Stop font properties from leaking

[gmischler]

Fixes #359 (and possibly others)

Checklist:

• The GitHub pipeline is OK (green),
  meaning that both pylint (static code analyzer) and black (code formatter) are happy with the changes of this PR.

• A unit test is covering the code added / modified by this PR

• This PR is ready to be merged

• [na] In case of a new feature, docstrings have been added, with also some documentation in the docs/ folder

• A mention of the change is present in CHANGELOG.md

• The PR description or comment contains a picture of a cute animal (not mandatory but encouraged 🐸)

This solves the problems with leaking state after certain combinations of text output. It replaces several previous attempts to solve similar problems, and does so in a more straightforward and complete way.

• Using qQ when the last fragment uses a different font/style than the current setting sometimes increases the size, sometimes just moves the q to a different location, but often also reduces it because the following text doesn't need to set the font.
• In some cases, Tw was added seperately to the rest of the content. Fixing that moved it to a different location, but usually didn't change the length (sometimes it elminated a newline).
• Found a way to save 3 bytes per word with unicode fonts (merging the leading space with the following word).

Since there were worries about increasing the file size of the resulting PDFs, here's a rundown of the affected test files:

same size:

• moving Tw - render_styled_newpos.pdf
• moving Tw - multi_cell_newpos.pdf
• moving Tw - multi_cell_ln_newpos.pdf
• moving Tw - multi_cell_markdown_with_fill_color.pdf
• moving Tw - multi_cell_markdown_justified.pdf
• moving Tw - simple_outline.pdf
• moving qQ - graphics_context.pdf
• moving qQ - links.pdf
• moving qQ - link_with_zoom_and_shift.pdf
• moving qQ - html_features.pdf
• moving qQ - html_justify_paragraph.pdf
• moving qQ - flextemplate_rotation.pdf
• moving qQ - flextemplate_multipage.pdf
• moving qQ - flextemplate_elements.pdf
• moving qQ - template_nominal_hardcoded.pdf
• moving qQ - template_nominal_csv.pdf
• moving qQ - template_nominal_multipage.pdf
• moving qQ - template_nominal_justify.pdf

shorter

• moving Tw - 2_pages_outline.pdf
• moving Tw - ln_positioning_and_page_breaking_for_multicell.pdf
• replace F with qQ - cell_markdown_bleeding.pdf
• replace F with qQ - cell_markdown_right_aligned.pdf
• ttf words - multi_cell_j_paragraphs
• moving qQ - test_img_inside_html_table_centered_with_align.pdf

longer

• add qQ - multi_cell_markdown
• add qQ /ttf words - multi_cell_justified_with_unicode_fonts
• add qQ /ttf words - multi_cell_markdown_with_ttf_fonts

[codecov]

Codecov Report

Merging #384 (b1c877c) into master (2df5409) will increase coverage by 0.06%.
The diff coverage is 97.95%.

❗ Current head b1c877c differs from pull request most recent head a1f581d. Consider uploading reports for the commit a1f581d to get more accurate results

@@            Coverage Diff             @@
##           master     #384      +/-   ##
==========================================
+ Coverage   90.76%   90.82%   +0.06%     
==========================================
  Files          21       20       -1     
  Lines        6097     6041      -56     
  Branches     1247     1231      -16     
==========================================
- Hits         5534     5487      -47     
+ Misses        331      326       -5     
+ Partials      232      228       -4     

Impacted Files	Coverage Δ
fpdf/fpdf.py	86.65% <97.95%> (+0.21%)	⬆️
fpdf/html.py	95.21% <0.00%> (ø)
fpdf/recorder.py	88.88% <0.00%> (ø)
fpdf/graphics_state.py	100.00% <0.00%> (ø)
fpdf/enums.py
fpdf/util.py	86.20% <0.00%> (+5.25%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2df5409...a1f581d. Read the comment docs.

[Lucas-C]

This PR includes 88 commits:
you probably should rebase your branch on the master branch of this repo

[gmischler]

This PR includes 88 commits: you probably should rebase your branch on the master branch of this repo

I've tried, but Git Desktop repeatedly got confused and I had to cancel the attempt...

[Lucas-C]

I performed the rebase. Could you check that all is right @gmischler please?

Also, seems like many reference PDF files have been impacted by this change:
could you explain how / why?

[gmischler]

I performed the rebase. Could you check that all is right @gmischler please?

Also, seems like many reference PDF files have been impacted by this change: could you explain how / why?

I was just doing another rebase attempt myself again, which may have conflicted with yours...

There now seem to be a few files included that shouldn't be.
I'll have to comb through everything to make sure it's actually in a useable state.

And PDFs that aren't listed and explained at the top here should probably not be included.

[gmischler]

Ok, I think it's clean now.

[Lucas-C]

The code seems quite clearer after this PR, nice and good job!

Thank you very much for the rundown on the affected test files sizes.

However it does not seem totally accurate...
render_styled_newpos.pdf for example is 15bytes larger.
Would you know why?

[gmischler]

Thank you very much for the rundown on the affected test files sizes.

However it does not seem totally accurate... render_styled_newpos.pdf for example is 15bytes larger. Would you know why?

Interesting.
I had compared the uncompressed files, which are the same length. Apparently in this case the compression is slightly less effective for the new version. I would expect such an effect to average out in total, though.

[Lucas-C]

I'm OK to merge this once the conflicts are resolved

[Lucas-C]

Just a heads-up: I'm planning to remove the useless \n characters from the calls to FPDF._out() in local_context() and rect_clip() as part of #386 This will generate several minor changes in reference PDF files, so I'll wait for your PR to be merged first

[Lucas-C]

Merged: thank you @gmischler !