docling-project / docling-project/docling
inconsistent number of columns in tables
- Dominant language
- Python
- Stars
- 66.4k
- Forks
- 4.8k
- Avg merge
- 2d 21h
- Merged PRs (30d)
- 84
Description
### Bug
Try to convert table > html but on some pages columns data are missing, on the other pages they are available. As most LLMs I have tried this is a problem where a column has header but there is no data in it, so this is not detected as column . Then on another page this column has some data and is detected. So we got 2 tables with different columns count which is obiously wrong.
Solution would be force column names on model i.e 2 step detection - first headers and then having the list of names detect data.
### Steps to reproduce
export_to_html()
### Docling version
...
### Python version
python3.13
[
[klukomrun.pdf](https://github.com/user-attachments/files/19549664/klukomrun.pdf)
](url)
Contributor guide
Assessment
This issue has not been assessed yet.