BUG: Format floats using their intrinsic decimal precision

[programmarchy]

Since FloatObject is represented as a decimal, format numbers using their intrinsic precision, instead of reducing the precision to 5 decimal places.

This fixes rendering issues for PDFs that contain coordinates, transformations, etc. with real numbers containing more than 5 decimal places of precision. For example, PDFs exported from Microsoft PowerPoint contain numbers with up to 11 decimal places.

Fixes: #1266

[codecov]

Codecov Report

Base: 94.63% // Head: 94.63% // Increases project coverage by +0.00% 🎉

Coverage data is based on head (6b7f7ef) compared to base (3cf80bf).
Patch coverage: 100.00% of modified lines in pull request are covered.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1267   +/-   ##
=======================================
  Coverage   94.63%   94.63%           
=======================================
  Files          30       30           
  Lines        5140     5141    +1     
  Branches     1058     1058           
=======================================
+ Hits         4864     4865    +1     
  Misses        164      164           
  Partials      112      112           

Impacted Files	Coverage Δ
PyPDF2/generic/_base.py	100.00% <ø> (ø)
PyPDF2/_writer.py	91.06% <100.00%> (+0.01%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[programmarchy]

Rebased this PR so tests are passing, and believe all the changes requested in the last review have been addressed.

Could you take another look, @MasterOdin?

[MartinThoma]

It looks good from my side - thank you for writing a unit test 🤗

[MartinThoma]

@MasterOdin You were a lot more involved in this PR than I was. What do you think?

[MasterOdin]

Looks good. Thanks for the work here @programmarchy! 👍

[MartinThoma]

Thank you both for all the work you put in it 🙏

I've just merged the PR and I will release it today to PyPI :-)

[MartinThoma]

@programmarchy If you want, I can add you to https://pypdf2.readthedocs.io/en/latest/meta/CONTRIBUTORS.html :-)

[programmarchy]

@programmarchy If you want, I can add you to https://pypdf2.readthedocs.io/en/latest/meta/CONTRIBUTORS.html :-)

That would be very cool, thank you @MartinThoma!

[mrknwk]

When you use page.scale_by() with a non-integer value in 2.10.9, Acrobat (22.002.20212) displays the transformed PDF pages as empty square drawing areas. Maybe one could leave the default precision at less than 20 digits (this is the Acrobat tipping point, I guess) for compatibility reasons and make it configurable?

[programmarchy]

Interesting, would you mind sharing how you came to find out 20 digits is the tipping point for Acrobat, @mrknwk?

One way I was thinking of to make this configurable would be to adopt context vars as implemented in decimal.Context for example. The context provides sane defaults with a central point for changing behavior.

It would allow us to write something like:

import PyPDF2
from PyPDF2 import PdfReader, PdfWriter
from PyPDF2.context import Context, StripExtraTrailingZeros, QuantizeInteger

ctx = StreamContext()
ctx.max_prec = 5  # specify maximum precision
ctx.flags = [
    StripExtraTrailingZeros,
    QuantizeInteger,
]  # could also specify additional format flags
PyPDF2.setcontext(ctx)

reader = PdfReader("./path/to/file.pdf")
reader.pages[0].scale_by(0.5)
writer = PdfWriter()
writer.add_page(reader.pages[0])
...

Or like this:

with PyPDF2.localcontext() as ctx:
    ctx.max_prec = 5  # specify maximum precision
    ctx.flags = [
        StripExtraTrailingZeros,
        QuantizeInteger,
    ]  # could also specify additional format flags
    ...

Or maybe this:

PyPDF2.setcontext(AdobeAcrobactContext)

Might make sense to open a separate issue to discuss further.

[mrknwk]

@programmarchy It really was just trial and error. 😊

But 20 digits is also the limit that one of the maintainers of PDF Arranger found in a test. He contacted Adobe about it and apparently it is an Acrobat "implementation level limitation". So maybe the third option would be a nice way to go then.

[programmarchy]

@mrknwk I'd be happy to take a stab at implementing the above. Could you please create a GitHub issue with a corresponding sample PDF, and tag me?