#pragma alloca
#endif
-#undef _GNU_SOURCE
-#define _GNU_SOURCE
-
#ifdef HAVE_CONFIG_H
# include <config.h>
#endif
# define POS_AS_IN_BUFFER(p) ((p) + (NILP (re_match_object) || BUFFERP (re_match_object)))
# define RE_MULTIBYTE_P(bufp) ((bufp)->multibyte)
+# define RE_TARGET_MULTIBYTE_P(bufp) ((bufp)->target_multibyte)
# define RE_STRING_CHAR(p, s) \
(multibyte ? (STRING_CHAR (p, s)) : (*(p)))
# define RE_STRING_CHAR_AND_LENGTH(p, s, len) \
(multibyte ? (STRING_CHAR_AND_LENGTH (p, s, len)) : ((len) = 1, *(p)))
-/* Set C a (possibly multibyte) character before P. P points into a
- string which is the virtual concatenation of STR1 (which ends at
- END1) or STR2 (which ends at END2). */
-# define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
- do { \
- if (multibyte) \
- { \
- re_char *dtemp = (p) == (str2) ? (end1) : (p); \
- re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \
- while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \
- c = STRING_CHAR (dtemp, (p) - dtemp); \
- } \
- else \
- (c = ((p) == (str2) ? (end1) : (p))[-1]); \
+/* Set C a (possibly converted to multibyte) character before P. P
+ points into a string which is the virtual concatenation of STR1
+ (which ends at END1) or STR2 (which ends at END2). */
+# define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
+ do { \
+ if (multibyte) \
+ { \
+ re_char *dtemp = (p) == (str2) ? (end1) : (p); \
+ re_char *dlimit = ((p) > (str2) && (p) <= (end2)) ? (str2) : (str1); \
+ while (dtemp-- > dlimit && !CHAR_HEAD_P (*dtemp)); \
+ c = STRING_CHAR (dtemp, (p) - dtemp); \
+ } \
+ else \
+ { \
+ (c = ((p) == (str2) ? (end1) : (p))[-1]); \
+ MAKE_CHAR_MULTIBYTE (c); \
+ } \
} while (0)
+/* Set C a (possibly converted to multibyte) character at P, and set
+ LEN to the byte length of that character. */
+# define GET_CHAR_AFTER(c, p, len) \
+ do { \
+ if (multibyte) \
+ c = STRING_CHAR_AND_LENGTH (p, 0, len); \
+ else \
+ { \
+ c = *p; \
+ len = 1; \
+ MAKE_CHAR_MULTIBYTE (c); \
+ } \
+ } while (0)
#else /* not emacs */
# define CHARSET_LEADING_CODE_BASE(c) 0
# define MAX_MULTIBYTE_LENGTH 1
# define RE_MULTIBYTE_P(x) 0
+# define RE_TARGET_MULTIBYTE_P(x) 0
# define WORD_BOUNDARY_P(c1, c2) (0)
# define CHAR_HEAD_P(p) (1)
# define SINGLE_BYTE_CHAR_P(c) (1)
# define SAME_CHARSET_P(c1, c2) (1)
# define MULTIBYTE_FORM_LENGTH(p, s) (1)
+# define PREV_CHAR_BOUNDARY(p, limit) ((p)--)
# define STRING_CHAR(p, s) (*(p))
# define RE_STRING_CHAR STRING_CHAR
# define CHAR_STRING(c, s) (*(s) = (c), 1)
# define RE_STRING_CHAR_AND_LENGTH STRING_CHAR_AND_LENGTH
# define GET_CHAR_BEFORE_2(c, p, str1, end1, str2, end2) \
(c = ((p) == (str2) ? *((end1) - 1) : *((p) - 1)))
+# define GET_CHAR_AFTER(c, p, len) \
+ (c = *p, len = 1)
# define MAKE_CHAR(charset, c1, c2) (c1)
# define BYTE8_TO_CHAR(c) (c)
# define CHAR_BYTE8_P(c) (0)
+# define MAKE_CHAR_MULTIBYTE(c) (c)
+# define MAKE_CHAR_UNIBYTE(c) (c)
+# define CHAR_LEADING_CODE(c) (c)
+
#endif /* not emacs */
#ifndef RE_TRANSLATE
# ifdef __GNUC__
# define alloca __builtin_alloca
# else /* not __GNUC__ */
-# if HAVE_ALLOCA_H
+# ifdef HAVE_ALLOCA_H
# include <alloca.h>
# endif /* HAVE_ALLOCA_H */
# endif /* not __GNUC__ */
if (start == NULL)
{
- printf ("(null)\n");
+ fprintf (stderr, "(null)\n");
return;
}
/* Loop over pattern commands. */
while (p < pend)
{
- printf ("%d:\t", p - start);
+ fprintf (stderr, "%d:\t", p - start);
switch ((re_opcode_t) *p++)
{
case no_op:
- printf ("/no_op");
+ fprintf (stderr, "/no_op");
break;
case succeed:
- printf ("/succeed");
+ fprintf (stderr, "/succeed");
break;
case exactn:
mcnt = *p++;
- printf ("/exactn/%d", mcnt);
+ fprintf (stderr, "/exactn/%d", mcnt);
do
{
- putchar ('/');
- putchar (*p++);
+ fprintf (stderr, "/%c", *p++);
}
while (--mcnt);
break;
case start_memory:
- printf ("/start_memory/%d", *p++);
+ fprintf (stderr, "/start_memory/%d", *p++);
break;
case stop_memory:
- printf ("/stop_memory/%d", *p++);
+ fprintf (stderr, "/stop_memory/%d", *p++);
break;
case duplicate:
- printf ("/duplicate/%d", *p++);
+ fprintf (stderr, "/duplicate/%d", *p++);
break;
case anychar:
- printf ("/anychar");
+ fprintf (stderr, "/anychar");
break;
case charset:
int length = CHARSET_BITMAP_SIZE (p - 1);
int has_range_table = CHARSET_RANGE_TABLE_EXISTS_P (p - 1);
- printf ("/charset [%s",
+ fprintf (stderr, "/charset [%s",
(re_opcode_t) *(p - 1) == charset_not ? "^" : "");
assert (p + *p < pend);
/* Are we starting a range? */
if (last + 1 == c && ! in_range)
{
- putchar ('-');
+ fprintf (stderr, "-");
in_range = 1;
}
/* Have we broken a range? */
else if (last + 1 != c && in_range)
{
- putchar (last);
+ fprintf (stderr, "%c", last);
in_range = 0;
}
if (! in_range)
- putchar (c);
+ fprintf (stderr, "%c", c);
last = c;
}
if (in_range)
- putchar (last);
+ fprintf (stderr, "%c", last);
- putchar (']');
+ fprintf (stderr, "]");
p += 1 + length;
if (has_range_table)
{
int count;
- printf ("has-range-table");
+ fprintf (stderr, "has-range-table");
/* ??? Should print the range table; for now, just skip it. */
p += 2; /* skip range table bits */
break;
case begline:
- printf ("/begline");
+ fprintf (stderr, "/begline");
break;
case endline:
- printf ("/endline");
+ fprintf (stderr, "/endline");
break;
case on_failure_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_jump to %d", p + mcnt - start);
+ fprintf (stderr, "/on_failure_jump to %d", p + mcnt - start);
break;
case on_failure_keep_string_jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_keep_string_jump to %d", p + mcnt - start);
+ fprintf (stderr, "/on_failure_keep_string_jump to %d", p + mcnt - start);
break;
case on_failure_jump_nastyloop:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_jump_nastyloop to %d", p + mcnt - start);
+ fprintf (stderr, "/on_failure_jump_nastyloop to %d", p + mcnt - start);
break;
case on_failure_jump_loop:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_jump_loop to %d", p + mcnt - start);
+ fprintf (stderr, "/on_failure_jump_loop to %d", p + mcnt - start);
break;
case on_failure_jump_smart:
extract_number_and_incr (&mcnt, &p);
- printf ("/on_failure_jump_smart to %d", p + mcnt - start);
+ fprintf (stderr, "/on_failure_jump_smart to %d", p + mcnt - start);
break;
case jump:
extract_number_and_incr (&mcnt, &p);
- printf ("/jump to %d", p + mcnt - start);
+ fprintf (stderr, "/jump to %d", p + mcnt - start);
break;
case succeed_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
+ fprintf (stderr, "/succeed_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
break;
case jump_n:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
+ fprintf (stderr, "/jump_n to %d, %d times", p - 2 + mcnt - start, mcnt2);
break;
case set_number_at:
extract_number_and_incr (&mcnt, &p);
extract_number_and_incr (&mcnt2, &p);
- printf ("/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2);
+ fprintf (stderr, "/set_number_at location %d to %d", p - 2 + mcnt - start, mcnt2);
break;
case wordbound:
- printf ("/wordbound");
+ fprintf (stderr, "/wordbound");
break;
case notwordbound:
- printf ("/notwordbound");
+ fprintf (stderr, "/notwordbound");
break;
case wordbeg:
- printf ("/wordbeg");
+ fprintf (stderr, "/wordbeg");
break;
case wordend:
- printf ("/wordend");
+ fprintf (stderr, "/wordend");
case syntaxspec:
- printf ("/syntaxspec");
+ fprintf (stderr, "/syntaxspec");
mcnt = *p++;
- printf ("/%d", mcnt);
+ fprintf (stderr, "/%d", mcnt);
break;
case notsyntaxspec:
- printf ("/notsyntaxspec");
+ fprintf (stderr, "/notsyntaxspec");
mcnt = *p++;
- printf ("/%d", mcnt);
+ fprintf (stderr, "/%d", mcnt);
break;
# ifdef emacs
case before_dot:
- printf ("/before_dot");
+ fprintf (stderr, "/before_dot");
break;
case at_dot:
- printf ("/at_dot");
+ fprintf (stderr, "/at_dot");
break;
case after_dot:
- printf ("/after_dot");
+ fprintf (stderr, "/after_dot");
break;
case categoryspec:
- printf ("/categoryspec");
+ fprintf (stderr, "/categoryspec");
mcnt = *p++;
- printf ("/%d", mcnt);
+ fprintf (stderr, "/%d", mcnt);
break;
case notcategoryspec:
- printf ("/notcategoryspec");
+ fprintf (stderr, "/notcategoryspec");
mcnt = *p++;
- printf ("/%d", mcnt);
+ fprintf (stderr, "/%d", mcnt);
break;
# endif /* emacs */
case begbuf:
- printf ("/begbuf");
+ fprintf (stderr, "/begbuf");
break;
case endbuf:
- printf ("/endbuf");
+ fprintf (stderr, "/endbuf");
break;
default:
- printf ("?%d", *(p-1));
+ fprintf (stderr, "?%d", *(p-1));
}
- putchar ('\n');
+ fprintf (stderr, "\n");
}
- printf ("%d:\tend of pattern.\n", p - start);
+ fprintf (stderr, "%d:\tend of pattern.\n", p - start);
}
/* Check that we are not stuck in an infinite loop. */
#define CHECK_INFINITE_LOOP(pat_cur, string_place) \
do { \
- int failure = TOP_FAILURE_HANDLE(); \
+ int failure = TOP_FAILURE_HANDLE (); \
/* Check for infinite matching loops */ \
- while (failure > 0 && \
- (FAILURE_STR (failure) == string_place \
- || FAILURE_STR (failure) == NULL)) \
+ while (failure > 0 \
+ && (FAILURE_STR (failure) == string_place \
+ || FAILURE_STR (failure) == NULL)) \
{ \
assert (FAILURE_PAT (failure) >= bufp->buffer \
&& FAILURE_PAT (failure) <= bufp->buffer + bufp->used); \
if (FAILURE_PAT (failure) == pat_cur) \
- goto fail; \
+ { \
+ cycle = 1; \
+ break; \
+ } \
DEBUG_PRINT2 (" Other pattern: %p\n", FAILURE_PAT (failure)); \
failure = NEXT_FAILURE_HANDLE(failure); \
} \
DEBUG_PRINT2 (" Other string: %p\n", FAILURE_STR (failure)); \
} while (0)
-
+
/* Push the information about the state we will need
if we ever fail back to it.
static int analyse_first _RE_ARGS ((re_char *p, re_char *pend,
char *fastmap, const int multibyte));
-/* Fetch the next character in the uncompiled pattern---translating it
- if necessary. */
-#define PATFETCH(c) \
- do { \
- PATFETCH_RAW (c); \
- c = TRANSLATE (c); \
- } while (0)
-
/* Fetch the next character in the uncompiled pattern, with no
translation. */
-#define PATFETCH_RAW(c) \
+#define PATFETCH(c) \
do { \
int len; \
if (p == pend) return REG_EEND; \
/* But patterns can have more than `MAX_REGNUM' registers. We just
ignore the excess. */
-typedef unsigned regnum_t;
+typedef int regnum_t;
/* Macros for the compile stack. */
/* The next available element. */
#define COMPILE_STACK_TOP (compile_stack.stack[compile_stack.avail])
-
+/* Explicit quit checking is only used on NTemacs. */
+#if defined WINDOWSNT && defined emacs && defined QUIT
+extern int immediate_quit;
+# define IMMEDIATE_QUIT_CHECK \
+ do { \
+ if (immediate_quit) QUIT; \
+ } while (0)
+#else
+# define IMMEDIATE_QUIT_CHECK ((void)0)
+#endif
+\f
/* Structure to manage work area for range table. */
struct range_table_work_area
{
int bits; /* flag to record character classes */
};
-/* Make sure that WORK_AREA can hold more N multibyte characters. */
-#define EXTEND_RANGE_TABLE_WORK_AREA(work_area, n) \
- do { \
- if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \
- { \
- (work_area).allocated += 16 * sizeof (int); \
- if ((work_area).table) \
- (work_area).table \
- = (int *) realloc ((work_area).table, (work_area).allocated); \
- else \
- (work_area).table \
- = (int *) malloc ((work_area).allocated); \
- if ((work_area).table == 0) \
- FREE_STACK_RETURN (REG_ESPACE); \
- } \
+/* Make sure that WORK_AREA can hold more N multibyte characters.
+ This is used only in set_image_of_range and set_image_of_range_1.
+ It expects WORK_AREA to be a pointer.
+ If it can't get the space, it returns from the surrounding function. */
+
+#define EXTEND_RANGE_TABLE(work_area, n) \
+ do { \
+ if (((work_area).used + (n)) * sizeof (int) > (work_area).allocated) \
+ { \
+ extend_range_table_work_area (&work_area); \
+ if ((work_area).table == 0) \
+ return (REG_ESPACE); \
+ } \
} while (0)
#define SET_RANGE_TABLE_WORK_AREA_BIT(work_area, bit) \
/* Set a range (RANGE_START, RANGE_END) to WORK_AREA. */
#define SET_RANGE_TABLE_WORK_AREA(work_area, range_start, range_end) \
do { \
- EXTEND_RANGE_TABLE_WORK_AREA ((work_area), 2); \
+ EXTEND_RANGE_TABLE ((work_area), 2); \
(work_area).table[(work_area).used++] = (range_start); \
(work_area).table[(work_area).used++] = (range_end); \
} while (0)
#define RANGE_TABLE_WORK_USED(work_area) ((work_area).used)
#define RANGE_TABLE_WORK_BITS(work_area) ((work_area).bits)
#define RANGE_TABLE_WORK_ELT(work_area, i) ((work_area).table[i])
-
+\f
/* Set the bit for character C in a list. */
#define SET_LIST_BIT(c) (b[((c)) / BYTEWIDTH] |= 1 << ((c) % BYTEWIDTH))
+#ifdef emacs
+
+/* Store characters in the rage range C0 to C1 in WORK_AREA while
+ translating them and paying attention to the continuity of
+ translated characters.
+
+ Implementation note: It is better to implement this fairly big
+ macro by a function, but it's not that easy because macros called
+ in this macro assume various local variables already declared. */
+
+#define SETUP_MULTIBYTE_RANGE(work_area, c0, c1) \
+ do { \
+ re_wchar_t c, t, t_last; \
+ int n; \
+ \
+ c = (c0); \
+ t_last = multibyte ? TRANSLATE (c) : TRANSLATE (MAKE_CHAR_MULTIBYTE (c)); \
+ for (c++, n = 1; c <= (c1); c++, n++) \
+ { \
+ t = multibyte ? TRANSLATE (c) : TRANSLATE (MAKE_CHAR_MULTIBYTE (c)); \
+ if (t_last + n == t) \
+ continue; \
+ SET_RANGE_TABLE_WORK_AREA ((work_area), t_last, t_last + n - 1); \
+ t_last = t; \
+ n = 0; \
+ } \
+ if (n > 0) \
+ SET_RANGE_TABLE_WORK_AREA ((work_area), t_last, t_last + n - 1); \
+ } while (0)
+
+#endif /* emacs */
+
/* Get the next unsigned number in the uncompiled pattern. */
#define GET_UNSIGNED_NUMBER(num) \
do { if (p != pend) \
{ \
PATFETCH (c); \
+ if (c == ' ') \
+ FREE_STACK_RETURN (REG_BADBR); \
while ('0' <= c && c <= '9') \
{ \
+ int prev; \
if (num < 0) \
- num = 0; \
+ num = 0; \
+ prev = num; \
num = num * 10 + c - '0'; \
+ if (num / 10 != prev) \
+ FREE_STACK_RETURN (REG_BADBR); \
if (p == pend) \
- break; \
+ break; \
PATFETCH (c); \
} \
+ if (c == ' ') \
+ FREE_STACK_RETURN (REG_BADBR); \
} \
} while (0)
-
+\f
#if WIDE_CHAR_SUPPORT
/* The GNU C library provides support for user-defined character classes
and the functions from ISO C amendement 1. */
}
}
#endif
+\f
+/* Filling in the work area of a range. */
-/* Explicit quit checking is only used on NTemacs. */
-#if defined WINDOWSNT && defined emacs && defined QUIT
-extern int immediate_quit;
-# define IMMEDIATE_QUIT_CHECK \
- do { \
- if (immediate_quit) QUIT; \
- } while (0)
-#else
-# define IMMEDIATE_QUIT_CHECK ((void)0)
+/* Actually extend the space in WORK_AREA. */
+
+static void
+extend_range_table_work_area (work_area)
+ struct range_table_work_area *work_area;
+{
+ work_area->allocated += 16 * sizeof (int);
+ if (work_area->table)
+ work_area->table
+ = (int *) realloc (work_area->table, work_area->allocated);
+ else
+ work_area->table
+ = (int *) malloc (work_area->allocated);
+}
+
+#if 0
+#ifdef emacs
+
+/* Carefully find the ranges of codes that are equivalent
+ under case conversion to the range start..end when passed through
+ TRANSLATE. Handle the case where non-letters can come in between
+ two upper-case letters (which happens in Latin-1).
+ Also handle the case of groups of more than 2 case-equivalent chars.
+
+ The basic method is to look at consecutive characters and see
+ if they can form a run that can be handled as one.
+
+ Returns -1 if successful, REG_ESPACE if ran out of space. */
+
+static int
+set_image_of_range_1 (work_area, start, end, translate)
+ RE_TRANSLATE_TYPE translate;
+ struct range_table_work_area *work_area;
+ re_wchar_t start, end;
+{
+ /* `one_case' indicates a character, or a run of characters,
+ each of which is an isolate (no case-equivalents).
+ This includes all ASCII non-letters.
+
+ `two_case' indicates a character, or a run of characters,
+ each of which has two case-equivalent forms.
+ This includes all ASCII letters.
+
+ `strange' indicates a character that has more than one
+ case-equivalent. */
+
+ enum case_type {one_case, two_case, strange};
+
+ /* Describe the run that is in progress,
+ which the next character can try to extend.
+ If run_type is strange, that means there really is no run.
+ If run_type is one_case, then run_start...run_end is the run.
+ If run_type is two_case, then the run is run_start...run_end,
+ and the case-equivalents end at run_eqv_end. */
+
+ enum case_type run_type = strange;
+ int run_start, run_end, run_eqv_end;
+
+ Lisp_Object eqv_table;
+
+ if (!RE_TRANSLATE_P (translate))
+ {
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = (start);
+ work_area->table[work_area->used++] = (end);
+ return -1;
+ }
+
+ eqv_table = XCHAR_TABLE (translate)->extras[2];
+
+ for (; start <= end; start++)
+ {
+ enum case_type this_type;
+ int eqv = RE_TRANSLATE (eqv_table, start);
+ int minchar, maxchar;
+
+ /* Classify this character */
+ if (eqv == start)
+ this_type = one_case;
+ else if (RE_TRANSLATE (eqv_table, eqv) == start)
+ this_type = two_case;
+ else
+ this_type = strange;
+
+ if (start < eqv)
+ minchar = start, maxchar = eqv;
+ else
+ minchar = eqv, maxchar = start;
+
+ /* Can this character extend the run in progress? */
+ if (this_type == strange || this_type != run_type
+ || !(minchar == run_end + 1
+ && (run_type == two_case
+ ? maxchar == run_eqv_end + 1 : 1)))
+ {
+ /* No, end the run.
+ Record each of its equivalent ranges. */
+ if (run_type == one_case)
+ {
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = run_start;
+ work_area->table[work_area->used++] = run_end;
+ }
+ else if (run_type == two_case)
+ {
+ EXTEND_RANGE_TABLE (work_area, 4);
+ work_area->table[work_area->used++] = run_start;
+ work_area->table[work_area->used++] = run_end;
+ work_area->table[work_area->used++]
+ = RE_TRANSLATE (eqv_table, run_start);
+ work_area->table[work_area->used++]
+ = RE_TRANSLATE (eqv_table, run_end);
+ }
+ run_type = strange;
+ }
+
+ if (this_type == strange)
+ {
+ /* For a strange character, add each of its equivalents, one
+ by one. Don't start a range. */
+ do
+ {
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = eqv;
+ work_area->table[work_area->used++] = eqv;
+ eqv = RE_TRANSLATE (eqv_table, eqv);
+ }
+ while (eqv != start);
+ }
+
+ /* Add this char to the run, or start a new run. */
+ else if (run_type == strange)
+ {
+ /* Initialize a new range. */
+ run_type = this_type;
+ run_start = start;
+ run_end = start;
+ run_eqv_end = RE_TRANSLATE (eqv_table, run_end);
+ }
+ else
+ {
+ /* Extend a running range. */
+ run_end = minchar;
+ run_eqv_end = RE_TRANSLATE (eqv_table, run_end);
+ }
+ }
+
+ /* If a run is still in progress at the end, finish it now
+ by recording its equivalent ranges. */
+ if (run_type == one_case)
+ {
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = run_start;
+ work_area->table[work_area->used++] = run_end;
+ }
+ else if (run_type == two_case)
+ {
+ EXTEND_RANGE_TABLE (work_area, 4);
+ work_area->table[work_area->used++] = run_start;
+ work_area->table[work_area->used++] = run_end;
+ work_area->table[work_area->used++]
+ = RE_TRANSLATE (eqv_table, run_start);
+ work_area->table[work_area->used++]
+ = RE_TRANSLATE (eqv_table, run_end);
+ }
+
+ return -1;
+}
+
+#endif /* emacs */
+
+/* Record the the image of the range start..end when passed through
+ TRANSLATE. This is not necessarily TRANSLATE(start)..TRANSLATE(end)
+ and is not even necessarily contiguous.
+ Normally we approximate it with the smallest contiguous range that contains
+ all the chars we need. However, for Latin-1 we go to extra effort
+ to do a better job.
+
+ This function is not called for ASCII ranges.
+
+ Returns -1 if successful, REG_ESPACE if ran out of space. */
+
+static int
+set_image_of_range (work_area, start, end, translate)
+ RE_TRANSLATE_TYPE translate;
+ struct range_table_work_area *work_area;
+ re_wchar_t start, end;
+{
+ re_wchar_t cmin, cmax;
+
+#ifdef emacs
+ /* For Latin-1 ranges, use set_image_of_range_1
+ to get proper handling of ranges that include letters and nonletters.
+ For a range that includes the whole of Latin-1, this is not necessary.
+ For other character sets, we don't bother to get this right. */
+ if (RE_TRANSLATE_P (translate) && start < 04400
+ && !(start < 04200 && end >= 04377))
+ {
+ int newend;
+ int tem;
+ newend = end;
+ if (newend > 04377)
+ newend = 04377;
+ tem = set_image_of_range_1 (work_area, start, newend, translate);
+ if (tem > 0)
+ return tem;
+
+ start = 04400;
+ if (end < 04400)
+ return -1;
+ }
#endif
+
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = (start);
+ work_area->table[work_area->used++] = (end);
+
+ cmin = -1, cmax = -1;
+
+ if (RE_TRANSLATE_P (translate))
+ {
+ int ch;
+
+ for (ch = start; ch <= end; ch++)
+ {
+ re_wchar_t c = TRANSLATE (ch);
+ if (! (start <= c && c <= end))
+ {
+ if (cmin == -1)
+ cmin = c, cmax = c;
+ else
+ {
+ cmin = MIN (cmin, c);
+ cmax = MAX (cmax, c);
+ }
+ }
+ }
+
+ if (cmin != -1)
+ {
+ EXTEND_RANGE_TABLE (work_area, 2);
+ work_area->table[work_area->used++] = (cmin);
+ work_area->table[work_area->used++] = (cmax);
+ }
+ }
+
+ return -1;
+}
+#endif /* 0 */
\f
#ifndef MATCH_MAY_ALLOCATE
/* If the object matched can contain multibyte characters. */
const boolean multibyte = RE_MULTIBYTE_P (bufp);
+ /* If a target of matching can contain multibyte characters. */
+ const boolean target_multibyte = RE_TARGET_MULTIBYTE_P (bufp);
+
#ifdef DEBUG
debug++;
DEBUG_PRINT1 ("\nCompiling pattern: ");
unsigned int startoffset = 0;
re_opcode_t ofj =
/* Check if the loop can match the empty string. */
- (simple || !analyse_first (laststart, b, NULL, 0)) ?
- on_failure_jump : on_failure_jump_loop;
+ (simple || !analyse_first (laststart, b, NULL, 0))
+ ? on_failure_jump : on_failure_jump_loop;
assert (skip_one_char (laststart) <= b);
-
+
if (!zero_times_ok && simple)
{ /* Since simple * loops can be made faster by using
on_failure_keep_string_jump, we turn simple P+
{
boolean emptyp = analyse_first (laststart, b, NULL, 0);
- /* The non-greedy multiple match looks like a repeat..until:
- we only need a conditional jump at the end of the loop */
+ /* The non-greedy multiple match looks like
+ a repeat..until: we only need a conditional jump
+ at the end of the loop. */
if (emptyp) BUF_PUSH (no_op);
STORE_JUMP (emptyp ? on_failure_jump_nastyloop
: on_failure_jump, b, laststart);
{
/* The repeat...until naturally matches one or more.
To also match zero times, we need to first jump to
- the end of the loop (its conditional jump). */
+ the end of the loop (its conditional jump). */
INSERT_JUMP (jump, laststart, b);
b += 3;
}
if (p == pend) FREE_STACK_RETURN (REG_EBRACK);
+ /* Don't translate yet. The range TRANSLATE(X..Y) cannot
+ always be determined from TRANSLATE(X) and TRANSLATE(Y)
+ So the translation is done later in a loop. Example:
+ (let ((case-fold-search t)) (string-match "[A-_]" "A")) */
PATFETCH (c);
/* \ might escape characters inside [...] and [^...]. */
break;
}
- /* What should we do for the character which is
- greater than 0x7F, but not BASE_LEADING_CODE_P?
- XXX */
-
/* See if we're at the beginning of a possible character
class. */
them). */
if (c == ':' && *p == ']')
{
- int ch;
+ re_wchar_t ch;
re_wctype_t cc;
+ int limit;
cc = re_wctype (str);
don't need to handle them for multibyte.
They are distinguished by a negative wctype. */
- if (multibyte)
- SET_RANGE_TABLE_WORK_AREA_BIT (range_table_work,
- re_wctype_to_bit (cc));
+ for (ch = 0; ch < 128; ++ch)
+ if (re_iswctype (btowc (ch), cc))
+ {
+ c = TRANSLATE (ch);
+ SET_LIST_BIT (c);
+ }
- for (ch = 0; ch < 1 << BYTEWIDTH; ++ch)
+ if (target_multibyte)
+ {
+ SET_RANGE_TABLE_WORK_AREA_BIT
+ (range_table_work, re_wctype_to_bit (cc));
+ }
+ else
{
- int translated = TRANSLATE (ch);
- if (re_iswctype (btowc (ch), cc))
- SET_LIST_BIT (translated);
+ for (ch = 0; ch < (1 << BYTEWIDTH); ++ch)
+ {
+ c = ch;
+ MAKE_CHAR_MULTIBYTE (c);
+ if (re_iswctype (btowc (c), cc))
+ {
+ c = TRANSLATE (c);
+ MAKE_CHAR_UNIBYTE (c);
+ SET_LIST_BIT (c);
+ }
+ }
}
/* Repeat the loop. */
/* Fetch the character which ends the range. */
PATFETCH (c1);
-
- if (SINGLE_BYTE_CHAR_P (c)
- && ! SINGLE_BYTE_CHAR_P (c1))
+ if (c > c1)
{
- /* Handle a range starting with a character
- fitting in a bitmap to a character not
- fitting in a bitmap (thus require range
- table). We use both a bitmap (for the
- range from C to 255) and a range table (for
- the remaining range). Here, we setup only
- a range table. A bitmap is setup later. */
- re_wchar_t c2
- = CHAR_BYTE8_P (c1) ? BYTE8_TO_CHAR (0x80) : 256;
-
- SET_RANGE_TABLE_WORK_AREA (range_table_work, c2, c1);
- c1 = 255;
+ if (syntax & RE_NO_EMPTY_RANGES)
+ FREE_STACK_RETURN (REG_ERANGE);
+ /* Else, repeat the loop. */
}
}
else
/* Range from C to C. */
c1 = c;
- /* Set the range ... */
- if (SINGLE_BYTE_CHAR_P (c))
- /* ... into bitmap. */
+#ifndef emacs
+ c = TRANSLATE (c);
+ c1 = TRANSLATE (c1);
+ /* Set the range into bitmap */
+ for (; c <= c1; c++)
+ SET_LIST_BIT (TRANSLATE (c));
+#else /* not emacs */
+ if (target_multibyte)
{
- re_wchar_t this_char;
- int range_start = c, range_end = c1;
-
- /* If the start is after the end, the range is empty. */
- if (range_start > range_end)
+ if (c1 >= 128)
{
- if (syntax & RE_NO_EMPTY_RANGES)
- FREE_STACK_RETURN (REG_ERANGE);
- /* Else, repeat the loop. */
+ re_wchar_t c0 = MAX (c, 128);
+
+ SETUP_MULTIBYTE_RANGE (range_table_work, c0, c1);
+ c1 = 127;
}
- else
+ for (; c <= c1; c++)
+ SET_LIST_BIT (TRANSLATE (c));
+ }
+ else
+ {
+ re_wchar_t c0;
+
+ for (; c <= c1; c++)
{
- for (this_char = range_start; this_char <= range_end;
- this_char++)
- SET_LIST_BIT (TRANSLATE (this_char));
+ c0 = c;
+ if (! multibyte)
+ MAKE_CHAR_MULTIBYTE (c0);
+ c0 = TRANSLATE (c0);
+ MAKE_CHAR_UNIBYTE (c0);
+ SET_LIST_BIT (c0);
}
}
- else
- /* ... into range table. */
- SET_RANGE_TABLE_WORK_AREA (range_table_work, c, c1);
+#endif /* not emacs */
}
/* Discard any (non)matching list bytes that are all 0 at the
/* Do not translate the character after the \, so that we can
distinguish, e.g., \B from \b, even if we normally would
translate, e.g., B to b. */
- PATFETCH_RAW (c);
+ PATFETCH (c);
switch (c)
{
goto unfetch_interval;
}
- if (upper_bound == 0)
- /* If the upper bound is zero, just drop the sub pattern
- altogether. */
- b = laststart;
- else if (lower_bound == 1 && upper_bound == 1)
- /* Just match it once: nothing to do here. */
- ;
-
- /* Otherwise, we have a nontrivial interval. When
- we're all done, the pattern will look like:
- set_number_at <jump count> <upper bound>
- set_number_at <succeed_n count> <lower bound>
-
succeed_n <after jump addr> <succeed_n count>
- <body of loop>
- jump_n <succeed_n addr> <jump count>
- (The upper bound and `jump_n' are omitted if
- `upper_bound' is 1, though.) */
- else
- { /* If the upper bound is > 1, we need to insert
- more at the end of the loop. */
- unsigned int nbytes = (upper_bound < 0 ? 3
- : upper_bound > 1 ? 5 : 0);
- unsigned int startoffset = 0;
-
- GET_BUFFER_SPACE (20); /* We might use less. */
-
- if (lower_bound == 0)
- {
- /* A succeed_n that starts with 0 is really a
- a simple on_failure_jump_loop. */
- INSERT_JUMP (on_failure_jump_loop, laststart,
- b + 3 + nbytes);
- b += 3;
- }
- else
- {
- /* Initialize lower bound of the `succeed_n', even
- though it will be set during matching by its
- attendant `set_number_at' (inserted next),
- because `re_compile_fastmap' needs to know.
- Jump to the `jump_n' we might insert below. */
- INSERT_JUMP2 (succeed_n, laststart,
- b + 5 + nbytes,
- lower_bound);
- b += 5;
-
- /* Code to initialize the lower bound. Insert
- before the `succeed_n'. The `5' is the last two
- bytes of this `set_number_at', plus 3 bytes of
- the following `succeed_n'. */
- insert_op2 (set_number_at, laststart, 5, lower_bound, b);
- b += 5;
- startoffset += 5;
- }
-
- if (upper_bound < 0)
- {
- /* A negative upper bound stands for infinity,
- in which case it degenerates to a plain jump. */
- STORE_JUMP (jump, b, laststart + startoffset);
- b += 3;
- }
- else if (upper_bound > 1)
- { /* More than one repetition is allowed, so
- append a backward jump to the `succeed_n'
- that starts this interval.
-
- When we've reached this during matching,
- we'll have matched the interval once, so
- jump back only `upper_bound - 1' times. */
- STORE_JUMP2 (jump_n, b, laststart + startoffset,
- upper_bound - 1);
- b += 5;
-
- /* The location we want to set is the second
- parameter of the `jump_n'; that is `b-2' as
- an absolute address. `laststart' will be
- the `set_number_at' we're about to insert;
- `laststart+3' the number to set, the source
- for the relative address. But we are
- inserting into the middle of the pattern --
- so everything is getting moved up by 5.
- Conclusion: (b - 2) - (laststart + 3) + 5,
- i.e., b - laststart.
-
- We insert this at the beginning of the loop
- so that if we fail during matching, we'll
- reinitialize the bounds. */
- insert_op2 (set_number_at, laststart, b - laststart,
- upper_bound - 1, b);
- b += 5;
- }
- }
+ if (upper_bound == 0)
+ /* If the upper bound is zero, just drop the sub pattern
+ altogether. */
+ b = laststart;
+ else if (lower_bound == 1 && upper_bound == 1)
+ /* Just match it once: nothing to do here. */
+ ;
+
+ /* Otherwise, we have a nontrivial interval. When
+ we're all done, the pattern will look like:
+ set_number_at <jump count> <upper bound>
+ set_number_at <succeed_n count> <lower bound>
+ succeed_n <after jump addr> <succeed_n count>
+ <body of loop>
+ jump_n <succeed_n addr> <jump count>
+ (The upper bound and `jump_n' are omitted if
+ `upper_bound' is 1, though.) */
+ else
+ { /* If the upper bound is > 1, we need to insert
+ more at the end of the loop. */
+ unsigned int nbytes = (upper_bound < 0 ? 3
+ : upper_bound > 1 ? 5 : 0);
+ unsigned int startoffset = 0;
+
+ GET_BUFFER_SPACE (20); /* We might use less. */
+
+ if (lower_bound == 0)
+ {
+ /* A succeed_n that starts with 0 is really a
+ a simple on_failure_jump_loop. */
+ INSERT_JUMP (on_failure_jump_loop, laststart,
+ b + 3 + nbytes);
+ b += 3;
+ }
+ else
+ {
+ /* Initialize lower bound of the `succeed_n', even
+ though it will be set during matching by its
+ attendant `set_number_at' (inserted next),
+ because `re_compile_fastmap' needs to know.
+ Jump to the `jump_n' we might insert below. */
+ INSERT_JUMP2 (succeed_n, laststart,
+ b + 5 + nbytes,
+ lower_bound);
+ b += 5;
+
+ /* Code to initialize the lower bound. Insert
+ before the `succeed_n'. The `5' is the last two
+ bytes of this `set_number_at', plus 3 bytes of
+ the following `succeed_n'. */
+ insert_op2 (set_number_at, laststart, 5, lower_bound, b);
+ b += 5;
+ startoffset += 5;
+ }
+
+ if (upper_bound < 0)
+ {
+ /* A negative upper bound stands for infinity,
+ in which case it degenerates to a plain jump. */
+ STORE_JUMP (jump, b, laststart + startoffset);
+ b += 3;
+ }
+ else if (upper_bound > 1)
+ { /* More than one repetition is allowed, so
+ append a backward jump to the `succeed_n'
+ that starts this interval.
+
+ When we've reached this during matching,
+ we'll have matched the interval once, so
+ jump back only `upper_bound - 1' times. */
+ STORE_JUMP2 (jump_n, b, laststart + startoffset,
+ upper_bound - 1);
+ b += 5;
+
+ /* The location we want to set is the second
+ parameter of the `jump_n'; that is `b-2' as
+ an absolute address. `laststart' will be
+ the `set_number_at' we're about to insert;
+ `laststart+3' the number to set, the source
+ for the relative address. But we are
+ inserting into the middle of the pattern --
+ so everything is getting moved up by 5.
+ Conclusion: (b - 2) - (laststart + 3) + 5,
+ i.e., b - laststart.
+
+ We insert this at the beginning of the loop
+ so that if we fail during matching, we'll
+ reinitialize the bounds. */
+ insert_op2 (set_number_at, laststart, b - laststart,
+ upper_bound - 1, b);
+ b += 5;
+ }
+ }
pending_exact = 0;
beg_interval = NULL;
}
case 'c':
laststart = b;
- PATFETCH_RAW (c);
+ PATFETCH (c);
BUF_PUSH_2 (categoryspec, c);
break;
case 'C':
laststart = b;
- PATFETCH_RAW (c);
+ PATFETCH (c);
BUF_PUSH_2 (notcategoryspec, c);
break;
#endif /* emacs */
/* You might think it would be useful for \ to mean
not to translate; but if we don't translate it
it will never match anything. */
- c = TRANSLATE (c);
goto normal_char;
}
break;
default:
/* Expects the character in `c'. */
normal_char:
- /* If no exactn currently being built. */
+ /* If no exactn currently being built. */
if (!pending_exact
/* If last exactn not at current position. */
{
int len;
- if (multibyte)
- len = CHAR_STRING (c, b);
+ if (! multibyte)
+ MAKE_CHAR_MULTIBYTE (c);
+ c = TRANSLATE (c);
+ if (target_multibyte)
+ {
+ len = CHAR_STRING (c, b);
+ b += len;
+ }
else
- *b = c, len = 1;
- b += len;
+ {
+ MAKE_CHAR_UNIBYTE (c);
+ *b++ = c;
+ len = 1;
+ }
(*pending_exact) += len;
}
/* We have succeeded; set the length of the buffer. */
bufp->used = b - bufp->buffer;
+#ifdef emacs
+ /* Now the buffer is adjusted for the multibyteness of a target. */
+ bufp->multibyte = bufp->target_multibyte;
+#endif
+
#ifdef DEBUG
if (debug > 0)
{
case exactn:
if (fastmap)
- {
- int c = RE_STRING_CHAR (p + 1, pend - p);
-
- if (SINGLE_BYTE_CHAR_P (c))
- fastmap[c] = 1;
- else
- fastmap[p[1]] = 1;
- }
+ /* If multibyte is nonzero, the first byte of each
+ character is an ASCII or a leading code. Otherwise,
+ each byte is a character. Thus, this works in both
+ cases. */
+ fastmap[p[1]] = 1;
break;
case charset_not:
- /* Chars beyond end of bitmap are possible matches.
- All the single-byte codes can occur in multibyte buffers.
- So any that are not listed in the charset
- are possible matches, even in multibyte buffers. */
if (!fastmap) break;
- for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH;
- j < (1 << BYTEWIDTH); j++)
- fastmap[j] = 1;
+ {
+ /* Chars beyond end of bitmap are possible matches. */
+ /* In a multibyte case, the bitmap is used only for ASCII
+ characters. */
+ int limit = multibyte ? 128 : (1 << BYTEWIDTH);
+
+ for (j = CHARSET_BITMAP_SIZE (&p[-1]) * BYTEWIDTH;
+ j < limit; j++)
+ fastmap[j] = 1;
+ }
+
/* Fallthrough */
case charset:
if (!fastmap) break;
fastmap[j] = 1;
if ((not && multibyte)
- /* Any character set can possibly contain a character
+ /* Any leading code can possibly start a character
which doesn't match the specified set of characters. */
|| (CHARSET_RANGE_TABLE_EXISTS_P (&p[-2])
&& CHARSET_RANGE_TABLE_BITS (&p[-2]) != 0))
/* If we can match a character class, we can match
- any character set. */
+ any multibyte characters. */
{
- set_fastmap_for_multibyte_characters:
if (match_any_multibyte_characters == false)
{
- for (j = 0x80; j < 0x100; j++) /* XXX */
- if (BASE_LEADING_CODE_P (j))
- fastmap[j] = 1;
+ for (j = 0x80; j < (1 << BYTEWIDTH); j++)
+ fastmap[j] = 1;
match_any_multibyte_characters = true;
}
}
else if (!not && CHARSET_RANGE_TABLE_EXISTS_P (&p[-2])
&& match_any_multibyte_characters == false)
{
- /* Set fastmap[I] to 1 where I is a base leading code of each
+ /* Set fastmap[I] to 1 where I is a leading code of each
multibyte characer in the range table. */
int c, count;
- unsigned char buf1[MAX_MULTIBYTE_LENGTH];
- unsigned char buf2[MAX_MULTIBYTE_LENGTH];
+ unsigned char lc1, lc2;
/* Make P points the range table. `+ 2' is to skip flag
bits for a character class. */
{
/* Extract the start and end of each range. */
EXTRACT_CHARACTER (c, p);
- CHAR_STRING (c, buf1);
+ lc1 = CHAR_LEADING_CODE (c);
p += 3;
EXTRACT_CHARACTER (c, p);
- CHAR_STRING (c, buf2);
- for (j = buf1[0]; j <= buf2[0]; j++)
+ lc2 = CHAR_LEADING_CODE (c);
+ for (j = lc1; j <= lc2; j++)
fastmap[j] = 1;
}
}
if (!fastmap) break;
not = (re_opcode_t)p[-1] == notcategoryspec;
k = *p++;
- for (j = 0; j < (1 << BYTEWIDTH); j++)
+ for (j = (multibyte ? 127 : (1 << BYTEWIDTH)); j >= 0; j--)
if ((CHAR_HAS_CATEGORY (j, k)) ^ not)
fastmap[j] = 1;
if (multibyte)
- /* Any character set can possibly contain a character
- whose category is K (or not). */
- goto set_fastmap_for_multibyte_characters;
+ {
+ /* Any character set can possibly contain a character
+ whose category is K (or not). */
+ if (match_any_multibyte_characters == false)
+ {
+ for (j = 0x80; j < (1 << BYTEWIDTH); j++)
+ fastmap[j] = 1;
+ match_any_multibyte_characters = true;
+ }
+ }
break;
/* All cases after this match the empty string. These end with
case has already been handled, so we only need to look at the
fallthrough case. */
continue;
-
+
case succeed_n:
/* If N == 0, it should be an on_failure_jump_loop instead. */
DEBUG_STATEMENT (EXTRACT_NUMBER (j, p + 2); assert (j > 0));
}
WEAK_ALIAS (__re_search, re_search)
+/* Head address of virtual concatenation of string. */
+#define HEAD_ADDR_VSTRING(P) \
+ (((P) >= size1 ? string2 : string1))
+
/* End address of virtual concatenation of string. */
#define STOP_ADDR_VSTRING(P) \
(((P) >= size1 ? string2 + size2 : string1 + size1))
int total_size = size1 + size2;
int endpos = startpos + range;
boolean anchored_start;
-
- /* Nonzero if we have to concern multibyte character. */
+ /* Nonzero if BUFP is setup for multibyte characters. We are sure
+ that it is the same as RE_TARGET_MULTIBYTE_P (bufp). */
const boolean multibyte = RE_MULTIBYTE_P (bufp);
/* Check for out-of-range STARTPOS. */
buf_ch = STRING_CHAR_AND_LENGTH (d, range - lim,
buf_charlen);
-
buf_ch = RE_TRANSLATE (translate, buf_ch);
- if (buf_ch >= 0400
- || fastmap[buf_ch])
+ if (fastmap[CHAR_LEADING_CODE (buf_ch)])
break;
range -= buf_charlen;
d += buf_charlen;
}
else
- while (range > lim
- && !fastmap[RE_TRANSLATE (translate, *d)])
+ while (range > lim)
{
+ buf_ch = *d;
+ MAKE_CHAR_MULTIBYTE (buf_ch);
+ buf_ch = RE_TRANSLATE (translate, buf_ch);
+ MAKE_CHAR_UNIBYTE (buf_ch);
+ if (fastmap[buf_ch])
+ break;
d++;
range--;
}
}
else
- while (range > lim && !fastmap[*d])
- {
- d++;
- range--;
- }
+ {
+ if (multibyte)
+ while (range > lim)
+ {
+ int buf_charlen;
+ buf_ch = STRING_CHAR_AND_LENGTH (d, range - lim,
+ buf_charlen);
+ if (fastmap[CHAR_LEADING_CODE (buf_ch)])
+ break;
+ range -= buf_charlen;
+ d += buf_charlen;
+ }
+ else
+ while (range > lim && !fastmap[*d])
+ {
+ d++;
+ range--;
+ }
+ }
startpos += irange - range;
}
else /* Searching backwards. */
int room = (startpos >= size1
? size2 + size1 - startpos
: size1 - startpos);
- buf_ch = RE_STRING_CHAR (d, room);
- buf_ch = TRANSLATE (buf_ch);
-
- if (! (buf_ch >= 0400
- || fastmap[buf_ch]))
- goto advance;
+ if (multibyte)
+ {
+ buf_ch = STRING_CHAR (d, room);
+ buf_ch = TRANSLATE (buf_ch);
+ if (! fastmap[CHAR_LEADING_CODE (buf_ch)])
+ goto advance;
+ }
+ else
+ {
+ if (! fastmap[TRANSLATE (*d)])
+ goto advance;
+ }
}
}
/* Update STARTPOS to the previous character boundary. */
if (multibyte)
{
- re_char *p = POS_ADDR_VSTRING (startpos);
- int len = 0;
+ re_char *p = POS_ADDR_VSTRING (startpos) + 1;
+ re_char *p0 = p;
+ re_char *phead = HEAD_ADDR_VSTRING (startpos);
/* Find the head of multibyte form. */
- while (!CHAR_HEAD_P (*p))
- p--, len++;
-
- /* Adjust it. */
-#if 0 /* XXX */
- if (MULTIBYTE_FORM_LENGTH (p, len + 1) != (len + 1))
- ;
- else
-#endif
- {
- range += len;
- if (range > 0)
- break;
+ PREV_CHAR_BOUNDARY (p, phead);
+ range += p0 - 1 - p;
+ if (range > 0)
+ break;
- startpos -= len;
- }
+ startpos -= p0 - 1 - p;
}
}
}
{
case anychar:
break;
-
+
case exactn:
p += *p + 1;
break;
else
p += 1 + CHARSET_BITMAP_SIZE (p - 1);
break;
-
+
case syntaxspec:
case notsyntaxspec:
#ifdef emacs
return 1;
}
break;
-
+
case endline:
case exactn:
{
/* Test if C is listed in charset (or charset_not)
at `p1'. */
- if (SINGLE_BYTE_CHAR_P (c))
+ if (! multibyte || IS_REAL_ASCII (c))
{
if (c < CHARSET_BITMAP_SIZE (p1) * BYTEWIDTH
&& p1[2 + c / BYTEWIDTH] & (1 << (c % BYTEWIDTH)))
size of bitmap table of P1 is extracted by
using macro `CHARSET_BITMAP_SIZE'.
- Since we know that all the character listed in
- P2 is ASCII, it is enough to test only bitmap
- table of P1. */
+ In a multibyte case, we know that all the character
+ listed in P2 is ASCII. In a unibyte case, P1 has only a
+ bitmap table. So, in both cases, it is enough to test
+ only the bitmap table of P1. */
if ((re_opcode_t) *p1 == charset)
{
}
}
break;
-
+
case charset_not:
switch (SWITCH_ENUM_CAST (*p1))
{
they don't overlap. The union of the two sets of excluded
chars should cover all possible chars, which, as a matter of
fact, is virtually impossible in multibyte buffers. */
- ;
+ break;
}
break;
}
WEAK_ALIAS (__re_match_2, re_match_2)
+#ifdef emacs
+#define TRANSLATE_VIA_MULTIBYTE(c) \
+ do { \
+ if (multibyte) \
+ (c) = TRANSLATE (c); \
+ else \
+ { \
+ MAKE_CHAR_MULTIBYTE (c); \
+ (c) = TRANSLATE (c); \
+ MAKE_CHAR_UNIBYTE (c); \
+ } \
+ } while (0)
+
+#else
+#define TRANSLATE_VIA_MULTIBYTE(c) ((c) = TRANSLATE (c))
+#endif
+
+
/* This is a separate function so that we can force an alloca cleanup
afterwards. */
static int
/* We use this to map every character in the string. */
RE_TRANSLATE_TYPE translate = bufp->translate;
- /* Nonzero if we have to concern multibyte character. */
+ /* Nonzero if BUFP is setup for multibyte characters. We are sure
+ that it is the same as RE_TARGET_MULTIBYTE_P (bufp). */
const boolean multibyte = RE_MULTIBYTE_P (bufp);
/* Failure point stack. Each place that can handle a failure further
/* Remember the start point to rollback upon failure. */
dfail = d;
+#ifndef emacs
/* This is written out as an if-else so we don't waste time
testing `translate' inside the loop. */
if (RE_TRANSLATE_P (translate))
- {
- if (multibyte)
- do
+ do
+ {
+ PREFETCH ();
+ if (RE_TRANSLATE (translate, *d) != *p++)
{
- int pat_charlen, buf_charlen;
- unsigned int pat_ch, buf_ch;
-
- PREFETCH ();
- pat_ch = STRING_CHAR_AND_LENGTH (p, pend - p, pat_charlen);
- buf_ch = STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen);
+ d = dfail;
+ goto fail;
+ }
+ d++;
+ }
+ while (--mcnt);
+ else
+ do
+ {
+ PREFETCH ();
+ if (*d++ != *p++)
+ {
+ d = dfail;
+ goto fail;
+ }
+ }
+ while (--mcnt);
+#else /* emacs */
+ /* The cost of testing `translate' is comparatively small. */
+ if (multibyte)
+ do
+ {
+ int pat_charlen, buf_charlen;
+ unsigned int pat_ch, buf_ch;
- if (RE_TRANSLATE (translate, buf_ch)
- != pat_ch)
- {
- d = dfail;
- goto fail;
- }
+ PREFETCH ();
+ pat_ch = STRING_CHAR_AND_LENGTH (p, pend - p, pat_charlen);
+ buf_ch = STRING_CHAR_AND_LENGTH (d, dend - d, buf_charlen);
- p += pat_charlen;
- d += buf_charlen;
- mcnt -= pat_charlen;
- }
- while (mcnt > 0);
- else
- do
+ if (TRANSLATE (buf_ch) != pat_ch)
{
- PREFETCH ();
- if (RE_TRANSLATE (translate, *d) != *p++)
- {
- d = dfail;
- goto fail;
- }
- d++;
+ d = dfail;
+ goto fail;
}
- while (--mcnt);
- }
+
+ p += pat_charlen;
+ d += buf_charlen;
+ mcnt -= pat_charlen;
+ }
+ while (mcnt > 0);
else
- {
- do
- {
- PREFETCH ();
- if (*d++ != *p++)
- {
- d = dfail;
- goto fail;
- }
- }
- while (--mcnt);
- }
+ do
+ {
+ unsigned int buf_ch;
+
+ PREFETCH ();
+ buf_ch = *d++;
+ TRANSLATE_VIA_MULTIBYTE (buf_ch);
+ if (buf_ch != *p++)
+ {
+ d = dfail;
+ goto fail;
+ }
+ }
+ while (--mcnt);
+#endif
break;
PREFETCH ();
c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);
- c = TRANSLATE (c); /* The character to match. */
+ TRANSLATE_VIA_MULTIBYTE (c); /* The character to match. */
- if (SINGLE_BYTE_CHAR_P (c))
+ if (! multibyte || IS_REAL_ASCII (c))
{ /* Lookup bitmap. */
/* Cast to `unsigned' instead of `unsigned char' in
case the bit list is a full 32 bytes long. */
assert (!REG_UNSET (regstart[*p]));
/* Strictly speaking, there should be code such as:
-
+
assert (REG_UNSET (regend[*p]));
PUSH_FAILURE_REGSTOP ((unsigned int)*p);
}
else
{
- unsigned char c;
+ unsigned c;
GET_CHAR_BEFORE_2 (c, d, string1, end1, string2, end2);
if (c == '\n')
break;
cycle detection cannot work. Worse yet, such a detection
can not only fail to detect a cycle, but it can also wrongly
detect a cycle (between different instantiations of the same
- loop.
+ loop).
So the method used for those nasty loops is a little different:
We use a special cycle-detection-stack-frame which is pushed
when the on_failure_jump_nastyloop failure-point is *popped*.
mcnt, p + mcnt);
assert ((re_opcode_t)p[-4] == no_op);
- CHECK_INFINITE_LOOP (p - 4, d);
- PUSH_FAILURE_POINT (p - 3, d);
+ {
+ int cycle = 0;
+ CHECK_INFINITE_LOOP (p - 4, d);
+ if (!cycle)
+ /* If there's a cycle, just continue without pushing
+ this failure point. The failure point is the "try again"
+ option, which shouldn't be tried.
+ We want (x?)*?y\1z to match both xxyz and xxyxz. */
+ PUSH_FAILURE_POINT (p - 3, d);
+ }
break;
-
/* Simple loop detecting on_failure_jump: just check on the
failure stack if the same spot was already hit earlier. */
case on_failure_jump_loop:
EXTRACT_NUMBER_AND_INCR (mcnt, p);
DEBUG_PRINT3 ("EXECUTING on_failure_jump_loop %d (to %p):\n",
mcnt, p + mcnt);
-
- CHECK_INFINITE_LOOP (p - 3, d);
- PUSH_FAILURE_POINT (p - 3, d);
+ {
+ int cycle = 0;
+ CHECK_INFINITE_LOOP (p - 3, d);
+ if (cycle)
+ /* If there's a cycle, get out of the loop, as if the matching
+ had failed. We used to just `goto fail' here, but that was
+ aborting the search a bit too early: we want to keep the
+ empty-loop-match and keep matching after the loop.
+ We want (x?)*y\1z to match both xxyz and xxyxz. */
+ p += mcnt;
+ else
+ PUSH_FAILURE_POINT (p - 3, d);
+ }
break;
is the character at D, and S2 is the syntax of C2. */
re_wchar_t c1, c2;
int s1, s2;
+ int dummy;
#ifdef emacs
int offset = PTR_TO_OFFSET (d - 1);
int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
UPDATE_SYNTAX_TABLE_FORWARD (charpos + 1);
#endif
PREFETCH_NOLIMIT ();
- c2 = RE_STRING_CHAR (d, dend - d);
+ GET_CHAR_AFTER (c2, d, dummy);
s2 = SYNTAX (c2);
if (/* Case 2: Only one of S1 and S2 is Sword. */
is the character at D, and S2 is the syntax of C2. */
re_wchar_t c1, c2;
int s1, s2;
+ int dummy;
#ifdef emacs
int offset = PTR_TO_OFFSET (d);
int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
UPDATE_SYNTAX_TABLE (charpos);
#endif
PREFETCH ();
- c2 = RE_STRING_CHAR (d, dend - d);
+ GET_CHAR_AFTER (c2, d, dummy);
s2 = SYNTAX (c2);
-
+
/* Case 2: S2 is not Sword. */
if (s2 != Sword)
goto fail;
is the character at D, and S2 is the syntax of C2. */
re_wchar_t c1, c2;
int s1, s2;
+ int dummy;
#ifdef emacs
int offset = PTR_TO_OFFSET (d) - 1;
int charpos = SYNTAX_TABLE_BYTE_TO_CHAR (offset);
if (!AT_STRINGS_END (d))
{
PREFETCH_NOLIMIT ();
- c2 = RE_STRING_CHAR (d, dend - d);
+ GET_CHAR_AFTER (c2, d, dummy);
#ifdef emacs
UPDATE_SYNTAX_TABLE_FORWARD (charpos);
#endif
int len;
re_wchar_t c;
- c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);
-
+ GET_CHAR_AFTER (c, d, len);
if ((SYNTAX (c) != (enum syntaxcode) mcnt) ^ not)
goto fail;
d += len;
int len;
re_wchar_t c;
- c = RE_STRING_CHAR_AND_LENGTH (d, dend - d, len);
-
+ GET_CHAR_AFTER (c, d, len);
if ((!CHAR_HAS_CATEGORY (c, mcnt)) ^ not)
goto fail;
d += len;
int p1_charlen, p2_charlen;
re_wchar_t p1_ch, p2_ch;
- p1_ch = RE_STRING_CHAR_AND_LENGTH (p1, p1_end - p1, p1_charlen);
- p2_ch = RE_STRING_CHAR_AND_LENGTH (p2, p2_end - p2, p2_charlen);
+ GET_CHAR_AFTER (p1_ch, p1, p1_charlen);
+ GET_CHAR_AFTER (p2_ch, p2, p2_charlen);
if (RE_TRANSLATE (translate, p1_ch)
!= RE_TRANSLATE (translate, p2_ch))
const regex_t *__restrict preg;
const char *__restrict string;
size_t nmatch;
- regmatch_t pmatch[];
+ regmatch_t pmatch[__restrict_arr];
int eflags;
{
int ret;
HCoop / bpt / emacs.git / blobdiff