Improve table recognition (renamed)

[rgosens2]

It cannot create html tables from tables in the pdf and totally skips 'em, including the text. Not very useful.

Tested on Intel manual (see: Attached file)

vol2a_2016_test1.pdf

[fmalina]

I added your document to the default test suite:

Before https://fmalina.github.io/unilex-transcript/tests/PDF/vol2a.pdf
After https://fmalina.github.io/unilex-transcript/tests/HTM/vol2a.htm

I see the tables are there including text, just search for "ASCII adjust AL after addition", which is the last cell in the first table. Table formatting was dropped for this table. I suppose grid_data for reconstruct_table didn't catch the grid.

[rgosens2]

Thanks for looking into it. I see the Agincourt House pdf gives proper tables, so it can be done. Intel's pdf must be a bit trickier.

[airtonix]

@fmalina are we saying that it does output <table/> or not ?

because i loaded up your example output there and it seems that tables end up like this:

[fmalina]

Yes, tables are supported and work in some examples. Obviously yours didn’t get recognised and there are various resons why it happens

[airtonix]

Yes, tables are supported and work in some examples. Obviously yours didn’t get recognised and there are various resons why it happens

Ok so i'm not saying there's a problem with my documents (there very well might be though).

What I am saying is:

your example is broken.

That screenshot is from your HTML output:

https://github.com/fmalina/unilex-transcript/blob/main/tests/HTM/vol2a.htm

there's no table tags in that html file.

I ask because your comment: #15 (comment)

seems to indicate that you're not aware that something is different now.