Bugfix of baseline_chars (missing comma)#87
Conversation
I downloaded it, extracted the zip, and did a diff between the PNGs, but there is no difference... |
pelson
left a comment
pelson
left a comment
There was a problem hiding this comment.
(Kerning is fluctuating, but it will be fixed in a separate PR.fixed.)
Now that I look deeper, I note that if we're auto-kerning all the letters then we should just go straight to fix the bounding box of the character itself, and leave kerning for specific pairs...
I have a few changes after this one that I'll propose in a PR.
|
|
||
|
|
||
| baseline_chars = ['a' 'e', 'm', 'A', 'E', 'M', '&', '@', '.', u'≪', u'É'] | ||
| baseline_chars = ['a', 'e', 'm', 'A', 'E', 'M', '&', '@', '.', u'≪', u'É'] |
There was a problem hiding this comment.
One little character missing has quite the effect... !
| fig = render_block(label, rows) | ||
| out = os.path.join(OUTDIR, f'charmap_{slugify(label)}.png') | ||
| fig.savefig(out, dpi=150, bbox_inches='tight', facecolor=fig.get_facecolor()) | ||
| Image.open(out).save(out) |
There was a problem hiding this comment.
Previously there was the matplotlib version in the metadata, which was a source of non-reproducible output (the version changed recently). Here, I drop that metadata.
2 participants
Fixes #67.
['a' 'e', ...]was interpreted as['ae', ...]and did not match a and e.With this fix, the recognition of the baseline is improved, and the variation within
_spansis also reduced. As a result, c and s get closer to the baseline, and their positions at low DPI are improved.(Kerning is fluctuating,
but it will be fixed in a separate PR.fixed.)