To: vim_dev@googlegroups.com
Subject: Patch 7.3.1011
Fcc: outbox
From: Bram Moolenaar <Bram@moolenaar.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
------------
Patch 7.3.1011
Problem: New regexp engine is inefficient with multi-byte characters.
Solution: Handle a character at a time instead of a byte at a time. Also
make \Z partly work.
Files: src/regexp_nfa.c, src/testdir/test95.in, src/testdir/test95.ok
*** ../vim-7.3.1010/src/regexp_nfa.c 2013-05-24 20:25:28.000000000 +0200
--- src/regexp_nfa.c 2013-05-24 21:49:43.000000000 +0200
***************
*** 46,54 ****
NFA_NCLOSE, /* End of subexpr. marked with \%( ... \) */
NFA_START_INVISIBLE,
NFA_END_INVISIBLE,
- NFA_MULTIBYTE, /* Next nodes in NFA are part of the same
- multibyte char */
- NFA_END_MULTIBYTE, /* End of multibyte char in the NFA */
NFA_COMPOSING, /* Next nodes in NFA are part of the
composing multibyte char */
NFA_END_COMPOSING, /* End of a composing char in the NFA */
--- 46,51 ----
***************
*** 195,220 ****
*post_ptr++ = c; \
} while (0)
- #define EMIT_MBYTE(c) \
- len = (*mb_char2bytes)(c, buf); \
- EMIT(buf[0]); \
- for (i = 1; i < len; i++) \
- { \
- EMIT(buf[i]); \
- EMIT(NFA_CONCAT); \
- } \
- EMIT(NFA_MULTIBYTE);
-
- #define EMIT_COMPOSING_UTF(input) \
- len = utfc_ptr2len(input); \
- EMIT(input[0]); \
- for (i = 1; i < len; i++) \
- { \
- EMIT(input[i]); \
- EMIT(NFA_CONCAT); \
- } \
- EMIT(NFA_COMPOSING);
-
/*
* Initialize internal variables before NFA compilation.
* Return OK on success, FAIL otherwise.
--- 192,197 ----
***************
*** 611,618 ****
#ifdef FEAT_MBYTE
char_u *old_regparse = regparse;
int clen;
- int len;
- static char_u buf[30];
int i;
#endif
int extra = 0;
--- 588,593 ----
***************
*** 845,858 ****
return FAIL;
c = coll_get_char();
! #ifdef FEAT_MBYTE
! if ((*mb_char2len)(c) > 1)
! {
! EMIT_MBYTE(c);
! }
! else
! #endif
! EMIT(c);
break;
/* Catch \%^ and \%$ regardless of where they appear in the
--- 820,826 ----
return FAIL;
c = coll_get_char();
! EMIT(c);
break;
/* Catch \%^ and \%$ regardless of where they appear in the
***************
*** 1135,1146 ****
* skip it. */
for (c = startc + 1; c <= endc; c++)
{
! if ((*mb_char2len)(c) > 1)
! {
! EMIT_MBYTE(c);
! }
! else
! EMIT(c);
TRY_NEG();
EMIT_GLUE();
}
--- 1103,1109 ----
* skip it. */
for (c = startc + 1; c <= endc; c++)
{
! EMIT(c);
TRY_NEG();
EMIT_GLUE();
}
***************
*** 1187,1200 ****
if (got_coll_char == TRUE && startc == 0)
EMIT(0x0a);
else
! #ifdef FEAT_MBYTE
! if ((*mb_char2len)(startc) > 1)
! {
! EMIT_MBYTE(startc);
! }
! else
! #endif
! EMIT(startc);
TRY_NEG();
EMIT_GLUE();
}
--- 1150,1156 ----
if (got_coll_char == TRUE && startc == 0)
EMIT(0x0a);
else
! EMIT(startc);
TRY_NEG();
EMIT_GLUE();
}
***************
*** 1242,1271 ****
int plen;
nfa_do_multibyte:
! /* length of current char, with composing chars,
! * from pointer */
! plen = (*mb_ptr2len)(old_regparse);
! if (enc_utf8 && clen != plen)
! {
! /* A composing character is always handled as a
! * separate atom, surrounded by NFA_COMPOSING and
! * NFA_END_COMPOSING. Note that right now we are
* building the postfix form, not the NFA itself;
* a composing char could be: a, b, c, NFA_COMPOSING
! * where 'a', 'b', 'c' are chars with codes > 256.
! */
! EMIT_COMPOSING_UTF(old_regparse);
regparse = old_regparse + plen;
}
else
- /* A multi-byte character is always handled as a
- * separate atom, surrounded by NFA_MULTIBYTE and
- * NFA_END_MULTIBYTE */
- if (plen > 1)
- {
- EMIT_MBYTE(c);
- }
- else
#endif
{
c = no_Magic(c);
--- 1198,1227 ----
int plen;
nfa_do_multibyte:
! /* Length of current char with composing chars. */
! if (enc_utf8 && clen != (plen = (*mb_ptr2len)(old_regparse)))
! {
! /* A base character plus composing characters.
! * This requires creating a separate atom as if enclosing
! * the characters in (), where NFA_COMPOSING is the ( and
! * NFA_END_COMPOSING is the ). Note that right now we are
* building the postfix form, not the NFA itself;
* a composing char could be: a, b, c, NFA_COMPOSING
! * where 'b' and 'c' are chars with codes > 256. */
! i = 0;
! for (;;)
! {
! EMIT(c);
! if (i > 0)
! EMIT(NFA_CONCAT);
! if (i += utf_char2len(c) >= plen)
! break;
! c = utf_ptr2char(old_regparse + i);
! }
! EMIT(NFA_COMPOSING);
regparse = old_regparse + plen;
}
else
#endif
{
c = no_Magic(c);
***************
*** 1702,1710 ****
case NFA_START_INVISIBLE: STRCPY(code, "NFA_START_INVISIBLE"); break;
case NFA_END_INVISIBLE: STRCPY(code, "NFA_END_INVISIBLE"); break;
- case NFA_MULTIBYTE: STRCPY(code, "NFA_MULTIBYTE"); break;
- case NFA_END_MULTIBYTE: STRCPY(code, "NFA_END_MULTIBYTE"); break;
-
case NFA_COMPOSING: STRCPY(code, "NFA_COMPOSING"); break;
case NFA_END_COMPOSING: STRCPY(code, "NFA_END_COMPOSING"); break;
--- 1658,1663 ----
***************
*** 2194,2200 ****
}
e1 = POP();
e1.start->negated = TRUE;
! if (e1.start->c == NFA_MULTIBYTE || e1.start->c == NFA_COMPOSING)
e1.start->out1->negated = TRUE;
PUSH(e1);
break;
--- 2147,2153 ----
}
e1 = POP();
e1.start->negated = TRUE;
! if (e1.start->c == NFA_COMPOSING)
e1.start->out1->negated = TRUE;
PUSH(e1);
break;
***************
*** 2311,2316 ****
--- 2264,2279 ----
PUSH(frag(s, list1(&s1->out)));
break;
+ case NFA_COMPOSING: /* char with composing char */
+ #if 0
+ /* TODO */
+ if (regflags & RF_ICOMBINE)
+ {
+ goto normalchar;
+ }
+ #endif
+ /* FALLTHROUGH */
+
case NFA_MOPEN + 0: /* Submatch */
case NFA_MOPEN + 1:
case NFA_MOPEN + 2:
***************
*** 2322,2329 ****
case NFA_MOPEN + 8:
case NFA_MOPEN + 9:
case NFA_NOPEN: /* \%( "Invisible Submatch" */
- case NFA_MULTIBYTE: /* mbyte char */
- case NFA_COMPOSING: /* composing char */
if (nfa_calc_size == TRUE)
{
nstate += 2;
--- 2285,2290 ----
***************
*** 2336,2344 ****
case NFA_NOPEN:
mclose = NFA_NCLOSE;
break;
- case NFA_MULTIBYTE:
- mclose = NFA_END_MULTIBYTE;
- break;
case NFA_COMPOSING:
mclose = NFA_END_COMPOSING;
break;
--- 2297,2302 ----
***************
*** 2377,2385 ****
goto theend;
patch(e.out, s1);
! if (mopen == NFA_MULTIBYTE || mopen == NFA_COMPOSING)
! /* MULTIBYTE->out1 = END_MULTIBYTE
! * COMPOSING->out1 = END_COMPOSING */
patch(list1(&s->out1), s1);
PUSH(frag(s, list1(&s1->out)));
--- 2335,2342 ----
goto theend;
patch(e.out, s1);
! if (mopen == NFA_COMPOSING)
! /* COMPOSING->out1 = END_COMPOSING */
patch(list1(&s->out1), s1);
PUSH(frag(s, list1(&s1->out)));
***************
*** 2540,2556 ****
case NFA_COMPOSING:
/* nfa_regmatch() will match all the bytes of this composing char. */
break;
-
- case NFA_MULTIBYTE:
- /* nfa_regmatch() will match all the bytes of this multibyte char. */
- break;
#endif
- case NFA_END_MULTIBYTE:
- /* Successfully matched this mbyte char */
- addstate(l, state->out, m, off, lid, match);
- break;
-
case NFA_NOPEN:
case NFA_NCLOSE:
addstate(l, state->out, m, off, lid, match);
--- 2497,2504 ----
***************
*** 2841,2847 ****
regsub_T *submatch;
regsub_T *m;
{
! int c = -1;
int n;
int i = 0;
int result;
--- 2789,2795 ----
regsub_T *submatch;
regsub_T *m;
{
! int c;
int n;
int i = 0;
int result;
***************
*** 2859,2865 ****
List *listtbl[2][2];
List *ll;
int listid = 1;
- int endnode;
List *thislist;
List *nextlist;
List *neglist;
--- 2807,2812 ----
***************
*** 3190,3222 ****
break;
}
! case NFA_MULTIBYTE:
case NFA_COMPOSING:
! endnode = t->state->c + 1;
result = OK;
sta = t->state->out;
! len = 1;
! while (sta->c != endnode && len <= n)
{
! if (reginput[len-1] != sta->c)
! {
! result = FAIL;
break;
! }
! len++;
sta = sta->out;
}
/* if input char length doesn't match regexp char length */
! if (len -1 < n || sta->c != endnode)
result = FAIL;
! end = t->state->out1; /* NFA_END_MULTIBYTE or
! NFA_END_COMPOSING */
/* If \Z was present, then ignore composing characters */
! if (ireg_icombine && endnode == NFA_END_COMPOSING)
result = 1 ^ sta->negated;
ADD_POS_NEG_STATE(end);
break;
case NFA_NEWL:
if (!reg_line_lbr && REG_MULTI
--- 3137,3171 ----
break;
}
! #ifdef FEAT_MBYTE
case NFA_COMPOSING:
! {
! int mc = c;
!
result = OK;
sta = t->state->out;
! len = 0;
! while (sta->c != NFA_END_COMPOSING && len < n)
{
! if (len > 0)
! mc = mb_ptr2char(reginput + len);
! if (mc != sta->c)
break;
! len += mb_char2len(mc);
sta = sta->out;
}
/* if input char length doesn't match regexp char length */
! if (len < n || sta->c != NFA_END_COMPOSING)
result = FAIL;
! end = t->state->out1; /* NFA_END_COMPOSING */
/* If \Z was present, then ignore composing characters */
! if (ireg_icombine)
result = 1 ^ sta->negated;
ADD_POS_NEG_STATE(end);
break;
+ }
+ #endif
case NFA_NEWL:
if (!reg_line_lbr && REG_MULTI
***************
*** 3425,3430 ****
--- 3374,3387 ----
if (!result)
result = ireg_ic == TRUE
&& MB_TOLOWER(t->state->c) == MB_TOLOWER(c);
+ #ifdef FEAT_MBYTE
+ /* 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
+ && n != utf_char2len(c))
+ result = FALSE;
+ #endif
ADD_POS_NEG_STATE(t->state);
break;
}
*** ../vim-7.3.1010/src/testdir/test95.in 2013-05-24 20:25:28.000000000 +0200
--- src/testdir/test95.in 2013-05-24 20:45:08.000000000 +0200
***************
*** 35,40 ****
--- 35,44 ----
:call add(tl, ['\f\+', '&*fname ', 'fname'])
:call add(tl, ['\%#=1\f\+', '&*fname ', 'fname'])
+ :"""" Test composing character matching
+ :call add(tl, ['.ม', 'xม่x yมy', 'yม'])
+ :call add(tl, ['.ม่', 'xม่x yมy', 'xม่'])
+
:"""" Test \Z
:call add(tl, ['ú\Z', 'x'])
*** ../vim-7.3.1010/src/testdir/test95.ok 2013-05-24 20:25:28.000000000 +0200
--- src/testdir/test95.ok 2013-05-24 20:44:41.000000000 +0200
***************
*** 9,13 ****
--- 9,15 ----
OK - \%#=1\i\+
OK - \f\+
OK - \%#=1\f\+
+ OK - .ม
+ OK - .ม่
OK - ú\Z
OK - [^[=a=]]\+
*** ../vim-7.3.1010/src/version.c 2013-05-24 20:25:28.000000000 +0200
--- src/version.c 2013-05-24 21:56:02.000000000 +0200
***************
*** 730,731 ****
--- 730,733 ----
{ /* Add new patch number below this line */
+ /**/
+ 1011,
/**/
--
If you had to identify, in one word, the reason why the
human race has not achieved, and never will achieve, its
full potential, that word would be "meetings."
/// Bram Moolenaar -- Bram@Moolenaar.net -- http://www.Moolenaar.net \\\
/// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/ \\\
\\\ an exciting new programming language -- http://www.Zimbu.org ///
\\\ help me help AIDS victims -- http://ICCF-Holland.org ///