jqueguiner / num2words2 / 25196818773 / 1

94%

main: 95%

push

github

Korean: fix ordinal conversion KeyError on hundreds (100, 200…) (#34)

Ports savoirfairelinux/num2words#556 by @dogzz9445.

The old to_ordinal logic split outwords on "백" (hundred) and assumed
the trailing slice contained either a tens word ("십") or a single-digit
ordinal-mappable form. For values like 100 / 200 the trailing slice was
empty, raising KeyError.

The new logic uses a positive-lookbehind regex split that keeps the
"천"/"백" markers attached to the preceding tokens, falls back to the
identity mapping when an entry is missing, and rejoins without inserting
extra spaces.

Test data updated for the new rejoin semantics:
- 101 → "백한 번째" (was "백 한 번째")
- 137 → "백서른일곱 번째" (was "백 서른일곱 번째")

Original PR: https://github.com/savoirfairelinux/num2words/pull/556
(Resolves upstream #555.)

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

7284 of 8308 branches covered (87.67%)

Branch coverage included in aggregate %.

40338 of 42103 relevant lines covered (95.81%)

0.96 hits per line