8000 Add caching of ctype.h/wctype.h results in regc_locale.c. · postgres/postgres@e00f68e · GitHub
[go: up one dir, main page]
Skip to content

Commit e00f68e

Browse files
tglsfdctglsfdc
committed
Add caching of ctype.h/wctype.h results in regc_locale.c.
While this doesn't save a huge amount of runtime, it still seems worth doing, especially since I realized that the data copying I did in my first draft was quite unnecessary. In this version, once we have the results cached, getting them back for re-use is really very cheap. Also, remove the hard-wired limitation to not consider wctype.h results for character codes above 255. It turns out that we can't push the limit as far up as I'd originally hoped, because the regex colormap code is not efficient enough to cope very well with character classes containing many thousand letters, which a Unicode locale is entirely capable of producing. Still, we can push it up to U+7FF (which I chose as the limit of 2-byte UTF8 characters), which will at least make Eastern Europeans happy pending a better solution. Thus, this commit resolves the specific complaint in bug #6457, but not the more general issue that letters of non-western alphabets are mostly not recognized as matching [[:alpha:]].
1 parent 27af914 commit e00f68e

File tree

2 files changed

+260
-81
lines changed

2 files changed

+260
-81
lines changed

‎src/backend/regex/regc_locale.c

Lines changed: 39 additions & 80 deletions
Original file line numberDiff line numberDiff line change
@@ -350,6 +350,16 @@ static const struct cname
350350
};
351351

352352

353+
/*
354+
* We do not use the hard-wired Unicode classification tables that Tcl does.
355+
* This is because (a) we need to deal with other encodings besides Unicode,
356+
* and (b) we want to track the behavior of the libc locale routines as
357+
* closely as possible. For example, it wouldn't be unreasonable for a
358+
* locale to not consider every Unicode letter as a letter. So we build
359+
* character classification cvecs by asking libc, even for Unicode.
360+
*/
361+
362+
353363
/*
354364
* element - map collating-element name to celt
355365
*/
@@ -489,7 +499,11 @@ eclass(struct vars * v, /* context */
489499
/*
490500
* cclass - supply cvec for a character class
491501
*
492-
* Must include case counterparts on request.
502+
* Must include case counterparts if "cases" is true.
503+
*
504+
* The returned cvec might be either a transient cvec gotten from getcvec(),
505+
* or a permanently cached one from pg_ctype_get_cache(). This is okay
506+
* because callers are not supposed to explicitly free the result either way.
493507
*/
494508
static struct cvec *
495509
cclass(struct vars * v, /* context */
@@ -548,79 +562,54 @@ cclass(struct vars * v, /* context */
548562
index = (int) CC_ALPHA;
549563

550564
/*
551-
* Now compute the character class contents.
552-
*
553-
* For the moment, assume that only char codes < 256 can be in these
554-
* classes.
565+
* Now compute the character class contents. For classes that are
566+
* based on the behavior of a <wctype.h> or <ctype.h> function, we use
567+
* pg_ctype_get_cache so that we can cache the results. Other classes
568+
* have definitions that are hard-wired here, and for those we just
569+
* construct a transient cvec on the fly.
555570
*/
556571

557572
switch ((enum classes) index)
558573
{
559574
case CC_PRINT:
560-
cv = getcvec(v, UCHAR_MAX, 0);
561-
if (cv)
562-
{
563-
for (i = 0; i <= UCHAR_MAX; i++)
564-
{
565-
if (pg_wc_isprint((chr) i))
566-
addchr(cv, (chr) i);
567-
}
568-
}
575+
cv = pg_ctype_get_cache(pg_wc_isprint);
569576
break;
570577
case CC_ALNUM:
571-
cv = getcvec(v, UCHAR_MAX, 0);
572-
if (cv)
573-
{
574-
for (i = 0; i <= UCHAR_MAX; i++)
575-
{
576-
if (pg_wc_isalnum((chr) i))
577-
addchr(cv, (chr) i);
578-
}
579-
}
578+
cv = pg_ctype_get_cache(pg_wc_isalnum);
580579
break;
581580
case CC_ALPHA:
582-
cv = getcvec(v, UCHAR_MAX, 0);
583-
if (cv)
584-
{
585-
for (i = 0; i <= UCHAR_MAX; i++)
586-
{
587-
if (pg_wc_isalpha((chr) i))
588-
addchr(cv, (chr) i);
589-
}
590-
}
581+
cv = pg_ctype_get_cache(pg_wc_isalpha);
591582
break;
592583
case CC_ASCII:
584+
/* hard-wired meaning */
593585
cv = getcvec(v, 0, 1);
594586
if (cv)
595587
addrange(cv, 0, 0x7f);
596588
break;
597589
case CC_BLANK:
590+
/* hard-wired meaning */
598591
cv = getcvec(v, 2, 0);
599592
addchr(cv, '\t');
600593
addchr(cv, ' ');
601594
break;
602595
case CC_CNTRL:
596+
/* hard-wired meaning */
603597
cv = getcvec(v, 0, 2);
604598
addrange(cv, 0x0, 0x1f);
605599
addrange(cv, 0x7f, 0x9f);
606600
break;
607601
case CC_DIGIT:
608-
cv = getcvec(v, 0, 1);
609-
if (cv)
610-
addrange(cv, (chr) '0', (chr) '9');
602+
cv = pg_ctype_get_cache(pg_wc_isdigit);
611603
break;
612604
case CC_PUNCT:
613-
cv = getcvec(v, UCHAR_MAX, 0);
614-
if (cv)
615-
{
616-
for (i = 0; i <= UCHAR_MAX; i++)
617-
{
618-
if (pg_wc_ispunct((chr) i))
619-
addchr(cv, (chr) i);
620-
}
621-
}
605+
cv = pg_ctype_get_cache(pg_wc_ispunct);
622606
break;
623607
case CC_XDIGIT:
608+
/*
609+
* It's not clear how to define this in non-western locales, and
610+
* even less clear that there's any particular use in trying.
611+
* So just hard-wire the meaning.
612+
*/
624613
cv = getcvec(v, 0, 3);
625614
if (cv)
626615
{
@@ -630,50 +619,20 @@ cclass(struct vars * v, /* context */
630619
}
631620
break;
632621
case CC_SPACE:
633-
cv = getcvec(v, UCHAR_MAX, 0);
634-
if (cv)
635-
{
636-
for (i = 0; i <= UCHAR_MAX; i++)
637-
{
638-
if (pg_wc_isspace((chr) i))
639-
addchr(cv, (chr) i);
640-
}
641-
}
622+
cv = pg_ctype_get_cache(pg_wc_isspace);
642623
break;
643624
case CC_LOWER:
644-
cv = getcvec(v, UCHAR_MAX, 0);
645-
if (cv)
646-
{
647-
for (i = 0; i <= UCHAR_MAX; i++)
648-
{
649-
if (pg_wc_islower((chr) i))
650-
addchr(cv, (chr) i);
651-
}
652-
}
625+
cv = pg_ctype_get_cache(pg_wc_islower);
653626
break;
654627
case CC_UPPER:
655-
cv = getcvec(v, UCHAR_MAX, 0);
656-
if (cv)
657-
{
658-
for (i = 0; i <= UCHAR_MAX; i++)
659-
{
660-
if (pg_wc_isupper((chr) i))
661-
addchr(cv, (chr) i);
662-
}
663-
}
628+
cv = pg_ctype_get_cache(pg_wc_isupper);
664629
break;
665630
case CC_GRAPH:
666-
cv = getcvec(v, UCHAR_MAX, 0);
667-
if (cv)
668-
{
669-
for (i = 0; i <= UCHAR_MAX; i++)
670-
{
671-
if (pg_wc_isgraph((chr) i))
672-
addchr(cv, (chr) i);
673-
}
674-
}
631+
cv = pg_ctype_get_cache(pg_wc_isgraph);
675632
break;
676633
}
634+
635+
/* If cv is NULL now, the reason must be "out of memory" */
677636
if (cv == NULL)
678637
ERR(REG_ESPACE);
679638
return cv;

0 commit comments

Comments
 (0)
0