[Jayant-kernel]

Metadata

• Reference Issue: Part of [MNT] Replace Live Server Calls with Mocks in Unit Tests #1649
• New Tests Added: Existing ontology tests updated
• Documentation Updated: N/A
• Change Log Entry: Replace live ontology writes in dataset tests with mocked requests.

Details

The ontology tests write links to dataset 1, and the multiple-feature test leaves them behind. This replaces the four write cases with mocked HTTP requests so they can run offline without changing server data.

The tests check the add/remove endpoints, posted feature details, and handling of error codes 1105 and 1106. The error cases check the client's response handling; URL validation still belongs to the server.

Checked locally: all 5 offline tests in test_dataset.py pass, along with pre-commit checks.

[shreyashsri79]

LLM generated content, by GPT-6 (Codex)

Independent local verification of Jayant's diff at fa05aebbaa94b78b405978ac2567618ae300933b:

• Exact PR head: 5 passed, 21 deselected.
• The same test diff applied three-way to current main f4fd28e7b55a44615cace15c9dd51afb0cb08f44: 4 passed, 21 deselected. Main now marks the unrelated _read_qualities test as test_server, explaining the count difference.

Both runs selected tests/test_datasets/test_dataset.py -m 'not production_server and not test_server' through pytest.main, with a process-local wrapper that rejects socket connection and DNS calls. Both recorded zero attempted network calls. Environment: Python 3.14.7, pandas 3.1.1, requests-mock 1.12.1, isolated .[test] installation and temporary OpenML configuration/cache directories.

The assertions exercise the add/remove endpoints, encoded payloads, feature indices and client handling of errors 1105/1106. The comment correctly distinguishes mocked rejection handling from actual server-side URL validation. These tests also pass through the refactored backend on current main.

No blocking finding in this narrow review. Server tests, the full suite and pre-commit were not run locally; this does not validate live server behavior. Codex performed this review and the local checks.

[shreyashsri79]

Follow-up on 167630a6709b0a0a6a1dec1a15c0db2f2e8ad2fe: the four changed ontology cases pass locally on this exact head.

There is one import-block conflict with current main (ccb524473f91b22a66b8680df569736161ff7b0d). In a private integration checkout, keeping both imports resolved it:

import requests
from requests_mock import Mocker

The same four cases also passed on current main plus this patch with that import-only resolution. The three ontology test functions, including parametrization, remained identical to the author head.

Validation used test_add_remove_ontology_to_dataset, test_add_same_ontology_multiple_features, and both test_add_illegal_ontology cases through a process-local socket-blocking pytest wrapper with -m "not production_server and not test_server". Both runs recorded zero network attempts. Environment: Python 3.14.7, pytest 9.2.0, requests-mock 1.12.1. The resolved diff also passed git diff --cached --check.

Full-suite, server, pre-commit and other-Python checks were not run locally; these results are not hosted CI. No additional finding within this scope.

Review and local checks by GPT-6 (Codex).

[Jayant-kernel]

Done the changes