nx_stem.nx

buildroot/runtime/_hdl_build/nx_stem.nx

14381 B368 linesdepth 2pulls 2 transitivereach 75 importersview sourcekind library

docs dependencies structs consts functions

about

nx_stem.nx -- SOVEREIGN morphological stemming: FULL Porter (1980), the estate's ONE stemmer. HISTORY: Step-1a-only -> PORTER-LITE 2026-07-03 (1a/1b/1c + a step-4 subset + -ly) -> FULL PORTER 2026-08-13 (all 5 steps, first-match-stops inside each step, the paper's suffix tables complete). The gate's own HONEST line named "fuller Porter steps" as the extension; the trigger was nx_beir_eval measuring our BM25 at nDCG@10 305 permil vs the stemmed Anserini baseline 320 -- its header named stemming as the gap. CONSUMERS (grep-verified 2026-08-13, coverage_complete=1): - nx_docportal_search_seg.nx: QUERY-TIME dictionary stem-expansion (both sides computed per query, nothing persisted -> a stemmer upgrade shifts conflation keys CONSISTENTLY; the change is DORMANT in the live daemon until its next rebuild -- RE-RUN THE MRR RULER on that rebuild). - nx_stem_gate.nx (R-UX-2): every pre-upgrade vector verified to hold under full Porter. - nx_beir_eval.nx: index+query tokenizer (the BEIR stemming rung). HONEST SCOPE: no irregulars (children->child needs a lexicon). Buffer contract: step 1b can APPEND one byte (-e restore) and nx_stem null-terminates, so out needs n+2 capacity. exports: nx_stem (buffered) + vr_stem (in-place, backward-compat) + pst_stem (in-place core). license_tier: ORIGINAL

dependencies 1 imports · 4 importers

imports: nx_syscalls.nx

imported by: nx_beir_eval.nxnx_docportal_search_seg.nxnx_stem_bound_gate.nxnx_stem_gate.nx

structs

none

consts

none

functions

22	func pst_cons(b: *u8, i: i64) -> i64 called by 5: pst_cons pst_m pst_vowelin pst_doublec pst_cvc calls 1: pst_cons
37	func pst_m(b: *u8, j: i64) -> i64 called by 7: pst_step1b2 pst_step1b pst_step2 pst_step3 pst_step4 pst_step5a+1 calls 1: pst_cons
52	func pst_vowelin(b: *u8, j: i64) -> i64 called by 2: pst_step1b pst_step1c calls 1: pst_cons
57	func pst_doublec(b: *u8, j: i64) -> i64 called by 2: pst_step1b2 pst_step5b calls 1: pst_cons
63	func pst_cvc(b: *u8, j: i64) -> i64 called by 2: pst_step1b2 pst_step5a calls 1: pst_cons
75	func pst_ends(b: u8, k: i64, s: u8, sl: i64) -> i64 called by 7: pst_step1a pst_step1b2 pst_step1b pst_step1c pst_step2 pst_step3+1
81	func pst_set(b: u8, js: i64, s: u8, sl: i64) -> i64 called by 3: pst_step1b2 pst_step2 pst_step3
87	func pst_step1a(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 1: pst_ends
103	func pst_step1b2(b: *u8, k: i64) -> i64 called by 1: pst_step1b calls 5: pst_ends pst_set pst_doublec pst_m pst_cvc
124	func pst_step1b(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 4: pst_ends pst_m pst_vowelin pst_step1b2
145	func pst_step1c(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 2: pst_ends pst_vowelin
151	func pst_step2(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 3: pst_ends pst_m pst_set
234	func pst_step3(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 3: pst_ends pst_m pst_set
262	func pst_step4(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 2: pst_ends pst_m
329	func pst_step5a(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 2: pst_m pst_cvc
336	func pst_step5b(b: *u8, k: i64) -> i64 called by 1: pst_stem calls 2: pst_doublec pst_m
343	func pst_stem(b: *u8, n: i64) -> i64 called by 2: be_tok nx_stem calls 8: pst_step1a pst_step1b pst_step1c pst_step2 pst_step3 pst_step4+2
358	func nx_stem(inp: u8, n0: i64, out: u8) -> i64 called by 3: dss_stem_expand vr_stem main calls 1: pst_stem
366	func vr_stem(w: *u8, n: i64) -> i64 called by 2: st_case main calls 1: nx_stem