fix(preprocessors): stop CSVDocumentSplitter recursing forever on multi-block CSVs (#12190)

[Anai-Guo]

Related Issues

Fixes #12190

Proposed Changes

CSVDocumentSplitter.run() could raise RecursionError when both row_split_threshold and column_split_threshold are set and a later row block still needs a column split.

Root cause: _find_split_indices reports empty rows/columns using pandas index/column labels, while _split_dataframe removes them with positional .iloc slicing. For the top-level DataFrame the labels happen to equal the positions (a RangeIndex), so it works. But on a sub-table produced by an earlier split, the labels no longer match positions — e.g. an empty row at label 6 sits at position 1 in a 3-row sub-table. df.iloc[0:6] then returns the whole table unchanged, the empty row is never removed, and _recursive_split keeps re-detecting it and recurses until the interpreter's recursion limit is hit.

The fix detects empty rows/columns by position (enumerate over the boolean mask) so the split points stay consistent with the positional .iloc slicing. Sub-table index/column labels are left untouched, so the row_idx_start / col_idx_start metadata (read from split_df.index[0] / split_df.columns[0]) is unchanged. For the top-level DataFrame, positions equal labels, so existing behavior — including the _find_split_indices unit tests — is preserved.

How did you test it?

• Added a regression test (test_recursive_split_later_block_needs_column_split) using the exact CSV from the issue; it previously hit RecursionError and now returns 4 sub-tables with the correct content and metadata.
• Verified the existing TestFindSplitIndices, test_recursive_split_one_level, and test_recursive_split_two_levels cases still produce identical results.

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added tests that demonstrate the correct behavior of the change
• I've used one of the conventional commit types for my PR title
• I documented my code
• I ran pre-commit hooks and fixed any issue

🤖 Generated with Claude Code

[vercel]

@Anai-Guo is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[HaystackBot]

Hi @Anai-Guo, thanks a lot for your contribution! 🙏

We noticed that the Contributor License Agreement (CLA) check (license/cla) hasn't passed yet, so we've temporarily moved this PR to draft and paused the review assignment.

To get your PR reviewed, please sign the CLA via the link in the license/cla check below (or in the CLA bot comment). As soon as the check turns green, this PR will automatically be marked ready for review again and a reviewer will be re-assigned.

[HaystackBot]

Hi @Anai-Guo, just a friendly reminder: this PR is still in draft because the Contributor License Agreement (CLA) hasn't been signed yet. We'd love to review your contribution! Please sign the CLA via the link in the license/cla check, and this PR will automatically be marked ready for review.