, repo.scm.root_dir):
+        repo_root_parts = repo.fs.path.relparts(repo.root_dir, repo.scm.root_dir)
+
+    cwd_parts = ()
+    if repo.fs.path.isin(repo.fs.path.getcwd(), repo.scm.root_dir):
+        cwd_parts = repo.fs.path.relparts(repo.fs.path.getcwd(), repo.scm.root_dir)
+
+    saved_fs = repo.fs
+    saved_root = repo.root_dir
+    try:
+        _switch_fs(repo, rev, repo_root_parts, cwd_parts)
+        yield rev
+    finally:
+        repo.fs = saved_fs
+        repo.root_dir = saved_root

dvc/repo/commit.py

@@ -43,6 +43,7 @@ def commit(
     force=False,
     allow_missing=False,
     data_only=False,
+    relink=True,
 ):
     stages_info = [
         info
@@ -60,6 +61,10 @@ def commit(
             if any(changes):
                 prompt_to_commit(stage, changes, force=force)
                 stage.save(allow_missing=allow_missing)
-        stage.commit(filter_info=stage_info.filter_info, allow_missing=allow_missing)
+        stage.commit(
+            filter_info=stage_info.filter_info,
+            allow_missing=allow_missing,
+            relink=relink,
+        )
         stage.dump(update_pipeline=False)
     return [s.stage for s in stages_info]

dvc/repo/data.py

@@ -13,17 +13,20 @@ from typing import (
     Union,
 )
 
+from dvc.ui import ui
+
 if TYPE_CHECKING:
     from dvc.repo import Repo
     from dvc.scm import Git, NoSCM
     from dvc_data.index import DataIndex
+    from dvc_data.index.diff import Change
 
 
 def posixpath_to_os_path(path: str) -> str:
     return path.replace(posixpath.sep, os.path.sep)
 
 
-def _adapt_typ(typ):
+def _adapt_typ(typ: str) -> str:
     from dvc_data.index.diff import ADD, DELETE, MODIFY
 
     if typ == MODIFY:
@@ -38,7 +41,7 @@ def _adapt_typ(typ):
     return typ
 
 
-def _adapt_path(change):
+def _adapt_path(change: "Change") -> str:
     isdir = False
     if change.new and change.new.meta:
         isdir = change.new.meta.isdir
@@ -54,9 +57,12 @@ def _diff(
     old: "DataIndex",
     new: "DataIndex",
     *,
-    granular: Optional[bool] = False,
-    with_missing: Optional[bool] = False,
+    granular: bool = False,
+    not_in_cache: bool = False,
+    not_in_remote: bool = False,
+    remote_refresh: bool = False,
 ) -> Dict[str, List[str]]:
+    from dvc_data.index import StorageError
     from dvc_data.index.diff import UNCHANGED, UNKNOWN, diff
 
     ret: Dict[str, List[str]] = {}
@@ -91,14 +97,27 @@ def _diff(
             continue
 
         if (
-            with_missing
+            not_in_cache
             and change.old
             and change.old.hash_info
-            and not old.storage_map[change.key].cache.exists(change.old.hash_info.value)
+            and not old.storage_map.cache_exists(change.old)
         ):
             # NOTE: emulating previous behaviour
             _add_change("not_in_cache", change)
 
+        try:
+            if (
+                not_in_remote
+                and change.old
+                and change.old.hash_info
+                and not old.storage_map.remote_exists(
+                    change.old, refresh=remote_refresh
+                )
+            ):
+                _add_change("not_in_remote", change)
+        except StorageError:
+            pass
+
         _add_change(change.typ, change)
 
     return ret
@@ -143,14 +162,19 @@ def _git_info(scm: Union["Git", "NoSCM"], untracked_files: str = "all") -> GitIn
 def _diff_index_to_wtree(repo: "Repo", **kwargs: Any) -> Dict[str, List[str]]:
     from .index import build_data_index
 
-    workspace = build_data_index(repo.index, repo.root_dir, repo.fs, compute_hash=True)
+    with ui.status("Building workspace index"):
+        workspace = build_data_index(
+            repo.index, repo.root_dir, repo.fs, compute_hash=True
+        )
 
-    return _diff(
-        repo.index.data["repo"],
-        workspace,
-        with_missing=True,
-        **kwargs,
-    )
+    with ui.status("Calculating diff between index/workspace"):
+        return _diff(
+            repo.index.data["repo"],
+            workspace,
+            not_in_cache=True,
+            not_in_remote=True,
+            **kwargs,
+        )
 
 
 def _diff_head_to_index(
@@ -158,17 +182,16 @@ def _diff_head_to_index(
 ) -> Dict[str, List[str]]:
     index = repo.index.data["repo"]
 
-    for rev in repo.brancher(revs=[head]):
-        if rev == "workspace":
-            continue
-
+    with repo.switch(head):
         head_index = repo.index.data["repo"]
 
-    return _diff(head_index, index, **kwargs)
+    with ui.status("Calculating diff between head/index"):
+        return _diff(head_index, index, **kwargs)
 
 
 class Status(TypedDict):
     not_in_cache: List[str]
+    not_in_remote: List[str]
     committed: Dict[str, List[str]]
     uncommitted: Dict[str, List[str]]
     untracked: List[str]
@@ -195,12 +218,18 @@ def _transform_git_paths_to_dvc(repo: "Repo", files: Iterable[str]) -> List[str]
     return [repo.fs.path.relpath(file, start) for file in files]
 
 
-def status(repo: "Repo", untracked_files: str = "no", **kwargs: Any) -> Status:
+def status(
+    repo: "Repo",
+    untracked_files: str = "no",
+    **kwargs: Any,
+) -> Status:
     from dvc.scm import NoSCMError, SCMError
 
     head = kwargs.pop("head", "HEAD")
-    uncommitted_diff = _diff_index_to_wtree(repo, **kwargs)
-    not_in_cache = uncommitted_diff.pop("not_in_cache", [])
+    uncommitted_diff = _diff_index_to_wtree(
+        repo,
+        **kwargs,
+    )
     unchanged = set(uncommitted_diff.pop("unchanged", []))
 
     try:
@@ -215,7 +244,8 @@ def status(repo: "Repo", untracked_files: str = "no", **kwargs: Any) -> Status:
     untracked = _transform_git_paths_to_dvc(repo, untracked)
     # order matters here
     return Status(
-        not_in_cache=not_in_cache,
+        not_in_cache=uncommitted_diff.pop("not_in_cache", []),
+        not_in_remote=uncommitted_diff.pop("not_in_remote", []),
         committed=committed_diff,
         uncommitted=uncommitted_diff,
         untracked=untracked,

dvc/repo/diff.py

@@ -1,39 +1,103 @@
 import logging
 import os
 from collections import defaultdict
-from typing import (
-    TYPE_CHECKING,
-    Any,
-    Dict,
-    Iterable,
-    Iterator,
-    List,
-    Optional,
-    Set,
-    Tuple,
-    cast,
-)
+from typing import Dict, List, Optional
 
 from dvc.exceptions import PathMissingError
 from dvc.repo import locked
-from dvc.utils.collections import ensure_list
-
-if TYPE_CHECKING:
-    from dvc.fs import DVCFileSystem, FileSystem
-    from dvc.output import Output
-    from dvc.repo import Repo
-    from dvc_data.hashfile.obj import HashFile
-    from dvc_data.hashfile.tree import Tree
+from dvc.ui import ui
 
 logger = logging.getLogger(__name__)
 
 
+def _path(entry):
+    if entry and entry.meta and entry.meta.isdir:
+        return os.path.join(*entry.key, "")
+    return os.path.join(*entry.key)
+
+
+def _hash(entry):
+    if entry and entry.hash_info:
+        return entry.hash_info.value
+    return None
+
+
+def _diff(old, new, with_missing=False):
+    from dvc_data.index.diff import ADD, DELETE, MODIFY, RENAME
+    from dvc_data.index.diff import diff as idiff
+
+    ret: "Dict[str, List[Dict]]" = {
+        "added": [],
+        "deleted": [],
+        "modified": [],
+        "renamed": [],
+        "not in cache": [],
+    }
+
+    for change in idiff(
+        old,
+        new,
+        with_renames=True,
+        hash_only=True,
+    ):
+        if change.typ == ADD:
+            ret["added"].append(
+                {
+                    "path": _path(change.new),
+                    "hash": _hash(change.new),
+                }
+            )
+        elif change.typ == DELETE:
+            ret["deleted"].append(
+                {
+                    "path": _path(change.old),
+                    "hash": _hash(change.old),
+                }
+            )
+        elif change.typ == MODIFY:
+            ret["modified"].append(
+                {
+                    "path": _path(change.old),
+                    "hash": {
+                        "old": _hash(change.old),
+                        "new": _hash(change.new),
+                    },
+                }
+            )
+        elif change.typ == RENAME:
+            ret["renamed"].append(
+                {
+                    "path": {
+                        "old": _path(change.old),
+                        "new": _path(change.new),
+                    },
+                    "hash": _hash(change.old),
+                }
+            )
+
+        if (
+            with_missing
+            and change.old
+            and change.old.hash_info
+            and not old.storage_map.cache_exists(change.old)
+        ):
+            ret["not in cache"].append(
+                {
+                    "path": _path(change.old),
+                    "hash": _hash(change.old),
+                }
+            )
+
+    return ret if any(ret.values()) else {}
+
+
 @locked
 def diff(
     self,
     a_rev: str = "HEAD",
     b_rev: Optional[str] = None,
     targets: Optional[List[str]] = None,
+    recursive: bool = False,
 ):
     """
     By default, it compares the workspace with the last commit's fs.
@@ -42,231 +106,58 @@ def diff(
     the concept of `index`, but it keeps the same interface, thus,
     `dvc diff` would be the same as `dvc diff HEAD`.
     """
-
     if self.scm.no_commits:
         return {}
 
-    from dvc.fs.dvc import DVCFileSystem
+    indexes = {}
+    missing_targets = defaultdict(set)
+    with_missing = False
+    if not b_rev:
+        b_rev = "workspace"
+        with_missing = True
 
-    dvcfs = DVCFileSystem(repo=self)
-    targets = [dvcfs.from_os_path(target) for target in ensure_list(targets)]
-
-    b_rev = b_rev if b_rev else "workspace"
-    results = {}
-    missing_targets = {}
     for rev in self.brancher(revs=[a_rev, b_rev]):
-        if rev == "workspace" and rev != b_rev:
+        if rev == "workspace" and b_rev != "workspace":
             # brancher always returns workspace, but we only need to compute
             # workspace paths/checksums if b_rev was None
             continue
 
-        targets_paths = None
-        if targets:
-            # convert targets to paths, and capture any missing targets
-            targets_paths, missing_targets[rev] = _targets_to_paths(dvcfs, targets)
+        def onerror(target, _exc):
+            # pylint: disable-next=cell-var-from-loop
+            missing_targets[rev].add(target)  # noqa: B023
 
-        results[rev] = _paths_checksums(self, targets_paths)
-
-    if targets:
-        # check for overlapping missing targets between a_rev and b_rev
-        for target in set(missing_targets[a_rev]) & set(missing_targets[b_rev]):
-            raise PathMissingError(target, self)
-
-    old = results[a_rev]
-    new = results[b_rev]
-
-    # Compare paths between the old and new fs.
-    # set() efficiently converts dict keys to a set
-    added = sorted(set(new) - set(old))
-    deleted_or_missing = set(old) - set(new)
-    if b_rev == "workspace":
-        # missing status is only applicable when diffing local workspace
-        # against a commit
-        missing = sorted(_filter_missing(dvcfs, deleted_or_missing))
-    else:
-        missing = []
-    deleted = sorted(deleted_or_missing - set(missing))
-    modified = sorted(set(old) & set(new))
-
-    # Cases when file was changed and renamed are resulted
-    # in having deleted and added record
-    # To cover such cases we need to change hashing function
-    # to produce rolling/chunking hash
-
-    renamed = _calculate_renamed(new, old, added, deleted)
-
-    for renamed_item in renamed:
-        added.remove(renamed_item["path"]["new"])
-        deleted.remove(renamed_item["path"]["old"])
-
-    ret = {
-        "added": [{"path": path, "hash": new[path]} for path in added],
-        "deleted": [{"path": path, "hash": old[path]} for path in deleted],
-        "modified": [
-            {"path": path, "hash": {"old": old[path], "new": new[path]}}
-            for path in modified
-            if old[path] != new[path]
-        ],
-        "renamed": renamed,
-        "not in cache": [{"path": path, "hash": old[path]} for path in missing],
-    }
-
-    return ret if any(ret.values()) else {}
-
-
-def _paths_checksums(
-    repo: "Repo", targets: Optional[List[str]]
-) -> Dict[str, Optional[str]]:
-    """
-    A dictionary of checksums addressed by relpaths collected from
-    the current fs outputs.
-
-    To help distinguish between a directory and a file output,
-    the former one will come with a trailing slash in the path:
-
-        directory: "data/"
-        file:      "data"
-    """
-
-    return dict(_output_paths(repo, targets))
-
-
-def _output_paths(  # noqa: C901
-    repo: "Repo", targets: Optional[List[str]]
-) -> Iterable[Tuple[str, Optional[str]]]:
-    from dvc.fs import LocalFileSystem
-    from dvc_data.hashfile.build import build
-
-    on_working_fs = isinstance(repo.fs, LocalFileSystem)
-
-    def _exists(output: "Output") -> bool:
-        if on_working_fs:
-            return output.exists
-        return True
-
-    def _to_path(output: "Output") -> str:
-        relparts = output.fs.path.relparts(output.fs_path)
-        base = os.path.join(*relparts)
-        if output.is_dir_checksum:
-            return os.path.join(base, "")
-        return base
-
-    for output in repo.index.outs:
-        if not _exists(output):
-            continue
-
-        yield_output = targets is None or any(
-            output.fs.path.isin_or_eq(output.fs_path, target) for target in targets
+        view = self.index.targets_view(
+            targets,
+            onerror=onerror,
+            recursive=recursive,
         )
 
-        obj: Optional["HashFile"]
-        if on_working_fs:
-            _, _, obj = build(
-                repo.cache.repo,
-                output.fs_path,
-                repo.cache.repo.fs,
-                "md5",
-                dry_run=True,
-                ignore=output.dvcignore,
-            )
-            hash_info = obj.hash_info
-        else:
-            hash_info = output.hash_info
-            obj = output.get_obj()
+        if rev == "workspace":
+            from .index import build_data_index
 
-        if yield_output:
-            yield _to_path(output), hash_info.value
-
-        if not obj:
-            continue
-
-        if output.is_dir_checksum and (
-            yield_output
-            or (
-                isinstance(targets, list)
-                and any(
-                    output.fs.path.isin(target, output.fs_path) for target in targets
+            with ui.status("Building workspace index"):
+                data = build_data_index(
+                    view,
+                    self.root_dir,
+                    self.fs,
+                    compute_hash=True,
                 )
-            )
-        ):
-            obj = cast("Tree", obj)
-            yield from _dir_output_paths(output.fs, output.fs_path, obj, targets)
-
-
-def _dir_output_paths(
-    fs: "FileSystem",
-    fs_path: str,
-    obj: "Tree",
-    targets: Optional[Iterable[str]] = None,
-) -> Iterator[Tuple[str, str]]:
-    base = os.path.join(*fs.path.relparts(fs_path))
-    for key, _, oid in obj:
-        fname = fs.path.join(fs_path, *key)
-        if targets is None or any(
-            fs.path.isin_or_eq(fname, target) for target in targets
-        ):
-            # pylint: disable=no-member
-            yield os.path.join(base, *key), oid.value
-
-
-def _filter_missing(dvcfs: "DVCFileSystem", paths: Set[str]) -> Iterator[str]:
-    for path in paths:
-        fs_path = dvcfs.from_os_path(path)
-        try:
-            info = dvcfs.info(fs_path)
-            dvc_info = info.get("dvc_info", {})
-            entry = dvc_info.get("entry")
-            if (
-                entry
-                and info["type"] == "directory"
-                and not dvcfs.repo.cache.local.exists(entry.hash_info.value)
-            ):
-                yield path
-        except FileNotFoundError:
-            pass
-
-
-def _targets_to_paths(
-    dvcfs: "DVCFileSystem", targets: List[str]
-) -> Tuple[List[str], List[str]]:
-    paths: List[str] = []
-    missing: List[str] = []
-
-    if not targets:
-        return paths, missing
-
-    for target in targets:
-        if dvcfs.exists(target):
-            paths.append(dvcfs.repo.fs.path.abspath(target))
         else:
-            missing.append(target)
+            data = view.data["repo"]
 
-    return paths, missing
+        assert rev not in indexes
+        indexes[rev] = data
 
+    if targets:
+        old_missing = missing_targets.get(a_rev, set())
+        new_missing = missing_targets.get(b_rev, set())
 
-def _calculate_renamed(
-    new: Dict[str, Any],
-    old: Dict[str, Any],
-    added: Iterable[str],
-    deleted: Iterable[str],
-) -> List[Dict[str, Dict[str, str]]]:
-    old_inverted: Dict[str, List[str]] = defaultdict(list)
-    # It is needed to be dict of lists to cover cases
-    # when repo has paths with same hash
-    for path, path_hash in old.items():
-        old_inverted[path_hash].append(path)
+        # check for overlapping missing targets between a_rev and b_rev
+        for target in old_missing & new_missing:
+            raise PathMissingError(target, self)
 
-    renamed = []
-    for path in added:
-        path_hash = new[path]
-        old_paths = old_inverted[path_hash]
-        try:
-            iterator = enumerate(old_paths)
-            index = next(idx for idx, path in iterator if path in deleted)
-        except StopIteration:
-            continue
+    old = indexes[a_rev]
+    new = indexes[b_rev]
 
-        old_path = old_paths.pop(index)
-        renamed.append({"path": {"old": old_path, "new": path}, "hash": path_hash})
-
-    return renamed
+    with ui.status("Calculating diff"):
+        return _diff(old, new, with_missing=with_missing)

dvc/repo/experiments/__init__.py

@@ -11,6 +11,7 @@ from dvc.ui import ui
 from dvc.utils import relpath
 from dvc.utils.objects import cached_property
 
+from .cache import ExpCache
 from .exceptions import (
     BaselineMismatchError,
     ExperimentExistsError,
@@ -22,6 +23,7 @@ from .queue.celery import LocalCeleryQueue
 from .queue.tempdir import TempDirQueue
 from .queue.workspace import WorkspaceQueue
 from .refs import (
+    APPLY_STASH,
     CELERY_FAILED_STASH,
     CELERY_STASH,
     EXEC_APPLY,
@@ -31,6 +33,7 @@ from .refs import (
     WORKSPACE_STASH,
     ExpRefInfo,
 )
+from .stash import ApplyStash
 from .utils import check_ref_format, exp_refs_by_rev, unlocked_repo
 
 if TYPE_CHECKING:
@@ -86,6 +89,14 @@ class Experiments:
     def celery_queue(self) -> LocalCeleryQueue:
         return LocalCeleryQueue(self.repo, CELERY_STASH, CELERY_FAILED_STASH)
 
+    @cached_property
+    def apply_stash(self) -> ApplyStash:
+        return ApplyStash(self.scm, APPLY_STASH)
+
+    @cached_property
+    def cache(self) -> ExpCache:
+        return ExpCache(self.repo)
+
     @property
     def stash_revs(self) -> Dict[str, "ExpStashEntry"]:
         revs = {}
@@ -96,19 +107,9 @@ class Experiments:
     def reproduce_one(
         self,
         tmp_dir: bool = False,
-        machine: Optional[str] = None,
         **kwargs,
     ):
         """Reproduce and checkout a single (standalone) experiment."""
-        if not (tmp_dir or machine):
-            staged, _, _ = self.scm.status(untracked_files="no")
-            if staged:
-                logger.warning(
-                    "Your workspace contains staged Git changes which will be "
-                    "unstaged before running this experiment."
-                )
-                self.scm.reset()
-
         exp_queue: "BaseStashQueue" = (
             self.tempdir_queue if tmp_dir else self.workspace_queue
         )
@@ -417,7 +418,7 @@ class Experiments:
                 try:
                     name = ExpRefInfo.from_ref(ref).name
                 except InvalidExpRefError:
-                    pass  # noqa: S110
+                    pass
             if not name:
                 if rev in self.stash_revs:
                     name = self.stash_revs[rev].name
@@ -493,3 +494,8 @@ class Experiments:
         from dvc.repo.experiments.remove import remove
 
         return remove(self.repo, *args, **kwargs)
+
+    def clean(self, *args, **kwargs):
+        from dvc.repo.experiments.clean import clean
+
+        return clean(self.repo, *args, **kwargs)

dvc/repo/experiments/apply.py

@@ -1,14 +1,14 @@
 import logging
 import os
-from contextlib import contextmanager
 from typing import TYPE_CHECKING, Optional
 
 from dvc.repo import locked
 from dvc.repo.scm_context import scm_context
 from dvc.scm import Git
+from dvc.ui import ui
 from dvc.utils.fs import remove
 
-from .exceptions import ApplyConflictError, BaselineMismatchError, InvalidExpRevError
+from .exceptions import BaselineMismatchError, InvalidExpRevError
 from .executor.base import BaseExecutor
 from .refs import EXEC_APPLY
 
@@ -21,11 +21,9 @@ logger = logging.getLogger(__name__)
 
 @locked
 @scm_context
-def apply(repo: "Repo", rev: str, force: bool = True, **kwargs):  # noqa: C901
-    from scmrepo.exceptions import SCMError as _SCMError
-
+def apply(repo: "Repo", rev: str, **kwargs):  # noqa: C901
     from dvc.repo.checkout import checkout as dvc_checkout
-    from dvc.scm import GitMergeError, RevError, resolve_rev
+    from dvc.scm import RevError, resolve_rev
 
     exps: "Experiments" = repo.experiments
 
@@ -34,7 +32,6 @@ def apply(repo: "Repo", rev: str, force: bool = True, **kwargs):  # noqa: C901
     assert isinstance(repo.scm, Git)
     try:
         exp_rev = resolve_rev(repo.scm, rev)
-        exps.check_baseline(exp_rev)
     except RevError as exc:
         (
             exp_ref_info,
@@ -42,14 +39,9 @@ def apply(repo: "Repo", rev: str, force: bool = True, **kwargs):  # noqa: C901
         ) = exps.celery_queue.get_ref_and_entry_by_names(
             rev
         )[rev]
-        baseline_sha = repo.scm.get_rev()
         if exp_ref_info:
-            if baseline_sha != exp_ref_info.baseline_sha:
-                raise InvalidExpRevError(rev)  # noqa: B904
             exp_rev = repo.scm.get_ref(str(exp_ref_info))
         elif queue_entry:
-            if queue_entry.baseline_rev != baseline_sha:
-                raise InvalidExpRevError(rev)
             exp_rev = queue_entry.stash_rev
             is_stash = True
         else:
@@ -57,68 +49,22 @@ def apply(repo: "Repo", rev: str, force: bool = True, **kwargs):  # noqa: C901
     except BaselineMismatchError as exc:
         raise InvalidExpRevError(rev) from exc
 
-    # NOTE: we don't use scmrepo's stash_workspace() here since we need
-    # finer control over the merge behavior when we unstash everything
-    with _apply_workspace(repo, rev, force):
-        try:
-            repo.scm.merge(exp_rev, commit=False, squash=True)
-        except _SCMError as exc:
-            raise GitMergeError(str(exc), scm=repo.scm)  # noqa: B904
-
-    repo.scm.reset()
-
-    if is_stash:
-        assert repo.tmp_dir is not None
-        args_path = os.path.join(repo.tmp_dir, BaseExecutor.PACKED_ARGS_FILE)
-        if os.path.exists(args_path):
-            remove(args_path)
-
+    _apply(repo, exp_rev, name=rev, is_stash=is_stash)
     dvc_checkout(repo, **kwargs)
 
     repo.scm.set_ref(EXEC_APPLY, exp_rev)
-    logger.info(
-        "Changes for experiment '%s' have been applied to your current workspace.",
-        rev,
+    ui.write(
+        f"Changes for experiment '{rev}' have been applied to your current workspace.",
     )
 
 
-@contextmanager
-def _apply_workspace(repo: "Repo", rev: str, force: bool):
-    from scmrepo.exceptions import MergeConflictError
-    from scmrepo.exceptions import SCMError as _SCMError
+def _apply(repo: "Repo", rev: str, name: Optional[str] = None, is_stash: bool = False):
+    exps: "Experiments" = repo.experiments
 
-    assert isinstance(repo.scm, Git)
-    if repo.scm.is_dirty(untracked_files=True):
-        logger.debug("Stashing workspace")
-        stash_rev: Optional[str] = repo.scm.stash.push(include_untracked=True)
-    else:
-        stash_rev = None
-    try:
-        yield
-    except Exception:  # pylint: disable=broad-except
-        if stash_rev:
-            _clean_and_pop(repo.scm)
-        raise
-    if not stash_rev:
-        return
-
-    try:
-        repo.scm.reset()
-        repo.scm.stash.apply(stash_rev, skip_conflicts=force)
-        repo.scm.stash.drop()
-    except (MergeConflictError, _SCMError) as exc:
-        _clean_and_pop(repo.scm)
-        raise ApplyConflictError(rev) from exc
-    except Exception:  # pylint: disable=broad-except
-        _clean_and_pop(repo.scm)
-        raise
-
-
-def _clean_and_pop(scm: "Git"):
-    """Revert any changes and pop the last stash entry."""
-    scm.reset(hard=True)
-    if scm.is_dirty(untracked_files=True):
-        # drop any changes to untracked files before popping stash
-        scm.stash.push(include_untracked=True)
-        scm.stash.drop()
-    scm.stash.pop()
+    with exps.apply_stash.preserve_workspace(rev, name=name):
+        with repo.scm.detach_head(rev, force=True):
+            if is_stash:
+                assert repo.tmp_dir is not None
+                args_path = os.path.join(repo.tmp_dir, BaseExecutor.PACKED_ARGS_FILE)
+                if os.path.exists(args_path):
+                    remove(args_path)

dvc/repo/experiments/cache.py

@@ -0,0 +1,72 @@
+import logging
+import os
+from typing import TYPE_CHECKING, Optional, Union
+
+from dvc.fs import localfs
+from dvc_objects.db import ObjectDB
+
+from .serialize import DeserializeError, SerializableError, SerializableExp
+from .utils import EXEC_TMP_DIR
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+logger = logging.getLogger(__name__)
+
+
+class ExpCache:
+    """Serialized experiment state cache.
+
+    ODB with git SHAs as keys. Objects can be either SerializableExp or
+    SerializableError.
+    """
+
+    CACHE_DIR = os.path.join(EXEC_TMP_DIR, "cache")
+
+    def __init__(self, repo: "Repo"):
+        path = os.path.join(repo.tmp_dir, self.CACHE_DIR)
+        self.odb = ObjectDB(localfs, path)
+
+    def delete(self, rev: str):
+        self.odb.delete(rev)
+
+    def put(
+        self,
+        exp: Union[SerializableExp, SerializableError],
+        rev: Optional[str] = None,
+        force: bool = False,
+    ):
+        rev = rev or getattr(exp, "rev", None)
+        assert rev
+        assert rev != "workspace"
+        if force or not self.odb.exists(rev):
+            try:
+                self.delete(rev)
+            except FileNotFoundError:
+                pass
+            self.odb.add_bytes(rev, exp.as_bytes())
+            logger.trace(  # type: ignore[attr-defined]
+                "ExpCache: cache put '%s'", rev[:7]
+            )
+
+    def get(self, rev: str) -> Optional[Union[SerializableExp, SerializableError]]:
+        obj = self.odb.get(rev)
+        try:
+            with obj.fs.open(obj.path, "rb") as fobj:
+                data = fobj.read()
+        except FileNotFoundError:
+            logger.trace(  # type: ignore[attr-defined]
+                "ExpCache: cache miss '%s'", rev[:7]
+            )
+            return None
+        for typ in (SerializableExp, SerializableError):
+            try:
+                exp = typ.from_bytes(data)  # type: ignore[attr-defined]
+                logger.trace(  # type: ignore[attr-defined]
+                    "ExpCache: cache load '%s'", rev[:7]
+                )
+                return exp
+            except DeserializeError:
+                continue
+        logger.debug("ExpCache: unknown object type for '%s'", rev)
+        return None

dvc/repo/experiments/clean.py

@@ -0,0 +1,12 @@
+from typing import TYPE_CHECKING
+
+from dvc.ui import ui
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+
+def clean(repo: "Repo"):
+    ui.write("Cleaning up dvc-task messages...")
+    repo.experiments.celery_queue.celery.clean()
+    ui.write("Done!")

dvc/repo/experiments/exceptions.py

@@ -66,18 +66,6 @@ class MultipleBranchError(DvcException):
         self.ref_infos = ref_infos
 
 
-class ApplyConflictError(InvalidArgumentError):
-    def __init__(self, name):
-        super().__init__(
-            f"Experiment '{name}' cannot be applied to because your current "
-            "workspace contains changes which would be overwritten. Either "
-            "'git stash' your current changes before applying this "
-            "experiment, or re-run this command with '--force' to overwrite "
-            "your current changes."
-        )
-        self.name = name
-
-
 class AmbiguousExpRefInfo(InvalidArgumentError):
     def __init__(
         self,

dvc/repo/experiments/executor/base.py

@@ -21,6 +21,7 @@ from typing import (
     Union,
 )
 
+from funcy import get_in
 from scmrepo.exceptions import SCMError
 
 from dvc.env import DVC_EXP_AUTO_PUSH, DVC_EXP_GIT_REMOTE
@@ -547,55 +548,55 @@ class BaseExecutor(ABC):
         from dvc.repo import Repo
         from dvc.stage.monitor import CheckpointKilledError
 
-        dvc = Repo(os.path.join(info.root_dir, info.dvc_dir))
-        info.status = TaskStatus.RUNNING
-        if infofile is not None:
-            info.dump_json(infofile)
-        if cls.QUIET:
-            dvc.scm_context.quiet = cls.QUIET
-        old_cwd = os.getcwd()
-        if info.wdir:
-            os.chdir(os.path.join(dvc.scm.root_dir, info.wdir))
-        else:
-            os.chdir(dvc.root_dir)
-
-        try:
-            post_live_metrics(
-                "start",
-                info.baseline_rev,
-                info.name,
-                "dvc",
-                params=to_studio_params(dvc.params.show()),
-            )
-            logger.debug("Running repro in '%s'", os.getcwd())
-            yield dvc
-            info.status = TaskStatus.SUCCESS
-        except CheckpointKilledError:
-            info.status = TaskStatus.FAILED
-            raise
-        except DvcException:
-            if log_errors:
-                logger.exception("")
-            info.status = TaskStatus.FAILED
-            raise
-        except Exception:
-            if log_errors:
-                logger.exception("unexpected error")
-            info.status = TaskStatus.FAILED
-            raise
-        finally:
-            post_live_metrics(
-                "done",
-                info.baseline_rev,
-                info.name,
-                "dvc",
-                experiment_rev=dvc.experiments.scm.get_ref(EXEC_BRANCH),
-            )
-
+        with Repo(os.path.join(info.root_dir, info.dvc_dir)) as dvc:
+            info.status = TaskStatus.RUNNING
             if infofile is not None:
                 info.dump_json(infofile)
-            dvc.close()
-            os.chdir(old_cwd)
+            if cls.QUIET:
+                dvc.scm_context.quiet = cls.QUIET
+            old_cwd = os.getcwd()
+            if info.wdir:
+                os.chdir(os.path.join(dvc.scm.root_dir, info.wdir))
+            else:
+                os.chdir(dvc.root_dir)
+
+            try:
+                post_live_metrics(
+                    "start",
+                    info.baseline_rev,
+                    info.name,
+                    "dvc",
+                    params=to_studio_params(dvc.params.show()),
+                )
+                logger.debug("Running repro in '%s'", os.getcwd())
+                yield dvc
+                info.status = TaskStatus.SUCCESS
+            except CheckpointKilledError:
+                info.status = TaskStatus.FAILED
+                raise
+            except DvcException:
+                if log_errors:
+                    logger.exception("")
+                info.status = TaskStatus.FAILED
+                raise
+            except Exception:
+                if log_errors:
+                    logger.exception("unexpected error")
+                info.status = TaskStatus.FAILED
+                raise
+            finally:
+                post_live_metrics(
+                    "done",
+                    info.baseline_rev,
+                    info.name,
+                    "dvc",
+                    experiment_rev=dvc.experiments.scm.get_ref(EXEC_BRANCH),
+                    metrics=get_in(dvc.metrics.show(), ["", "data"]),
+                )
+
+                if infofile is not None:
+                    info.dump_json(infofile)
+                os.chdir(old_cwd)
 
     @classmethod
     def _repro_args(cls, dvc):

dvc/repo/experiments/executor/local.py

@@ -5,7 +5,6 @@ from tempfile import mkdtemp
 from typing import TYPE_CHECKING, Optional, Union
 
 from funcy import retry
-from scmrepo.exceptions import SCMError as _SCMError
 from shortuuid import uuid
 
 from dvc.lock import LockError
@@ -20,7 +19,7 @@ from dvc.repo.experiments.refs import (
     EXPS_TEMP,
 )
 from dvc.repo.experiments.utils import EXEC_TMP_DIR, get_exp_rwlock
-from dvc.scm import SCM, Git, GitMergeError
+from dvc.scm import SCM, Git
 from dvc.utils.fs import remove
 from dvc.utils.objects import cached_property
 
@@ -134,10 +133,7 @@ class TempDirExecutor(BaseLocalExecutor):
         self.scm.checkout(head, detach=True)
         merge_rev = self.scm.get_ref(EXEC_MERGE)
 
-        try:
-            self.scm.merge(merge_rev, squash=True, commit=False)
-        except _SCMError as exc:
-            raise GitMergeError(str(exc), scm=self.scm)  # noqa: B904
+        self.scm.stash.apply(merge_rev)
 
     def _config(self, cache_dir):
         local_config = os.path.join(
@@ -229,10 +225,7 @@ class WorkspaceExecutor(BaseLocalExecutor):
                 )
             )
             merge_rev = self.scm.get_ref(EXEC_MERGE)
-            try:
-                self.scm.merge(merge_rev, squash=True, commit=False)
-            except _SCMError as exc:
-                raise GitMergeError(str(exc), scm=self.scm)  # noqa: B904
+            self.scm.stash.apply(merge_rev)
             if branch:
                 self.scm.set_ref(EXEC_BRANCH, branch, symbolic=True)
             elif scm.get_ref(EXEC_BRANCH):

dvc/repo/experiments/executor/ssh.py

@@ -169,7 +169,7 @@ class SSHExecutor(BaseExecutor):
             head = EXEC_BRANCH if branch else EXEC_HEAD
             self._ssh_cmd(fs, f"git checkout {head}")
             merge_rev = scm.get_ref(EXEC_MERGE)
-            self._ssh_cmd(fs, f"git merge --squash --no-commit {merge_rev}")
+            self._ssh_cmd(fs, f"git stash apply {merge_rev}")
 
             if self._setup_script:
                 self._init_setup_script(fs)

dvc/repo/experiments/init.py

@@ -21,9 +21,9 @@ from dvc.exceptions import DvcException
 from dvc.stage import PipelineStage
 
 if TYPE_CHECKING:
+    from dvc.dependency import Dependency
     from dvc.dvcfile import ProjectFile, SingleStageFile
     from dvc.repo import Repo
-    from dvc.dependency import Dependency
 
 from dvc.ui import ui
 
@@ -218,13 +218,12 @@ def init(
             provided=overrides,
             stream=stream,
         )
+    elif "live" in overrides:
+        # suppress `metrics`/`plots` if live is selected.
+        defaults.pop("metrics", None)
+        defaults.pop("plots", None)
     else:
-        if "live" in overrides:
-            # suppress `metrics`/`plots` if live is selected.
-            defaults.pop("metrics", None)
-            defaults.pop("plots", None)
-        else:
-            defaults.pop("live", None)  # suppress live otherwise
+        defaults.pop("live", None)  # suppress live otherwise
 
     context: Dict[str, str] = {**defaults, **overrides}
     assert "cmd" in context

dvc/repo/experiments/queue/base.py

@@ -295,7 +295,7 @@ class BaseStashQueue(ABC):
                 output.
         """
 
-    def _stash_exp(  # noqa: C901
+    def _stash_exp(  # noqa: PLR0915, C901
         self,
         *args,
         params: Optional[Dict[str, List[str]]] = None,
@@ -323,12 +323,12 @@ class BaseStashQueue(ABC):
         .. _Hydra Override:
             https://hydra.cc/docs/next/advanced/override_grammar/basic/
         """
-        with self.scm.detach_head(client="dvc") as orig_head:
-            stash_head = orig_head
-            if baseline_rev is None:
-                baseline_rev = orig_head
+        with self.scm.stash_workspace(reinstate_index=True) as workspace:
+            with self.scm.detach_head(client="dvc") as orig_head:
+                stash_head = orig_head
+                if baseline_rev is None:
+                    baseline_rev = orig_head
 
-            with self.scm.stash_workspace() as workspace:
                 try:
                     if workspace:
                         self.stash.apply(workspace)
@@ -472,6 +472,7 @@ class BaseStashQueue(ABC):
                 force=True,
                 allow_missing=True,
                 data_only=True,
+                relink=False,
             )
 
     @staticmethod

dvc/repo/experiments/queue/celery.py

@@ -90,7 +90,7 @@ class LocalCeleryQueue(BaseStashQueue):
                 "dvc_task.proc.tasks",
             ],
         )
-        app.conf.update({"task_acks_late": True})
+        app.conf.update({"task_acks_late": True, "result_expires": None})
         return app
 
     @cached_property
@@ -138,6 +138,9 @@ class LocalCeleryQueue(BaseStashQueue):
         wdir_hash = hashlib.sha256(self.wdir.encode("utf-8")).hexdigest()[:6]
         node_name = f"dvc-exp-{wdir_hash}-{num}@localhost"
         cmd = ["exp", "queue-worker", node_name]
+        if num == 1:
+            # automatically run celery cleanup when primary worker shuts down
+            cmd.append("--clean")
         name = f"dvc-exp-worker-{num}"
 
         logger.debug("start a new worker: %s, node: %s", name, node_name)
@@ -336,7 +339,8 @@ class LocalCeleryQueue(BaseStashQueue):
                         task_id,
                         remained_entries[entry],
                     )
-                    self.celery.backend.mark_as_failure(task_id, None)
+                    backend = self.celery.backend
+                    backend.mark_as_failure(task_id, None)  # type: ignore[attr-defined]
 
         if remained_revs:
             raise CannotKillTasksError(remained_revs)

dvc/repo/experiments/queue/remove.py

@@ -9,7 +9,7 @@ if TYPE_CHECKING:
     from dvc.repo.experiments.stash import ExpStashEntry
 
 
-def remove_tasks(  # noqa: C901
+def remove_tasks(  # noqa: C901, PLR0912
     celery_queue: "LocalCeleryQueue",
     queue_entries: Iterable["QueueEntry"],
 ):
@@ -40,7 +40,7 @@ def remove_tasks(  # noqa: C901
             msg,
             queue_entry,
         ) in celery_queue._iter_queued():  # pylint: disable=protected-access
-            if queue_entry.stash_rev in stash_revs:
+            if queue_entry.stash_rev in stash_revs and msg.delivery_tag:
                 celery_queue.celery.reject(msg.delivery_tag)
     finally:
         celery_queue.stash.remove_revs(list(stash_revs.values()))
@@ -56,7 +56,8 @@ def remove_tasks(  # noqa: C901
             result: AsyncResult = AsyncResult(task_id)
             if result is not None:
                 result.forget()
-            celery_queue.celery.purge(msg.delivery_tag)
+            if msg.delivery_tag:
+                celery_queue.celery.purge(msg.delivery_tag)
     finally:
         if celery_queue.failed_stash:
             celery_queue.failed_stash.remove_revs(failed_stash_revs)

dvc/repo/experiments/queue/tasks.py

@@ -28,17 +28,17 @@ def setup_exp(entry_dict: Dict[str, Any]) -> "BaseExecutor":
     from dvc.repo import Repo
 
     entry = QueueEntry.from_dict(entry_dict)
-    repo = Repo(entry.dvc_root)
-    # TODO: split executor.init_cache into separate subtask - we can release
-    # exp.scm_lock before DVC push
-    executor = BaseStashQueue.init_executor(
-        repo.experiments,
-        entry,
-        TempDirExecutor,
-        location="dvc-task",
-    )
-    infofile = repo.experiments.celery_queue.get_infofile_path(entry.stash_rev)
-    executor.info.dump_json(infofile)
+    with Repo(entry.dvc_root) as repo:
+        # TODO: split executor.init_cache into separate subtask - we can release
+        # exp.scm_lock before DVC push
+        executor = BaseStashQueue.init_executor(
+            repo.experiments,
+            entry,
+            TempDirExecutor,
+            location="dvc-task",
+        )
+        infofile = repo.experiments.celery_queue.get_infofile_path(entry.stash_rev)
+        executor.info.dump_json(infofile)
     return executor
 
 
@@ -59,23 +59,23 @@ def collect_exp(
     from dvc.repo import Repo
 
     entry = QueueEntry.from_dict(entry_dict)
-    repo = Repo(entry.dvc_root)
-    celery_queue = repo.experiments.celery_queue
-    infofile = celery_queue.get_infofile_path(entry.stash_rev)
-    executor_info = ExecutorInfo.load_json(infofile)
-    logger.debug("Collecting experiment info '%s'", str(executor_info))
-    executor = TempDirExecutor.from_info(executor_info)
-    exec_result = executor_info.result
-    try:
-        if exec_result is not None:
-            BaseStashQueue.collect_executor(repo.experiments, executor, exec_result)
-        else:
-            logger.debug("Experiment failed (Exec result was None)")
-            celery_queue.stash_failed(entry)
-    except Exception:  # pylint: disable=broad-except
-        # Log exceptions but do not re-raise so that task chain execution
-        # continues
-        logger.exception("Failed to collect experiment")
+    with Repo(entry.dvc_root) as repo:
+        celery_queue = repo.experiments.celery_queue
+        infofile = celery_queue.get_infofile_path(entry.stash_rev)
+        executor_info = ExecutorInfo.load_json(infofile)
+        logger.debug("Collecting experiment info '%s'", str(executor_info))
+        executor = TempDirExecutor.from_info(executor_info)
+        exec_result = executor_info.result
+        try:
+            if exec_result is not None:
+                BaseStashQueue.collect_executor(repo.experiments, executor, exec_result)
+            else:
+                logger.debug("Experiment failed (Exec result was None)")
+                celery_queue.stash_failed(entry)
+        except Exception:  # pylint: disable=broad-except
+            # Log exceptions but do not re-raise so that task chain execution
+            # continues
+            logger.exception("Failed to collect experiment")
     return executor.root_dir
 
 
@@ -102,9 +102,9 @@ def run_exp(entry_dict: Dict[str, Any]) -> None:
     from dvc.repo import Repo
 
     entry = QueueEntry.from_dict(entry_dict)
-    repo = Repo(entry.dvc_root)
-    queue = repo.experiments.celery_queue
-    infofile = queue.get_infofile_path(entry.stash_rev)
+    with Repo(entry.dvc_root) as repo:
+        queue = repo.experiments.celery_queue
+        infofile = queue.get_infofile_path(entry.stash_rev)
     executor = setup_exp.s(entry_dict)()
     try:
         cmd = ["dvc", "exp", "exec-run", "--infofile", infofile]

dvc/repo/experiments/refs.py

@@ -7,6 +7,9 @@ from .exceptions import InvalidExpRefError
 EXPS_NAMESPACE = "refs/exps"
 EXPS_STASH = f"{EXPS_NAMESPACE}/stash"
 WORKSPACE_STASH = EXPS_STASH
+APPLY_NAMESPACE = f"{EXPS_NAMESPACE}/apply"
+APPLY_HEAD = f"{APPLY_NAMESPACE}/ORIG_HEAD"
+APPLY_STASH = f"{APPLY_NAMESPACE}/stash"
 CELERY_STASH = f"{EXPS_NAMESPACE}/celery/stash"
 CELERY_FAILED_STASH = f"{EXPS_NAMESPACE}/celery/failed"
 EXEC_NAMESPACE = f"{EXPS_NAMESPACE}/exec"
@@ -18,6 +21,7 @@ EXEC_HEAD = f"{EXEC_NAMESPACE}/EXEC_HEAD"
 EXEC_MERGE = f"{EXEC_NAMESPACE}/EXEC_MERGE"
 EXPS_TEMP = f"{EXPS_NAMESPACE}/temp"
 STASHES = {WORKSPACE_STASH, CELERY_STASH}
+ITER_SKIP_NAMESPACES = {APPLY_NAMESPACE, EXEC_NAMESPACE}
 
 
 class ExpRefInfo:

dvc/repo/experiments/save.py

@@ -2,6 +2,8 @@ import logging
 import os
 from typing import TYPE_CHECKING, List, Optional
 
+from pathspec import PathSpec
+
 from dvc.scm import Git
 
 from .exceptions import ExperimentExistsError
@@ -22,7 +24,7 @@ def _save_experiment(
     name: Optional[str],
     include_untracked: Optional[List[str]],
 ) -> str:
-    repo.commit([], force=True)
+    repo.commit([], force=True, relink=False)
 
     name = name or get_random_exp_name(repo.scm, baseline_rev)
     ref_info = ExpRefInfo(baseline_rev, name)
@@ -59,7 +61,8 @@ def save(
 
     _, _, untracked = repo.scm.status()
     if include_untracked:
-        untracked = [file for file in untracked if file not in include_untracked]
+        spec = PathSpec.from_lines("gitwildmatch", include_untracked)
+        untracked = [file for file in untracked if not spec.match_file(file)]
     if untracked:
         logger.warning(
             (

dvc/repo/experiments/serialize.py

@@ -0,0 +1,164 @@
+import json
+from dataclasses import asdict, dataclass, field
+from datetime import datetime
+from typing import TYPE_CHECKING, Any, Callable, Dict, Optional
+
+from dvc.exceptions import DvcException
+from dvc.repo.metrics.show import _gather_metrics
+from dvc.repo.params.show import _gather_params
+from dvc.utils import onerror_collect, relpath
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+
+class DeserializeError(DvcException):
+    pass
+
+
+class _ISOEncoder(json.JSONEncoder):
+    def default(self, o: object) -> Any:
+        if isinstance(o, datetime):
+            return o.isoformat()
+        return super().default(o)
+
+
+@dataclass(frozen=True)
+class SerializableExp:
+    """Serializable experiment state."""
+
+    rev: str
+    timestamp: Optional[datetime] = None
+    params: Dict[str, Any] = field(default_factory=dict)
+    metrics: Dict[str, Any] = field(default_factory=dict)
+    deps: Dict[str, "_ExpDep"] = field(default_factory=dict)
+    outs: Dict[str, "_ExpOut"] = field(default_factory=dict)
+    status: Optional[str] = None
+    executor: Optional[str] = None
+    error: Optional["SerializableError"] = None
+
+    @classmethod
+    def from_repo(
+        cls,
+        repo: "Repo",
+        rev: Optional[str] = None,
+        onerror: Optional[Callable] = None,
+        **kwargs,
+    ) -> "SerializableExp":
+        """Returns a SerializableExp from the current repo state.
+
+        Params, metrics, deps, outs are filled via repo fs/index, all other fields
+        should be passed via kwargs.
+        """
+        from dvc.dependency import ParamsDependency, RepoDependency
+
+        if not onerror:
+            onerror = onerror_collect
+
+        rev = rev or repo.get_rev()
+        assert rev
+        status: Optional[str] = kwargs.get("status")
+        # NOTE: _gather_params/_gather_metrics return defaultdict which is not
+        # supported in dataclasses.asdict() on all python releases
+        # see https://bugs.python.org/issue35540
+        params = dict(_gather_params(repo, onerror=onerror))
+        if status and status.lower() in ("queued", "failed"):
+            metrics: Dict[str, Any] = {}
+        else:
+            metrics = dict(
+                _gather_metrics(
+                    repo,
+                    targets=None,
+                    rev=rev[:7],
+                    recursive=False,
+                    onerror=onerror_collect,
+                )
+            )
+        return cls(
+            rev=rev,
+            params=params,
+            metrics=metrics,
+            deps={
+                relpath(dep.fs_path, repo.root_dir): _ExpDep(
+                    hash=dep.hash_info.value if dep.hash_info else None,
+                    size=dep.meta.size if dep.meta else None,
+                    nfiles=dep.meta.nfiles if dep.meta else None,
+                )
+                for dep in repo.index.deps
+                if not isinstance(dep, (ParamsDependency, RepoDependency))
+            },
+            outs={
+                relpath(out.fs_path, repo.root_dir): _ExpOut(
+                    hash=out.hash_info.value if out.hash_info else None,
+                    size=out.meta.size if out.meta else None,
+                    nfiles=out.meta.nfiles if out.meta else None,
+                    use_cache=out.use_cache,
+                    is_data_source=out.stage.is_data_source,
+                )
+                for out in repo.index.outs
+                if not (out.is_metric or out.is_plot)
+            },
+            **kwargs,
+        )
+
+    def dumpd(self) -> Dict[str, Any]:
+        return asdict(self)
+
+    def as_bytes(self) -> bytes:
+        return _ISOEncoder().encode(self.dumpd()).encode("utf-8")
+
+    @classmethod
+    def from_bytes(cls, data: bytes):
+        try:
+            parsed = json.loads(data)
+            if "timestamp" in parsed:
+                parsed["timestamp"] = datetime.fromisoformat(parsed["timestamp"])
+            return cls(**parsed)
+        except (TypeError, json.JSONDecodeError) as exc:
+            raise DeserializeError("failed to load SerializableExp") from exc
+
+    @property
+    def contains_error(self) -> bool:
+        return (
+            self.error is not None
+            or self.params.get("error")
+            or any(value.get("error") for value in self.params.values())
+            or self.metrics.get("error")
+            or any(value.get("error") for value in self.metrics.values())
+        )
+
+
+@dataclass(frozen=True)
+class _ExpDep:
+    hash: Optional[str]  # noqa: A003
+    size: Optional[int]
+    nfiles: Optional[int]
+
+
+@dataclass(frozen=True)
+class _ExpOut:
+    hash: Optional[str]  # noqa: A003
+    size: Optional[int]
+    nfiles: Optional[int]
+    use_cache: bool
+    is_data_source: bool
+
+
+@dataclass(frozen=True)
+class SerializableError:
+    msg: str
+    type: str = ""  # noqa: A003
+
+    def dumpd(self) -> Dict[str, Any]:
+        return asdict(self)
+
+    def as_bytes(self) -> bytes:
+        return json.dumps(self.dumpd()).encode("utf-8")
+
+    @classmethod
+    def from_bytes(cls, data: bytes):
+        try:
+            parsed = json.loads(data)
+            return cls(**parsed)
+        except (TypeError, json.JSONDecodeError) as exc:
+            raise DeserializeError("failed to load SerializableError") from exc

dvc/repo/experiments/show.py

@@ -18,12 +18,11 @@ from typing import (
 
 from scmrepo.exceptions import SCMError as InnerScmError
 
-from dvc.repo.metrics.show import _gather_metrics
-from dvc.repo.params.show import _gather_params
+from dvc.exceptions import DvcException
 from dvc.scm import Git, SCMError, iter_revs, resolve_rev
-from dvc.utils import error_handler, onerror_collect, relpath
 
 from .refs import ExpRefInfo
+from .serialize import SerializableError, SerializableExp
 
 if TYPE_CHECKING:
     from scmrepo.git.objects import GitCommit
@@ -40,85 +39,116 @@ class ExpStatus(Enum):
     Failed = 3
 
 
+class _CachedError(DvcException):
+    def __init__(self, msg, typ, *args):
+        super().__init__(msg, *args)
+        self.typ = typ
+
+
 def _is_scm_error(collected_exp: Dict[str, Any]) -> bool:
     if "error" in collected_exp and (
-        isinstance(collected_exp["error"], (SCMError, InnerScmError))
+        isinstance(collected_exp["error"], (_CachedError, SCMError, InnerScmError))
     ):
         return True
     return False
 
 
-def _show_onerror_collect(result: Dict, exception: Exception, *args, **kwargs):
-    onerror_collect(result, exception, *args, **kwargs)
-    result["data"] = {}
+def _format_exp(exp: SerializableExp) -> Dict[str, Any]:
+    # SerializableExp always includes error but we need to strip it from show
+    # output when it is false-y to maintain compatibility w/tools that consume
+    # json output and assume that "error" key presence means there was an error
+    exp_dict = exp.dumpd()
+    if "error" in exp_dict and not exp_dict["error"]:
+        del exp_dict["error"]
+    return {"data": exp_dict}
+
+
+def _format_error(error: SerializableError):
+    msg = error.msg or "None"
+    return {"data": {}, "error": _CachedError(msg, error.type)}
 
 
-@error_handler
 def collect_experiment_commit(
     repo: "Repo",
     exp_rev: str,
     status: ExpStatus = ExpStatus.Success,
-    param_deps=False,
-    running: Optional[Dict[str, Any]] = None,
-    onerror: Optional[Callable] = None,
+    param_deps: bool = False,
+    force: bool = False,
+    **kwargs,
 ) -> Dict[str, Any]:
-    from dvc.dependency import ParamsDependency, RepoDependency
+    cache = repo.experiments.cache
+    # TODO: support filtering serialized exp when param_deps is set
+    if exp_rev != "workspace" and not (force or param_deps):
+        cached_exp = cache.get(exp_rev)
+        if cached_exp:
+            if status == ExpStatus.Running or (
+                isinstance(cached_exp, SerializableExp)
+                and cached_exp.status == ExpStatus.Running.name
+            ):
+                # expire cached queued exp entry once we start running it
+                cache.delete(exp_rev)
+            elif isinstance(cached_exp, SerializableError):
+                return _format_error(cached_exp)
+            else:
+                return _format_exp(cached_exp)
+    try:
+        exp = _collect_from_repo(
+            repo,
+            exp_rev,
+            status=status,
+            param_deps=param_deps,
+            force=force,
+            **kwargs,
+        )
+        if not (
+            exp_rev == "workspace"
+            or status == ExpStatus.Running
+            or param_deps
+            or exp.contains_error
+        ):
+            cache.put(exp, force=True)
+        return _format_exp(exp)
+    except Exception as exc:  # noqa: BLE001, pylint: disable=broad-except
+        logger.debug("", exc_info=True)
+        error = SerializableError(str(exc), type(exc).__name__)
+        return _format_error(error)
 
-    result: Dict[str, Any] = defaultdict(dict)
+
+def _collect_from_repo(
+    repo: "Repo",
+    exp_rev: str,
+    status: ExpStatus = ExpStatus.Success,
+    running: Optional[Dict[str, Any]] = None,
+    **kwargs,
+) -> "SerializableExp":
     running = running or {}
-    for rev in repo.brancher(revs=[exp_rev]):
+    with repo.switch(exp_rev) as rev:
         if rev == "workspace":
-            if exp_rev != "workspace":
-                continue
-            result["timestamp"] = None
+            timestamp: Optional[datetime] = None
         else:
             commit = repo.scm.resolve_commit(rev)
-            result["timestamp"] = datetime.fromtimestamp(commit.commit_time)
+            timestamp = datetime.fromtimestamp(commit.commit_time)
 
-        params = _gather_params(repo, targets=None, deps=param_deps, onerror=onerror)
-        if params:
-            result["params"] = params
-
-        result["deps"] = {
-            relpath(dep.fs_path, repo.root_dir): {
-                "hash": dep.hash_info.value,
-                "size": dep.meta.size,
-                "nfiles": dep.meta.nfiles,
-            }
-            for dep in repo.index.deps
-            if not isinstance(dep, (ParamsDependency, RepoDependency))
-        }
-        result["outs"] = {
-            relpath(out.fs_path, repo.root_dir): {
-                "hash": out.hash_info.value,
-                "size": out.meta.size,
-                "nfiles": out.meta.nfiles,
-                "use_cache": out.use_cache,
-                "is_data_source": out.stage.is_data_source,
-            }
-            for out in repo.index.outs
-            if not (out.is_metric or out.is_plot)
-        }
-
-        result["status"] = status.name
         if status == ExpStatus.Running:
-            result["executor"] = running.get(exp_rev, {}).get("location", None)
+            executor: Optional[str] = running.get(exp_rev, {}).get("location", None)
         else:
-            result["executor"] = None
+            executor = None
 
         if status == ExpStatus.Failed:
-            result["error"] = {
-                "msg": "Experiment run failed.",
-                "type": "",
-            }
-
-        if status not in {ExpStatus.Queued, ExpStatus.Failed}:
-            vals = _gather_metrics(
-                repo, targets=None, rev=rev, recursive=False, onerror=onerror
+            error: Optional["SerializableError"] = SerializableError(
+                "Experiment run failed."
             )
-            result["metrics"] = vals
+        else:
+            error = None
 
-    return result
+        return SerializableExp.from_repo(
+            repo,
+            rev=exp_rev,
+            timestamp=timestamp,
+            status=status.name,
+            executor=executor,
+            error=error,
+        )
 
 
 def _collect_complete_experiment(
@@ -170,6 +200,7 @@ def _collect_branch(
     repo: "Repo",
     baseline: str,
     running: Dict[str, Any],
+    refs: Optional[Iterable[str]] = None,
     **kwargs,
 ) -> Dict[str, Dict[str, Any]]:
     results: Dict[str, Dict[str, Any]] = OrderedDict()
@@ -188,7 +219,11 @@ def _collect_branch(
     commits: List[Tuple[str, "GitCommit", str, List[str]]] = []
 
     assert isinstance(repo.scm, Git)
-    for ref in repo.scm.iter_refs(base=str(ref_info)):
+    if refs:
+        ref_it = (ref for ref in iter(refs) if ref.startswith(str(ref_info)))
+    else:
+        ref_it = repo.scm.iter_refs(base=str(ref_info))
+    for ref in ref_it:
         try:
             commit = repo.scm.resolve_commit(ref)
             exp_rev = resolve_rev(repo.scm, ref)
@@ -217,19 +252,22 @@ def _collect_branch(
 
 
 def get_branch_names(
-    scm: "Git", baseline_set: Iterable[str]
+    scm: "Git", baseline_set: Iterable[str], refs: Optional[Iterable[str]] = None
 ) -> Dict[str, Optional[str]]:
     names: Dict[str, Optional[str]] = {}
-    for base in [
+    bases = [
         f"refs/exps/{baseline[:2]}/{baseline[2:]}/" for baseline in baseline_set
-    ] + ["refs/heads/", "refs/tags/"]:
-        for ref in scm.iter_refs(base=base):
-            if ref:
+    ] + ["refs/heads/", "refs/tags/"]
+    ref_it = iter(refs) if refs else scm.iter_refs()
+    for ref in ref_it:
+        for base in bases:
+            if ref.startswith(base):
                 try:
                     rev = scm.get_ref(ref)
                     names[rev] = ref[len(base) :]
                 except KeyError:
-                    logger.debug("unreosolved ref %s", ref)
+                    logger.debug("unresolved ref %s", ref)
+                break
     logger.debug("found refs for revs: %s", names)
     return names
 
@@ -377,13 +415,12 @@ def move_properties_to_head(result: Dict[str, Dict[str, Dict[str, Any]]]):
                 if rev_result["status"] == ExpStatus.Running.name:
                     head["status"] = ExpStatus.Running.name
                     rev_result["status"] = ExpStatus.Success.name
-            else:
-                if rev_result["checkpoint_tip"] == rev:
-                    head = rev_result
-                    checkpoint = True
+            elif rev_result["checkpoint_tip"] == rev:
+                head = rev_result
+                checkpoint = True
 
 
-def show(
+def show(  # noqa: PLR0913
     repo: "Repo",
     all_branches=False,
     all_tags=False,
@@ -396,12 +433,11 @@ def show(
     param_deps=False,
     onerror: Optional[Callable] = None,
     fetch_running: bool = True,
+    force: bool = False,
 ):
     if repo.scm.no_commits:
         return {}
 
-    onerror = onerror or _show_onerror_collect
-
     res: Dict[str, Dict] = defaultdict(OrderedDict)
 
     if not any([revs, all_branches, all_tags, all_commits]):
@@ -415,7 +451,8 @@ def show(
     found_revs.update(
         iter_revs(repo.scm, revs, num, all_branches, all_tags, all_commits)
     )
-    branch_names = get_branch_names(repo.scm, found_revs)
+    cached_refs = list(repo.scm.iter_refs())
+    branch_names = get_branch_names(repo.scm, found_revs, refs=cached_refs)
 
     running: Dict[str, Dict] = repo.experiments.get_running_exps(
         fetch_refs=fetch_running
@@ -427,7 +464,7 @@ def show(
             found_revs,
             running,
             param_deps=param_deps,
-            onerror=onerror,
+            force=force,
         )
         if not hide_queued
         else {}
@@ -439,6 +476,7 @@ def show(
         running,
         param_deps=param_deps,
         onerror=onerror,
+        force=force,
     )
 
     failed_experiments = (
@@ -448,6 +486,7 @@ def show(
             running,
             param_deps=param_deps,
             onerror=onerror,
+            force=force,
         )
         if not hide_failed
         else {}
@@ -460,6 +499,8 @@ def show(
             running=running,
             param_deps=param_deps,
             onerror=onerror,
+            force=force,
+            refs=cached_refs,
         )
 
     update_new(res, failed_experiments)

dvc/repo/experiments/stash.py

@@ -1,8 +1,18 @@
+import logging
 import re
-from typing import Dict, Iterable, NamedTuple, Optional
+from contextlib import contextmanager
+from typing import Dict, Iterable, Iterator, NamedTuple, Optional
 
 from scmrepo.git import Stash
 
+from dvc.exceptions import DvcException
+from dvc_objects.fs.local import localfs
+from dvc_objects.fs.utils import as_atomic
+
+from .refs import APPLY_HEAD, APPLY_STASH
+
+logger = logging.getLogger(__name__)
+
 
 class ExpStashEntry(NamedTuple):
     """Experiment stash entry.
@@ -72,3 +82,122 @@ class ExpStash(Stash):
             reverse=True,
         ):
             self.drop(index)
+
+
+class ApplyStashEntry(NamedTuple):
+    """Apply stash entry.
+
+    stash_index: Stash index for this entry. Can be None if this commit
+        is not pushed onto the stash ref.
+    head_rev: HEAD Git commit prior to exp apply.
+    rev: Applied experiment commit.
+    name: Optional applied exp name.
+    """
+
+    stash_index: Optional[int]
+    head_rev: str
+    rev: str
+    name: Optional[str]
+
+
+class ApplyStash(Stash):
+    DEFAULT_STASH = APPLY_STASH
+    MESSAGE_FORMAT = "dvc-exp-apply:{head_rev}:{rev}:{name}"
+    MESSAGE_RE = re.compile(
+        r"(?:commit: )"
+        r"dvc-exp-apply:(?P<head_rev>[0-9a-f]+):(?P<rev>[0-9a-f]+)"
+        r":(?P<name>[^~^:\\?\[\]*]*)"
+    )
+
+    @property
+    def stash_revs(self) -> Dict[str, ApplyStashEntry]:
+        revs = {}
+        for i, entry in enumerate(self):
+            msg = entry.message.decode("utf-8").strip()
+            m = self.MESSAGE_RE.match(msg)
+            if m:
+                revs[entry.new_sha.decode("utf-8")] = ApplyStashEntry(
+                    i,
+                    m.group("head_rev"),
+                    m.group("rev"),
+                    m.group("name"),
+                )
+        return revs
+
+    @classmethod
+    def format_message(
+        cls,
+        head_rev: str,
+        rev: str,
+        name: Optional[str] = None,
+    ) -> str:
+        return cls.MESSAGE_FORMAT.format(
+            head_rev=head_rev, rev=rev, name=name if name else ""
+        )
+
+    @contextmanager
+    def preserve_workspace(
+        self, rev: str, name: Optional[str] = None
+    ) -> Iterator[Optional[str]]:
+        if len(self):
+            logger.debug("Clearing existing exp-apply stash")
+            self.clear()
+        head = self.scm.get_rev()
+        self.scm.set_ref(APPLY_HEAD, head)
+        message = self.format_message(head, rev, name=name)
+        stash_rev = self.push(message=message, include_untracked=True)
+        try:
+            yield stash_rev
+            if stash_rev:
+                self._apply_difference(stash_rev, rev)
+        except Exception:  # pylint: disable=broad-except
+            self.revert_workspace()
+            raise
+
+    def _apply_difference(self, stash_rev: str, rev: str):
+        """Selectively apply changes from stash_rev.
+
+        Only changes to files from left which do not exist in right will be applied.
+        """
+        self._copy_difference(stash_rev, rev)
+        commit = self.scm.resolve_commit(stash_rev)
+        for parent_rev in commit.parents:
+            parent_commit = self.scm.resolve_commit(parent_rev)
+            if parent_commit.message.startswith("untracked files on "):
+                self._copy_difference(parent_rev, rev)
+
+    def _copy_difference(self, left_rev: str, right_rev: str):
+        left_fs = self.scm.get_fs(left_rev)
+        right_fs = self.scm.get_fs(right_rev)
+        paths = [path for path in left_fs.find("/") if not right_fs.exists(path)]
+        dest_paths = [
+            localfs.path.join(self.scm.root_dir, left_fs.path.relpath(path, "/"))
+            for path in paths
+        ]
+        for src, dest in zip(paths, dest_paths):
+            with as_atomic(localfs, dest, create_parents=True) as tmp_file:
+                left_fs.get_file(src, tmp_file)
+
+    def revert_workspace(self):
+        apply_head = self.scm.get_ref(self.ref)
+        head = self.scm.get_rev()
+        if apply_head != head:
+            raise DvcException(
+                f"Cannot revert workspace, current HEAD '{head[:7]}' does not match the"
+                f" pre-apply HEAD '{apply_head[:7]}'"
+            )
+        self.scm.reset(hard=True)
+        if len(self):
+            # In the event that the apply-stash and current workspace contain
+            # conflicting untracked files, we do:
+            #   1. stash the current untracked files
+            #   2. restore/pop the apply-stash (with untracked files)
+            #   3. restore/pop the untracked files from (1) and ignore any conflicts
+            #      (forcefully reverting to the apply-stash version)
+            workspace_rev = self.scm.stash.push(include_untracked=True)
+            try:
+                self.pop()
+            finally:
+                if workspace_rev:
+                    self.scm.stash.pop(skip_conflicts=True)
+        self.scm.remove_ref(self.ref)

dvc/repo/fetch.py

@@ -19,7 +19,7 @@ logger = logging.getLogger(__name__)
 
 
 @locked
-def fetch(  # noqa: C901
+def fetch(  # noqa: C901, PLR0913
     self,
     targets=None,
     jobs=None,

dvc/repo/gc.py

@@ -23,7 +23,7 @@ def _validate_args(**kwargs):
 
 
 @locked
-def gc(
+def gc(  # noqa: PLR0913
     self: "Repo",
     all_branches: bool = False,
     cloud: bool = False,

dvc/repo/graph.py

@@ -33,10 +33,10 @@ def get_pipeline(pipelines, node):
     return found[0]
 
 
-def get_pipelines(G: "DiGraph"):
+def get_pipelines(graph: "DiGraph"):
     import networkx as nx
 
-    return [G.subgraph(c).copy() for c in nx.weakly_connected_components(G)]
+    return [graph.subgraph(c).copy() for c in nx.weakly_connected_components(graph)]
 
 
 def collect_pipeline(stage: "Stage", graph: "DiGraph") -> Iterator["Stage"]:
@@ -97,7 +97,7 @@ def build_graph(stages, outs_trie=None):
 
     from .trie import build_outs_trie
 
-    G = nx.DiGraph()
+    graph = nx.DiGraph()
 
     # Use trie to efficiently find overlapping outs and deps
     outs_trie = outs_trie or build_outs_trie(stages)
@@ -108,7 +108,7 @@ def build_graph(stages, outs_trie=None):
             raise StagePathAsOutputError(stage, str(out))
 
     # Building graph
-    G.add_nodes_from(stages)
+    graph.add_nodes_from(stages)
     for stage in stages:
         for dep in stage.deps:
             if dep.fs_path is None:
@@ -119,10 +119,10 @@ def build_graph(stages, outs_trie=None):
             if outs_trie.has_subtrie(dep_key):
                 overlapping.extend(outs_trie.values(prefix=dep_key))
 
-            G.add_edges_from((stage, out.stage) for out in overlapping)
-    check_acyclic(G)
+            graph.add_edges_from((stage, out.stage) for out in overlapping)
+    check_acyclic(graph)
 
-    return G
+    return graph
 
 
 # NOTE: using stage graph instead of just list of stages to make sure that it
@@ -131,9 +131,9 @@ def build_graph(stages, outs_trie=None):
 def build_outs_graph(graph, outs_trie):
     import networkx as nx
 
-    G = nx.DiGraph()
+    outs_graph = nx.DiGraph()
 
-    G.add_nodes_from(outs_trie.values())
+    outs_graph.add_nodes_from(outs_trie.values())
     for stage in graph.nodes():
         for dep in stage.deps:
             if dep.fs_path is None:
@@ -145,5 +145,5 @@ def build_outs_graph(graph, outs_trie):
                 overlapping.extend(outs_trie.values(prefix=dep_key))
 
             for from_out in stage.outs:
-                G.add_edges_from((from_out, out) for out in overlapping)
-    return G
+                outs_graph.add_edges_from((from_out, out) for out in overlapping)
+    return outs_graph

dvc/repo/imp_url.py

@@ -14,7 +14,7 @@ from . import locked
 
 @locked
 @scm_context
-def imp_url(  # noqa: C901
+def imp_url(  # noqa: C901, PLR0913
     self: "Repo",
     url,
     out=None,

dvc/repo/imports.py

@@ -1,7 +1,7 @@
 import logging
 import os
 from tempfile import TemporaryDirectory
-from typing import TYPE_CHECKING, List, Set, Tuple, Union, cast
+from typing import TYPE_CHECKING, List, Set, Tuple, Union
 
 if TYPE_CHECKING:
     from dvc.dependency.base import Dependency
@@ -9,7 +9,8 @@ if TYPE_CHECKING:
     from dvc.repo.index import Index, IndexView
     from dvc.stage import Stage
     from dvc.types import TargetType
-    from dvc_data.hashfile import HashInfo
+    from dvc_data.hashfile.hash_info import HashInfo
+
 
 logger = logging.getLogger(__name__)
 
@@ -63,6 +64,8 @@ def unpartial_imports(index: Union["Index", "IndexView"]) -> int:
     Returns:
         Total number of files which were unpartialed.
     """
+    from dvc_data.hashfile.hash_info import HashInfo
+
     updated = 0
     for out in index.outs:
         # we need to use view[key] here and since the out fields have not been
@@ -75,7 +78,8 @@ def unpartial_imports(index: Union["Index", "IndexView"]) -> int:
                 out.hash_info = dep.get_obj().hash_info
                 out.meta = dep.get_meta()
             else:
-                out.hash_info = cast("HashInfo", entry.hash_info)
+                assert isinstance(entry.hash_info, HashInfo)
+                out.hash_info = entry.hash_info
                 out.meta = entry.meta
             out.stage.md5 = out.stage.compute_md5()
             out.stage.dump()
@@ -115,7 +119,7 @@ def save_imports(
                 desc="Downloading imports from source",
                 unit="files",
             ) as cb:
-                checkout(data_view, tmpdir, cache.fs, callback=cb)
+                checkout(data_view, tmpdir, cache.fs, callback=cb, storage="data")
             md5(data_view)
             save(data_view, odb=cache, hardlink=True)
 

dvc/repo/index.py

@@ -112,6 +112,60 @@ def collect_files(
         dirs[:] = [d for d in dirs if not is_out_or_ignored(root, d)]
 
 
+def _load_data_from_outs(index, prefix, outs):
+    from dvc_data.index import DataIndexEntry
+
+    for out in outs:
+        if not out.use_cache:
+            continue
+
+        ws, key = out.index_key
+
+        entry = DataIndexEntry(
+            key=key,
+            meta=out.meta,
+            hash_info=out.hash_info,
+        )
+
+        if out.stage.is_import and not out.stage.is_repo_import:
+            dep = out.stage.deps[0]
+            entry.meta = dep.meta
+            entry.hash_info = dep.hash_info
+
+        # FIXME PyGTrie-based DataIndex doesn't remove entry.key during
+        # index.add, so we have to set the entry manually here to make
+        # index.view() work correctly.
+        index[(*prefix, ws, *key)] = entry
+
+
+def _load_storage_from_out(storage_map, key, out):
+    from dvc.config import NoRemoteError
+    from dvc_data.index import FileStorage, ObjectStorage
+
+    if out.odb:
+        storage_map.add_data(ObjectStorage(key, out.odb))
+    storage_map.add_cache(ObjectStorage(key, out.cache))
+    try:
+        remote = out.repo.cloud.get_remote(out.remote)
+        if remote.fs.version_aware:
+            storage_map.add_remote(
+                FileStorage(
+                    key=key,
+                    fs=remote.fs,
+                    path=remote.fs.path.join(remote.path, *key),
+                    index=remote.index,
+                )
+            )
+        else:
+            storage_map.add_remote(ObjectStorage(key, remote.odb, index=remote.index))
+    except NoRemoteError:
+        pass
+
+    if out.stage.is_import and not out.stage.is_repo_import:
+        dep = out.stage.deps[0]
+        storage_map.add_data(FileStorage(key, dep.fs, dep.fs_path))
+
+
 class Index:
     def __init__(
         self,
@@ -281,51 +335,53 @@ class Index:
 
         return dict(by_workspace)
 
+    @cached_property
+    def data_tree(self):
+        from dvc_data.hashfile.tree import Tree
+
+        tree = Tree()
+        for out in self.outs:
+            if not out.use_cache:
+                continue
+
+            ws, key = out.index_key
+
+            tree.add((ws, *key), out.meta, out.hash_info)
+
+        tree.digest()
+
+        return tree
+
     @cached_property
     def data(self) -> "Dict[str, DataIndex]":
-        from collections import defaultdict
+        prefix: "DataIndexKey"
+        loaded = False
 
-        from dvc.config import NoRemoteError
-        from dvc_data.index import DataIndex, DataIndexEntry, Storage
+        index = self.repo.data_index
+        prefix = ("tree", self.data_tree.hash_info.value)
+        if index.has_node(prefix):
+            loaded = True
 
-        by_workspace: dict = defaultdict(DataIndex)
+        if not loaded:
+            _load_data_from_outs(index, prefix, self.outs)
+            index.commit()
 
-        by_workspace["repo"] = DataIndex()
-        by_workspace["local"] = DataIndex()
+        by_workspace = {}
+        by_workspace["repo"] = index.view((*prefix, "repo"))
+        by_workspace["local"] = index.view((*prefix, "local"))
 
         for out in self.outs:
             if not out.use_cache:
                 continue
 
-            workspace, key = out.index_key
-            data_index = by_workspace[workspace]
+            ws, key = out.index_key
+            if ws not in by_workspace:
+                by_workspace[ws] = index.view((*prefix, ws))
 
-            entry = DataIndexEntry(
-                key=key,
-                meta=out.meta,
-                hash_info=out.hash_info,
-            )
+            data_index = by_workspace[ws]
+            _load_storage_from_out(data_index.storage_map, key, out)
 
-            storage = Storage(odb=out.odb, cache=out.cache)
-            try:
-                storage.remote = self.repo.cloud.get_remote_odb(out.remote)
-            except NoRemoteError:
-                pass
-
-            if out.stage.is_import and not out.stage.is_repo_import:
-                dep = out.stage.deps[0]
-
-                entry.meta = dep.meta
-                entry.hash_info = dep.hash_info
-
-                storage.odb = dep.odb
-                storage.fs = dep.fs
-                storage.path = dep.fs_path
-
-            data_index.add(entry)
-            data_index.storage_map[key] = storage
-
-        return dict(by_workspace)
+        return by_workspace
 
     @staticmethod
     def _hash_targets(
@@ -526,6 +582,10 @@ def build_data_index(
     from dvc_data.index.build import build_entries, build_entry
     from dvc_data.index.save import build_tree
 
+    ignore = None
+    if workspace == "repo" and isinstance(fs, LocalFileSystem):
+        ignore = index.repo.dvcignore
+
     data = DataIndex()
     for key in index.data_keys.get(workspace, set()):
         out_path = fs.path.join(path, *key)
@@ -547,7 +607,11 @@ def build_data_index(
             continue
 
         for entry in build_entries(
-            out_path, fs, compute_hash=compute_hash, state=index.repo.state
+            out_path,
+            fs,
+            compute_hash=compute_hash,
+            state=index.repo.state,
+            ignore=ignore,
         ):
             if not entry.key or entry.key == ("",):
                 # NOTE: whether the root will be returned by build_entries

dvc/repo/metrics/show.py

@@ -143,7 +143,7 @@ def show(
         except (SCMError, NoSCMError):
             # SCMError - detached head
             # NoSCMError - no repo case
-            pass  # noqa: S110
+            pass
         else:
             if res.get("workspace") == res.get(active_branch):
                 res.pop("workspace", None)

dvc/repo/move.py

@@ -32,11 +32,11 @@ def move(self, from_path, to_path):
     It only works with outputs generated by `add` or `import`,
     also known as data sources.
     """
-    import dvc.output as Output
+    from dvc import output
     from dvc.dvcfile import DVC_FILE_SUFFIX
     from dvc.stage import Stage
 
-    from_out = Output.loads_from(Stage(self), [from_path])[0]
+    from_out = output.loads_from(Stage(self), [from_path])[0]
     assert from_out.protocol == "local"
 
     to_path = _expand_target_path(from_path, to_path)
@@ -71,7 +71,7 @@ def move(self, from_path, to_path):
     else:
         to_path = os.path.relpath(to_path, stage.wdir)
 
-    to_out = Output.loads_from(stage, [to_path], out.use_cache, out.metric)[0]
+    to_out = output.loads_from(stage, [to_path], out.use_cache, out.metric)[0]
 
     out.move(to_out)
     stage.save()

dvc/repo/params/show.py

@@ -153,7 +153,7 @@ def show(
         except (SCMError, NoSCMError):
             # SCMError - detached head
             # NoSCMError - no repo case
-            pass  # noqa: S110
+            pass
         else:
             if res.get("workspace") == res.get(active_branch):
                 res.pop("workspace", None)

dvc/repo/plots/__init__.py

@@ -2,7 +2,7 @@ import csv
 import io
 import logging
 import os
-from collections import OrderedDict, defaultdict
+from collections import defaultdict
 from copy import deepcopy
 from functools import partial
 from multiprocessing import cpu_count
@@ -18,8 +18,8 @@ from typing import (
     Union,
 )
 
+import dpath
 import dpath.options
-import dpath.util
 from funcy import distinct, first, project
 
 from dvc.exceptions import DvcException
@@ -74,7 +74,6 @@ class Plots:
         recursive: bool = False,
         onerror: Optional[Callable] = None,
         props: Optional[Dict] = None,
-        config_files: Optional[Set[str]] = None,
     ) -> Iterator[Dict]:
         """Collects plots definitions and data sources.
 
@@ -131,7 +130,6 @@ class Plots:
                 targets=targets,
                 revision=rev,
                 onerror=onerror,
-                config_files=config_files,
                 props=props,
             )
             if definitions:
@@ -186,7 +184,6 @@ class Plots:
         props=None,
         recursive=False,
         onerror=None,
-        config_files: Optional[Set[str]] = None,
     ):
         if onerror is None:
             onerror = onerror_collect
@@ -198,7 +195,6 @@ class Plots:
             recursive,
             onerror=onerror,
             props=props,
-            config_files=config_files,
         ):
             _resolve_data_sources(data)
             result.update(data)
@@ -389,7 +385,7 @@ def _collect_output_plots(repo, targets, props, onerror: Optional[Callable] = No
                 onerror=onerror,
             )
 
-            dpath.util.merge(
+            dpath.merge(
                 result,
                 {"": unpacked},
             )
@@ -436,16 +432,13 @@ def _resolve_definitions(
                     props={**plot_props, **props},
                     onerror=onerror,
                 )
-                dpath.util.merge(
+                dpath.merge(
                     result,
                     unpacked,
                 )
-        else:
-            if _matches(targets, config_path, plot_id):
-                adjusted_props = _adjust_sources(fs, plot_props, config_dir)
-                dpath.util.merge(
-                    result, {"data": {plot_id: {**adjusted_props, **props}}}
-                )
+        elif _matches(targets, config_path, plot_id):
+            adjusted_props = _adjust_sources(fs, plot_props, config_dir)
+            dpath.merge(result, {"data": {plot_id: {**adjusted_props, **props}}})
 
     return result
 
@@ -474,7 +467,7 @@ def _collect_pipeline_files(repo, targets: List[str], props, onerror=None):
             dvcfile_defs_dict,
             onerror=onerror,
         )
-        dpath.util.merge(
+        dpath.merge(
             result,
             {dvcfile_path: resolved},
         )
@@ -485,7 +478,6 @@ def _collect_pipeline_files(repo, targets: List[str], props, onerror=None):
 def _collect_definitions(
     repo: "Repo",
     targets=None,
-    config_files: Optional[Set[str]] = None,
     props: Optional[Dict] = None,
     onerror: Optional[Callable] = None,
     **kwargs,
@@ -497,39 +489,20 @@ def _collect_definitions(
 
     fs = DVCFileSystem(repo=repo)
 
-    if not config_files:
-        dpath.util.merge(
-            result,
-            _collect_pipeline_files(repo, targets, props, onerror=onerror),
-        )
+    dpath.merge(
+        result,
+        _collect_pipeline_files(repo, targets, props, onerror=onerror),
+    )
 
-    if targets or (not targets and not config_files):
-        dpath.util.merge(
-            result,
-            _collect_output_plots(repo, targets, props, onerror=onerror),
-        )
-
-    if config_files:
-        for path in config_files:
-            definitions = parse(fs, path).get("data", {})
-            if definitions:
-                resolved = _resolve_definitions(
-                    repo.dvcfs,
-                    targets,
-                    props,
-                    path,
-                    definitions,
-                    onerror=onerror,
-                )
-                dpath.util.merge(
-                    result,
-                    {path: resolved},
-                )
+    dpath.merge(
+        result,
+        _collect_output_plots(repo, targets, props, onerror=onerror),
+    )
 
     for target in targets:
         if not result or fs.exists(target):
             unpacked = unpack_if_dir(fs, target, props=props, onerror=onerror)
-            dpath.util.merge(result[""], unpacked)
+            dpath.merge(result[""], unpacked)
 
     return dict(result)
 
@@ -584,21 +557,14 @@ def _load_sv(path, fs, delimiter=",", header=True):
     with fs.open(path, "r") as fd:
         content = fd.read()
 
-    first_row = first(csv.reader(io.StringIO(content)))
-
     if header:
         reader = csv.DictReader(io.StringIO(content), delimiter=delimiter)
     else:
+        first_row = first(csv.reader(io.StringIO(content)))
         reader = csv.DictReader(
             io.StringIO(content),
             delimiter=delimiter,
             fieldnames=[str(i) for i in range(len(first_row))],
         )
 
-    fieldnames = reader.fieldnames or []
-    data = list(reader)
-
-    return [
-        OrderedDict([(field, data_point[field]) for field in fieldnames])
-        for data_point in data
-    ]
+    return list(reader)

dvc/repo/pull.py

@@ -11,7 +11,7 @@ logger = logging.getLogger(__name__)
 
 
 @locked
-def pull(
+def pull(  # noqa: PLR0913
     self,
     targets=None,
     jobs=None,
@@ -25,6 +25,7 @@ def pull(
     run_cache=False,
     glob=False,
     odb: Optional["ObjectDB"] = None,
+    allow_missing=False,
 ):
     if isinstance(targets, str):
         targets = [targets]
@@ -48,6 +49,7 @@ def pull(
         with_deps=with_deps,
         force=force,
         recursive=recursive,
+        allow_missing=allow_missing,
     )
 
     stats["fetched"] = processed_files_count

dvc/repo/push.py

@@ -1,5 +1,7 @@
+from contextlib import suppress
 from typing import TYPE_CHECKING, Optional, Sequence
 
+from dvc.config import NoRemoteError
 from dvc.exceptions import InvalidArgumentError, UploadError
 from dvc.utils import glob_targets
 
@@ -13,7 +15,7 @@ if TYPE_CHECKING:
 
 
 @locked
-def push(  # noqa: C901
+def push(  # noqa: C901, PLR0913
     self,
     targets=None,
     jobs=None,
@@ -30,9 +32,10 @@ def push(  # noqa: C901
     include_imports=False,
 ):
     worktree_remote: Optional["Remote"] = None
-    _remote = self.cloud.get_remote(name=remote)
-    if _remote.worktree or _remote.fs.version_aware:
-        worktree_remote = _remote
+    with suppress(NoRemoteError):
+        _remote = self.cloud.get_remote(name=remote)
+        if _remote and (_remote.worktree or _remote.fs.version_aware):
+            worktree_remote = _remote
 
     pushed = 0
     used_run_cache = self.stage_cache.push(remote, odb=odb) if run_cache else []

dvc/repo/reproduce.py

@@ -129,7 +129,7 @@ def reproduce(  # noqa: C901
 
 
 def _reproduce_stages(  # noqa: C901
-    G, stages, downstream=False, single_item=False, on_unchanged=None, **kwargs
+    graph, stages, downstream=False, single_item=False, on_unchanged=None, **kwargs
 ):
     r"""Derive the evaluation of the given node for the given graph.
 
@@ -166,7 +166,7 @@ def _reproduce_stages(  # noqa: C901
 
     The derived evaluation of _downstream_ B would be: [B, D, E]
     """
-    steps = _get_steps(G, stages, downstream, single_item)
+    steps = _get_steps(graph, stages, downstream, single_item)
 
     force_downstream = kwargs.pop("force_downstream", False)
     result = []
@@ -222,16 +222,16 @@ def _reproduce_stages(  # noqa: C901
     return result
 
 
-def _get_steps(G, stages, downstream, single_item):
+def _get_steps(graph, stages, downstream, single_item):
     import networkx as nx
 
-    active = G.copy()
+    active = graph.copy()
     if not single_item:
         # NOTE: frozen stages don't matter for single_item
-        for stage in G:
+        for stage in graph:
             if stage.frozen:
                 # NOTE: disconnect frozen stage from its dependencies
-                active.remove_edges_from(G.out_edges(stage))
+                active.remove_edges_from(graph.out_edges(stage))
 
     all_pipelines: List["Stage"] = []
     for stage in stages:

dvc/repo/worktree.py

@@ -1,6 +1,6 @@
 import logging
 from functools import partial
-from typing import TYPE_CHECKING, Any, Dict, Iterable, Optional, Set, Tuple, Union, cast
+from typing import TYPE_CHECKING, Any, Dict, Iterable, Optional, Set, Tuple, Union
 
 from funcy import first
 
@@ -17,7 +17,6 @@ if TYPE_CHECKING:
     from dvc.stage import Stage
     from dvc.types import TargetType
     from dvc_data.hashfile.meta import Meta
-    from dvc_data.hashfile.tree import Tree
     from dvc_data.index import DataIndex, DataIndexView
     from dvc_objects.fs.base import FileSystem
 
@@ -29,7 +28,8 @@ logger = logging.getLogger(__name__)
 def _meta_checksum(fs: "FileSystem", meta: "Meta") -> Any:
     if not meta or meta.isdir:
         return meta
-    return getattr(meta, cast(str, fs.PARAM_CHECKSUM))
+    assert fs.PARAM_CHECKSUM
+    return getattr(meta, fs.PARAM_CHECKSUM)
 
 
 def worktree_view_by_remotes(
@@ -119,12 +119,6 @@ def fetch_worktree(
     ):
         remote_obj = _get_remote(repo, remote_name, remote, "fetch")
         index = view.data["repo"]
-        for key in index.storage_map:
-            storage = index.storage_map[key]
-            storage.fs = remote_obj.fs
-            storage.path = remote_obj.fs.path.join(remote_obj.path, *key)
-            index.storage_map[key] = storage
-
         total = len(index)
         with Callback.as_tqdm_callback(
             unit="file",
@@ -132,7 +126,7 @@ def fetch_worktree(
             disable=total == 0,
         ) as cb:
             cb.set_size(total)
-            transferred += save(index, callback=cb, jobs=jobs)
+            transferred += save(index, callback=cb, jobs=jobs, storage="remote")
     return transferred
 
 
@@ -196,6 +190,11 @@ def push_worktree(
                     f"remote {remote_obj.name!r} does not support versioning"
                 ) from None
 
+        if remote_obj.index is not None:
+            for key, entry in new_index.iteritems():
+                remote_obj.index[key] = entry
+            remote_obj.index.commit()
+
         for out in view.outs:
             workspace, _key = out.index_key
             _merge_push_meta(out, repo.index.data[workspace], remote_obj.name)
@@ -217,13 +216,15 @@ def _merge_push_meta(
     conflicts (i.e. the DVC output's version ID may not match the pushed/latest
     version ID as long when the file content of both versions is the same).
     """
+    from dvc_data.hashfile.tree import Tree
     from dvc_data.index.save import build_tree
 
     _, key = out.index_key
     entry = index[key]
     repo = out.stage.repo
     if out.isdir():
-        old_tree = cast("Tree", out.get_obj())
+        old_tree = out.get_obj()
+        assert isinstance(old_tree, Tree)
         entry.hash_info = old_tree.hash_info
         entry.meta = out.meta
         for subkey, entry in index.iteritems(key):
@@ -357,6 +358,7 @@ def _fetch_out_changes(
             update_meta=False,
             meta_only=True,
             meta_cmp_key=partial(_meta_checksum, remote.fs),
+            storage="data",
             callback=cb,
         )
         out.save()
@@ -419,6 +421,7 @@ def _get_update_diff_index(
     remote_index: Union["DataIndex", "DataIndexView"],
     remote: "Remote",
 ) -> "DataIndex":
+    from dvc_data.hashfile.tree import Tree
     from dvc_data.index import DataIndex
     from dvc_data.index.diff import ADD, MODIFY, UNCHANGED, diff
 
@@ -438,7 +441,8 @@ def _get_update_diff_index(
             if out.isdir():
                 if not entry.meta.isdir:
                     fs_path = repo.fs.path.join(repo.root_dir, *entry.key)
-                    tree = cast("Tree", out.obj)
+                    tree = out.obj
+                    assert isinstance(tree, Tree)
                     _, entry.hash_info = tree.get(  # type: ignore[misc]
                         repo.fs.path.relparts(fs_path, out.fs_path)
                     )

dvc/rwlock.py

@@ -74,7 +74,7 @@ def _infos_to_str(infos):
     )
 
 
-def _check_blockers(tmp_dir, lock, info, *, mode, waiters):  # noqa: C901
+def _check_blockers(tmp_dir, lock, info, *, mode, waiters):  # noqa: C901, PLR0912
     from .lock import LockError
 
     non_existing_pid = set()

dvc/scm.py

@@ -20,7 +20,6 @@ from scmrepo.noscm import NoSCM
 
 from dvc.exceptions import DvcException
 from dvc.progress import Tqdm
-from dvc.utils import format_link
 
 if TYPE_CHECKING:
     from scmrepo.progress import GitProgressEvent
@@ -58,25 +57,6 @@ class GitAuthError(SCMError):
         super().__init__(f"{reason}\n{doc}")
 
 
-class GitMergeError(SCMError):
-    def __init__(self, msg: str, scm: Optional["Git"] = None) -> None:
-        if scm and self._is_shallow(scm):
-            url = format_link(
-                "https://dvc.org/doc/user-guide/troubleshooting#git-shallow"
-            )
-            msg = (
-                f"{msg}: `dvc exp` does not work in shallow Git repos. "
-                f"See {url} for more information."
-            )
-        super().__init__(msg)
-
-    @staticmethod
-    def _is_shallow(scm: "Git") -> bool:
-        if os.path.exists(os.path.join(scm.root_dir, Git.GIT_DIR, "shallow")):
-            return True
-        return False
-
-
 @contextmanager
 def map_scm_exception(with_cause: bool = False) -> Iterator[None]:
     from scmrepo.exceptions import SCMError as InternalSCMError
@@ -185,7 +165,7 @@ def clone(url: str, to_path: str, **kwargs):
             raise CloneError("SCM error") from exc
 
 
-def resolve_rev(scm: "Git", rev: str) -> str:
+def resolve_rev(scm: Union["Git", "NoSCM"], rev: str) -> str:
     from scmrepo.exceptions import RevError as InternalRevError
 
     from dvc.repo.experiments.utils import fix_exp_head
@@ -193,6 +173,7 @@ def resolve_rev(scm: "Git", rev: str) -> str:
     try:
         return scm.resolve_rev(fix_exp_head(scm, rev))
     except InternalRevError as exc:
+        assert isinstance(scm, Git)
         # `scm` will only resolve git branch and tag names,
         # if rev is not a sha it may be an abbreviated experiment name
         if not (rev == "HEAD" or rev.startswith("refs/")):
@@ -227,7 +208,7 @@ def _get_n_commits(scm: "Git", revs: List[str], num: int) -> List[str]:
     return results
 
 
-def iter_revs(  # noqa: C901
+def iter_revs(
     scm: "Git",
     revs: Optional[List[str]] = None,
     num: int = 1,

dvc/stage/__init__.py

@@ -140,7 +140,7 @@ class Stage(params.StageParams):
     # pylint:disable=no-value-for-parameter
     # rwlocked() confuses pylint
 
-    def __init__(
+    def __init__(  # noqa: PLR0913
         self,
         repo,
         path=None,
@@ -545,13 +545,13 @@ class Stage(params.StageParams):
         )
 
     @rwlocked(write=["outs"])
-    def commit(self, allow_missing=False, filter_info=None) -> None:
+    def commit(self, allow_missing=False, filter_info=None, **kwargs) -> None:
         from dvc.output import OutputDoesNotExistError
 
         link_failures = []
         for out in self.filter_outs(filter_info):
             try:
-                out.commit(filter_info=filter_info)
+                out.commit(filter_info=filter_info, **kwargs)
             except OutputDoesNotExistError:
                 if not (allow_missing or out.checkpoint):
                     raise

dvc/stage/cache.py

@@ -1,6 +1,5 @@
 import logging
 import os
-import tempfile
 from contextlib import contextmanager
 from typing import TYPE_CHECKING, List, Optional, Tuple
 
@@ -167,9 +166,10 @@ class StageCache:
         COMPILED_LOCK_FILE_STAGE_SCHEMA(cache)
 
         path = self._get_cache_path(cache_key, cache_value)
-        parent = self.repo.cache.local.fs.path.parent(path)
+        local_fs = self.repo.cache.local.fs
+        parent = local_fs.path.parent(path)
         self.repo.cache.local.makedirs(parent)
-        tmp = tempfile.NamedTemporaryFile(delete=False, dir=parent).name
+        tmp = local_fs.path.join(parent, fs.utils.tmp_fname())
         assert os.path.exists(parent)
         assert os.path.isdir(parent)
         dump_yaml(tmp, cache)

dvc/stage/monitor.py

@@ -29,7 +29,7 @@ class MonitorTask:
         raise NotImplementedError
 
     @property
-    def SIGNAL_FILE(self) -> str:
+    def SIGNAL_FILE(self) -> str:  # noqa: N802
         raise NotImplementedError
 
     @property

dvc/stage/serialize.py

@@ -8,7 +8,6 @@ from typing import (
     List,
     Optional,
     Union,
-    cast,
     no_type_check,
 )
 
@@ -24,7 +23,6 @@ from .utils import resolve_wdir, split_params_deps
 
 if TYPE_CHECKING:
     from dvc.stage import PipelineStage, Stage
-    from dvc_data.hashfile.tree import Tree
 
 PARAM_PARAMS = ParamsDependency.PARAM_PARAMS
 PARAM_PATH = ParamsDependency.PARAM_PATH
@@ -155,6 +153,7 @@ def to_pipeline_file(stage: "PipelineStage"):
 
 def to_single_stage_lockfile(stage: "Stage", **kwargs) -> dict:
     from dvc.output import _serialize_tree_obj_to_files, split_file_meta_from_cloud
+    from dvc_data.hashfile.tree import Tree
 
     assert stage.cmd
 
@@ -163,7 +162,7 @@ def to_single_stage_lockfile(stage: "Stage", **kwargs) -> dict:
         if item.hash_info.isdir and kwargs.get("with_files"):
             obj = item.obj or item.get_obj()
             if obj:
-                obj = cast("Tree", obj)
+                assert isinstance(obj, Tree)
                 ret[item.PARAM_FILES] = [
                     split_file_meta_from_cloud(f)
                     for f in _serialize_tree_obj_to_files(obj)

dvc/testing/fixtures.py

@@ -26,8 +26,6 @@ __all__ = [
     "workspace",
     "make_workspace",
     "local_workspace",
-    "docker",
-    "docker_compose",
     "docker_compose_project_name",
     "docker_services",
 ]
@@ -215,48 +213,47 @@ def local_workspace(make_workspace):
     return make_workspace("workspace", typ="local")
 
 
-@pytest.fixture(scope="session")
-def docker():
-    # See https://travis-ci.community/t/docker-linux-containers-on-windows/301
-    if os.environ.get("CI") and os.name == "nt":
-        pytest.skip("disabled for Windows on Github Actions")
-
-    try:
-        subprocess.check_output("docker ps", shell=True)
-    except (subprocess.CalledProcessError, OSError):
-        pytest.skip("no docker installed")
-
-
-@pytest.fixture(scope="session")
-def docker_compose(docker):  # noqa: ARG001
-    try:
-        subprocess.check_output("docker-compose version", shell=True)
-    except (subprocess.CalledProcessError, OSError):
-        pytest.skip("no docker-compose installed")
-
-
 @pytest.fixture(scope="session")
 def docker_compose_project_name():
     return "pytest-dvc-test"
 
 
 @pytest.fixture(scope="session")
-def docker_services(docker_compose_file, docker_compose_project_name, tmp_path_factory):
-    # overriding `docker_services` fixture from `pytest_docker` plugin to
-    # only launch docker images once.
-
+def docker_services(
+    tmp_path_factory,
+    docker_compose_command,
+    docker_compose_file,
+    docker_compose_project_name,
+    docker_setup,
+):
     from filelock import FileLock
-
-    # pylint: disable-next=import-error
     from pytest_docker.plugin import DockerComposeExecutor, Services
 
-    executor = DockerComposeExecutor(docker_compose_file, docker_compose_project_name)
+    if os.environ.get("CI") and os.name == "nt":
+        pytest.skip("disabled for Windows on CI")
+
+    try:
+        subprocess.check_output("docker ps", stderr=subprocess.STDOUT, shell=True)
+    except subprocess.CalledProcessError as err:
+        out = (err.output or b"").decode("utf-8")
+        pytest.skip(f"docker is not installed or the daemon is not running: {out}")
+
+    try:
+        cmd = "docker-compose version"
+        subprocess.check_output(cmd, stderr=subprocess.STDOUT, shell=True)
+    except subprocess.CalledProcessError as err:
+        out = (err.output or b"").decode("utf-8")
+        pytest.skip(f"docker-compose is not installed: {out}")
+
+    executor = DockerComposeExecutor(
+        docker_compose_command, docker_compose_file, docker_compose_project_name
+    )
 
     # making sure we don't accidentally launch docker-compose in parallel,
     # as it might result in network conflicts. Inspired by:
     # https://github.com/pytest-dev/pytest-xdist#making-session-scoped-fixtures-execute-only-once
     lockfile = tmp_path_factory.getbasetemp().parent / "docker-compose.lock"
-    with FileLock(str(lockfile)):  # pylint:disable=abstract-class-instantiated
-        executor.execute("up --build -d")
-
-    return Services(executor)
+    with FileLock(os.fspath(lockfile)):
+        executor.execute(docker_setup)
+        # note: we are not tearing down the containers here
+        return Services(executor)

dvc/testing/workspace_tests.py

@@ -282,3 +282,85 @@ class TestGetUrl:
     def test_get_url_nonexistent(self, cloud):
         with pytest.raises(URLMissingError):
             Repo.get_url(str(cloud / "nonexistent"), config=cloud.config)
+
+
+class TestToRemote:
+    def test_add_to_remote(self, tmp_dir, dvc, remote, workspace):
+        workspace.gen("foo", "foo")
+
+        url = "remote://workspace/foo"
+        [stage] = dvc.add(url, to_remote=True)
+
+        assert not (tmp_dir / "foo").exists()
+        assert (tmp_dir / "foo.dvc").exists()
+
+        assert len(stage.deps) == 0
+        assert len(stage.outs) == 1
+
+        hash_info = stage.outs[0].hash_info
+        meta = stage.outs[0].meta
+        assert hash_info.name == "md5"
+        assert hash_info.value == "acbd18db4cc2f85cedef654fccc4a4d8"
+        assert (remote / "ac" / "bd18db4cc2f85cedef654fccc4a4d8").read_text() == "foo"
+        assert meta.size == len("foo")
+
+    def test_import_url_to_remote_file(self, tmp_dir, dvc, workspace, remote):
+        workspace.gen("foo", "foo")
+
+        url = "remote://workspace/foo"
+        stage = dvc.imp_url(url, to_remote=True)
+
+        assert stage.deps[0].hash_info.value is not None
+        assert not (tmp_dir / "foo").exists()
+        assert (tmp_dir / "foo.dvc").exists()
+
+        assert len(stage.deps) == 1
+        assert stage.deps[0].def_path == url
+        assert len(stage.outs) == 1
+
+        hash_info = stage.outs[0].hash_info
+        assert hash_info.name == "md5"
+        assert hash_info.value == "acbd18db4cc2f85cedef654fccc4a4d8"
+        assert (remote / "ac" / "bd18db4cc2f85cedef654fccc4a4d8").read_text() == "foo"
+        assert stage.outs[0].meta.size == len("foo")
+
+    def test_import_url_to_remote_dir(self, tmp_dir, dvc, workspace, remote):
+        import json
+
+        workspace.gen(
+            {
+                "data": {
+                    "foo": "foo",
+                    "bar": "bar",
+                    "sub_dir": {"baz": "sub_dir/baz"},
+                }
+            }
+        )
+
+        url = "remote://workspace/data"
+        stage = dvc.imp_url(url, to_remote=True)
+
+        assert not (tmp_dir / "data").exists()
+        assert (tmp_dir / "data.dvc").exists()
+
+        assert len(stage.deps) == 1
+        assert stage.deps[0].def_path == url
+        assert len(stage.outs) == 1
+
+        hash_info = stage.outs[0].hash_info
+        assert hash_info.name == "md5"
+        assert hash_info.value == "55d05978954d1b2cd7b06aedda9b9e43.dir"
+        file_parts = json.loads(
+            (remote / "55" / "d05978954d1b2cd7b06aedda9b9e43.dir").read_text()
+        )
+
+        assert len(file_parts) == 3
+        assert {file_part["relpath"] for file_part in file_parts} == {
+            "foo",
+            "bar",
+            "sub_dir/baz",
+        }
+
+        for file_part in file_parts:
+            md5 = file_part["md5"]
+            assert (remote / md5[:2] / md5[2:]).read_text() == file_part["relpath"]

dvc/ui/__init__.py

@@ -17,7 +17,9 @@ from dvc.utils.objects import cached_property
 
 if TYPE_CHECKING:
     from rich.console import Console as RichConsole
+    from rich.console import JustifyMethod, OverflowMethod
     from rich.status import Status
+    from rich.style import Style
     from rich.text import Text as RichText
 
     from dvc.progress import Tqdm
@@ -87,7 +89,7 @@ class Console:
             styled=styled,
         )
 
-    def write_json(
+    def write_json(  # noqa: PLR0913
         self,
         data: Any,
         indent: Optional[int] = None,
@@ -105,8 +107,9 @@ class Console:
         if indent is None and self.isatty():
             indent = 2
 
-        console = self.error_console if stderr else self.rich_console
-        return console.print_json(
+        from rich.json import JSON
+
+        json = JSON.from_data(
             data=data,
             indent=indent,
             highlight=bool(highlight),
@@ -117,6 +120,50 @@ class Console:
             default=default,
             sort_keys=sort_keys,
         )
+        if not highlight:
+            return self.write(json.text, stderr=stderr)
+        return self.rich_print(json, stderr=stderr, soft_wrap=True)
+
+    def rich_print(
+        self,
+        *objects: Any,
+        sep: str = " ",
+        end: str = "\n",
+        stderr: bool = False,
+        style: Optional[Union[str, "Style"]] = None,
+        justify: Optional["JustifyMethod"] = None,
+        overflow: Optional["OverflowMethod"] = None,
+        no_wrap: Optional[bool] = None,
+        emoji: Optional[bool] = None,
+        markup: Optional[bool] = None,
+        highlight: Optional[bool] = None,
+        width: Optional[int] = None,
+        height: Optional[int] = None,
+        crop: bool = True,
+        soft_wrap: Optional[bool] = None,
+        new_line_start: bool = False,
+    ) -> None:
+        if stderr:
+            console = self.error_console
+        else:
+            console = self.rich_console
+        return console.print(
+            *objects,
+            sep=sep,
+            end=end,
+            style=style,
+            justify=justify,
+            overflow=overflow,
+            no_wrap=no_wrap,
+            emoji=emoji,
+            markup=markup,
+            highlight=highlight,
+            width=width,
+            height=height,
+            crop=crop,
+            soft_wrap=soft_wrap,
+            new_line_start=new_line_start,
+        )
 
     def write(
         self,
@@ -142,10 +189,21 @@ class Console:
         with Tqdm.external_write_mode(file=file):
             # if we are inside pager context, send the output to rich's buffer
             if styled or self._paginate.get():
-                console = self.error_console if stderr else self.rich_console
                 if styled:
-                    return console.print(*objects, sep=sep, end=end)
-                return console.out(*objects, sep=sep, end=end, highlight=False)
+                    return self.rich_print(*objects, sep=sep, end=end, stderr=stderr)
+                return self.rich_print(
+                    sep.join(str(_object) for _object in objects),
+                    style=None,
+                    highlight=False,
+                    emoji=False,
+                    markup=False,
+                    no_wrap=True,
+                    overflow="ignore",
+                    crop=False,
+                    sep=sep,
+                    end=end,
+                    stderr=stderr,
+                )
 
             values = (self.formatter.format(obj, style) for obj in objects)
             return print(*values, sep=sep, end=end, file=file)

dvc/utils/collections.py

@@ -14,8 +14,8 @@ def apply_diff(src, dest):  # noqa: C901
 
     Used in Stage load/dump cycle to preserve comments and custom formatting.
     """
-    Seq = (list, tuple)
-    Container = (Mapping, list, tuple)
+    Seq = (list, tuple)  # noqa: N806
+    Container = (Mapping, list, tuple)  # noqa: N806
 
     def is_same_type(a, b):
         return any(

dvc/utils/serialize/_common.py

@@ -11,7 +11,6 @@ from typing import (
     Protocol,
     TextIO,
     Union,
-    cast,
 )
 
 from funcy import reraise
@@ -97,8 +96,7 @@ def _modify_data(
     dumper: DumpersFn,
     fs: Optional["FileSystem"] = None,
 ):
-    exists_fn = cast(Callable[["StrPath"], bool], fs.exists if fs else os.path.exists)
-    file_exists = exists_fn(cast(str, path))
+    file_exists = fs.exists(os.fspath(path)) if fs else os.path.exists(path)
     data = _load_data(path, parser=parser, fs=fs) if file_exists else {}
     yield data
     _dump_data(path, data, dumper=dumper, fs=fs)