1berry/ruby
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

String#gsub! Elide MatchData allocation when we know it can't escape

Browse Source 
In gsub is used with a string replacement or a map that doesn't
have a default proc, we know for sure no code can cause the MatchData
to escape the `gsub` call.

In such case, we still have to allocate a new MatchData because we
don't know what is the lifetime of the backref, but for any subsequent
match we can re-use the MatchData we allocated ourselves, reducing
allocations significantly.

This partially fixes [Misc #20652], except when a block is used,
and partially reduce the performance impact of
abc0304cb28cb9dcc3476993bc487884c139fd11 / [Bug #17507]

```
compare-ruby: ruby 3.5.0dev (2025-02-24T09:44:57Z master 5cf146399f) +PRISM [arm64-darwin24]
built-ruby: ruby 3.5.0dev (2025-02-24T10:58:27Z gsub-elude-match da966636e9) +PRISM [arm64-darwin24]
warming up....

|                 |compare-ruby|built-ruby|
|:----------------|-----------:|---------:|
|escape           |      3.577k|    3.697k|
|                 |           -|     1.03x|
|escape_bin       |      5.869k|    6.743k|
|                 |           -|     1.15x|
|escape_utf8      |      3.448k|    3.738k|
|                 |           -|     1.08x|
|escape_utf8_bin  |      6.361k|    7.267k|
|                 |           -|     1.14x|
```

Co-Authored-By: Étienne Barrié <etienne.barrie@gmail.com>
This commit is contained in:
Jean Boussier 2025-02-24 11:39:00 +01:00
parent 21ac0a3a64
commit 87f9c3c65e
3 changed files with 44 additions and 20 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
internal/re.h
View File

@ -14,12 +14,12 @@
/* re.c */
VALUE rb_reg_compile(VALUE str, int options, const char *sourcefile, int sourceline);
VALUE rb_reg_check_preprocess(VALUE);
long rb_reg_search0(VALUE, VALUE, long, int, int);
long rb_reg_search0(VALUE, VALUE, long, int, int, VALUE *);
VALUE rb_reg_match_p(VALUE re, VALUE str, long pos);
bool rb_reg_start_with_p(VALUE re, VALUE str);
VALUE rb_reg_hash(VALUE re);
VALUE rb_reg_equal(VALUE re1, VALUE re2);
void rb_backref_set_string(VALUE string, long pos, long len);
VALUE rb_backref_set_string(VALUE string, long pos, long len);
void rb_match_unbusy(VALUE);
int rb_match_count(VALUE match);
VALUE rb_reg_new_ary(VALUE ary, int options);

19
re.c
View File

@ -1521,7 +1521,7 @@ match_set_string(VALUE m, VALUE string, long pos, long len)
rmatch->regs.end[0] = pos + len;
}
void
VALUE
rb_backref_set_string(VALUE string, long pos, long len)
{
VALUE match = rb_backref_get();
@ -1530,6 +1530,7 @@ rb_backref_set_string(VALUE string, long pos, long len)
}
match_set_string(match, string, pos, len);
rb_backref_set(match);
return match;
}
/*
@ -1812,12 +1813,20 @@ rb_reg_search_set_match(VALUE re, VALUE str, long pos, int reverse, int set_back
return ONIG_MISMATCH;
}
VALUE match = match_alloc(rb_cMatch);
VALUE match;
if (set_match && RTEST(*set_match)) {
match = *set_match;
}
else {
match = match_alloc(rb_cMatch);
}
rb_matchext_t *rm = RMATCH_EXT(match);
rm->regs = regs;
if (set_backref_str) {
RB_OBJ_WRITE(match, &RMATCH(match)->str, rb_str_new4(str));
rb_obj_reveal(match, rb_cMatch);
}
else {
/* Note that a MatchData object with RMATCH(match)->str == 0 is incomplete!
@ -1835,15 +1844,15 @@ rb_reg_search_set_match(VALUE re, VALUE str, long pos, int reverse, int set_back
}
long
rb_reg_search0(VALUE re, VALUE str, long pos, int reverse, int set_backref_str)
rb_reg_search0(VALUE re, VALUE str, long pos, int reverse, int set_backref_str, VALUE *match)
{
return rb_reg_search_set_match(re, str, pos, reverse, set_backref_str, NULL);
return rb_reg_search_set_match(re, str, pos, reverse, set_backref_str, match);
}
long
rb_reg_search(VALUE re, VALUE str, long pos, int reverse)
{
return rb_reg_search0(re, str, pos, reverse, 1);
return rb_reg_search_set_match(re, str, pos, reverse, 1, NULL);
}
static OnigPosition

41
string.c
View File

@ -6122,14 +6122,17 @@ get_pat_quoted(VALUE pat, int check)
}
static long
rb_pat_search(VALUE pat, VALUE str, long pos, int set_backref_str)
rb_pat_search0(VALUE pat, VALUE str, long pos, int set_backref_str, VALUE *match)
{
if (BUILTIN_TYPE(pat) == T_STRING) {
pos = rb_str_byteindex(str, pat, pos);
if (set_backref_str) {
if (pos >= 0) {
str = rb_str_new_frozen_String(str);
rb_backref_set_string(str, pos, RSTRING_LEN(pat));
VALUE match_data = rb_backref_set_string(str, pos, RSTRING_LEN(pat));
if (match) {
*match = match_data;
}
}
else {
rb_backref_set(Qnil);
@ -6138,10 +6141,16 @@ rb_pat_search(VALUE pat, VALUE str, long pos, int set_backref_str)
return pos;
}
else {
return rb_reg_search0(pat, str, pos, 0, set_backref_str);
return rb_reg_search0(pat, str, pos, 0, set_backref_str, match);
}
}
static long
rb_pat_search(VALUE pat, VALUE str, long pos, int set_backref_str)
{
return rb_pat_search0(pat, str, pos, set_backref_str, NULL);
}
/*
* call-seq:
@ -6293,12 +6302,12 @@ rb_str_sub(int argc, VALUE *argv, VALUE str)
static VALUE
str_gsub(int argc, VALUE *argv, VALUE str, int bang)
{
VALUE pat, val = Qnil, repl, match0 = Qnil, dest, hash = Qnil;
VALUE pat, val = Qnil, repl, match0 = Qnil, dest, hash = Qnil, match = Qnil;
long beg, beg0, end0;
long offset, blen, slen, len, last;
enum {STR, ITER, FAST_MAP, MAP} mode = STR;
char *sp, *cp;
int need_backref = -1;
int need_backref_str = -1;
rb_encoding *str_enc;
switch (argc) {
@ -6324,7 +6333,8 @@ str_gsub(int argc, VALUE *argv, VALUE str, int bang)
}
pat = get_pat_quoted(argv[0], 1);
beg = rb_pat_search(pat, str, 0, need_backref);
beg = rb_pat_search0(pat, str, 0, need_backref_str, &match);
if (beg < 0) {
if (bang) return Qnil; /* no match, no substitution */
return str_duplicate(rb_cString, str);
@ -6341,7 +6351,6 @@ str_gsub(int argc, VALUE *argv, VALUE str, int bang)
ENC_CODERANGE_SET(dest, rb_enc_asciicompat(str_enc) ? ENC_CODERANGE_7BIT : ENC_CODERANGE_VALID);
do {
VALUE match = rb_backref_get();
struct re_registers *regs = RMATCH_REGS(match);
if (RB_TYPE_P(pat, T_STRING)) {
beg0 = beg;
@ -6354,7 +6363,7 @@ str_gsub(int argc, VALUE *argv, VALUE str, int bang)
if (mode == ITER) match0 = rb_reg_nth_match(0, match);
}
if (mode) {
if (mode != STR) {
if (mode == ITER) {
val = rb_obj_as_string(rb_yield(match0));
}
@ -6378,10 +6387,10 @@ str_gsub(int argc, VALUE *argv, VALUE str, int bang)
rb_raise(rb_eRuntimeError, "block should not cheat");
}
}
else if (need_backref) {
else if (need_backref_str) {
val = rb_reg_regsub(repl, str, regs, RB_TYPE_P(pat, T_STRING) ? Qnil : pat);
if (need_backref < 0) {
need_backref = val != repl;
if (need_backref_str < 0) {
need_backref_str = val != repl;
}
}
else {
@ -6409,14 +6418,20 @@ str_gsub(int argc, VALUE *argv, VALUE str, int bang)
}
cp = RSTRING_PTR(str) + offset;
if (offset > RSTRING_LEN(str)) break;
beg = rb_pat_search(pat, str, offset, need_backref);
// In FAST_MAP and STR mode the backref can't escape so we can re-use the MatchData safely.
if (mode != FAST_MAP && mode != STR) {
match = Qnil;
}
beg = rb_pat_search0(pat, str, offset, need_backref_str, &match);
RB_GC_GUARD(match);
} while (beg >= 0);
if (RSTRING_LEN(str) > offset) {
rb_enc_str_buf_cat(dest, cp, RSTRING_LEN(str) - offset, str_enc);
}
rb_pat_search(pat, str, last, 1);
rb_pat_search0(pat, str, last, 1, &match);
if (bang) {
str_shared_replace(str, dest);
}