summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorNobuyoshi Nakada <[email protected]>2023-06-09 16:10:30 +0900
committerNobuyoshi Nakada <[email protected]>2023-06-09 20:22:30 +0900
commitab6eb3786c94e69c561080cbb796c2381702a3a4 ()
tree2fe551edd62566939a7760439851fd7ad90b2d13
parentd54f66d1b4ce32d78b526b1ea9e3f213a763d07c (diff)
Optimize `Regexp#dup` and `Regexp.new(/RE/)`
When copying from another regexp, copy already built `regex_t` instead of re-compiling its source.
Notes: Merged: https://.com/ruby/ruby/pull/7922
-rw-r--r--benchmark/regexp_dup.yml6
-rw-r--r--benchmark/regexp_new.yml7
-rw-r--r--include/ruby/onigmo.h2
-rw-r--r--re.c43
-rw-r--r--regcomp.c74
5 files changed, 125 insertions, 7 deletions
@@ -0,0 +1,6 @@
@@ -0,0 +1,7 @@
@@ -844,6 +844,8 @@ void onig_free(OnigRegex);
ONIG_EXTERN
void onig_free_body(OnigRegex);
ONIG_EXTERN
OnigPosition onig_scan(OnigRegex reg, const OnigUChar* str, const OnigUChar* end, OnigRegion* region, OnigOptionType option, int (*scan_callback)(OnigPosition, OnigPosition, OnigRegion*, void*), void* callback_arg);
ONIG_EXTERN
OnigPosition onig_search(OnigRegex, const OnigUChar* str, const OnigUChar* end, const OnigUChar* start, const OnigUChar* range, OnigRegion* region, OnigOptionType option);
@@ -3213,6 +3213,15 @@ rb_reg_preprocess_dregexp(VALUE ary, int options)
return result;
}
static int
rb_reg_initialize(VALUE obj, const char *s, long len, rb_encoding *enc,
int options, onig_errmsg_buffer err,
@@ -3223,10 +3232,7 @@ rb_reg_initialize(VALUE obj, const char *s, long len, rb_encoding *enc,
rb_encoding *fixed_enc = 0;
rb_encoding *a_enc = rb_ascii8bit_encoding();
- rb_check_frozen(obj);
- if (re->ptr)
- rb_raise(rb_eTypeError, "already initialized regexp");
- re->ptr = 0;
if (rb_enc_dummy_p(enc)) {
errcpy(err, "can't make regexp with dummy encoding");
@@ -3862,6 +3868,24 @@ set_timeout(rb_hrtime_t *hrt, VALUE timeout)
double2hrtime(hrt, timeout_d);
}
struct reg_init_args {
VALUE str;
VALUE timeout;
@@ -3931,9 +3955,14 @@ static VALUE
rb_reg_initialize_m(int argc, VALUE *argv, VALUE self)
{
struct reg_init_args args;
- reg_extract_args(argc, argv, &args);
- reg_init_args(self, args.str, args.enc, args.flags);
set_timeout(&RREGEXP_PTR(self)->timelimit, args.timeout);
@@ -4356,7 +4385,7 @@ rb_reg_init_copy(VALUE copy, VALUE re)
{
if (!OBJ_INIT_COPY(copy, re)) return copy;
rb_reg_check(re);
- return rb_reg_init_str(copy, RREGEXP_SRC(re), rb_reg_options(re));
}
VALUE
@@ -5671,6 +5671,80 @@ onig_free(regex_t* reg)
}
}
#ifdef RUBY
size_t
onig_memsize(const regex_t *reg)