fix(CSVDocumentSplitter): use positional column index instead of column label

[Tyagiquamar]

Problem

CSVDocumentSplitter accepts read_csv_kwargs, but setting header=0 makes a normal CSV crash during splitting. With header=0, pandas uses the first row as string column labels, and the splitter later executed int(split_df.columns[0]) for the col_idx_start metadata, raising on the string label. The sub-table sort also ordered by label instead of position.

Solution

Resolve the first column of each sub-table to its positional index in the original DataFrame with df.columns.get_loc(...) for both the sort key and the col_idx_start metadata. Behaviour with the default header=None is unchanged, where labels are already positional integers.

Testing

• python -m pytest test/components/preprocessors/test_csv_document_splitter.py -q — 30 passed (Docker, python:3.12-bookworm), including a new regression test running the exact scenario from the issue: read_csv_kwargs={"header": 0} with a blank-row split, asserting col_idx_start == 0, correct row_idx_start values and split contents.

Fixes #12965

Note: this change was prepared with the help of an AI coding agent under human direction, per the AI-assistants contribution guidance.

[vercel]

@Tyagiquamar 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 @Tyagiquamar, 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.