dvc: bump to 2.8.1

Per https://packaging.python.org/specifications/core-metadata/#description-content-type

.github/workflows/benchmarks.yaml

@@ -0,0 +1,23 @@
+name: benchmarks
+on: [pull_request, workflow_dispatch]
+
+jobs:
+  bench:
+    name: run benchmarks
+    runs-on: ${{ matrix.os }}
+    strategy:
+      fail-fast: false
+      matrix:
+        os: [ubuntu-18.04]
+    steps:
+      - name: Cancel Previous Runs
+        uses: styfle/[email protected]
+        with:
+          access_token: ${{ github.token }}
+      - uses: actions/setup-python@v2
+        with:
+            python-version: 3.7
+      - uses: iterative/dvc-bench@master
+        with:
+            pytest_options: "-k 'test_init or test_help'"
+

.github/workflows/tests.yaml

@@ -16,6 +16,10 @@ jobs:
     timeout-minutes: 10
     runs-on: ubuntu-latest
     steps:
+    - name: Cancel Previous Runs
+      uses: styfle/[email protected]
+      with:
+        access_token: ${{ github.token }}
     - uses: actions/[email protected]
     - name: Set up Python 3.8
       uses: actions/[email protected]
@@ -38,17 +42,12 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-18.04, windows-2019, macos-10.15]
-        pyv: ["3.6", "3.7", "3.8", "3.9"]
-        include:
-          - os: "macos-10.15"
-            pyv: "3.10.0-rc.1"
+        os: [ubuntu-20.04, windows-latest, macos-latest]
+        pyv: ["3.6", "3.7", "3.8", "3.9", "3.10"]
         exclude:
-          - os: macos-10.15
-            pyv: "3.9"
-          - os: macos-10.15
-            pyv: "3.7"
-
+          # no wheels for pygit2 yet
+          - os: windows-latest
+            pyv: "3.10"
     steps:
     - uses: actions/[email protected]
     - name: Set up Python
@@ -64,13 +63,9 @@ jobs:
       uses: actions/[email protected]
       with:
         path: ${{ steps.pip-cache-dir.outputs.dir }}
-        key: ${{ runner.os }}-pip-${{ hashFiles('setup.py') }}
+        key: ${{ runner.os }}-pip-${{ hashFiles('setup.cfg') }}
         restore-keys: |
           ${{ runner.os }}-pip-
-    - name: install libgit2 on Python3.10 in macOS
-      if: matrix.pyv == '3.10.0-rc.1' && matrix.os == 'macos-10.15'
-      run: |
-        brew install libgit2
     - name: install
       if: steps.cache.pip-cache-dir.cache-hit != 'true'
       run: |
@@ -99,7 +94,7 @@ jobs:
       run: |
         echo "extra_test_args=--disable-gdrive" >> $GITHUB_ENV
     - name: run tests
-      timeout-minutes: 30
+      timeout-minutes: 40
       env:
         GDRIVE_CREDENTIALS_DATA: ${{ secrets.GDRIVE_CREDENTIALS_DATA }}
         AWS_ACCESS_KEY_ID: ${{ secrets.AWS_ACCESS_KEY_ID }}
@@ -112,7 +107,7 @@ jobs:
         --cov-report=xml --cov-report=term
         --all ${{ env.extra_test_args }}
     - name: upload coverage report
-      uses: codecov/codecov-action@v2.0.3
+      uses: codecov/codecov-action@v2.1.0
       with:
         file: ./coverage.xml
         fail_ci_if_error: false

.pre-commit-config.yaml

@@ -6,7 +6,7 @@ repos:
       - id: black
         language_version: python3
     repo: https://github.com/ambv/black
-    rev: 21.8b0
+    rev: 21.9b0
   - hooks:
       - id: codespell
         args: 

LICENSE

@@ -186,7 +186,7 @@
       same "printed page" as the copyright notice for easier
       identification within third-party archives.
 
-   Copyright {yyyy} {name of copyright owner}
+   Copyright 2017-2021 Iterative, Inc.
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

MANIFEST.in

@@ -1,3 +1,2 @@
 include fastentrypoints.py
 include LICENSE
-include test_requirements.txt

dvc/_debug.py

@@ -1,10 +1,39 @@
 from contextlib import ExitStack, contextmanager
-from typing import TYPE_CHECKING
+from typing import TYPE_CHECKING, Callable, Union
 
 if TYPE_CHECKING:
     from argparse import Namespace
 
 
+@contextmanager
+def yappi_profile(
+    path: Union[Callable[[], str], str] = None, wall_clock: bool = True
+):
+    try:
+        import yappi  # pylint: disable=import-error
+    except ImportError:
+        print("Failed to run profiler, yappi is not installed")
+        yield
+        return
+
+    yappi.set_clock_type("wall" if wall_clock else "cpu")
+
+    yappi.start()
+    yield
+    yappi.stop()
+
+    # pylint:disable=no-member
+    if path:
+        stats = yappi.get_func_stats()
+        fpath = path() if callable(path) else path
+        stats.save(fpath, "callgrind")
+    else:
+        yappi.get_func_stats().print_all()
+        yappi.get_thread_stats().print_all()
+
+    yappi.clear_stats()
+
+
 @contextmanager
 def instrument(html_output=False):
     """Run a statistical profiler"""
@@ -70,6 +99,13 @@ def debugtools(args: "Namespace" = None, **kwargs):
             stack.enter_context(profile(kw.get("cprofile_dump")))
         if kw.get("instrument") or kw.get("instrument_open"):
             stack.enter_context(instrument(kw.get("instrument_open", False)))
+        if kw.get("yappi"):
+            from datetime import datetime
+
+            output = "callgrind.dvc-{0:%Y%m%d}_{0:%H%M%S}.out"
+            stack.enter_context(
+                yappi_profile(path=lambda: output.format(datetime.now()))
+            )
         yield
 
 
@@ -79,6 +115,9 @@ def add_debugging_flags(parser):
     parser.add_argument(
         "--cprofile", action="store_true", default=False, help=SUPPRESS
     )
+    parser.add_argument(
+        "--yappi", action="store_true", default=False, help=SUPPRESS
+    )
     parser.add_argument("--cprofile-dump", help=SUPPRESS)
     parser.add_argument(
         "--pdb", action="store_true", default=False, help=SUPPRESS

dvc/command/config.py

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

dvc/command/dag.py

@@ -103,9 +103,8 @@ class CmdDAG(CmdBase):
         if self.args.dot:
             ui.write(_show_dot(G))
         else:
-            from dvc.utils.pager import pager
-
-            pager(_show_ascii(G))
+            with ui.pager():
+                ui.write(_show_ascii(G))
 
         return 0
 

dvc/command/data_sync.py

@@ -393,6 +393,7 @@ def add_parser(subparsers, _parent_parser):
         help="Show status of all stages in the specified directory.",
     )
     status_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,

dvc/command/diff.py

@@ -1,5 +1,4 @@
 import argparse
-import json
 import logging
 import os
 
@@ -16,7 +15,7 @@ def _digest(checksum):
     return "{}..{}".format(checksum["old"][0:8], checksum["new"][0:8])
 
 
-def _show_md(diff, show_hash=False, hide_missing=False):
+def _show_markdown(diff, show_hash=False, hide_missing=False):
     headers = ["Status", "Hash", "Path"] if show_hash else ["Status", "Path"]
     rows = []
     statuses = ["added", "deleted", "renamed", "modified"]
@@ -142,10 +141,10 @@ class CmdDiff(CmdBase):
                         del entry["hash"]
                 diff[key] = entries
 
-            if self.args.show_json:
-                ui.write(json.dumps(diff))
-            elif self.args.show_md:
-                _show_md(diff, show_hash, hide_missing)
+            if self.args.json:
+                ui.write_json(diff)
+            elif self.args.markdown:
+                _show_markdown(diff, show_hash, hide_missing)
             elif diff:
                 self._show_diff(diff, hide_missing)
 
@@ -188,6 +187,7 @@ def add_parser(subparsers, parent_parser):
         nargs="?",
     )
     diff_parser.add_argument(
+        "--json",
         "--show-json",
         help="Format the output into a JSON",
         action="store_true",
@@ -200,9 +200,11 @@ def add_parser(subparsers, parent_parser):
         default=False,
     )
     diff_parser.add_argument(
+        "--md",
         "--show-md",
         help="Show tabulated output in the Markdown format (GFM).",
         action="store_true",
+        dest="markdown",
         default=False,
     )
     diff_parser.add_argument(

dvc/command/experiments.py

@@ -5,9 +5,8 @@ from datetime import date, datetime
 from fnmatch import fnmatch
 from typing import TYPE_CHECKING, Dict, Iterable, Optional
 
-from funcy import lmap
+from funcy import compact, lmap
 
-from dvc import prompt
 from dvc.command import completion
 from dvc.command.base import CmdBase, append_doc_link, fix_subparsers
 from dvc.command.metrics import DEFAULT_PRECISION
@@ -19,9 +18,8 @@ from dvc.utils.flatten import flatten
 from dvc.utils.serialize import encode_exception
 
 if TYPE_CHECKING:
-    from rich.text import Text
-
     from dvc.compare import TabularData
+    from dvc.ui import RichText
 
 
 logger = logging.getLogger(__name__)
@@ -260,11 +258,12 @@ def _sort_column(sort_by, metric_names, param_names):
 def _sort_exp(experiments, sort_path, sort_name, typ, reverse):
     def _sort(item):
         rev, exp = item
-        tip = exp.get("checkpoint_tip")
+        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.get(typ, {}).get(sort_path, {})
+        data = exp_data.get(typ, {}).get(sort_path, {}).get("data", {})
         val = flatten(data).get(sort_name)
         return val is None, val
 
@@ -291,8 +290,6 @@ def _format_time(datetime_obj, fill_value=FILL_VALUE, iso=False):
 
 
 def _extend_row(row, names, items, precision, fill_value=FILL_VALUE):
-    from rich.text import Text
-
     from dvc.compare import _format_field, with_value
 
     if not items:
@@ -307,10 +304,10 @@ def _extend_row(row, names, items, precision, fill_value=FILL_VALUE):
                 item.get(name),
                 FILL_VALUE_ERRORED if data.get("error", None) else fill_value,
             )
-            # wrap field data in rich.Text, otherwise rich may
+            # wrap field data in ui.rich_text, otherwise rich may
             # interpret unescaped braces from list/dict types as rich
             # markup tags
-            row.append(Text(str(_format_field(value, precision))))
+            row.append(ui.rich_text(str(_format_field(value, precision))))
 
 
 def _parse_filter_list(param_list):
@@ -361,17 +358,15 @@ def experiments_table(
     return td
 
 
-def prepare_exp_id(kwargs) -> "Text":
-    from rich.text import Text
-
+def prepare_exp_id(kwargs) -> "RichText":
     exp_name = kwargs["Experiment"]
     rev = kwargs["rev"]
     typ = kwargs.get("typ", "baseline")
 
     if typ == "baseline" or not exp_name:
-        text = Text(exp_name or rev)
+        text = ui.rich_text(exp_name or rev)
     else:
-        text = Text.assemble(rev, " [", (exp_name, "bold"), "]")
+        text = ui.rich_text.assemble(rev, " [", (exp_name, "bold"), "]")
 
     parent = kwargs.get("parent")
     suff = f" ({parent})" if parent else ""
@@ -379,7 +374,7 @@ def prepare_exp_id(kwargs) -> "Text":
 
     tree = experiment_types[typ]
     pref = f"{tree} " if tree else ""
-    return Text(pref) + text
+    return ui.rich_text(pref) + text
 
 
 def baseline_styler(typ):
@@ -387,7 +382,7 @@ def baseline_styler(typ):
 
 
 def show_experiments(
-    all_experiments, pager=True, no_timestamp=False, show_csv=False, **kwargs
+    all_experiments, pager=True, no_timestamp=False, csv=False, **kwargs
 ):
     from funcy.seqs import flatten as flatten_list
 
@@ -443,7 +438,7 @@ def show_experiments(
 
     row_styles = lmap(baseline_styler, td.column("typ"))
 
-    if not show_csv:
+    if not csv:
         merge_headers = ["Experiment", "rev", "typ", "parent"]
         td.column("Experiment")[:] = map(
             prepare_exp_id, td.as_dict(merge_headers)
@@ -477,7 +472,7 @@ def show_experiments(
         rich_table=True,
         header_styles=styles,
         row_styles=row_styles,
-        show_csv=show_csv,
+        csv=csv,
     )
 
 
@@ -510,18 +505,16 @@ class CmdExperimentsShow(CmdBase):
             logger.exception("failed to show experiments")
             return 1
 
-        if self.args.show_json:
-            import json
-
-            ui.write(json.dumps(all_experiments, default=_format_json))
+        if self.args.json:
+            ui.write_json(all_experiments, default=_format_json)
         else:
             precision = (
                 self.args.precision or None
-                if self.args.show_csv
+                if self.args.csv
                 else DEFAULT_PRECISION
             )
-            fill_value = "" if self.args.show_csv else FILL_VALUE
-            iso = True if self.args.show_csv else False
+            fill_value = "" if self.args.csv else FILL_VALUE
+            iso = True if self.args.csv else False
 
             show_experiments(
                 all_experiments,
@@ -536,7 +529,7 @@ class CmdExperimentsShow(CmdBase):
                 fill_value=fill_value,
                 iso=iso,
                 pager=not self.args.no_pager,
-                show_csv=self.args.show_csv,
+                csv=self.args.csv,
             )
         return 0
 
@@ -565,10 +558,8 @@ class CmdExperimentsDiff(CmdBase):
             logger.exception("failed to show experiments diff")
             return 1
 
-        if self.args.show_json:
-            import json
-
-            ui.write(json.dumps(diff))
+        if self.args.json:
+            ui.write_json(diff)
         else:
             from dvc.compare import show_diff
 
@@ -583,7 +574,7 @@ class CmdExperimentsDiff(CmdBase):
                 show_diff(
                     diff[key],
                     title=title,
-                    markdown=self.args.show_md,
+                    markdown=self.args.markdown,
                     no_path=self.args.no_path,
                     old=self.args.old,
                     on_empty_diff="diff not supported",
@@ -668,7 +659,7 @@ class CmdExperimentsGC(CmdRepro):
         logger.warning(msg)
 
         msg = "Are you sure you want to proceed?"
-        if not self.args.force and not prompt.confirm(msg):
+        if not self.args.force and not ui.confirm(msg):
             return 1
 
         removed = self.repo.experiments.gc(
@@ -789,6 +780,77 @@ class CmdExperimentsRemove(CmdBase):
         return 0
 
 
+class CmdExperimentsInit(CmdBase):
+    CODE = "src"
+    DATA = "data"
+    MODELS = "models"
+    DEFAULT_METRICS = "metrics.json"
+    DEFAULT_PARAMS = "params.yaml"
+    PLOTS = "plots"
+    DVCLIVE = "dvclive"
+    DEFAULTS = {
+        "code": CODE,
+        "data": DATA,
+        "models": MODELS,
+        "metrics": DEFAULT_METRICS,
+        "params": DEFAULT_PARAMS,
+        "plots": PLOTS,
+        "live": DVCLIVE,
+    }
+
+    def run(self):
+        from dvc.command.stage import parse_cmd
+
+        cmd = parse_cmd(self.args.cmd)
+        if not self.args.interactive and not cmd:
+            raise InvalidArgumentError("command is not specified")
+
+        from dvc.repo.experiments.init import init
+
+        defaults = {}
+        if not self.args.explicit:
+            config = self.repo.config["exp"]
+            defaults.update({**self.DEFAULTS, **config})
+
+        cli_args = compact(
+            {
+                "cmd": cmd,
+                "code": self.args.code,
+                "data": self.args.data,
+                "models": self.args.models,
+                "metrics": self.args.metrics,
+                "params": self.args.params,
+                "plots": self.args.plots,
+                "live": self.args.live,
+            }
+        )
+
+        initialized_stage = init(
+            self.repo,
+            name=self.args.name,
+            type=self.args.type,
+            defaults=defaults,
+            overrides=cli_args,
+            interactive=self.args.interactive,
+            force=self.args.force,
+        )
+        if self.args.run:
+            return self.repo.experiments.run(
+                targets=[initialized_stage.addressing]
+            )
+        return 0
+
+
+class RawDefaultsHelpFormatter(
+    argparse.RawDescriptionHelpFormatter,
+    argparse.ArgumentDefaultsHelpFormatter,
+):
+    def _get_help_string(self, action: argparse.Action) -> Optional[str]:
+        if action.default:
+            return super()._get_help_string(action)
+        return action.help
+
+
 def add_parser(subparsers, parent_parser):
     EXPERIMENTS_HELP = "Commands to run and compare experiments."
 
@@ -911,12 +973,14 @@ def add_parser(subparsers, parent_parser):
         help="Always show git commit SHAs instead of branch/tag names.",
     )
     experiments_show_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,
         help="Print output in JSON format instead of a human-readable table.",
     )
     experiments_show_parser.add_argument(
+        "--csv",
         "--show-csv",
         action="store_true",
         default=False,
@@ -985,15 +1049,18 @@ def add_parser(subparsers, parent_parser):
         help="Show only params that are stage dependencies.",
     )
     experiments_diff_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,
         help="Show output in JSON format.",
     )
     experiments_diff_parser.add_argument(
+        "--md",
         "--show-md",
         action="store_true",
         default=False,
+        dest="markdown",
         help="Show tabulated output in the Markdown format (GFM).",
     )
     experiments_diff_parser.add_argument(
@@ -1151,9 +1218,9 @@ def add_parser(subparsers, parent_parser):
         nargs="?",
         default=None,
         help=(
-            "Optional Git remote name or Git URL. If provided, experiments "
-            "from the specified Git repository will be listed instead of "
-            "local experiments."
+            "Optional Git remote name or Git URL. "
+            "If provided, experiments from the specified Git repository "
+            " will be listed instead of local ones."
         ),
         metavar="[<git_remote>]",
     )
@@ -1303,6 +1370,92 @@ def add_parser(subparsers, parent_parser):
     )
     experiments_remove_parser.set_defaults(func=CmdExperimentsRemove)
 
+    EXPERIMENTS_INIT_HELP = "Initialize experiments."
+    experiments_init_parser = experiments_subparsers.add_parser(
+        "init",
+        parents=[parent_parser],
+        description=append_doc_link(EXPERIMENTS_INIT_HELP, "exp/init"),
+        formatter_class=RawDefaultsHelpFormatter,
+    )
+    experiments_init_parser.add_argument(
+        "cmd",
+        nargs=argparse.REMAINDER,
+        help="Command to execute.",
+        metavar="command",
+    )
+    experiments_init_parser.add_argument(
+        "--run",
+        action="store_true",
+        help="Run the experiment after initializing it",
+    )
+    experiments_init_parser.add_argument(
+        "--interactive",
+        "-i",
+        action="store_true",
+        help="Prompt for values that are not provided",
+    )
+    experiments_init_parser.add_argument(
+        "-f",
+        "--force",
+        action="store_true",
+        default=False,
+        help="Overwrite existing stage",
+    )
+
+    experiments_init_parser.add_argument(
+        "--explicit",
+        action="store_true",
+        default=False,
+        help="Only use the path values explicitly provided",
+    )
+    experiments_init_parser.add_argument(
+        "--name", "-n", help="Name of the stage to create"
+    )
+    experiments_init_parser.add_argument(
+        "--code",
+        default=CmdExperimentsInit.CODE,
+        help="Path to the source file or directory "
+        "which your experiments depend",
+    )
+    experiments_init_parser.add_argument(
+        "--data",
+        default=CmdExperimentsInit.DATA,
+        help="Path to the data file or directory "
+        "which your experiments depend",
+    )
+    experiments_init_parser.add_argument(
+        "--models",
+        default=CmdExperimentsInit.MODELS,
+        help="Path to the model file or directory for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--params",
+        default=CmdExperimentsInit.DEFAULT_PARAMS,
+        help="Path to the parameters file for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--metrics",
+        default=CmdExperimentsInit.DEFAULT_METRICS,
+        help="Path to the metrics file for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--plots",
+        default=CmdExperimentsInit.PLOTS,
+        help="Path to the plots file or directory for your experiments",
+    )
+    experiments_init_parser.add_argument(
+        "--live",
+        help="Path to log dvclive outputs for your experiments "
+        f"(default: {CmdExperimentsInit.DVCLIVE})",
+    )
+    experiments_init_parser.add_argument(
+        "--type",
+        choices=["default", "live"],
+        default="default",
+        help="Select type of stage to create",
+    )
+    experiments_init_parser.set_defaults(func=CmdExperimentsInit)
+
 
 def _add_run_common(parser):
     """Add common args for 'exp run' and 'exp resume'."""

dvc/command/ls/__init__.py

@@ -34,10 +34,8 @@ class CmdList(CmdBaseNoRepo):
                 recursive=self.args.recursive,
                 dvc_only=self.args.dvc_only,
             )
-            if self.args.show_json:
-                import json
-
-                ui.write(json.dumps(entries))
+            if self.args.json:
+                ui.write_json(entries)
             elif entries:
                 entries = _prettify(entries, with_color=True)
                 ui.write("\n".join(entries))
@@ -54,6 +52,7 @@ def add_parser(subparsers, parent_parser):
     )
     list_parser = subparsers.add_parser(
         "list",
+        aliases=["ls"],
         parents=[parent_parser],
         description=append_doc_link(LIST_HELP, "list"),
         help=LIST_HELP,
@@ -70,7 +69,10 @@ def add_parser(subparsers, parent_parser):
         "--dvc-only", action="store_true", help="Show only DVC outputs."
     )
     list_parser.add_argument(
-        "--show-json", action="store_true", help="Show output in JSON format."
+        "--json",
+        "--show-json",
+        action="store_true",
+        help="Show output in JSON format.",
     )
     list_parser.add_argument(
         "--rev",

dvc/command/machine.py

@@ -69,6 +69,22 @@ class CmdMachineRemove(CmdMachineConfig):
         return 0
 
 
+class CmdMachineList(CmdMachineConfig):
+    def run(self):
+        levels = [self.args.level] if self.args.level else self.config.LEVELS
+        for level in levels:
+            conf = self.config.read(level)["machine"]
+            if self.args.name:
+                conf = conf.get(self.args.name, {})
+            prefix = self._config_file_prefix(
+                self.args.show_origin, self.config, level
+            )
+            configs = list(self._format_config(conf, prefix))
+            if configs:
+                ui.write("\n".join(configs))
+        return 0
+
+
 class CmdMachineModify(CmdMachineConfig):
     def run(self):
         from dvc.config import merge
@@ -219,6 +235,27 @@ def add_parser(subparsers, parent_parser):
     )
     machine_default_parser.set_defaults(func=CmdMachineDefault)
 
+    machine_LIST_HELP = "List the configuration of one/all machines."
+    machine_list_parser = machine_subparsers.add_parser(
+        "list",
+        parents=[parent_config_parser, parent_parser],
+        description=append_doc_link(machine_LIST_HELP, "machine/list"),
+        help=machine_LIST_HELP,
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+    )
+    machine_list_parser.add_argument(
+        "--show-origin",
+        default=False,
+        action="store_true",
+        help="Show the source file containing each config value.",
+    )
+    machine_list_parser.add_argument(
+        "name",
+        nargs="?",
+        type=str,
+        help="name of machine to specify",
+    )
+    machine_list_parser.set_defaults(func=CmdMachineList)
     machine_MODIFY_HELP = "Modify the configuration of an machine."
     machine_modify_parser = machine_subparsers.add_parser(
         "modify",

dvc/command/metrics.py

@@ -31,16 +31,14 @@ class CmdMetricsShow(CmdMetricsBase):
             logger.exception("")
             return 1
 
-        if self.args.show_json:
-            import json
-
-            ui.write(json.dumps(metrics, default=encode_exception))
+        if self.args.json:
+            ui.write_json(metrics, default=encode_exception)
         else:
             from dvc.compare import show_metrics
 
             show_metrics(
                 metrics,
-                markdown=self.args.show_md,
+                markdown=self.args.markdown,
                 all_branches=self.args.all_branches,
                 all_tags=self.args.all_tags,
                 all_commits=self.args.all_commits,
@@ -65,17 +63,15 @@ class CmdMetricsDiff(CmdMetricsBase):
             logger.exception("failed to show metrics diff")
             return 1
 
-        if self.args.show_json:
-            import json
-
-            ui.write(json.dumps(diff))
+        if self.args.json:
+            ui.write_json(diff)
         else:
             from dvc.compare import show_diff
 
             show_diff(
                 diff,
                 title="Metric",
-                markdown=self.args.show_md,
+                markdown=self.args.markdown,
                 no_path=self.args.no_path,
                 precision=self.args.precision or DEFAULT_PRECISION,
                 round_digits=True,
@@ -140,15 +136,18 @@ def add_parser(subparsers, parent_parser):
         help="Show metrics for all commits.",
     )
     metrics_show_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,
         help="Show output in JSON format.",
     )
     metrics_show_parser.add_argument(
+        "--md",
         "--show-md",
         action="store_true",
         default=False,
+        dest="markdown",
         help="Show tabulated output in the Markdown format (GFM).",
     )
     metrics_show_parser.add_argument(
@@ -220,15 +219,18 @@ def add_parser(subparsers, parent_parser):
         help="Show unchanged metrics as well.",
     )
     metrics_diff_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,
         help="Show output in JSON format.",
     )
     metrics_diff_parser.add_argument(
+        "--md",
         "--show-md",
         action="store_true",
         default=False,
+        dest="markdown",
         help="Show tabulated output in the Markdown format (GFM).",
     )
     metrics_diff_parser.add_argument(

dvc/command/params.py

@@ -25,17 +25,15 @@ class CmdParamsDiff(CmdBase):
             logger.exception("failed to show params diff")
             return 1
 
-        if self.args.show_json:
-            import json
-
-            ui.write(json.dumps(diff))
+        if self.args.json:
+            ui.write_json(diff)
         else:
             from dvc.compare import show_diff
 
             show_diff(
                 diff,
                 title="Param",
-                markdown=self.args.show_md,
+                markdown=self.args.markdown,
                 no_path=self.args.no_path,
                 show_changes=False,
             )
@@ -103,15 +101,18 @@ def add_parser(subparsers, parent_parser):
         help="Show only params that are stage dependencies.",
     )
     params_diff_parser.add_argument(
+        "--json",
         "--show-json",
         action="store_true",
         default=False,
         help="Show output in JSON format.",
     )
     params_diff_parser.add_argument(
+        "--md",
         "--show-md",
         action="store_true",
         default=False,
+        dest="markdown",
         help="Show tabulated output in the Markdown format (GFM).",
     )
     params_diff_parser.add_argument(

dvc/command/remote.py

@@ -86,7 +86,7 @@ class CmdRemoteDefault(CmdRemote):
         if self.args.name is None and not self.args.unset:
             conf = self.config.read(self.args.level)
             try:
-                print(conf["core"]["remote"])
+                ui.write(conf["core"]["remote"])
             except KeyError:
                 ui.write("No default remote set")
                 return 1

dvc/command/status.py

@@ -69,10 +69,8 @@ class CmdDataStatus(CmdDataBase):
             if self.args.quiet:
                 return bool(st)
 
-            if self.args.show_json:
-                import json
-
-                ui.write(json.dumps(st))
+            if self.args.json:
+                ui.write_json(st)
                 return 0
 
             if st:

dvc/compare.py

@@ -168,7 +168,7 @@ class TabularData(MutableSequence[Sequence["CellT"]]):
     def render(self, **kwargs: Any):
         from dvc.ui import ui
 
-        if kwargs.pop("show_csv", False):
+        if kwargs.pop("csv", False):
             ui.write(self.to_csv(), end="")
         else:
             ui.table(self, headers=self.keys(), **kwargs)

dvc/config.py

@@ -76,8 +76,10 @@ class Config(dict):
     APPNAME = "dvc"
     APPAUTHOR = "iterative"
 
+    SYSTEM_LEVELS = ("system", "global")
+    REPO_LEVELS = ("repo", "local")
     # In the order they shadow each other
-    LEVELS = ("system", "global", "repo", "local")
+    LEVELS = SYSTEM_LEVELS + REPO_LEVELS
 
     CONFIG = "config"
     CONFIG_LOCAL = "config.local"
@@ -286,7 +288,7 @@ class Config(dict):
     def edit(self, level=None, validate=True):
         # NOTE: we write to repo config by default, same as git config
         level = level or "repo"
-        if level in {"repo", "local"} and self.dvc_dir is None:
+        if self.dvc_dir is None and level in self.REPO_LEVELS:
             raise ConfigError("Not inside a DVC repo")
 
         conf = self.load_one(level)

dvc/config_schema.py

@@ -156,6 +156,8 @@ SCHEMA = {
                     "grant_write_acp": str,
                     "grant_full_control": str,
                     "cache_regions": bool,
+                    "read_timeout": Coerce(int),
+                    "connect_timeout": Coerce(int),
                     **REMOTE_COMMON,
                 },
                 "gs": {
@@ -223,13 +225,27 @@ SCHEMA = {
         )
     },
     "state": {
+        "dir": str,
         "row_limit": All(Coerce(int), Range(1)),  # obsoleted
         "row_cleanup_quota": All(Coerce(int), Range(0, 100)),  # obsoleted
     },
+    "index": {
+        "dir": str,
+    },
     "machine": {
         str: {
             "cloud": All(Lower, Choices("aws", "azure")),
+            "region": All(
+                Lower, Choices("us-west", "us-east", "eu-west", "eu-north")
+            ),
+            "image": str,
+            "name": str,
+            "spot": Bool,
+            "spot_price": Coerce(float),
+            "instance_hdd_size": Coerce(int),
             "instance_type": Lower,
+            "instance_gpu": Lower,
+            "ssh_private": str,
             "startup_script": str,
         },
     },
@@ -240,4 +256,13 @@ SCHEMA = {
         Optional("parametrization", default=True): Bool,
     },
     "plots": {"html_template": str},
+    "exp": {
+        "code": str,
+        "data": str,
+        "models": str,
+        "metrics": str,
+        "params": str,
+        "plots": str,
+        "live": str,
+    },
 }

dvc/data_cloud.py

@@ -59,7 +59,7 @@ class DataCloud:
         from dvc.objects.db import get_odb
 
         cls, config, path_info = get_cloud_fs(self.repo, name=name)
-        config["tmp_dir"] = self.repo.tmp_dir
+        config["tmp_dir"] = self.repo.index_db_dir
         return get_odb(cls(**config), path_info, **config)
 
     def push(

dvc/dependency/repo.py

@@ -63,7 +63,7 @@ class RepoDependency(Dependency):
         return {self.PARAM_PATH: self.def_path, self.PARAM_REPO: self.def_repo}
 
     def download(self, to, jobs=None):
-        from dvc.checkout import checkout
+        from dvc.objects.checkout import checkout
 
         for odb, objs in self.get_used_objs().items():
             self.repo.cloud.pull(objs, jobs=jobs, odb=odb)

dvc/dvcfile.py

@@ -1,10 +1,7 @@
-import collections
 import contextlib
 import logging
 import os
-from typing import TYPE_CHECKING, Any, Union
-
-from voluptuous import MultipleInvalid
+from typing import TYPE_CHECKING, Any, Callable, Tuple, TypeVar, Union
 
 from dvc.exceptions import DvcException
 from dvc.parsing.versions import LOCKFILE_VERSION, SCHEMA_KWD
@@ -12,23 +9,18 @@ from dvc.stage import serialize
 from dvc.stage.exceptions import (
     StageFileBadNameError,
     StageFileDoesNotExistError,
-    StageFileFormatError,
     StageFileIsNotDvcFileError,
 )
 from dvc.types import AnyPath
 from dvc.utils import relpath
 from dvc.utils.collections import apply_diff
-from dvc.utils.serialize import (
-    dump_yaml,
-    modify_yaml,
-    parse_yaml,
-    parse_yaml_for_update,
-)
+from dvc.utils.serialize import dump_yaml, modify_yaml
 
 if TYPE_CHECKING:
     from dvc.repo import Repo
 
 logger = logging.getLogger(__name__)
+_T = TypeVar("_T")
 
 DVC_FILE = "Dvcfile"
 DVC_FILE_SUFFIX = ".dvc"
@@ -45,10 +37,6 @@ class FileIsGitIgnored(DvcException):
         )
 
 
-class LockfileCorruptedError(DvcException):
-    pass
-
-
 class ParametrizedDumpError(DvcException):
     pass
 
@@ -96,7 +84,7 @@ def check_dvcfile_path(repo, path):
 
 
 class FileMixin:
-    SCHEMA = None
+    SCHEMA: Callable[[_T], _T]
 
     def __init__(self, repo, path, verify=True, **kwargs):
         self.repo = repo
@@ -138,7 +126,11 @@ class FileMixin:
         if self._is_git_ignored():
             raise FileIsGitIgnored(self.path)
 
-    def _load(self):
+    def load(self, **kwargs: Any) -> Any:
+        d, _ = self._load(**kwargs)
+        return d
+
+    def _load(self, **kwargs: Any) -> Tuple[Any, str]:
         # it raises the proper exceptions by priority:
         # 1. when the file doesn't exists
         # 2. filename is not a DVC file
@@ -155,18 +147,23 @@ class FileMixin:
             raise StageFileIsNotDvcFileError(self.path)
 
         self._check_gitignored()
-        with self.repo.fs.open(self.path, encoding="utf-8") as fd:
-            stage_text = fd.read()
-        d = parse_yaml(stage_text, self.path)
-        return self.validate(d, self.relpath), stage_text
+        return self._load_yaml(**kwargs)
 
     @classmethod
-    def validate(cls, d, fname=None):
-        assert isinstance(cls.SCHEMA, collections.abc.Callable)
-        try:
-            return cls.SCHEMA(d)  # pylint: disable=not-callable
-        except MultipleInvalid as exc:
-            raise StageFileFormatError(f"'{fname}' format error: {exc}")
+    def validate(cls, d: _T, fname: str = None) -> _T:
+        from dvc.utils.strictyaml import validate
+
+        return validate(d, cls.SCHEMA, path=fname)  # type: ignore[arg-type]
+
+    def _load_yaml(self, **kwargs: Any) -> Tuple[Any, str]:
+        from dvc.utils import strictyaml
+
+        return strictyaml.load(
+            self.path,
+            self.SCHEMA,  # type: ignore[arg-type]
+            self.repo.fs,
+            **kwargs,
+        )
 
     def remove(self, force=False):  # pylint: disable=unused-argument
         with contextlib.suppress(FileNotFoundError):
@@ -298,10 +295,7 @@ class PipelineFile(FileMixin):
         if not self.exists():
             return
 
-        with open(self.path, encoding="utf-8") as f:
-            d = parse_yaml_for_update(f.read(), self.path)
-
-        self.validate(d, self.path)
+        d, _ = self._load_yaml(round_trip=True)
         if stage.name not in d.get("stages", {}):
             return
 
@@ -343,34 +337,25 @@ def migrate_lock_v1_to_v2(d, version_info):
     d["stages"] = stages
 
 
+def lockfile_schema(data: _T) -> _T:
+    schema = get_lockfile_schema(data)
+    return schema(data)
+
+
 class Lockfile(FileMixin):
-    @classmethod
-    def validate(cls, d, fname=None):
-        schema = get_lockfile_schema(d)
-        try:
-            return schema(d)
-        except MultipleInvalid as exc:
-            raise StageFileFormatError(f"'{fname}' format error: {exc}")
+    SCHEMA = staticmethod(lockfile_schema)  # type: ignore[assignment]
 
     def _verify_filename(self):
         pass  # lockfile path is hardcoded, so no need to verify here
 
-    def _load(self):
+    def _load(self, **kwargs: Any):
         try:
-            return super()._load()
+            return super()._load(**kwargs)
         except StageFileDoesNotExistError:
             # we still need to account for git-ignored dvc.lock file
             # even though it may not exist or have been .dvcignored
             self._check_gitignored()
             return {}, ""
-        except StageFileFormatError as exc:
-            raise LockfileCorruptedError(
-                f"Lockfile '{self.relpath}' is corrupted."
-            ) from exc
-
-    def load(self):
-        d, _ = self._load()
-        return d
 
     @property
     def latest_version_info(self):
@@ -409,10 +394,7 @@ class Lockfile(FileMixin):
         if not self.exists():
             return
 
-        with open(self.path, encoding="utf-8") as f:
-            d = parse_yaml_for_update(f.read(), self.path)
-        self.validate(d, self.path)
-
+        d, _ = self._load_yaml(round_trip=True)
         version = LOCKFILE_VERSION.from_dict(d)
         data = d if version == LOCKFILE_VERSION.V1 else d.get("stages", {})
         if stage.name not in data:

dvc/exceptions.py

@@ -364,3 +364,8 @@ class CircularImportError(DvcException):
             f"'{dep}' contains invalid circular import. "
             f"DVC repo '{a}' already imports from '{b}'."
         )
+
+
+class PrettyDvcException(DvcException):
+    def __pretty_exc__(self, **kwargs):
+        """Print prettier exception message."""

dvc/fs/azure.py

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

dvc/fs/base.py

@@ -1,12 +1,17 @@
+import contextlib
 import logging
+import os
 from concurrent.futures import ThreadPoolExecutor, as_completed
-from functools import partial, partialmethod
+from functools import partialmethod
 from multiprocessing import cpu_count
 from typing import Any, ClassVar, Dict, FrozenSet, Optional
 
+from tqdm.utils import CallbackIOWrapper
+
 from dvc.exceptions import DvcException
 from dvc.path_info import URLInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK, FsspecCallback
+from dvc.ui import ui
 from dvc.utils import tmp_fname
 from dvc.utils.fs import makedirs, move
 
@@ -170,12 +175,6 @@ class BaseFileSystem:
         """
         raise NotImplementedError
 
-    def ls(self, path_info, detail=False):
-        raise RemoteActionNotImplemented("ls", self.scheme)
-
-    def find(self, path_info, detail=False, prefix=None):
-        raise RemoteActionNotImplemented("find", self.scheme)
-
     def is_empty(self, path_info):
         return False
 
@@ -217,43 +216,58 @@ class BaseFileSystem:
             return False
         return hash_.endswith(cls.CHECKSUM_DIR_SUFFIX)
 
-    def upload(self, from_info, to_info, name=None, no_progress_bar=False):
-        if not hasattr(self, "_upload"):
-            raise RemoteActionNotImplemented("upload", self.scheme)
+    def upload(
+        self,
+        from_info,
+        to_info,
+        total=None,
+        desc=None,
+        callback=None,
+        no_progress_bar=False,
+        **pbar_args,
+    ):
+        is_file_obj = hasattr(from_info, "read")
+        method = "upload_fobj" if is_file_obj else "put_file"
+        if not hasattr(self, method):
+            raise RemoteActionNotImplemented(method, self.scheme)
 
         if to_info.scheme != self.scheme:
             raise NotImplementedError
 
-        if from_info.scheme != "local":
-            raise NotImplementedError
+        if not is_file_obj:
+            desc = desc or from_info.name
 
-        logger.debug("Uploading '%s' to '%s'", from_info, to_info)
+        stack = contextlib.ExitStack()
+        if not callback:
+            pbar = ui.progress(
+                desc=desc,
+                disable=no_progress_bar,
+                bytes=True,
+                total=total or -1,
+                **pbar_args,
+            )
+            stack.enter_context(pbar)
+            callback = pbar.as_callback()
+            if total:
+                callback.set_size(total)
 
-        name = name or from_info.name
+        with stack:
+            if is_file_obj:
+                wrapped = CallbackIOWrapper(
+                    callback.relative_update, from_info
+                )
+                # `size` is used to provide hints to the WebdavFileSystem
+                # for legacy servers.
+                # pylint: disable=no-member
+                return self.upload_fobj(wrapped, to_info, size=total)
 
-        self._upload(  # noqa, pylint: disable=no-member
-            from_info.fspath,
-            to_info,
-            name=name,
-            no_progress_bar=no_progress_bar,
-        )
+            if from_info.scheme != "local":
+                raise NotImplementedError
 
-    def upload_fobj(
-        self, fobj, to_info, no_progress_bar=False, size=None, **pbar_args
-    ):
-        if not hasattr(self, "_upload_fobj"):
-            raise RemoteActionNotImplemented("upload_fobj", self.scheme)
-
-        with Tqdm.wrapattr(
-            fobj,
-            "read",
-            disable=no_progress_bar,
-            bytes=True,
-            total=size,
-            **pbar_args,
-        ) as wrapped:
-            self._upload_fobj(  # pylint: disable=no-member
-                wrapped, to_info, size=size
+            logger.debug("Uploading '%s' to '%s'", from_info, to_info)
+            # pylint: disable=no-member
+            return self.put_file(
+                os.fspath(from_info), to_info, callback=callback
             )
 
     def download(
@@ -261,84 +275,105 @@ class BaseFileSystem:
         from_info,
         to_info,
         name=None,
+        callback=None,
         no_progress_bar=False,
         jobs=None,
         _only_file=False,
         **kwargs,
     ):
-        if not hasattr(self, "_download"):
-            raise RemoteActionNotImplemented("download", self.scheme)
+        if not hasattr(self, "get_file"):
+            raise RemoteActionNotImplemented("get_file", self.scheme)
 
         if from_info.scheme != self.scheme:
             raise NotImplementedError
 
-        if to_info.scheme == self.scheme != "local":
-            self.copy(from_info, to_info)
-            return 0
-
         if to_info.scheme != "local":
             raise NotImplementedError
 
-        if not _only_file and self.isdir(from_info):
-            return self._download_dir(
-                from_info, to_info, name, no_progress_bar, jobs, **kwargs
+        download_dir = not _only_file and self.isdir(from_info)
+
+        desc = name or to_info.name
+        stack = contextlib.ExitStack()
+        if not callback:
+            pbar_kwargs = {"unit": "Files"} if download_dir else {}
+            pbar = ui.progress(
+                total=-1,
+                desc="Downloading directory" if download_dir else desc,
+                bytes=not download_dir,
+                disable=no_progress_bar,
+                **pbar_kwargs,
             )
-        return self._download_file(from_info, to_info, name, no_progress_bar)
+            stack.enter_context(pbar)
+            callback = pbar.as_callback()
+
+        with stack:
+            if download_dir:
+                return self._download_dir(
+                    from_info, to_info, callback=callback, jobs=jobs, **kwargs
+                )
+            return self._download_file(from_info, to_info, callback=callback)
 
     download_file = partialmethod(download, _only_file=True)
 
     def _download_dir(
-        self, from_info, to_info, name, no_progress_bar, jobs, **kwargs
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        jobs=None,
+        **kwargs,
     ):
         from_infos = list(self.walk_files(from_info, **kwargs))
         if not from_infos:
             makedirs(to_info, exist_ok=True)
             return None
+
         to_infos = (
             to_info / info.relative_to(from_info) for info in from_infos
         )
+        callback.set_size(len(from_infos))
 
-        with Tqdm(
-            total=len(from_infos),
-            desc="Downloading directory",
-            unit="Files",
-            disable=no_progress_bar,
-        ) as pbar:
-            download_files = pbar.wrap_fn(
-                partial(self._download_file, name=name, no_progress_bar=True)
-            )
-            max_workers = jobs or self.jobs
-            with ThreadPoolExecutor(max_workers=max_workers) as executor:
-                futures = [
-                    executor.submit(download_files, from_info, to_info)
-                    for from_info, to_info in zip(from_infos, to_infos)
-                ]
+        download_files = FsspecCallback.wrap_fn(callback, self._download_file)
+        max_workers = jobs or self.jobs
+        with ThreadPoolExecutor(max_workers=max_workers) as executor:
+            futures = [
+                executor.submit(download_files, from_info, to_info)
+                for from_info, to_info in zip(from_infos, to_infos)
+            ]
 
-                # NOTE: unlike pulling/fetching cache, where we need to
-                # download everything we can, not raising an error here might
-                # turn very ugly, as the user might think that he has
-                # downloaded a complete directory, while having a partial one,
-                # which might cause unexpected results in his pipeline.
-                for future in as_completed(futures):
-                    # NOTE: executor won't let us raise until all futures that
-                    # it has are finished, so we need to cancel them ourselves
-                    # before re-raising.
-                    exc = future.exception()
-                    if exc:
-                        for entry in futures:
-                            entry.cancel()
-                        raise exc
+            # NOTE: unlike pulling/fetching cache, where we need to
+            # download everything we can, not raising an error here might
+            # turn very ugly, as the user might think that he has
+            # downloaded a complete directory, while having a partial one,
+            # which might cause unexpected results in his pipeline.
+            for future in as_completed(futures):
+                # NOTE: executor won't let us raise until all futures that
+                # it has are finished, so we need to cancel them ourselves
+                # before re-raising.
+                exc = future.exception()
+                if exc:
+                    for entry in futures:
+                        entry.cancel()
+                    raise exc
 
-    def _download_file(self, from_info, to_info, name, no_progress_bar):
+    def _download_file(
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+    ):
         makedirs(to_info.parent, exist_ok=True)
-
-        logger.debug("Downloading '%s' to '%s'", from_info, to_info)
-        name = name or to_info.name
-
         tmp_file = tmp_fname(to_info)
 
-        self._download(  # noqa, pylint: disable=no-member
-            from_info, tmp_file, name=name, no_progress_bar=no_progress_bar
-        )
+        logger.debug("Downloading '%s' to '%s'", from_info, to_info)
+        try:
+            # noqa, pylint: disable=no-member
+            self.get_file(from_info, tmp_file, callback=callback)
+        except Exception:  # pylint: disable=broad-except
+            # do we need to rollback makedirs for previously not-existing
+            # directories?
+            with contextlib.suppress(FileNotFoundError):
+                os.unlink(tmp_file)
+            raise
 
         move(tmp_file, to_info)

dvc/fs/dvc.py

@@ -6,6 +6,7 @@ from dvc.exceptions import OutputNotFoundError
 from dvc.path_info import PathInfo
 from dvc.utils import relpath
 
+from ..progress import DEFAULT_CALLBACK
 from ._metadata import Metadata
 from .base import BaseFileSystem
 
@@ -256,10 +257,12 @@ class DvcFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
 
         return ret
 
-    def _download(self, from_info, to_file, **kwargs):
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         fs, path = self._get_fs_path(from_info)
-        fs._download(  # pylint: disable=protected-access
-            path, to_file, **kwargs
+        fs.get_file(  # pylint: disable=protected-access
+            path, to_file, callback=callback, **kwargs
         )
 
     def checksum(self, path_info):

dvc/fs/fsspec_wrapper.py

@@ -1,13 +1,37 @@
 import os
 import shutil
 from functools import lru_cache
+from typing import (
+    IO,
+    TYPE_CHECKING,
+    Any,
+    Dict,
+    Iterator,
+    Optional,
+    Union,
+    overload,
+)
 
 from funcy import cached_property
+from tqdm.utils import CallbackIOWrapper
 
-from dvc.progress import Tqdm
+from dvc.path_info import URLInfo
+from dvc.progress import DEFAULT_CALLBACK
 
 from .base import BaseFileSystem
-from .local import LocalFileSystem
+
+FSPath = URLInfo
+AnyFSPath = Union[FSPath, str]
+
+if TYPE_CHECKING:
+    from typing import Iterable
+
+    from typing_extensions import Literal
+
+
+# An info() entry, might evolve to a TypedDict
+# in the future (e.g for properly type 'size' etc).
+Entry = Dict[str, Any]
 
 
 # pylint: disable=no-member
@@ -19,18 +43,8 @@ class FSSpecWrapper(BaseFileSystem):
         self.fs_args = {"skip_instance_cache": True}
         self.fs_args.update(self._prepare_credentials(**kwargs))
 
-    @cached_property
-    def fs(self):
-        raise NotImplementedError
-
-    def _with_bucket(self, path):
-        return str(path)
-
-    def _strip_bucket(self, entry):
-        return entry
-
     @staticmethod
-    def _get_kwargs_from_urls(urlpath):
+    def _get_kwargs_from_urls(urlpath: str) -> "Dict[str, Any]":
         from fsspec.utils import infer_storage_options
 
         options = infer_storage_options(urlpath)
@@ -38,6 +52,28 @@ class FSSpecWrapper(BaseFileSystem):
         options.pop("protocol", None)
         return options
 
+    @cached_property
+    def fs(self):
+        raise NotImplementedError
+
+    def _with_bucket(self, path: AnyFSPath) -> str:
+        return str(path)
+
+    def _strip_bucket(self, entry: str) -> str:
+        return entry
+
+    @overload
+    def _strip_buckets(
+        self, entries: "Iterable[Entry]", detail: "Literal[True]"
+    ) -> "Iterator[Entry]":
+        ...
+
+    @overload
+    def _strip_buckets(
+        self, entries: "Iterable[str]", detail: "Literal[False]" = False
+    ) -> "Iterator[str]":
+        ...
+
     def _strip_buckets(self, entries, detail=False):
         for entry in entries:
             if detail:
@@ -47,59 +83,91 @@ class FSSpecWrapper(BaseFileSystem):
                 entry = self._strip_bucket(entry)
             yield entry
 
-    def _entry_hook(self, entry):
+    def _entry_hook(self, entry: "Entry") -> "Entry":
         """Simple hook method to be overridden when wanted to process
         entries within info() and ls(detail=True) calls"""
         return entry
 
     def _prepare_credentials(
-        self, **config
-    ):  # pylint: disable=unused-argument
+        self, **config: Dict[str, Any]  # pylint: disable=unused-argument
+    ) -> Dict[str, Any]:
         """Prepare the arguments for authentication to the
         host filesystem"""
         return {}
 
-    def _isdir(self, path_info):
+    def _isdir(self, path_info: FSPath) -> bool:
         return self.fs.isdir(self._with_bucket(path_info))
 
-    def isdir(self, path_info):
+    def isdir(self, path_info: FSPath) -> bool:
         try:
             return self._isdir(path_info)
         except FileNotFoundError:
             return False
 
-    def isfile(self, path_info):
+    def isfile(self, path_info: FSPath) -> bool:
         try:
             return not self._isdir(path_info)
         except FileNotFoundError:
             return False
 
-    def is_empty(self, path_info):
+    def is_empty(self, path_info: FSPath) -> bool:
         entry = self.info(path_info)
         if entry["type"] == "directory":
             return not self.fs.ls(self._with_bucket(path_info))
         return entry["size"] == 0
 
     def open(
-        self, path_info, mode="r", **kwargs
-    ):  # pylint: disable=arguments-differ
+        self,
+        path_info: FSPath,
+        mode: str = "r",
+        encoding: Optional[str] = None,
+        **kwargs,
+    ) -> "IO":  # pylint: disable=arguments-differ
         return self.fs.open(self._with_bucket(path_info), mode=mode)
 
-    def checksum(self, path_info):
+    def checksum(self, path_info: FSPath) -> str:
         return self.fs.checksum(self._with_bucket(path_info))
 
-    def copy(self, from_info, to_info):
+    def copy(self, from_info: FSPath, to_info: FSPath) -> None:
         self.makedirs(to_info.parent)
         self.fs.copy(self._with_bucket(from_info), self._with_bucket(to_info))
 
-    def exists(self, path_info) -> bool:
+    def exists(self, path_info: FSPath) -> bool:
         return self.fs.exists(self._with_bucket(path_info))
 
+    @overload
+    def ls(
+        self, path_info: FSPath, detail: "Literal[True]"
+    ) -> "Iterator[Entry]":
+        ...
+
+    @overload
+    def ls(self, path_info: FSPath, detail: "Literal[False]") -> Iterator[str]:
+        ...
+
     def ls(self, path_info, detail=False):
         path = self._with_bucket(path_info)
         files = self.fs.ls(path, detail=detail)
         yield from self._strip_buckets(files, detail=detail)
 
+    @overload
+    def find(
+        self,
+        path_info: FSPath,
+        detail: "Literal[True]",
+        prefix: Optional[bool],
+    ) -> "Iterator[Entry]":
+        ...
+
+    @overload
+    def find(
+        self,
+        path_info: FSPath,
+        detail: "Literal[False]",
+        prefix: Optional[bool],
+    ) -> Iterator[str]:
+        ...
+
     # pylint: disable=unused-argument
     def find(self, path_info, detail=False, prefix=None):
         path = self._with_bucket(path_info)
@@ -109,63 +177,56 @@ class FSSpecWrapper(BaseFileSystem):
 
         yield from self._strip_buckets(files, detail=detail)
 
-    def walk_files(self, path_info, **kwargs):
+    def walk_files(self, path_info: FSPath, **kwargs) -> Iterator[FSPath]:
         for file in self.find(path_info, **kwargs):
             yield path_info.replace(path=file)
 
-    def move(self, from_info, to_info):
+    def move(self, from_info: FSPath, to_info: FSPath) -> None:
         self.fs.move(self._with_bucket(from_info), self._with_bucket(to_info))
 
-    def remove(self, path_info):
+    def remove(self, path_info: FSPath) -> None:
         self.fs.rm_file(self._with_bucket(path_info))
 
-    def info(self, path_info):
+    def info(self, path_info: FSPath) -> "Entry":
         info = self.fs.info(self._with_bucket(path_info)).copy()
         info = self._entry_hook(info)
         info["name"] = self._strip_bucket(info["name"])
         return info
 
-    def makedirs(self, path_info, **kwargs):
+    def makedirs(self, path_info: FSPath, **kwargs) -> None:
         self.fs.makedirs(
             self._with_bucket(path_info), exist_ok=kwargs.pop("exist_ok", True)
         )
 
-    def _upload_fobj(self, fobj, to_info, size=None):
-        self.makedirs(to_info.parent)
-        with self.open(to_info, "wb") as fdest:
-            shutil.copyfileobj(fobj, fdest, length=fdest.blocksize)
-
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **kwargs
-    ):
-        self.makedirs(to_info.parent)
-        size = os.path.getsize(from_file)
-        with open(from_file, "rb") as fobj:
-            self.upload_fobj(
-                fobj,
-                to_info,
-                size=size,
-                desc=name,
-                no_progress_bar=no_progress_bar,
-            )
+    def put_file(
+        self,
+        from_file: FSPath,
+        to_info: FSPath,
+        callback: Any = DEFAULT_CALLBACK,
+        **kwargs,
+    ) -> None:
+        self.fs.put_file(
+            from_file, self._with_bucket(to_info), callback=callback, **kwargs
+        )
         self.fs.invalidate_cache(self._with_bucket(to_info.parent))
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
-    ):
-        total = self.getsize(self._with_bucket(from_info))
-        with self.open(from_info, "rb") as fobj:
-            with Tqdm.wrapattr(
-                fobj,
-                "read",
-                desc=name,
-                disable=no_progress_bar,
-                bytes=True,
-                total=total,
-                **pbar_args,
-            ) as wrapped:
-                with open(to_file, "wb") as fdest:
-                    shutil.copyfileobj(wrapped, fdest, length=fobj.blocksize)
+    def get_file(
+        self,
+        from_info: FSPath,
+        to_info: FSPath,
+        callback: Any = DEFAULT_CALLBACK,
+        **kwargs,
+    ) -> None:
+        self.fs.get_file(
+            self._with_bucket(from_info), to_info, callback=callback, **kwargs
+        )
+
+    def upload_fobj(self, fobj: IO, to_info: FSPath, **kwargs) -> None:
+        self.makedirs(to_info.parent)
+        with self.open(to_info, "wb") as fdest:
+            shutil.copyfileobj(
+                fobj, fdest, length=getattr(fdest, "blocksize", None)
+            )
 
 
 # pylint: disable=abstract-method
@@ -173,21 +234,21 @@ class ObjectFSWrapper(FSSpecWrapper):
     TRAVERSE_PREFIX_LEN = 3
 
     @lru_cache(512)
-    def _with_bucket(self, path):
+    def _with_bucket(self, path: AnyFSPath) -> str:
         if isinstance(path, self.PATH_CLS):
             return f"{path.bucket}/{path.path}"
-        return path
+        return str(path)
 
-    def _strip_bucket(self, entry):
+    def _strip_bucket(self, entry: str) -> str:
         try:
             bucket, path = entry.split("/", 1)
         except ValueError:
             # If there is no path attached, only returns
             # the bucket (top-level).
-            bucket, path = entry, None
+            bucket, path = entry, ""
         return path or bucket
 
-    def makedirs(self, path_info, **kwargs):
+    def makedirs(self, path_info: FSPath, **kwargs) -> None:
         # For object storages make this method a no-op. The original
         # fs.makedirs() method will only check if the bucket exists
         # and create if it doesn't though we don't want to support
@@ -195,7 +256,7 @@ class ObjectFSWrapper(FSSpecWrapper):
         # simply ignore all mkdir()/makedirs() calls.
         return None
 
-    def _isdir(self, path_info):
+    def _isdir(self, path_info: FSPath) -> bool:
         # Directory in object storages are interpreted differently
         # among different fsspec providers, so this logic is a temporary
         # measure for us to adapt as of now. It checks whether it is a
@@ -230,7 +291,9 @@ class ObjectFSWrapper(FSSpecWrapper):
         if len(files) == 1 and files[0] == path and self.isdir(path_info):
             return None
 
-        yield from self._strip_buckets(files, detail=detail)
+        yield from self._strip_buckets(  # pylint: disable=not-an-iterable
+            files, detail=detail
+        )
 
 
 # pylint: disable=arguments-differ
@@ -254,42 +317,38 @@ class NoDirectoriesMixin:
         raise NotImplementedError
 
 
-_LOCAL_FS = LocalFileSystem()
-
-
 class CallbackMixin:
-    """Use the native ``get_file()``/``put_file()`` APIs
-    if the target filesystem supports callbacks."""
+    """Provides callback support for the filesystem that don't support yet."""
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **pbar_args
+    def put_file(
+        self,
+        from_file,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        **kwargs,
     ):
-        with Tqdm(
-            desc=name,
-            disable=no_progress_bar,
-            bytes=True,
-            total=-1,
-            **pbar_args,
-        ) as pbar:
-            self.fs.put_file(
-                os.fspath(from_file),
-                self._with_bucket(to_info),
-                callback=pbar.as_callback(_LOCAL_FS, from_file),
-            )
-        self.fs.invalidate_cache(self._with_bucket(to_info.parent))
+        """Add compatibility support for Callback."""
+        # pylint: disable=protected-access
+        self.makedirs(to_info.parent)
+        size = os.path.getsize(from_file)
+        with open(from_file, "rb") as fobj:
+            callback.set_size(size)
+            wrapped = CallbackIOWrapper(callback.relative_update, fobj)
+            self.upload_fobj(wrapped, to_info)
+            self.fs.invalidate_cache(self._with_bucket(to_info.parent))
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
+    def get_file(
+        self,
+        from_info,
+        to_info,
+        callback=DEFAULT_CALLBACK,
+        **kwargs,
     ):
-        with Tqdm(
-            desc=name,
-            disable=no_progress_bar,
-            bytes=True,
-            total=-1,
-            **pbar_args,
-        ) as pbar:
-            self.fs.get_file(
-                self._with_bucket(from_info),
-                os.fspath(to_file),
-                callback=pbar.as_callback(self, from_info),
-            )
+        # pylint: disable=protected-access
+        total: int = self.getsize(from_info)
+        if total:
+            callback.set_size(total)
+
+        with self.open(from_info, "rb") as fobj, open(to_info, "wb") as fdest:
+            wrapped = CallbackIOWrapper(callback.relative_update, fobj)
+            shutil.copyfileobj(wrapped, fdest, length=fobj.blocksize)

dvc/fs/gdrive.py

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

dvc/fs/git.py

@@ -1,8 +1,11 @@
 import errno
 import os
 
+from tqdm.utils import CallbackIOWrapper
+
 from dvc.utils import is_exec, relpath
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 
 
@@ -116,16 +119,18 @@ class GitFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
                 # NOTE: os.path.join is ~5.5 times slower
                 yield f"{root}{os.sep}{file}"
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **kwargs
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
         import shutil
 
-        from dvc.progress import Tqdm
+        total = self.getsize(from_info)
+        if total:
+            callback.set_size(total)
 
-        with open(to_file, "wb+") as to_fobj:
-            with Tqdm.wrapattr(
-                to_fobj, "write", desc=name, disable=no_progress_bar
-            ) as wrapped:
-                with self.open(from_info, "rb", **kwargs) as from_fobj:
-                    shutil.copyfileobj(from_fobj, wrapped)
+        with self.open(from_info, "rb", **kwargs) as from_fobj:
+            with open(to_file, "wb+") as to_fobj:
+                wrapped = CallbackIOWrapper(
+                    callback.relative_update, from_fobj
+                )
+                shutil.copyfileobj(wrapped, to_fobj)

dvc/fs/gs.py

@@ -6,17 +6,16 @@ from funcy import cached_property, wrap_prop
 from dvc.path_info import CloudURLInfo
 from dvc.scheme import Schemes
 
-from .fsspec_wrapper import ObjectFSWrapper
-
-
 # pylint:disable=abstract-method
-class GSFileSystem(ObjectFSWrapper):
+from .fsspec_wrapper import CallbackMixin, ObjectFSWrapper
+
+
+class GSFileSystem(CallbackMixin, ObjectFSWrapper):
     scheme = Schemes.GS
     PATH_CLS = CloudURLInfo
     REQUIRES = {"gcsfs": "gcsfs"}
     PARAM_CHECKSUM = "etag"
     DETAIL_FIELDS = frozenset(("etag", "size"))
-    TRAVERSE_PREFIX_LEN = 2
 
     def _prepare_credentials(self, **config):
         login_info = {"consistency": None}

dvc/fs/hdfs.py

@@ -1,285 +1,90 @@
-import errno
-import io
-import logging
 import os
 import re
-import shutil
 import subprocess
-from collections import deque
-from contextlib import closing, contextmanager
+import threading
+
+from funcy import cached_property, wrap_prop
 
-from dvc.hash_info import HashInfo
-from dvc.progress import Tqdm
 from dvc.scheme import Schemes
-from dvc.utils import fix_env, tmp_fname
+from dvc.utils import fix_env
 
-from .base import BaseFileSystem, RemoteCmdError
-from .pool import get_connection
+from .fsspec_wrapper import CallbackMixin, FSSpecWrapper
 
-logger = logging.getLogger(__name__)
+CHECKSUM_REGEX = re.compile(r".*\t.*\t(?P<checksum>.*)")
 
 
-def _hadoop_fs(cmd, user=None):
-    cmd = "hadoop fs -" + cmd
-    if user:
-        cmd = f"HADOOP_USER_NAME={user} " + cmd
-
-    # NOTE: close_fds doesn't work with redirected stdin/stdout/stderr.
-    # See https://github.com/iterative/dvc/issues/1197.
-    close_fds = os.name != "nt"
-
-    executable = os.getenv("SHELL") if os.name != "nt" else None
-    p = subprocess.Popen(
-        cmd,
-        shell=True,
-        close_fds=close_fds,
-        executable=executable,
-        env=fix_env(os.environ),
-        stdin=subprocess.PIPE,
-        stdout=subprocess.PIPE,
-        stderr=subprocess.PIPE,
-    )
-    out, err = p.communicate()
-    if p.returncode != 0:
-        raise RemoteCmdError("hdfs", cmd, p.returncode, err)
-    return out.decode("utf-8")
-
-
-def _group(regex, s, gname):
-    match = re.match(regex, s)
-    assert match is not None
-    return match.group(gname)
-
-
-def _hadoop_fs_checksum(path_info):
-    # NOTE: pyarrow doesn't support checksum, so we need to use hadoop
-    regex = r".*\t.*\t(?P<checksum>.*)"
-    stdout = _hadoop_fs(f"checksum {path_info.url}", user=path_info.user)
-    return _group(regex, stdout, "checksum")
-
-
-class HDFSFileSystem(BaseFileSystem):
+# pylint: disable=abstract-method
+class HDFSFileSystem(CallbackMixin, FSSpecWrapper):
     scheme = Schemes.HDFS
-    REQUIRES = {"pyarrow": "pyarrow"}
-    REGEX = r"^hdfs://((?P<user>.*)@)?.*$"
+    REQUIRES = {"fsspec": "fsspec", "pyarrow": "pyarrow"}
     PARAM_CHECKSUM = "checksum"
-    TRAVERSE_PREFIX_LEN = 2
 
-    BLOCK_SIZE = 2 ** 26  # 128MB
-
-    def __init__(self, **config):
-        super().__init__(**config)
-
-        self.host = config["host"]
-        self.user = config.get("user")
-        self.port = config.get("port")
-
-        self.kerb_ticket = config.get("kerb_ticket")
+    def _with_bucket(self, path):
+        if isinstance(path, self.PATH_CLS):
+            return path.path
+        return super()._with_bucket(path)
 
     @staticmethod
     def _get_kwargs_from_urls(urlpath):
-        from fsspec.implementations.hdfs import PyArrowHDFS
+        from fsspec.implementations.arrow import HadoopFileSystem
 
         # pylint:disable=protected-access
-        return PyArrowHDFS._get_kwargs_from_urls(urlpath)
+        return HadoopFileSystem._get_kwargs_from_urls(urlpath)
 
-    def hdfs(self, path_info):
-        import pyarrow.fs
+    def _prepare_credentials(self, **config):
+        return config
 
-        # NOTE: HadoopFileSystem is not meant to be closed by us and doesn't
-        # have close() method or alternative, so we add a noop one to satisfy
-        # our connection pool.
-        class HDFSConnection(pyarrow.fs.HadoopFileSystem):
-            def close(self):
-                pass
+    @wrap_prop(threading.Lock())
+    @cached_property
+    def fs(self):
+        from fsspec.implementations.arrow import HadoopFileSystem
 
-        return get_connection(
-            HDFSConnection,
-            path_info.host,
-            path_info.port,
-            user=path_info.user,
-            kerb_ticket=self.kerb_ticket,
-        )
-
-    @contextmanager
-    def open(self, path_info, mode="r", encoding=None, **kwargs):
-        assert mode in {"r", "rt", "rb"}
-
-        try:
-            with self.hdfs(path_info) as hdfs, closing(
-                hdfs.open_input_stream(path_info.path)
-            ) as fd:
-                if mode == "rb":
-                    yield fd
-                else:
-                    yield io.TextIOWrapper(fd, encoding=encoding)
-        except OSError as e:
-            # Empty .errno and not specific enough error class in pyarrow,
-            # see https://issues.apache.org/jira/browse/ARROW-6248
-            if "file does not exist" in str(e):
-                raise FileNotFoundError(*e.args)
-            raise
-
-    def exists(self, path_info) -> bool:
-        assert not isinstance(path_info, list)
-        assert path_info.scheme == "hdfs"
-        with self.hdfs(path_info) as hdfs:
-            import pyarrow.fs
-
-            file_info = hdfs.get_file_info(path_info.path)
-            return file_info.type != pyarrow.fs.FileType.NotFound
-
-    def _walk(self, hdfs, root, topdown=True):
-        import posixpath
-
-        from pyarrow.fs import FileSelector, FileType
-
-        dirs = deque()
-        nondirs = deque()
-
-        selector = FileSelector(root)
-        for entry in hdfs.get_file_info(selector):
-            if entry.type == FileType.Directory:
-                dirs.append(entry.base_name)
-            else:
-                nondirs.append(entry.base_name)
-
-        if topdown:
-            yield root, dirs, nondirs
-
-        for dname in dirs:
-            # NOTE: posixpath.join() is slower
-            yield from self._walk(
-                hdfs, f"{root}{posixpath.sep}{dname}", topdown=topdown
-            )
-
-        if not topdown:
-            yield root, dirs, nondirs
-
-    def walk(self, top, topdown=True, onerror=None, **kwargs):
-        if not self.isdir(top):
-            return
-
-        with self.hdfs(top) as hdfs:
-            for root, dnames, fnames in self._walk(
-                hdfs, top.path, topdown=topdown
-            ):
-                yield top.replace(path=root), dnames, fnames
-
-    def walk_files(self, path_info, **kwargs):
-        for root, _, fnames in self.walk(path_info):
-            yield from (root / fname for fname in fnames)
-
-    def remove(self, path_info):
-        if path_info.scheme != "hdfs":
-            raise NotImplementedError
-
-        if self.exists(path_info):
-            logger.debug(f"Removing {path_info.path}")
-            with self.hdfs(path_info) as hdfs:
-                import pyarrow.fs
-
-                file_info = hdfs.get_file_info(path_info.path)
-                if file_info.type == pyarrow.fs.FileType.Directory:
-                    hdfs.delete_dir(path_info.path)
-                else:
-                    hdfs.delete_file(path_info.path)
-
-    def makedirs(self, path_info, **kwargs):
-        with self.hdfs(path_info) as hdfs:
-            # NOTE: fs.create_dir creates parents by default
-            hdfs.create_dir(path_info.path)
-
-    def copy(self, from_info, to_info, **_kwargs):
-        # NOTE: hdfs.copy_file is not supported yet in pyarrow
-        with self.hdfs(to_info) as hdfs:
-            # NOTE: this is how `hadoop fs -cp` works too: it copies through
-            # your local machine.
-            with closing(hdfs.open_input_stream(from_info.path)) as from_fobj:
-                tmp_info = to_info.parent / tmp_fname(to_info.name)
-                try:
-                    with closing(
-                        hdfs.open_output_stream(tmp_info.path)
-                    ) as tmp_fobj:
-                        shutil.copyfileobj(from_fobj, tmp_fobj)
-                    hdfs.move(tmp_info.path, to_info.path)
-                except Exception:
-                    self.remove(tmp_info)
-                    raise
-
-    def isfile(self, path_info):
-        with self.hdfs(path_info) as hdfs:
-            import pyarrow.fs
-
-            file_info = hdfs.get_file_info(path_info.path)
-            return file_info.type == pyarrow.fs.FileType.File
-
-    def isdir(self, path_info):
-        with self.hdfs(path_info) as hdfs:
-            import pyarrow.fs
-
-            file_info = hdfs.get_file_info(path_info.path)
-            return file_info.type == pyarrow.fs.FileType.Directory
-
-    def info(self, path_info):
-        with self.hdfs(path_info) as hdfs:
-            from pyarrow.fs import FileType
-
-            file_info = hdfs.get_file_info(path_info.path)
-            if file_info.type is FileType.Directory:
-                kind = "directory"
-            elif file_info.type is FileType.File:
-                kind = "file"
-            else:
-                raise FileNotFoundError(
-                    errno.ENOENT, os.strerror(errno.ENOENT), path_info
-                )
-
-            return {"size": file_info.size, "type": kind}
+        return HadoopFileSystem(**self.fs_args)
 
     def checksum(self, path_info):
-        return HashInfo(
-            "checksum",
-            _hadoop_fs_checksum(path_info),
-            size=self.getsize(path_info),
+        return self._checksum(path_info)
+
+    def _checksum(self, path_info, **kwargs):
+        # PyArrow doesn't natively support retrieving the
+        # checksum, so we have to use hadoop fs
+
+        result = self._run_command(
+            f"checksum {path_info.url}",
+            env=fix_env(os.environ),
+            user=path_info.user,
         )
+        if result is None:
+            return None
 
-    def _upload_fobj(self, fobj, to_info, **kwargs):
-        with self.hdfs(to_info) as hdfs:
-            with hdfs.open_output_stream(to_info.path) as fdest:
-                shutil.copyfileobj(fobj, fdest, self.BLOCK_SIZE)
+        match = CHECKSUM_REGEX.match(result)
+        if match is None:
+            return None
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
-    ):
-        with self.hdfs(to_info) as hdfs:
-            tmp_file = tmp_fname(to_info.path)
-            total = os.path.getsize(from_file)
-            with open(from_file, "rb") as fobj:
-                with Tqdm.wrapattr(
-                    fobj,
-                    "read",
-                    desc=name,
-                    total=total,
-                    disable=no_progress_bar,
-                ) as wrapped:
-                    with hdfs.open_output_stream(tmp_file) as sobj:
-                        shutil.copyfileobj(wrapped, sobj, self.BLOCK_SIZE)
-            hdfs.move(tmp_file, to_info.path)
+        return match.group("checksum")
 
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **_kwargs
-    ):
-        with self.hdfs(from_info) as hdfs:
-            file_info = hdfs.get_file_info(from_info.path)
-            total = file_info.size
-            with open(to_file, "wb+") as fobj:
-                with Tqdm.wrapattr(
-                    fobj,
-                    "write",
-                    desc=name,
-                    total=total,
-                    disable=no_progress_bar,
-                ) as wrapped:
-                    with hdfs.open_input_stream(from_info.path) as sobj:
-                        shutil.copyfileobj(sobj, wrapped, self.BLOCK_SIZE)
+    def _run_command(self, cmd, env=None, user=None):
+        cmd = "hadoop fs -" + cmd
+        if user:
+            cmd = f"HADOOP_USER_NAME={user} " + cmd
+
+        # NOTE: close_fds doesn't work with redirected stdin/stdout/stderr.
+        # See https://github.com/iterative/dvc/issues/1197.
+        close_fds = os.name != "nt"
+
+        executable = os.getenv("SHELL") if os.name != "nt" else None
+        p = subprocess.Popen(
+            cmd,
+            shell=True,
+            close_fds=close_fds,
+            executable=executable,
+            env=env or os.environ,
+            stdin=subprocess.PIPE,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+        )
+        out, err = p.communicate()
+
+        if p.returncode != 0:
+            raise subprocess.CalledProcessError(p.returncode, cmd, out, err)
+        else:
+            return out.decode("utf-8")

dvc/fs/http.py

@@ -6,7 +6,7 @@ from dvc import prompt
 from dvc.path_info import HTTPURLInfo
 from dvc.scheme import Schemes
 
-from .fsspec_wrapper import CallbackMixin, FSSpecWrapper, NoDirectoriesMixin
+from .fsspec_wrapper import FSSpecWrapper, NoDirectoriesMixin
 
 
 @wrap_with(threading.Lock())
@@ -32,7 +32,7 @@ def make_context(ssl_verify):
 
 
 # pylint: disable=abstract-method
-class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
+class HTTPFileSystem(NoDirectoriesMixin, FSSpecWrapper):
     scheme = Schemes.HTTP
     PATH_CLS = HTTPURLInfo
     PARAM_CHECKSUM = "checksum"
@@ -87,6 +87,9 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
                     ssl=make_context(config["ssl_verify"])
                 )
 
+        # Allow reading proxy configurations from the environment.
+        client_kwargs["trust_env"] = True
+
         credentials["get_client"] = self.get_client
         self.upload_method = config.get("method", "POST")
         return credentials
@@ -99,6 +102,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
             attempts=self.SESSION_RETRIES,
             factor=self.SESSION_BACKOFF_FACTOR,
             max_timeout=self.REQUEST_TIMEOUT,
+            exceptions=[aiohttp.ClientError],
         )
 
         # The default timeout for the aiohttp is 300 seconds

dvc/fs/local.py

@@ -7,6 +7,7 @@ from dvc.system import System
 from dvc.utils import is_exec, tmp_fname
 from dvc.utils.fs import copy_fobj_to_file, copyfile, makedirs, move, remove
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 
 logger = logging.getLogger(__name__)
@@ -108,7 +109,7 @@ class LocalFileSystem(BaseFileSystem):
             self.remove(tmp_info)
             raise
 
-    def _upload_fobj(self, fobj, to_info, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         self.makedirs(to_info.parent)
         tmp_info = to_info.parent / tmp_fname("")
         try:
@@ -158,21 +159,15 @@ class LocalFileSystem(BaseFileSystem):
     def info(self, path_info):
         return self.fs.info(path_info)
 
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         makedirs(to_info.parent, exist_ok=True)
-
         tmp_file = tmp_fname(to_info)
-        copyfile(
-            from_file, tmp_file, name=name, no_progress_bar=no_progress_bar
-        )
+        copyfile(from_file, tmp_file, callback=callback)
         os.replace(tmp_file, to_info)
 
-    @staticmethod
-    def _download(
-        from_info, to_file, name=None, no_progress_bar=False, **_kwargs
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
     ):
-        copyfile(
-            from_info, to_file, no_progress_bar=no_progress_bar, name=name
-        )
+        copyfile(from_info, to_file, callback=callback)

dvc/fs/oss.py

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

dvc/fs/repo.py

@@ -9,6 +9,7 @@ from funcy import lfilter, wrap_with
 
 from dvc.path_info import PathInfo
 
+from ..progress import DEFAULT_CALLBACK
 from .base import BaseFileSystem
 from .dvc import DvcFileSystem
 
@@ -448,19 +449,21 @@ class RepoFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             for fname in files:
                 yield PathInfo(root) / fname
 
-    def _download(self, from_info, to_file, **kwargs):
+    def get_file(
+        self, from_info, to_file, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         fs, dvc_fs = self._get_fs_pair(from_info)
         try:
-            fs._download(  # pylint: disable=protected-access
-                from_info, to_file, **kwargs
+            fs.get_file(  # pylint: disable=protected-access
+                from_info, to_file, callback=callback, **kwargs
             )
             return
         except FileNotFoundError:
             if not dvc_fs:
                 raise
 
-        dvc_fs._download(  # pylint: disable=protected-access
-            from_info, to_file, **kwargs
+        dvc_fs.get_file(  # pylint: disable=protected-access
+            from_info, to_file, callback=callback, **kwargs
         )
 
     def metadata(self, path):

dvc/fs/s3.py

@@ -6,7 +6,7 @@ from collections import defaultdict
 from funcy import cached_property, wrap_prop
 
 from dvc.path_info import CloudURLInfo
-from dvc.progress import Tqdm
+from dvc.progress import DEFAULT_CALLBACK
 from dvc.scheme import Schemes
 
 from .fsspec_wrapper import ObjectFSWrapper
@@ -99,6 +99,11 @@ class BaseS3FileSystem(ObjectFSWrapper):
         client["endpoint_url"] = config.get("endpointurl")
         client["verify"] = config.get("ssl_verify")
 
+        # timeout configuration
+        config_kwargs = login_info["config_kwargs"]
+        config_kwargs["read_timeout"] = config.get("read_timeout")
+        config_kwargs["connect_timeout"] = config.get("connect_timeout")
+
         # encryptions
         additional = login_info["s3_additional_kwargs"]
         additional["ServerSideEncryption"] = config.get("sse")
@@ -204,36 +209,23 @@ class S3FileSystem(BaseS3FileSystem):  # pylint:disable=abstract-method
         return bucket.Object(path_info.path)
 
     @_translate_exceptions
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **pbar_args
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
-        total = os.path.getsize(from_file)
-        with Tqdm(
-            disable=no_progress_bar,
-            total=total,
-            bytes=True,
-            desc=name,
-            **pbar_args,
-        ) as pbar:
-            obj = self._get_obj(to_info)
-            obj.upload_file(
-                from_file,
-                Callback=pbar.update,
-                ExtraArgs=self.fs_args.get("s3_additional_kwargs"),
-                Config=self._transfer_config,
-            )
+        callback.set_size(os.path.getsize(from_file))
+        obj = self._get_obj(to_info)
+        obj.upload_file(
+            from_file,
+            Callback=callback.relative_update,
+            ExtraArgs=self.fs_args.get("s3_additional_kwargs"),
+            Config=self._transfer_config,
+        )
         self.fs.invalidate_cache(self._with_bucket(to_info.parent))
 
     @_translate_exceptions
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **pbar_args
+    def get_file(
+        self, from_info, to_info, callback=DEFAULT_CALLBACK, **kwargs
     ):
         obj = self._get_obj(from_info)
-        with Tqdm(
-            disable=no_progress_bar,
-            total=obj.content_length,
-            bytes=True,
-            desc=name,
-            **pbar_args,
-        ) as pbar:
-            obj.download_file(to_file, Callback=pbar.update)
+        callback.set_size(obj.content_length)
+        obj.download_file(to_info, Callback=callback.relative_update)

dvc/fs/ssh.py

@@ -8,7 +8,8 @@ from dvc import prompt
 from dvc.scheme import Schemes
 from dvc.utils.fs import as_atomic
 
-from .fsspec_wrapper import CallbackMixin, FSSpecWrapper
+from ..progress import DEFAULT_CALLBACK
+from .fsspec_wrapper import FSSpecWrapper
 
 _SSH_TIMEOUT = 60 * 30
 _SSH_CONFIG_FILE = os.path.expanduser(os.path.join("~", ".ssh", "config"))
@@ -26,7 +27,7 @@ def ask_password(host, user, port):
 
 
 # pylint:disable=abstract-method
-class SSHFileSystem(CallbackMixin, FSSpecWrapper):
+class SSHFileSystem(FSSpecWrapper):
     scheme = Schemes.SSH
     REQUIRES = {"sshfs": "sshfs"}
 
@@ -35,10 +36,8 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
 
     def _with_bucket(self, path):
         if isinstance(path, self.PATH_CLS):
-            path_str = path.path
-        else:
-            path_str = super()._with_bucket(path)
-        return os.path.normpath(path_str)
+            return path.path
+        return super()._with_bucket(path)
 
     def _prepare_credentials(self, **config):
         self.CAN_TRAVERSE = True
@@ -66,7 +65,13 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
             or self.DEFAULT_PORT
         )
 
+        if config.get("ask_password") and config.get("password") is None:
+            config["password"] = ask_password(
+                login_info["host"], login_info["username"], login_info["port"]
+            )
+
         login_info["password"] = config.get("password")
+        login_info["passphrase"] = config.get("password")
 
         raw_keys = []
         if config.get("keyfile"):
@@ -99,11 +104,6 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
         login_info["agent_forwarding"] = config.get("agent_forwarding", True)
         login_info["proxy_command"] = user_ssh_config.get("ProxyCommand")
 
-        if config.get("ask_password") and login_info["password"] is None:
-            login_info["password"] = ask_password(
-                login_info["host"], login_info["username"], login_info["port"]
-            )
-
         # We are going to automatically add stuff to known_hosts
         # something like paramiko's AutoAddPolicy()
         login_info["known_hosts"] = None
@@ -119,10 +119,12 @@ class SSHFileSystem(CallbackMixin, FSSpecWrapper):
     # Ensure that if an interrupt happens during the transfer, we don't
     # pollute the cache.
 
-    def _upload_fobj(self, fobj, to_info, *args, **kwargs):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         with as_atomic(self, to_info) as tmp_file:
-            super()._upload_fobj(fobj, tmp_file, *args, **kwargs)
+            super().upload_fobj(fobj, tmp_file, **kwargs)
 
-    def _upload(self, from_file, to_info, *args, **kwargs):
+    def put_file(
+        self, from_file, to_info, callback=DEFAULT_CALLBACK, **kwargs
+    ):
         with as_atomic(self, to_info) as tmp_file:
-            super()._upload(from_file, tmp_file, *args, **kwargs)
+            super().put_file(from_file, tmp_file, callback=callback, **kwargs)

dvc/fs/utils.py

@@ -33,7 +33,8 @@ def transfer(
             return from_fs.download_file(from_info, to_info)
 
         with from_fs.open(from_info, mode="rb") as fobj:
-            return to_fs.upload_fobj(fobj, to_info)
+            size = from_fs.getsize(from_info)
+            return to_fs.upload(fobj, to_info, total=size)
     except OSError as exc:
         # If the target file already exists, we are going to simply
         # ignore the exception (#4992).

dvc/fs/webdav.py

@@ -1,5 +1,4 @@
 import logging
-import os
 import threading
 from functools import lru_cache
 
@@ -71,15 +70,13 @@ class WebDAVFileSystem(FSSpecWrapper):  # pylint:disable=abstract-method
 
         return WebdavFileSystem(**self.fs_args)
 
-    def _upload_fobj(self, fobj, to_info, size: int = None):
+    def upload_fobj(self, fobj, to_info, **kwargs):
         rpath = self.translate_path_info(to_info)
-        self.makedirs(os.path.dirname(rpath))
+        size = kwargs.get("size")
         # using upload_fileobj to directly upload fileobj rather than buffering
         # and using overwrite=True to avoid check for an extra exists call,
         # as caller should ensure that the file does not exist beforehand.
-        return self.fs.client.upload_fileobj(
-            fobj, rpath, overwrite=True, size=size
-        )
+        return self.fs.upload_fileobj(fobj, rpath, overwrite=True, size=size)
 
     @lru_cache(512)
     def translate_path_info(self, path):
@@ -87,7 +84,7 @@ class WebDAVFileSystem(FSSpecWrapper):  # pylint:disable=abstract-method
             return path.path[len(self.prefix) :].lstrip("/")
         return path
 
-    _with_bucket = translate_path_info
+    _with_bucket = translate_path_info  # type: ignore
 
 
 class WebDAVSFileSystem(WebDAVFileSystem):  # pylint:disable=abstract-method

dvc/fs/webhdfs.py

@@ -1,51 +1,24 @@
-import logging
-import os
-import posixpath
-import shutil
 import threading
-from contextlib import contextmanager
 
 from funcy import cached_property, wrap_prop
 
-from dvc.hash_info import HashInfo
 from dvc.path_info import CloudURLInfo
-from dvc.progress import Tqdm
 from dvc.scheme import Schemes
 
-from .base import BaseFileSystem
-
-logger = logging.getLogger(__name__)
+# pylint:disable=abstract-method
+from .fsspec_wrapper import CallbackMixin, FSSpecWrapper
 
 
-def update_pbar(pbar, total):
-    """Update pbar to accept the two arguments passed by hdfs"""
-
-    def update(_, bytes_transfered):
-        if bytes_transfered == -1:
-            pbar.update_to(total)
-            return
-        pbar.update_to(bytes_transfered)
-
-    return update
-
-
-class WebHDFSFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
+class WebHDFSFileSystem(CallbackMixin, FSSpecWrapper):
     scheme = Schemes.WEBHDFS
     PATH_CLS = CloudURLInfo
-    REQUIRES = {"hdfs": "hdfs"}
+    REQUIRES = {"fsspec": "fsspec"}
     PARAM_CHECKSUM = "checksum"
-    TRAVERSE_PREFIX_LEN = 2
 
-    def __init__(self, **config):
-        super().__init__(**config)
-
-        self.host = config["host"]
-        self.user = config.get("user")
-        self.port = config.get("port")
-
-        self.hdfscli_config = config.get("hdfscli_config")
-        self.token = config.get("webhdfs_token")
-        self.alias = config.get("webhdfs_alias")
+    def _with_bucket(self, path):
+        if isinstance(path, self.PATH_CLS):
+            return f"/{path.path.rstrip('/')}"
+        return path
 
     @staticmethod
     def _get_kwargs_from_urls(urlpath):
@@ -57,115 +30,20 @@ class WebHDFSFileSystem(BaseFileSystem):  # pylint:disable=abstract-method
             )
         )
 
+    def _prepare_credentials(self, **config):
+        if "webhdfs_token" in config:
+            config["token"] = config.pop("webhdfs_token")
+
+        return config
+
     @wrap_prop(threading.Lock())
     @cached_property
-    def hdfs_client(self):
-        import hdfs
+    def fs(self):
+        from fsspec.implementations.webhdfs import WebHDFS
 
-        logger.debug("HDFSConfig: %s", self.hdfscli_config)
-
-        try:
-            return hdfs.config.Config(self.hdfscli_config).get_client(
-                self.alias
-            )
-        except hdfs.util.HdfsError as exc:
-            exc_msg = str(exc)
-            errors = (
-                "No alias specified",
-                "Invalid configuration file",
-                f"Alias {self.alias} not found",
-            )
-            if not any(err in exc_msg for err in errors):
-                raise
-
-            http_url = f"http://{self.host}:{self.port}"
-            logger.debug("URL: %s", http_url)
-
-            if self.token is not None:
-                client = hdfs.TokenClient(http_url, token=self.token, root="/")
-            else:
-                client = hdfs.InsecureClient(
-                    http_url, user=self.user, root="/"
-                )
-
-        return client
-
-    @contextmanager
-    def open(self, path_info, mode="r", encoding=None, **kwargs):
-        assert mode in {"r", "rt", "rb"}
-
-        with self.hdfs_client.read(
-            path_info.path, encoding=encoding
-        ) as reader:
-            yield reader
-
-    def walk_files(self, path_info, **kwargs):
-        if not self.exists(path_info):
-            return
-
-        root = path_info.path
-        for path, _, fnames in self.hdfs_client.walk(root):
-            for fname in fnames:
-                yield path_info.replace(path=posixpath.join(path, fname))
-
-    def remove(self, path_info):
-        if path_info.scheme != self.scheme:
-            raise NotImplementedError
-
-        self.hdfs_client.delete(path_info.path)
-
-    def exists(self, path_info) -> bool:
-        assert not isinstance(path_info, list)
-        assert path_info.scheme == "webhdfs"
-
-        status = self.hdfs_client.status(path_info.path, strict=False)
-        return status is not None
-
-    def info(self, path_info):
-        st = self.hdfs_client.status(path_info.path)
-        return {"size": st["length"], "type": "file"}
+        return WebHDFS(**self.fs_args)
 
     def checksum(self, path_info):
-        return HashInfo(
-            "checksum",
-            self.hdfs_client.checksum(path_info.path)["bytes"],
-            size=self.hdfs_client.status(path_info.path)["length"],
-        )
-
-    def copy(self, from_info, to_info, **_kwargs):
-        with self.hdfs_client.read(from_info.path) as reader:
-            with self.hdfs_client.write(to_info.path) as writer:
-                shutil.copyfileobj(reader, writer)
-
-    def move(self, from_info, to_info):
-        self.hdfs_client.makedirs(to_info.parent.path)
-        self.hdfs_client.rename(from_info.path, to_info.path)
-
-    def _upload_fobj(self, fobj, to_info, **kwargs):
-        with self.hdfs_client.write(to_info.path) as fdest:
-            shutil.copyfileobj(fobj, fdest)
-
-    def _upload(
-        self, from_file, to_info, name=None, no_progress_bar=False, **_kwargs
-    ):
-        total = os.path.getsize(from_file)
-        with Tqdm(
-            desc=name, total=total, disable=no_progress_bar, bytes=True
-        ) as pbar:
-            self.hdfs_client.upload(
-                to_info.path,
-                from_file,
-                overwrite=True,
-                progress=update_pbar(pbar, total),
-            )
-
-    def _download(
-        self, from_info, to_file, name=None, no_progress_bar=False, **_kwargs
-    ):
-        total = self.getsize(from_info)
-        with Tqdm(
-            desc=name, total=total, disable=no_progress_bar, bytes=True
-        ) as pbar:
-            self.hdfs_client.download(
-                from_info.path, to_file, progress=update_pbar(pbar, total)
-            )
+        path = self._with_bucket(path_info)
+        ukey = self.fs.ukey(path)
+        return ukey["bytes"]

dvc/ignore.py

@@ -11,7 +11,6 @@ from dvc.fs.base import BaseFileSystem
 from dvc.path_info import PathInfo
 from dvc.pathspec_math import PatternInfo, merge_patterns
 from dvc.scheme import Schemes
-from dvc.system import System
 from dvc.types import AnyPath, List, Optional
 from dvc.utils import relpath
 from dvc.utils.collections import PathStringTrie
@@ -84,7 +83,7 @@ class DvcIgnorePatterns(DvcIgnore):
         else:
             return False
 
-        if not System.is_unix():
+        if os.name == "nt":
             path = normalize_file(path)
         return path
 

dvc/info.py

@@ -136,11 +136,14 @@ def _get_supported_remotes():
 
 
 def get_fs_type(path):
-
-    partition = {
-        pathlib.Path(part.mountpoint): (part.fstype + " on " + part.device)
-        for part in psutil.disk_partitions(all=True)
-    }
+    partition = {}
+    for part in psutil.disk_partitions(all=True):
+        if part.fstype != "":
+            try:
+                mountpoint = pathlib.Path(part.mountpoint).resolve()
+                partition[mountpoint] = part.fstype + " on " + part.device
+            except PermissionError:
+                pass
 
     # need to follow the symlink: https://github.com/iterative/dvc/issues/5065
     path = pathlib.Path(path).resolve()

dvc/logger.py

@@ -142,9 +142,22 @@ class LoggerHandler(logging.StreamHandler):
         super().handleError(record)
         raise LoggingException(record)
 
+    def emit_pretty_exception(self, exc, verbose: bool = False):
+        return exc.__pretty_exc__(verbose=verbose)
+
     def emit(self, record):
         """Write to Tqdm's stream so as to not break progress-bars"""
         try:
+            if record.exc_info:
+                _, exc, *_ = record.exc_info
+                if hasattr(exc, "__pretty_exc__"):
+                    try:
+                        self.emit_pretty_exception(exc, verbose=_is_verbose())
+                        if not _is_verbose():
+                            return
+                    except Exception:  # noqa, pylint: disable=broad-except
+                        pass
+
             msg = self.format(record)
             Tqdm.write(
                 msg, file=self.stream, end=getattr(self, "terminator", "\n")

dvc/objects/checkout.py

@@ -1,4 +1,5 @@
 import logging
+from itertools import chain
 
 from shortuuid import uuid
 
@@ -10,52 +11,19 @@ from dvc.exceptions import (
     DvcException,
 )
 from dvc.ignore import DvcIgnoreFilter
-from dvc.objects import check, load
 from dvc.objects.db.slow_link_detection import (  # type: ignore[attr-defined]
     slow_link_guard,
 )
-from dvc.objects.errors import ObjectFormatError
+from dvc.objects.diff import ROOT
+from dvc.objects.diff import diff as odiff
 from dvc.objects.stage import stage
+from dvc.objects.tree import Tree
 from dvc.types import Optional
 
 logger = logging.getLogger(__name__)
 
 
-def _changed(path_info, fs, obj, cache, state=None):
-    logger.trace("checking if '%s'('%s') has changed.", path_info, obj)
-
-    try:
-        check(cache, obj)
-    except (FileNotFoundError, ObjectFormatError):
-        logger.debug(
-            "cache for '%s'('%s') has changed.", path_info, obj.hash_info
-        )
-        return True
-
-    actual = state.get(path_info, fs) if state else None
-    if actual is None:
-        try:
-            actual = stage(
-                cache, path_info, fs, obj.hash_info.name, dry_run=True
-            )[1].hash_info
-        except FileNotFoundError:
-            logger.debug("'%s' doesn't exist.", path_info)
-            return True
-
-    if obj.hash_info != actual:
-        logger.debug(
-            "hash value '%s' for '%s' has changed (actual '%s').",
-            obj.hash_info,
-            actual,
-            path_info,
-        )
-        return True
-
-    logger.trace("'%s' hasn't changed.", path_info)
-    return False
-
-
-def _remove(path_info, fs, cache, force=False):
+def _remove(path_info, fs, in_cache, force=False):
     if not fs.exists(path_info):
         return
 
@@ -63,13 +31,7 @@ def _remove(path_info, fs, cache, force=False):
         fs.remove(path_info)
         return
 
-    current = stage(cache, path_info, fs, fs.PARAM_CHECKSUM, dry_run=True)[
-        1
-    ].hash_info
-    try:
-        obj = load(cache, current)
-        check(cache, obj)
-    except (FileNotFoundError, ObjectFormatError):
+    if not in_cache:
         msg = (
             f"file/directory '{path_info}' is going to be removed. "
             "Are you sure you want to proceed?"
@@ -98,7 +60,7 @@ def _verify_link(cache, path_info, link_type):
 
 def _do_link(cache, from_info, to_info, link_method):
     if cache.fs.exists(to_info):
-        raise DvcException(f"Link '{to_info}' already exists!")
+        cache.fs.remove(to_info)  # broken symlink
 
     link_method(from_info, to_info)
 
@@ -130,7 +92,7 @@ def _link(cache, from_info, to_info):
     try:
         _try_links(cache, from_info, to_info, cache.cache_types)
     except FileNotFoundError as exc:
-        raise CheckoutError([to_info]) from exc
+        raise CheckoutError([str(to_info)]) from exc
 
 
 def _cache_is_copy(cache, path_info):
@@ -160,7 +122,7 @@ def _cache_is_copy(cache, path_info):
 def _checkout_file(
     path_info,
     fs,
-    obj,
+    change,
     cache,
     force,
     progress_callback=None,
@@ -169,24 +131,24 @@ def _checkout_file(
 ):
     """The file is changed we need to checkout a new copy"""
     modified = False
-    cache_info = cache.hash_to_path_info(obj.hash_info.value)
-    if fs.exists(path_info):
-        if not relink and _changed(path_info, fs, obj, cache, state=state):
-            modified = True
-            _remove(path_info, fs, cache, force=force)
-            _link(cache, cache_info, path_info)
-        else:
+    cache_info = cache.hash_to_path_info(change.new.obj.hash_info.value)
+    if change.old.obj:
+        if relink:
             if fs.iscopy(path_info) and _cache_is_copy(cache, path_info):
                 cache.unprotect(path_info)
             else:
-                _remove(path_info, fs, cache, force=force)
+                _remove(path_info, fs, change.old.in_cache, force=force)
                 _link(cache, cache_info, path_info)
+        else:
+            modified = True
+            _remove(path_info, fs, change.old.in_cache, force=force)
+            _link(cache, cache_info, path_info)
     else:
         _link(cache, cache_info, path_info)
         modified = True
 
     if state:
-        state.save(path_info, fs, obj.hash_info)
+        state.save(path_info, fs, change.new.obj.hash_info)
 
     if progress_callback:
         progress_callback(str(path_info))
@@ -194,109 +156,83 @@ def _checkout_file(
     return modified
 
 
-def _remove_redundant_files(
-    path_info, fs, obj, cache, force, dvcignore: Optional[DvcIgnoreFilter]
-):
-    if dvcignore:
-        existing_files = set(dvcignore.walk_files(fs, path_info))
-    else:
-        existing_files = set(fs.walk_files(path_info))
-
-    needed_files = {path_info.joinpath(*key) for key, _ in obj}
-    redundant_files = existing_files - needed_files
-    for path in redundant_files:
-        _remove(path, fs, cache, force)
-
-    return bool(redundant_files)
-
-
-def _checkout_dir(
+def _diff(
     path_info,
     fs,
     obj,
     cache,
-    force,
-    progress_callback=None,
     relink=False,
     dvcignore: Optional[DvcIgnoreFilter] = None,
-    state=None,
 ):
-    modified = False
-    # Create dir separately so that dir is created
-    # even if there are no files in it
-    if not fs.exists(path_info):
-        modified = True
-        fs.makedirs(path_info)
-
-    logger.debug("Linking directory '%s'.", path_info)
-
-    failed = []
-    for entry_key, entry_obj in obj:
-        try:
-            entry_modified = _checkout_file(
-                path_info.joinpath(*entry_key),
-                fs,
-                entry_obj,
-                cache,
-                force,
-                progress_callback,
-                relink,
-                state=None,
-            )
-            if entry_modified:
-                modified = True
-        except CheckoutError as exc:
-            failed.extend(exc.target_infos)
-
-    modified = (
-        _remove_redundant_files(
-            path_info, fs, obj, cache, force, dvcignore=dvcignore
+    old = None
+    try:
+        _, old = stage(
+            cache,
+            path_info,
+            fs,
+            obj.hash_info.name if obj else cache.fs.PARAM_CHECKSUM,
+            dry_run=True,
+            dvcignore=dvcignore,
         )
-        or modified
-    )
+    except FileNotFoundError:
+        pass
 
-    if failed:
-        raise CheckoutError(failed)
+    diff = odiff(old, obj, cache)
 
-    if state:
-        state.save(path_info, fs, obj.hash_info)
+    if relink:
+        diff.modified.extend(diff.unchanged)
 
-    # relink is not modified, assume it as nochange
-    return modified and not relink
+    return diff
 
 
 def _checkout(
+    diff,
     path_info,
     fs,
-    obj,
     cache,
     force=False,
     progress_callback=None,
     relink=False,
-    dvcignore: Optional[DvcIgnoreFilter] = None,
     state=None,
 ):
-    if not obj.hash_info.isdir:
-        ret = _checkout_file(
-            path_info, fs, obj, cache, force, progress_callback, relink, state
-        )
-    else:
-        ret = _checkout_dir(
-            path_info,
-            fs,
-            obj,
-            cache,
-            force,
-            progress_callback,
-            relink,
-            dvcignore=dvcignore,
-            state=state,
-        )
+    if not diff:
+        return
 
-    if state:
-        state.save_link(path_info, fs)
+    for change in diff.deleted:
+        entry_path = (
+            path_info.joinpath(*change.old.key)
+            if change.old.key != ROOT
+            else path_info
+        )
+        _remove(entry_path, fs, change.old.in_cache, force=force)
 
-    return ret
+    failed = []
+    for change in chain(diff.added, diff.modified):
+        entry_path = (
+            path_info.joinpath(*change.new.key)
+            if change.new.key != ROOT
+            else path_info
+        )
+        if isinstance(change.new.obj, Tree):
+            fs.makedirs(entry_path)
+            continue
+
+        try:
+            _checkout_file(
+                entry_path,
+                fs,
+                change,
+                cache,
+                force,
+                progress_callback,
+                relink,
+                state=state,
+            )
+        except CheckoutError as exc:
+            failed.extend(exc.target_infos)
+
+    if failed:
+        raise CheckoutError(failed)
 
 
 def checkout(
@@ -311,41 +247,54 @@ def checkout(
     dvcignore: Optional[DvcIgnoreFilter] = None,
     state=None,
 ):
+
     if path_info.scheme not in ["local", cache.fs.scheme]:
         raise NotImplementedError
 
-    failed = None
-    skip = False
+    diff = _diff(
+        path_info,
+        fs,
+        obj,
+        cache,
+        relink=relink,
+        dvcignore=dvcignore,
+    )
+
+    failed = []
     if not obj:
         if not quiet:
             logger.warning(
                 "No file hash info found for '%s'. It won't be created.",
                 path_info,
             )
-        _remove(path_info, fs, cache, force=force)
-        failed = path_info
-
-    elif not relink and not _changed(path_info, fs, obj, cache, state=state):
+        failed.append(str(path_info))
+    elif not diff:
         logger.trace("Data '%s' didn't change.", path_info)  # type: ignore
-        skip = True
 
-    if failed or skip:
+    try:
+        _checkout(
+            diff,
+            path_info,
+            fs,
+            cache,
+            force=force,
+            progress_callback=progress_callback,
+            relink=relink,
+            state=state,
+        )
+    except CheckoutError as exc:
+        failed.extend(exc.target_infos)
+
+    if diff and state:
+        state.save_link(path_info, fs)
+        if not failed:
+            state.save(path_info, fs, obj.hash_info)
+
+    if failed or not diff:
         if progress_callback and obj:
             progress_callback(str(path_info), len(obj))
         if failed:
-            raise CheckoutError([failed])
+            raise CheckoutError(failed)
         return
 
-    logger.debug("Checking out '%s' with cache '%s'.", path_info, obj)
-
-    return _checkout(
-        path_info,
-        fs,
-        obj,
-        cache,
-        force,
-        progress_callback,
-        relink,
-        dvcignore=dvcignore,
-        state=state,
-    )
+    return bool(diff) and not relink

dvc/objects/db/base.py

@@ -97,7 +97,6 @@ class ObjectDB:
         hash_info: "HashInfo",
         move: bool = True,
         verify: Optional[bool] = None,
-        **kwargs,
     ):
         if self.read_only:
             raise ObjectDBPermissionError("Cannot add to read-only ODB")
@@ -121,10 +120,6 @@ class ObjectDB:
         except (ObjectFormatError, FileNotFoundError):
             pass
 
-        callback = kwargs.get("download_callback")
-        if callback:
-            callback(1)
-
     def hash_to_path_info(self, hash_) -> "DvcPath":
         return self.path_info / hash_[0:2] / hash_[2:]
 

dvc/objects/db/reference.py

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

dvc/objects/diff.py

@@ -0,0 +1,135 @@
+from dataclasses import dataclass, field
+from typing import TYPE_CHECKING, List, Optional, Tuple
+
+if TYPE_CHECKING:
+    from .file import HashFile
+
+ADD = "add"
+MODIFY = "modify"
+DELETE = "delete"
+UNCHANGED = "unchanged"
+
+
+@dataclass
+class TreeEntry:
+    in_cache: bool
+    key: Tuple[str]
+    obj: Optional["HashFile"] = field(default=None)
+
+    def __bool__(self):
+        return bool(self.obj)
+
+    def __eq__(self, other):
+        if not isinstance(other, TreeEntry):
+            return False
+
+        if self.key != other.key or bool(self.obj) != bool(other.obj):
+            return False
+
+        if not self.obj:
+            return False
+
+        return self.obj.hash_info == other.obj.hash_info
+
+
+@dataclass
+class Change:
+    old: TreeEntry
+    new: TreeEntry
+
+    @property
+    def typ(self):
+        if not self.old and not self.new:
+            return UNCHANGED
+
+        if self.old and not self.new:
+            return DELETE
+
+        if not self.old and self.new:
+            return ADD
+
+        if self.old != self.new:
+            return MODIFY
+
+        return UNCHANGED
+
+    def __bool__(self):
+        return self.typ != UNCHANGED
+
+
+@dataclass
+class DiffResult:
+    added: List[Change] = field(default_factory=list, compare=True)
+    modified: List[Change] = field(default_factory=list, compare=True)
+    deleted: List[Change] = field(default_factory=list, compare=True)
+    unchanged: List[Change] = field(default_factory=list, compare=True)
+
+    def __bool__(self):
+        return bool(self.added or self.modified or self.deleted)
+
+
+ROOT = ("",)
+
+
+def diff(
+    old: Optional["HashFile"], new: Optional["HashFile"], cache
+) -> DiffResult:
+    from .tree import Tree
+
+    if old is None and new is None:
+        return DiffResult()
+
+    def _get_keys(obj):
+        if not obj:
+            return []
+        return [ROOT] + (
+            [key for key, _ in obj] if isinstance(obj, Tree) else []
+        )
+
+    old_keys = set(_get_keys(old))
+    new_keys = set(_get_keys(new))
+
+    def _get_obj(obj, key):
+        if not obj or key == ROOT:
+            return obj
+
+        return obj.get(key)
+
+    def _in_cache(obj, cache):
+        from . import check
+        from .errors import ObjectFormatError
+
+        if not obj:
+            return False
+
+        try:
+            check(cache, obj)
+            return True
+        except (FileNotFoundError, ObjectFormatError):
+            return False
+
+    ret = DiffResult()
+    for key in old_keys | new_keys:
+        old_obj = _get_obj(old, key)
+        new_obj = _get_obj(new, key)
+
+        change = Change(
+            old=TreeEntry(_in_cache(old_obj, cache), key, old_obj),
+            new=TreeEntry(_in_cache(new_obj, cache), key, new_obj),
+        )
+
+        if change.typ == ADD:
+            ret.added.append(change)
+        elif change.typ == MODIFY:
+            ret.modified.append(change)
+        elif change.typ == DELETE:
+            ret.deleted.append(change)
+        else:
+            assert change.typ == UNCHANGED
+            if not change.new.in_cache and not isinstance(
+                change.new.obj, Tree
+            ):
+                ret.modified.append(change)
+            else:
+                ret.unchanged.append(change)
+    return ret

dvc/objects/stage.py

@@ -34,9 +34,8 @@ def _upload_file(path_info, fs, odb, upload_odb):
     tmp_info = upload_odb.path_info / tmp_fname()
     with fs.open(path_info, mode="rb", chunk_size=fs.CHUNK_SIZE) as stream:
         stream = HashedStreamReader(stream)
-        upload_odb.fs.upload_fobj(
-            stream, tmp_info, desc=path_info.name, size=fs.getsize(path_info)
-        )
+        size = fs.getsize(path_info)
+        upload_odb.fs.upload(stream, tmp_info, desc=path_info.name, total=size)
 
     odb.add(tmp_info, upload_odb.fs, stream.hash_info)
     return path_info, odb.get(stream.hash_info)
@@ -46,18 +45,16 @@ def _get_file_hash(path_info, fs, name):
     info = fs.info(path_info)
     if name in info:
         assert not info[name].endswith(".dir")
-        return HashInfo(name, info[name], size=info["size"])
+        hash_value = info[name]
+    elif hasattr(fs, name):
+        func = getattr(fs, name)
+        hash_value = func(path_info)
+    elif name == "md5":
+        hash_value = file_md5(path_info, fs)
+    else:
+        raise NotImplementedError
 
-    func = getattr(fs, name, None)
-    if func:
-        return func(path_info)
-
-    if name == "md5":
-        return HashInfo(
-            name, file_md5(path_info, fs), size=fs.getsize(path_info)
-        )
-
-    raise NotImplementedError
+    return HashInfo(name, hash_value, size=info["size"])
 
 
 def get_file_hash(path_info, fs, name, state=None):

dvc/objects/transfer.py

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

dvc/output.py

@@ -9,7 +9,6 @@ from funcy import collecting, project
 from voluptuous import And, Any, Coerce, Length, Lower, Required, SetTo
 
 from dvc import objects, prompt
-from dvc.checkout import checkout
 from dvc.exceptions import (
     CheckoutError,
     CollectCacheError,
@@ -17,6 +16,7 @@ from dvc.exceptions import (
     MergeError,
     RemoteCacheRequiredError,
 )
+from dvc.objects.checkout import checkout
 
 from .fs import get_cloud_fs
 from .fs.hdfs import HDFSFileSystem
@@ -849,8 +849,9 @@ class Output:
         try:
             objects.check(self.odb, obj)
         except FileNotFoundError:
-            remote = self.repo.cloud.get_remote_odb(self.remote)
-            self.repo.cloud.pull([obj.hash_info], odb=remote, **kwargs)
+            if self.remote:
+                kwargs["remote"] = self.remote
+            self.repo.cloud.pull([obj.hash_info], **kwargs)
 
         if self.obj:
             return self.obj

dvc/progress.py

@@ -1,5 +1,4 @@
 """Manages progress bars for DVC repo."""
-
 import logging
 import sys
 from threading import RLock
@@ -137,8 +136,8 @@ class Tqdm(tqdm):
 
         return wrapped
 
-    def as_callback(self, fs, path_info):
-        return FsspecCallback(fs, path_info, self)
+    def as_callback(self):
+        return FsspecCallback(self)
 
     def close(self):
         self.postfix["info"] = ""
@@ -168,23 +167,46 @@ class Tqdm(tqdm):
 
 
 class FsspecCallback(fsspec.Callback):
-    def __init__(self, fs, path_info, progress_bar):
-        self.fs = fs
-        self.path_info = path_info
+    def __init__(self, progress_bar):
         self.progress_bar = progress_bar
         super().__init__()
 
     def set_size(self, size):
-        """``set_size()`` is an API that might be called with ``None`` if the
-        information is not already present on the caller. In that case, we'll
-        retrieve the size from an ``info()`` call."""
-        if size is None:
-            size = self.fs.info(self.path_info)["size"]
-        self.progress_bar.total = size
-        self.progress_bar.refresh()
+        if size is not None:
+            self.progress_bar.total = size
+            self.progress_bar.refresh()
+            super().set_size(size)
 
     def relative_update(self, inc=1):
         self.progress_bar.update(inc)
+        super().relative_update(inc)
 
     def absolute_update(self, value):
         self.progress_bar.update_to(value)
+        super().absolute_update(value)
+
+    @staticmethod
+    def wrap_fn(cb, fn):
+        def wrapped(*args, **kwargs):
+            res = fn(*args, **kwargs)
+            cb.relative_update()
+            return res
+
+        return wrapped
+
+
+def tdqm_or_callback_wrapped(
+    fobj, method, total, callback=None, **pbar_kwargs
+):
+    if callback:
+        from funcy import nullcontext
+        from tqdm.utils import CallbackIOWrapper
+
+        callback.set_size(total)
+        wrapper = CallbackIOWrapper(callback.relative_update, fobj, method)
+        return nullcontext(wrapper)
+
+    return Tqdm.wrapattr(fobj, method, total=total, bytes=True, **pbar_kwargs)
+
+
+DEFAULT_CALLBACK = fsspec.callbacks.NoOpCallback()

dvc/render/image.py

@@ -30,7 +30,6 @@ class ImageRenderer(Renderer):
         img_path = os.path.join(
             static, f"{revision}_{filename.replace(os.sep, '_')}"
         )
-        rel_img_path = relpath(img_path, page_dir_path)
         with open(img_path, "wb") as fd:
             fd.write(image_data)
         return """
@@ -38,7 +37,7 @@ class ImageRenderer(Renderer):
             <p>{title}</p>
             <img src="{src}">
         </div>""".format(
-            title=revision, src=rel_img_path
+            title=revision, src=(relpath(img_path, page_dir_path))
         )
 
     def _convert(self, path: "StrPath"):

dvc/render/utils.py

@@ -1,48 +1,28 @@
 import os.path
-from typing import Dict, List, Set
+from typing import Dict, List
+
+import dpath
 
 
-def get_files(data: Dict) -> Set:
+def get_files(data: Dict) -> List:
     files = set()
     for rev in data.keys():
         for file in data[rev].get("data", {}).keys():
             files.add(file)
-    return files
+    sorted_files = sorted(files)
+    return sorted_files
 
 
 def group_by_filename(plots_data: Dict) -> List[Dict]:
-    # TODO use dpath.util.search once
-    #  https://github.com/dpath-maintainers/dpath-python/issues/147 is released
-    #  now cannot search when errors are present in data
     files = get_files(plots_data)
     grouped = []
-
     for file in files:
-        tmp: Dict = {}
-        for revision, revision_data in plots_data.items():
-            if file in revision_data.get("data", {}):
-                if "data" not in tmp:
-                    tmp[revision] = {"data": {}}
-                tmp[revision]["data"].update(
-                    {file: revision_data["data"][file]}
-                )
-        grouped.append(tmp)
-
+        grouped.append(dpath.util.search(plots_data, ["*", "*", file]))
     return grouped
 
 
 def find_vega(repo, plots_data, target):
-    # TODO same as group_by_filename
-    grouped = group_by_filename(plots_data)
-    found = None
-    for plot_group in grouped:
-        files = get_files(plot_group)
-        assert len(files) == 1
-        file = files.pop()
-        if file == target:
-            found = plot_group
-            break
-
+    found = dpath.util.search(plots_data, ["*", "*", target])
     from dvc.render.vega import VegaRenderer
 
     if found and VegaRenderer.matches(found):

dvc/render/vega.py

@@ -28,7 +28,7 @@ class PlotDataStructureError(DvcException):
     def __init__(self):
         super().__init__(
             "Plot data extraction failed. Please see "
-            "https://man.dvc.org/plot for supported data formats."
+            "https://man.dvc.org/plots for supported data formats."
         )
 
 

dvc/repo/__init__.py

@@ -120,6 +120,34 @@ class Repo:
 
         return root_dir, dvc_dir, tmp_dir
 
+    def _get_database_dir(self, db_name):
+        # NOTE: by default, store SQLite-based remote indexes and state's
+        # `links` and `md5s` caches in the repository itself to avoid any
+        # possible state corruption in 'shared cache dir' scenario, but allow
+        # user to override this through config when, say, the repository is
+        # located on a mounted volume — see
+        # https://github.com/iterative/dvc/issues/4420
+        base_db_dir = self.config.get(db_name, {}).get("dir", None)
+        if not base_db_dir:
+            return self.tmp_dir
+
+        import hashlib
+
+        from dvc.utils.fs import makedirs
+
+        root_dir_hash = hashlib.sha224(
+            self.root_dir.encode("utf-8")
+        ).hexdigest()
+
+        db_dir = os.path.join(
+            base_db_dir,
+            self.DVC_DIR,
+            f"{os.path.basename(self.root_dir)}-{root_dir_hash[0:7]}",
+        )
+
+        makedirs(db_dir, exist_ok=True)
+        return db_dir
+
     def __init__(
         self,
         root_dir=None,
@@ -183,10 +211,8 @@ class Repo:
                 friendly=True,
             )
 
-            # NOTE: storing state and link_state in the repository itself to
-            # avoid any possible state corruption in 'shared cache dir'
-            # scenario.
-            self.state = State(self.root_dir, self.tmp_dir, self.dvcignore)
+            state_db_dir = self._get_database_dir("state")
+            self.state = State(self.root_dir, state_db_dir, self.dvcignore)
             self.odb = ODBManager(self)
 
             self.stage_cache = StageCache(self)
@@ -452,6 +478,10 @@ class Repo:
 
         return RepoFileSystem(self, subrepos=self.subrepos, **self._fs_conf)
 
+    @cached_property
+    def index_db_dir(self):
+        return self._get_database_dir("index")
+
     @contextmanager
     def open_by_relpath(self, path, remote=None, mode="r", encoding=None):
         """Opens a specified resource as a file descriptor"""

dvc/repo/diff.py

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

dvc/repo/experiments/init.py

@@ -0,0 +1,193 @@
+import os
+from typing import TYPE_CHECKING, Dict, Iterable
+
+from funcy import compact, lremove, post_processing
+from rich.prompt import Prompt
+
+from dvc.types import OptStr
+
+if TYPE_CHECKING:
+    from dvc.repo import Repo
+    from dvc.dvcfile import DVCFile
+    from dvc.stage import Stage
+
+from dvc.ui import ui
+
+PROMPTS = {
+    "cmd": "[b]Command[/b] to execute",
+    "code": "Path to a [b]code[/b] file/directory",
+    "data": "Path to a [b]data[/b] file/directory",
+    "models": "Path to a [b]model[/b] file/directory",
+    "params": "Path to a [b]parameters[/b] file",
+    "metrics": "Path to a [b]metrics[/b] file",
+    "plots": "Path to a [b]plots[/b] file/directory",
+    "live": "Path to log [b]dvclive[/b] outputs",
+}
+
+
+class RequiredPrompt(Prompt):
+    def process_response(self, value: str):
+        from rich.prompt import InvalidResponse
+
+        ret = super().process_response(value)
+        if not ret:
+            raise InvalidResponse(
+                "[prompt.invalid]Response required. Please try again."
+            )
+        return ret
+
+    @classmethod
+    def get_input(cls, *args, **kwargs) -> str:
+        try:
+            return super().get_input(*args, **kwargs)
+        except (KeyboardInterrupt, EOFError):
+            ui.error_write()
+            raise
+
+    def render_default(self, default):
+        from rich.text import Text
+
+        return Text(f"{default!s}", "green")
+
+
+class SkippablePrompt(RequiredPrompt):
+    skip_value: str = "n"
+
+    def process_response(self, value: str):
+        ret = super().process_response(value)
+        return None if ret == self.skip_value else ret
+
+    def make_prompt(self, default):
+        prompt = self.prompt.copy()
+        prompt.end = ""
+
+        prompt.append(" [")
+        if (
+            default is not ...
+            and self.show_default
+            and isinstance(default, (str, self.response_type))
+        ):
+            _default = self.render_default(default)
+            prompt.append(_default)
+            prompt.append(", ")
+
+        prompt.append(f"{self.skip_value} to skip", style="italic")
+        prompt.append("]")
+        prompt.append(self.prompt_suffix)
+        return prompt
+
+
+@post_processing(dict)
+def _prompts(keys: Iterable[str], defaults: Dict[str, OptStr]):
+    for key in keys:
+        if key == "cmd":
+            prompt_cls = RequiredPrompt
+        else:
+            prompt_cls = SkippablePrompt
+        kwargs = {"default": defaults[key]} if key in defaults else {}
+        prompt = PROMPTS.get(key)
+        value = prompt_cls.ask(prompt, console=ui.error_console, **kwargs)
+        yield key, value
+
+
+def init_interactive(
+    defaults: Dict[str, str],
+    provided: Iterable[str],
+    show_heading: bool = False,
+    live: bool = False,
+) -> Dict[str, str]:
+    primary = lremove(provided, ["cmd", "code", "data", "models", "params"])
+    secondary = lremove(provided, ["live"] if live else ["metrics", "plots"])
+
+    if not (primary or secondary):
+        return {}
+
+    message = (
+        "This command will guide you to set up your first stage in "
+        "[green]dvc.yaml[/green].\n"
+    )
+    if show_heading:
+        ui.error_write(message, styled=True)
+
+    return compact(
+        {
+            **_prompts(primary, defaults),
+            **_prompts(secondary, defaults),
+        }
+    )
+
+
+def _check_stage_exists(
+    dvcfile: "DVCFile", name: str, force: bool = False
+) -> None:
+    if not force and dvcfile.exists() and name in dvcfile.stages:
+        from dvc.stage.exceptions import DuplicateStageName
+
+        hint = "Use '--force' to overwrite."
+        raise DuplicateStageName(
+            f"Stage '{name}' already exists in 'dvc.yaml'. {hint}"
+        )
+
+
+def init(
+    repo: "Repo",
+    name: str = None,
+    type: str = "default",  # pylint: disable=redefined-builtin
+    defaults: Dict[str, str] = None,
+    overrides: Dict[str, str] = None,
+    interactive: bool = False,
+    force: bool = False,
+) -> "Stage":
+    from dvc.dvcfile import make_dvcfile
+
+    dvcfile = make_dvcfile(repo, "dvc.yaml")
+    name = name or type
+
+    _check_stage_exists(dvcfile, name, force=force)
+
+    defaults = defaults or {}
+    overrides = overrides or {}
+
+    with_live = type == "live"
+    if interactive:
+        defaults = init_interactive(
+            defaults=defaults or {},
+            show_heading=not dvcfile.exists(),
+            live=with_live,
+            provided=overrides.keys(),
+        )
+    else:
+        if with_live:
+            # suppress `metrics`/`params` if live is selected, unless
+            # it is also provided via overrides/cli.
+            # This makes output to be a checkpoint as well.
+            defaults.pop("metrics")
+            defaults.pop("params")
+        else:
+            defaults.pop("live")  # suppress live otherwise
+
+    context: Dict[str, str] = {**defaults, **overrides}
+    assert "cmd" in context
+
+    params_kv = []
+    if context.get("params"):
+        from dvc.utils.serialize import LOADERS
+
+        path = context["params"]
+        assert isinstance(path, str)
+        _, ext = os.path.splitext(path)
+        params_kv = [{path: list(LOADERS[ext](path))}]
+
+    checkpoint_out = bool(context.get("live"))
+    models = context.get("models")
+    return repo.stage.add(
+        name=name,
+        cmd=context["cmd"],
+        deps=compact([context.get("code"), context.get("data")]),
+        params=params_kv,
+        metrics_no_cache=compact([context.get("metrics")]),
+        plots_no_cache=compact([context.get("plots")]),
+        live=context.get("live"),
+        force=force,
+        **{"checkpoints" if checkpoint_out else "outs": compact([models])},
+    )

dvc/schema.py

@@ -1,3 +1,5 @@
+from collections.abc import Mapping
+
 from voluptuous import Any, Optional, Required, Schema
 
 from dvc import dependency, output
@@ -94,11 +96,27 @@ STAGE_DEFINITION = {
     Optional(StageParams.PARAM_LIVE): Any(str, LIVE_PSTAGE_SCHEMA),
 }
 
+
+def either_or(primary, fallback, fallback_includes=None):
+    def validator(data):
+        schema = primary
+        if (
+            isinstance(data, Mapping)
+            and set(fallback_includes or []) & data.keys()
+        ):
+            schema = fallback
+        return Schema(schema)(data)
+
+    return validator
+
+
 FOREACH_IN = {
     Required(FOREACH_KWD): Any(dict, list, str),
     Required(DO_KWD): STAGE_DEFINITION,
 }
-SINGLE_PIPELINE_STAGE_SCHEMA = {str: Any(FOREACH_IN, STAGE_DEFINITION)}
+SINGLE_PIPELINE_STAGE_SCHEMA = {
+    str: either_or(STAGE_DEFINITION, FOREACH_IN, [FOREACH_KWD, DO_KWD])
+}
 MULTI_STAGE_SCHEMA = {
     STAGES: SINGLE_PIPELINE_STAGE_SCHEMA,
     VARS_KWD: VARS_SCHEMA,

dvc/scm/base.py

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

dvc/scm/git/backend/dulwich.py

@@ -20,7 +20,7 @@ from funcy import cached_property
 
 from dvc.path_info import PathInfo
 from dvc.progress import Tqdm
-from dvc.scm.base import InvalidRemoteSCMRepo, SCMError
+from dvc.scm.base import GitAuthError, InvalidRemoteSCMRepo, SCMError
 from dvc.utils import relpath
 
 from ..objects import GitObject
@@ -355,7 +355,7 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 yield os.fsdecode(key)
 
     def iter_remote_refs(self, url: str, base: Optional[str] = None):
-        from dulwich.client import get_transport_and_path
+        from dulwich.client import HTTPUnauthorized, get_transport_and_path
         from dulwich.errors import NotGitRepository
         from dulwich.porcelain import get_remote_repo
 
@@ -376,6 +376,8 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 yield from (os.fsdecode(ref) for ref in client.get_refs(path))
         except NotGitRepository as exc:
             raise InvalidRemoteSCMRepo(url) from exc
+        except HTTPUnauthorized:
+            raise GitAuthError(url)
 
     def get_refs_containing(self, rev: str, pattern: Optional[str] = None):
         raise NotImplementedError
@@ -388,7 +390,7 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
         force: bool = False,
         on_diverged: Optional[Callable[[str, str], bool]] = None,
     ):
-        from dulwich.client import get_transport_and_path
+        from dulwich.client import HTTPUnauthorized, get_transport_and_path
         from dulwich.errors import NotGitRepository, SendPackError
         from dulwich.porcelain import (
             DivergedBranches,
@@ -447,6 +449,8 @@ class DulwichBackend(BaseGitBackend):  # pylint:disable=abstract-method
                 )
         except (NotGitRepository, SendPackError) as exc:
             raise SCMError("Git failed to push '{src}' to '{url}'") from exc
+        except HTTPUnauthorized:
+            raise GitAuthError(url)
 
     def _push_dest_refs(
         self, src: Optional[str], dest: str

dvc/stage/__init__.py

@@ -273,7 +273,7 @@ class Stage(params.StageParams):
             from dvc.output import Output
             from dvc.schema import LIVE_PROPS
 
-            env[DVCLIVE_PATH] = str(out.path_info)
+            env[DVCLIVE_PATH] = relpath(out.path_info, self.wdir)
             if isinstance(out.live, dict):
                 config = project(out.live, LIVE_PROPS)
 

dvc/stage/exceptions.py

@@ -9,10 +9,6 @@ class StageCmdFailedError(DvcException):
         super().__init__(msg)
 
 
-class StageFileFormatError(DvcException):
-    pass
-
-
 class StageFileDoesNotExistError(DvcException):
     DVC_IGNORED = "is dvc-ignored"
     DOES_NOT_EXIST = "does not exist"

dvc/state.py

@@ -53,7 +53,6 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         self.tmp_dir = tmp_dir
         self.root_dir = root_dir
         self.dvcignore = dvcignore
-        self.fs = LocalFileSystem()
 
         if not tmp_dir:
             return
@@ -80,12 +79,7 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         if not isinstance(fs, LocalFileSystem):
             return
 
-        assert isinstance(path_info, str) or path_info.scheme == "local"
-        assert hash_info
-        assert isinstance(hash_info, HashInfo)
-        assert os.path.exists(path_info)
-
-        mtime, size = get_mtime_and_size(path_info, self.fs, self.dvcignore)
+        mtime, size = get_mtime_and_size(path_info, fs, self.dvcignore)
         inode = get_inode(path_info)
 
         logger.debug(
@@ -112,17 +106,12 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         if not isinstance(fs, LocalFileSystem):
             return None
 
-        assert isinstance(path_info, str) or path_info.scheme == "local"
-        path = os.fspath(path_info)
-
-        # NOTE: use os.path.exists instead of LocalFileSystem.exists
-        # because it uses lexists() and will return True for broken
-        # symlinks that we cannot stat() in get_mtime_and_size
-        if not os.path.exists(path):
+        try:
+            mtime, size = get_mtime_and_size(path_info, fs, self.dvcignore)
+        except FileNotFoundError:
             return None
 
-        mtime, size = get_mtime_and_size(path, self.fs, self.dvcignore)
-        inode = get_inode(path)
+        inode = get_inode(path_info)
 
         value = self.md5s.get(inode)
 
@@ -141,12 +130,11 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
         if not isinstance(fs, LocalFileSystem):
             return
 
-        assert isinstance(path_info, str) or path_info.scheme == "local"
-
-        if not self.fs.exists(path_info):
+        try:
+            mtime, _ = get_mtime_and_size(path_info, fs, self.dvcignore)
+        except FileNotFoundError:
             return
 
-        mtime, _ = get_mtime_and_size(path_info, self.fs, self.dvcignore)
         inode = get_inode(path_info)
         relative_path = relpath(path_info, self.root_dir)
 
@@ -168,11 +156,11 @@ class State(StateBase):  # pylint: disable=too-many-instance-attributes
             for relative_path in ref:
                 path = os.path.join(self.root_dir, relative_path)
 
-                if path in used or not self.fs.exists(path):
+                if path in used or not fs.exists(path):
                     continue
 
                 inode = get_inode(path)
-                mtime, _ = get_mtime_and_size(path, self.fs, self.dvcignore)
+                mtime, _ = get_mtime_and_size(path, fs, self.dvcignore)
 
                 if ref[relative_path] == (inode, mtime):
                     logger.debug("Removing '%s' as unused link.", path)

dvc/system.py

@@ -2,30 +2,13 @@ import errno
 import logging
 import os
 import platform
-import sys
 
 from dvc.exceptions import DvcException
 
 logger = logging.getLogger(__name__)
 
-if (
-    sys.platform == "win32"
-    and sys.version_info < (3, 8)
-    and sys.getwindowsversion() >= (6, 2)
-):
-    try:
-        import speedcopy
-
-        speedcopy.patch_copyfile()
-    except ImportError:
-        pass
-
 
 class System:
-    @staticmethod
-    def is_unix():
-        return os.name != "nt"
-
     @staticmethod
     def hardlink(source, link_name):
         try:
@@ -117,99 +100,14 @@ class System:
         # so we need to chmod it to look like a normal copy.
         os.chmod(link_name, 0o666 & ~umask)
 
-    @staticmethod
-    def _getdirinfo(path):
-        from collections import namedtuple
-
-        from win32file import (  # pylint: disable=import-error
-            FILE_FLAG_BACKUP_SEMANTICS,
-            FILE_FLAG_OPEN_REPARSE_POINT,
-            FILE_SHARE_READ,
-            OPEN_EXISTING,
-            CreateFileW,
-            GetFileInformationByHandle,
-        )
-
-        # NOTE: use FILE_FLAG_OPEN_REPARSE_POINT to open symlink itself and not
-        # the target See https://docs.microsoft.com/en-us/windows/desktop/api/
-        # fileapi/nf-fileapi-createfilew#symbolic-link-behavior
-        flags = FILE_FLAG_BACKUP_SEMANTICS | FILE_FLAG_OPEN_REPARSE_POINT
-
-        hfile = CreateFileW(
-            path, 0, FILE_SHARE_READ, None, OPEN_EXISTING, flags, None
-        )
-
-        # See BY_HANDLE_FILE_INFORMATION structure from fileapi.h
-        Info = namedtuple(
-            "BY_HANDLE_FILE_INFORMATION",
-            [
-                "dwFileAttributes",
-                "ftCreationTime",
-                "ftLastAccessTime",
-                "ftLastWriteTime",
-                "dwVolumeSerialNumber",
-                "nFileSizeHigh",
-                "nFileSizeLow",
-                "nNumberOfLinks",
-                "nFileIndexHigh",
-                "nFileIndexLow",
-            ],
-        )
-
-        return Info(*GetFileInformationByHandle(hfile))
-
     @staticmethod
     def inode(path):
-        path = os.fspath(path)
-
-        if System.is_unix():
-            import ctypes
-
-            inode = os.lstat(path).st_ino
-            # NOTE: See https://bugs.python.org/issue29619 and
-            # https://stackoverflow.com/questions/34643289/
-            # pythons-os-stat-is-returning-wrong-inode-value
-            inode = ctypes.c_ulong(inode).value
-        else:
-            # getdirinfo from ntfsutils works on both files and dirs
-            info = System._getdirinfo(path)
-            inode = abs(
-                hash(
-                    (
-                        info.dwVolumeSerialNumber,
-                        info.nFileIndexHigh,
-                        info.nFileIndexLow,
-                    )
-                )
-            )
-        assert inode >= 0
-        assert inode < 2 ** 64
-        return inode
+        return os.lstat(path).st_ino
 
     @staticmethod
     def is_symlink(path):
-        path = os.fspath(path)
-
-        if System.is_unix():
-            return os.path.islink(path)
-
-        # https://docs.microsoft.com/en-us/windows/desktop/fileio/
-        # file-attribute-constants
-        from winnt import (  # pylint: disable=import-error
-            FILE_ATTRIBUTE_REPARSE_POINT,
-        )
-
-        if os.path.lexists(path):
-            info = System._getdirinfo(path)
-            return info.dwFileAttributes & FILE_ATTRIBUTE_REPARSE_POINT
-        return False
+        return os.path.islink(path)
 
     @staticmethod
     def is_hardlink(path):
-        path = os.fspath(path)
-
-        if System.is_unix():
-            return os.stat(path).st_nlink > 1
-
-        info = System._getdirinfo(path)
-        return info.nNumberOfLinks > 1
+        return os.stat(path).st_nlink > 1

dvc/ui/__init__.py

@@ -1,12 +1,15 @@
-import sys
+from contextlib import contextmanager
 from typing import (
     TYPE_CHECKING,
     Any,
+    Callable,
     Dict,
     Iterable,
+    Iterator,
     Optional,
     Sequence,
     TextIO,
+    Type,
     Union,
 )
 
@@ -14,6 +17,7 @@ from funcy import cached_property
 
 if TYPE_CHECKING:
     from rich.status import Status
+    from rich.text import Text as RichText
 
     from dvc.progress import Tqdm
     from dvc.ui.table import Headers, Styles, TableData
@@ -40,8 +44,13 @@ class Console:
     def __init__(
         self, formatter: Formatter = None, enable: bool = False
     ) -> None:
+        from contextvars import ContextVar
+
         self.formatter: Formatter = formatter or Formatter()
         self._enabled: bool = enable
+        self._paginate: ContextVar[bool] = ContextVar(
+            "_paginate", default=False
+        )
 
     def enable(self) -> None:
         self._enabled = True
@@ -74,6 +83,23 @@ class Console:
             styled=styled,
         )
 
+    def write_json(
+        self, data: Any, indent: int = 2, default: Callable = None
+    ) -> None:
+        import json
+
+        if self.isatty():
+            from rich.highlighter import JSONHighlighter
+
+            j = json.dumps(data, indent=indent, default=default)
+            highlighter = JSONHighlighter()
+            text = highlighter(j)
+            text.no_wrap = True
+            text.overflow = None
+            return self.write(text, styled=True)
+
+        return self.write(json.dumps(data))
+
     def write(
         self,
         *objects: Any,
@@ -85,6 +111,8 @@ class Console:
         styled: bool = False,
         file: TextIO = None,
     ) -> None:
+        import sys
+
         from dvc.progress import Tqdm
 
         sep = " " if sep is None else sep
@@ -94,21 +122,39 @@ class Console:
 
         file = file or (sys.stderr if stderr else sys.stdout)
         with Tqdm.external_write_mode(file=file):
-            if styled:
+            # if we are inside pager context, send the output to rich's buffer
+            if styled or self._paginate.get():
                 console = self.error_console if stderr else self.rich_console
-                return console.print(*objects, sep=sep, end=end)
+                if styled:
+                    return console.print(*objects, sep=sep, end=end)
+                return console.out(*objects, sep=sep, end=end, highlight=False)
 
-            values = (
-                self.formatter.format(obj, style=style) for obj in objects
-            )
+            values = (self.formatter.format(obj, style) for obj in objects)
             return print(*values, sep=sep, end=end, file=file)
 
+    @property
+    def rich_text(self) -> "Type[RichText]":
+        from rich.text import Text
+
+        return Text
+
     @staticmethod
     def progress(*args, **kwargs) -> "Tqdm":
         from dvc.progress import Tqdm
 
         return Tqdm(*args, **kwargs)
 
+    @contextmanager
+    def pager(self, styles: bool = True) -> Iterator[None]:
+        from .pager import DvcPager
+
+        tok = self._paginate.set(True)
+        try:
+            with self.rich_console.pager(pager=DvcPager(), styles=styles):
+                yield
+        finally:
+            self._paginate.reset(tok)
+
     def prompt(
         self, text: str, choices: Iterable[str] = None, password: bool = False
     ) -> Optional[str]:
@@ -195,6 +241,11 @@ class Console:
     def status(self, status: str, **kwargs: Any) -> "Status":
         return self.error_console.status(status, **kwargs)
 
+    def isatty(self) -> bool:
+        import sys
+
+        return sys.stdout.isatty()
+
 
 ui = Console()
 

dvc/ui/table.py

@@ -1,5 +1,5 @@
 from collections import abc
-from contextlib import contextmanager
+from contextlib import ExitStack, contextmanager
 from itertools import zip_longest
 from typing import TYPE_CHECKING, Dict, Iterator, Sequence, Union
 
@@ -8,15 +8,13 @@ from dvc.types import DictStrAny
 if TYPE_CHECKING:
     from rich.console import Console as RichConsole
     from rich.table import Table
-    from rich.text import Text
-
-    from dvc.ui import Console
 
+    from dvc.ui import Console, RichText
 
 SHOW_MAX_WIDTH = 1024
 
 
-CellT = Union[str, "Text"]  # Text is mostly compatible with str
+CellT = Union[str, "RichText"]  # RichText is mostly compatible with str
 Row = Sequence[CellT]
 TableData = Sequence[Row]
 Headers = Sequence[str]
@@ -31,6 +29,7 @@ def plain_table(
     pager: bool = False,
     force: bool = True,
 ) -> None:
+    from funcy import nullcontext
     from tabulate import tabulate
 
     text: str = tabulate(
@@ -45,11 +44,8 @@ def plain_table(
         # NOTE: md table is incomplete without the trailing newline
         text += "\n"
 
-    if pager:
-        from dvc.utils.pager import pager as _pager
-
-        _pager(text)
-    else:
+    cm = ui.pager() if pager else nullcontext()
+    with cm:
         ui.write(text, force=force)
 
 
@@ -111,14 +107,13 @@ def rich_table(
     for row, style in zip_longest(data, rs):
         table.add_row(*row, **(style or {}))
 
-    console = ui.rich_console
+    stack = ExitStack()
+    if pager:
+        stack.enter_context(
+            console_width(table, ui.rich_console, SHOW_MAX_WIDTH)
+        )
+        stack.enter_context(ui.pager())
 
-    if not pager:
-        console.print(table)
+    with stack:
+        ui.write(table, styled=True)
         return
-
-    from dvc.utils.pager import DvcPager
-
-    with console_width(table, console, SHOW_MAX_WIDTH):
-        with console.pager(pager=DvcPager(), styles=True):
-            console.print(table)

dvc/updater.py

@@ -10,7 +10,7 @@ from dvc import __version__
 from dvc.utils.pkg import PKG
 
 if TYPE_CHECKING:
-    from rich.text import Text
+    from dvc.ui import RichText
 
 logger = logging.getLogger(__name__)
 
@@ -123,20 +123,23 @@ class Updater:
         current: str = None,
         color: str = "yellow",
         pkg: Optional[str] = None,
-    ) -> "Text":
-        from rich.text import Text
+    ) -> "RichText":
+        from dvc.ui import ui
 
         current = current or self.current
-        update_message = Text.from_markup(
+        update_message = ui.rich_text.from_markup(
             f"You are using dvc version [bold]{current}[/]; "
             f"however, version [bold]{latest}[/] is available."
         )
-        instruction = Text.from_markup(self._get_update_instructions(pkg=pkg))
-        return Text.assemble(
+        instruction = ui.rich_text.from_markup(
+            self._get_update_instructions(pkg=pkg)
+        )
+        return ui.rich_text.assemble(
             "\n", update_message, "\n", instruction, style=color
         )
 
-    def _get_update_instructions(self, pkg: Optional[str] = None) -> str:
+    @staticmethod
+    def _get_update_instructions(pkg: Optional[str] = None) -> str:
         if pkg in ("osxpkg", "exe", "binary"):
             return (
                 "To upgrade, uninstall dvc and reinstall from "

dvc/utils/_benedict.py

@@ -0,0 +1,26 @@
+"""
+Rollbacks monkeypatching of `json.encoder` by benedict.
+
+It monkeypatches json to use Python-based encoder instead of C-based
+during import.
+
+We rollback that monkeypatch by keeping reference to that C-based
+encoder and reinstate them after importing benedict.
+See: https://github.com/iterative/dvc/issues/6423
+     https://github.com/fabiocaccamo/python-benedict/issues/62
+and the source of truth:
+https://github.com/fabiocaccamo/python-benedict/blob/c98c471065/benedict/dicts/__init__.py#L282-L285
+"""
+from json import encoder
+
+try:
+    c_make_encoder = encoder.c_make_encoder  # type: ignore[attr-defined]
+except AttributeError:
+    c_make_encoder = None
+
+
+from benedict import benedict  # noqa: E402
+
+encoder.c_make_encoder = c_make_encoder  # type: ignore[attr-defined]
+# Please import benedict from here lazily
+__all__ = ["benedict"]

dvc/utils/collections.py

@@ -82,19 +82,10 @@ def chunk_dict(d: Dict[_KT, _VT], size: int = 1) -> List[Dict[_KT, _VT]]:
 
 def merge_params(src: Dict, to_update: Dict) -> Dict:
     """Recursively merges params with benedict's syntax support in-place."""
-    from benedict import benedict
+    from ._benedict import benedict
 
     data = benedict(src)
-    if src:
-        data.merge(to_update, overwrite=True)
-    else:
-        # NOTE: the following line may seem like an unnecessary duplication
-        # data.merge might affect the `src` if it's not empty, so we cannot
-        # check `if src` later, as it may have been mutated already.
-        data.merge(to_update, overwrite=True)
-        # benedict has issues keeping references to an empty dictionary
-        # see: https://github.com/iterative/dvc/issues/6374.
-        src.update(data)
+    data.merge(to_update, overwrite=True)
     return src
 
 

dvc/utils/fs.py

@@ -193,33 +193,39 @@ def makedirs(path, exist_ok=False, mode=None):
         logger.trace("failed to chmod '%o' '%s'", mode, path, exc_info=True)
 
 
-def copyfile(src, dest, no_progress_bar=False, name=None):
+def copyfile(src, dest, callback=None, no_progress_bar=False, name=None):
     """Copy file with progress bar"""
-    from dvc.progress import Tqdm
-
     name = name if name else os.path.basename(dest)
     total = os.stat(src).st_size
 
     if os.path.isdir(dest):
         dest = os.path.join(dest, os.path.basename(src))
 
+    if callback:
+        callback.set_size(total)
+
     try:
         System.reflink(src, dest)
     except DvcException:
+        from dvc.progress import tdqm_or_callback_wrapped
+
         with open(src, "rb") as fsrc, open(dest, "wb+") as fdest:
-            with Tqdm.wrapattr(
+            with tdqm_or_callback_wrapped(
                 fdest,
                 "write",
-                desc=name,
+                total,
+                callback=callback,
                 disable=no_progress_bar,
-                total=total,
-                bytes=True,
-            ) as fdest_wrapped:
+                desc=name,
+            ) as wrapped:
                 while True:
                     buf = fsrc.read(LOCAL_CHUNK_SIZE)
                     if not buf:
                         break
-                    fdest_wrapped.write(buf)
+                    wrapped.write(buf)
+
+    if callback:
+        callback.absolute_update(total)
 
 
 def copy_fobj_to_file(fsrc, dest):

dvc/utils/http.py

@@ -1,67 +0,0 @@
-import io
-from contextlib import contextmanager
-
-from dvc.utils.stream import IterStream
-
-
-@contextmanager
-def open_url(url, mode="r", encoding=None, **iter_opts):
-    """Opens an url as a readable stream.
-
-    Resumes on connection error.
-    Url could be a string or a callable returning a string.
-    """
-    assert mode in {"r", "rt", "rb"}
-
-    with iter_url(url, **iter_opts) as (response, it):
-        bytes_stream = IterStream(it)
-
-        if mode == "rb":
-            yield bytes_stream
-        else:
-            encoding = encoding or response.encoding
-            yield io.TextIOWrapper(bytes_stream, encoding=encoding)
-
-
-@contextmanager
-def iter_url(url, chunk_size=io.DEFAULT_BUFFER_SIZE, auth=None):
-    """Iterate over chunks requested from url."""
-    import requests
-
-    def request(headers=None):
-        the_url = url() if callable(url) else url
-        response = requests.get(
-            the_url, stream=True, headers=headers, auth=auth
-        )
-        if response.status_code == 404:
-            raise FileNotFoundError(f"Can't open {the_url}")
-        response.raise_for_status()
-        return response
-
-    def gen(response):
-        try:
-            pos = 0
-            while True:
-                try:
-                    for chunk in response.iter_content(chunk_size):
-                        pos += len(chunk)
-                        yield chunk
-                    break
-                except requests.ConnectionError:
-                    response.close()
-                    if response.headers.get("Accept-Ranges") != "bytes":
-                        raise
-
-                    # Reopen request from where we stopped
-                    headers = {"Range": f"bytes={pos}-"}
-                    response = request(headers)
-        finally:
-            response.close()
-
-    response = request()
-    it = gen(response)
-    try:
-        yield response, it
-    finally:
-        # Ensure connection is closed
-        it.close()

dvc/utils/serialize/__init__.py

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

dvc/utils/serialize/_common.py

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

dvc/utils/serialize/_json.py

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

dvc/utils/serialize/_toml.py

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

dvc/utils/serialize/_yaml.py

@@ -58,8 +58,8 @@ def _dump(data, stream):
     return yaml.dump(data, stream)
 
 
-def dump_yaml(path, data, fs=None):
-    return _dump_data(path, data, dumper=_dump, fs=fs)
+def dump_yaml(path, data, fs=None, **kwargs):
+    return _dump_data(path, data, dumper=_dump, fs=fs, **kwargs)
 
 
 def loads_yaml(s, typ="safe"):

dvc/utils/stream.py

@@ -8,61 +8,6 @@ from dvc.istextfile import DEFAULT_CHUNK_SIZE, istextblock
 from dvc.utils import dos2unix
 
 
-class IterStream(io.RawIOBase):
-    """Wraps an iterator yielding bytes as a file object"""
-
-    def __init__(self, iterator):  # pylint: disable=super-init-not-called
-        self.iterator = iterator
-        self.leftover = b""
-
-    def readable(self):
-        return True
-
-    def writable(self) -> bool:
-        return False
-
-    # Python 3 requires only .readinto() method, it still uses other ones
-    # under some circumstances and falls back if those are absent. Since
-    # iterator already constructs byte strings for us, .readinto() is not the
-    # most optimal, so we provide .read1() too.
-
-    def readinto(self, b):
-        try:
-            n = len(b)  # We're supposed to return at most this much
-            chunk = self.leftover or next(self.iterator)
-            output, self.leftover = chunk[:n], chunk[n:]
-
-            n_out = len(output)
-            b[:n_out] = output
-            return n_out
-        except StopIteration:
-            return 0  # indicate EOF
-
-    readinto1 = readinto
-
-    def read1(self, n=-1):
-        try:
-            chunk = self.leftover or next(self.iterator)
-        except StopIteration:
-            return b""
-
-        # Return an arbitrary number or bytes
-        if n <= 0:
-            self.leftover = b""
-            return chunk
-
-        output, self.leftover = chunk[:n], chunk[n:]
-        return output
-
-    def peek(self, n):
-        while len(self.leftover) < n:
-            try:
-                self.leftover += next(self.iterator)
-            except StopIteration:
-                break
-        return self.leftover[:n]
-
-
 class HashedStreamReader(io.IOBase):
 
     PARAM_CHECKSUM = "md5"

dvc/utils/strictyaml.py

@@ -0,0 +1,288 @@
+"""
+This module combines schema and yaml parser into one, to provide better error
+messages through a single entrypoint `load`.
+
+Used for parsing dvc.yaml, dvc.lock and .dvc files.
+
+Not to be confused with strictyaml, a python library with similar motivations.
+"""
+import re
+import typing
+from contextlib import suppress
+from typing import TYPE_CHECKING, Any, Callable, List, Optional, Tuple, TypeVar
+
+from dvc.exceptions import PrettyDvcException
+from dvc.ui import ui
+from dvc.utils.serialize import (
+    EncodingError,
+    YAMLFileCorruptedError,
+    parse_yaml,
+    parse_yaml_for_update,
+)
+
+if TYPE_CHECKING:
+    from rich.syntax import Syntax
+    from ruamel.yaml import StreamMark
+    from voluptuous import MultipleInvalid
+
+    from dvc.fs.base import BaseFileSystem
+    from dvc.ui import RichText
+
+
+_T = TypeVar("_T")
+merge_conflict_marker = re.compile("^([<=>]{7}) .*$", re.MULTILINE)
+
+
+def make_relpath(path: str) -> str:
+    import os
+
+    from dvc.utils import relpath
+
+    rel = relpath(path)
+    prefix = ""
+    if not rel.startswith(".."):
+        prefix = "./" if os.name == "posix" else ".\\"
+    return prefix + rel
+
+
+def _prepare_message(message: str) -> "RichText":
+    return ui.rich_text(message, style="red")
+
+
+def _prepare_cause(cause: str) -> "RichText":
+    return ui.rich_text(cause, style="bold")
+
+
+def _prepare_code_snippets(
+    code: str, start_line: int = 1, **kwargs: Any
+) -> "Syntax":
+    from rich.syntax import Syntax
+
+    kwargs.setdefault("start_line", start_line)
+    return Syntax(
+        code,
+        "yaml",
+        theme="ansi_dark",
+        word_wrap=True,
+        line_numbers=True,
+        indent_guides=True,
+        **kwargs,
+    )
+
+
+class YAMLSyntaxError(PrettyDvcException, YAMLFileCorruptedError):
+    def __init__(self, path: str, yaml_text: str, exc: Exception) -> None:
+        self.path: str = path
+        self.yaml_text: str = yaml_text
+        self.exc: Exception = exc
+
+        merge_conflicts = merge_conflict_marker.search(self.yaml_text)
+        self.hint = " (possible merge conflicts)" if merge_conflicts else ""
+        super().__init__(self.path)
+
+    def __pretty_exc__(self, **kwargs: Any) -> None:
+        from ruamel.yaml.error import MarkedYAMLError
+
+        exc = self.exc.__cause__
+
+        if not isinstance(exc, MarkedYAMLError):
+            raise ValueError("nothing to pretty-print here. :)")
+
+        source = self.yaml_text.splitlines()
+
+        def prepare_linecol(mark: "StreamMark") -> str:
+            return f"in line {mark.line + 1}, column {mark.column + 1}"
+
+        def prepare_message(
+            message: str, mark: "StreamMark" = None
+        ) -> "RichText":
+            cause = ", ".join(
+                [message.capitalize(), prepare_linecol(mark) if mark else ""]
+            )
+            return _prepare_cause(cause)
+
+        def prepare_code(mark: "StreamMark") -> "Syntax":
+            line = mark.line + 1
+            code = "" if line > len(source) else source[line - 1]
+            return _prepare_code_snippets(code, line)
+
+        lines: List[object] = []
+        if hasattr(exc, "context"):
+            if exc.context_mark is not None:
+                lines.append(
+                    prepare_message(str(exc.context), exc.context_mark)
+                )
+            if exc.context_mark is not None and (
+                exc.problem is None
+                or exc.problem_mark is None
+                or exc.context_mark.name != exc.problem_mark.name
+                or exc.context_mark.line != exc.problem_mark.line
+                or exc.context_mark.column != exc.problem_mark.column
+            ):
+                lines.extend([prepare_code(exc.context_mark), ""])
+            if exc.problem is not None:
+                lines.append(
+                    prepare_message(str(exc.problem), exc.problem_mark)
+                )
+            if exc.problem_mark is not None:
+                lines.append(prepare_code(exc.problem_mark))
+
+        if lines:
+            # we should not add a newline after the main message
+            # if there are no other outputs
+            lines.insert(0, "")
+
+        rel = make_relpath(self.path)
+        lines.insert(0, _prepare_message(f"'{rel}' is invalid{self.hint}."))
+        for line in lines:
+            ui.error_write(line, styled=True)
+
+
+def determine_linecol(
+    data, paths, max_steps=5
+) -> Tuple[Optional[int], Optional[int], int]:
+    """Determine linecol from the CommentedMap for the `paths` location.
+
+    CommentedMap from `ruamel.yaml` has `.lc` property from which we can read
+    `.line` and `.col`. This is available in the collections type,
+    i.e. list and dictionaries.
+
+    But this may fail on non-collection types. For example, if the `paths` is
+    ['stages', 'metrics'], metrics being a boolean type does not have `lc`
+    prop.
+    ```
+    stages:
+      metrics: true
+    ```
+
+    To provide some context to the user, we step up to the
+    path ['stages'], which being a collection type, will have `lc` prop
+    with which we can find line and col.
+
+    This may end up being not accurate, so we try to show the same amount of
+    lines of code for `n` number of steps taken upwards. In a worst case,
+    it may be just 1 step (as non-collection item cannot have child items),
+    but `schema validator` may provide us arbitrary path. So, this caps the
+    number of steps upward to just 5. If it does not find any linecols, it'll
+    abort.
+    """
+    from dpath.util import get
+
+    step = 1
+    line, col = None, None
+    while paths and step < max_steps:
+        value = get(data, paths, default=None)
+        if value is not None:
+            with suppress(AttributeError, TypeError):
+                line = value.lc.line + 1
+                col = value.lc.col + 1
+                break
+        step += 1
+        *paths, _ = paths
+
+    return line, col, step
+
+
+class YAMLValidationError(PrettyDvcException):
+    def __init__(
+        self,
+        exc: "MultipleInvalid",
+        path: str = None,
+        text: str = None,
+    ) -> None:
+        self.text = text or ""
+        self.exc = exc
+
+        rel = make_relpath(path) if path else ""
+        self.path = path or ""
+
+        message = f"'{rel}' validation failed"
+        if len(self.exc.errors) > 1:
+            message += f": {len(self.exc.errors)} errors"
+        super().__init__(f"{message}")
+
+    def _prepare_context(self, data: typing.Mapping) -> List[object]:
+        lines: List[object] = []
+        for index, error in enumerate(self.exc.errors):
+            if index and lines[-1]:
+                lines.append("")
+            line, col, step = determine_linecol(data, error.path)
+            parts = [error.error_message]
+            if error.path:
+                parts.append("in " + " -> ".join(str(p) for p in error.path))
+            if line:
+                parts.append(f"line {line}")
+            if col:
+                parts.append(f"column {col}")
+            lines.append(_prepare_cause(", ".join(parts)))
+
+            if line:
+                # we show one line above the error
+                # we try to show few more lines if we could not
+                # reliably figure out where the error was
+                lr = (line - 1, line + step - 1)
+                code = _prepare_code_snippets(self.text, line_range=lr)
+                lines.append(code)
+        return lines
+
+    def __pretty_exc__(self, **kwargs: Any) -> None:
+        """Prettify exception message."""
+        from collections.abc import Mapping
+
+        lines: List[object] = []
+        data = parse_yaml_for_update(self.text, self.path)
+        if isinstance(data, Mapping):
+            lines.extend(self._prepare_context(data))
+
+        cause = ""
+        if lines:
+            # we should not add a newline after the main message
+            # if there are no other outputs
+            lines.insert(0, "")
+        else:
+            # if we don't have any context to show, we'll fallback to what we
+            # got from voluptuous and print them in the same line.
+            cause = f": {self.exc}"
+
+        lines.insert(0, _prepare_message(f"{self}{cause}."))
+        for line in lines:
+            ui.error_write(line, styled=True)
+
+
+def validate(
+    data: _T,
+    schema: Callable[[_T], _T],
+    text: str = None,
+    path: str = None,
+) -> _T:
+    from voluptuous import MultipleInvalid
+
+    try:
+        return schema(data)
+    except MultipleInvalid as exc:
+        raise YAMLValidationError(exc, path, text) from exc
+
+
+def load(
+    path: str,
+    schema: Callable[[_T], _T] = None,
+    fs: "BaseFileSystem" = None,
+    encoding: str = "utf-8",
+    round_trip: bool = False,
+) -> Any:
+    open_fn = fs.open if fs else open
+    try:
+        with open_fn(path, encoding=encoding) as fd:  # type: ignore
+            text = fd.read()
+        data = parse_yaml(text, path, typ="rt" if round_trip else "safe")
+    except UnicodeDecodeError as exc:
+        raise EncodingError(path, encoding) from exc
+    except YAMLFileCorruptedError as exc:
+        cause = exc.__cause__
+        raise YAMLSyntaxError(path, text, exc) from cause
+
+    if schema:
+        # not returning validated data, as it may remove
+        # details from CommentedMap that we get from roundtrip parser
+        validate(data, schema, text=text, path=path)
+    return data, text

dvc/utils/threadpool.py

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

dvc/version.py

@@ -6,7 +6,7 @@
 import os
 import subprocess
 
-_BASE_VERSION = "2.7.2"
+_BASE_VERSION = "2.8.1"
 
 
 def _generate_version(base_version):

scripts/pyinstaller/hooks/hook-dvc.py

@@ -7,7 +7,6 @@ datas = copy_metadata("adlfs", recursive=True)
 datas += copy_metadata("knack")
 datas += copy_metadata("gcsfs")
 datas += copy_metadata("pyarrow")
-datas += copy_metadata("hdfs")
 datas += copy_metadata("pydrive2")
 datas += copy_metadata("s3fs", recursive=True)
 datas += copy_metadata("boto3")

scripts/pyinstaller/hooks/hook-win32file.py

@@ -1 +0,0 @@
-hiddenimports = ["win32timezone"]

setup.cfg

@@ -1,3 +1,171 @@
+[metadata]
+description = Git for data scientists - manage your code and data together
+name = dvc
+long_description = file: README.rst
+long_description_content_type = text/x-rst
+license = Apache License 2.0
+license_file = LICENSE
+url = http://dvc.org
+project_urls =
+    Documentation = https://dvc.org/doc
+    Source = https://github.com/iterative/dvc
+download_url = https://github.com/iterative/dvc
+author = Dmitry Petrov
+author_email = [email protected]
+maintainer = Iterative
+maintainer_email = [email protected]
+keywords = data-science, data-version-control, machine-learning, git
+    developer-tools, reproducibility, collaboration, ai
+classifiers =
+    Development Status :: 4 - Beta
+    Programming Language :: Python :: 3
+    Programming Language :: Python :: 3.6
+    Programming Language :: Python :: 3.7
+    Programming Language :: Python :: 3.8
+    Programming Language :: Python :: 3.9
+    Programming Language :: Python :: 3.10
+
+[options]
+python_requires = >=3.6
+zip_safe = False
+packages = find:
+include_package_data = True
+install_requires =
+    # See https://github.com/pyinstaller/pyinstaller/issues/1945
+    ply>=3.9
+    colorama>=0.3.9
+    configobj>=5.0.6
+    gitpython>3
+    dulwich>=0.20.23
+    pygit2>=1.5.0,<1.7; python_version < '3.7'
+    pygit2>=1.5.0; python_version >= '3.7'
+    setuptools>=34.0.0
+    nanotime>=0.5.2
+    pyasn1>=0.4.1
+    voluptuous>=0.11.7
+    requests>=2.22.0
+    # See https://github.com/bdcht/grandalf/issues/26
+    grandalf==0.6
+    distro>=1.3.0
+    appdirs>=1.4.3
+    ruamel.yaml>=0.17.11
+    toml>=0.10.1
+    funcy>=1.14
+    pathspec>=0.6.0,<0.9.0
+    shortuuid>=0.5.0
+    tqdm>=4.45.0,<5
+    packaging>=19.0
+    zc.lockfile>=1.2.1
+    flufl.lock>=3.2,<4
+    win-unicode-console>=0.5; sys_platform == 'win32'
+    networkx>=2.5
+    psutil>=5.8.0
+    pydot>=1.2.4
+    dataclasses>=0.7; python_version < '3.7'
+    contextvars>=2.1; python_version < '3.7'
+    importlib-metadata>=1.4; python_version < '3.8'
+    flatten_dict>=0.4.1,<1
+    tabulate>=0.8.7
+    pygtrie>=2.3.2
+    dpath>=2.0.2,<3
+    shtab>=1.3.4,<2
+    rich>=10.9.0
+    dictdiffer>=0.8.1
+    python-benedict>=0.24.2
+    pyparsing==2.4.7
+    typing_extensions>=3.7.4
+    fsspec[http]>=2021.10.0
+    aiohttp-retry>=2.4.5
+    diskcache>=5.2.1
+
+[options.extras_require]
+# gssapi should not be included in all_remotes, because it doesn't have wheels
+# for linux and mac, so it will fail to compile if user doesn't have all the
+# requirements, including kerberos itself. Once all the wheels are available,
+# we can start shipping it by default.
+all =
+    %(azure)s
+    %(gdrive)s
+    %(gs)s
+    %(hdfs)s
+    %(oss)s
+    %(s3)s
+    %(ssh)s
+    %(webdav)s
+dev =
+    %(all)s
+    %(tests)s
+# Extra dependencies for remote integrations
+azure =
+    adlfs==2021.9.1
+    azure-identity>=1.4.0
+    knack
+gdrive = pydrive2[fsspec]>=1.9.4
+gs = gcsfs==2021.10.0
+hdfs =
+    fsspec[arrow]; python_version < '3.10'
+oss = ossfs==2021.8.0
+s3 =
+    s3fs==2021.10.0
+    aiobotocore[boto3]>1.0.1
+ssh = sshfs[bcrypt]>=2021.8.1
+ssh_gssapi = sshfs[gssapi]>=2021.8.1
+webdav = webdav4>=0.9.1
+# not to break `dvc[webhdfs]`
+webhdfs =
+terraform = tpi[ssh]>=0.0.0
+tests =
+    %(terraform)s
+    wheel==0.37.0
+    # Test requirements
+    pytest==6.2.5
+    pytest-cov==3.0.0
+    pytest-xdist==2.4.0
+    pytest-mock==3.6.1
+    pytest-lazy-fixture==0.6.3
+    pytest-docker==0.10.3
+    flaky==3.7.0
+    mock==4.0.3
+    rangehttpserver==1.2.0
+    mock-ssh-server==0.9.1
+    wget==3.2
+    filelock==3.3.0
+    wsgidav==3.1.1
+    crc32c==2.2.post0
+    xmltodict==0.12.0
+    google-compute-engine==2.8.13
+    google-cloud-storage==1.42.3
+    dvclive==0.4.0
+    pywin32>=225; sys_platform == 'win32' and python_version < '3.10'
+    hdfs==2.6.0
+    # required by collective.checkdocs
+    Pygments==2.10.0
+    collective.checkdocs==0.2
+    pydocstyle==6.1.1
+    jaraco.windows==5.7.0
+    # pylint requirements
+    pylint==2.9.6
+    # we use this to suppress pytest-related false positives in our tests.
+    pylint-pytest==1.0.3
+    # we use this to suppress some messages in tests, eg: foo/bar naming,
+    # and, protected method calls in our tests
+    pylint-plugin-utils==0.6
+    # type-checking
+    mypy==0.910
+    types-requests==2.25.9
+    types-paramiko==2.7.1
+    types-tabulate==0.8.2
+    types-toml==0.10.0
+
+[options.packages.find]
+exclude =
+    tests
+    tests.*
+
+[options.entry_points]
+console_scripts =
+    dvc = dvc.main:main
+
 [flake8]
 ignore=
     E203, # Whitespace before ':'
@@ -17,8 +185,6 @@ known_first_party=dvc,tests
 line_length=79
 
 [tool:pytest]
-timeout = 200
-timeout_method = thread
 log_level = debug
 addopts = -ra
 markers =

setup.py

@@ -1,8 +1,9 @@
 import importlib.util
 import os
 from pathlib import Path
+from textwrap import dedent
 
-from setuptools import find_packages, setup
+from setuptools import setup
 from setuptools.command.build_py import build_py as _build_py
 
 # Prevents pkg_resources import in entry point script,
@@ -16,12 +17,23 @@ except ImportError:
 
 # Read package meta-data from version.py
 # see https://packaging.python.org/guides/single-sourcing-package-version/
-pkg_dir = os.path.dirname(os.path.abspath(__file__))
-version_path = os.path.join(pkg_dir, "dvc", "version.py")
-spec = importlib.util.spec_from_file_location("dvc.version", version_path)
-dvc_version = importlib.util.module_from_spec(spec)
-spec.loader.exec_module(dvc_version)
-version = dvc_version.__version__  # noqa: F821
+try:
+    pkg_dir = os.path.dirname(os.path.abspath(__file__))
+    version_path = os.path.join(pkg_dir, "dvc", "version.py")
+    spec = importlib.util.spec_from_file_location("dvc.version", version_path)
+    dvc_version = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(dvc_version)
+    version = dvc_version.__version__  # noqa: F821
+except Exception as exc:  # pylint: disable=broad-except
+    # Dependabot seem to stop working when we don't handle this except block.
+    # Most likely, it's because of the restrictions in execution of the module.
+    # We workaround it, and print the error message if it also happens on other
+    # installations (though this message may likely be suppressed by build
+    # tools, eg: pip only shows this message on `--verbose` mode).
+    import sys
+
+    print("Could not load version info: ", exc, file=sys.stderr)
+    version = "UNKNOWN"
 
 
 # To achieve consistency between the build version and the one provided
@@ -35,134 +47,20 @@ version = dvc_version.__version__  # noqa: F821
 #   - https://github.com/python/mypy
 class build_py(_build_py):
     def pin_version(self):
-        path = os.path.join(self.build_lib, "dvc")
-        self.mkpath(path)
-        with open(os.path.join(path, "version.py"), "w") as fobj:
-            fobj.write("# AUTOGENERATED at build time by setup.py\n")
-            fobj.write(f'__version__ = "{version}"\n')
+        ver_file = Path(self.build_lib) / "dvc" / "version.py"
+        self.mkpath(os.fspath(ver_file.parent))
+        ver_file.write_text(
+            dedent(
+                f"""\
+                # AUTOGENERATED at build time by setup.py
+                __version__ = "{version}"
+                """
+            )
+        )
 
     def run(self):
         self.execute(self.pin_version, ())
         _build_py.run(self)
 
 
-install_requires = [
-    "ply>=3.9",  # See https://github.com/pyinstaller/pyinstaller/issues/1945
-    "colorama>=0.3.9",
-    "configobj>=5.0.6",
-    "gitpython>3",
-    "dulwich>=0.20.23",
-    "pygit2>=1.5.0",
-    "setuptools>=34.0.0",
-    "nanotime>=0.5.2",
-    "pyasn1>=0.4.1",
-    "voluptuous>=0.11.7",
-    "requests>=2.22.0",
-    "grandalf==0.6",
-    "distro>=1.3.0",
-    "appdirs>=1.4.3",
-    "ruamel.yaml>=0.17.11",
-    "toml>=0.10.1",
-    "funcy>=1.14",
-    "pathspec>=0.6.0,<0.9.0",
-    "shortuuid>=0.5.0",
-    "tqdm>=4.45.0,<5",
-    "packaging>=19.0",
-    "zc.lockfile>=1.2.1",
-    "flufl.lock>=3.2,<4",
-    "win-unicode-console>=0.5; sys_platform == 'win32'",
-    "pywin32>=225; sys_platform == 'win32' and python_version < '3.10'",
-    "networkx>=2.5",
-    "psutil>=5.8.0",
-    "pydot>=1.2.4",
-    "speedcopy>=2.0.1; python_version < '3.8' and sys_platform == 'win32'",
-    "dataclasses==0.7; python_version < '3.7'",
-    "importlib-metadata>=1.4; python_version < '3.8'",
-    "flatten_dict>=0.4.1,<1",
-    "tabulate>=0.8.7",
-    "pygtrie>=2.3.2",
-    "dpath>=2.0.1,<3",
-    "shtab>=1.3.4,<2",
-    "rich>=10.0.0",
-    "dictdiffer>=0.8.1",
-    "python-benedict>=0.21.1",
-    "pyparsing==2.4.7",
-    "typing_extensions>=3.7.4; python_version < '3.10'",
-    # until https://github.com/python/typing/issues/865 is fixed for python3.10
-    "typing_extensions==3.10.0.0; python_version >= '3.10'",
-    "fsspec[http]>=2021.8.1",
-    "aiohttp-retry==2.4.5",
-    "diskcache>=5.2.1",
-]
-
-
-# Extra dependencies for remote integrations
-
-gs = ["gcsfs==2021.8.1"]
-gdrive = ["pydrive2[fsspec]>=1.9.2"]
-s3 = ["s3fs==2021.8.1", "aiobotocore[boto3]>1.0.1"]
-azure = ["adlfs==2021.8.2", "azure-identity>=1.4.0", "knack"]
-oss = ["ossfs==2021.8.0"]
-ssh = ["sshfs>=2021.8.1"]
-
-hdfs = ["pyarrow>=2.0.0; python_version < '3.10'"]
-webhdfs = ["hdfs==2.5.8"]
-webdav = ["webdav4>=0.9.1"]
-# gssapi should not be included in all_remotes, because it doesn't have wheels
-# for linux and mac, so it will fail to compile if user doesn't have all the
-# requirements, including kerberos itself. Once all the wheels are available,
-# we can start shipping it by default.
-ssh_gssapi = ["sshfs[gssapi]>=2021.8.1"]
-all_remotes = gs + s3 + azure + ssh + oss + gdrive + hdfs + webhdfs + webdav
-
-terraform = ["tpi[ssh]>=0.0.0"]
-
-tests_requirements = (
-    Path("test_requirements.txt").read_text().strip().splitlines()
-) + terraform
-
-setup(
-    name="dvc",
-    version=version,
-    description="Git for data scientists - manage your code and data together",
-    long_description=open("README.rst", encoding="UTF-8").read(),
-    author="Dmitry Petrov",
-    author_email="[email protected]",
-    maintainer="Iterative",
-    maintainer_email="[email protected]",
-    download_url="https://github.com/iterative/dvc",
-    license="Apache License 2.0",
-    install_requires=install_requires,
-    extras_require={
-        "all": all_remotes,
-        "gs": gs,
-        "gdrive": gdrive,
-        "s3": s3,
-        "azure": azure,
-        "oss": oss,
-        "ssh": ssh,
-        "ssh_gssapi": ssh_gssapi,
-        "hdfs": hdfs,
-        "webhdfs": webhdfs,
-        "webdav": webdav,
-        "terraform": terraform,
-        "tests": tests_requirements,
-    },
-    keywords="data-science data-version-control machine-learning git"
-    " developer-tools reproducibility collaboration ai",
-    python_requires=">=3.6",
-    classifiers=[
-        "Development Status :: 4 - Beta",
-        "Programming Language :: Python :: 3",
-        "Programming Language :: Python :: 3.6",
-        "Programming Language :: Python :: 3.7",
-        "Programming Language :: Python :: 3.8",
-        "Programming Language :: Python :: 3.9",
-    ],
-    packages=find_packages(exclude=["tests"]),
-    include_package_data=True,
-    url="http://dvc.org",
-    entry_points={"console_scripts": ["dvc = dvc.main:main"]},
-    cmdclass={"build_py": build_py},
-    zip_safe=False,
-)
+setup(version=version, cmdclass={"build_py": build_py})

test_requirements.txt

@@ -1,46 +0,0 @@
-wheel==0.37.0
-
-# Test requirements
-pytest==6.2.5
-pytest-cov==2.12.1
-pytest-xdist==2.3.0
-pytest-mock==3.6.1
-pytest-lazy-fixture==0.6.3
-pytest-timeout==1.4.2
-pytest-docker==0.10.3
-
-flaky==3.7.0
-mock==4.0.3
-rangehttpserver==1.2.0
-mock-ssh-server==0.9.0
-beautifulsoup4==4.9.3
-wget==3.2
-filelock==3.0.12
-wsgidav==3.1.1
-crc32c==2.2.post0
-xmltodict==0.12.0
-google-compute-engine==2.8.13
-google-cloud-storage==1.42.0
-urllib3==1.26.6
-dvclive>=0.2.1
-
-# required by collective.checkdocs
-Pygments==2.10.0
-collective.checkdocs==0.2
-pydocstyle==6.1.1
-jaraco.windows==5.6.0
-
-# pylint requirements
-pylint==2.9.6
-# we use this to suppress pytest-related false positives in our tests.
-pylint-pytest==1.0.3
-# we use this to suppress some messages in tests, eg: foo/bar naming,
-# and, protected method calls in our tests
-pylint-plugin-utils==0.6
-
-# type-checking
-mypy==0.910
-types-requests
-types-paramiko
-types-tabulate
-types-toml

tests/conftest.py

@@ -5,6 +5,7 @@ import pytest
 
 from .dir_helpers import *  # noqa, pylint: disable=wildcard-import
 from .remotes import *  # noqa, pylint: disable=wildcard-import
+from .utils.scriptify import scriptify
 
 # Prevent updater and analytics from running their processes
 os.environ["DVC_TEST"] = "true"
@@ -18,11 +19,12 @@ REMOTES = {
     "azure": False,
     "gdrive": False,
     "gs": False,
-    "hdfs": False,
+    "hdfs": True,
+    "real_hdfs": False,
     "http": True,
     "oss": False,
     "s3": False,
-    "ssh": True,
+    "ssh": False,
     "webdav": True,
 }
 
@@ -164,3 +166,6 @@ def custom_template(tmp_dir, dvc):
     template = tmp_dir / "custom_template.json"
     shutil.copy(tmp_dir / ".dvc" / "plots" / "default.json", template)
     return template
+
+
+scriptify_fixture = pytest.fixture(lambda: scriptify, name="scriptify")

tests/dir_helpers.py

@@ -49,12 +49,14 @@ import os
 import pathlib
 import sys
 from contextlib import contextmanager
+from functools import partialmethod
 from textwrap import dedent
 
 import pytest
 from funcy import lmap, retry
 
 from dvc.logger import disable_other_loggers
+from dvc.utils import serialize
 from dvc.utils.fs import makedirs
 
 __all__ = [
@@ -250,6 +252,22 @@ class TmpDir(pathlib.Path):
     def hash_to_path_info(self, hash_):
         return self / hash_[0:2] / hash_[2:]
 
+    def dump(self, *args, **kwargs):
+        return serialize.DUMPERS[self.suffix](self, *args, **kwargs)
+
+    def parse(self, *args, **kwargs):
+        return serialize.LOADERS[self.suffix](self, *args, **kwargs)
+
+    def modify(self, *args, **kwargs):
+        return serialize.MODIFIERS[self.suffix](self, *args, **kwargs)
+
+    load_yaml = partialmethod(serialize.load_yaml)
+    dump_yaml = partialmethod(serialize.dump_yaml)
+    load_json = partialmethod(serialize.load_json)
+    dump_json = partialmethod(serialize.dump_json)
+    load_toml = partialmethod(serialize.load_toml)
+    dump_toml = partialmethod(serialize.dump_toml)
+
 
 def _coerce_filenames(filenames):
     if isinstance(filenames, (str, bytes, pathlib.PurePath)):

tests/docker-compose.yml

@@ -2,14 +2,16 @@
 version: '3.2'
 services:
   azurite:
-    image: mcr.microsoft.com/azure-storage/azurite:3.13.1
+    image: mcr.microsoft.com/azure-storage/azurite:3.14.2
     command: azurite -L -l /data --blobHost 0.0.0.0 --queueHost 0.0.0.0
     ports:
       - "10000"
+
   motoserver:
     image: motoserver/moto
     ports:
       - "5000"
+
   oss:
     image: rkuprieiev/oss-emulator
     ports:
@@ -25,3 +27,13 @@ services:
       - "50010:50010"
       - "50075:50075"
       - "50070"
+
+  openssh-server:
+    image: ghcr.io/linuxserver/openssh-server
+    environment:
+      - USER_NAME=user
+      - PUBLIC_KEY_FILE=/tmp/key
+    ports:
+      - 2222
+    volumes:
+      - ./remotes/user.key.pub:/tmp/key

tests/func/experiments/conftest.py

@@ -88,3 +88,17 @@ def checkpoint_stage(tmp_dir, scm, dvc, mocker):
     scm.commit("init")
     stage.iterations = DEFAULT_ITERATIONS
     return stage
+
+
+@pytest.fixture
+def http_auth_patch(mocker):
+    from dulwich.client import HTTPUnauthorized
+
+    url = "https://0.0.0.0"
+    client = mocker.MagicMock()
+    client.get_refs.side_effect = HTTPUnauthorized("", url)
+    client.send_pack.side_effect = HTTPUnauthorized("", url)
+
+    patch = mocker.patch("dulwich.client.get_transport_and_path")
+    patch.return_value = (client, url)
+    return url

tests/func/experiments/test_experiments.py

@@ -8,7 +8,7 @@ from funcy import first
 
 from dvc.dvcfile import PIPELINE_FILE
 from dvc.repo.experiments.utils import exp_refs_by_rev
-from dvc.utils.serialize import PythonFileCorruptedError, load_yaml
+from dvc.utils.serialize import PythonFileCorruptedError
 from tests.func.test_repro_multistage import COPY_SCRIPT
 
 
@@ -660,9 +660,9 @@ def test_modified_data_dep(tmp_dir, scm, dvc, workspace, params, target):
 
 
 def test_exp_run_recursive(tmp_dir, scm, dvc, run_copy_metrics):
-    tmp_dir.dvc_gen("metric_t.json", "foo: 1")
+    tmp_dir.dvc_gen("metric_t.json", '{"foo": 1}')
     run_copy_metrics(
         "metric_t.json", "metric.json", metrics=["metric.json"], no_exec=True
     )
     assert dvc.experiments.run(".", recursive=True)
-    assert load_yaml(tmp_dir / "metric.json") == {"foo": 1}
+    assert (tmp_dir / "metric.json").parse() == {"foo": 1}

tests/func/experiments/test_init.py

@@ -0,0 +1,33 @@
+import os
+
+from dvc.command.experiments import CmdExperimentsInit
+from dvc.main import main
+from dvc.utils.serialize import load_yaml
+
+
+def test_init(tmp_dir, dvc):
+    tmp_dir.gen(
+        {
+            CmdExperimentsInit.CODE: {"copy.py": ""},
+            "data": "data",
+            "params.yaml": '{"foo": 1}',
+            "dvclive": {},
+            "plots": {},
+        }
+    )
+    code_path = os.path.join(CmdExperimentsInit.CODE, "copy.py")
+    script = f"python {code_path}"
+
+    assert main(["exp", "init", script]) == 0
+    assert load_yaml(tmp_dir / "dvc.yaml") == {
+        "stages": {
+            "default": {
+                "cmd": script,
+                "deps": ["data", "src"],
+                "metrics": [{"metrics.json": {"cache": False}}],
+                "outs": ["models"],
+                "params": ["foo"],
+                "plots": [{"plots": {"cache": False}}],
+            }
+        }
+    }

tests/func/experiments/test_remote.py

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

tests/func/experiments/test_show.py

@@ -11,7 +11,7 @@ from dvc.repo.experiments.base import EXPS_STASH, ExpRefInfo
 from dvc.repo.experiments.executor.base import BaseExecutor, ExecutorInfo
 from dvc.repo.experiments.utils import exp_refs_by_rev
 from dvc.utils.fs import makedirs
-from dvc.utils.serialize import YAMLFileCorruptedError, dump_yaml
+from dvc.utils.serialize import YAMLFileCorruptedError
 from tests.func.test_repro_multistage import COPY_SCRIPT
 
 
@@ -286,7 +286,7 @@ def test_show_filter(
         "train/bar": 1,
         "nested": {"foo": 1, "bar": 1},
     }
-    dump_yaml(params_file, params_data)
+    (tmp_dir / params_file).dump(params_data)
 
     dvc.run(
         cmd="python copy.py params.yaml metrics.yaml",
@@ -386,7 +386,7 @@ def test_show_running_workspace(tmp_dir, scm, dvc, exp_stage, capsys):
     makedirs(pid_dir, True)
     info = ExecutorInfo(None, None, None, BaseExecutor.DEFAULT_LOCATION)
     pidfile = os.path.join(pid_dir, f"workspace{BaseExecutor.PIDFILE_EXT}")
-    dump_yaml(pidfile, info.to_dict())
+    (tmp_dir / pidfile).dump(info.to_dict())
 
     assert dvc.experiments.show()["workspace"] == {
         "baseline": {
@@ -417,7 +417,7 @@ def test_show_running_executor(tmp_dir, scm, dvc, exp_stage):
     makedirs(pid_dir, True)
     info = ExecutorInfo(None, None, None, BaseExecutor.DEFAULT_LOCATION)
     pidfile = os.path.join(pid_dir, f"{exp_rev}{BaseExecutor.PIDFILE_EXT}")
-    dump_yaml(pidfile, info.to_dict())
+    (tmp_dir / pidfile).dump(info.to_dict())
 
     results = dvc.experiments.show()
     exp_data = get_in(results, [baseline_rev, exp_rev, "data"])
@@ -455,7 +455,7 @@ def test_show_running_checkpoint(
     info = ExecutorInfo(123, "foo.git", baseline_rev, executor)
     rev = "workspace" if workspace else stash_rev
     pidfile = os.path.join(pid_dir, f"{rev}{BaseExecutor.PIDFILE_EXT}")
-    dump_yaml(pidfile, info.to_dict())
+    (tmp_dir / pidfile).dump(info.to_dict())
 
     mocker.patch.object(
         BaseExecutor, "fetch_exps", return_value=[str(exp_ref)]
@@ -495,6 +495,8 @@ def test_show_with_broken_repo(tmp_dir, scm, dvc, exp_stage, caplog):
 
 
 def test_show_csv(tmp_dir, scm, dvc, exp_stage, capsys):
+    import time
+
     baseline_rev = scm.get_rev()
 
     def _get_rev_isotimestamp(rev):
@@ -505,12 +507,13 @@ def test_show_csv(tmp_dir, scm, dvc, exp_stage, capsys):
     result1 = dvc.experiments.run(exp_stage.addressing, params=["foo=2"])
     rev1 = first(result1)
     ref_info1 = first(exp_refs_by_rev(scm, rev1))
+    time.sleep(1)
     result2 = dvc.experiments.run(exp_stage.addressing, params=["foo=3"])
     rev2 = first(result2)
     ref_info2 = first(exp_refs_by_rev(scm, rev2))
 
     capsys.readouterr()
-    assert main(["exp", "show", "--show-csv"]) == 0
+    assert main(["exp", "show", "--csv"]) == 0
     cap = capsys.readouterr()
     assert (
         "Experiment,rev,typ,Created,parent,metrics.yaml:foo,params.yaml:foo"

tests/func/machine/test_machine_config.py

@@ -0,0 +1,123 @@
+import os
+import textwrap
+
+import pytest
+
+from dvc.main import main
+
+CONFIG_TEXT = textwrap.dedent(
+    """\
+        [feature]
+            machine = true
+        ['machine \"foo\"']
+            cloud = aws
+    """
+)
+
+
+@pytest.mark.parametrize(
+    "slot,value",
+    [
+        ("region", "us-west"),
+        ("image", "iterative-cml"),
+        ("name", "iterative_test"),
+        ("spot", "True"),
+        ("spot_price", "1.2345"),
+        ("spot_price", "12345"),
+        ("instance_hdd_size", "10"),
+        ("instance_type", "l"),
+        ("instance_gpu", "tesla"),
+        ("ssh_private", "secret"),
+    ],
+)
+def test_machine_modify_susccess(tmp_dir, dvc, slot, value):
+    (tmp_dir / ".dvc" / "config").write_text(CONFIG_TEXT)
+    assert main(["machine", "modify", "foo", slot, value]) == 0
+    assert (
+        tmp_dir / ".dvc" / "config"
+    ).read_text() == CONFIG_TEXT + f"    {slot} = {value}\n"
+    assert main(["machine", "modify", "--unset", "foo", slot]) == 0
+    assert (tmp_dir / ".dvc" / "config").read_text() == CONFIG_TEXT
+
+
+def test_machine_modify_startup_script(tmp_dir, dvc):
+    slot, value = "startup_script", "start.sh"
+    (tmp_dir / ".dvc" / "config").write_text(CONFIG_TEXT)
+    assert main(["machine", "modify", "foo", slot, value]) == 0
+    assert (
+        tmp_dir / ".dvc" / "config"
+    ).read_text() == CONFIG_TEXT + f"    {slot} = ../{value}\n"
+    assert main(["machine", "modify", "--unset", "foo", slot]) == 0
+    assert (tmp_dir / ".dvc" / "config").read_text() == CONFIG_TEXT
+
+
+@pytest.mark.parametrize(
+    "slot,value,msg",
+    [
+        (
+            "region",
+            "other-west",
+            "expected one of us-west, us-east, eu-west, eu-north",
+        ),
+        ("spot_price", "NUM", "expected float"),
+        ("instance_hdd_size", "BIG", "expected int"),
+    ],
+)
+def test_machine_modify_fail(tmp_dir, dvc, caplog, slot, value, msg):
+    (tmp_dir / ".dvc" / "config").write_text(CONFIG_TEXT)
+
+    assert main(["machine", "modify", "foo", slot, value]) == 251
+    assert (tmp_dir / ".dvc" / "config").read_text() == CONFIG_TEXT
+    assert msg in caplog.text
+
+
+FULL_CONFIG_TEXT = textwrap.dedent(
+    """\
+        [feature]
+            machine = true
+        ['machine \"bar\"']
+            cloud = azure
+        ['machine \"foo\"']
+            cloud = aws
+            region = us-west
+            image = iterative-cml
+            name = iterative_test
+            spot = True
+            spot_price = 1.2345
+            instance_hdd_size = 10
+            instance_type = l
+            instance_gpu = tesla
+            ssh_private = secret
+            startup_script = {}
+    """.format(
+        os.path.join("..", "start.sh")
+    )
+)
+
+
+def test_machine_list(tmp_dir, dvc, capsys):
+    (tmp_dir / ".dvc" / "config").write_text(FULL_CONFIG_TEXT)
+
+    assert main(["machine", "list"]) == 0
+    cap = capsys.readouterr()
+    assert "cloud=azure" in cap.out
+
+    assert main(["machine", "list", "foo"]) == 0
+    cap = capsys.readouterr()
+    assert "cloud=azure" not in cap.out
+    assert "cloud=aws" in cap.out
+    assert "region=us-west" in cap.out
+    assert "image=iterative-cml" in cap.out
+    assert "name=iterative_test" in cap.out
+    assert "spot=True" in cap.out
+    assert "spot_price=1.2345" in cap.out
+    assert "instance_hdd_size=10" in cap.out
+    assert "instance_type=l" in cap.out
+    assert "instance_gpu=tesla" in cap.out
+    assert "ssh_private=secret" in cap.out
+    assert (
+        "startup_script={}".format(
+            os.path.join(tmp_dir, ".dvc", "..", "start.sh")
+        )
+        in cap.out
+    )

tests/func/metrics/test_diff.py

@@ -1,7 +1,6 @@
 import json
 
 from dvc.main import main
-from dvc.utils.serialize import dump_yaml
 
 
 def test_metrics_diff_simple(tmp_dir, scm, dvc, run_copy_metrics):
@@ -22,7 +21,7 @@ def test_metrics_diff_simple(tmp_dir, scm, dvc, run_copy_metrics):
 def test_metrics_diff_yaml(tmp_dir, scm, dvc, run_copy_metrics):
     def _gen(val):
         metrics = {"a": {"b": {"c": val, "d": 1, "e": str(val)}}}
-        dump_yaml("m_temp.yaml", metrics)
+        (tmp_dir / "m_temp.yaml").dump(metrics)
         run_copy_metrics(
             "m_temp.yaml", "m.yaml", metrics=["m.yaml"], commit=str(val)
         )
@@ -39,7 +38,7 @@ def test_metrics_diff_yaml(tmp_dir, scm, dvc, run_copy_metrics):
 def test_metrics_diff_json(tmp_dir, scm, dvc, run_copy_metrics):
     def _gen(val):
         metrics = {"a": {"b": {"c": val, "d": 1, "e": str(val)}}}
-        tmp_dir.gen({"m_temp.json": json.dumps(metrics)})
+        (tmp_dir / "m_temp.json").dump(metrics)
         run_copy_metrics(
             "m_temp.json", "m.json", metrics=["m.json"], commit=str(val)
         )
@@ -55,7 +54,7 @@ def test_metrics_diff_json(tmp_dir, scm, dvc, run_copy_metrics):
 def test_metrics_diff_json_unchanged(tmp_dir, scm, dvc, run_copy_metrics):
     def _gen(val):
         metrics = {"a": {"b": {"c": val, "d": 1, "e": str(val)}}}
-        tmp_dir.gen({"m_temp.json": json.dumps(metrics)})
+        (tmp_dir / "m_temp.json").dump(metrics)
         run_copy_metrics(
             "m_temp.json", "m.json", metrics=["m.json"], commit=str(val)
         )
@@ -69,7 +68,7 @@ def test_metrics_diff_json_unchanged(tmp_dir, scm, dvc, run_copy_metrics):
 
 def test_metrics_diff_broken_json(tmp_dir, scm, dvc, run_copy_metrics):
     metrics = {"a": {"b": {"c": 1, "d": 1, "e": "3"}}}
-    tmp_dir.gen({"m_temp.json": json.dumps(metrics)})
+    (tmp_dir / "m_temp.json").dump(metrics)
     run_copy_metrics(
         "m_temp.json",
         "m.json",
@@ -94,7 +93,7 @@ def test_metrics_diff_no_metrics(tmp_dir, scm, dvc):
 
 def test_metrics_diff_new_metric(tmp_dir, scm, dvc, run_copy_metrics):
     metrics = {"a": {"b": {"c": 1, "d": 1, "e": "3"}}}
-    tmp_dir.gen({"m_temp.json": json.dumps(metrics)})
+    (tmp_dir / "m_temp.json").dump(metrics)
     run_copy_metrics("m_temp.json", "m.json", metrics_no_cache=["m.json"])
 
     assert dvc.metrics.diff() == {
@@ -107,7 +106,7 @@ def test_metrics_diff_new_metric(tmp_dir, scm, dvc, run_copy_metrics):
 
 def test_metrics_diff_deleted_metric(tmp_dir, scm, dvc, run_copy_metrics):
     metrics = {"a": {"b": {"c": 1, "d": 1, "e": "3"}}}
-    tmp_dir.gen({"m_temp.json": json.dumps(metrics)})
+    (tmp_dir / "m_temp.json").dump(metrics)
     run_copy_metrics(
         "m_temp.json",
         "m.json",

tests/func/metrics/test_show.py

@@ -6,10 +6,9 @@ from funcy import get_in
 
 from dvc.dvcfile import PIPELINE_FILE
 from dvc.exceptions import OverlappingOutputPathsError
-from dvc.path_info import PathInfo
 from dvc.repo import Repo
 from dvc.utils.fs import remove
-from dvc.utils.serialize import YAMLFileCorruptedError, dump_yaml, modify_yaml
+from dvc.utils.serialize import YAMLFileCorruptedError
 
 
 def test_show_simple(tmp_dir, dvc, run_copy_metrics):
@@ -227,10 +226,10 @@ def test_show_no_metrics_files(tmp_dir, dvc, caplog):
 def test_metrics_show_overlap(
     tmp_dir, dvc, run_copy_metrics, clear_before_run
 ):
-    data_dir = PathInfo("data")
-    (tmp_dir / data_dir).mkdir()
+    data_dir = tmp_dir / "data"
+    data_dir.mkdir()
 
-    dump_yaml(data_dir / "m1_temp.yaml", {"a": {"b": {"c": 2, "d": 1}}})
+    (data_dir / "m1_temp.yaml").dump({"a": {"b": {"c": 2, "d": 1}}})
     run_copy_metrics(
         str(data_dir / "m1_temp.yaml"),
         str(data_dir / "m1.yaml"),
@@ -239,7 +238,7 @@ def test_metrics_show_overlap(
         name="cp-m1",
         metrics=[str(data_dir / "m1.yaml")],
     )
-    with modify_yaml("dvc.yaml") as d:
+    with (tmp_dir / "dvc.yaml").modify() as d:
         # trying to make an output overlaps error
         d["stages"]["corrupted-stage"] = {
             "cmd": "mkdir data",

tests/func/objects/db/test_index.py

@@ -76,7 +76,7 @@ def test_partial_upload(tmp_dir, dvc, index, mocker):
     tmp_dir.dvc_gen({"foo": "foo content"})
     tmp_dir.dvc_gen({"bar": {"baz": "baz content"}})
 
-    original = LocalFileSystem._upload
+    original = LocalFileSystem.upload
 
     def unreliable_upload(self, from_file, to_info, name=None, **kwargs):
         if "baz" in name: