dvcfs: handle NotADirectoryError from datafs.ls #9746
Conversation
|
Any idea for where to test this? |
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## main #9746 +/- ##
==========================================
- Coverage 90.48% 90.47% -0.01%
==========================================
Files 480 480
Lines 36528 36534 +6
Branches 5252 5252
==========================================
+ Hits 33051 33053 +2
- Misses 2885 2887 +2
- Partials 592 594 +2
☔ View full report in Codecov by Sentry. |
|
For the record: I'll add a test, no worries. Thank you! 🙏 |
| def test_ls_dirty(tmp_dir, dvc): | ||
| tmp_dir.dvc_gen({"data": "data"}) | ||
| (tmp_dir / "data").unlink() | ||
|
|
||
| tmp_dir.gen({"data": {"foo": "foo", "bar": "bar"}}) | ||
|
|
||
| fs = DVCFileSystem(repo=dvc) | ||
| assert set(fs.ls("data")) == {"data/foo", "data/bar"} |
There was a problem hiding this comment.
This is a little different from what's described in the issue (because there is dvc.lock there), but it is the same situation of having a dirty workspace compared to what is recorded (or missing like with missing dvc.lock) in dvc files.
efiop
changed the title
|
EDIT: fixed in iterative/dvc-s3#46 |
Before
dvc_fs.ls(path)withoutdvc.lockwould returnpath/path. Now it won't return anything sincedvc_fshas no info about what's in that path.Fixes #9745