aboutsummaryrefslogtreecommitdiff
path: root/posix/regex.c
diff options
context:
space:
mode:
authorUlrich Drepper <drepper@redhat.com>2000-08-13 06:38:11 +0000
committerUlrich Drepper <drepper@redhat.com>2000-08-13 06:38:11 +0000
commit8868f97b04da96174316ca6f46d1c3354c19e3e8 (patch)
tree2379280b4e147e18b4300ca330a0952bda92a8e3 /posix/regex.c
parent451fc03299e4b317aad3965bf81ec772e49c51ce (diff)
downloadglibc-8868f97b04da96174316ca6f46d1c3354c19e3e8.zip
glibc-8868f97b04da96174316ca6f46d1c3354c19e3e8.tar.gz
glibc-8868f97b04da96174316ca6f46d1c3354c19e3e8.tar.bz2
Update.
2000-08-12 Ulrich Drepper <drepper@redhat.com> * posix/regex.c (compile_range): Implement collation sequence handling for glibc. Use simple character ordering otherwise and not strcoll.
Diffstat (limited to 'posix/regex.c')
-rw-r--r--posix/regex.c45
1 files changed, 28 insertions, 17 deletions
diff --git a/posix/regex.c b/posix/regex.c
index f1d066b..b4adfcf 100644
--- a/posix/regex.c
+++ b/posix/regex.c
@@ -3507,23 +3507,19 @@ compile_range (range_start_char, p_ptr, pend, translate, syntax, b)
unsigned char *b;
{
unsigned this_char;
-
const char *p = *p_ptr;
reg_errcode_t ret;
- char range_start[2];
- char range_end[2];
- char ch[2];
+#if _LIBC
+ const unsigned char *collseq;
+ unsigned int start_colseq;
+ unsigned int end_colseq;
+#else
+ unsigned end_char;
+#endif
if (p == pend)
return REG_ERANGE;
- /* Fetch the endpoints without translating them; the
- appropriate translation is done in the bit-setting loop below. */
- range_start[0] = TRANSLATE (range_start_char);
- range_start[1] = '\0';
- range_end[0] = TRANSLATE (p[0]);
- range_end[1] = '\0';
-
/* Have to increment the pointer into the pattern string, so the
caller isn't still at the ending character. */
(*p_ptr)++;
@@ -3531,19 +3527,34 @@ compile_range (range_start_char, p_ptr, pend, translate, syntax, b)
/* Report an error if the range is empty and the syntax prohibits this. */
ret = syntax & RE_NO_EMPTY_RANGES ? REG_ERANGE : REG_NOERROR;
- /* Here we see why `this_char' has to be larger than an `unsigned
- char' -- we would otherwise go into an infinite loop, since all
- characters <= 0xff. */
- ch[1] = '\0';
+#if _LIBC
+ collseq = (const unsigned char *) _NL_CURRENT (LC_COLLATE,
+ _NL_COLLATE_COLLSEQMB);
+
+ start_colseq = collseq[TRANSLATE (range_start_char)];
+ end_colseq = collseq[TRANSLATE (p[0])];
for (this_char = 0; this_char <= (unsigned char) -1; ++this_char)
{
- ch[0] = this_char;
- if (strcoll (range_start, ch) <= 0 && strcoll (ch, range_end) <= 0)
+ unsigned int this_colseq = collseq[TRANSLATE (this_char)];
+
+ if (start_colseq <= this_colseq && this_colseq <= end_colseq)
{
SET_LIST_BIT (TRANSLATE (this_char));
ret = REG_NOERROR;
}
}
+#else
+ /* Here we see why `this_char' has to be larger than an `unsigned
+ char' -- we would otherwise go into an infinite loop, since all
+ characters <= 0xff. */
+ range_start_char = TRANSLATE (range_start_char);
+ end_char = TRANSLATE (p[0]);
+ for (this_char = range_start_char; this_char <= end_char; ++this_char)
+ {
+ SET_LIST_BIT (TRANSLATE (this_char));
+ ret = REG_NOERROR;
+ }
+#endif
return ret;
}