nx_hyphen.nx

nx_hyphen.nx -- LIANG HYPHENATION (the algorithm behind TeX's \hyphenation), integer-exact. WHY: nx_kp_break optimises the breakpoints it is GIVEN. Knuth-Plass in TeX is strictly stronger than the version I shipped first, because TeX also gets to break INSIDE words -- hyphenation points are extra candidate breaks fed into the same optimisation. Without them a narrow column has too few places to break and even an optimal breaker is forced into loose lines. This organ supplies those points. HOW (Liang 1983): wrap the word in dots (".hyphenation."), slide every pattern over it, and where a pattern matches, its digits vote on the inter-letter positions it covers. Each position takes the MAXIMUM vote it receives. An ODD final value means "a break is allowed here", even means "not allowed". Odd/even rather than a threshold is what lets a later, longer, more specific pattern VETO a shorter general one -- the mechanism that makes the pattern set composable instead of order-dependent. ★ SCOPE, STATED PLAINLY RATHER THAN IMPLIED: TeX's English pattern file holds 4447 patterns. This organ ships a SUBSET (see HY_PAT below). The gate therefore proves two DIFFERENT things and keeps them apart: (a) the ALGORITHM is correct -- it reproduces the exact expected points for words the loaded patterns cover, and it correctly refuses the ones they veto. That is a pass/fail claim. (b) COVERAGE over a wider word list is REPORTED AS A NUMBER with no verdict attached, because a subset cannot honestly be graded against a bar nobody calibrated. Reporting (b) as a verdict is how a partial capability starts being counted as a whole one. nx_hyphen word <word> -- show the hyphenation points nx_hyphen selftest -- the gate license_tier: ORIGINAL