Deprecate setting text kerning factor to any non-None value #30322
Conversation
github-actions
bot
added
topic: rcparams
topic: text/fonts
Documentation: user guide
QuLogic
force-pushed
the
deprecate-kerning
branch
2 times, most recently
from
76132af to
30280d0
Compare
| .. plot:: | ||
|
|
||
| import matplotlib.pyplot as plt | ||
|
|
||
| # Use old kerning values: | ||
| plt.rcParams['text.kerning_factor'] = 6 | ||
| fig, ax = plt.subplots() | ||
| ax.text(0.0, 0.05, 'BRAVO\nAWKWARD\nVAT\nW.Test', fontsize=56) | ||
| ax.set_title('Before (text.kerning_factor = 6)') | ||
|
|
||
| Note how the spacing between characters is uniform between their bounding boxes |
There was a problem hiding this comment.
I decided to remove this plot from the old what's new, as it will no longer show any difference between it and the plot below (with the correct results) once we switch to libraqm, and so it might be misleading.
| @@ -188,7 +191,7 @@ def test_ft2font_clear(): | |||
|
|
|||
| def test_ft2font_set_size(): | |||
| file = fm.findfont('DejaVu Sans') | |||
| font = ft2font.FT2Font(file, hinting_factor=1, _kerning_factor=1) | |||
There was a problem hiding this comment.
This being 1 seems to have been a typo. Since the test checks relative results, it made no difference here.
github-project-automation
bot
moved this to Waiting for other PR
in Font and text overhaul
QuLogic
moved this from Waiting for other PR
to Ready for Review
in Font and text overhaul
|
The rcParams depreciation machinery was removed per #29563. But you can still dreprecate rcParams if you want to
|
QuLogic
force-pushed
the
deprecate-kerning
branch
from
30280d0 to
99d357f
Compare
|
OK, I've added a deprecation for the rcParam directly, then. |
|
|
||
| Due to internal changes to support complex text rendering, the kerning factor on fonts is | ||
| no longer used. Setting the ``text.kerning_factor`` rcParam (which existed only for | ||
| backwards-compatibility) to any value other than 0 is deprecated, and the rcParam will be |
There was a problem hiding this comment.
It looks like in the deprecation warning it warns for a value !=None, not a value !=0. Am I missing something, or is this changelog entry wrong with respect to the current code?
There was a problem hiding this comment.
No, quite right. It originally warned on non-0, but once I added the rcParam as well, I switched to non-None instead.
This factor existed only to preserve test images, but as of matplotlib#29816, it is set to 0 (i.e., disabled and providing default behaviour). In the future, with libraqm, it will have no effect no matter its setting (because we won't be applying kerning ourselves at all.)
QuLogic
force-pushed
the
deprecate-kerning
branch
from
99d357f to
779af59
Compare
QuLogic
changed the title
PR summary
This factor existed only to preserve test images, but as of #29816, it is set to 0 (i.e., disabled and providing default behaviour). In the future, with libraqm, it will have no effect no matter its setting (because we won't be applying kerning ourselves at all.)
Since we seem to have removed most of the rcParam deprecation machinery, instead of warning from there, warn from the
FT2Fontconstructor, which is always passed_kerning_factorin the normal code path (i.e., anywhere outside of direct tests.) This does mean the warning won't occur directly where you set it, nor if you don't use any text, but I think that's a reasonable compromise.PR checklist