fix(PLU-373): make notion database test row-order insensitive (#714)

## Summary
- Notion's database query endpoint doesn't guarantee stable row order,
but `test_notion_source_database` compared the downloaded HTML
byte-for-byte. Every prior "fix" (#693, #605) was just reshuffling the
same four rows.
- Add `unordered_table_html_equality_check` that compares the header
positionally and data rows as a multiset of their text content, and wire
it into the database test only (the page test isn't a table).
- On mismatch the new check prints the symmetric difference so the next
real regression won't be a black box like
https://github.com/Unstructured-IO/unstructured-ingest/actions/runs/26238030726/job/77216703553
was.
- Version bumped to `1.6.2-dev` to satisfy `check-version` without
claiming a release slot for a test-only change.

## Test plan
- [x] `uncategorized_connectors_int_test` Notion tests pass (real
verification — needs the CI's `NOTION_API_KEY`)
- [x] Smoke-tested locally against the real fixture: reordered rows
compare equal, a tampered row fails with a readable diff

<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Make the Notion database integration test row-order insensitive and
correct a misleading helper comment, eliminating flaky failures when
Notion returns rows in different orders. Addresses PLU-373.

- **Bug Fixes**
- Added `unordered_table_html_equality_check` to compare the header
positionally and treat table rows as a multiset; prints a readable
symmetric diff on mismatch.
- Used in `test_notion_source_database` only; the page test remains
unchanged.
- Fixed the helper comment to state it scans all `<tr>` elements, not
just the first table.
- Bumped version to `1.6.6-dev` and updated `CHANGELOG.md`; test-only
change with no runtime impact.

<sup>Written for commit 87d92cc.
Summary will update on new commits. <a
href="https://cubic.dev/pr/Unstructured-IO/unstructured-ingest/pull/714?utm_source=github">Review
in cubic</a></sup>

<!-- End of auto-generated description by cubic. -->

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: awalker4

CHANGELOG.md

@@ -1,3 +1,9 @@
+## [1.6.6-dev]
+
+### Fixes
+
+- **test(notion): make `test_notion_source_database` row-order insensitive.** Test-only change; no published behavior.
+
 ## [1.6.5]
 
 ### Fixes

test/integration/connectors/test_notion.py

@@ -3,6 +3,9 @@
 import pytest
 
 from test.integration.connectors.utils.constants import SOURCE_TAG, UNCATEGORIZED_TAG
+from test.integration.connectors.utils.validation.equality import (
+    unordered_table_html_equality_check,
+)
 from test.integration.connectors.utils.validation.source import (
     SourceValidationConfigs,
     get_all_file_data,
@@ -59,6 +62,7 @@ def test_notion_source_database(temp_dir):
             exclude_fields_extend=["metadata.date_created", "metadata.date_modified"],
             predownload_file_data_check=source_filedata_display_name_set_check,
             postdownload_file_data_check=source_filedata_display_name_set_check,
+            file_equality_check=unordered_table_html_equality_check,
         ),
     )
 

test/integration/connectors/utils/validation/equality.py

@@ -1,4 +1,5 @@
 import json
+from collections import Counter
 from pathlib import Path
 
 from bs4 import BeautifulSoup
@@ -47,6 +48,49 @@ def html_equality_check(expected_filepath: Path, current_filepath: Path) -> bool
     return expected_soup.text == current_soup.text
 
 
+def unordered_table_html_equality_check(
+    expected_filepath: Path, current_filepath: Path
+) -> bool:
+    # Equality check for HTML files whose rows arrive in arbitrary order.
+    # The first <tr> in the document is compared positionally as a header;
+    # remaining <tr>s are compared as a multiset of their text content. Used
+    # for connectors whose upstream API doesn't guarantee stable row ordering
+    # (e.g. Notion's database query response).
+    with expected_filepath.open() as expected_f:
+        expected_soup = BeautifulSoup(expected_f, "html.parser")
+    with current_filepath.open() as current_f:
+        current_soup = BeautifulSoup(current_f, "html.parser")
+
+    def split_rows(soup: BeautifulSoup) -> tuple[str, list[str]]:
+        rows = soup.find_all("tr")
+        if not rows:
+            return "", []
+        header = rows[0].get_text(" ", strip=True)
+        data = sorted(r.get_text(" ", strip=True) for r in rows[1:])
+        return header, data
+
+    expected_header, expected_data = split_rows(expected_soup)
+    current_header, current_data = split_rows(current_soup)
+
+    if expected_header != current_header:
+        print("table header differs:")
+        print(f"  expected: {expected_header}")
+        print(f"  current:  {current_header}")
+        return False
+    if expected_data != current_data:
+        expected_counts = Counter(expected_data)
+        current_counts = Counter(current_data)
+        only_in_expected = expected_counts - current_counts
+        only_in_current = current_counts - expected_counts
+        print("table rows differ (order-insensitive):")
+        for row, n in only_in_expected.items():
+            print(f"  only in expected (x{n}): {row}")
+        for row, n in only_in_current.items():
+            print(f"  only in current  (x{n}): {row}")
+        return False
+    return True
+
+
 def txt_equality_check(expected_filepath: Path, current_filepath: Path) -> bool:
     with expected_filepath.open() as expected_f:
         expected_text_lines = expected_f.readlines()

unstructured_ingest/__version__.py

@@ -1 +1 @@
-__version__ = "1.6.5"  # pragma: no cover
+__version__ = "1.6.6-dev"  # pragma: no cover