BUG: PDF size increases because of too high float writing precision

[pubpub-zz]

closes #1910
address regression from #2203

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (91b6dcd) 94.38% compared to head (d076f76) 94.38%.
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2213   +/-   ##
=======================================
  Coverage   94.38%   94.38%           
=======================================
  Files          43       43           
  Lines        7588     7589    +1     
  Branches     1497     1497           
=======================================
+ Hits         7162     7163    +1     
  Misses        262      262           
  Partials      164      164           

Files Changed	Coverage Δ
pypdf/generic/_base.py	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[MartinThoma]

I just made a little test:

import pikepdf
from os import stat
import pypdf.generic._base

INPUT_PDF = "pypdf/sample-files/009-pdflatex-geotopo/GeoTopo.pdf"
OUTPUT_PDF = "output_example.pdf"

pypdf.generic._base.FLOAT_WRITE_PRECISION = 1


def test_filesize(INPUT_PDF, OUTPUT_PDF):
    reader = pypdf.PdfReader(INPUT_PDF)

    writer = pypdf.PdfWriter(clone_from=reader)

    for page in writer.pages:
        page.compress_content_streams(level=9)

    with open(OUTPUT_PDF, 'wb') as f:
        writer.write(f)

    orig_size = stat(INPUT_PDF).st_size / (1024)
    pypdf_size = stat(OUTPUT_PDF).st_size / (1024)

    pikepdf.settings.set_flate_compression_level(9)

    with pikepdf.Pdf.open(INPUT_PDF,
                          allow_overwriting_input=True,
                          suppress_warnings=True) as pdf:

        pdf.save(OUTPUT_PDF,
                 object_stream_mode=pikepdf.
                 ObjectStreamMode.generate,
                 compress_streams=True,
                 stream_decode_level=pikepdf.
                 StreamDecodeLevel.specialized)

    pikepdf_size = stat(OUTPUT_PDF).st_size / (1024)

    print(f'{"input file:":<15}{INPUT_PDF}')
    print(f'{"original size:":<15}{orig_size:.4f} KB')
    print(f'{"pypdf size:":<15}{pypdf_size:.4f} KB')
    print(f'{"pikepdf size:":<15}{pikepdf_size:.4f} KB')

test_filesize(INPUT_PDF, OUTPUT_PDF)

The output is:

input file:    pypdf/sample-files/009-pdflatex-geotopo/GeoTopo.pdf
original size: 5196.4180 KB
pypdf size:    5574.2656 KB
pikepdf size:  5185.7314 KB

For comparison:

* pypdf before this PR                       : 5702.3506 KB
* pypdf with a presision of 8 (as in this PR): 5594.0645 KB
* Smallpdf                                   : 1793.479 KB

Most interesting is that I wasn't able to spot a difference.

[pubpub-zz]

From my analysis, you should not put a value lower than 5 : If you take some colorspace which are using some FloatObject, you may get some too big rounding that will affect the colors and which may not be visible on most of the pdf viewers.

[stefan6419846]

Some part of the comment seems to have got lost.