dvc: bump to 2.7.2

* Show csv format of experiments

fix #5989 and #5446 

1. add --show-csv to dvc exp show
2. add a new functional test for csv format
3. add a new unit test for show_experiements
4. solve name duplication in table common headers (Created,Parent, etc... ), metrics files and param files.

Co-authored-by: Peter Rowlands (변기호) <[email protected]>

README.rst

@@ -222,7 +222,9 @@ Contributing
 |Maintainability| |Donate|
 
 Contributions are welcome! Please see our `Contributing Guide <https://dvc.org/doc/user-guide/contributing/core>`_ for more
-details.
+details. Thanks to all our contributors!
+
+|Contribs|
 
 Mailing List
 ============
@@ -299,3 +301,7 @@ Barrak, A., Eghan, E.E. and Adams, B. `On the Co-evolution of ML Pipelines and S
 .. |Flowchart| image:: https://dvc.org/img/flow.gif
    :target: https://dvc.org/img/flow.gif
    :alt: how_dvc_works
+
+.. |Contribs| image:: https://contrib.rocks/image?repo=iterative/dvc
+   :target: https://github.com/iterative/dvc/graphs/contributors
+   :alt: Contributors

dvc/command/experiments.py

@@ -141,6 +141,8 @@ def _collect_rows(
     precision=DEFAULT_PRECISION,
     sort_by=None,
     sort_order=None,
+    fill_value=FILL_VALUE,
+    iso=False,
 ):
     from dvc.scm.git import Git
 
@@ -161,8 +163,8 @@ def _collect_rows(
         elif exp.get("queued"):
             state = "Queued"
         else:
-            state = FILL_VALUE
-        executor = exp.get("executor", FILL_VALUE)
+            state = fill_value
+        executor = exp.get("executor", fill_value)
         is_baseline = rev == "baseline"
 
         if is_baseline:
@@ -202,12 +204,12 @@ def _collect_rows(
             exp_name,
             name_rev,
             typ,
-            _format_time(exp.get("timestamp")),
+            _format_time(exp.get("timestamp"), fill_value, iso),
             parent,
             state,
             executor,
         ]
-        fill_value = FILL_VALUE_ERRORED if results.get("error") else FILL_VALUE
+        fill_value = FILL_VALUE_ERRORED if results.get("error") else fill_value
         _extend_row(
             row,
             metric_names,
@@ -274,14 +276,18 @@ def _sort_exp(experiments, sort_path, sort_name, typ, reverse):
     return ret
 
 
-def _format_time(timestamp):
-    if timestamp is None:
-        return FILL_VALUE
-    if timestamp.date() == date.today():
+def _format_time(datetime_obj, fill_value=FILL_VALUE, iso=False):
+    if datetime_obj is None:
+        return fill_value
+
+    if iso:
+        return datetime_obj.isoformat()
+
+    if datetime_obj.date() == date.today():
         fmt = "%I:%M %p"
     else:
         fmt = "%b %d, %Y"
-    return timestamp.strftime(fmt)
+    return datetime_obj.strftime(fmt)
 
 
 def _extend_row(row, names, items, precision, fill_value=FILL_VALUE):
@@ -320,6 +326,7 @@ def _parse_filter_list(param_list):
 
 def experiments_table(
     all_experiments,
+    headers,
     metric_headers,
     metric_names,
     param_headers,
@@ -327,22 +334,15 @@ def experiments_table(
     sort_by=None,
     sort_order=None,
     precision=DEFAULT_PRECISION,
+    fill_value=FILL_VALUE,
+    iso=False,
 ) -> "TabularData":
     from funcy import lconcat
 
     from dvc.compare import TabularData
 
-    headers = [
-        "Experiment",
-        "rev",
-        "typ",
-        "Created",
-        "parent",
-        "State",
-        "Executor",
-    ]
     td = TabularData(
-        lconcat(headers, metric_headers, param_headers), fill_value=FILL_VALUE
+        lconcat(headers, metric_headers, param_headers), fill_value=fill_value
     )
     for base_rev, experiments in all_experiments.items():
         rows = _collect_rows(
@@ -353,6 +353,8 @@ def experiments_table(
             sort_by=sort_by,
             sort_order=sort_order,
             precision=precision,
+            fill_value=fill_value,
+            iso=iso,
         )
         td.extend(rows)
 
@@ -385,8 +387,10 @@ def baseline_styler(typ):
 
 
 def show_experiments(
-    all_experiments, pager=True, no_timestamp=False, **kwargs
+    all_experiments, pager=True, no_timestamp=False, show_csv=False, **kwargs
 ):
+    from funcy.seqs import flatten as flatten_list
+
     include_metrics = _parse_filter_list(kwargs.pop("include_metrics", []))
     exclude_metrics = _parse_filter_list(kwargs.pop("exclude_metrics", []))
     include_params = _parse_filter_list(kwargs.pop("include_params", []))
@@ -399,11 +403,26 @@ def show_experiments(
         include_params=include_params,
         exclude_params=exclude_params,
     )
-    metric_headers = _normalize_headers(metric_names)
-    param_headers = _normalize_headers(param_names)
+
+    headers = [
+        "Experiment",
+        "rev",
+        "typ",
+        "Created",
+        "parent",
+        "State",
+        "Executor",
+    ]
+
+    names = {**metric_names, **param_names}
+    counter = Counter(flatten_list([list(a.keys()) for a in names.values()]))
+    counter.update(headers)
+    metric_headers = _normalize_headers(metric_names, counter)
+    param_headers = _normalize_headers(param_names, counter)
 
     td = experiments_table(
         all_experiments,
+        headers,
         metric_headers,
         metric_names,
         param_headers,
@@ -411,6 +430,8 @@ def show_experiments(
         kwargs.get("sort_by"),
         kwargs.get("sort_order"),
         kwargs.get("precision"),
+        kwargs.get("fill_value"),
+        kwargs.get("iso"),
     )
 
     if no_timestamp:
@@ -422,9 +443,12 @@ def show_experiments(
 
     row_styles = lmap(baseline_styler, td.column("typ"))
 
-    merge_headers = ["Experiment", "rev", "typ", "parent"]
-    td.column("Experiment")[:] = map(prepare_exp_id, td.as_dict(merge_headers))
-    td.drop(*merge_headers[1:])
+    if not show_csv:
+        merge_headers = ["Experiment", "rev", "typ", "parent"]
+        td.column("Experiment")[:] = map(
+            prepare_exp_id, td.as_dict(merge_headers)
+        )
+        td.drop(*merge_headers[1:])
 
     headers = {"metrics": metric_headers, "params": param_headers}
     styles = {
@@ -453,13 +477,11 @@ def show_experiments(
         rich_table=True,
         header_styles=styles,
         row_styles=row_styles,
+        show_csv=show_csv,
     )
 
 
-def _normalize_headers(names):
-    count = Counter(
-        name for path in names for name in names[path] for path in names
-    )
+def _normalize_headers(names, count):
     return [
         name if count[name] == 1 else f"{path}:{name}"
         for path in names
@@ -493,6 +515,14 @@ class CmdExperimentsShow(CmdBase):
 
             ui.write(json.dumps(all_experiments, default=_format_json))
         else:
+            precision = (
+                self.args.precision or None
+                if self.args.show_csv
+                else DEFAULT_PRECISION
+            )
+            fill_value = "" if self.args.show_csv else FILL_VALUE
+            iso = True if self.args.show_csv else False
+
             show_experiments(
                 all_experiments,
                 include_metrics=self.args.include_metrics,
@@ -502,8 +532,11 @@ class CmdExperimentsShow(CmdBase):
                 no_timestamp=self.args.no_timestamp,
                 sort_by=self.args.sort_by,
                 sort_order=self.args.sort_order,
-                precision=self.args.precision or DEFAULT_PRECISION,
+                precision=precision,
+                fill_value=fill_value,
+                iso=iso,
                 pager=not self.args.no_pager,
+                show_csv=self.args.show_csv,
             )
         return 0
 
@@ -883,6 +916,12 @@ def add_parser(subparsers, parent_parser):
         default=False,
         help="Print output in JSON format instead of a human-readable table.",
     )
+    experiments_show_parser.add_argument(
+        "--show-csv",
+        action="store_true",
+        default=False,
+        help="Print output in csv format instead of a human-readable table.",
+    )
     experiments_show_parser.add_argument(
         "--precision",
         type=int,

dvc/compare.py

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

dvc/fs/http.py

@@ -50,7 +50,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
         from dvc.config import ConfigError
 
         credentials = {}
-        client_args = credentials.setdefault("client_args", {})
+        client_kwargs = credentials.setdefault("client_kwargs", {})
 
         if config.get("auth"):
             user = config.get("user")
@@ -68,7 +68,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
                         "'user' and 'password'"
                     )
 
-                client_args["auth"] = aiohttp.BasicAuth(user, password)
+                client_kwargs["auth"] = aiohttp.BasicAuth(user, password)
             elif auth_method == "custom":
                 if custom_auth_header is None or password is None:
                     raise ConfigError(
@@ -83,7 +83,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
 
         if "ssl_verify" in config:
             with fsspec_loop():
-                client_args["connector"] = aiohttp.TCPConnector(
+                client_kwargs["connector"] = aiohttp.TCPConnector(
                     ssl=make_context(config["ssl_verify"])
                 )
 
@@ -92,6 +92,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
         return credentials
 
     async def get_client(self, **kwargs):
+        import aiohttp
         from aiohttp_retry import ExponentialRetry, RetryClient
 
         kwargs["retry_options"] = ExponentialRetry(
@@ -100,6 +101,18 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
             max_timeout=self.REQUEST_TIMEOUT,
         )
 
+        # The default timeout for the aiohttp is 300 seconds
+        # which is too low for DVC's interactions (especially
+        # on the read) when dealing with large data blobs. We
+        # unlimit the total time to read, and only limit the
+        # time that is spent when connecting to the remote server.
+        kwargs["timeout"] = aiohttp.ClientTimeout(
+            total=None,
+            connect=self.REQUEST_TIMEOUT,
+            sock_connect=self.REQUEST_TIMEOUT,
+            sock_read=None,
+        )
+
         return RetryClient(**kwargs)
 
     @cached_property
@@ -108,7 +121,7 @@ class HTTPFileSystem(CallbackMixin, NoDirectoriesMixin, FSSpecWrapper):
             HTTPFileSystem as _HTTPFileSystem,
         )
 
-        return _HTTPFileSystem(timeout=self.REQUEST_TIMEOUT)
+        return _HTTPFileSystem(**self.fs_args)
 
     def _entry_hook(self, entry):
         entry["checksum"] = entry.get("ETag") or entry.get("Content-MD5")

dvc/machine/backend/base.py

@@ -1,15 +1,38 @@
-from abc import abstractmethod
+from abc import ABC, abstractmethod
 from contextlib import contextmanager
 from typing import TYPE_CHECKING, Iterator, Optional
 
-from tpi import base
-
 if TYPE_CHECKING:
     from dvc.fs.ssh import SSHFileSystem
     from dvc.repo.experiments.executor.base import BaseExecutor
+    from dvc.types import StrPath
 
 
-class BaseMachineBackend(base.BaseMachineBackend):
+class BaseMachineBackend(ABC):
+    def __init__(self, tmp_dir: "StrPath", **kwargs):
+        raise NotImplementedError
+
+    @abstractmethod
+    def create(self, name: Optional[str] = None, **config):
+        """Create and start an instance of the specified machine."""
+
+    @abstractmethod
+    def destroy(self, name: Optional[str] = None, **config):
+        """Stop and destroy all instances of the specified machine."""
+
+    @abstractmethod
+    def instances(
+        self, name: Optional[str] = None, **config
+    ) -> Iterator[dict]:
+        """Iterate over status of all instances of the specified machine."""
+
+    def close(self):
+        pass
+
+    @abstractmethod
+    def run_shell(self, name: Optional[str] = None, **config):
+        """Spawn an interactive SSH shell for the specified machine."""
+
     @abstractmethod
     def get_executor(
         self, name: Optional[str] = None, **config

dvc/machine/backend/terraform.py

@@ -5,11 +5,17 @@ from tpi import TerraformProviderIterative, terraform
 
 from dvc.fs.ssh import SSHFileSystem
 
+from .base import BaseMachineBackend
+
 if TYPE_CHECKING:
     from dvc.repo.experiments.executor.base import BaseExecutor
+    from dvc.types import StrPath
 
 
-class TerraformBackend(terraform.TerraformBackend):
+class TerraformBackend(terraform.TerraformBackend, BaseMachineBackend):
+    def __init__(self, tmp_dir: "StrPath", **kwargs):
+        super().__init__(tmp_dir)
+
     def get_executor(
         self, name: Optional[str] = None, **config
     ) -> "BaseExecutor":

dvc/version.py

@@ -6,7 +6,7 @@
 import os
 import subprocess
 
-_BASE_VERSION = "2.7.1"
+_BASE_VERSION = "2.7.2"
 
 
 def _generate_version(base_version):

setup.py

@@ -107,7 +107,7 @@ ssh = ["sshfs>=2021.8.1"]
 
 hdfs = ["pyarrow>=2.0.0; python_version < '3.10'"]
 webhdfs = ["hdfs==2.5.8"]
-webdav = ["webdav4>=0.9.0"]
+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,

tests/func/experiments/test_show.py

@@ -9,6 +9,7 @@ from dvc.exceptions import InvalidArgumentError
 from dvc.main import main
 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 tests.func.test_repro_multistage import COPY_SCRIPT
@@ -269,6 +270,10 @@ def test_show_filter(
     included,
     excluded,
 ):
+    from contextlib import contextmanager
+
+    from dvc.ui import ui
+
     capsys.readouterr()
     div = "│" if os.name == "nt" else "┃"
 
@@ -312,13 +317,30 @@ def test_show_filter(
     if e_params is not None:
         command.append(f"--exclude-params={e_params}")
 
-    assert main(command) == 0
+    @contextmanager
+    def console_with(console, width):
+        console_options = console.options
+        original = console_options.max_width
+        con_width = console._width
+
+        try:
+            console_options.max_width = width
+            console._width = width
+            yield
+        finally:
+            console_options.max_width = original
+            console._width = con_width
+
+    with console_with(ui.rich_console, 255):
+        assert main(command) == 0
     cap = capsys.readouterr()
 
     for i in included:
-        assert f"{div} {i} {div}" in cap.out
+        assert f"{div} params.yaml:{i} {div}" in cap.out
+        assert f"{div} metrics.yaml:{i} {div}" in cap.out
     for e in excluded:
-        assert f"{div} {e} {div}" not in cap.out
+        assert f"{div} params.yaml:{e} {div}" not in cap.out
+        assert f"{div} metrics.yaml:{e} {div}" not in cap.out
 
 
 def test_show_multiple_commits(tmp_dir, scm, dvc, exp_stage):
@@ -412,7 +434,6 @@ def test_show_running_checkpoint(
 ):
     from dvc.repo.experiments.base import EXEC_BRANCH
     from dvc.repo.experiments.executor.local import TempDirExecutor
-    from dvc.repo.experiments.utils import exp_refs_by_rev
 
     baseline_rev = scm.get_rev()
     dvc.experiments.run(
@@ -471,3 +492,46 @@ def test_show_with_broken_repo(tmp_dir, scm, dvc, exp_stage, caplog):
 
     paths = ["workspace", "baseline", "error"]
     assert isinstance(get_in(result, paths), YAMLFileCorruptedError)
+
+
+def test_show_csv(tmp_dir, scm, dvc, exp_stage, capsys):
+    baseline_rev = scm.get_rev()
+
+    def _get_rev_isotimestamp(rev):
+        return datetime.fromtimestamp(
+            scm.gitpython.repo.rev_parse(rev).committed_date
+        ).isoformat()
+
+    result1 = dvc.experiments.run(exp_stage.addressing, params=["foo=2"])
+    rev1 = first(result1)
+    ref_info1 = first(exp_refs_by_rev(scm, rev1))
+    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
+    cap = capsys.readouterr()
+    assert (
+        "Experiment,rev,typ,Created,parent,metrics.yaml:foo,params.yaml:foo"
+        in cap.out
+    )
+    assert ",workspace,baseline,,,3,3" in cap.out
+    assert (
+        "master,{},baseline,{},,1,1".format(
+            baseline_rev[:7], _get_rev_isotimestamp(baseline_rev)
+        )
+        in cap.out
+    )
+    assert (
+        "{},{},branch_base,{},,2,2".format(
+            ref_info1.name, rev1[:7], _get_rev_isotimestamp(rev1)
+        )
+        in cap.out
+    )
+    assert (
+        "{},{},branch_commit,{},,3,3".format(
+            ref_info2.name, rev2[:7], _get_rev_isotimestamp(rev2)
+        )
+        in cap.out
+    )

tests/unit/command/test_experiments.py

@@ -1,3 +1,5 @@
+from datetime import datetime
+
 import pytest
 
 from dvc.cli import parse_args
@@ -12,6 +14,7 @@ from dvc.command.experiments import (
     CmdExperimentsRemove,
     CmdExperimentsRun,
     CmdExperimentsShow,
+    show_experiments,
 )
 from dvc.exceptions import InvalidArgumentError
 
@@ -277,3 +280,128 @@ def test_experiments_remove(dvc, scm, mocker, queue, clear_all, remote):
         clear_all=clear_all,
         remote="myremote" if remote else None,
     )
+
+
+all_experiments = {
+    "workspace": {
+        "baseline": {
+            "data": {
+                "timestamp": None,
+                "params": {
+                    "params.yaml": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "parent": 20170428,
+                            "train": {
+                                "n_est": 100,
+                                "min_split": 36,
+                            },
+                        }
+                    }
+                },
+                "queued": False,
+                "running": False,
+                "executor": None,
+                "metrics": {
+                    "scores.json": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "avg_prec": 0.5843640011189556,
+                            "roc_auc": 0.9544670443829399,
+                        }
+                    }
+                },
+            }
+        }
+    },
+    "b05eecc666734e899f79af228ff49a7ae5a18cc0": {
+        "baseline": {
+            "data": {
+                "timestamp": datetime(2021, 8, 2, 16, 48, 14),
+                "params": {
+                    "params.yaml": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "parent": 20170428,
+                            "train": {
+                                "n_est": 100,
+                                "min_split": 2,
+                            },
+                        }
+                    }
+                },
+                "queued": False,
+                "running": False,
+                "executor": None,
+                "metrics": {
+                    "scores.json": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "avg_prec": 0.5325162867864254,
+                            "roc_auc": 0.9106964878520005,
+                        }
+                    }
+                },
+                "name": "master",
+            }
+        },
+        "ae99936461d6c3092934160f8beafe66a294f98d": {
+            "data": {
+                "timestamp": datetime(2021, 8, 31, 14, 56, 55),
+                "params": {
+                    "params.yaml": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "parent": 20170428,
+                            "train": {
+                                "n_est": 100,
+                                "min_split": 36,
+                            },
+                        }
+                    }
+                },
+                "queued": True,
+                "running": True,
+                "executor": None,
+                "metrics": {
+                    "scores.json": {
+                        "data": {
+                            "featurize": {"max_features": 3000, "ngrams": 1},
+                            "avg_prec": 0.5843640011189556,
+                            "roc_auc": 0.9544670443829399,
+                        }
+                    }
+                },
+                "name": "exp-44136",
+            }
+        },
+    },
+}
+
+
+def test_show_experiments(capsys):
+    show_experiments(
+        all_experiments, precision=None, fill_value="", iso=True, show_csv=True
+    )
+    cap = capsys.readouterr()
+    assert (
+        "Experiment,rev,typ,Created,parent,State,scores.json:"
+        "featurize.max_features,scores.json:featurize.ngrams,"
+        "avg_prec,roc_auc,params.yaml:featurize.max_features,"
+        "params.yaml:featurize.ngrams,params.yaml:parent,"
+        "train.n_est,train.min_split" in cap.out
+    )
+    assert (
+        ",workspace,baseline,,,,3000,1,0.5843640011189556,0.9544670443829399,"
+        "3000,1,20170428,100,36" in cap.out
+    )
+    assert (
+        "master,b05eecc,baseline,2021-08-02T16:48:14,,,3000,1,"
+        "0.5325162867864254,0.9106964878520005,3000,1,20170428,100,2"
+        in cap.out
+    )
+    assert (
+        "exp-44136,ae99936,branch_base,2021-08-31T14:56:55,,Running,"
+        "3000,1,0.5843640011189556,0.9544670443829399,3000,1,20170428,100,36"
+        in cap.out
+    )

tests/unit/remote/test_http.py

@@ -23,7 +23,7 @@ def test_public_auth_method(dvc):
 
     fs = HTTPFileSystem(**config)
 
-    assert "auth" not in fs.fs_args["client_args"]
+    assert "auth" not in fs.fs_args["client_kwargs"]
     assert "headers" not in fs.fs_args
 
 
@@ -40,8 +40,8 @@ def test_basic_auth_method(dvc):
 
     fs = HTTPFileSystem(**config)
 
-    assert fs.fs_args["client_args"]["auth"].login == user
-    assert fs.fs_args["client_args"]["auth"].password == password
+    assert fs.fs_args["client_kwargs"]["auth"].login == user
+    assert fs.fs_args["client_kwargs"]["auth"].password == password
 
 
 def test_custom_auth_method(dvc):
@@ -70,7 +70,7 @@ def test_ssl_verify_disable(dvc):
     }
 
     fs = HTTPFileSystem(**config)
-    assert not fs.fs_args["client_args"]["connector"]._ssl
+    assert not fs.fs_args["client_kwargs"]["connector"]._ssl
 
 
 @patch("ssl.SSLContext.load_verify_locations")
@@ -84,7 +84,7 @@ def test_ssl_verify_custom_cert(dvc, mocker):
     fs = HTTPFileSystem(**config)
 
     assert isinstance(
-        fs.fs_args["client_args"]["connector"]._ssl, ssl.SSLContext
+        fs.fs_args["client_kwargs"]["connector"]._ssl, ssl.SSLContext
     )