jqueguiner / num2words2 / 25217939620

95%

master: 94%

push

github

feat(fraction): add to_fraction + 'n/d' string routing (#113)

Implements savoirfairelinux/num2words#584 — fraction support across
the dispatcher.

Public API:
  - num2words('1/3', lang='en')        # 'one third'
  - num2words('-3/4', lang='fr')       # 'moins trois quarts'
  - num2words('1/100', lang='de')      # 'ein Hundertstel'
  - explicit:  to_fraction(num, denom) on any converter

Plumbing:
  - Num2Word_Base.to_fraction() default = '<cardinal num> <ordinal denom>[s]'
  - Dispatcher detects 'n/d' string pattern (with whitespace tolerance)
    and routes directly to converter.to_fraction(int, int)
  - 'fraction' added to CONVERTES_TYPES

Per-language idiomatic forms:
  - English:    half/halves, quarter/quarters; ordinal+s otherwise
  - French:     demi(s), tiers (invariant), quart(s); ordinal+s otherwise
  - Spanish:    medio(s), tercio(s), cuarto(s); 'un' for numerator 1
  - Italian:    mezzo/mezzi; ordinal -o → -i for plural
  - Portuguese: meio(s), terço(s); inherited by pt_BR
  - German:     halb/halbe + Drittel/Viertel/...el (invariant plural);
                'ein' rather than 'eins' for numerator 1; 'Hundertstel'
                drops the redundant 'ein' prefix from einhundert

Edge cases handled:
  - 0/3 → 'zero' (collapses to cardinal)
  - 1/1 → 'one' (denominator 1 collapses to cardinal)
  - 1/-3 and -1/3 both → 'minus one third'
  - 1/0 → ZeroDivisionError

Tests: tests/test_584_fractions.py — 21 cases covering all six
languages, idiomatic vs ordinal-fallback denominators, signs, edge
cases, and the dispatcher's pattern recognition (whitespace, currency
not falsely routed).

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

7988 of 8770 branches covered (91.08%)

Branch coverage included in aggregate %.

203 of 232 new or added lines in 9 files covered. (87.5%)

42165 of 43912 relevant lines covered (96.02%)

5.76 hits per line