how to extract table with invisible lines

[cqluohong]

Like the table below, there are 4 columns per row, but I can't get the correct results when I use pdfpumber to extract.

[cqluohong]

[OisinMoran]

Have you tried cropping to just the table you want to extract before running the extraction?
Another thing to try would be playing around with the parameters (it looks like increasing the snap_tolerance might work here).

[cqluohong]

Have you tried cropping to just the table you want to extract before running the extraction?
Another thing to try would be playing around with the parameters (it looks like increasing the snap_tolerance might work here).

yeah,I try,in vertical direction,Extracting the form is OK,is there ocr technique in pdfplumer?I also try camelot that use it ,but handle Chinese,there is a bug,I am confused!

[OisinMoran]

So pdfplumber "works best on machine-generated, rather than scanned, PDFs". There is no OCR capability. If you can share the actual PDF you are trying to extract a table from it can help with debugging the issue.

[cqluohong]

So pdfplumber "works best on machine-generated, rather than scanned, PDFs". There is no OCR capability. If you can share the actual PDF you are trying to extract a table from it can help with debugging the issue.

430027-北科光大-2017年年度报告.pdf
I want to extract tableswith page 47- 57

[luoqygit]

If you don't mind using an alpha version, you can switch to 0.6-alpha, and use 'snap_y_tolerance' when calling 'page.extract_tables()'.
tabs = page.extract_tables( { 'snap_y_tolerance': 6} )
You can replace 6 with whatever value you want.

Please refer to 0.6-alpha documents and #51 for details.

[jsvine]

Thanks @luoqygit and @OisinMoran! Cleaning up old issues. Feel free to reopen @cqluohong if you'd like to continue the discussion.

[clj55]

Hi I'm facing issues extracting the invisible tables too. I can't crop the page to specific coordinates of the table because I'm running the program on multiple PDFs where the table can appear in different positions.

Customer Information.pdf

I tried it with these table_settings:

table_settings={
    "vertical_strategy":"text",
                "text_keep_blank_chars":True,
                "horizontal_strategy":"text",
                }

But it recognised the paragraph text as a table too