Improve Type-1 font parsing by jkseppan · Pull Request #20715 · matplotlib/matplotlib · GitHub

PR Summary

Parse font properties also from the encrypted part of the file, and reimplement the parsing so it understands more of PostScript's syntax. This fixes a bug where Type1Font.transform would not remove the UniqueID key but break some PostScript code referring to UniqueID instead.

Incidentally, fix the bug where every font had a weight property with value 'Normal' - the correct property is spelled Weight with a capital letter.

This is a prerequisite for subsetting Type-1 fonts (#127).

PR Checklist

Has pytest style unit tests (and pytest passes).
Is Flake 8 compliant (run flake8 on changed files to check).
New features are documented, with examples if plot related.
Documentation is sphinx and numpydoc compliant (the docs should build without error).
Conforms to Matplotlib style conventions (install flake8-docstrings and run flake8 --docstring-convention=all).
New features have an entry in doc/users/next_whats_new/ (follow instructions in README.rst there).
API changes documented in doc/api/next_api_changes/ (follow instructions in README.rst there).

With this I can produce smaller pdf files with usetex in some small tests, but this obviously needs more extensive testing, thus marking as draft. On top of matplotlib#20634 and matplotlib#20715. Closes matplotlib#127.

With this I can produce smaller pdf files with usetex in some small tests, but this obviously needs more extensive testing, thus marking as draft. On top of matplotlib#20715. Closes matplotlib#127.

Thanks for the review @anntzer!

Move Type1Font._tokens into a top-level function _tokenize that is a coroutine. The parsing stage consuming the tokens can instruct the tokenizer to return a binary token - this is necessary when decrypting the CharStrings and Subrs arrays, since the preceding context determines which parts of the data need to be decrypted. The function now also parses the encrypted portion of the font file. To support usage as a coroutine, move the whitespace filtering into the function, since passing the information about binary tokens would not easily work through a filter. The function now returns tokens as subclasses of a new _Token class, which carry the position and value of the token and can have token-specific helper methods. The position data will be needed when modifying the file, as the font is transformed or subsetted. A new helper function _expression can be used to consume tokens that form a balanced subexpression delimited by [] or {}. This helps fix a bug in UniqueID removal: if the font includes PostScript code that checks if the UniqueID is set in the current dictionary, the previous code broke that code instead of removing the UniqueID definition. Fonts can include UniqueID in the encrypted portion as well as the cleartext one, and removal is now done in both portions. Fix a bug related to font weight: the key is title-cased and not lower-cased, so font.prop['weight'] should not exist.

Type-1 fonts are required to have subroutines with specific contents but their names may vary. They are usually ND, NP and RD but names like | and |- appear too.

I added some tests and realized that the string escaping was slightly wrong. Now the code also parses string values, although it is unlikely that font properties would include escaped whitespace characters.

I admit I still haven't followed all the logic, but we can always check again later :)

With this I can produce smaller pdf files with usetex in some small tests, but this obviously needs more extensive testing, thus marking as draft. Give dviread.DviFont a fake filename attribute for character tracking. On top of matplotlib#20715. Closes matplotlib#127.

I guess that here you don't really care about handling the backslash escapes, and all you want to do is simply to match the (unescaped) parentheses, so you could perhaps just replace instring_re by something like (?<!\\)[()] (parentheses not preceded by a backslash)?

Ah, I gave this a try but it wouldn't work because the parenthesis can be preceded by backslashes if the backslash is itself escaped (i.e. one would really need to search for "parenthesis not preceded by an even number of backslashes"). So let's forget about this for now.

I'll merge based on @anntzer review. This should not go into 3.5 so that it has time to be used on master...

Improve Type-1 font parsing

With this I can produce smaller pdf files with usetex in some small tests, but this obviously needs more extensive testing, thus marking as draft. Give dviread.DviFont a fake filename attribute for character tracking. On top of matplotlib#20715. Closes matplotlib#127.

jkseppan force-pushed the type1-improved-parsing branch 4 times, most recently from 981ef67 to 90b5889 Compare July 22, 2021 12:46

jkseppan mentioned this pull request Jul 22, 2021

Type-1 font subsetting #20716

Merged

7 tasks

jklymak added topic: text/fonts status: waiting for other PR labels Jul 22, 2021

jkseppan force-pushed the type1-improved-parsing branch 2 times, most recently from e35728b to 9418b35 Compare July 22, 2021 15:59

jklymak removed the status: waiting for other PR label Jul 22, 2021

anntzer reviewed Aug 22, 2021

View reviewed changes