Use multiprocessing to parallely process PDF pages

[vinayak-mehta]

>>> camelot.read_pdf('filename.pdf', pages='all', parallel=True)

We could try and use all cores present on the machine using multiprocessing. More ideas are welcome.

[satheeshkatipomu]

Hi @vinayak-mehta ,

Even I thought of implementing this. dramatiq or celery are my suggestions for asynchronous processing of pages.

[jontis]

I'm doing this with dask but it's chosen out of habit.

[selcukusta]

Is there any improvement in there? I have a file that has only one page. The page has a table (25 rows x 13 columns). read_pdf function takes 10 seconds after that to_excel takes only 100-150 ms. I'm thinking about 10 seconds is too long, am I wrong?

[NixBiks]

@selcukusta I think this is more about running multiple pages at the same time rather than speeding up the extraction of a single page.

But does anyone have a solution for multiple pages in parallel?

[vinayak-mehta]

@selcukusta I think this is more about running multiple pages at the same time rather than speeding up the extraction of a single page.

Yes!

But does anyone have a solution for multiple pages in parallel?

Using multiprocessing, we should be able to distribute multiple pages on all cores, processing them in parallel.

[NixBiks]

I get this though

objc[53475]: +[__NSPlaceholderDate initialize] may have been in progress in another thread when fork() was called. We cannot safely call it or ignore it in the fork() child process. Crashing instead. Set a breakpoint on objc_initializeAfterForkError to debug.

Oh; what is the difference between https://github.com/atlanhq/camelot and https://github.com/camelot-dev/camelot ? Didn't notice two repos before now...

[selcukusta]

@selcukusta I think this is more about running multiple pages at the same time rather than speeding up the extraction of a single page.

But does anyone have a solution for multiple pages in parallel?

Yeah, I know. Actually it's related with that but the issue was closed and referenced to it.

[rawsh-bt]

Does anyone have an update? I've tried inheriting PageHandler and making pages multithreaded / multicore, and multi threading processing multiple pdfs, but I'm running into a ghostscript error (seems like it's not thread safe?)

[phoewass]

I did implement a multi-threading layer above camelot.read_pdf using multiprocessing library.
I faced a couple of pitfalls doing it, so I can help on this if I may.

[vinayak-mehta]

@phoewass That would be awesome if you're still interested!

[RickyGunawan09]

can anyone tell me how to use multiprocess in camelot ? or did this issues still on progress ?

[phoewass]

Hi all. Sorry it took me a while to publish the PR while the code was already available.
Now the PR is there to be reviewed, I'm looking forward for your feedback.

[vinayak-mehta]

👀

[Siddharth1India]

Any update on this? My PDFs are 100s of pages and I can really use this feature.

[mlbrothers]

@phoewass @vinayak-mehta is this feature part of library now? If not, is there any way I can utilize multiprocessing to read multipage PDF?

[deepakagrawal]

Any updates on this features?

[bosd]

Hey!

As camelot is dead, we try to build a maintained fork at pypdf_table_extraction.

There is a discussion about this in:

py-pdf#8 (reply in thread)