Fix exception when guessing the AFM familyname #9760
Conversation
|
|
||
| import matplotlib.afm as afm | ||
|
|
||
|
|
||
| AFM_TEST_DATA = b"""StartFontMetrics 2.0 |
There was a problem hiding this comment.
Can we use one of the afm files in https://github.com/matplotlib/matplotlib/tree/master/lib/matplotlib/mpl-data/fonts ? I am (with out thinking about it too much) concerned about copyright on including this in the source.
Is there something about this afm file that triggers the bug that does not trigger it otherwise?
There was a problem hiding this comment.
These files are a bit lengthy in their CharMetrics block, which would make it a bit cumbersome to test their char metrics. I've changed the AFM_TEST_DATA to a shorter one with invented data. That should prevent any copyright issues.
| @@ -503,8 +505,8 @@ def get_familyname(self): | |||
|
|
|||
| # FamilyName not specified so we'll make a guess | |||
| name = self.get_fullname() | |||
| extras = (br'(?i)([ -](regular|plain|italic|oblique|bold|semibold|' | |||
| br'light|ultralight|extra|condensed))+$') | |||
| extras = (r'(?i)([ -](regular|plain|italic|oblique|bold|semibold|' | |||
There was a problem hiding this comment.
The issue is that what ever font was being used in the OP did not have a b'FamilyName' entry, hence hit this code block, where as most afm do have it and return from this function early?
There was a problem hiding this comment.
Yes. the re.sub would always fail. But it's not reached if the entry has a b'FamilyName'.
|
Other than my concern about copyright issues 👍 Thanks for working on this (and if I recall correctly you have at least one other PR in flight?). |
7c9e385 to
712c393
Compare
|
Thanks! |
|
This was milestoned for 2.2 but targetted |
|
Is there a reason, this was milestoned for 2.2 and not for 2.1.x? AFAICS, this is a pure bug fix without any side-effects. I don't see why it should not be released with the next bugfix release. |
|
as it is already on 2.1.x so re-milestoned. |
PR Summary
Fixes #8347.
As opposed to the suggestion in #8347, it's not the replacement argument that needs to be
bytes. Instead, the regexp has to bestrbecause all header values (here 'FullName') are converted to 'appropriate python types' (herestr).Cleaned up a docstring and added some tests on the way. The actual fix is just
afm.pyl.508fPR Checklist