Difficulties with Multi-line headers. Rows shifted down.

[poetaster]

Describe the bug
This pdf, https://poetaster.de/misc/118.pdf (which I'm not uploading here since it may be a copyright issue) is read well but camelot shifts the rows under the multi-header controllability, down.

Steps to reproduce the bug

Load the above file and try both stream and lattice reading. I tried a lot of variations:

stream with different row tolerances:
dfs = camelot.read_pdf('118.pdf', flavor='stream', row_tol=20,flag_size=True)

and lattice with many scale and shift variations.

dfs = camelot.read_pdf('118.pdf', flavor='lattice', shift_text=['r','t', 'r', 't'], line_scale=20)

Lattice appears to get it right:

camelot.plot(dfs[0], kind='grid').show()

Which seems correct. But it always shifts the rows in the controllability part.

Expected behavior

Rows should not be shifted.

Code

Began with:

import camelot
dfs = camelot.read_pdf('118.pdf') 

And tried many variation, most recent lattice being:
dfs = camelot.read_pdf('118.pdf', flavor='lattice', shift_text=['r','t', 'r', 't'], line_scale=20)

PDF
See above.

Screenshots
See above.

Environment

• OS: ubuntu 22
• Python version: 3.10.12
• Numpy version: 1.24.0
• OpenCV version: 4.8.1.78
• Ghostscript version: 0.7
• Camelot version: 0.11.0

Additional context

[wolfassi123]

@poetaster Did you manage to find a way to fix the issues with multi-row headers?

[paulobarrera14]

@poetaster Pinging this again, where you able to find a fix for the multi-row headers?

[poetaster]

I hadn't worked on this (ended up reading excel files directly for that project) since then. I've looked now, but thought i should probably update camelot? What version would be best to test with?

[poetaster]

I wasn't sure if I had done the original on my PC or on my jupyter lab server. On this pc, camelot is at 0.9.0 and the results are the same.

[poetaster]

Ok, updated to 0.11.0 and same same. I'm not sure if it's just that I haven't understood the shifting 'foo', but even without, camelot get's the grid correct, but shifts the content of the 'controlability' columns 2 down.