diff options
Diffstat (limited to 'src/nvim/regexp_nfa.c')
| -rw-r--r-- | src/nvim/regexp_nfa.c | 41 |
1 files changed, 34 insertions, 7 deletions
diff --git a/src/nvim/regexp_nfa.c b/src/nvim/regexp_nfa.c index 21581d3823..2659eac762 100644 --- a/src/nvim/regexp_nfa.c +++ b/src/nvim/regexp_nfa.c @@ -85,6 +85,7 @@ enum { NFA_COMPOSING, /* Next nodes in NFA are part of the composing multibyte char */ NFA_END_COMPOSING, /* End of a composing char in the NFA */ + NFA_ANY_COMPOSING, // \%C: Any composing characters. NFA_OPT_CHARS, /* \%[abc] */ /* The following are used only in the postfix form, not in the NFA */ @@ -1350,6 +1351,10 @@ static int nfa_regatom(void) EMIT(NFA_VISUAL); break; + case 'C': + EMIT(NFA_ANY_COMPOSING); + break; + case '[': { int n; @@ -2259,6 +2264,7 @@ static void nfa_set_code(int c) case NFA_MARK_LT: STRCPY(code, "NFA_MARK_LT "); break; case NFA_CURSOR: STRCPY(code, "NFA_CURSOR "); break; case NFA_VISUAL: STRCPY(code, "NFA_VISUAL "); break; + case NFA_ANY_COMPOSING: STRCPY(code, "NFA_ANY_COMPOSING "); break; case NFA_STAR: STRCPY(code, "NFA_STAR "); break; case NFA_STAR_NONGREEDY: STRCPY(code, "NFA_STAR_NONGREEDY "); break; @@ -2716,6 +2722,7 @@ static int nfa_max_width(nfa_state_T *startstate, int depth) case NFA_NLOWER_IC: case NFA_UPPER_IC: case NFA_NUPPER_IC: + case NFA_ANY_COMPOSING: /* possibly non-ascii */ if (has_mbyte) len += 3; @@ -3714,6 +3721,7 @@ static int match_follows(nfa_state_T *startstate, int depth) continue; case NFA_ANY: + case NFA_ANY_COMPOSING: case NFA_IDENT: case NFA_SIDENT: case NFA_KWORD: @@ -3943,7 +3951,7 @@ skip_add: #endif switch (state->c) { case NFA_MATCH: - nfa_match = TRUE; + //nfa_match = TRUE; break; case NFA_SPLIT: @@ -4573,6 +4581,7 @@ static int failure_chance(nfa_state_T *state, int depth) case NFA_MATCH: case NFA_MCLOSE: + case NFA_ANY_COMPOSING: /* empty match works always */ return 0; @@ -4951,6 +4960,11 @@ static int nfa_regmatch(nfa_regprog_T *prog, nfa_state_T *start, regsubs_T *subm switch (t->state->c) { case NFA_MATCH: { + // If the match ends before a composing characters and + // ireg_icombine is not set, that is not really a match. + if (enc_utf8 && !ireg_icombine && utf_iscomposing(curc)) { + break; + } nfa_match = TRUE; copy_sub(&submatch->norm, &t->subs.norm); if (nfa_has_zsubexpr) @@ -5430,6 +5444,18 @@ static int nfa_regmatch(nfa_regprog_T *prog, nfa_state_T *start, regsubs_T *subm } break; + case NFA_ANY_COMPOSING: + // On a composing character skip over it. Otherwise do + // nothing. Always matches. + if (enc_utf8 && utf_iscomposing(curc)) { + add_off = clen; + } else { + add_here = TRUE; + add_off = 0; + } + add_state = t->state->out; + break; + /* * Character classes like \a for alpha, \d for digit etc. */ @@ -5769,12 +5795,13 @@ static int nfa_regmatch(nfa_regprog_T *prog, nfa_state_T *start, regsubs_T *subm if (!result && ireg_ic) result = vim_tolower(c) == vim_tolower(curc); - /* If there is a composing character which is not being - * ignored there can be no match. Match with composing - * character uses NFA_COMPOSING above. */ - if (result && enc_utf8 && !ireg_icombine - && clen != utf_char2len(curc)) - result = FALSE; + + // If ireg_icombine is not set only skip over the character + // itself. When it is set skip over composing characters. + if (result && enc_utf8 && !ireg_icombine) { + clen = utf_char2len(curc); + } + ADD_STATE_IF_MATCH(t->state); break; } |