feat(iter_submodules): add match_containing mode

This alters the `pathspec` evaluation to yield submodules that
*may* have content matching any pathspec, rather than the
pathspec having to match a submodule item directly.

This feature can be useful for implementing submodule recursion
around Git commands that do not support submodule recursion
directly.

datalad_next/iter_collections/gitworktree.py

@@ -244,25 +244,59 @@ def iter_submodules(
     path: Path,
     *,
     pathspecs: list[str | GitPathSpec] | None = None,
+    match_containing: bool = False,
 ) -> Generator[GitTreeItem, None, None]:
     """Given a path, report all submodules of a repository worktree underneath
 
-    This is a thin convenience wrapper around ``iter_gitworktree()``.
+    With ``match_containing`` set to the default ``False``, this is merely a
+    convenience wrapper around ``iter_gitworktree()`` that selectively reports
+    on submodules. With ``match_containing=True`` and ``pathspecs`` given, the
+    yielded items corresponding to submodules where the given ``pathsspecs``
+    *could* match content. This includes submodules that are not available
+    locally, because no actual matching of pathspecs to submodule content is
+    performed -- only an evaluation of the submodule item itself.
     """
+    if not pathspecs:
+        # force flag to be sensible to simplify internal logic
+        match_containing = False
+    else:
+        pathspecs = [GitPathSpec.from_pathspec_str(ps) for ps in pathspecs]
+
     for item in iter_gitworktree(
         path,
         untracked=None,
         link_target=False,
         fp=False,
         recursive='repository',
-        pathspecs=pathspecs,
+        # if we want to match submodules that contain pathspecs matches
+        # we cannot give the job to Git, it won't report anything,
+        # but we need to match manually below
+        pathspecs=None if match_containing else pathspecs,
     ):
         # exclude non-submodules, or a submodule that was found at
         # the root path -- which would indicate that the submodule
         # itself it not around, only its record in the parent
-        if item.gittype == GitTreeItemType.submodule \
-                and item.name != PurePath('.'):
+        if item.gittype != GitTreeItemType.submodule \
+                or item.name == PurePath('.'):
+            continue
+
+        if not match_containing:
             yield item
+            continue
+
+        # does any pathspec match the "inside" of the current submodule's
+        # path
+        # we are using any() here to return as fast as possible.
+        # theoretically, we could also port all of them and enhance
+        # GitTreeItem to carry them outside, but we have no idea
+        # about the outside use case here, and cannot assume the additional
+        # cost is worth it
+        if any(ps.for_subdir(str(item.name)) for ps in pathspecs):
+            yield item
+            continue
+
+        # no match
+        continue
 
 
 def _get_item(

datalad_next/iter_collections/tests/test_itergitworktree.py

@@ -274,3 +274,19 @@ def test_iter_submodules(modified_dataset):
     res = list(iter_submodules(p, pathspecs=[':(exclude)*/sm_c']))
     assert len(res) == len(all_sm) - 1
     assert not any(str(r.name) == 'dir_sm/sm_c' for r in res)
+
+    # test pathspecs matching inside submodules
+    # baseline, pointing inside a submodule gives no matching results
+    assert not list(iter_submodules(p, pathspecs=['dir_sm/sm_c/.datalad']))
+    # we can discover the submodule that could have content that matches
+    # the pathspec
+    res = list(iter_submodules(p, pathspecs=['dir_sm/sm_c/.datalad'],
+                               match_containing=True))
+    assert len(res) == 1
+    assert str(res[0].name) == 'dir_sm/sm_c'
+    # if we use a wildcard that matches any submodule, we also get all of them
+    # and this includes the dropped submodule, because iter_submodules()
+    # make no assumptions on what this information will be used for
+    res = list(iter_submodules(p, pathspecs=['*/.datalad'],
+                               match_containing=True))
+    assert len(res) == len(all_sm)