Latest Results
Bugfix Virama, matching uucode (ghostty) (#222)
After careful review of Virama width calculation and study of uucode, two related changes:
- Bugfix Invisible_Stacker viramas now form conjuncts (Burmese, Khmer, etc.)
Discovered while analyzing Burmese, found in ``IndicSyllabicCategory.txt``, "causes conjunct
formation", added in Unicode 14.0:
# Indic_Syllabic_Category=Invisible_Stacker
#
# Invisible stacker (usually kills inherent vowel of consonant; is not visible
# by itself; causes conjunct formation or consonant
# stacking)
So this logically belongs with the others in ``_ISC_VIRAMA_SET``. Use code generation instead of
the copy and paste done before, this set would have changed in several past unicode releases and
so may change again in the future.
- Change Virama+consonant width calculation to match @jacobsandlund's uucode (ghostty)
Our previous implementation was perhaps too complex, but more "giving" for width compared to
uucode's implementation, and neither of ours really has any basis in standards, seems to be of
taste. contour's libunicode does not make concern for virama, and uucode's does, measurement
is simpler to implement and will probably improve performance, see code comment::
# Virama codepoints (Indic_Syllabic_Category=Virama|Invisible_Stacker)
# are treated as zero-width combining marks, matching Ghostty's uucode
# library. Whether a virama+consonant sequence should collapse to 1 or
# 2 cells is undefined by Unicode; we follow uucode's (1). Latest Branches
-16%
+14%
+40%
jq/python-315-lazy-imports © 2026 CodSpeed Technology