Performance about paper2remarkable HOT 7 CLOSED

Hi @sirupsen, thanks for the bug reports! Yes, performance is a bit of an issue. The majority of the time is spent on cropping the pages of the pdf to remove excess white space. After some profiling it seems that the expensive part of that operation is creating an image of a pdf page using the pdfplumber package and I don't see a way to avoid that (the reason this is done in Python at all is that the bounding box returned by pdfcrop isn't very accurate (4cb3af8)).

I did just push a commit (a2e833a) that speeds up using the --center flag, but you don't use that in your example.

Alternatively, I could add a --no-crop flag, which would speed up the process but lead to less nice results. Would that help?

from paper2remarkable.

Might it be possible to remove whitespace without converting to an image first? 👂

from paper2remarkable.

I'm not sure, it might be necessary to render the pdf to figure out what the page looks like. That said, there is a pdfparser package that links directly to libpoppler, which seems a lot faster but is not a drop-in replacement for the current method with pdfplumber. I'll look into this a bit more and see what I can do.

from paper2remarkable.

@sirupsen Thanks again for reporting this issue! I've just pushed some changes that give about an 8x speedup. I'll prepare a new release of the package soon.

from paper2remarkable.

Excellent!!! I’ll upgrade as soon as it’s released.

I do think a no crop flag would be useful. Some may prefer to have tools (change pen, highlight, etc) visible at all times — or perhaps the crop could leave just enough space to always have the tools available on the left side?

from paper2remarkable.

I do think a no crop flag would be useful. Some may prefer to have tools (change pen, highlight, etc) visible at all times — or perhaps the crop could leave just enough space to always have the tools available on the left side?

Try version 0.5.4, both a --no-crop and a --right option are now available! :)

from paper2remarkable.

😍

from paper2remarkable.