Implement full support for fp=True in iter_annexworktree()

Previously, it would only open annex objects.

Now regular files (tracked or untracked) and symlink targets
(via the symlink) are also opened, if they actually exist.

The corresponding test is extended appropriately.

Author: mih

datalad_next/iter_collections/annexworktree.py

@@ -31,6 +31,7 @@
     GitWorktreeFileSystemItem,
     iter_gitworktree
 )
+from .utils import FileSystemItemType
 
 
 lgr = logging.getLogger('datalad.ext.next.iter_collections.annexworktree')
@@ -83,7 +84,7 @@ def iter_annexworktree(
 
     The basic semantics of all arguments are identical to
     :func:`~datalad_next.iter_collections.gitworktree.iter_gitworktree`.
-    Importantly, with ``fp=True``, the annex object is opened directly,
+    Importantly, with ``fp=True``, an annex object is opened directly,
     if available. If not available, no attempt is made to open the associated
     symlink or pointer file.
 
@@ -240,19 +241,31 @@ def iter_annexworktree(
         path = Path(path)
         for res in results:
             item = _get_worktree_item(path, get_fs_info=True, **res)
-            annexobjpath = res.get('annexobjpath')
-            if not annexobjpath:
-                # this is not an annexed file
-                yield item
-                continue
-            full_annexobjpath = path / annexobjpath
-            if not full_annexobjpath.exists():
-                # annexed object is not present
-                yield item
-                continue
-            with (full_annexobjpath).open('rb') as active_fp:
-                item.fp = active_fp
+            # determine would file we would open
+            fp_src = None
+            if item.annexobjpath is not None:
+                # this is an annexed file
+                fp_src = item.annexobjpath
+            elif item.type == FileSystemItemType.file \
+                    and item.annexkey is None:
+                # regular file (untracked or tracked)
+                fp_src = item.name
+            elif item.type == FileSystemItemType.symlink \
+                    and item.annexkey is None:
+                # regular symlink
+                fp_src = item.name
+            if fp_src is None:
+                # nothing to open
                 yield item
+            else:
+                fp_src_fullpath = path / fp_src
+                if not fp_src_fullpath.exists():
+                    # nothing there to open (would resolve through a symlink)
+                    yield item
+                else:
+                    with fp_src_fullpath.open('rb') as active_fp:
+                        item.fp = active_fp
+                        yield item
 
 
 def _get_worktree_item(

datalad_next/iter_collections/tests/test_iterannexworktree.py

@@ -5,6 +5,7 @@
 from datalad import cfg as dlcfg
 
 from datalad_next.datasets import Dataset
+from datalad_next.utils import check_symlink_capability
 
 from ..annexworktree import (
     iter_annexworktree,
@@ -98,9 +99,21 @@ def test_iter_annexworktree_basic_fp(existing_dataset, no_result_rendering):
     for i in ('untracked', 'ingit'):
         (ds.pathobj / f'file_{i}').write_text(content_tmpl.format(i))
     ds.save('file_ingit', to_git=True)
+    # and add symlinks (untracked and in git)
+    if check_symlink_capability(
+        ds.pathobj / '_dummy', ds.pathobj / '_dummy_target'
+    ):
+        for i in ('symlinkuntracked', 'symlinkingit'):
+            tpath = ds.pathobj / f'target_{i}'
+            lpath = ds.pathobj / f'file_{i}'
+            tpath.write_text(content_tmpl.format(i))
+            lpath.symlink_to(tpath)
+    ds.save('file_symlinkingit', to_git=True)
 
     # we expect to process an exact number of files below
-    fcount = 5
+    # 3 annexed files, 1 untracked, 1 in git, 1 symlink in git,
+    # 1 symlink untracked
+    fcount = 7
     for ai in filter(
         lambda i: str(i.name.name).startswith('file_'),
         iter_annexworktree(ds.pathobj, fp=True)
@@ -110,5 +123,7 @@ def test_iter_annexworktree_basic_fp(existing_dataset, no_result_rendering):
             assert content_tmpl.format(
                 ai.name.name[5:]) == ai.fp.read().decode()
         else:
-            assert (ds.pathobj / ai.annexobjpath).exists() is False
+            assert (ai.annexobjpath and (
+                ds.pathobj / ai.annexobjpath).exists() is False) or (
+                    ai.name.exists() is False)
     assert not fcount