Misdetection of columns with narrow gap between them

[saidakyuz]

When I was trying to extract table in screenshot from pdf, you can download from given link, I got the table object which extracted as first and second columns are combined together.

Steps to reproduce the bug

Run the following code after installing Camelot and Ghostscript etc.

Expected behavior

I was expecting to get the table extracted correctly.

Here is the code I used to extract:

tables, layout, dim = self.extract_tables(linescale=30, flag_size= True)
acc_tables = self.filter_acc_tables(tables=tables, min_accuracy=85, max_whitespace=30)

PDF

Screenshots

Environment

• OS: Windows
• Python version: 3.9.12
• Numpy version: 1.22.3
• OpenCV version: 4.5.5.64
• Ghostscript version: 0.7
• Camelot version: 0.10.1

Link for PDF

https://www.irf.com/product-info/datasheets/data/irhm9150.pdf

[saidakyuz]

According to my observation, this happens only if the cells are spanned vertically and horizontally and there are some other cells that are not spanned horizontally on the same column with the cells two-dimensional spanned. Somehow each cell in the same row could have opposite values of vspan. (True or False) The issue caused by this attribute, but I still have no solution for it. For example following tables has the issue: