extract_tables() returning empty with explicit vertical lines

[juliamyndshft]

Hi, I'm having an issue with this pdf:
pa-matrix-2018.pdf

I'm using this code:

bounding_box = (0, 600, 900, 2455)
ts = {"vertical_strategy": "explicit",
         "horizontal_strategy": "text",
         "explicit_vertical_lines": "[150, 200, 850, 900]"}
cropped_page = pdf.pages[0].crop(bounding_box)
table = cropped_page.extract_tables(table_settings=ts)

I'm pretty sure the vertical lines are correct for grabbing the first and fourth columns on the page, but the table comes up empty. It's worth noting that when I use extract_text(), all the text in the document shows up fine, but I need it to be separated by column. What could be happening here?

Thanks!
Julia

[jsvine]

Hi @juliamyndshft, and thanks for your interest in this library. One issue I see is the quoting of the list in explicit_vertical_lines — instead of "[150, 200, 850, 900]", try [150, 200, 850, 900]. Does that fix things?

[juliamyndshft]

Hi @jsvine, thanks for your response! Apologies, that was a typo on my part; it's without the quotes in my code, so I do get the same result.

[jsvine]

Ah, looks like the lines might not quite be where you wanted them:

im = cropped_page.to_image()
im.reset().draw_vlines(ts["explicit_vertical_lines"])

...produces this:

I don't know what process you used to select the line positions, but one easy way to trial-and-error it is to use use the PageImage.draw_*(...) methods documented here.

[jsvine]

One other thing that will be handy is using PageImage.debug_tablefinder(...). E.g., the current settings:

im.reset().debug_tablefinder(ts)

... produce this:

[juliamyndshft]

Ah! I didn't realize the lines needed to be so tight on the edges. This worked with the explicit vertical lines of [175, 210, 850, 900]. Thanks so much!