deps: bumps dvc-azure>=2.21.2

Closes #9506 via https://github.com/iterative/dvc-azure/pull/48

(cherry picked from commit a1b891ce01)

.github/mergify.yml

@@ -0,0 +1,9 @@
+pull_request_rules:
+  - name: backport patches to 2.x
+    conditions:
+      - base=main
+      - label=backport
+    actions:
+      backport:
+        branches:
+          - "2.x"

.github/workflows/codeql.yml

@@ -2,7 +2,7 @@ name: "CodeQL"
 
 on:
   push:
-    branches: ["main"]
+    branches: ["main", "2.x"]
   pull_request:
     branches: ["main"]
   schedule:

.github/workflows/packages.yaml

@@ -73,6 +73,8 @@ jobs:
 
   pip:
     runs-on: ubuntu-20.04
+    permissions:
+      id-token: write
     steps:
     - uses: actions/checkout@v3
       with:
@@ -100,13 +102,10 @@ jobs:
     - name: Publish packages to PyPI
       if: github.event_name == 'release'
       uses: pypa/gh-action-pypi-publish@release/v1
-      with:
-        password: ${{ secrets.PYPI_TOKEN }}
 
     - name: Publish to Test PyPI
       if: ${{ github.event_name == 'release' || (github.event_name == 'push' && github.ref == 'refs/heads/main') }}
       uses: pypa/gh-action-pypi-publish@release/v1
       with:
-        password: ${{ secrets.TEST_PYPI_TOKEN }}
         repository_url: https://test.pypi.org/legacy/
         skip_existing: true

.github/workflows/tests.yaml

@@ -2,7 +2,7 @@ name: Tests
 
 on:
   push:
-    branches: [main]
+    branches: [main, 2.x]
   pull_request:
   schedule:
     - cron: '5 1 * * *'  # every day at 01:05
@@ -42,6 +42,8 @@ jobs:
     - uses: pre-commit/[email protected]
     - name: run tests
       timeout-minutes: 40
+      env:
+        PYTHONUTF8: 1
       run: >
         pytest -n=logical --timeout=300 --durations=100
         --cov --cov-report=xml --cov-report=term

.pre-commit-config.yaml

@@ -43,7 +43,7 @@ repos:
     hooks:
       - id: isort
   - repo: https://github.com/charliermarsh/ruff-pre-commit
-    rev: 'v0.0.260'
+    rev: 'v0.0.261'
     hooks:
       - id: ruff
   - repo: https://github.com/codespell-project/codespell

dvc/annotations.py

@@ -28,9 +28,21 @@ class Annotation:
 
 
 @dataclass
-class Artifact(Annotation):
+class Artifact:
     PARAM_PATH: ClassVar[str] = "path"
-    path: Optional[str] = None
+    PARAM_DESC: ClassVar[str] = "desc"
+    PARAM_TYPE: ClassVar[str] = "type"
+    PARAM_LABELS: ClassVar[str] = "labels"
+    PARAM_META: ClassVar[str] = "meta"
+
+    path: str
+    desc: Optional[str] = None
+    type: Optional[str] = None  # noqa: A003
+    labels: List[str] = field(default_factory=list)
+    meta: Dict[str, Any] = field(default_factory=dict)
+
+    def to_dict(self) -> Dict[str, str]:
+        return compact(asdict(self))
 
 
 ANNOTATION_FIELDS = [field.name for field in fields(Annotation)]

dvc/api/__init__.py

@@ -2,11 +2,16 @@ 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 exp_save, make_checkpoint
+from .experiments import exp_save, exp_show, make_checkpoint
+from .scm import all_branches, all_commits, all_tags
 from .show import metrics_show, params_show
 
 __all__ = [
+    "all_branches",
+    "all_commits",
+    "all_tags",
     "exp_save",
+    "exp_show",
     "get_url",
     "make_checkpoint",
     "open",

dvc/api/experiments.py

@@ -1,10 +1,13 @@
 import builtins
 import os
 from time import sleep
-from typing import List, Optional
+from typing import Dict, List, Optional, Union
+
+from rich.text import Text
 
 from dvc.env import DVC_CHECKPOINT, DVC_ROOT
 from dvc.repo import Repo
+from dvc.repo.experiments.show import tabulate
 from dvc.stage.monitor import CheckpointTask
 
 
@@ -69,3 +72,77 @@ def exp_save(
         return repo.experiments.save(
             name=name, force=force, include_untracked=include_untracked
         )
+
+
+def _postprocess(exp_rows):
+    for exp_row in exp_rows:
+        for k, v in exp_row.items():
+            if isinstance(v, Text):
+                v_str = str(v)
+                try:
+                    exp_row[k] = float(v_str)
+                except ValueError:
+                    exp_row[k] = v_str
+
+            if not exp_row[k]:
+                exp_row[k] = None
+
+    return exp_rows
+
+
+def exp_show(
+    repo: Optional[str] = None,
+    revs: Optional[Union[str, List[str]]] = None,
+    num: int = 1,
+    param_deps: bool = False,
+    force: bool = False,
+) -> List[Dict]:
+    """Get DVC experiments tracked in `repo`.
+
+    Without arguments, this function will retrieve all experiments derived from
+    the Git `HEAD`.
+
+    See the options below to customize the experiments retrieved.
+
+    Args:
+        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).
+        revs (Union[str, List[str]], optional): Git revision(s) (e.g. branch,
+            tag, SHA commit) to use as a reference point to start listing
+            experiments.
+            Defaults to `None`, which will use `HEAD` as starting point.
+        num (int, optional): show experiments from the last `num` commits
+            (first parents) starting from the `revs` baseline.
+            Give a negative value to include all first-parent commits (similar
+            to `git log -n`).
+            Defaults to 1.
+        param_deps (bool, optional): include only parameters that are stage
+            dependencies.
+            Defaults to `False`.
+        force (bool, optional): force re-collection of experiments instead of
+            loading from internal experiments cache.
+            DVC caches `exp_show` data for completed experiments to improve
+            performance of subsequent calls.
+            When `force` is specified, DVC will reload all experiment data and
+            ignore any previously cached results.
+            Defaults to `False`.
+
+    Returns:
+        List[Dict]: Each item in the list will contain a dictionary with
+            the info for an individual experiment.
+            See Examples below.
+    """
+    with Repo.open(repo) as _repo:
+        experiments = _repo.experiments.show(
+            revs=revs,
+            num=num,
+            param_deps=param_deps,
+            force=force,
+        )
+        td, _ = tabulate(experiments, fill_value=None)
+
+        return _postprocess(td.as_dict())

dvc/api/scm.py

@@ -0,0 +1,54 @@
+from typing import List, Optional
+
+from dvc.repo import Repo
+
+
+def all_branches(repo: Optional[str] = None) -> List[str]:
+    """Get all Git branches in a DVC repository.
+
+    Args:
+        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).
+    Returns:
+        List[str]: Names of the Git branches.
+    """
+    with Repo.open(repo) as _repo:
+        return _repo.scm.list_branches()
+
+
+def all_commits(repo: Optional[str] = None) -> List[str]:
+    """Get all Git commits in a DVC repository.
+
+    Args:
+        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).
+    Returns:
+        List[str]: SHAs of the Git commits.
+    """
+    with Repo.open(repo) as _repo:
+        return _repo.scm.list_all_commits()
+
+
+def all_tags(repo: Optional[str] = None) -> List[str]:
+    """Get all Git tags in a DVC repository.
+
+    Args:
+        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).
+    Returns:
+        List[str]: Names of the Git tags.
+    """
+    with Repo.open(repo) as _repo:
+        return _repo.scm.list_tags()

dvc/cli/__init__.py

@@ -245,7 +245,7 @@ def main(argv=None):  # noqa: C901, PLR0912, PLR0915
     finally:
         logger.setLevel(outer_log_level)
 
-        from dvc.external_repo import clean_repos
+        from dvc.repo.open_repo import clean_repos
 
         # Remove cached repos in the end of the call, these are anonymous
         # so won't be reused by any other subsequent run anyway.

dvc/commands/add.py

@@ -62,8 +62,11 @@ class CmdAdd(CmdBase):
                 remote=self.args.remote,
                 to_remote=self.args.to_remote,
                 jobs=self.args.jobs,
+                force=self.args.force,
             )
-
+        except FileNotFoundError:
+            logger.exception("")
+            return 1
         except DvcException:
             logger.exception("")
             return 1
@@ -140,6 +143,13 @@ def add_parser(subparsers, parent_parser):
         ),
         metavar="<number>",
     )
+    parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Override local file or folder if exists.",
+    )
 
     _add_annotating_args(parser)
     parser.add_argument(

dvc/commands/data.py

@@ -117,6 +117,7 @@ class CmdDataStatus(CmdBase):
             status = self.repo.data_status(
                 granular=self.args.granular,
                 untracked_files=self.args.untracked_files,
+                not_in_remote=self.args.not_in_remote,
                 remote_refresh=self.args.remote_refresh,
             )
 
@@ -244,10 +245,16 @@ def add_parser(subparsers, parent_parser):
         help="Show untracked files.",
     )
     data_status_parser.add_argument(
-        "--remote-refresh",
+        "--not-in-remote",
         action="store_true",
         default=False,
-        help="Refresh remote index.",
+        help="Show files not in remote.",
+    )
+    data_status_parser.add_argument(
+        "--no-remote-refresh",
+        dest="remote_refresh",
+        action="store_false",
+        help="Use cached remote index (don't check remote).",
     )
     data_status_parser.set_defaults(func=CmdDataStatus)
 

dvc/commands/experiments/__init__.py

@@ -82,6 +82,7 @@ def add_rev_selection_flags(
     experiments_subcmd_parser.add_argument(
         "--rev",
         type=str,
+        action="append",
         default=None,
         help=msg,
         metavar="<commit>",

dvc/commands/experiments/exec_run.py

@@ -20,6 +20,7 @@ class CmdExecutorRun(CmdBaseNoRepo):
             log_level=logger.getEffectiveLevel(),
             infofile=self.args.infofile,
             copy_paths=self.args.copy_paths,
+            message=self.args.message,
         )
         return 0
 
@@ -47,4 +48,11 @@ def add_parser(experiments_subparsers, parent_parser):
             " Only used if `--temp` or `--queue` is specified."
         ),
     )
+    exec_run_parser.add_argument(
+        "-M",
+        "--message",
+        type=str,
+        default=None,
+        help="Custom commit message to use when committing the experiment.",
+    )
     exec_run_parser.set_defaults(func=CmdExecutorRun)

dvc/commands/experiments/push.py

@@ -52,7 +52,9 @@ class CmdExperimentsPush(CmdBase):
             ui.write(humanize.get_summary(stats.items()))
 
         if project_url := result.get("url"):
-            ui.write("[yellow]View your experiments at", project_url, styled=True)
+            ui.rich_print(
+                "View your experiments at", project_url, style="yellow", soft_wrap=True
+            )
 
     def run(self):
         from dvc.repo.experiments.push import UploadError

dvc/commands/experiments/run.py

@@ -37,6 +37,7 @@ class CmdExperimentsRun(CmdRepro):
             tmp_dir=self.args.tmp_dir,
             machine=self.args.machine,
             copy_paths=self.args.copy_paths,
+            message=self.args.message,
             **self._common_kwargs,
         )
 
@@ -147,3 +148,10 @@ def _add_run_common(parser):
             " Only used if `--temp` or `--queue` is specified."
         ),
     )
+    parser.add_argument(
+        "-M",
+        "--message",
+        type=str,
+        default=None,
+        help="Custom commit message to use when committing the experiment.",
+    )

dvc/commands/experiments/save.py

@@ -16,6 +16,7 @@ class CmdExperimentsSave(CmdBase):
                 name=self.args.name,
                 force=self.args.force,
                 include_untracked=self.args.include_untracked,
+                message=self.args.message,
             )
         except DvcException:
             logger.exception("failed to save experiment")
@@ -75,4 +76,11 @@ def add_parser(experiments_subparsers, parent_parser):
         help="List of untracked paths to include in the experiment.",
         metavar="<path>",
     )
+    save_parser.add_argument(
+        "-M",
+        "--message",
+        type=str,
+        default=None,
+        help="Custom commit message to use when committing the experiment.",
+    )
     save_parser.set_defaults(func=CmdExperimentsSave)

dvc/commands/experiments/show.py

@@ -2,9 +2,8 @@ import argparse
 import logging
 import os
 import re
-from collections import Counter, OrderedDict, defaultdict
 from datetime import date, datetime
-from typing import TYPE_CHECKING, Any, Dict, Set
+from typing import TYPE_CHECKING, Dict, Iterable
 
 from funcy import lmap
 
@@ -12,9 +11,8 @@ from dvc.cli import completion
 from dvc.cli.command import CmdBase
 from dvc.cli.utils import append_doc_link
 from dvc.commands.metrics import DEFAULT_PRECISION
-from dvc.exceptions import DvcException, InvalidArgumentError
+from dvc.exceptions import DvcException
 from dvc.ui import ui
-from dvc.utils.flatten import flatten
 from dvc.utils.serialize import encode_exception
 
 if TYPE_CHECKING:
@@ -28,30 +26,6 @@ FILL_VALUE_ERRORED = "!"
 logger = logging.getLogger(__name__)
 
 
-def _update_names(names, items):
-    for name, item in items:
-        item = item.get("data", {})
-        if isinstance(item, dict):
-            item = flatten(item)
-            names[name].update({key: None for key in item})
-
-
-def _collect_names(all_experiments, **kwargs):
-    metric_names: Dict[str, Dict[str, Any]] = defaultdict(dict)
-    param_names: Dict[str, Dict[str, Any]] = defaultdict(dict)
-    deps_names: Set[str] = set()
-
-    for _, experiments in all_experiments.items():
-        for exp_data in experiments.values():
-            exp = exp_data.get("data", {})
-            _update_names(metric_names, exp.get("metrics", {}).items())
-            _update_names(param_names, exp.get("params", {}).items())
-            for dep_name in exp.get("deps", {}):
-                deps_names.add(dep_name)
-
-    return metric_names, param_names, sorted(deps_names)
-
-
 experiment_types = {
     "checkpoint_tip": "│ ╓",
     "checkpoint_commit": "│ ╟",
@@ -62,229 +36,6 @@ experiment_types = {
 }
 
 
-def _collect_rows(  # noqa: C901, PLR0912, PLR0913, PLR0915
-    base_rev,
-    experiments,
-    all_headers,
-    metric_headers,
-    param_headers,
-    metric_names,
-    param_names,
-    deps_names,
-    precision=DEFAULT_PRECISION,
-    sort_by=None,
-    sort_order=None,
-    fill_value=FILL_VALUE,
-    iso=False,
-):
-    from dvc.scm import Git
-
-    if sort_by:
-        sort_path, sort_name, sort_type = _sort_column(
-            sort_by, metric_names, param_names
-        )
-        reverse = sort_order == "desc"
-        experiments = _sort_exp(experiments, sort_path, sort_name, sort_type, reverse)
-
-    new_checkpoint = True
-    for i, (rev, results) in enumerate(experiments.items()):
-        fill_value = FILL_VALUE_ERRORED if results.get("error") else fill_value
-        row_dict = {k: fill_value for k in all_headers}
-
-        exp = results.get("data", {})
-
-        state = exp.get("status")
-        if state == "Success":
-            state = fill_value
-
-        is_baseline = rev == "baseline"
-
-        if is_baseline:
-            name_rev = base_rev[:7] if Git.is_sha(base_rev) else base_rev
-        else:
-            name_rev = rev[:7]
-
-        tip = exp.get("checkpoint_tip")
-        parent_rev = exp.get("checkpoint_parent", "")
-        parent_exp = experiments.get(parent_rev, {}).get("data", {})
-        parent_tip = parent_exp.get("checkpoint_tip")
-
-        parent = ""
-        if is_baseline:
-            typ = "baseline"
-        elif tip:
-            if tip == parent_tip:
-                typ = "checkpoint_tip" if new_checkpoint else "checkpoint_commit"
-            elif parent_rev == base_rev:
-                typ = "checkpoint_base"
-            else:
-                typ = "checkpoint_commit"
-                parent = parent_rev[:7]
-        elif i < len(experiments) - 1:
-            typ = "branch_commit"
-        else:
-            typ = "branch_base"
-
-        if not is_baseline:
-            new_checkpoint = not (tip and tip == parent_tip)
-
-        row_dict["Experiment"] = exp.get("name", "")
-        row_dict["rev"] = name_rev
-        row_dict["typ"] = typ
-        row_dict["Created"] = format_time(exp.get("timestamp"), fill_value, iso)
-        row_dict["parent"] = parent
-        row_dict["State"] = state
-        row_dict["Executor"] = exp.get("executor", fill_value)
-
-        _extend_row(
-            row_dict,
-            metric_names,
-            metric_headers,
-            exp.get("metrics", {}).items(),
-            precision,
-            fill_value=fill_value,
-        )
-        _extend_row(
-            row_dict,
-            param_names,
-            param_headers,
-            exp.get("params", {}).items(),
-            precision,
-            fill_value=fill_value,
-        )
-        for dep in deps_names:
-            hash_info = exp.get("deps", {}).get(dep, {}).get("hash")
-            if hash_info is not None:
-                hash_info = hash_info[:7]
-            row_dict[dep] = hash_info
-        yield list(row_dict.values())
-
-
-def _sort_column(sort_by, metric_names, param_names):
-    path, _, sort_name = sort_by.rpartition(":")
-    matches = set()
-
-    if path:
-        if path in metric_names and sort_name in metric_names[path]:
-            matches.add((path, sort_name, "metrics"))
-        if path in param_names and sort_name in param_names[path]:
-            matches.add((path, sort_name, "params"))
-    else:
-        for path in metric_names:
-            if sort_name in metric_names[path]:
-                matches.add((path, sort_name, "metrics"))
-        for path in param_names:
-            if sort_name in param_names[path]:
-                matches.add((path, sort_name, "params"))
-
-    if len(matches) == 1:
-        return matches.pop()
-    if len(matches) > 1:
-        raise InvalidArgumentError(
-            "Ambiguous sort column '{}' matched '{}'".format(
-                sort_by,
-                ", ".join([f"{path}:{name}" for path, name, _ in matches]),
-            )
-        )
-    raise InvalidArgumentError(f"Unknown sort column '{sort_by}'")
-
-
-def _sort_exp(experiments, sort_path, sort_name, typ, reverse):
-    def _sort(item):
-        rev, exp = item
-        exp_data = exp.get("data", {})
-        tip = exp_data.get("checkpoint_tip")
-        if tip and tip != rev:
-            # Sort checkpoint experiments by tip commit
-            return _sort((tip, experiments[tip]))
-        data = exp_data.get(typ, {}).get(sort_path, {}).get("data", {})
-        val = flatten(data).get(sort_name)
-        return val is None, val
-
-    ret = OrderedDict()
-    if "baseline" in experiments:
-        ret["baseline"] = experiments.pop("baseline")
-
-    ret.update(sorted(experiments.items(), key=_sort, reverse=reverse))
-    return ret
-
-
-def format_time(datetime_obj, fill_value=FILL_VALUE, iso=False):
-    if datetime_obj is None:
-        return fill_value
-
-    if iso:
-        return datetime_obj.isoformat()
-
-    if datetime_obj.date() == date.today():
-        fmt = "%I:%M %p"
-    else:
-        fmt = "%b %d, %Y"
-    return datetime_obj.strftime(fmt)
-
-
-def _extend_row(row, names, headers, items, precision, fill_value=FILL_VALUE):
-    from dvc.compare import _format_field, with_value
-
-    for fname, data in items:
-        item = data.get("data", {})
-        item = flatten(item) if isinstance(item, dict) else {fname: item}
-        for name in names[fname]:
-            value = with_value(
-                item.get(name),
-                FILL_VALUE_ERRORED if data.get("error", None) else fill_value,
-            )
-            # wrap field data in ui.rich_text, otherwise rich may
-            # interpret unescaped braces from list/dict types as rich
-            # markup tags
-            value = ui.rich_text(str(_format_field(value, precision)))
-            if name in headers:
-                row[name] = value
-            else:
-                row[f"{fname}:{name}"] = value
-
-
-def experiments_table(  # noqa: PLR0913
-    all_experiments,
-    headers,
-    metric_headers,
-    metric_names,
-    param_headers,
-    param_names,
-    deps_names,
-    sort_by=None,
-    sort_order=None,
-    precision=DEFAULT_PRECISION,
-    fill_value=FILL_VALUE,
-    iso=False,
-) -> "TabularData":
-    from funcy import lconcat
-
-    from dvc.compare import TabularData
-
-    all_headers = lconcat(headers, metric_headers, param_headers, deps_names)
-    td = TabularData(all_headers, fill_value=fill_value)
-    for base_rev, experiments in all_experiments.items():
-        rows = _collect_rows(
-            base_rev,
-            experiments,
-            all_headers,
-            metric_headers,
-            param_headers,
-            metric_names,
-            param_names,
-            deps_names,
-            sort_by=sort_by,
-            sort_order=sort_order,
-            precision=precision,
-            fill_value=fill_value,
-            iso=iso,
-        )
-        td.extend(rows)
-
-    return td
-
-
 def prepare_exp_id(kwargs) -> "RichText":
     exp_name = kwargs["Experiment"]
     rev = kwargs["rev"]
@@ -309,7 +60,8 @@ def baseline_styler(typ):
 
 
 def show_experiments(  # noqa: C901, PLR0912
-    all_experiments,
+    td: "TabularData",
+    headers: Dict[str, Iterable[str]],
     keep=None,
     drop=None,
     pager=True,
@@ -318,40 +70,6 @@ def show_experiments(  # noqa: C901, PLR0912
     pcp=False,
     **kwargs,
 ):
-    from funcy.seqs import flatten as flatten_list
-
-    metric_names, param_names, deps_names = _collect_names(all_experiments)
-
-    headers_list = [
-        "Experiment",
-        "rev",
-        "typ",
-        "Created",
-        "parent",
-        "State",
-        "Executor",
-    ]
-
-    names = {**metric_names, **param_names}
-    counter = Counter(flatten_list([list(a.keys()) for a in names.values()]))
-    counter.update(headers_list)
-    metric_headers = _normalize_headers(metric_names, counter)
-    param_headers = _normalize_headers(param_names, counter)
-
-    td = experiments_table(
-        all_experiments,
-        headers_list,
-        metric_headers,
-        metric_names,
-        param_headers,
-        param_names,
-        deps_names,
-        kwargs.get("sort_by"),
-        kwargs.get("sort_order"),
-        kwargs.get("precision"),
-        kwargs.get("fill_value"),
-        kwargs.get("iso"),
-    )
     if keep:
         for col in td.keys():  # noqa: SIM118
             if re.match(keep, col):
@@ -368,11 +86,6 @@ def show_experiments(  # noqa: C901, PLR0912
         td.column("Experiment")[:] = map(prepare_exp_id, td.as_dict(merge_headers))
         td.drop(*merge_headers[1:])
 
-    headers = {
-        "metrics": metric_headers,
-        "params": param_headers,
-        "deps": deps_names,
-    }
     styles = {
         "Experiment": {"no_wrap": True, "header_style": "black on grey93"},
         "Created": {"header_style": "black on grey93"},
@@ -453,19 +166,17 @@ 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)
 
 
 class CmdExperimentsShow(CmdBase):
     def run(self):
+        from dvc.repo.experiments.show import tabulate
+
         try:
-            all_experiments = self.repo.experiments.show(
+            exps = self.repo.experiments.show(
                 all_branches=self.args.all_branches,
                 all_tags=self.args.all_tags,
                 all_commits=self.args.all_commits,
@@ -483,23 +194,29 @@ class CmdExperimentsShow(CmdBase):
             return 1
 
         if self.args.json:
-            ui.write_json(all_experiments, default=_format_json)
+            ui.write_json([exp.dumpd() for exp in exps], default=_format_json)
         else:
             precision = (
                 self.args.precision or None if self.args.csv else DEFAULT_PRECISION
             )
             fill_value = "" if self.args.csv else FILL_VALUE
             iso = self.args.csv
+            td, headers = tabulate(
+                exps,
+                precision=precision,
+                fill_value=fill_value,
+                iso=iso,
+                sort_by=self.args.sort_by,
+                sort_order=self.args.sort_order,
+            )
 
             show_experiments(
-                all_experiments,
+                td,
+                headers,
                 keep=self.args.keep,
                 drop=self.args.drop,
                 sort_by=self.args.sort_by,
                 sort_order=self.args.sort_order,
-                precision=precision,
-                fill_value=fill_value,
-                iso=iso,
                 pager=not self.args.no_pager,
                 csv=self.args.csv,
                 markdown=self.args.markdown,

dvc/commands/gc.py

@@ -10,7 +10,7 @@ logger = logging.getLogger(__name__)
 
 
 class CmdGC(CmdBase):
-    def run(self):  # noqa: C901
+    def run(self):  # noqa: C901, PLR0912
         from dvc.repo.gc import _validate_args
 
         _validate_args(
@@ -22,6 +22,8 @@ class CmdGC(CmdBase):
             workspace=self.args.workspace,
             rev=self.args.rev,
             num=self.args.num,
+            cloud=self.args.cloud,
+            not_in_remote=self.args.not_in_remote,
         )
 
         if self.args.rev:
@@ -46,6 +48,10 @@ class CmdGC(CmdBase):
 
         if self.args.all_experiments:
             msg += " and all experiments"
+
+        if self.args.not_in_remote:
+            msg += " that are present in the DVC remote"
+
         if self.args.repos:
             msg += " of the current and the following repos:"
 
@@ -74,6 +80,7 @@ class CmdGC(CmdBase):
             workspace=self.args.workspace,
             rev=self.args.rev,
             num=self.args.num,
+            not_in_remote=self.args.not_in_remote,
         )
         return 0
 
@@ -157,12 +164,18 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Keep data files for all experiments.",
     )
+    gc_parser.add_argument(
+        "--not-in-remote",
+        action="store_true",
+        default=False,
+        help="Keep data files that are not present in the remote.",
+    )
     gc_parser.add_argument(
         "-c",
         "--cloud",
         action="store_true",
         default=False,
-        help="Collect garbage in remote repository.",
+        help="Collect garbage in remote storage in addition to local cache.",
     )
     gc_parser.add_argument(
         "-r",

dvc/commands/get.py

@@ -36,6 +36,7 @@ class CmdGet(CmdBaseNoRepo):
                 out=self.args.out,
                 rev=self.args.rev,
                 jobs=self.args.jobs,
+                force=self.args.force,
             )
             return 0
         except CloneError:
@@ -94,4 +95,11 @@ def add_parser(subparsers, parent_parser):
         ),
         metavar="<number>",
     )
+    get_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Override local file or folder if exists.",
+    )
     get_parser.set_defaults(func=CmdGet)

dvc/commands/get_url.py

@@ -14,7 +14,12 @@ class CmdGetUrl(CmdBaseNoRepo):
         from dvc.repo import Repo
 
         try:
-            Repo.get_url(self.args.url, out=self.args.out, jobs=self.args.jobs)
+            Repo.get_url(
+                self.args.url,
+                out=self.args.out,
+                jobs=self.args.jobs,
+                force=self.args.force,
+            )
             return 0
         except DvcException:
             logger.exception("failed to get '%s'", self.args.url)
@@ -46,4 +51,11 @@ def add_parser(subparsers, parent_parser):
         ),
         metavar="<number>",
     )
+    get_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Override local file or folder if exists.",
+    )
     get_parser.set_defaults(func=CmdGetUrl)

dvc/commands/imp_url.py

@@ -25,6 +25,7 @@ class CmdImportUrl(CmdBase):
                 labels=self.args.labels,
                 meta=self.args.meta,
                 jobs=self.args.jobs,
+                force=self.args.force,
                 version_aware=self.args.version_aware,
             )
         except DvcException:
@@ -112,6 +113,13 @@ def add_parser(subparsers, parent_parser):
         ),
         metavar="<number>",
     )
+    import_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Override local file or folder if exists.",
+    )
     import_parser.add_argument(
         "--version-aware",
         action="store_true",

dvc/commands/queue/status.py

@@ -1,10 +1,8 @@
 import argparse
 import logging
-from typing import Dict, List, Optional
 
 from dvc.cli.command import CmdBase
 from dvc.cli.utils import append_doc_link
-from dvc.commands.experiments.show import format_time
 from dvc.compare import TabularData
 from dvc.ui import ui
 
@@ -15,9 +13,9 @@ class CmdQueueStatus(CmdBase):
     """Show queue task and worker status."""
 
     def run(self) -> int:
-        result: List[
-            Dict[str, Optional[str]]
-        ] = self.repo.experiments.celery_queue.status()
+        from dvc.repo.experiments.show import format_time
+
+        result = self.repo.experiments.celery_queue.status()
         if result:
             all_headers = ["Task", "Name", "Created", "Status"]
             td = TabularData(all_headers)

dvc/commands/repro.py

@@ -38,6 +38,7 @@ class CmdRepro(CmdBase):
             "recursive": self.args.recursive,
             "force_downstream": self.args.force_downstream,
             "pull": self.args.pull,
+            "allow_missing": self.args.allow_missing,
         }
 
     @property
@@ -135,6 +136,12 @@ and then the stage name name.
             "from the run-cache."
         ),
     )
+    repro_parser.add_argument(
+        "--allow-missing",
+        action="store_true",
+        default=False,
+        help=("Skip stages with missing data but no other changes."),
+    )
     repro_parser.add_argument(
         "--dry",
         action="store_true",

dvc/commands/stage.py

@@ -1,5 +1,6 @@
 import argparse
 import logging
+from contextlib import contextmanager
 from itertools import chain, filterfalse
 from typing import TYPE_CHECKING, Dict, Iterable, List
 
@@ -119,8 +120,22 @@ def parse_cmd(commands: List[str]) -> str:
     return " ".join(map(quote_argument, commands))
 
 
+@contextmanager
+def _disable_logging(highest_level=logging.CRITICAL):
+    previous_level = logging.root.manager.disable
+
+    logging.disable(highest_level)
+
+    try:
+        yield
+    finally:
+        logging.disable(previous_level)
+
+
 class CmdStageAdd(CmdBase):
     def run(self):
+        from dvc.repo import lock_repo
+
         kwargs = vars(self.args)
         kwargs.update(
             {
@@ -128,7 +143,15 @@ class CmdStageAdd(CmdBase):
                 "params": parse_params(self.args.params),
             }
         )
-        self.repo.stage.add(**kwargs)
+
+        with self.repo.scm_context, lock_repo(self.repo):
+            with _disable_logging(logging.INFO):
+                stage = self.repo.stage.add(**kwargs)
+            logger.info("Added stage %r in %r", stage.addressing, stage.relpath)
+            if self.args.run:
+                stage.run()
+                stage.dump(update_pipeline=False)
+
         return 0
 
 
@@ -257,6 +280,12 @@ def _add_common_args(parser):
             "This doesn't affect any DVC operations."
         ),
     )
+    parser.add_argument(
+        "--run",
+        action="store_true",
+        default=False,
+        help="Execute the stage after generating it.",
+    )
     parser.add_argument(
         "command",
         nargs=argparse.REMAINDER,

dvc/compare.py

@@ -35,7 +35,7 @@ def with_value(value, default):
 
 
 class TabularData(MutableSequence[Sequence["CellT"]]):
-    def __init__(self, columns: Sequence[str], fill_value: str = ""):
+    def __init__(self, columns: Sequence[str], fill_value: Optional[str] = ""):
         self._columns: Dict[str, Column] = {name: Column() for name in columns}
         self._keys: List[str] = list(columns)
         self._fill_value = fill_value

dvc/config.py

@@ -79,9 +79,6 @@ class Config(dict):
         ConfigError: thrown if config has an invalid format.
     """
 
-    APPNAME = "dvc"
-    APPAUTHOR = "iterative"
-
     SYSTEM_LEVELS = ("system", "global")
     REPO_LEVELS = ("repo", "local")
     # In the order they shadow each other
@@ -121,14 +118,14 @@ class Config(dict):
 
     @classmethod
     def get_dir(cls, level):
-        from platformdirs import site_config_dir, user_config_dir
+        from dvc.dirs import global_config_dir, system_config_dir
 
         assert level in ("global", "system")
 
         if level == "global":
-            return user_config_dir(cls.APPNAME, cls.APPAUTHOR)
+            return global_config_dir()
         if level == "system":
-            return site_config_dir(cls.APPNAME, cls.APPAUTHOR)
+            return system_config_dir()
 
     @cached_property
     def files(self) -> Dict[str, str]:

dvc/config_schema.py

@@ -215,6 +215,7 @@ SCHEMA = {
                     "timeout": Coerce(int),
                     "gss_auth": Bool,
                     "allow_agent": Bool,
+                    "max_sessions": Coerce(int),
                     Optional("verify", default=False): Bool,
                     **REMOTE_COMMON,
                 },
@@ -301,9 +302,6 @@ SCHEMA = {
     "feature": FeatureSchema(
         {
             Optional("machine", default=False): Bool,
-            Optional("push_exp_to_studio", default=False): Bool,
-            "studio_token": str,
-            "studio_url": str,
         },
     ),
     "plots": {
@@ -329,4 +327,10 @@ SCHEMA = {
         "config_dir": str,
         "config_name": str,
     },
+    "studio": {
+        "token": str,
+        "url": str,
+        "repo_url": str,
+        Optional("offline", default=False): Bool,
+    },
 }

dvc/dependency/repo.py

@@ -79,10 +79,14 @@ class RepoDependency(Dependency):
     ) -> "DVCFileSystem":
         from dvc.fs import DVCFileSystem
 
+        config = {"cache": self.repo.config["cache"]}
+        config["cache"]["dir"] = self.repo.cache.local.path
+
         return DVCFileSystem(
             url=self.def_repo[self.PARAM_URL],
             rev=rev or self._get_rev(locked=locked),
             subrepos=True,
+            config=config,
         )
 
     def _get_rev(self, locked: bool = True):

dvc/dirs.py

@@ -0,0 +1,26 @@
+import os
+
+import platformdirs
+
+from . import env
+
+APPNAME = "dvc"
+APPAUTHOR = "iterative"
+
+
+def system_config_dir():
+    return os.getenv(env.DVC_SYSTEM_CONFIG_DIR) or platformdirs.site_config_dir(
+        APPNAME, APPAUTHOR
+    )
+
+
+def global_config_dir():
+    return os.getenv(env.DVC_SYSTEM_CONFIG_DIR) or platformdirs.user_config_dir(
+        APPNAME, APPAUTHOR
+    )
+
+
+def site_cache_dir():
+    return os.getenv(env.DVC_SITE_CACHE_DIR) or platformdirs.site_cache_dir(
+        APPNAME, APPAUTHOR, opinion=True
+    )

dvc/env.py

@@ -1,12 +1,19 @@
+DVCLIVE_RESUME = "DVCLIVE_RESUME"
 DVC_CHECKPOINT = "DVC_CHECKPOINT"
 DVC_DAEMON = "DVC_DAEMON"
+DVC_EXP_AUTO_PUSH = "DVC_EXP_AUTO_PUSH"
+DVC_EXP_BASELINE_REV = "DVC_EXP_BASELINE_REV"
+DVC_EXP_GIT_REMOTE = "DVC_EXP_GIT_REMOTE"
+DVC_EXP_NAME = "DVC_EXP_NAME"
+DVC_GLOBAL_CONFIG_DIR = "DVC_GLOBAL_CONFIG_DIR"
+DVC_IGNORE_ISATTY = "DVC_IGNORE_ISATTY"
+DVC_NO_ANALYTICS = "DVC_NO_ANALYTICS"
 DVC_PAGER = "DVC_PAGER"
 DVC_ROOT = "DVC_ROOT"
-DVCLIVE_RESUME = "DVCLIVE_RESUME"
-DVC_IGNORE_ISATTY = "DVC_IGNORE_ISATTY"
-DVC_EXP_BASELINE_REV = "DVC_EXP_BASELINE_REV"
-DVC_EXP_NAME = "DVC_EXP_NAME"
-DVC_EXP_GIT_REMOTE = "DVC_EXP_GIT_REMOTE"
-DVC_EXP_AUTO_PUSH = "DVC_EXP_AUTO_PUSH"
-DVC_NO_ANALYTICS = "DVC_NO_ANALYTICS"
 DVC_SHOW_TRACEBACK = "DVC_SHOW_TRACEBACK"
+DVC_SITE_CACHE_DIR = "DVC_SITE_CACHE_DIR"
+DVC_STUDIO_OFFLINE = "DVC_STUDIO_OFFLINE"
+DVC_STUDIO_REPO_URL = "DVC_STUDIO_REPO_URL"
+DVC_STUDIO_TOKEN = "DVC_STUDIO_TOKEN"  # noqa: S105 # nosec B105
+DVC_STUDIO_URL = "DVC_STUDIO_URL"
+DVC_SYSTEM_CONFIG_DIR = "DVC_SYSTEM_CONFIG_DIR"

dvc/exceptions.py

@@ -1,4 +1,5 @@
 """Exceptions raised by the dvc."""
+import errno
 from typing import Dict, List
 
 from dvc.utils import format_link
@@ -206,6 +207,18 @@ class ETagMismatchError(DvcException):
         )
 
 
+class FileExistsLocallyError(FileExistsError, DvcException):
+    def __init__(self, path, hint=None):
+        import os.path
+
+        self.path = path
+        hint = "" if hint is None else f". {hint}"
+        path_typ = "directory" if os.path.isdir(path) else "file"
+        msg = f"The {path_typ} '{path}' already exists locally{hint}"
+        super().__init__(msg)
+        self.errno = errno.EEXIST
+
+
 class FileMissingError(DvcException):
     def __init__(self, path, hint=None):
         self.path = path

dvc/info.py

@@ -32,6 +32,7 @@ def get_dvc_info():
         f"Platform: Python {platform.python_version()} on {platform.platform()}",
         f"Subprojects:{_get_subprojects()}",
         f"Supports:{_get_supported_remotes()}",
+        f"Config:{_get_config_dirs()}",
     ]
 
     try:
@@ -42,7 +43,7 @@ def get_dvc_info():
             # `dvc config cache.shared group`.
             if os.path.exists(repo.cache.local.path):
                 info.append(f"Cache types: {_get_linktype_support_info(repo)}")
-                fs_type = get_fs_type(repo.cache.local.path)
+                fs_type = _get_fs_type(repo.cache.local.path)
                 info.append(f"Cache directory: {fs_type}")
             else:
                 info.append("Cache types: " + error_link("no-dvc-cache"))
@@ -51,7 +52,7 @@ def get_dvc_info():
             info.append(f"Remotes: {_get_remotes(repo.config)}")
 
             root_directory = repo.root_dir
-            fs_root = get_fs_type(os.path.abspath(root_directory))
+            fs_root = _get_fs_type(os.path.abspath(root_directory))
             info.append(f"Workspace directory: {fs_root}")
             info.append(f"Repo: {_get_dvc_repo_info(repo)}")
             info.append(f"Repo.site_cache_dir: {repo.site_cache_dir}")
@@ -136,7 +137,18 @@ def _get_supported_remotes():
     return "\n\t" + ",\n\t".join(supported_remotes)
 
 
-def get_fs_type(path):
+def _get_config_dirs():
+    from dvc.config import Config
+
+    dirs = [
+        f"Global: {Config.get_dir('global')}",
+        f"System: {Config.get_dir('system')}",
+    ]
+
+    return "\n\t" + "\n\t".join(dirs)
+
+
+def _get_fs_type(path):
     partition = {}
     for part in psutil.disk_partitions(all=True):
         if part.fstype:
@@ -155,11 +167,11 @@ def get_fs_type(path):
     return ("unknown", "none")
 
 
-def _get_dvc_repo_info(self):
-    if self.config.get("core", {}).get("no_scm", False):
+def _get_dvc_repo_info(repo):
+    if repo.config.get("core", {}).get("no_scm", False):
         return "dvc (no_scm)"
 
-    if self.root_dir != self.scm.root_dir:
+    if repo.root_dir != repo.scm.root_dir:
         return "dvc (subdir), git"
 
     return "dvc, git"

dvc/output.py

@@ -1,10 +1,11 @@
+import errno
 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
+from typing import TYPE_CHECKING, Dict, List, Optional, Set, Tuple, Type, Union
 from urllib.parse import urlparse
 
 from funcy import collecting, first, project
@@ -25,26 +26,25 @@ from dvc_data.hashfile import check as ocheck
 from dvc_data.hashfile import load as oload
 from dvc_data.hashfile.build import build
 from dvc_data.hashfile.checkout import checkout
+from dvc_data.hashfile.db import HashFileDB, add_update_tree
 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_data.hashfile.tree import Tree, du
 from dvc_objects.errors import ObjectFormatError
 
 from .annotations import ANNOTATION_FIELDS, ANNOTATION_SCHEMA, Annotation
 from .fs import LocalFileSystem, RemoteMissingDepsError, Schemes, get_cloud_fs
-from .fs.callbacks import DEFAULT_CALLBACK
+from .fs.callbacks import DEFAULT_CALLBACK, Callback, TqdmCallback
 from .utils import relpath
 from .utils.fs import path_isin
 
 if TYPE_CHECKING:
-    from dvc_data.hashfile.db import HashFileDB
     from dvc_data.hashfile.obj import HashFile
     from dvc_data.index import DataIndexKey
     from dvc_objects.db import ObjectDB
 
-    from .fs.callbacks import Callback
     from .ignore import DvcIgnoreFilter
 
 logger = logging.getLogger(__name__)
@@ -276,6 +276,11 @@ class OutputIsIgnoredError(DvcException):
         super().__init__(f"Path '{match.file}' is ignored by\n{lines}")
 
 
+class CheckoutCallback(TqdmCallback):
+    # disable branching for checkouts
+    branch = Callback.branch  # type: ignore[assignment]
+
+
 class Output:
     IS_DEPENDENCY = False
 
@@ -417,8 +422,6 @@ class Output:
 
     def _compute_meta_hash_info_from_files(self) -> None:
         if self.files:
-            from dvc_data.hashfile.db import HashFileDB
-
             tree = Tree.from_list(self.files, hash_name=self.hash_name)
             tree.digest(with_meta=True)
             self.odb = HashFileDB(tree.fs, tree.path + ".odb")
@@ -741,15 +744,18 @@ class Output:
                     hardlink=hardlink,
                 )
             if relink:
-                self._checkout(
-                    filter_info or self.fs_path,
-                    self.fs,
-                    obj,
-                    self.cache,
-                    relink=True,
-                    state=self.repo.state,
-                    prompt=prompt.confirm,
-                )
+                rel = self.fs.path.relpath(filter_info or self.fs_path)
+                with CheckoutCallback(desc=f"Checking out {rel}", unit="files") as cb:
+                    self._checkout(
+                        filter_info or self.fs_path,
+                        self.fs,
+                        obj,
+                        self.cache,
+                        relink=True,
+                        state=self.repo.state,
+                        prompt=prompt.confirm,
+                        progress_callback=cb,
+                    )
                 self.set_exec()
 
     def _commit_granular_dir(self, filter_info, hardlink) -> Optional["HashFile"]:
@@ -891,9 +897,16 @@ class Output:
         checkpoint_reset: bool = False,
         **kwargs,
     ) -> Optional[Tuple[bool, Optional[bool]]]:
+        # callback passed act as a aggregate callback.
+        # do not let checkout to call set_size and change progressbar.
+        class CallbackProxy(Callback):
+            def relative_update(self, inc: int = 1) -> None:
+                progress_callback.relative_update(inc)
+                return super().relative_update(inc)
+
+        callback = CallbackProxy()
         if not self.use_cache:
-            if progress_callback != DEFAULT_CALLBACK:
-                progress_callback.relative_update(self.get_files_number(filter_info))
+            callback.relative_update(self.get_files_number(filter_info))
             return None
 
         obj = self.get_obj(filter_info=filter_info)
@@ -915,7 +928,7 @@ class Output:
                 obj,
                 self.cache,
                 force=force,
-                progress_callback=progress_callback,
+                progress_callback=callback,
                 relink=relink,
                 state=self.repo.state,
                 prompt=prompt.confirm,
@@ -1008,7 +1021,7 @@ class Output:
         if self.exists:
             self.cache.unprotect(self.fs_path)
 
-    def get_dir_cache(self, **kwargs):
+    def get_dir_cache(self, **kwargs) -> Optional["Tree"]:
         if not self.is_dir_checksum:
             raise DvcException("cannot get dir cache for file checksum")
 
@@ -1022,14 +1035,17 @@ class Output:
                 self.repo.cloud.pull([obj.hash_info], **kwargs)
 
         if self.obj:
+            assert isinstance(self.obj, Tree)
             return self.obj
 
         try:
-            self.obj = oload(self.cache, self.hash_info)
+            obj = oload(self.cache, self.hash_info)
+            assert isinstance(obj, Tree)
         except (FileNotFoundError, ObjectFormatError):
-            self.obj = None
+            obj = None
 
-        return self.obj
+        self.obj = obj
+        return obj
 
     def _collect_used_dir_cache(
         self, remote=None, force=False, jobs=None, filter_info=None
@@ -1165,7 +1181,7 @@ class Output:
 
     def merge(self, ancestor, other, allowed=None):
         from dvc_data.hashfile.tree import MergeError as TreeMergeError
-        from dvc_data.hashfile.tree import du, merge
+        from dvc_data.hashfile.tree import merge
 
         assert other
 
@@ -1198,6 +1214,167 @@ class Output:
             nfiles=len(merged),
         )
 
+    def unstage(self, path: str) -> Tuple["Meta", "Tree"]:
+        from pygtrie import Trie
+
+        from dvc_objects.fs.path import Path
+
+        assert isinstance(self.fs.path, Path)
+        rel_key = tuple(self.fs.path.parts(self.fs.path.relpath(path, self.fs_path)))
+
+        if self.hash_info:
+            tree = self.get_dir_cache()
+            if tree is None:
+                raise DvcException(f"could not read {self.hash_info.value!r}")
+        else:
+            tree = Tree()
+
+        trie = tree.as_trie()
+        assert isinstance(trie, Trie)
+
+        try:
+            del trie[rel_key:]  # type: ignore[misc]
+        except KeyError:
+            raise FileNotFoundError(  # noqa: B904
+                errno.ENOENT,
+                os.strerror(errno.ENOENT),
+                self.fs.path.relpath(path),
+            )
+
+        new = tree.from_trie(trie)
+        new.digest()
+        return Meta(nfiles=len(new), isdir=True), new
+
+    def apply(
+        self,
+        path: str,
+        obj: Union["Tree", "HashFile"],
+        meta: "Meta",
+    ) -> Tuple["Meta", "Tree"]:
+        from pygtrie import Trie
+
+        from dvc_objects.fs.path import Path
+
+        assert isinstance(self.fs.path, Path)
+        append_only = True
+        rel_key = tuple(self.fs.path.parts(self.fs.path.relpath(path, self.fs_path)))
+
+        if self.hash_info:
+            tree = self.get_dir_cache()
+            if tree is None:
+                raise DvcException(f"could not read {self.hash_info.value!r}")
+        else:
+            tree = Tree()
+
+        trie = tree.as_trie()
+        assert isinstance(trie, Trie)
+
+        try:
+            del trie[rel_key:]  # type: ignore[misc]
+        except KeyError:
+            pass
+        else:
+            append_only = False
+
+        items = {}
+        if isinstance(obj, Tree):
+            items = {(*rel_key, *key): (m, o) for key, m, o in obj}
+        else:
+            items = {rel_key: (meta, obj.hash_info)}
+        trie.update(items)
+
+        new = Tree.from_trie(trie)
+        new.digest()
+
+        size = self.meta.size if self.meta and self.meta.size else None
+        if append_only and size and meta.size is not None:
+            # if files were only appended, we can sum to the existing size
+            size += meta.size
+        elif self.hash_info and self.hash_info == new.hash_info:
+            # if hashes are same, sizes must have been the same
+            size = self.meta.size
+        else:
+            size = None
+
+        meta = Meta(nfiles=len(new), size=size, isdir=True)
+        return meta, new
+
+    def add(  # noqa: C901
+        self, path: Optional[str] = None, no_commit: bool = False, relink: bool = True
+    ) -> Optional["HashFile"]:
+        path = path or self.fs_path
+        if self.hash_info and not self.is_dir_checksum and self.fs_path != path:
+            raise DvcException(
+                f"Cannot modify '{self}' which is being tracked as a file"
+            )
+
+        assert self.repo
+        cache = self.cache if self.use_cache else self.repo.cache.local
+        assert isinstance(cache, HashFileDB)
+
+        new: "HashFile"
+        try:
+            assert self.hash_name
+            staging, meta, obj = build(
+                cache,
+                path,
+                self.fs,
+                self.hash_name,
+                ignore=self.dvcignore,
+                dry_run=not self.use_cache,
+            )
+        except FileNotFoundError as exc:
+            if self.fs_path == path:
+                raise self.DoesNotExistError(self) from exc
+            if not self.is_dir_checksum:
+                raise
+
+            meta, new = self.unstage(path)
+            staging, obj = None, None
+        else:
+            assert obj
+            assert staging
+            if self.fs_path != path:
+                meta, new = self.apply(path, obj, meta)
+                add_update_tree(staging, new)
+            else:
+                new = obj
+
+        self.obj = new
+        self.hash_info = self.obj.hash_info
+        self.meta = meta
+        self.files = None
+        self.ignore()
+
+        if no_commit or not self.use_cache:
+            return obj
+
+        if isinstance(new, Tree):
+            add_update_tree(cache, new)
+
+        if not obj:
+            return obj
+
+        assert staging
+        assert obj.hash_info
+        otransfer(staging, self.cache, {obj.hash_info}, hardlink=relink, shallow=False)
+
+        if relink:
+            rel = self.fs.path.relpath(path)
+            with CheckoutCallback(desc=f"Checking out {rel}", unit="files") as callback:
+                self._checkout(
+                    path,
+                    self.fs,
+                    obj,
+                    self.cache,
+                    relink=True,
+                    state=self.repo.state,
+                    prompt=prompt.confirm,
+                    progress_callback=callback,
+                )
+            self.set_exec()
+        return obj
+
     @property
     def fspath(self):
         return self.fs_path

dvc/repo/__init__.py

@@ -287,18 +287,9 @@ class Repo:
 
     @staticmethod
     def open(url, *args, **kwargs):  # noqa: A003
-        if url is None:
-            url = os.getcwd()
+        from .open_repo import open_repo
 
-        if os.path.exists(url):
-            try:
-                return Repo(url, *args, **kwargs)
-            except NotDvcRepoError:
-                pass  # fallthrough to external_repo
-
-        from dvc.external_repo import external_repo
-
-        return external_repo(url, *args, **kwargs)
+        return open_repo(url, *args, **kwargs)
 
     @cached_property
     def scm(self) -> Union["Git", "NoSCM"]:
@@ -563,12 +554,10 @@ class Repo:
         import getpass
         import hashlib
 
-        import platformdirs
-
+        from dvc.dirs import site_cache_dir
         from dvc.fs import GitFileSystem
 
-        default = platformdirs.site_cache_dir("dvc", "iterative", opinion=True)
-        cache_dir = self.config["core"].get("site_cache_dir") or default
+        cache_dir = self.config["core"].get("site_cache_dir") or site_cache_dir()
 
         if isinstance(self.fs, GitFileSystem):
             relparts = ()

dvc/repo/add.py

@@ -2,14 +2,16 @@ import logging
 import os
 from contextlib import contextmanager
 from itertools import tee
-from typing import TYPE_CHECKING, Any, Iterator, List, Optional
+from typing import TYPE_CHECKING, Any, Iterator, List, NamedTuple, Optional
 
 import colorama
 
 from dvc.exceptions import (
     CacheLinkError,
+    DvcException,
     InvalidArgumentError,
     OutputDuplicationError,
+    OutputNotFoundError,
     OverlappingOutputPathsError,
     RecursiveAddingWhileUsingFilename,
 )
@@ -29,13 +31,25 @@ Stages = List["Stage"]
 logger = logging.getLogger(__name__)
 
 
-OVERLAPPING_OUTPUT_FMT = (
+class StageInfo(NamedTuple):
+    stage: "Stage"
+    output_exists: bool
+
+
+OVERLAPPING_CHILD_FMT = (
     "Cannot add '{out}', because it is overlapping with other "
     "DVC tracked output: '{parent}'.\n"
     "To include '{out}' in '{parent}', run "
     "'dvc commit {parent_stage}'"
 )
 
+OVERLAPPING_PARENT_FMT = (
+    "Cannot add '{parent}', because it is overlapping with other "
+    "DVC tracked output: '{out}'.\n"
+    "To include '{out}' in '{parent}', run "
+    "'dvc remove {out_stage}' and then 'dvc add {parent}'"
+)
+
 
 def check_recursive_and_fname(args):
     if args.recursive and args.fname:
@@ -86,11 +100,18 @@ def translate_graph_error(stages: Stages) -> Iterator[None]:
     try:
         yield
     except OverlappingOutputPathsError as exc:
-        msg = OVERLAPPING_OUTPUT_FMT.format(
-            out=exc.overlapping_out.fs_path,
-            parent=exc.parent.fs_path,
-            parent_stage=exc.parent.stage.addressing,
-        )
+        if exc.parent in [o for s in stages for o in s.outs]:
+            msg = OVERLAPPING_PARENT_FMT.format(
+                out=exc.overlapping_out,
+                parent=exc.parent,
+                out_stage=exc.overlapping_out.stage.addressing,
+            )
+        else:
+            msg = OVERLAPPING_CHILD_FMT.format(
+                out=exc.overlapping_out,
+                parent=exc.parent,
+                parent_stage=exc.parent.stage.addressing,
+            )
         raise OverlappingOutputPathsError(  # noqa: B904
             exc.parent, exc.overlapping_out, msg
         )
@@ -100,7 +121,7 @@ def translate_graph_error(stages: Stages) -> Iterator[None]:
         )
 
 
-def progress_iter(stages: Stages) -> Iterator["Stage"]:
+def progress_iter(stages: List[StageInfo]) -> Iterator["StageInfo"]:
     total = len(stages)
     desc = "Adding..."
     with ui.progress(stages, total=total, desc=desc, unit="file", leave=True) as pbar:
@@ -108,10 +129,10 @@ def progress_iter(stages: Stages) -> Iterator["Stage"]:
             pbar.bar_format = desc
             pbar.refresh()
 
-        for stage in pbar:
+        for item in pbar:
             if total > 1:
-                pbar.set_msg(f"{stage.outs[0]}")
-            yield stage
+                pbar.set_msg(f"{item.stage.outs[0]}")
+            yield item
             if total == 1:  # restore bar format for stats
                 # pylint: disable=no-member
                 pbar.bar_format = pbar.BAR_FMT_DEFAULT
@@ -169,12 +190,14 @@ def add(
     desc = "Collecting targets"
     stages_it = create_stages(repo, add_targets, fname, transfer, **kwargs)
     stages = list(ui.progress(stages_it, desc=desc, unit="file"))
+
+    stages_list = [stage for stage, _ in stages]
     msg = "Collecting stages from the workspace"
-    with translate_graph_error(stages), ui.status(msg) as status:
+    with translate_graph_error(stages_list), ui.status(msg) as status:
         # remove existing stages that are to-be replaced with these
         # new stages for the graph checks.
         repo.check_graph(
-            stages=stages, callback=lambda: status.update("Checking graph")
+            stages=stages_list, callback=lambda: status.update("Checking graph")
         )
 
     odb = None
@@ -182,18 +205,18 @@ def add(
         odb = repo.cloud.get_remote_odb(kwargs.get("remote"), "add")
 
     with warn_link_failures() as link_failures:
-        for stage, source in zip(progress_iter(stages), sources):
+        for (stage, output_exists), source in zip(progress_iter(stages), sources):
+            out = stage.outs[0]
             if to_remote or to_cache:
                 stage.transfer(source, to_remote=to_remote, odb=odb, **kwargs)
             else:
                 try:
-                    stage.save()
-                    if not no_commit:
-                        stage.commit()
+                    path = out.fs.path.abspath(source) if output_exists else None
+                    stage.add_outs(path, no_commit=no_commit)
                 except CacheLinkError:
                     link_failures.append(str(stage.relpath))
             stage.dump()
-    return stages
+    return stages_list
 
 
 LARGE_DIR_RECURSIVE_ADD_WARNING = (
@@ -247,24 +270,34 @@ def create_stages(
     fname: Optional[str] = None,
     transfer: bool = False,
     external: bool = False,
+    force: bool = False,
     **kwargs: Any,
-) -> Iterator["Stage"]:
+) -> Iterator[StageInfo]:
     for target in targets:
         if kwargs.get("out"):
-            target = resolve_output(target, kwargs["out"])
+            target = resolve_output(target, kwargs["out"], force=force)
         path, wdir, out = resolve_paths(
             repo, target, always_local=transfer and not kwargs.get("out")
         )
 
-        stage = repo.stage.create(
-            single_stage=True,
-            validate=False,
-            fname=fname or path,
-            wdir=wdir,
-            outs=[out],
-            external=external,
-        )
+        try:
+            (out_obj,) = repo.find_outs_by_path(target, strict=False)
+            stage = out_obj.stage
+            if not stage.is_data_source:
+                raise DvcException(f"cannot update {out!r}: not a data source")
+            output_exists = True
+        except OutputNotFoundError:
+            stage = repo.stage.create(
+                single_stage=True,
+                validate=False,
+                fname=fname or path,
+                wdir=wdir,
+                outs=[out],
+                external=external,
+                force=force,
+            )
+            output_exists = False
 
         out_obj = stage.outs[0]
         out_obj.annot.update(**kwargs)
-        yield stage
+        yield StageInfo(stage, output_exists)

dvc/repo/artifacts.py

@@ -1,19 +1,26 @@
 import logging
-import os
 import re
-from typing import TYPE_CHECKING, Dict
+from pathlib import Path
+from typing import Dict, Optional
 
 from dvc.annotations import Artifact
-from dvc.dvcfile import FileMixin
+from dvc.dvcfile import PROJECT_FILE
+from dvc.exceptions import InvalidArgumentError
+from dvc.repo import Repo
 from dvc.utils import relpath
-
-if TYPE_CHECKING:
-    from dvc.repo import Repo
+from dvc.utils.serialize import modify_yaml
 
 logger = logging.getLogger(__name__)
 
 
-NAME_RE = re.compile(r"^[a-z]([a-z0-9-]*[a-z0-9])?$")
+# Constants are taken from GTO.
+# When we make it a dependency, we can import them instead
+SEPARATOR_IN_NAME = ":"
+DIRNAME = r"[a-z0-9-_./]+"
+NAME = r"[a-z0-9]([a-z0-9-/]*[a-z0-9])?"
+NAME_RE = re.compile(f"^{NAME}$")
+FULLNAME = f"((?P<dirname>{DIRNAME}){SEPARATOR_IN_NAME})?(?P<name>{NAME})"
+FULLNAME_RE = re.compile(f"^{FULLNAME}$")
 
 
 def name_is_compatible(name: str) -> bool:
@@ -22,22 +29,23 @@ def name_is_compatible(name: str) -> bool:
 
 def check_name_format(name: str) -> None:
     if not name_is_compatible(name):
-        logger.warning(
-            "Can't use '%s' as artifact name (ID)."
+        raise InvalidArgumentError(
+            f"Can't use '{name}' as artifact name (ID)."
             " You can use letters and numbers, and use '-' as separator"
-            " (but not at the start or end). The first character must be a letter.",
-            name,
+            " (but not at the start or end)."
         )
 
 
-class ArtifactsFile(FileMixin):
-    from dvc.schema import SINGLE_ARTIFACT_SCHEMA as SCHEMA
-
-    def dump(self, stage, **kwargs):
-        raise NotImplementedError
-
-    def merge(self, ancestor, other, allowed=None):
-        raise NotImplementedError
+def check_for_nested_dvc_repo(dvcfile: Path):
+    if dvcfile.is_absolute():
+        raise InvalidArgumentError("Use relative path to dvc.yaml.")
+    path = dvcfile.parent
+    while path.name:
+        if (path / Repo.DVC_DIR).is_dir():
+            raise InvalidArgumentError(
+                f"Nested DVC repos like {path} are not supported."
+            )
+        path = path.parent
 
 
 class Artifacts:
@@ -50,18 +58,30 @@ class Artifacts:
             dvcfile,
             dvcfile_artifacts,
         ) in self.repo.index._artifacts.items():  # pylint: disable=protected-access
-            # read the artifacts.yaml file if needed
-            if isinstance(dvcfile_artifacts, str):
-                dvcfile_artifacts = ArtifactsFile(
-                    self.repo,
-                    os.path.join(os.path.dirname(dvcfile), dvcfile_artifacts),
-                    verify=False,
-                ).load()
-            if not dvcfile_artifacts:
-                continue
             dvcyaml = relpath(dvcfile, self.repo.root_dir)
             artifacts[dvcyaml] = {}
             for name, value in dvcfile_artifacts.items():
-                check_name_format(name)
+                try:
+                    check_name_format(name)
+                except InvalidArgumentError as e:
+                    logger.warning(e.msg)
                 artifacts[dvcyaml][name] = Artifact(**value)
         return artifacts
+
+    def add(self, name: str, artifact: Artifact, dvcfile: Optional[str] = None):
+        with self.repo.scm_context(quiet=True):
+            check_name_format(name)
+            dvcyaml = Path(dvcfile or PROJECT_FILE)
+            check_for_nested_dvc_repo(
+                dvcyaml.relative_to(self.repo.root_dir)
+                if dvcyaml.is_absolute()
+                else dvcyaml
+            )
+
+            with modify_yaml(dvcyaml) as data:
+                artifacts = data.setdefault("artifacts", {})
+                artifacts.update({name: artifact.to_dict()})
+
+            self.repo.scm_context.track_file(dvcfile)
+
+        return artifacts.get(name)

dvc/repo/data.py

@@ -172,7 +172,6 @@ def _diff_index_to_wtree(repo: "Repo", **kwargs: Any) -> Dict[str, List[str]]:
             repo.index.data["repo"],
             workspace,
             not_in_cache=True,
-            not_in_remote=True,
             **kwargs,
         )
 

dvc/repo/experiments/__init__.py

@@ -119,6 +119,7 @@ class Experiments:
         self,
         tmp_dir: bool = False,
         copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
         **kwargs,
     ):
         """Reproduce and checkout a single (standalone) experiment."""
@@ -126,7 +127,9 @@ class Experiments:
             self.tempdir_queue if tmp_dir else self.workspace_queue
         )
         self.queue_one(exp_queue, **kwargs)
-        results = self._reproduce_queue(exp_queue, copy_paths=copy_paths)
+        results = self._reproduce_queue(
+            exp_queue, copy_paths=copy_paths, message=message
+        )
         exp_rev = first(results)
         if exp_rev is not None:
             self._log_reproduced(results, tmp_dir=tmp_dir)
@@ -349,7 +352,11 @@ class Experiments:
 
     @unlocked_repo
     def _reproduce_queue(
-        self, queue: "BaseStashQueue", copy_paths: Optional[List[str]] = None, **kwargs
+        self,
+        queue: "BaseStashQueue",
+        copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
+        **kwargs,
     ) -> Dict[str, str]:
         """Reproduce queued experiments.
 
@@ -360,7 +367,7 @@ class Experiments:
             dict mapping successfully reproduced experiment revs to their
             results.
         """
-        exec_results = queue.reproduce(copy_paths=copy_paths)
+        exec_results = queue.reproduce(copy_paths=copy_paths, message=message)
 
         results: Dict[str, str] = {}
         for _, exp_result in exec_results.items():
@@ -442,17 +449,6 @@ class Experiments:
             result[rev] = name
         return result
 
-    def get_running_exps(self, fetch_refs: bool = True) -> Dict[str, Dict]:
-        """Return info for running experiments."""
-        result = {}
-        for queue in (
-            self.workspace_queue,
-            self.tempdir_queue,
-            self.celery_queue,
-        ):
-            result.update(queue.get_running_exps(fetch_refs))
-        return result
-
     def apply(self, *args, **kwargs):
         from dvc.repo.experiments.apply import apply
 

dvc/repo/experiments/brancher.py

@@ -0,0 +1,47 @@
+from contextlib import ExitStack, contextmanager
+from typing import TYPE_CHECKING, Iterator, Tuple
+
+from dvc.repo.experiments.exceptions import InvalidExpRevError
+from dvc.scm import RevError
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+
+@contextmanager
+def switch_repo(
+    repo: "Repo",
+    rev: str,
+) -> Iterator[Tuple["Repo", str]]:
+    """Return a repo instance (brancher) switched to rev.
+
+    If rev is the name of a running experiment, the returned instance will be
+    the live repo wherever the experiment is running.
+
+    NOTE: This will not resolve git SHA's that only exist in queued exp workspaces
+    (it will only match queued exp names).
+    """
+    try:
+        with repo.switch(rev):
+            yield repo, rev
+        return
+    except RevError as exc:
+        orig_exc = exc
+    exps = repo.experiments
+
+    if rev == exps.workspace_queue.get_running_exp():
+        yield repo, "workspace"
+        return
+
+    for queue in (exps.tempdir_queue, exps.celery_queue):
+        try:
+            active_repo = queue.active_repo(rev)
+        except InvalidExpRevError:
+            continue
+        stack = ExitStack()
+        stack.enter_context(active_repo)
+        stack.enter_context(active_repo.switch("workspace"))
+        with stack:
+            yield active_repo, rev
+        return
+    raise orig_exc

dvc/repo/experiments/collect.py

@@ -0,0 +1,380 @@
+import itertools
+import logging
+import os
+from datetime import datetime
+from typing import (
+    TYPE_CHECKING,
+    Collection,
+    Dict,
+    Iterable,
+    Iterator,
+    List,
+    Optional,
+    Union,
+)
+
+from funcy import first
+from scmrepo.exceptions import SCMError as InnerSCMError
+
+from dvc.scm import Git, SCMError, iter_revs
+
+from .exceptions import InvalidExpRefError
+from .refs import EXEC_BRANCH, ExpRefInfo
+from .serialize import ExpRange, ExpState, SerializableError, SerializableExp
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+
+    from .cache import ExpCache
+
+logger = logging.getLogger(__name__)
+
+
+def collect_rev(
+    repo: "Repo",
+    rev: str,
+    param_deps: bool = False,
+    force: bool = False,
+    cache: Optional["ExpCache"] = None,
+    **kwargs,
+) -> ExpState:
+    """Collect experiment state for the given revision.
+
+    Exp will be loaded from cache when available unless rev is 'workspace' or
+    force is set.
+    """
+    from dvc.fs import LocalFileSystem
+
+    cache = cache or repo.experiments.cache
+    assert cache
+    # TODO: support filtering serialized exp when param_deps is set
+    if rev != "workspace" and not (force or param_deps):
+        cached_exp = cache.get(rev)
+        if cached_exp:
+            if isinstance(cached_exp, SerializableError):
+                return ExpState(rev=rev, error=cached_exp)
+            return ExpState(rev=rev, data=cached_exp)
+    if rev == "workspace" and isinstance(repo.fs, LocalFileSystem):
+        orig_cwd: Optional[str] = os.getcwd()
+        os.chdir(repo.root_dir)
+    else:
+        orig_cwd = None
+    try:
+        data = _collect_rev(
+            repo,
+            rev,
+            param_deps=param_deps,
+            force=force,
+            **kwargs,
+        )
+        if not (rev == "workspace" or param_deps or data.contains_error):
+            cache.put(data, force=True)
+        return ExpState(rev=rev, data=data)
+    except Exception as exc:  # noqa: BLE001, pylint: disable=broad-except
+        logger.debug("", exc_info=True)
+        error = SerializableError(str(exc), type(exc).__name__)
+        return ExpState(rev=rev, error=error)
+    finally:
+        if orig_cwd:
+            os.chdir(orig_cwd)
+
+
+def _collect_rev(
+    repo: "Repo",
+    rev: str,
+    param_deps: bool = False,
+    **kwargs,
+) -> SerializableExp:
+    with repo.switch(rev) as rev:
+        if rev == "workspace":
+            timestamp: Optional[datetime] = None
+        else:
+            commit = repo.scm.resolve_commit(rev)
+            timestamp = datetime.fromtimestamp(commit.commit_time)
+
+        return SerializableExp.from_repo(
+            repo,
+            rev=rev,
+            param_deps=param_deps,
+            timestamp=timestamp,
+        )
+
+
+def collect_branch(
+    repo: "Repo",
+    rev: str,
+    end_rev: Optional[str] = None,
+    **kwargs,
+) -> Iterator["ExpState"]:
+    """Iterate over exp states in a Git branch.
+
+    Git branch will be traversed in reverse, starting from rev.
+
+    Args:
+        rev: Branch tip (head).
+        end_rev: If specified, traversal will stop when end_rev is reached
+            (exclusive, end_rev will not be collected).
+    """
+    try:
+        for branch_rev in repo.scm.branch_revs(rev, end_rev):
+            yield collect_rev(repo, branch_rev, **kwargs)
+    except (SCMError, InnerSCMError):
+        pass
+
+
+def collect_exec_branch(
+    repo: "Repo",
+    baseline_rev: str,
+    **kwargs,
+) -> Iterator["ExpState"]:
+    """Iterate over active experiment branch for the current executor."""
+    last_rev = repo.scm.get_ref(EXEC_BRANCH) or repo.scm.get_rev()
+    last_rev = repo.scm.get_rev()
+    yield collect_rev(repo, "workspace", **kwargs)
+    if last_rev != baseline_rev:
+        yield from collect_branch(repo, last_rev, baseline_rev, **kwargs)
+
+
+def collect_queued(
+    repo: "Repo",
+    baseline_revs: Collection[str],
+    **kwargs,
+) -> Dict[str, List["ExpRange"]]:
+    """Collect queued experiments derived from the specified revisions.
+
+    Args:
+        repo: Repo.
+        baseline_revs: Resolved baseline Git SHAs.
+
+    Returns:
+        Dict mapping baseline revision to list of queued experiments.
+    """
+    if not baseline_revs:
+        return {}
+    return repo.experiments.celery_queue.collect_queued_data(baseline_revs, **kwargs)
+
+
+def collect_active(
+    repo: "Repo",
+    baseline_revs: Collection[str],
+    **kwargs,
+) -> Dict[str, List["ExpRange"]]:
+    """Collect active (running) experiments derived from the specified revisions.
+
+    Args:
+        repo: Repo.
+        baseline_revs: Resolved baseline Git SHAs.
+
+    Returns:
+        Dict mapping baseline revision to list of active experiments.
+    """
+    if not baseline_revs:
+        return {}
+    result: Dict[str, List["ExpRange"]] = {}
+    exps = repo.experiments
+    for queue in (exps.workspace_queue, exps.tempdir_queue, exps.celery_queue):
+        for baseline, active_exps in queue.collect_active_data(
+            baseline_revs, **kwargs
+        ).items():
+            if baseline in result:
+                result[baseline].extend(active_exps)
+            else:
+                result[baseline] = list(active_exps)
+    return result
+
+
+def collect_failed(
+    repo: "Repo",
+    baseline_revs: Collection[str],
+    **kwargs,
+) -> Dict[str, List["ExpRange"]]:
+    """Collect failed experiments derived from the specified revisions.
+
+    Args:
+        repo: Repo.
+        baseline_revs: Resolved baseline Git SHAs.
+
+    Returns:
+        Dict mapping baseline revision to list of active experiments.
+    """
+    if not baseline_revs:
+        return {}
+    return repo.experiments.celery_queue.collect_failed_data(baseline_revs, **kwargs)
+
+
+def collect_successful(
+    repo: "Repo",
+    baseline_revs: Collection[str],
+    **kwargs,
+) -> Dict[str, List["ExpRange"]]:
+    """Collect successful experiments derived from the specified revisions.
+
+    Args:
+        repo: Repo.
+        baseline_revs: Resolved baseline Git SHAs.
+
+    Returns:
+        Dict mapping baseline revision to successful experiments.
+    """
+    result: Dict[str, List["ExpRange"]] = {}
+    for baseline_rev in baseline_revs:
+        result[baseline_rev] = list(_collect_baseline(repo, baseline_rev, **kwargs))
+    return result
+
+
+def _collect_baseline(
+    repo: "Repo",
+    baseline_rev: str,
+    **kwargs,
+) -> Iterator["ExpRange"]:
+    """Iterate over experiments derived from a baseline revision.
+
+    Args:
+        repo: Repo.
+        baseline_revs: Resolved baseline Git SHAs.
+
+    Yields:
+        Tuple of (timestamp, exp_range).
+    """
+    ref_info = ExpRefInfo(baseline_sha=baseline_rev)
+    refs: Optional[Iterable[str]] = kwargs.get("refs")
+    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:
+            ref_info = ExpRefInfo.from_ref(ref)
+            exp_rev = repo.scm.get_ref(ref)
+            if not exp_rev:
+                continue
+        except (InvalidExpRefError, SCMError, InnerSCMError):
+            continue
+        exps = list(collect_branch(repo, exp_rev, baseline_rev, **kwargs))
+        if exps:
+            exps[0].name = ref_info.name
+            yield ExpRange(exps, name=ref_info.name)
+
+
+def collect(
+    repo: "Repo",
+    revs: Union[List[str], str, None] = None,
+    all_branches: bool = False,
+    all_tags: bool = False,
+    all_commits: bool = False,
+    num: int = 1,
+    hide_queued: bool = False,
+    hide_failed: bool = False,
+    sha_only: bool = False,
+    **kwargs,
+) -> List["ExpState"]:
+    """Collect baseline revisions and derived experiments."""
+    assert isinstance(repo.scm, Git)
+    if repo.scm.no_commits:
+        return []
+    if not any([revs, all_branches, all_tags, all_commits]):
+        revs = ["HEAD"]
+    if isinstance(revs, str):
+        revs = [revs]
+    cached_refs = list(repo.scm.iter_refs())
+    baseline_revs = list(
+        iter_revs(
+            repo.scm,
+            revs=revs,
+            num=num,
+            all_branches=all_branches,
+            all_tags=all_tags,
+            all_commits=all_commits,
+        )
+    )
+    if sha_only:
+        baseline_names: Dict[str, Optional[str]] = {}
+    else:
+        baseline_names = _describe(repo.scm, baseline_revs, refs=cached_refs)
+
+    workspace_data = collect_rev(repo, "workspace", **kwargs)
+    result: List["ExpState"] = [workspace_data]
+    queued = collect_queued(repo, baseline_revs, **kwargs) if not hide_queued else {}
+    active = collect_active(repo, baseline_revs, **kwargs)
+    failed = collect_failed(repo, baseline_revs, **kwargs) if not hide_failed else {}
+    successful = collect_successful(repo, baseline_revs, **kwargs)
+
+    for baseline_rev in baseline_revs:
+        baseline_data = collect_rev(repo, baseline_rev)
+        experiments = list(
+            itertools.chain.from_iterable(
+                _sorted_ranges(collected.get(baseline_rev, []))
+                for collected in (active, successful, queued, failed)
+            )
+        )
+        result.append(
+            ExpState(
+                rev=baseline_rev,
+                name=baseline_names.get(baseline_rev),
+                data=baseline_data.data,
+                error=baseline_data.error,
+                experiments=experiments if experiments else None,
+            )
+        )
+    return result
+
+
+def _describe(
+    scm: "Git",
+    revs: Iterable[str],
+    refs: Optional[Iterable[str]] = None,
+) -> Dict[str, Optional[str]]:
+    """Describe revisions using a tag, branch.
+
+    The first matching name will be returned for each rev. Names are preferred in this
+    order:
+        - current branch (if rev matches HEAD and HEAD is a branch)
+        - tags
+        - branches
+
+    Returns:
+        Dict mapping revisions from revs to a name.
+    """
+
+    head_rev = scm.get_rev()
+    head_ref = scm.get_ref("HEAD", follow=False)
+    if head_ref and head_ref.startswith("refs/heads/"):
+        head_branch = head_ref[len("refs/heads/") :]
+    else:
+        head_branch = None
+
+    tags = {}
+    branches = {}
+    ref_it = iter(refs) if refs else scm.iter_refs()
+    for ref in ref_it:
+        is_tag = ref.startswith("refs/tags/")
+        is_branch = ref.startswith("refs/heads/")
+        if not (is_tag or is_branch):
+            continue
+        rev = scm.get_ref(ref)
+        if not rev:
+            logger.debug("unresolved ref %s", ref)
+            continue
+        if is_tag and rev not in tags:
+            tags[rev] = ref[len("refs/tags/") :]
+        if is_branch and rev not in branches:
+            branches[rev] = ref[len("refs/heads/") :]
+    names: Dict[str, Optional[str]] = {}
+    for rev in revs:
+        if rev == head_rev and head_branch:
+            names[rev] = head_branch
+        else:
+            names[rev] = tags.get(rev) or branches.get(rev)
+    return names
+
+
+def _sorted_ranges(exp_ranges: Iterable["ExpRange"]) -> List["ExpRange"]:
+    """Return list of ExpRange sorted by timestamp."""
+
+    def _head_timestamp(exp_range: "ExpRange") -> datetime:
+        head_exp = first(exp_range.revs)
+        if head_exp and head_exp.data and head_exp.data.timestamp:
+            return head_exp.data.timestamp
+        return datetime.fromtimestamp(0)
+
+    return sorted(exp_ranges, key=_head_timestamp, reverse=True)

dvc/repo/experiments/diff.py

@@ -7,7 +7,7 @@ logger = logging.getLogger(__name__)
 
 
 def diff(repo, *args, a_rev=None, b_rev=None, param_deps=False, **kwargs):
-    from dvc.repo.experiments.show import collect_experiment_commit
+    from dvc.repo.experiments.collect import collect_rev
     from dvc.scm import resolve_rev
 
     if repo.scm.no_commits:
@@ -15,22 +15,21 @@ def diff(repo, *args, a_rev=None, b_rev=None, param_deps=False, **kwargs):
 
     if a_rev:
         rev = resolve_rev(repo.scm, a_rev)
-        old = collect_experiment_commit(repo, rev, param_deps=param_deps)
     else:
-        old = collect_experiment_commit(repo, "HEAD", param_deps=param_deps)
+        rev = resolve_rev(repo.scm, "HEAD")
+    old = collect_rev(repo, rev, param_deps=param_deps)
 
     if b_rev:
         rev = resolve_rev(repo.scm, b_rev)
-        new = collect_experiment_commit(repo, rev, param_deps=param_deps)
     else:
-        new = collect_experiment_commit(repo, "workspace", param_deps=param_deps)
+        rev = "workspace"
+    new = collect_rev(repo, rev, param_deps=param_deps)
 
     with_unchanged = kwargs.pop("all", False)
-
     return {
         key: _diff(
-            format_dict(old.get("data", {}).get(key, {})),
-            format_dict(new.get("data", {}).get(key, {})),
+            format_dict(getattr(old.data, key, {})),
+            format_dict(getattr(new.data, key, {})),
             with_unchanged=with_unchanged,
         )
         for key in ["metrics", "params"]

dvc/repo/experiments/exceptions.py

@@ -114,3 +114,10 @@ class UnresolvedRunningExpNamesError(UnresolvedExpNamesError):
 
 class ExpQueueEmptyError(DvcException):
     pass
+
+
+class ExpNotStartedError(DvcException):
+    def __init__(self, name: str):
+        super().__init__(
+            f"Queued experiment '{name}' exists but has not started running yet"
+        )

dvc/repo/experiments/executor/base.py

@@ -7,12 +7,14 @@ from contextlib import contextmanager
 from dataclasses import asdict, dataclass
 from enum import IntEnum
 from functools import partial
+from itertools import chain
 from typing import (
     TYPE_CHECKING,
     Any,
     Callable,
     Dict,
     Iterable,
+    Iterator,
     List,
     NamedTuple,
     Optional,
@@ -41,6 +43,7 @@ from dvc.stage.serialize import to_lockfile
 from dvc.ui import ui
 from dvc.utils import dict_sha256, env2bool, relpath
 from dvc.utils.fs import remove
+from dvc.utils.studio import env_to_config
 
 if TYPE_CHECKING:
     from queue import Queue
@@ -259,12 +262,23 @@ class BaseExecutor(ABC):
             **kwargs,
         )
 
+    @classmethod
+    def _get_top_level_paths(cls, repo: "Repo") -> List["str"]:
+        return list(
+            chain(
+                _collect_top_level_metrics(repo),
+                _collect_top_level_params(repo),
+                repo.index._plot_sources,  # pylint: disable=protected-access
+            )
+        )
+
     @classmethod
     def save(
         cls,
         info: "ExecutorInfo",
         force: bool = False,
         include_untracked: Optional[List[str]] = None,
+        message: Optional[str] = None,
     ) -> ExecutorResult:
         from dvc.dvcfile import LOCK_FILE
         from dvc.repo import Repo
@@ -280,11 +294,7 @@ class BaseExecutor(ABC):
             os.chdir(dvc.root_dir)
 
         include_untracked = include_untracked or []
-        include_untracked.extend(_collect_top_level_metrics(dvc))
-        include_untracked.extend(_collect_top_level_params(dvc))
-        include_untracked.extend(
-            dvc.index._plot_sources  # pylint: disable=protected-access
-        )
+        include_untracked.extend(cls._get_top_level_paths(dvc))
         # dvc repro automatically stages dvc.lock. Running redundant `git add`
         # on it causes an error when exiting the detached head context.
         if LOCK_FILE in dvc.scm.untracked_files():
@@ -300,6 +310,7 @@ class BaseExecutor(ABC):
                 exp_hash,
                 exp_name=info.name,
                 force=force,
+                message=message,
             )
             ref: Optional[str] = dvc.scm.get_ref(EXEC_BRANCH, follow=False)
             exp_ref = ExpRefInfo.from_ref(ref) if ref else None
@@ -453,6 +464,7 @@ class BaseExecutor(ABC):
         log_errors: bool = True,
         log_level: Optional[int] = None,
         copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
         **kwargs,
     ) -> "ExecutorResult":
         """Run dvc repro and return the result.
@@ -485,11 +497,15 @@ class BaseExecutor(ABC):
         exp_ref: Optional["ExpRefInfo"] = None
         repro_force: bool = False
 
+        if info.name:
+            ui.write(f"Reproducing experiment '{info.name}'")
+
         with cls._repro_dvc(
             info,
             infofile,
             log_errors=log_errors,
             copy_paths=copy_paths,
+            message=message,
             **kwargs,
         ) as dvc:
             if auto_push:
@@ -540,6 +556,7 @@ class BaseExecutor(ABC):
                 info.name,
                 repro_force or checkpoint_reset,
             )
+
             stages = dvc_reproduce(
                 dvc,
                 *args,
@@ -547,6 +564,9 @@ class BaseExecutor(ABC):
                 checkpoint_func=checkpoint_func,
                 **kwargs,
             )
+            if paths := cls._get_top_level_paths(dvc):
+                logger.debug("Staging top-level files: %s", paths)
+                dvc.scm_context.add(paths)
 
             exp_hash = cls.hash_exp(stages)
             if not repro_dry:
@@ -558,6 +578,7 @@ class BaseExecutor(ABC):
                     auto_push,
                     git_remote,
                     repro_force,
+                    message=message,
                 )
                 info.result_hash = exp_hash
                 info.result_ref = ref
@@ -578,6 +599,7 @@ class BaseExecutor(ABC):
         auto_push,
         git_remote,
         repro_force,
+        message: Optional[str] = None,
     ) -> Tuple[Optional[str], Optional["ExpRefInfo"], bool]:
         is_checkpoint = any(stage.is_checkpoint for stage in stages)
         cls.commit(
@@ -586,6 +608,7 @@ class BaseExecutor(ABC):
             exp_name=info.name,
             force=repro_force,
             checkpoint=is_checkpoint,
+            message=message,
         )
         if auto_push:
             cls._auto_push(dvc, dvc.scm, git_remote)
@@ -613,8 +636,9 @@ class BaseExecutor(ABC):
         infofile: Optional[str] = None,
         log_errors: bool = True,
         copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
         **kwargs,
-    ):
+    ) -> Iterator["Repo"]:
         from dvc_studio_client.post_live_metrics import post_live_metrics
 
         from dvc.repo import Repo
@@ -636,13 +660,23 @@ class BaseExecutor(ABC):
             else:
                 os.chdir(dvc.root_dir)
 
+            args_path = os.path.join(dvc.tmp_dir, cls.PACKED_ARGS_FILE)
+            if os.path.exists(args_path):
+                _, kwargs = cls.unpack_repro_args(args_path)
+            dvc_studio_config = dvc.config.get("studio")
+            # set missing config options using saved config
+            # inferring repo url will fail if not set here
+            run_env_config = env_to_config(kwargs.get("run_env", {}))
+            dvc_studio_config = {**run_env_config, **dvc_studio_config}
             try:
                 post_live_metrics(
                     "start",
                     info.baseline_rev,
-                    info.name,
+                    info.name,  # type: ignore[arg-type]
                     "dvc",
                     params=to_studio_params(dvc.params.show()),
+                    dvc_studio_config=dvc_studio_config,
+                    message=message,
                 )
                 logger.debug("Running repro in '%s'", os.getcwd())
                 yield dvc
@@ -664,10 +698,11 @@ class BaseExecutor(ABC):
                 post_live_metrics(
                     "done",
                     info.baseline_rev,
-                    info.name,
+                    info.name,  # type: ignore[arg-type]
                     "dvc",
                     experiment_rev=dvc.experiments.scm.get_ref(EXEC_BRANCH),
                     metrics=get_in(dvc.metrics.show(), ["", "data"]),
+                    dvc_studio_config=dvc_studio_config,
                 )
 
                 if infofile is not None:
@@ -681,7 +716,7 @@ class BaseExecutor(ABC):
             args, kwargs = cls.unpack_repro_args(args_path)
             remove(args_path)
             # explicitly git rm/unstage the args file
-            dvc.scm.add([args_path])
+            dvc.scm.add([args_path], force=True)
         else:
             args = []
             kwargs = {}
@@ -739,6 +774,7 @@ class BaseExecutor(ABC):
         exp_name: Optional[str] = None,
         force: bool = False,
         checkpoint: bool = False,
+        message: Optional[str] = None,
     ):
         """Commit stages as an experiment and return the commit SHA."""
         rev = scm.get_rev()
@@ -768,7 +804,8 @@ class BaseExecutor(ABC):
                 logger.debug("Commit to new experiment branch '%s'", branch)
 
         scm.add([], update=True)
-        scm.commit(f"dvc: commit experiment {exp_hash}", no_verify=True)
+        message = message or f"dvc: commit experiment {exp_hash}"
+        scm.commit(message, no_verify=True)
         new_rev = scm.get_rev()
         if check_conflict:
             new_rev = cls._raise_ref_conflict(scm, branch, new_rev, checkpoint)

dvc/repo/experiments/executor/local.py

@@ -17,7 +17,7 @@ from dvc.repo.experiments.refs import (
     EXEC_HEAD,
     EXEC_MERGE,
     EXEC_NAMESPACE,
-    EXPS_TEMP,
+    TEMP_NAMESPACE,
 )
 from dvc.repo.experiments.utils import EXEC_TMP_DIR, get_exp_rwlock
 from dvc.scm import SCM, Git
@@ -90,9 +90,9 @@ class TempDirExecutor(BaseLocalExecutor):
         if infofile:
             self.info.dump_json(infofile)
 
-        temp_head = f"{EXPS_TEMP}/head-{uuid()}"
-        temp_merge = f"{EXPS_TEMP}/merge-{uuid()}"
-        temp_baseline = f"{EXPS_TEMP}/baseline-{uuid()}"
+        temp_head = f"{TEMP_NAMESPACE}/head-{uuid()}"
+        temp_merge = f"{TEMP_NAMESPACE}/merge-{uuid()}"
+        temp_baseline = f"{TEMP_NAMESPACE}/baseline-{uuid()}"
 
         temp_ref_dict = {
             temp_head: entry.head_rev,

dvc/repo/experiments/executor/ssh.py

@@ -243,6 +243,7 @@ class SSHExecutor(BaseExecutor):
         log_errors: bool = True,
         log_level: Optional[int] = None,
         copy_paths: Optional[List[str]] = None,  # noqa: ARG003
+        message: Optional[str] = None,  # noqa: ARG003
         **kwargs,
     ) -> "ExecutorResult":
         """Reproduce an experiment on a remote machine over SSH.

dvc/repo/experiments/ls.py

@@ -1,6 +1,6 @@
 import logging
 from collections import defaultdict
-from typing import Optional
+from typing import List, Optional, Union
 
 from dvc.repo import locked
 from dvc.repo.scm_context import scm_context
@@ -15,14 +15,17 @@ logger = logging.getLogger(__name__)
 @scm_context
 def ls(
     repo,
-    rev: Optional[str] = None,
+    rev: Optional[Union[List[str], str]] = None,
     all_commits: bool = False,
     num: int = 1,
     git_remote: Optional[str] = None,
 ):
     rev_set = None
     if not all_commits:
-        revs = iter_revs(repo.scm, [rev or "HEAD"], num)
+        rev = rev or "HEAD"
+        if isinstance(rev, str):
+            rev = [rev]
+        revs = iter_revs(repo.scm, rev, num)
         rev_set = set(revs.keys())
     ref_info_dict = exp_refs_by_baseline(repo.scm, rev_set, git_remote)
 

dvc/repo/experiments/pull.py

@@ -23,7 +23,7 @@ def pull(  # noqa: C901
     git_remote: str,
     exp_names: Union[Iterable[str], str],
     all_commits=False,
-    rev: Optional[str] = None,
+    rev: Optional[Union[List[str], str]] = None,
     num=1,
     force: bool = False,
     pull_cache: bool = False,
@@ -49,7 +49,9 @@ def pull(  # noqa: C901
                 raise UnresolvedExpNamesError(unresolved_exp_names)
 
         if rev:
-            rev_dict = iter_revs(repo.scm, [rev], num)
+            if isinstance(rev, str):
+                rev = [rev]
+            rev_dict = iter_revs(repo.scm, rev, num)
             rev_set = set(rev_dict.keys())
             ref_info_dict = exp_refs_by_baseline(repo.scm, rev_set, git_remote)
             for _, ref_info_list in ref_info_dict.items():

dvc/repo/experiments/push.py

@@ -14,11 +14,13 @@ from typing import (
 from funcy import compact, group_by
 from scmrepo.git.backend.base import SyncStatus
 
+from dvc.env import DVC_STUDIO_TOKEN, DVC_STUDIO_URL
 from dvc.exceptions import DvcException
 from dvc.repo import locked
 from dvc.repo.scm_context import scm_context
 from dvc.scm import Git, TqdmGit, iter_revs
 from dvc.utils import env2bool
+from dvc.utils.collections import ensure_list
 
 from .exceptions import UnresolvedExpNamesError
 from .refs import ExpRefInfo
@@ -39,21 +41,18 @@ class UploadError(DvcException):
 def notify_refs_to_studio(
     repo: "Repo", git_remote: str, **refs: List[str]
 ) -> Optional[str]:
-    config = repo.config["feature"]
+    import os
+
+    config = repo.config["studio"]
     refs = compact(refs)
     if not refs or env2bool("DVC_TEST"):
         return None
 
-    if not (config.get("studio_token") or config["push_exp_to_studio"]):
-        logger.debug(
-            "Either feature.studio_token or feature.push_exp_to_studio config "
-            "needs to be set."
-        )
-        return None
-
-    import os
-
-    token = os.environ.get("STUDIO_TOKEN") or config.get("studio_token")
+    token = (
+        os.environ.get(DVC_STUDIO_TOKEN)
+        or os.environ.get("STUDIO_TOKEN")
+        or config.get("token")
+    )
     if not token:
         logger.debug("Studio token not found.")
         return None
@@ -63,19 +62,44 @@ def notify_refs_to_studio(
     from dvc.utils import studio
 
     _, repo_url = get_remote_repo(repo.scm.dulwich.repo, git_remote)
-    studio_url = config.get("studio_url")
+    studio_url = os.environ.get(DVC_STUDIO_URL) or config.get("url")
     d = studio.notify_refs(repo_url, token, base_url=studio_url, **refs)
     return d.get("url")
 
 
+def exp_refs_from_names(scm: "Git", exp_names: List[str]) -> Set["ExpRefInfo"]:
+    exp_ref_set = set()
+    exp_ref_dict = resolve_name(scm, exp_names)
+    unresolved_exp_names = []
+    for exp_name, exp_ref in exp_ref_dict.items():
+        if exp_ref is None:
+            unresolved_exp_names.append(exp_name)
+        else:
+            exp_ref_set.add(exp_ref)
+
+    if unresolved_exp_names:
+        raise UnresolvedExpNamesError(unresolved_exp_names)
+    return exp_ref_set
+
+
+def exp_refs_from_rev(scm: "Git", rev: List[str], num: int = 1) -> Set["ExpRefInfo"]:
+    exp_ref_set = set()
+    rev_dict = iter_revs(scm, rev, num)
+    rev_set = set(rev_dict.keys())
+    ref_info_dict = exp_refs_by_baseline(scm, rev_set)
+    for _, ref_info_list in ref_info_dict.items():
+        exp_ref_set.update(ref_info_list)
+    return exp_ref_set
+
+
 @locked
 @scm_context
-def push(  # noqa: C901
+def push(
     repo: "Repo",
     git_remote: str,
-    exp_names: Union[Iterable[str], str],
+    exp_names: Union[List[str], str],
     all_commits: bool = False,
-    rev: Optional[str] = None,
+    rev: Optional[Union[List[str], str]] = None,
     num: int = 1,
     force: bool = False,
     push_cache: bool = False,
@@ -85,29 +109,12 @@ def push(  # noqa: C901
     assert isinstance(repo.scm, Git)
     if all_commits:
         exp_ref_set.update(exp_refs(repo.scm))
-
-    else:
-        if exp_names:
-            if isinstance(exp_names, str):
-                exp_names = [exp_names]
-            exp_ref_dict = resolve_name(repo.scm, exp_names)
-
-            unresolved_exp_names = []
-            for exp_name, exp_ref in exp_ref_dict.items():
-                if exp_ref is None:
-                    unresolved_exp_names.append(exp_name)
-                else:
-                    exp_ref_set.add(exp_ref)
-
-            if unresolved_exp_names:
-                raise UnresolvedExpNamesError(unresolved_exp_names)
-
-        if rev:
-            rev_dict = iter_revs(repo.scm, [rev], num)
-            rev_set = set(rev_dict.keys())
-            ref_info_dict = exp_refs_by_baseline(repo.scm, rev_set)
-            for _, ref_info_list in ref_info_dict.items():
-                exp_ref_set.update(ref_info_list)
+    if exp_names:
+        exp_ref_set.update(exp_refs_from_names(repo.scm, ensure_list(exp_names)))
+    if rev:
+        if isinstance(rev, str):
+            rev = [rev]
+        exp_ref_set.update(exp_refs_from_rev(repo.scm, rev, num=num))
 
     push_result = _push(repo, git_remote, exp_ref_set, force)
 
@@ -120,15 +127,20 @@ def push(  # noqa: C901
     pushed_refs_info = (
         push_result[SyncStatus.UP_TO_DATE] + push_result[SyncStatus.SUCCESS]
     )
+
+    e = None
     if push_cache:
         try:
             result["uploaded"] = _push_cache(repo, pushed_refs_info, **kwargs)
         except Exception as exc:  # noqa: BLE001, pylint: disable=broad-except
-            raise UploadError("failed to push cache", result) from exc
+            e = exc
 
     pushed_refs = [str(r) for r in pushed_refs_info]
-    url = notify_refs_to_studio(repo, git_remote, pushed=pushed_refs)
-    return {**result, "url": url}
+    result["url"] = notify_refs_to_studio(repo, git_remote, pushed=pushed_refs)
+
+    if e:
+        raise UploadError("failed to push cache", result) from e
+    return result
 
 
 def _push(

dvc/repo/experiments/queue/base.py

@@ -17,8 +17,7 @@ from typing import (
     Union,
 )
 
-from dvc_studio_client.env import STUDIO_REPO_URL
-from dvc_studio_client.post_live_metrics import get_studio_repo_url
+from dvc_studio_client.post_live_metrics import get_studio_config
 from funcy import retry
 
 from dvc.dependency import ParamsDependency
@@ -39,6 +38,7 @@ from dvc.repo.experiments.utils import (
 )
 from dvc.ui import ui
 from dvc.utils.objects import cached_property
+from dvc.utils.studio import config_to_env
 
 from .utils import get_remote_executor_refs
 
@@ -46,6 +46,7 @@ if TYPE_CHECKING:
     from dvc.repo import Repo
     from dvc.repo.experiments import Experiments
     from dvc.repo.experiments.executor.base import ExecutorResult
+    from dvc.repo.experiments.serialize import ExpRange
     from dvc.scm import Git
 
 logger = logging.getLogger(__name__)
@@ -253,7 +254,7 @@ class BaseStashQueue(ABC):
 
     @abstractmethod
     def reproduce(
-        self, copy_paths: Optional[List[str]] = None
+        self, copy_paths: Optional[List[str]] = None, message: Optional[str] = None
     ) -> Mapping[str, Mapping[str, str]]:
         """Reproduce queued experiments sequentially."""
 
@@ -419,12 +420,13 @@ class BaseStashQueue(ABC):
                     run_env[DVC_EXP_NAME] = name
                     if resume_rev:
                         run_env[DVCLIVE_RESUME] = "1"
-                    studio_repo_url = get_studio_repo_url()
-                    if studio_repo_url is not None:
-                        run_env[STUDIO_REPO_URL] = studio_repo_url
 
+                    # save studio config to read later by dvc and dvclive
+                    studio_config = get_studio_config(
+                        dvc_studio_config=self.repo.config.get("studio")
+                    )
+                    run_env = {**config_to_env(studio_config), **run_env}
                     self._pack_args(*args, run_env=run_env, **kwargs)
-
                     # save experiment as a stash commit
                     msg = self._stash_msg(
                         stash_head,
@@ -520,7 +522,7 @@ class BaseStashQueue(ABC):
         else:
             extra = None
         BaseExecutor.pack_repro_args(self.args_file, *args, extra=extra, **kwargs)
-        self.scm.add(self.args_file)
+        self.scm.add(self.args_file, force=True)
 
     @staticmethod
     def _format_new_params_msg(new_params, config_path):
@@ -728,9 +730,78 @@ class BaseStashQueue(ABC):
             )
 
     @abstractmethod
-    def get_running_exps(self, fetch_refs: bool = True) -> Dict[str, Dict]:
-        """Get the execution info of the currently running experiments
+    def collect_active_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        fetch_refs: bool = False,
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        """Collect data for active (running) experiments.
 
         Args:
-            fetch_ref (bool): fetch completed checkpoints or not.
+            baseline_revs: Optional resolved baseline Git SHAs. If set, only experiments
+                derived from the specified revisions will be collected. Defaults to
+                collecting all experiments.
+            fetch_refs: Whether or not to fetch completed checkpoint commits from Git
+                remote.
+
+        Returns:
+            Dict mapping baseline revision to list of active experiments.
         """
+
+    @abstractmethod
+    def collect_queued_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        """Collect data for queued experiments.
+
+        Args:
+            baseline_revs: Optional resolved baseline Git SHAs. If set, only experiments
+                derived from the specified revisions will be collected. Defaults to
+                collecting all experiments.
+
+        Returns:
+            Dict mapping baseline revision to list of queued experiments.
+        """
+
+    @abstractmethod
+    def collect_failed_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        """Collect data for failed experiments.
+
+        Args:
+            baseline_revs: Optional resolved baseline Git SHAs. If set, only experiments
+                derived from the specified revisions will be collected. Defaults to
+                collecting all experiments.
+
+        Returns:
+            Dict mapping baseline revision to list of queued experiments.
+        """
+
+    def active_repo(self, name: str) -> "Repo":
+        """Return a Repo for the specified active experiment if it exists."""
+        from dvc.repo import Repo
+        from dvc.repo.experiments.exceptions import (
+            ExpNotStartedError,
+            InvalidExpRevError,
+        )
+        from dvc.repo.experiments.executor.base import ExecutorInfo, TaskStatus
+
+        for entry in self.iter_active():
+            if entry.name != name:
+                continue
+            infofile = self.get_infofile_path(entry.stash_rev)
+            executor_info = ExecutorInfo.load_json(infofile)
+            if executor_info.status < TaskStatus.RUNNING:
+                raise ExpNotStartedError(name)
+            dvc_root = os.path.join(executor_info.root_dir, executor_info.dvc_dir)
+            try:
+                return Repo(dvc_root)
+            except (FileNotFoundError, DvcException) as exc:
+                raise InvalidExpRevError(name) from exc
+        raise InvalidExpRevError(name)

dvc/repo/experiments/queue/celery.py

@@ -2,6 +2,7 @@ import hashlib
 import locale
 import logging
 import os
+from collections import defaultdict
 from typing import (
     TYPE_CHECKING,
     Collection,
@@ -40,6 +41,7 @@ if TYPE_CHECKING:
 
     from dvc.repo.experiments.executor.base import ExecutorResult
     from dvc.repo.experiments.refs import ExpRefInfo
+    from dvc.repo.experiments.serialize import ExpRange
     from dvc_task.app import FSApp
     from dvc_task.proc.manager import ProcessManager
     from dvc_task.worker import TemporaryWorker
@@ -123,6 +125,7 @@ class LocalCeleryQueue(BaseStashQueue):
             without_mingle=True,
             without_gossip=True,
             timeout=10,
+            loglevel="debug" if logger.getEffectiveLevel() <= logging.DEBUG else "info",
         )
 
     def _spawn_worker(self, num: int = 1):
@@ -141,6 +144,8 @@ class LocalCeleryQueue(BaseStashQueue):
         if num == 1:
             # automatically run celery cleanup when primary worker shuts down
             cmd.append("--clean")
+        if logger.getEffectiveLevel() <= logging.DEBUG:
+            cmd.append("-v")
         name = f"dvc-exp-worker-{num}"
 
         logger.debug("start a new worker: %s, node: %s", name, node_name)
@@ -175,12 +180,18 @@ class LocalCeleryQueue(BaseStashQueue):
         return started
 
     def put(
-        self, *args, copy_paths: Optional[List[str]] = None, **kwargs
+        self,
+        *args,
+        copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
+        **kwargs,
     ) -> QueueEntry:
         """Stash an experiment and add it to the queue."""
         with get_exp_rwlock(self.repo, writes=["workspace", CELERY_STASH]):
             entry = self._stash_exp(*args, **kwargs)
-        self.celery.signature(run_exp.s(entry.asdict(), copy_paths=copy_paths)).delay()
+        self.celery.signature(
+            run_exp.s(entry.asdict(), copy_paths=copy_paths, message=message)
+        ).delay()
         return entry
 
     # NOTE: Queue consumption should not be done directly. Celery worker(s)
@@ -198,7 +209,9 @@ class LocalCeleryQueue(BaseStashQueue):
                 continue
             args, kwargs, _embed = msg.decode()
             entry_dict = kwargs.get("entry_dict", args[0])
-            logger.debug("Found queued task %s", entry_dict["stash_rev"])
+            logger.trace(  # type: ignore[attr-defined]
+                "Found queued task %s", entry_dict["stash_rev"]
+            )
             yield _MessageEntry(msg, QueueEntry.from_dict(entry_dict))
 
     def _iter_processed(self) -> Generator[_MessageEntry, None, None]:
@@ -214,7 +227,9 @@ class LocalCeleryQueue(BaseStashQueue):
             task_id = msg.headers["id"]
             result: AsyncResult = AsyncResult(task_id)
             if not result.ready():
-                logger.debug("Found active task %s", entry.stash_rev)
+                logger.trace(  # type: ignore[attr-defined]
+                    "Found active task %s", entry.stash_rev
+                )
                 yield _TaskEntry(result, entry)
 
     def _iter_done_tasks(self) -> Generator[_TaskEntry, None, None]:
@@ -222,7 +237,9 @@ class LocalCeleryQueue(BaseStashQueue):
             task_id = msg.headers["id"]
             result: AsyncResult = AsyncResult(task_id)
             if result.ready():
-                logger.debug("Found done task %s", entry.stash_rev)
+                logger.trace(  # type: ignore[attr-defined]
+                    "Found done task %s", entry.stash_rev
+                )
                 yield _TaskEntry(result, entry)
 
     def iter_active(self) -> Generator[QueueEntry, None, None]:
@@ -253,7 +270,7 @@ class LocalCeleryQueue(BaseStashQueue):
                 yield QueueDoneResult(queue_entry, exp_result)
 
     def reproduce(
-        self, copy_paths: Optional[List[str]] = None
+        self, copy_paths: Optional[List[str]] = None, message: Optional[str] = None
     ) -> Mapping[str, Mapping[str, str]]:
         raise NotImplementedError
 
@@ -472,19 +489,6 @@ class LocalCeleryQueue(BaseStashQueue):
 
         return celery_remove(self, *args, **kwargs)
 
-    def get_running_exps(self, fetch_refs: bool = True) -> Dict[str, Dict]:
-        """Get the execution info of the currently running experiments
-
-        Args:
-            fetch_ref (bool): fetch completed checkpoints or not.
-        """
-        result: Dict[str, Dict] = {}
-        for entry in self.iter_active():
-            result.update(
-                fetch_running_exp_from_temp_dir(self, entry.stash_rev, fetch_refs)
-            )
-        return result
-
     def get_ref_and_entry_by_names(
         self,
         exp_names: Union[str, List[str]],
@@ -512,3 +516,124 @@ class LocalCeleryQueue(BaseStashQueue):
             queue_entry = None if git_remote else queue_entry_match[exp_name]
             results[exp_name] = ExpRefAndQueueEntry(exp_ref, queue_entry)
         return results
+
+    def collect_active_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        fetch_refs: bool = False,
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        from dvc.repo import Repo
+        from dvc.repo.experiments.collect import collect_exec_branch
+        from dvc.repo.experiments.serialize import (
+            ExpExecutor,
+            ExpRange,
+            LocalExpExecutor,
+        )
+
+        result: Dict[str, List[ExpRange]] = defaultdict(list)
+        for entry in self.iter_active():
+            if baseline_revs and entry.baseline_rev not in baseline_revs:
+                continue
+            if fetch_refs:
+                fetch_running_exp_from_temp_dir(self, entry.stash_rev, fetch_refs)
+            proc_info = self.proc.get(entry.stash_rev)
+            executor_info = self._load_info(entry.stash_rev)
+            if proc_info:
+                local_exec: Optional[LocalExpExecutor] = LocalExpExecutor(
+                    root=executor_info.root_dir,
+                    log=proc_info.stdout,
+                    pid=proc_info.pid,
+                    task_id=entry.stash_rev,
+                )
+            else:
+                local_exec = None
+            dvc_root = os.path.join(executor_info.root_dir, executor_info.dvc_dir)
+            with Repo(dvc_root) as exec_repo:
+                kwargs["cache"] = self.repo.experiments.cache
+                exps = list(
+                    collect_exec_branch(exec_repo, executor_info.baseline_rev, **kwargs)
+                )
+            exps[0].rev = entry.stash_rev
+            exps[0].name = entry.name
+            result[entry.baseline_rev].append(
+                ExpRange(
+                    exps,
+                    executor=ExpExecutor(
+                        "running",
+                        name=executor_info.location,
+                        local=local_exec,
+                    ),
+                    name=entry.name,
+                )
+            )
+        return result
+
+    def collect_queued_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        from dvc.repo.experiments.collect import collect_rev
+        from dvc.repo.experiments.serialize import (
+            ExpExecutor,
+            ExpRange,
+            LocalExpExecutor,
+        )
+
+        result: Dict[str, List[ExpRange]] = defaultdict(list)
+        for entry in self.iter_queued():
+            if baseline_revs and entry.baseline_rev not in baseline_revs:
+                continue
+            exp = collect_rev(self.repo, entry.stash_rev, **kwargs)
+            exp.name = entry.name
+            local_exec: Optional[LocalExpExecutor] = LocalExpExecutor(
+                task_id=entry.stash_rev,
+            )
+            result[entry.baseline_rev].append(
+                ExpRange(
+                    [exp],
+                    executor=ExpExecutor("queued", name="dvc-task", local=local_exec),
+                    name=entry.name,
+                )
+            )
+        return result
+
+    def collect_failed_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        from dvc.repo.experiments.collect import collect_rev
+        from dvc.repo.experiments.serialize import (
+            ExpExecutor,
+            ExpRange,
+            LocalExpExecutor,
+            SerializableError,
+        )
+
+        result: Dict[str, List[ExpRange]] = defaultdict(list)
+        for entry, _ in self.iter_failed():
+            if baseline_revs and entry.baseline_rev not in baseline_revs:
+                continue
+            proc_info = self.proc.get(entry.stash_rev)
+            if proc_info:
+                local_exec: Optional[LocalExpExecutor] = LocalExpExecutor(
+                    log=proc_info.stdout,
+                    pid=proc_info.pid,
+                    returncode=proc_info.returncode,
+                    task_id=entry.stash_rev,
+                )
+            else:
+                local_exec = None
+            exp = collect_rev(self.repo, entry.stash_rev, **kwargs)
+            exp.name = entry.name
+            exp.error = SerializableError("Experiment run failed")
+            result[entry.baseline_rev].append(
+                ExpRange(
+                    [exp],
+                    executor=ExpExecutor("failed", local=local_exec),
+                    name=entry.name,
+                )
+            )
+        return result

dvc/repo/experiments/queue/tasks.py

@@ -91,7 +91,11 @@ def cleanup_exp(executor: TempDirExecutor, infofile: str) -> None:
 
 
 @shared_task
-def run_exp(entry_dict: Dict[str, Any], copy_paths: Optional[List[str]] = None) -> None:
+def run_exp(
+    entry_dict: Dict[str, Any],
+    copy_paths: Optional[List[str]] = None,
+    message: Optional[str] = None,
+) -> None:
     """Run a full experiment.
 
     Experiment subtasks are executed inline as one atomic operation.
@@ -111,6 +115,8 @@ def run_exp(entry_dict: Dict[str, Any], copy_paths: Optional[List[str]] = None)
         if copy_paths:
             for path in copy_paths:
                 cmd.extend(["--copy-paths", path])
+        if message:
+            cmd.extend(["--message", message])
         proc_dict = queue.proc.run_signature(cmd, name=entry.stash_rev)()
         collect_exp.s(proc_dict, entry_dict)()
     finally:

dvc/repo/experiments/queue/tempdir.py

@@ -1,7 +1,7 @@
 import logging
 import os
 from collections import defaultdict
-from typing import TYPE_CHECKING, Dict, Generator, List, Optional
+from typing import TYPE_CHECKING, Collection, Dict, Generator, List, Optional
 
 from funcy import first
 
@@ -19,6 +19,7 @@ from .workspace import WorkspaceQueue
 if TYPE_CHECKING:
     from dvc.repo.experiments import Experiments
     from dvc.repo.experiments.executor.base import BaseExecutor, ExecutorResult
+    from dvc.repo.experiments.serialize import ExpRange
     from dvc_task.proc.manager import ProcessManager
 
 logger = logging.getLogger(__name__)
@@ -96,6 +97,7 @@ class TempDirQueue(WorkspaceQueue):
         entry: QueueEntry,
         executor: "BaseExecutor",
         copy_paths: Optional[List[str]] = None,
+        message: Optional[str] = None,
         **kwargs,
     ) -> Dict[str, Dict[str, str]]:
         from dvc.stage.monitor import CheckpointKilledError
@@ -112,6 +114,7 @@ class TempDirQueue(WorkspaceQueue):
                 log_level=logger.getEffectiveLevel(),
                 log_errors=True,
                 copy_paths=copy_paths,
+                message=message,
             )
             if not exec_result.exp_hash:
                 raise DvcException(f"Failed to reproduce experiment '{rev[:7]}'")
@@ -141,10 +144,53 @@ class TempDirQueue(WorkspaceQueue):
     ) -> Dict[str, str]:
         return BaseStashQueue.collect_executor(exp, executor, exec_result)
 
-    def get_running_exps(self, fetch_refs: bool = True) -> Dict[str, Dict]:
-        result: Dict[str, Dict] = {}
+    def collect_active_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        fetch_refs: bool = False,
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        from dvc.repo import Repo
+        from dvc.repo.experiments.collect import collect_exec_branch
+        from dvc.repo.experiments.serialize import (
+            ExpExecutor,
+            ExpRange,
+            LocalExpExecutor,
+        )
+
+        result: Dict[str, List[ExpRange]] = defaultdict(list)
         for entry in self.iter_active():
-            result.update(
+            if baseline_revs and entry.baseline_rev not in baseline_revs:
+                continue
+            if fetch_refs:
                 fetch_running_exp_from_temp_dir(self, entry.stash_rev, fetch_refs)
+            proc_info = self.proc.get(entry.stash_rev)
+            infofile = self.get_infofile_path(entry.stash_rev)
+            executor_info = ExecutorInfo.load_json(infofile)
+            if proc_info:
+                local_exec: Optional[LocalExpExecutor] = LocalExpExecutor(
+                    root=executor_info.root_dir,
+                    log=proc_info.stdout,
+                    pid=proc_info.pid,
+                )
+            else:
+                local_exec = None
+            dvc_root = os.path.join(executor_info.root_dir, executor_info.dvc_dir)
+            with Repo(dvc_root) as repo:
+                exps = list(
+                    collect_exec_branch(repo, executor_info.baseline_rev, **kwargs)
+                )
+            exps[0].rev = entry.stash_rev
+            exps[0].name = entry.name
+            result[entry.baseline_rev].append(
+                ExpRange(
+                    exps,
+                    executor=ExpExecutor(
+                        "running",
+                        name=executor_info.location,
+                        local=local_exec,
+                    ),
+                    name=entry.name,
+                )
             )
         return result

dvc/repo/experiments/queue/workspace.py

@@ -1,28 +1,27 @@
 import json
 import logging
+import os
 from collections import defaultdict
 from typing import TYPE_CHECKING, Collection, Dict, Generator, List, Optional
 
 import psutil
 from funcy import first
-from voluptuous import Invalid
 
 from dvc.exceptions import DvcException
-from dvc.lock import make_lock
 from dvc.repo.experiments.exceptions import ExpQueueEmptyError
 from dvc.repo.experiments.executor.base import ExecutorInfo, TaskStatus
 from dvc.repo.experiments.executor.local import WorkspaceExecutor
 from dvc.repo.experiments.refs import EXEC_BRANCH, WORKSPACE_STASH
 from dvc.repo.experiments.utils import get_exp_rwlock
-from dvc.rwlock import RWLOCK_FILE, RWLOCK_LOCK, SCHEMA
-from dvc.utils import relpath
 from dvc.utils.fs import remove
+from dvc.utils.serialize import load_json
 
 from .base import BaseStashQueue, QueueEntry, QueueGetResult
 
 if TYPE_CHECKING:
     from dvc.repo.experiments import Experiments
     from dvc.repo.experiments.executor.base import BaseExecutor, ExecutorResult
+    from dvc.repo.experiments.serialize import ExpRange
 
     from .base import QueueDoneResult
 
@@ -83,14 +82,16 @@ class WorkspaceQueue(BaseStashQueue):
         raise NotImplementedError
 
     def reproduce(
-        self, copy_paths: Optional[List[str]] = None
+        self, copy_paths: Optional[List[str]] = None, message: Optional[str] = None
     ) -> Dict[str, Dict[str, str]]:
         results: Dict[str, Dict[str, str]] = defaultdict(dict)
         try:
             while True:
                 entry, executor = self.get()
                 results.update(
-                    self._reproduce_entry(entry, executor, copy_paths=copy_paths)
+                    self._reproduce_entry(
+                        entry, executor, copy_paths=copy_paths, message=message
+                    )
                 )
         except ExpQueueEmptyError:
             pass
@@ -101,9 +102,14 @@ class WorkspaceQueue(BaseStashQueue):
     ) -> Dict[str, Dict[str, str]]:
         kwargs.pop("copy_paths", None)
         from dvc.stage.monitor import CheckpointKilledError
+        from dvc_task.proc.process import ProcessInfo
 
         results: Dict[str, Dict[str, str]] = defaultdict(dict)
         exec_name = self._EXEC_NAME or entry.stash_rev
+        proc_info = ProcessInfo(os.getpid(), None, None, None, None)
+        proc_info_path = self._proc_info_path(exec_name)
+        os.makedirs(os.path.dirname(proc_info_path), exist_ok=True)
+        proc_info.dump(proc_info_path)
         infofile = self.get_infofile_path(exec_name)
         try:
             rev = entry.stash_rev
@@ -113,6 +119,7 @@ class WorkspaceQueue(BaseStashQueue):
                 infofile=infofile,
                 log_level=logger.getEffectiveLevel(),
                 log_errors=not isinstance(executor, WorkspaceExecutor),
+                message=kwargs.get("message"),
             )
             if not exec_result.exp_hash:
                 raise DvcException(f"Failed to reproduce experiment '{rev[:7]}'")
@@ -129,8 +136,29 @@ class WorkspaceQueue(BaseStashQueue):
             raise DvcException(f"Failed to reproduce experiment '{rev[:7]}'") from exc
         finally:
             executor.cleanup(infofile)
+            remove(self._proc_info_path(exec_name))
         return results
 
+    def _proc_info_path(self, name: str) -> str:
+        return os.path.join(self.pid_dir, name, f"{name}.json")
+
+    @property
+    def _active_pid(self) -> Optional[int]:
+        from dvc_task.proc.process import ProcessInfo
+
+        assert self._EXEC_NAME
+        name = self._EXEC_NAME
+        try:
+            proc_info = ProcessInfo.load(self._proc_info_path(name))
+            pid = proc_info.pid
+            if psutil.pid_exists(pid):
+                return pid
+            logger.debug("Workspace exec PID '%d' no longer exists, removing.", pid)
+            remove(self._proc_info_path(name))
+        except (FileNotFoundError, json.JSONDecodeError):
+            pass
+        return None
+
     @staticmethod
     def collect_executor(  # pylint: disable=unused-argument
         exp: "Experiments",
@@ -163,87 +191,71 @@ class WorkspaceQueue(BaseStashQueue):
     ):
         raise NotImplementedError
 
-    def check_rwlock(
-        self,
-        hardlink: bool = False,
-        autocorrect: bool = False,
-    ) -> bool:
-        """Check and autocorrect the RWLock status for file paths.
-
-        Args:
-            hardlink (bool): use hardlink lock to guard rwlock file when on
-                            edit.
-            autocorrect (bool): autocorrect corrupted rwlock file.
-
-        Return:
-            (bool): if the pid alive.
-        """
-        path = self.repo.fs.path.join(self.repo.tmp_dir, RWLOCK_FILE)
-
-        rwlock_guard = make_lock(
-            self.repo.fs.path.join(self.repo.tmp_dir, RWLOCK_LOCK),
-            tmp_dir=self.repo.tmp_dir,
-            hardlink_lock=hardlink,
-        )
-        with rwlock_guard:
-            try:
-                with self.repo.fs.open(path, encoding="utf-8") as fobj:
-                    lock: Dict[str, List[Dict]] = SCHEMA(json.load(fobj))
-                file_path = first(lock["read"])
-                if not file_path:
-                    return False
-                lock_info = first(lock["read"][file_path])
-                pid = int(lock_info["pid"])
-                if psutil.pid_exists(pid):
-                    return True
-                cmd = lock_info["cmd"]
-                logger.warning(
-                    "Process '%s' with (Pid %s), in RWLock-file '%s' had been killed.",
-                    cmd,
-                    pid,
-                    relpath(path),
-                )
-            except FileNotFoundError:
-                return False
-            except json.JSONDecodeError:
-                logger.warning(
-                    "Unable to read RWLock-file '%s'. JSON structure is corrupted",
-                    relpath(path),
-                )
-            except Invalid:
-                logger.warning("RWLock-file '%s' format error.", relpath(path))
-            if autocorrect:
-                logger.warning("Delete corrupted RWLock-file '%s'", relpath(path))
-                remove(path)
-            return False
-
-    def get_running_exps(
-        self,
-        fetch_refs: bool = True,  # noqa: ARG002
-    ) -> Dict[str, Dict]:
-        from dvc.utils.serialize import load_json
-
+    def get_running_exp(self) -> Optional[str]:
+        """Return the name of the exp running in workspace (if it exists)."""
         assert self._EXEC_NAME
-        result: Dict[str, Dict] = {}
-
-        if not self.check_rwlock(autocorrect=True):
-            return result
+        if self._active_pid is None:
+            return None
 
         infofile = self.get_infofile_path(self._EXEC_NAME)
+        try:
+            info = ExecutorInfo.from_dict(load_json(infofile))
+        except OSError:
+            return None
+        return info.name
 
+    def collect_active_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        fetch_refs: bool = False,  # noqa: ARG002
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        from dvc.repo.experiments.collect import collect_exec_branch
+        from dvc.repo.experiments.serialize import (
+            ExpExecutor,
+            ExpRange,
+            LocalExpExecutor,
+        )
+
+        result: Dict[str, List[ExpRange]] = defaultdict(list)
+        pid = self._active_pid
+        if pid is None:
+            return result
+
+        assert self._EXEC_NAME
+        infofile = self.get_infofile_path(self._EXEC_NAME)
         try:
             info = ExecutorInfo.from_dict(load_json(infofile))
         except OSError:
             return result
 
-        if info.status < TaskStatus.FAILED:
-            # If we are appending to a checkpoint branch in a workspace
-            # run, show the latest checkpoint as running.
-            if info.status == TaskStatus.SUCCESS:
-                return result
-            last_rev = self.scm.get_ref(EXEC_BRANCH)
-            if last_rev:
-                result[last_rev] = info.asdict()
-            else:
-                result[self._EXEC_NAME] = info.asdict()
+        if (
+            (not baseline_revs or info.baseline_rev in baseline_revs)
+            and info.status < TaskStatus.FAILED
+            and info.status != TaskStatus.SUCCESS
+        ):
+            local_exec = LocalExpExecutor(root=info.root_dir, pid=pid)
+            exps = list(collect_exec_branch(self.repo, info.baseline_rev, **kwargs))
+            exps[0].name = info.name
+            result[info.baseline_rev] = [
+                ExpRange(
+                    exps,
+                    executor=ExpExecutor("running", name="workspace", local=local_exec),
+                    name=info.name,
+                )
+            ]
         return result
+
+    def collect_queued_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        raise NotImplementedError
+
+    def collect_failed_data(
+        self,
+        baseline_revs: Optional[Collection[str]],
+        **kwargs,
+    ) -> Dict[str, List["ExpRange"]]:
+        raise NotImplementedError

dvc/repo/experiments/refs.py

@@ -10,8 +10,9 @@ 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"
+CELERY_NAMESPACE = f"{EXPS_NAMESPACE}/celery"
+CELERY_STASH = f"{CELERY_NAMESPACE}/stash"
+CELERY_FAILED_STASH = f"{CELERY_NAMESPACE}/failed"
 EXEC_NAMESPACE = f"{EXPS_NAMESPACE}/exec"
 EXEC_APPLY = f"{EXEC_NAMESPACE}/EXEC_APPLY"
 EXEC_CHECKPOINT = f"{EXEC_NAMESPACE}/EXEC_CHECKPOINT"
@@ -19,9 +20,14 @@ EXEC_BRANCH = f"{EXEC_NAMESPACE}/EXEC_BRANCH"
 EXEC_BASELINE = f"{EXEC_NAMESPACE}/EXEC_BASELINE"
 EXEC_HEAD = f"{EXEC_NAMESPACE}/EXEC_HEAD"
 EXEC_MERGE = f"{EXEC_NAMESPACE}/EXEC_MERGE"
-EXPS_TEMP = f"{EXPS_NAMESPACE}/temp"
+TEMP_NAMESPACE = f"{EXPS_NAMESPACE}/temp"
 STASHES = {WORKSPACE_STASH, CELERY_STASH}
-ITER_SKIP_NAMESPACES = {APPLY_NAMESPACE, EXEC_NAMESPACE}
+ITER_SKIP_NAMESPACES = {
+    APPLY_NAMESPACE,
+    CELERY_NAMESPACE,
+    EXEC_NAMESPACE,
+    TEMP_NAMESPACE,
+}
 
 
 class ExpRefInfo:

dvc/repo/experiments/remove.py

@@ -24,7 +24,7 @@ logger = logging.getLogger(__name__)
 def remove(  # noqa: C901, PLR0912
     repo: "Repo",
     exp_names: Union[None, str, List[str]] = None,
-    rev: Optional[str] = None,
+    rev: Optional[Union[List[str], str]] = None,
     all_commits: bool = False,
     num: int = 1,
     queue: bool = False,
@@ -60,6 +60,8 @@ def remove(  # noqa: C901, PLR0912
         if remained:
             raise UnresolvedExpNamesError(remained, git_remote=git_remote)
     elif rev:
+        if isinstance(rev, str):
+            rev = [rev]
         exp_ref_dict = _resolve_exp_by_baseline(repo, rev, num, git_remote)
         removed.extend(exp_ref_dict.keys())
         exp_ref_list.extend(exp_ref_dict.values())
@@ -85,14 +87,14 @@ def remove(  # noqa: C901, PLR0912
 
 def _resolve_exp_by_baseline(
     repo: "Repo",
-    rev: str,
+    rev: List[str],
     num: int,
     git_remote: Optional[str] = None,
 ) -> Dict[str, "ExpRefInfo"]:
     assert isinstance(repo.scm, Git)
 
     commit_ref_dict: Dict[str, "ExpRefInfo"] = {}
-    rev_dict = iter_revs(repo.scm, [rev], num)
+    rev_dict = iter_revs(repo.scm, rev, num)
     rev_set = set(rev_dict.keys())
     ref_info_dict = exp_refs_by_baseline(repo.scm, rev_set, git_remote)
     for _, ref_info_list in ref_info_dict.items():

dvc/repo/experiments/run.py

@@ -20,6 +20,7 @@ def run(  # noqa: C901, PLR0912
     tmp_dir: bool = False,
     queue: bool = False,
     copy_paths: Optional[Iterable[str]] = None,
+    message: Optional[str] = None,
     **kwargs,
 ) -> Dict[str, str]:
     """Reproduce the specified targets as an experiment.
@@ -74,6 +75,7 @@ def run(  # noqa: C901, PLR0912
             params=path_overrides,
             tmp_dir=tmp_dir,
             copy_paths=copy_paths,
+            message=message,
             **kwargs,
         )
 
@@ -96,6 +98,7 @@ def run(  # noqa: C901, PLR0912
             targets=targets,
             params=sweep_overrides,
             copy_paths=copy_paths,
+            message=message,
             **kwargs,
         )
         if sweep_overrides:

dvc/repo/experiments/save.py

@@ -16,6 +16,7 @@ def save(
     name: Optional[str] = None,
     force: bool = False,
     include_untracked: Optional[List[str]] = None,
+    message: Optional[str] = None,
 ) -> Optional[str]:
     """Save the current workspace status as an experiment.
 
@@ -29,7 +30,10 @@ def save(
 
     try:
         save_result = executor.save(
-            executor.info, force=force, include_untracked=include_untracked
+            executor.info,
+            force=force,
+            include_untracked=include_untracked,
+            message=message,
         )
         result = queue.collect_executor(repo.experiments, executor, save_result)
     finally:

dvc/repo/experiments/serialize.py

@@ -1,7 +1,7 @@
 import json
 from dataclasses import asdict, dataclass, field
 from datetime import datetime
-from typing import TYPE_CHECKING, Any, Callable, Dict, Optional
+from typing import TYPE_CHECKING, Any, Callable, Dict, Iterator, List, Literal, Optional
 
 from dvc.exceptions import DvcException
 from dvc.repo.metrics.show import _gather_metrics
@@ -25,17 +25,15 @@ class _ISOEncoder(json.JSONEncoder):
 
 @dataclass(frozen=True)
 class SerializableExp:
-    """Serializable experiment state."""
+    """Serializable experiment data."""
 
     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
+    deps: Dict[str, "ExpDep"] = field(default_factory=dict)
+    outs: Dict[str, "ExpOut"] = field(default_factory=dict)
+    meta: Dict[str, Any] = field(default_factory=dict)
 
     @classmethod
     def from_repo(
@@ -43,6 +41,7 @@ class SerializableExp:
         repo: "Repo",
         rev: Optional[str] = None,
         onerror: Optional[Callable] = None,
+        param_deps: bool = False,
         **kwargs,
     ) -> "SerializableExp":
         """Returns a SerializableExp from the current repo state.
@@ -57,29 +56,26 @@ class SerializableExp:
 
         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,
-                )
+        params = dict(_gather_params(repo, deps=param_deps, onerror=onerror))
+        metrics = dict(
+            _gather_metrics(
+                repo,
+                targets=None,
+                rev=rev[:7],
+                recursive=False,
+                onerror=onerror_collect,
             )
+        )
+        meta = cls._gather_meta(repo)
         return cls(
             rev=rev,
             params=params,
             metrics=metrics,
             deps={
-                relpath(dep.fs_path, repo.root_dir): _ExpDep(
+                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,
@@ -88,7 +84,7 @@ class SerializableExp:
                 if not isinstance(dep, (ParamsDependency, RepoDependency))
             },
             outs={
-                relpath(out.fs_path, repo.root_dir): _ExpOut(
+                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,
@@ -98,9 +94,16 @@ class SerializableExp:
                 for out in repo.index.outs
                 if not (out.is_metric or out.is_plot)
             },
+            meta=meta,
             **kwargs,
         )
 
+    @staticmethod
+    def _gather_meta(repo: "Repo") -> Dict[str, Any]:
+        return {
+            "has_checkpoints": any(stage.is_checkpoint for stage in repo.index.stages)
+        }
+
     def dumpd(self) -> Dict[str, Any]:
         return asdict(self)
 
@@ -113,6 +116,10 @@ class SerializableExp:
             parsed = json.loads(data)
             if "timestamp" in parsed:
                 parsed["timestamp"] = datetime.fromisoformat(parsed["timestamp"])
+            if "deps" in parsed:
+                parsed["deps"] = {k: ExpDep(**v) for k, v in parsed["deps"].items()}
+            if "outs" in parsed:
+                parsed["outs"] = {k: ExpOut(**v) for k, v in parsed["outs"].items()}
             return cls(**parsed)
         except (TypeError, json.JSONDecodeError) as exc:
             raise DeserializeError("failed to load SerializableExp") from exc
@@ -120,8 +127,7 @@ class SerializableExp:
     @property
     def contains_error(self) -> bool:
         return (
-            self.error is not None
-            or self.params.get("error")
+            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())
@@ -129,14 +135,14 @@ class SerializableExp:
 
 
 @dataclass(frozen=True)
-class _ExpDep:
+class ExpDep:
     hash: Optional[str]  # noqa: A003
     size: Optional[int]
     nfiles: Optional[int]
 
 
 @dataclass(frozen=True)
-class _ExpOut:
+class ExpOut:
     hash: Optional[str]  # noqa: A003
     size: Optional[int]
     nfiles: Optional[int]
@@ -162,3 +168,52 @@ class SerializableError:
             return cls(**parsed)
         except (TypeError, json.JSONDecodeError) as exc:
             raise DeserializeError("failed to load SerializableError") from exc
+
+
+@dataclass
+class ExpState:
+    """Git/DVC experiment state."""
+
+    rev: str
+    name: Optional[str] = None
+    data: Optional[SerializableExp] = None
+    error: Optional[SerializableError] = None
+    experiments: Optional[List["ExpRange"]] = None
+
+    def dumpd(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+@dataclass
+class ExpRange:
+    revs: List["ExpState"]
+    executor: Optional["ExpExecutor"] = None
+    name: Optional[str] = None
+
+    def __len__(self) -> int:
+        return len(self.revs)
+
+    def __iter__(self) -> Iterator["ExpState"]:
+        return iter(self.revs)
+
+    def __getitem__(self, index: int) -> "ExpState":
+        return self.revs[index]
+
+    def dumpd(self) -> Dict[str, Any]:
+        return asdict(self)
+
+
+@dataclass
+class LocalExpExecutor:
+    root: Optional[str] = None
+    log: Optional[str] = None
+    pid: Optional[int] = None
+    returncode: Optional[int] = None
+    task_id: Optional[str] = None
+
+
+@dataclass
+class ExpExecutor:
+    state: Literal["success", "queued", "running", "failed"]
+    name: Optional[str] = None
+    local: Optional[LocalExpExecutor] = None

dvc/repo/experiments/show.py

@@ -1,517 +1,390 @@
 import logging
-from collections import OrderedDict, defaultdict
-from datetime import datetime
-from enum import Enum
-from itertools import chain
+from collections import Counter, defaultdict
+from datetime import date, datetime
 from typing import (
     TYPE_CHECKING,
     Any,
-    Callable,
     Dict,
     Iterable,
+    Iterator,
     List,
+    Literal,
+    Mapping,
+    NamedTuple,
     Optional,
     Set,
     Tuple,
     Union,
 )
 
-from scmrepo.exceptions import SCMError as InnerScmError
+from dvc.exceptions import InvalidArgumentError
+from dvc.scm import Git
+from dvc.ui import ui
+from dvc.utils.flatten import flatten
 
-from dvc.exceptions import DvcException
-from dvc.scm import Git, SCMError, iter_revs, resolve_rev
-
-from .refs import ExpRefInfo
-from .serialize import SerializableError, SerializableExp
+from .collect import collect
 
 if TYPE_CHECKING:
-    from scmrepo.git.objects import GitCommit
-
+    from dvc.compare import TabularData
     from dvc.repo import Repo
+    from dvc.ui.table import CellT
+
+    from .serialize import ExpRange, ExpState
 
 logger = logging.getLogger(__name__)
 
 
-class ExpStatus(Enum):
-    Success = 0
-    Queued = 1
-    Running = 2
-    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"], (_CachedError, SCMError, InnerScmError))
-    ):
-        return True
-    return False
-
-
-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)}
-
-
-def collect_experiment_commit(
+def show(
     repo: "Repo",
-    exp_rev: str,
-    status: ExpStatus = ExpStatus.Success,
-    param_deps: bool = False,
-    force: bool = False,
+    revs: Union[List[str], str, None] = None,
+    all_branches: bool = False,
+    all_tags: bool = False,
+    all_commits: bool = False,
+    num: int = 1,
+    hide_queued: bool = False,
+    hide_failed: bool = False,
+    sha_only: bool = False,
     **kwargs,
-) -> Dict[str, Any]:
-    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)
-
-
-def _collect_from_repo(
-    repo: "Repo",
-    exp_rev: str,
-    status: ExpStatus = ExpStatus.Success,
-    running: Optional[Dict[str, Any]] = None,
-    **kwargs,
-) -> "SerializableExp":
-    running = running or {}
-    with repo.switch(exp_rev) as rev:
-        if rev == "workspace":
-            timestamp: Optional[datetime] = None
-        else:
-            commit = repo.scm.resolve_commit(rev)
-            timestamp = datetime.fromtimestamp(commit.commit_time)
-
-        if status == ExpStatus.Running:
-            executor: Optional[str] = running.get(exp_rev, {}).get("location", None)
-        else:
-            executor = None
-
-        if status == ExpStatus.Failed:
-            error: Optional["SerializableError"] = SerializableError(
-                "Experiment run failed."
-            )
-        else:
-            error = None
-
-        return SerializableExp.from_repo(
-            repo,
-            rev=exp_rev,
-            timestamp=timestamp,
-            status=status.name,
-            executor=executor,
-            error=error,
-        )
-
-
-def _collect_complete_experiment(
-    repo: "Repo",
-    baseline: str,
-    exp_rev: str,
-    running: Dict[str, Any],
-    revs: List[str],
-    **kwargs,
-) -> Dict[str, Dict[str, Any]]:
-    results: "OrderedDict[str, Dict[str, Any]]" = OrderedDict()
-
-    checkpoint = len(revs) > 1
-    prev = ""
-
-    for rev in revs:
-        status = ExpStatus.Running if rev in running else ExpStatus.Success
-        collected_exp = collect_experiment_commit(
-            repo,
-            rev,
-            status=status,
-            running=running,
-            **kwargs,
-        )
-        if _is_scm_error(collected_exp):
-            return {}
-        if checkpoint:
-            exp = {"checkpoint_tip": exp_rev}
-            if prev:
-                results[prev]["data"][  # type: ignore[unreachable]
-                    "checkpoint_parent"
-                ] = rev
-            if rev in results:
-                results[rev]["data"].update(exp)
-                results.move_to_end(rev)
-            else:
-                exp.update(collected_exp["data"])
-        else:
-            exp = collected_exp["data"]
-        if rev not in results:
-            results[rev] = {"data": exp}
-        prev = rev
-    if checkpoint and prev:
-        results[prev]["data"]["checkpoint_parent"] = baseline
-    return results
-
-
-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()
-    status = ExpStatus.Running if baseline in running else ExpStatus.Success
-    results["baseline"] = collect_experiment_commit(
+) -> List["ExpState"]:
+    return collect(
         repo,
-        baseline,
-        status=status,
-        running=running,
+        revs=revs,
+        all_branches=all_branches,
+        all_tags=all_tags,
+        all_commits=all_commits,
+        num=num,
+        hide_queued=hide_queued,
+        hide_failed=hide_failed,
+        sha_only=sha_only,
         **kwargs,
     )
-    if baseline == "workspace" or _is_scm_error(results["baseline"]):
-        return results
 
-    ref_info = ExpRefInfo(baseline_sha=baseline)
-    commits: List[Tuple[str, "GitCommit", str, List[str]]] = []
 
-    assert isinstance(repo.scm, Git)
-    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)
-            revs = list(repo.scm.branch_revs(exp_rev, baseline))
-        except (SCMError, InnerScmError):
-            continue
-        commits.append((ref, commit, exp_rev, revs))
+def tabulate(
+    baseline_states: Iterable["ExpState"],
+    fill_value: Optional[str] = "-",
+    error_value: str = "!",
+    **kwargs,
+) -> Tuple["TabularData", Dict[str, Iterable[str]]]:
+    """Return table data for experiments.
 
-    for exp_ref, _, exp_rev, revs in sorted(
-        commits, key=lambda x: x[1].commit_time, reverse=True
-    ):
-        ref_info = ExpRefInfo.from_ref(exp_ref)
-        assert ref_info.baseline_sha == baseline
-        collected_exp = _collect_complete_experiment(
-            repo,
-            baseline=baseline,
-            exp_rev=exp_rev,
-            running=running,
-            revs=revs,
+    Returns:
+        Tuple of (table_data, data_headers)
+    """
+    from funcy import lconcat
+    from funcy.seqs import flatten as flatten_list
+
+    from dvc.compare import TabularData
+
+    data_names = _collect_names(baseline_states)
+    metrics_names = data_names.metrics
+    params_names = data_names.params
+    deps_names = data_names.sorted_deps
+
+    headers = [
+        "Experiment",
+        "rev",
+        "typ",
+        "Created",
+        "parent",
+        "State",
+        "Executor",
+    ]
+    names = {**metrics_names, **params_names}
+    counter = Counter(flatten_list([list(a.keys()) for a in names.values()]))
+    counter.update(headers)
+    metrics_headers = _normalize_headers(metrics_names, counter)
+    params_headers = _normalize_headers(params_names, counter)
+
+    all_headers = lconcat(headers, metrics_headers, params_headers, deps_names)
+    td = TabularData(all_headers, fill_value=fill_value)
+    td.extend(
+        _build_rows(
+            baseline_states,
+            all_headers=all_headers,
+            metrics_headers=metrics_headers,
+            params_headers=params_headers,
+            metrics_names=metrics_names,
+            params_names=params_names,
+            deps_names=deps_names,
+            fill_value=fill_value,
+            error_value=error_value,
             **kwargs,
         )
-        if _is_scm_error(collected_exp):
-            continue
-        results.update(collected_exp)
-    return results
+    )
+    data_headers: Dict[str, Iterable[str]] = {
+        "metrics": metrics_headers,
+        "params": params_headers,
+        "deps": deps_names,
+    }
+    return td, data_headers
 
 
-def get_branch_names(
-    scm: "Git", baseline_set: Iterable[str], refs: Optional[Iterable[str]] = None
-) -> Dict[str, Optional[str]]:
-    names: Dict[str, Optional[str]] = {}
-    bases = [
-        f"refs/exps/{baseline[:2]}/{baseline[2:]}/" for baseline in baseline_set
-    ] + ["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("unresolved ref %s", ref)
-                break
-    logger.debug("found refs for revs: %s", names)
-    return names
-
-
-def update_names(  # noqa: C901
-    repo: "Repo",
-    branch_names: Dict[str, Optional[str]],
-    result: Dict[str, Dict[str, Any]],
-):
-    rev_set = set()
-    for baseline in result:
-        for rev in result[baseline]:
-            if rev == "baseline":
-                rev = baseline
-            if rev != "workspace":
-                rev_set.add(rev)
-
-    if rev_set:
-        rev_set.difference_update(branch_names.keys())
-
-    exact_name = repo.experiments.get_exact_name(rev_set)
-
-    for baseline, baseline_results in result.items():
-        name_set: Set[str] = set()
-        for rev, rev_result in baseline_results.items():
-            name: Optional[str] = None
-            if rev == "baseline":
-                rev = baseline
-                if rev == "workspace":
-                    continue
-            name = branch_names.get(rev, None) or exact_name[rev]
-            if name and name not in name_set:
-                name_set.add(name)
-                rev_result["data"]["name"] = name
-
-
-def _collect_active_experiment(
-    repo: "Repo",
-    found_revs: Dict[str, List[str]],
-    running: Dict[str, Any],
+def _build_rows(
+    baseline_states: Iterable["ExpState"],
+    *,
+    all_headers: Iterable[str],
+    fill_value: Optional[str],
+    sort_by: Optional[str] = None,
+    sort_order: Optional[Literal["asc", "desc"]] = None,
     **kwargs,
-) -> Dict[str, Dict[str, Any]]:
-    result: Dict[str, Dict] = defaultdict(OrderedDict)
-    for entry in chain(
-        repo.experiments.tempdir_queue.iter_active(),
-        repo.experiments.celery_queue.iter_active(),
-    ):
-        stash_rev = entry.stash_rev
-        if entry.baseline_rev in found_revs and (
-            stash_rev not in running or not running[stash_rev].get("last")
+) -> Iterator[Tuple["CellT", ...]]:
+    for baseline in baseline_states:
+        row: Dict[str, "CellT"] = {k: fill_value for k in all_headers}
+        row["Experiment"] = ""
+        if baseline.name:
+            row["rev"] = baseline.name
+        elif Git.is_sha(baseline.rev):
+            row["rev"] = baseline.rev[:7]
+        else:
+            row["rev"] = baseline.rev
+        row["typ"] = "baseline"
+        row["parent"] = ""
+        if baseline.data:
+            row["Created"] = format_time(
+                baseline.data.timestamp, fill_value=fill_value, **kwargs
+            )
+            row.update(
+                _data_cells(  # pylint: disable=missing-kwoa
+                    baseline, fill_value=fill_value, **kwargs
+                )
+            )
+        yield tuple(row.values())
+        if baseline.experiments:
+            if sort_by:
+                metrics_names: Mapping[str, Iterable[str]] = kwargs.get(
+                    "metrics_names", {}
+                )
+                params_names: Mapping[str, Iterable[str]] = kwargs.get(
+                    "params_names", {}
+                )
+                sort_path, sort_name, sort_type = _sort_column(
+                    sort_by, metrics_names, params_names
+                )
+                reverse = sort_order == "desc"
+                experiments = _sort_exp(
+                    baseline.experiments, sort_path, sort_name, sort_type, reverse
+                )
+            else:
+                experiments = baseline.experiments
+            for i, child in enumerate(experiments):
+                yield from _exp_range_rows(
+                    child,
+                    all_headers=all_headers,
+                    fill_value=fill_value,
+                    is_base=i == len(baseline.experiments) - 1,
+                    **kwargs,
+                )
+
+
+def _sort_column(
+    sort_by: str,
+    metric_names: Mapping[str, Iterable[str]],
+    param_names: Mapping[str, Iterable[str]],
+) -> Tuple[str, str, str]:
+    path, _, sort_name = sort_by.rpartition(":")
+    matches: Set[Tuple[str, str, str]] = set()
+
+    if path:
+        if path in metric_names and sort_name in metric_names[path]:
+            matches.add((path, sort_name, "metrics"))
+        if path in param_names and sort_name in param_names[path]:
+            matches.add((path, sort_name, "params"))
+    else:
+        for path in metric_names:
+            if sort_name in metric_names[path]:
+                matches.add((path, sort_name, "metrics"))
+        for path in param_names:
+            if sort_name in param_names[path]:
+                matches.add((path, sort_name, "params"))
+
+    if len(matches) == 1:
+        return matches.pop()
+    if len(matches) > 1:
+        raise InvalidArgumentError(
+            "Ambiguous sort column '{}' matched '{}'".format(
+                sort_by,
+                ", ".join([f"{path}:{name}" for path, name, _ in matches]),
+            )
+        )
+    raise InvalidArgumentError(f"Unknown sort column '{sort_by}'")
+
+
+def _sort_exp(
+    experiments: Iterable["ExpRange"],
+    sort_path: str,
+    sort_name: str,
+    typ: str,
+    reverse: bool,
+) -> List["ExpRange"]:
+    from funcy import first
+
+    def _sort(exp_range: "ExpRange"):
+        exp = first(exp_range.revs)
+        if not exp:
+            return True
+        data = exp.data.dumpd().get(typ, {}).get(sort_path, {}).get("data", {})
+        val = flatten(data).get(sort_name)
+        return val is None, val
+
+    return sorted(experiments, key=_sort, reverse=reverse)
+
+
+def _exp_range_rows(
+    exp_range: "ExpRange",
+    *,
+    all_headers: Iterable[str],
+    fill_value: Optional[str],
+    is_base: bool = False,
+    **kwargs,
+) -> Iterator[Tuple["CellT", ...]]:
+    for i, exp in enumerate(exp_range.revs):
+        row: Dict[str, "CellT"] = {k: fill_value for k in all_headers}
+        row["Experiment"] = exp.name or ""
+        row["rev"] = exp.rev[:7] if Git.is_sha(exp.rev) else exp.rev
+        if len(exp_range.revs) > 1:
+            if i == 0:
+                row["typ"] = "checkpoint_tip"
+            elif i == len(exp_range.revs) - 1:
+                row["typ"] = "checkpoint_base"
+            else:
+                row["typ"] = "checkpoint_commit"
+        else:
+            row["typ"] = "branch_base" if is_base else "branch_commit"
+        row["parent"] = ""
+        if exp_range.executor:
+            row["State"] = exp_range.executor.state.capitalize()
+            if exp_range.executor.name:
+                row["Executor"] = exp_range.executor.name.capitalize()
+        if exp.data:
+            row["Created"] = format_time(
+                exp.data.timestamp, fill_value=fill_value, **kwargs
+            )
+            row.update(
+                _data_cells(  # pylint: disable=missing-kwoa
+                    exp, fill_value=fill_value, **kwargs
+                )
+            )
+        yield tuple(row.values())
+
+
+def _data_cells(
+    exp: "ExpState",
+    *,
+    metrics_headers: Iterable[str],
+    params_headers: Iterable[str],
+    metrics_names: Mapping[str, Iterable[str]],
+    params_names: Mapping[str, Iterable[str]],
+    deps_names: Iterable[str],
+    fill_value: Optional[str] = "-",
+    error_value: str = "!",
+    precision: Optional[int] = None,
+    **kwargs,
+) -> Iterator[Tuple[str, "CellT"]]:
+    def _d_cells(
+        d: Mapping[str, Any],
+        names: Mapping[str, Iterable[str]],
+        headers: Iterable[str],
+    ) -> Iterator[Tuple[str, "CellT"]]:
+        from dvc.compare import _format_field, with_value
+
+        for fname, data in d.items():
+            item = data.get("data", {})
+            item = flatten(item) if isinstance(item, dict) else {fname: item}
+            for name in names[fname]:
+                value = with_value(
+                    item.get(name),
+                    error_value if data.get("error") else fill_value,
+                )
+                # wrap field data in ui.rich_text, otherwise rich may
+                # interpret unescaped braces from list/dict types as rich
+                # markup tags
+                value = ui.rich_text(str(_format_field(value, precision)))
+                if name in headers:
+                    yield name, value
+                else:
+                    yield f"{fname}:{name}", value
+
+    if not exp.data:
+        return
+    yield from _d_cells(exp.data.metrics, metrics_names, metrics_headers)
+    yield from _d_cells(exp.data.params, params_names, params_headers)
+    for name in deps_names:
+        dep = exp.data.deps.get(name)
+        if dep:
+            yield name, dep.hash or fill_value
+
+
+def format_time(
+    timestamp: Optional[datetime],
+    fill_value: Optional[str] = "-",
+    iso: bool = False,
+    **kwargs,
+) -> Optional[str]:
+    if not timestamp:
+        return fill_value
+    if iso:
+        return timestamp.isoformat()
+    if timestamp.date() == date.today():
+        fmt = "%I:%M %p"
+    else:
+        fmt = "%b %d, %Y"
+    return timestamp.strftime(fmt)
+
+
+class _DataNames(NamedTuple):
+    # NOTE: we use nested dict instead of set for metrics/params names to
+    # preserve key ordering
+    metrics: Dict[str, Dict[str, Any]]
+    params: Dict[str, Dict[str, Any]]
+    deps: Set[str]
+
+    @property
+    def sorted_deps(self):
+        return sorted(self.deps)
+
+    def update(self, other: "_DataNames"):
+        def _update_d(
+            d: Dict[str, Dict[str, Any]], other_d: Mapping[str, Mapping[str, Any]]
         ):
-            collected_exp = collect_experiment_commit(
-                repo,
-                stash_rev,
-                status=ExpStatus.Running,
-                running=running,
-                **kwargs,
-            )
-            if _is_scm_error(collected_exp):
-                continue
-            result[entry.baseline_rev][stash_rev] = collected_exp
+            for k, v in other_d.items():
+                if k in d:
+                    d[k].update(v)
+                else:
+                    d[k] = dict(v)
+
+        _update_d(self.metrics, other.metrics)
+        _update_d(self.params, other.params)
+        self.deps.update(other.deps)
+
+
+def _collect_names(exp_states: Iterable["ExpState"]) -> _DataNames:
+    result = _DataNames(defaultdict(dict), defaultdict(dict), set())
+
+    def _collect_d(result_d: Dict[str, Dict[str, Any]], data_d: Dict[str, Any]):
+        for path, item in data_d.items():
+            item = item.get("data", {})
+            if isinstance(item, dict):
+                item = flatten(item)
+                result_d[path].update((key, None) for key in item)
+
+    for exp in exp_states:
+        if exp.data:
+            _collect_d(result.metrics, exp.data.metrics)
+            _collect_d(result.params, exp.data.params)
+            result.deps.update(exp.data.deps)
+        if exp.experiments:
+            for child in exp.experiments:
+                result.update(_collect_names(child.revs))
+
     return result
 
 
-def _collect_queued_experiment(
-    repo: "Repo",
-    found_revs: Dict[str, List[str]],
-    running: Dict[str, Any],
-    **kwargs,
-) -> Dict[str, Dict[str, Any]]:
-    result: Dict[str, Dict] = defaultdict(OrderedDict)
-    for entry in repo.experiments.celery_queue.iter_queued():
-        stash_rev = entry.stash_rev
-        if entry.baseline_rev in found_revs:
-            collected_exp = collect_experiment_commit(
-                repo,
-                stash_rev,
-                status=ExpStatus.Queued,
-                running=running,
-                **kwargs,
-            )
-            if _is_scm_error(collected_exp):
-                continue
-            result[entry.baseline_rev][stash_rev] = collected_exp
-    return result
-
-
-def _collect_failed_experiment(
-    repo: "Repo",
-    found_revs: Dict[str, List[str]],
-    running: Dict[str, Any],
-    **kwargs,
-) -> Dict[str, Dict[str, Any]]:
-    result: Dict[str, Dict] = defaultdict(OrderedDict)
-    for queue_done_result in repo.experiments.celery_queue.iter_failed():
-        entry = queue_done_result.entry
-        stash_rev = entry.stash_rev
-        if entry.baseline_rev in found_revs:
-            collected_exp = collect_experiment_commit(
-                repo,
-                stash_rev,
-                status=ExpStatus.Failed,
-                running=running,
-                **kwargs,
-            )
-            if _is_scm_error(collected_exp):
-                continue
-            result[entry.baseline_rev][stash_rev] = collected_exp
-    return result
-
-
-def update_new(
-    to_dict: Dict[str, Dict[str, Any]], from_dict: Dict[str, Dict[str, Any]]
-):
-    for baseline, experiments in from_dict.items():
-        for rev, experiment in experiments.items():
-            to_dict[baseline][rev] = to_dict[baseline].get(rev, experiment)
-
-
-def move_properties_to_head(result: Dict[str, Dict[str, Dict[str, Any]]]):
-    for _, baseline_results in result.items():
-        checkpoint: bool = False
-        head: Dict[str, Any] = {}
-        for rev, rev_data in baseline_results.items():
-            if (
-                "data" not in rev_data
-                or rev_data["data"].get("checkpoint_tip", None) is None
-            ):
-                checkpoint = False
-                head = {}
-                continue
-
-            rev_result: Dict[str, Any] = rev_data["data"]
-            if (
-                checkpoint is True
-                and rev_result["checkpoint_tip"] == head["checkpoint_tip"]
-            ):
-                if "name" in rev_result and "name" not in head:
-                    head["name"] = rev_result["name"]
-                    del rev_result["name"]
-                if rev_result["executor"]:
-                    if not head["executor"]:
-                        head["executor"] = rev_result["executor"]
-                    rev_result["executor"] = None
-                if rev_result["status"] == ExpStatus.Running.name:
-                    head["status"] = ExpStatus.Running.name
-                    rev_result["status"] = ExpStatus.Success.name
-            elif rev_result["checkpoint_tip"] == rev:
-                head = rev_result
-                checkpoint = True
-
-
-def show(  # noqa: PLR0913
-    repo: "Repo",
-    all_branches=False,
-    all_tags=False,
-    revs: Union[List[str], str, None] = None,
-    all_commits=False,
-    hide_queued=False,
-    hide_failed=False,
-    sha_only=False,
-    num=1,
-    param_deps=False,
-    onerror: Optional[Callable] = None,
-    fetch_running: bool = True,
-    force: bool = False,
-):
-    if repo.scm.no_commits:
-        return {}
-
-    res: Dict[str, Dict] = defaultdict(OrderedDict)
-
-    if not any([revs, all_branches, all_tags, all_commits]):
-        revs = ["HEAD"]
-    if isinstance(revs, str):
-        revs = [revs]
-
-    assert isinstance(repo.scm, Git)
-
-    found_revs: Dict[str, List[str]] = {"workspace": []}
-    found_revs.update(
-        iter_revs(repo.scm, revs, num, all_branches, all_tags, all_commits)
-    )
-    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
-    )
-
-    queued_experiment = (
-        _collect_queued_experiment(
-            repo,
-            found_revs,
-            running,
-            param_deps=param_deps,
-            force=force,
-        )
-        if not hide_queued
-        else {}
-    )
-
-    active_experiment = _collect_active_experiment(
-        repo,
-        found_revs,
-        running,
-        param_deps=param_deps,
-        onerror=onerror,
-        force=force,
-    )
-
-    failed_experiments = (
-        _collect_failed_experiment(
-            repo,
-            found_revs,
-            running,
-            param_deps=param_deps,
-            onerror=onerror,
-            force=force,
-        )
-        if not hide_failed
-        else {}
-    )
-
-    for baseline in found_revs:
-        res[baseline] = _collect_branch(
-            repo,
-            baseline,
-            running=running,
-            param_deps=param_deps,
-            onerror=onerror,
-            force=force,
-            refs=cached_refs,
-        )
-
-    update_new(res, failed_experiments)
-
-    update_new(res, active_experiment)
-
-    update_new(res, queued_experiment)
-
-    if not sha_only:
-        update_names(repo, branch_names, res)
-
-    move_properties_to_head(res)
-
-    return res
+def _normalize_headers(
+    names: Mapping[str, Mapping[str, Any]], count: Mapping[str, int]
+) -> List[str]:
+    return [
+        name if count[name] == 1 else f"{path}:{name}"
+        for path in names
+        for name in names[path]
+    ]

dvc/repo/gc.py

@@ -1,5 +1,5 @@
 import logging
-from typing import TYPE_CHECKING, List, Optional
+from typing import TYPE_CHECKING, List, Optional, Set, Tuple
 
 from dvc.exceptions import InvalidArgumentError
 
@@ -7,11 +7,22 @@ from . import locked
 
 if TYPE_CHECKING:
     from dvc.repo import Repo
+    from dvc_data.hashfile.hash_info import HashInfo
+    from dvc_objects.db import ObjectDB
 
 logger = logging.getLogger(__name__)
 
 
 def _validate_args(**kwargs):
+    not_in_remote = kwargs.pop("not_in_remote", None)
+    cloud = kwargs.pop("cloud", None)
+    remote = kwargs.pop("remote", None)
+    if remote and not (cloud or not_in_remote):
+        raise InvalidArgumentError("`--remote` requires `--cloud` or `--not-in-remote`")
+    if not_in_remote and cloud:
+        raise InvalidArgumentError(
+            "`--not-in-remote` and `--cloud` are mutually exclusive"
+        )
     if not any(kwargs.values()):
         raise InvalidArgumentError(
             "Either of `-w|--workspace`, `-a|--all-branches`, `-T|--all-tags` "
@@ -22,8 +33,22 @@ def _validate_args(**kwargs):
         raise InvalidArgumentError("`--num` can only be used alongside `--rev`")
 
 
+def _used_obj_ids_not_in_remote(repo, default_remote, jobs, remote_odb_to_obj_ids):
+    used_obj_ids = set()
+    remote_oids = set()
+    for remote_odb, obj_ids in remote_odb_to_obj_ids:
+        if remote_odb is None:
+            remote_odb = repo.cloud.get_remote_odb(default_remote, "gc --not-in-remote")
+
+        remote_oids.update(
+            remote_odb.list_oids_exists({x.value for x in obj_ids}, jobs=jobs)
+        )
+        used_obj_ids.update(obj_ids)
+    return {obj for obj in used_obj_ids if obj.value not in remote_oids}
+
+
 @locked
-def gc(  # noqa: PLR0913
+def gc(  # noqa: PLR0912, PLR0913, C901
     self: "Repo",
     all_branches: bool = False,
     cloud: bool = False,
@@ -39,6 +64,7 @@ def gc(  # noqa: PLR0913
     commit_date: Optional[str] = None,
     rev: Optional[str] = None,
     num: Optional[int] = None,
+    not_in_remote: bool = False,
 ):
     # require `workspace` to be true to come into effect.
     # assume `workspace` to be enabled if any of `all_tags`, `all_commits`,
@@ -52,6 +78,8 @@ def gc(  # noqa: PLR0913
         commit_date=commit_date,
         rev=rev,
         num=num,
+        cloud=cloud,
+        not_in_remote=not_in_remote,
     )
 
     from contextlib import ExitStack
@@ -64,13 +92,14 @@ def gc(  # noqa: PLR0913
         repos = []
     all_repos = [Repo(path) for path in repos]
 
-    used_obj_ids = set()
+    odb_to_obj_ids: List[Tuple[Optional["ObjectDB"], Set["HashInfo"]]] = []
+
     with ExitStack() as stack:
         for repo in all_repos:
             stack.enter_context(repo.lock)
 
         for repo in [*all_repos, self]:
-            for obj_ids in repo.used_objs(
+            for odb, obj_ids in repo.used_objs(
                 all_branches=all_branches,
                 with_deps=with_deps,
                 all_tags=all_tags,
@@ -82,13 +111,22 @@ def gc(  # noqa: PLR0913
                 jobs=jobs,
                 revs=[rev] if rev else None,
                 num=num or 1,
-            ).values():
-                used_obj_ids.update(obj_ids)
+            ).items():
+                odb_to_obj_ids.append((odb, obj_ids))
+
+    if not odb_to_obj_ids:
+        odb_to_obj_ids = [(None, set())]
+
+    if not_in_remote:
+        used_obj_ids = _used_obj_ids_not_in_remote(self, remote, jobs, odb_to_obj_ids)
+    else:
+        used_obj_ids = set()
+        for _, obj_ids in odb_to_obj_ids:
+            used_obj_ids.update(obj_ids)
 
     for scheme, odb in self.cache.by_scheme():
         if not odb:
             continue
-
         removed = ogc(odb, used_obj_ids, jobs=jobs)
         if not removed:
             logger.info("No unused '%s' cache to remove.", scheme)
@@ -96,9 +134,11 @@ def gc(  # noqa: PLR0913
     if not cloud:
         return
 
-    odb = self.cloud.get_remote_odb(remote, "gc -c")
-    removed = ogc(odb, used_obj_ids, jobs=jobs)
-    if removed:
-        get_index(odb).clear()
-    else:
-        logger.info("No unused cache to remove from remote.")
+    for remote_odb, obj_ids in odb_to_obj_ids:
+        if remote_odb is None:
+            remote_odb = repo.cloud.get_remote_odb(remote, "gc -c")
+        removed = ogc(remote_odb, obj_ids, jobs=jobs)
+        if removed:
+            get_index(remote_odb).clear()
+        else:
+            logger.info("No unused cache to remove from remote.")

dvc/repo/get.py

@@ -4,7 +4,6 @@ from typing import TYPE_CHECKING, Union
 
 from dvc.exceptions import DvcException
 from dvc.utils import resolve_output
-from dvc.utils.fs import remove
 
 if TYPE_CHECKING:
     from dvc.fs.dvc import DVCFileSystem
@@ -20,64 +19,39 @@ class GetDVCFileError(DvcException):
         )
 
 
-def get(url, path, out=None, rev=None, jobs=None):
-    import shortuuid
-
+def get(url, path, out=None, rev=None, jobs=None, force=False):
     from dvc.dvcfile import is_valid_filename
-    from dvc.external_repo import external_repo
     from dvc.fs.callbacks import Callback
+    from dvc.repo import Repo
 
-    out = resolve_output(path, out)
+    out = resolve_output(path, out, force=force)
 
     if is_valid_filename(out):
         raise GetDVCFileError()
 
-    # Creating a directory right beside the output to make sure that they
-    # are on the same filesystem, so we could take the advantage of
-    # reflink and/or hardlink. Not using tempfile.TemporaryDirectory
-    # because it will create a symlink to tmpfs, which defeats the purpose
-    # and won't work with reflink/hardlink.
-    dpath = os.path.dirname(os.path.abspath(out))
-    tmp_dir = os.path.join(dpath, "." + str(shortuuid.uuid()))
+    with Repo.open(
+        url=url,
+        rev=rev,
+        subrepos=True,
+        uninitialized=True,
+    ) as repo:
+        from dvc.fs.data import DataFileSystem
 
-    # Try any links possible to avoid data duplication.
-    #
-    # Not using symlink, because we need to remove cache after we
-    # are done, and to make that work we would have to copy data
-    # over anyway before removing the cache, so we might just copy
-    # it right away.
-    #
-    # Also, we can't use theoretical "move" link type here, because
-    # the same cache file might be used a few times in a directory.
-    cache_types = ["reflink", "hardlink", "copy"]
-    try:
-        with external_repo(
-            url=url,
-            rev=rev,
-            subrepos=True,
-            uninitialized=True,
-            cache_dir=tmp_dir,
-            cache_types=cache_types,
-        ) as repo:
-            from dvc.fs.data import DataFileSystem
+        fs: Union[DataFileSystem, "DVCFileSystem"]
+        if os.path.isabs(path):
+            fs = DataFileSystem(index=repo.index.data["local"])
+            fs_path = fs.from_os_path(path)
+        else:
+            fs = repo.dvcfs
+            fs_path = fs.from_os_path(path)
 
-            fs: Union[DataFileSystem, "DVCFileSystem"]
-            if os.path.isabs(path):
-                fs = DataFileSystem(index=repo.index.data["local"])
-                fs_path = fs.from_os_path(path)
-            else:
-                fs = repo.dvcfs
-                fs_path = fs.from_os_path(path)
-
-            with Callback.as_tqdm_callback(
-                desc=f"Downloading {fs.path.name(path)}",
-                unit="files",
-            ) as cb:
-                fs.get(
-                    fs_path,
-                    os.path.abspath(out),
-                    batch_size=jobs,
-                    callback=cb,
-                )
-    finally:
-        remove(tmp_dir)
+        with Callback.as_tqdm_callback(
+            desc=f"Downloading {fs.path.name(path)}",
+            unit="files",
+        ) as cb:
+            fs.get(
+                fs_path,
+                os.path.abspath(out),
+                batch_size=jobs,
+                callback=cb,
+            )

dvc/repo/get_url.py

@@ -6,8 +6,8 @@ from dvc.fs import download, parse_external_url
 from dvc.utils import resolve_output
 
 
-def get_url(url, out=None, *, config=None, jobs=None):
-    out = resolve_output(url, out)
+def get_url(url, out=None, *, config=None, jobs=None, force=False):
+    out = resolve_output(url, out, force=force)
     out = os.path.abspath(out)
     (out,) = output.loads_from(None, [out], use_cache=False)
 

dvc/repo/imp_url.py

@@ -30,15 +30,17 @@ def imp_url(  # noqa: C901, PLR0913
     labels=None,
     meta=None,
     jobs=None,
+    force=False,
     fs_config=None,
     version_aware: bool = False,
 ):
-    out = resolve_output(url, out)
+    out = resolve_output(url, out, force=force)
     path, wdir, out = resolve_paths(self, out, always_local=to_remote and not out)
 
-    if to_remote and (no_exec or no_download):
+    if to_remote and (no_exec or no_download or version_aware):
         raise InvalidArgumentError(
-            "--no-exec/--no-download cannot be combined with --to-remote"
+            "--no-exec/--no-download/--version-aware cannot be combined with "
+            "--to-remote"
         )
 
     if not to_remote and remote:

dvc/repo/metrics/show.py

@@ -47,7 +47,7 @@ def _collect_metrics(repo, targets, recursive):
 
 
 def _extract_metrics(metrics, path, rev):
-    if isinstance(metrics, (int, float)):
+    if isinstance(metrics, (int, float, str)):
         return metrics
 
     if not isinstance(metrics, dict):

dvc/repo/open_repo.py

@@ -3,7 +3,7 @@ import os
 import tempfile
 import threading
 from contextlib import contextmanager
-from typing import TYPE_CHECKING, Collection, Dict, Iterator, Optional, Tuple
+from typing import TYPE_CHECKING, Dict, Iterator, Optional, Tuple
 
 from funcy import retry, wrap_with
 
@@ -13,21 +13,17 @@ from dvc.scm import CloneError, map_scm_exception
 from dvc.utils import relpath
 
 if TYPE_CHECKING:
-    from dvc.types import StrPath
-
-    from .scm import Git
+    from dvc.scm import Git
 
 logger = logging.getLogger(__name__)
 
 
 @contextmanager
 @map_scm_exception()
-def external_repo(
+def _external_repo(
     url,
     rev: Optional[str] = None,
     for_write: bool = False,
-    cache_dir: Optional["StrPath"] = None,
-    cache_types: Optional[Collection[str]] = None,
     **kwargs,
 ) -> Iterator["Repo"]:
     logger.debug("Creating external repo %s@%s", url, rev)
@@ -37,12 +33,8 @@ def external_repo(
     # the tip of the default branch
     rev = rev or "refs/remotes/origin/HEAD"
 
-    cache_config = {
-        "cache": {"dir": cache_dir or _get_cache_dir(url), "type": cache_types}
-    }
-
     config = _get_remote_config(url) if os.path.isdir(url) else {}
-    config.update(cache_config)
+    config.update({"cache": {"dir": _get_cache_dir(url)}})
     config.update(kwargs.pop("config", None) or {})
 
     main_root = "/"
@@ -55,7 +47,7 @@ def external_repo(
         root_dir=path,
         url=url,
         config=config,
-        repo_factory=erepo_factory(url, main_root, cache_config),
+        repo_factory=erepo_factory(url, main_root, {"cache": config["cache"]}),
         rev=rev,
         **kwargs,
     )
@@ -70,6 +62,22 @@ def external_repo(
             _remove(path)
 
 
+def open_repo(url, *args, **kwargs):
+    if url is None:
+        url = os.getcwd()
+
+    if os.path.exists(url):
+        try:
+            config = _get_remote_config(url)
+            config.update(kwargs.get("config") or {})
+            kwargs["config"] = config
+            return Repo(url, *args, **kwargs)
+        except NotDvcRepoError:
+            pass  # fallthrough to _external_repo
+
+    return _external_repo(url, *args, **kwargs)
+
+
 def erepo_factory(url, root_dir, cache_config):
     from dvc.fs import localfs
 

dvc/repo/plots/__init__.py

@@ -121,51 +121,58 @@ class Plots:
 
             }
         """
+        from dvc.repo.experiments.brancher import switch_repo
         from dvc.utils.collections import ensure_list
 
         targets = ensure_list(targets)
         targets = [self.repo.dvcfs.from_os_path(target) for target in targets]
 
-        for rev in self.repo.brancher(revs=revs):
-            # .brancher() adds unwanted workspace
-            if revs is not None and rev not in revs:
-                continue
-            rev = rev or "workspace"
-
-            res: Dict = {}
-            definitions = _collect_definitions(
-                self.repo,
-                targets=targets,
-                revision=rev,
-                onerror=onerror,
-                props=props,
-            )
-            if definitions:
-                res[rev] = {"definitions": definitions}
-
-                data_targets = _get_data_targets(definitions)
-
-                res[rev]["sources"] = self._collect_data_sources(
-                    targets=data_targets,
-                    recursive=recursive,
-                    props=props,
+        if revs is None:
+            revs = ["workspace"]
+        else:
+            revs = list(revs)
+            if "workspace" in revs:
+                # reorder revs to match repo.brancher ordering
+                revs.remove("workspace")
+                revs = ["workspace"] + revs
+        for rev in revs:
+            with switch_repo(self.repo, rev) as (repo, _):
+                res: Dict = {}
+                definitions = _collect_definitions(
+                    repo,
+                    targets=targets,
+                    revision=rev,
                     onerror=onerror,
+                    props=props,
                 )
-            yield res
+                if definitions:
+                    res[rev] = {"definitions": definitions}
+
+                    data_targets = _get_data_targets(definitions)
+
+                    res[rev]["sources"] = self._collect_data_sources(
+                        repo,
+                        targets=data_targets,
+                        recursive=recursive,
+                        props=props,
+                        onerror=onerror,
+                    )
+                yield res
 
     @error_handler
     def _collect_data_sources(
         self,
+        repo: "Repo",
         targets: Optional[List[str]] = None,
         recursive: bool = False,
         props: Optional[Dict] = None,
         onerror: Optional[Callable] = None,
     ):
-        fs = self.repo.dvcfs
+        fs = repo.dvcfs
 
         props = props or {}
 
-        plots = _collect_plots(self.repo, targets, recursive)
+        plots = _collect_plots(repo, targets, recursive)
         res: Dict[str, Any] = {}
         for fs_path, rev_props in plots.items():
             joined_props = {**rev_props, **props}

dvc/repo/reproduce.py

@@ -3,9 +3,8 @@ from functools import partial
 from typing import TYPE_CHECKING, Iterator, List
 
 from dvc.exceptions import DvcException, ReproductionError
-from dvc.repo.metrics.show import _collect_top_level_metrics
-from dvc.repo.params.show import _collect_top_level_params
 from dvc.repo.scm_context import scm_context
+from dvc.stage.cache import RunCacheNotSupported
 from dvc.stage.exceptions import CheckpointKilledError
 
 from . import locked
@@ -76,20 +75,9 @@ def _track_stage(stage: "Stage") -> None:
     return context.track_changed_files()
 
 
-def _track_top_level(repo: "Repo") -> None:
-    context = repo.scm_context
-    for metric_file in _collect_top_level_metrics(repo):
-        context.track_file(metric_file)
-    for param_file in _collect_top_level_params(repo):
-        context.track_file(param_file)
-    for plot_file in repo.index._plot_sources:  # pylint: disable=protected-access
-        context.track_file(plot_file)
-    context.track_changed_files()
-
-
 @locked
 @scm_context
-def reproduce(  # noqa: C901
+def reproduce(  # noqa: C901, PLR0912
     self: "Repo",
     targets=None,
     recursive=False,
@@ -138,13 +126,23 @@ def reproduce(  # noqa: C901
                 )
             )
 
-    result = _reproduce_stages(self.index.graph, list(stages), **kwargs)
-    _track_top_level(self)
-    return result
+    if kwargs.get("pull", False) and kwargs.get("run_cache", True):
+        logger.debug("Pulling run cache")
+        try:
+            self.stage_cache.pull(None)
+        except RunCacheNotSupported as e:
+            logger.warning("Failed to pull run cache: %s", e)
+
+    return _reproduce_stages(self.index.graph, list(stages), **kwargs)
 
 
 def _reproduce_stages(  # noqa: C901
-    graph, 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.
 

dvc/repo/stage.py

@@ -122,16 +122,15 @@ class StageLoad:
             force=force,
             **stage_data,
         )
-        with self.repo.scm_context:
-            stage.dump(update_lock=update_lock)
-            try:
-                stage.ignore_outs()
-            except FileNotFoundError as exc:
-                ui.warn(
-                    f"Could not create .gitignore entry in {exc.filename}."
-                    " DVC will attempt to create .gitignore entry again when"
-                    " the stage is run."
-                )
+        stage.dump(update_lock=update_lock)
+        try:
+            stage.ignore_outs()
+        except FileNotFoundError as exc:
+            ui.warn(
+                f"Could not create .gitignore entry in {exc.filename}."
+                " DVC will attempt to create .gitignore entry again when"
+                " the stage is run."
+            )
 
         return stage
 

dvc/repo/worktree.py

@@ -173,7 +173,7 @@ def push_worktree(
         ) as cb:
             cb.set_size(total)
             try:
-                pushed += checkout(
+                stats = checkout(
                     new_index,
                     remote_obj.path,
                     remote_obj.fs,
@@ -184,6 +184,7 @@ def push_worktree(
                     jobs=jobs,
                     **diff_kwargs,
                 )
+                pushed += sum(len(changes) for changes in stats.values())
             except VersioningNotSupported:
                 logger.exception("")
                 raise DvcException(

dvc/schema.py

@@ -116,7 +116,6 @@ PLOT_DEFINITION = {
 SINGLE_PLOT_SCHEMA = {str: Any(PLOT_DEFINITION, None)}
 ARTIFACTS = "artifacts"
 SINGLE_ARTIFACT_SCHEMA = Schema({str: ARTIFACT_SCHEMA})
-ARTIFACTS_SCHEMA = Any(str, SINGLE_ARTIFACT_SCHEMA)
 FOREACH_IN = {
     Required(FOREACH_KWD): Any(dict, list, str),
     Required(DO_KWD): STAGE_DEFINITION,
@@ -130,7 +129,7 @@ MULTI_STAGE_SCHEMA = {
     VARS_KWD: VARS_SCHEMA,
     StageParams.PARAM_PARAMS: [str],
     StageParams.PARAM_METRICS: [str],
-    ARTIFACTS: ARTIFACTS_SCHEMA,
+    ARTIFACTS: SINGLE_ARTIFACT_SCHEMA,
 }
 
 COMPILED_SINGLE_STAGE_SCHEMA = Schema(SINGLE_STAGE_SCHEMA)

dvc/stage/__init__.py

@@ -319,20 +319,22 @@ class Stage(params.StageParams):
             env.update(current)
         return env
 
-    def changed_deps(self) -> bool:
+    def changed_deps(self, allow_missing: bool = False) -> bool:
         if self.frozen:
             return False
 
         if self.is_callback or self.always_changed or self.is_checkpoint:
             return True
 
-        return self._changed_deps()
+        return self._changed_deps(allow_missing=allow_missing)
 
     @rwlocked(read=["deps"])
-    def _changed_deps(self) -> bool:
+    def _changed_deps(self, allow_missing: bool = False) -> bool:
         for dep in self.deps:
             status = dep.status()
             if status:
+                if allow_missing and status[str(dep)] == "deleted":
+                    continue
                 logger.debug(
                     "Dependency '%s' of %s changed because it is '%s'.",
                     dep,
@@ -343,10 +345,12 @@ class Stage(params.StageParams):
         return False
 
     @rwlocked(read=["outs"])
-    def changed_outs(self) -> bool:
+    def changed_outs(self, allow_missing: bool = False) -> bool:
         for out in self.outs:
             status = out.status()
             if status:
+                if allow_missing and status[str(out)] == "not in cache":
+                    continue
                 logger.debug(
                     "Output '%s' of %s changed because it is '%s'.",
                     out,
@@ -364,13 +368,13 @@ class Stage(params.StageParams):
         return changed
 
     @rwlocked(read=["deps", "outs"])
-    def changed(self) -> bool:
+    def changed(self, allow_missing: bool = False) -> bool:
         is_changed = (
             # Short-circuit order: stage md5 is fast,
             # deps are expected to change
             self.changed_stage()
-            or self.changed_deps()
-            or self.changed_outs()
+            or self.changed_deps(allow_missing=allow_missing)
+            or self.changed_outs(allow_missing=allow_missing)
         )
         if is_changed:
             logger.debug("%s changed.", self)
@@ -416,12 +420,15 @@ class Stage(params.StageParams):
         (out,) = self.outs
         out.transfer(source, odb=odb, jobs=kwargs.get("jobs"))
         if not to_remote:
-            out.checkout()
+            out.checkout(force=kwargs.get("force"))
             out.ignore()
 
     @rwlocked(read=["deps"], write=["outs"])
     def reproduce(self, interactive=False, **kwargs) -> Optional["Stage"]:
-        if not (kwargs.get("force", False) or self.changed()):
+        if not (
+            kwargs.get("force", False)
+            or self.changed(kwargs.get("allow_missing", False))
+        ):
             if not isinstance(self, PipelineStage) and self.is_data_source:
                 logger.info("'%s' didn't change, skipping", self.addressing)
             else:
@@ -495,34 +502,34 @@ class Stage(params.StageParams):
                 if not allow_missing:
                     raise
 
-    def save_outs(self, allow_missing: bool = False):
-        from dvc.output import OutputDoesNotExistError
-
+    def get_versioned_outs(self) -> Dict[str, "Output"]:
         from .exceptions import StageFileDoesNotExistError, StageNotFound
 
         try:
             old = self.reload()
-            old_outs = {out.def_path: out for out in old.outs}
-            merge_versioned = any(
-                (
-                    out.files is not None
-                    or (out.meta is not None and out.meta.version_id is not None)
-                )
-                for out in old_outs.values()
-            )
         except (StageFileDoesNotExistError, StageNotFound):
-            merge_versioned = False
+            return {}
 
+        return {
+            out.def_path: out
+            for out in old.outs
+            if out.files is not None
+            or (out.meta is not None and out.meta.version_id is not None)
+        }
+
+    def save_outs(self, allow_missing: bool = False):
+        from dvc.output import OutputDoesNotExistError
+
+        old_versioned_outs = self.get_versioned_outs()
         for out in self.outs:
             try:
                 out.save()
             except OutputDoesNotExistError:
                 if not (allow_missing or out.checkpoint):
                     raise
-            if merge_versioned:
-                old_out = old_outs.get(out.def_path)
-                if old_out is not None:
-                    out.merge_version_meta(old_out)
+
+            if old_out := old_versioned_outs.get(out.def_path):
+                out.merge_version_meta(old_out)
 
     def ignore_outs(self) -> None:
         for out in self.outs:
@@ -560,8 +567,31 @@ class Stage(params.StageParams):
         if link_failures:
             raise CacheLinkError(link_failures)
 
+    @rwlocked(write=["outs"])
+    def add_outs(  # noqa: C901
+        self, filter_info=None, allow_missing: bool = False, **kwargs
+    ):
+        from dvc.output import OutputDoesNotExistError
+
+        link_failures = []
+        old_versioned_outs = self.get_versioned_outs()
+        for out in self.filter_outs(filter_info):
+            try:
+                out.add(filter_info, **kwargs)
+            except (FileNotFoundError, OutputDoesNotExistError):
+                if not (allow_missing or out.checkpoint):
+                    raise
+            except CacheLinkError:
+                link_failures.append(filter_info or out.fs_path)
+
+            if old_out := old_versioned_outs.get(out.def_path):
+                out.merge_version_meta(old_out)
+
+        if link_failures:
+            raise CacheLinkError(link_failures)
+
     @rwlocked(read=["deps", "outs"])
-    def run(
+    def run(  # noqa: C901
         self,
         dry=False,
         no_commit=False,
@@ -573,15 +603,20 @@ class Stage(params.StageParams):
         if (self.cmd or self.is_import) and not self.frozen and not dry:
             self.remove_outs(ignore_remove=False, force=False)
 
-        if (not self.frozen and self.is_import) or self.is_partial_import:
+        if (
+            self.is_import and (not self.frozen or kwargs.get("pull"))
+        ) or self.is_partial_import:
             self._sync_import(dry, force, kwargs.get("jobs", None), no_download)
         elif not self.frozen and self.cmd:
-            self._run_stage(dry, force, **kwargs)
-        else:
+            self._run_stage(dry, force, allow_missing=allow_missing, **kwargs)
+        elif kwargs.get("pull"):
+            logger.info("Pulling data for %s", self)
+            self.repo.pull(self.addressing, jobs=kwargs.get("jobs", None))
+            self.checkout()
+        elif not dry:
             args = ("outputs", "frozen ") if self.frozen else ("data sources", "")
             logger.info("Verifying %s in %s%s", *args, self)
-            if not dry:
-                self._check_missing_outputs()
+            self._check_missing_outputs()
 
         if not dry:
             if kwargs.get("checkpoint_func", None) or no_download:

dvc/stage/run.py

@@ -133,10 +133,25 @@ def cmd_run(stage, dry=False, checkpoint_func=None, run_env=None):
         _run(stage, executable, cmd, checkpoint_func=checkpoint_func, **kwargs)
 
 
+def _pull_missing_deps(stage):
+    for dep in stage.deps:
+        if not dep.exists:
+            stage.repo.pull(dep.def_path)
+
+
 def run_stage(
-    stage, dry=False, force=False, checkpoint_func=None, run_env=None, **kwargs
+    stage,
+    dry=False,
+    force=False,
+    checkpoint_func=None,
+    run_env=None,
+    allow_missing: bool = False,
+    **kwargs,
 ):
     if not (force or checkpoint_func):
+        if allow_missing and kwargs.get("pull") and not dry:
+            _pull_missing_deps(stage)
+
         from .cache import RunCacheNotFoundError
 
         try:

dvc/testing/README.rst

@@ -0,0 +1,19 @@
+DVC pytest plugin
+
+dvc.testing.benchmarks
+======================
+
+Benchmark test definitions are now part of ``dvc.testing``.
+For CLI usage and `bench.dvc.org <https://bench.dvc.org>`_ details see `dvc-bench <https://github.com/iterative/dvc-bench>`_.
+
+``dvc.testing.benchmarks`` structure:
+
+* cli: should be able to run these with any dvc (rpm, deb, pypi, snap, etc) (could be used in dvc-test repo too)
+
+  * commands: granular tests for individual commands. These should have a cached setup, so that we could use them during rapid development instead of our hand-written scripts. Every test could be run in a separate machine.
+  * stories: multistage start-to-end benchmarks, useful for testing workflows (e.g. in documentation, see test_sharing inspired by `Storing and sharing <https://dvc.org/doc/start/data-management/data-versioning#storing-and-sharing>`_. Every full story could be run in a separate machine.
+
+* api: for python api only.
+
+  * methods: granular tests for individual methods (e.g. ``api.open/read``). Same reasoning as in ``cli.commands``
+  * stories: same as ``cli.stories`` but for our api. E.g. imagine using our api with pandas or smth like that.

dvc/testing/benchmarks/cli/commands/test_add.py

@@ -0,0 +1,3 @@
+# pylint: disable=unused-argument
+def test_add(bench_dvc, tmp_dir, dvc, dataset):
+    bench_dvc("add", dataset)

dvc/testing/benchmarks/cli/commands/test_checkout.py

@@ -0,0 +1,39 @@
+# pylint: disable=unused-argument
+import os
+
+import pytest
+
+from dvc.fs import localfs
+from dvc_objects.fs import generic
+from dvc_objects.fs.utils import tmp_fname
+
+
+def _skip_unsupported_link(src, dest, link_type):
+    src_test_file = os.path.join(src, tmp_fname())
+    dest_test_file = os.path.join(dest, tmp_fname())
+    if not generic.test_links(
+        [link_type], localfs, src_test_file, localfs, dest_test_file
+    ):
+        pytest.skip(f"{link_type} not supported")
+
+
+def generate_test(*, link_type="copy"):
+    def _test_checkout_func(bench_dvc, tmp_dir, dvc, make_dataset):
+        dataset = make_dataset(dvcfile=True, cache=True, files=False)
+
+        _skip_unsupported_link((tmp_dir / ".dvc" / "cache"), tmp_dir, link_type)
+
+        with dvc.config.edit() as conf:
+            conf["cache"]["type"] = link_type
+
+        bench_dvc("checkout", dataset)
+        bench_dvc("checkout", name="noop")
+        (dataset / "new").write_text("new")
+        bench_dvc("checkout", "--force", name="update")
+
+    return _test_checkout_func
+
+
+test_checkout_copy = generate_test(link_type="copy")
+test_checkout_symlink = generate_test(link_type="symlink")
+test_checkout_hardlink = generate_test(link_type="hardlink")

dvc/testing/benchmarks/cli/commands/test_data_status.py

@@ -0,0 +1,49 @@
+# pylint: disable=unused-argument
+from shutil import rmtree
+
+import pytest
+
+pytestmark = pytest.mark.requires(minversion=(2, 15, 0), reason="new command")
+
+
+def test_data_status(
+    bench_dvc,
+    tmp_dir,
+    scm,
+    dvc,
+    make_dataset,
+):
+    args = ("data", "status")
+    dataset = make_dataset(cache=True, files=True, dvcfile=True, commit=False)
+    rmtree(dvc.tmp_dir)
+
+    bench_dvc(*args, name="new")
+    bench_dvc(*args, name="noop")
+
+    tmp_dir.scm_add(dataset.with_suffix(".dvc").name, commit="add dataset")
+
+    (dataset / "new").write_text("new")
+    bench_dvc(*args, name="changed")
+    bench_dvc(*args, name="changed-noop")
+
+
+def test_data_status_all_flags(bench_dvc, tmp_dir, scm, dvc, make_dataset):
+    args = (
+        "data",
+        "status",
+        "--granular",
+        "--unchanged",
+        "--untracked-files",
+        "--json",
+    )
+    dataset = make_dataset(cache=True, files=True, dvcfile=True, commit=False)
+    rmtree(dvc.tmp_dir)
+
+    bench_dvc(*args, name="new")
+    bench_dvc(*args, name="noop")
+
+    tmp_dir.scm_add(dataset.with_suffix(".dvc").name, commit="add dataset")
+
+    (dataset / "new").write_text("new")
+    bench_dvc(*args, name="changed")
+    bench_dvc(*args, name="changed-noop")

dvc/testing/benchmarks/cli/commands/test_diff.py

@@ -0,0 +1,9 @@
+# pylint: disable=unused-argument
+def test_diff(bench_dvc, tmp_dir, scm, dvc, make_dataset):
+    dataset = make_dataset(cache=True, files=True, dvcfile=True, commit=True)
+    bench_dvc("diff")
+    bench_dvc("diff", name="noop")
+
+    (dataset / "new").write_text("new")
+    bench_dvc("diff", name="changed")
+    bench_dvc("diff", name="changed-noop")

dvc/testing/benchmarks/cli/commands/test_exp_show.py

@@ -0,0 +1,14 @@
+# pylint: disable=unused-argument
+import pytest
+
+
+@pytest.mark.requires(minversion=(2, 10, 0))
+def test_exp_show(make_project, monkeypatch, bench_dvc, dvc_bin):
+    url = "https://github.com/efiop/lstm_seq2seq"
+    rev = "dvc-bench"
+    path = make_project(url, rev=rev)
+    monkeypatch.chdir(path)
+
+    dvc_bin("exp", "pull", "-A", "--no-cache", "origin")
+
+    bench_dvc("exp", "show", "-A", "--no-pager")

dvc/testing/benchmarks/cli/commands/test_gc.py

@@ -0,0 +1,4 @@
+# pylint: disable=unused-argument
+def test_gc(bench_dvc, tmp_dir, dvc, make_dataset):
+    make_dataset(files=False, cache=True, dvcfile=False)
+    bench_dvc("gc", "-f", "-w")

dvc/testing/benchmarks/cli/commands/test_get.py

@@ -0,0 +1,6 @@
+# pylint: disable=unused-argument
+def test_get(bench_dvc, tmp_dir, scm, dvc, make_dataset, remote):
+    dataset = make_dataset(
+        cache=False, files=False, dvcfile=True, commit=True, remote=True
+    )
+    bench_dvc("get", tmp_dir, dataset.name, "-o", "new")

dvc/testing/benchmarks/cli/commands/test_get_url.py

@@ -0,0 +1,6 @@
+# pylint: disable=unused-argument
+def test_get_url(bench_dvc, tmp_dir, scm, dvc, make_dataset):
+    dataset = make_dataset(
+        cache=False, files=True, dvcfile=False, commit=False, remote=False
+    )
+    bench_dvc("get-url", str(dataset), "new")

dvc/testing/benchmarks/cli/commands/test_help.py

@@ -0,0 +1,2 @@
+def test_help(bench_dvc):
+    bench_dvc("--help", rounds=100)

dvc/testing/benchmarks/cli/commands/test_import.py

@@ -0,0 +1,6 @@
+# pylint: disable=unused-argument
+def test_import(bench_dvc, tmp_dir, scm, dvc, make_dataset, remote):
+    dataset = make_dataset(
+        cache=False, files=False, dvcfile=True, commit=True, remote=True
+    )
+    bench_dvc("import", tmp_dir, dataset.name, "-o", "new")

dvc/testing/benchmarks/cli/commands/test_import_url.py

@@ -0,0 +1,6 @@
+# pylint: disable=unused-argument
+def test_import_url(bench_dvc, tmp_dir, scm, dvc, make_dataset):
+    dataset = make_dataset(
+        cache=False, files=True, dvcfile=False, commit=False, remote=False
+    )
+    bench_dvc("import-url", str(dataset), "new")

dvc/testing/benchmarks/cli/commands/test_init.py

@@ -0,0 +1,14 @@
+# pylint: disable=unused-argument
+import shutil
+
+
+def test_init(bench_dvc, tmp_dir, scm):
+    def _cleanup_dir():
+        for item in tmp_dir.iterdir():
+            if item.is_dir():
+                if item.name != ".git":
+                    shutil.rmtree(item)
+            else:
+                item.unlink()
+
+    bench_dvc("init", setup=_cleanup_dir, rounds=100, warmup_rounds=1)

dvc/testing/benchmarks/cli/commands/test_ls.py

@@ -0,0 +1,7 @@
+# pylint: disable=unused-argument
+def test_list(bench_dvc, tmp_dir, scm, dvc, make_dataset, remote):
+    make_dataset(cache=False, files=False, dvcfile=True, commit=True, remote=True)
+    bench_dvc("list", tmp_dir)
+    bench_dvc("list", tmp_dir, "--dvc-only", name="dvc-only")
+    bench_dvc("list", tmp_dir, "--recursive", name="recursive")
+    bench_dvc("list", tmp_dir, "dataset", name="shallow")

dvc/testing/benchmarks/cli/commands/test_plots.py

@@ -0,0 +1,20 @@
+import pytest
+
+from dvc.repo import Repo
+from dvc.testing.benchmarks.fixtures import _pull
+
+# pylint: disable=redefined-outer-name,unused-argument
+
+# pylint: disable=unused-argument
+
+
+@pytest.mark.requires(minversion=(2, 34, 0), reason="top-level plots not supported")
+def test_plots(project, bench_dvc):
+    with Repo() as dvc:
+        _pull(dvc)
+
+    kwargs = {"rounds": 5, "iterations": 3, "warmup_rounds": 2}
+    bench_dvc("plots", "show", name="show", **kwargs)
+    bench_dvc("plots", "show", "--json", name="show-json", **kwargs)
+    bench_dvc("plots", "diff", "HEAD", name="diff", **kwargs)
+    bench_dvc("plots", "diff", "HEAD", "--json", name="diff-json", **kwargs)

dvc/testing/benchmarks/cli/commands/test_pull.py

@@ -0,0 +1,4 @@
+# pylint: disable=unused-argument
+def test_pull(bench_dvc, tmp_dir, dvc, make_dataset, remote):
+    make_dataset(cache=False, dvcfile=True, files=False, remote=True)
+    bench_dvc("pull")

dvc/testing/benchmarks/cli/commands/test_push.py

@@ -0,0 +1,4 @@
+# pylint: disable=unused-argument
+def test_push(bench_dvc, tmp_dir, dvc, make_dataset, remote):
+    make_dataset(cache=True, dvcfile=True, files=False)
+    bench_dvc("push")

dvc/testing/benchmarks/cli/commands/test_status.py

@@ -0,0 +1,9 @@
+# pylint: disable=unused-argument
+def test_status(bench_dvc, tmp_dir, dvc, make_dataset):
+    dataset = make_dataset(files=True, dvcfile=True, cache=True)
+    bench_dvc("status")
+    bench_dvc("status", name="noop")
+
+    (dataset / "new").write_text("new")
+    bench_dvc("status", name="changed")
+    bench_dvc("status", name="changed-noop")

dvc/testing/benchmarks/cli/commands/test_update.py

@@ -0,0 +1,9 @@
+# pylint: disable=unused-argument
+def test_update(bench_dvc, tmp_dir, dvc, make_dataset):
+    dataset = make_dataset(
+        cache=False, files=True, dvcfile=False, commit=False, remote=False
+    )
+    dvc.imp_url(str(dataset), "new")
+    (dataset / "new").write_text("new")
+    bench_dvc("update", "new.dvc")
+    bench_dvc("update", "new.dvc", name="noop")

dvc/testing/benchmarks/cli/stories/test_modify_data.py

@@ -0,0 +1,57 @@
+"""
+Tests for modifications to an existing dataset.
+"""
+import glob
+import os
+import random
+import shutil
+import sys
+
+import pytest
+
+# pylint: disable=unused-argument,unexpected-keyword-arg
+
+
+@pytest.mark.skipif(sys.version_info < (3, 10), reason="requires 3.10 glob.glob")
+def test_partial_add(bench_dvc, tmp_dir, dvc, dataset, remote):
+    random.seed(4231)
+    # Move some files to create a partial dataset
+    os.makedirs("partial-copy")
+    for f in glob.glob("*", root_dir=dataset, recursive=True):  # type: ignore[call-arg]
+        if random.random() > 0.5:  # noqa: S311 # nosec
+            shutil.move(dataset / f, tmp_dir / "partial-copy" / f)
+
+    # Add/push partial dataset
+    dvc.add(str(dataset))
+    dvc.push()
+
+    # Add more files to the dataset
+    shutil.copytree("partial-copy", dataset, dirs_exist_ok=True)
+
+    # Benchmark operations for adding files to a dataset
+    bench_dvc("add", dataset, name="partial-add")
+    bench_dvc("push", name="partial-add")
+    bench_dvc("gc", "-f", "-w", name="noop")
+    bench_dvc("gc", "-f", "-w", "-c", name="cloud-noop")
+
+
+@pytest.mark.skipif(sys.version_info < (3, 10), reason="requires 3.10 glob.glob")
+def test_partial_remove(bench_dvc, tmp_dir, dvc, dataset, remote):
+    random.seed(5232)
+    # Add/push full dataset
+    dvc.add(str(dataset))
+    dvc.push()
+
+    # Remove some files
+    for f in glob.glob("*", root_dir=dataset, recursive=True):  # type: ignore[call-arg]
+        if random.random() > 0.5:  # noqa: S311 # nosec
+            if os.path.isfile(dataset / f):
+                os.remove(dataset / f)
+            elif os.path.isdir(dataset / f):
+                shutil.rmtree(dataset / f)
+
+    # Benchmark operations for removing files from dataset
+    bench_dvc("add", dataset)
+    bench_dvc("push")
+    bench_dvc("gc", "-f", "-w")
+    bench_dvc("gc", "-f", "-w", "-c", name="cloud")