Extracting table with no vertical lines (only horizontal lines) doesn't work

[jjjhill]

Describe the bug

extract_table with a table with no vertical lines fails to recognize the entire table. It cuts off some of the right side.

Have you tried repairing the PDF?

Yes

Code to reproduce the problem

table_settings = {
    "vertical_strategy": "text",
    "horizontal_strategy": "lines",
}
left = p.crop((0, 0, p.bbox[2]/2, p.bbox[3]))

img = left.to_image(resolution=400)
img.debug_tablefinder(table_settings)

PDF file

example3.pdf

Expected behavior

Entire table was captured.

Actual behavior

Partial table was captured.

Screenshots

Here is the result of debug_tablefinder:

NOTE: I've also tried cropping the image strictly to include the table only. Same result.

Environment

• pdfplumber version: 0.10.4
• Python version: 3.10.6
• OS: Windows

Additional context

I've tried all relevant parameters table_settings with no luck.
Any suggestions?

[jsvine]

Hi @jjjhill, and thanks for your interest in pdfplumber. The text strategy is, unfortunately, a bit finicky and layout-dependent. But I don't see an actual bug at play here, so I'm moving this to a discussion.

For this particular PDF, I'd recommend using the "explicit_vertical_lines": [...] setting, and then using some logic to determine what those lines should be. One common strategy is to examine the horizontal lines/rects, and using their x0/x1 positioning as input. You might also use the the text in the header to identify the x position that divides the two columns.