feat: revize TripletTableSerializer

[wanadzhar913]

Why

TripletTableSerializer relied on first-row/first-column assumptions, causing triplets with vague semantic meaning when converted to text, and sometimes skipped cells (especially in 0,0).

The solution I came up with is very verbose e.g., row_1, col_1 = Value, etc. but was of the opinion that it's better to be precise with positioning (instead of assuming).

As the code edits the tests & test data, please let me know if it's acceptable as it will disturb with the main docling package downstream.

Issues

• closes Review the serialization of tables in triplet format #528

Changes

• Updated TripletTableSerializer in docling_core/transforms/chunker/hierarchical_chunker.py:
  • metadata-first header detection derived from TableItem's row_header / col_header attribute e.g., has_row_headers = any(getattr(cell, "row_header", False) for row in TableItem.data.grid for cell in row)
  • safe fallback when metadata is missing
  • consistent row_i/col_j labeling when headers are absent
  • empty/NaN filtering
• Improved nested table detection/serialization:
  • handles nested DataFrame cells
  • detects serialized inner-table strings and emits -> relation (instead of =
  • recursion depth cap (max_depth=3)
• Added focused tests in test/test_triplet_serializer.py (8 scenarios, including nested tables and merged cells)
• Refreshed affected chunker/hybrid golden fixtures datasets (e.g., uv run env DOCLING_GEN_TEST_DATA=1 pytest -q test/test_hierarchical_chunker.py)

Sample Output

I'm comparing it to 0c_out_chunks.json.

Old:

cell 0,0, 1 = cell 0,1. cell 1,0, 1 = <em><p>text in italic</p></em>. <ul>\n<li>list item 1</li>\n<li>list item 2</li>\n</ul>, 1 = cell 2,1. cell 3,0, 1 = inner cell 0,0, 1 = inner cell 0,1. inner cell 0,0, 2 = inner cell 0,2. inner cell 1,0, 1 = inner cell 1,1. inner cell 1,0, 2 = inner cell 1,2. <p>Some text in a generic group.</p>\n<p>More text in the group.</p>, 1 = cell 4,1

New:

row_0, col_0 = cell 0,0. row_0, col_1 = cell 0,1. row_1, col_0 = cell 1,0. row_1, col_1 = <em><p>text in italic</p></em>. row_2, col_0 = <ul>\n<li>list item 1</li>\n<li>list item 2</li>\n</ul>. row_2, col_1 = cell 2,1. row_3, col_0 = cell 3,0. row_3, col_1 -> row_0, col_0 = inner cell 0,0. row_0, col_1 = inner cell 0,1. row_0, col_2 = inner cell 0,2. row_1, col_0 = inner cell 1,0. row_1, col_1 = inner cell 1,1. row_1, col_2 = inner cell 1,2. row_4, col_0 = <p>Some text in a generic group.</p>\n<p>More text in the group.</p>. row_4, col_1 = cell 4,1

[github-actions]

✅ DCO Check Passed

Thanks @wanadzhar913, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

Your pull request matches the following merge protections and will not be merged until they are valid.

🔴 Require two reviewer for test updates

Waiting for

• #approved-reviews-by >= 2

This rule is failing.

When test data is updated, we require two reviewers

• #approved-reviews-by >= 2

🟢 Enforce conventional commit

Wonderful, this rule succeeded.

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[wanadzhar913]

Hi @ceberam! Hope the general approach I'm taking for this PR is correct. Tysm!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[wanadzhar913]

Hi @ceberam, I've added wider test coverage! Apols and tysm.

[ceberam]

Hi @ceberam, I've added wider test coverage! Apols and tysm.

Thanks @wanadzhar913 , I'll review in the next couple of days. In the meantime, can you please remediate the Developer Certificate of Origin (DCO) check failure?

[wanadzhar913]

Hi @ceberam, I've signed the DCO commit checks.

Am also aware of an open issue in docling-project/docling#3335. Am happy to rebase and reimplement if that fix is more pertinent.