Show csv format of experiments

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(item):
     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):
@@ -327,6 +333,8 @@ def experiments_table(
     sort_by=None,
     sort_order=None,
     precision=DEFAULT_PRECISION,
+    fill_value=FILL_VALUE,
+    iso=False,
 ) -> "TabularData":
     from funcy import lconcat
 
@@ -342,7 +350,7 @@ def experiments_table(
         "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 +361,8 @@ def experiments_table(
             sort_by=sort_by,
             sort_order=sort_order,
             precision=precision,
+            fill_value=fill_value,
+            iso=iso,
         )
         td.extend(rows)
 
@@ -385,7 +395,7 @@ 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
 ):
     include_metrics = _parse_filter_list(kwargs.pop("include_metrics", []))
     exclude_metrics = _parse_filter_list(kwargs.pop("exclude_metrics", []))
@@ -399,8 +409,13 @@ def show_experiments(
         include_params=include_params,
         exclude_params=exclude_params,
     )
-    metric_headers = _normalize_headers(metric_names)
-    param_headers = _normalize_headers(param_names)
+
+    names = {**metric_names, **param_names}
+    counter = Counter(
+        name for path in names for name in names[path] for path in names
+    )
+    metric_headers = _normalize_headers(metric_names, counter)
+    param_headers = _normalize_headers(param_names, counter)
 
     td = experiments_table(
         all_experiments,
@@ -411,6 +426,8 @@ 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 +439,12 @@ 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 +473,11 @@ 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 +511,10 @@ def run(self):
 
             ui.write(json.dumps(all_experiments, default=_format_json))
         else:
+            precision = None if self.args.show_csv else self.args.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 +524,11 @@ def run(self):
                 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
 
@@ -882,9 +907,16 @@ 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,
+        default=DEFAULT_PRECISION,
         help=(
             "Round metrics/params to `n` digits precision after the decimal "
             f"point. Rounds to {DEFAULT_PRECISION} digits by default."

dvc/compare.py

@@ -168,7 +168,10 @@ def row_from_dict(self, d: Mapping[str, "CellT"]) -> None:
     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

tests/func/experiments/test_show.py

@@ -9,6 +9,7 @@
 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 +492,47 @@ 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()
+    print(cap.out)
+    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
+    )