Repository navigation
MNT: Update tests to reflect server side changes - #1765
Conversation
| "contributor, collection_date, language, citation, " | ||
| "original_data_url, default_target_attribute, row_id_attribute, " | ||
| "ignore_attribute or paper_url to edit.", | ||
| ) |
There was a problem hiding this comment.
We used to rely on the server to provide this error response, but since this behavior was even encoded into a unit test, it seems that it should rather be expected of the client to catch the error. (I think it also makes more sense - no need to do a network round trip or force work onto the server)
|
|
||
| def _get_servers(mode: ServerMode) -> dict[APIVersion, dict[str, str | None]]: | ||
| if mode not in ServerMode: | ||
| if mode not in ServerMode._value2member_map_: |
There was a problem hiding this comment.
This change was made to support Python 3.10 and 3.11, which is still part of the CI matrix. When we drop support for these Python versions (which should do as soon as possible) we should revert this change.
| assert len(attribute_names) == X.shape[1] | ||
|
|
||
| @pytest.mark.test_server() | ||
| def test_data_edit_non_critical_field(self): |
There was a problem hiding this comment.
After OpenML#1290 we no longer allow editing of any field by users that do not own the dataset, so this test is no longer useful. There already existed a test that asserts the handling of an authorization error when trying to modify a field the user isn't allowed to.
| task_ids = [6] | ||
| uploader_ids = [1] | ||
| flow_ids = [6969] | ||
| runs = [6070667, 6070666] |
There was a problem hiding this comment.
While the batch limit of 1000 is fine, list_evaluations also has a (server-side) internal result set which is checked, and this was still exceeded. Since we do not care for getting such a large set per se, I filtered the set down to 2 particular runs.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #1765 +/- ##
==========================================
- Coverage 81.50% 80.89% -0.61%
==========================================
Files 63 63
Lines 5238 5240 +2
==========================================
- Hits 4269 4239 -30
- Misses 969 1001 +32 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
PGijsbers
changed the title
|
On the CI results:
|
3 participants
Metadata
Details
Various fixes to make tests pass.
The REST API has had a few updates after security and availability incidents, and the openml-python tests weren't always updated to reflect those changes. For changes related to #1757 I mostly added a skip instruction, though for some tests it was immediately obvious that even requesting 1000 results (the new limit) would be unnecessary. In those cases I just updated the test.