ultimate-notion / ultimate-notion / 28161288553

88%

push

github

Capture block_type on unsupported blocks (#356) (#357)

* Capture block_type on unsupported blocks (#356)

Notion returns unsupported blocks (e.g. button or AI blocks) as
`{"type": "unsupported", "unsupported": {"block_type": "button"}}`.
`UnsupportedBlockTypeData` was an empty model, so on dev/CI installs
(`extra='forbid'`) the `block_type` key was rejected with a pydantic
ValidationError. This also made every sibling block after the
unsupported one inaccessible, since the whole children list failed to
validate.

Capture the read-only `block_type` field and add a regression test.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Make to_markdown cassette carry block_type on unsupported blocks (#356)

The existing full-path VCR test `test_page_to_markdown` renders two
unsupported blocks but its cassette stored an empty `unsupported: {}`
payload, so it never exercised the validation crash. Notion actually
sends `unsupported: {block_type: ...}`; patch the recorded response to
match reality (a button block and an ai_block) so the existing
read->render test now fails on the unfixed model and proves sibling
blocks after an unsupported one stay accessible.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* Add dedicated VCR regression test for unsupported blocks (#356)

Besides the model-level unit test, add an explicit @pytest.mark.vcr()
test that reads a real page (the "Markdown Test" fixture page, which
contains a button and an AI block) and asserts the unsupported blocks
load without raising, expose their read-only `block_type`, and render to
the placeholder. The cassette is a single block-children interaction
derived from the patched to_markdown cassette; the block_type values
(`button`, `ai_block`) were confirmed against the live Notion API.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

1 of 1 new or added line in 1 file covered. (100.0%)

6611 of 7522 relevant lines covered (87.89%)

6.12 hits per line