archive/ruby
1
0
Fork 0
mirror of https://github.com/ruby/ruby.git synced 2025-08-26 22:45:03 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 8

[ruby/json] json_string_unescape: assume the string doesn't need escaping

Browse source 
If that assumption holds true, then we don't need to copy the
string into a buffer to unescape it. For small string is just saves
copying, but for large ones it also saves a malloc/free combo.

Before:

```
== Parsing twitter.json (567916 bytes)
ruby 3.3.4 (2024-07-09 revision be1089c8ec) +YJIT [arm64-darwin23]
Warming up --------------------------------------
                json    52.000 i/100ms
                  oj    61.000 i/100ms
           oj strict    70.000 i/100ms
          Oj::Parser    71.000 i/100ms
           rapidjson    55.000 i/100ms
Calculating -------------------------------------
                json    510.111 (± 2.9%) i/s    (1.96 ms/i) -      2.548k in   5.000029s
                  oj    610.232 (± 3.1%) i/s    (1.64 ms/i) -      3.050k in   5.003725s
           oj strict    713.231 (± 3.2%) i/s    (1.40 ms/i) -      3.570k in   5.010902s
          Oj::Parser    762.598 (± 3.0%) i/s    (1.31 ms/i) -      3.834k in   5.033130s
           rapidjson    553.029 (± 7.4%) i/s    (1.81 ms/i) -      2.750k in   5.022630s

Comparison:
                json:      510.1 i/s
          Oj::Parser:      762.6 i/s - 1.49x  faster
           oj strict:      713.2 i/s - 1.40x  faster
                  oj:      610.2 i/s - 1.20x  faster
           rapidjson:      553.0 i/s - same-ish: difference falls within error

== Parsing citm_catalog.json (1727030 bytes)
ruby 3.3.4 (2024-07-09 revision be1089c8ec) +YJIT [arm64-darwin23]
Warming up --------------------------------------
                json    28.000 i/100ms
                  oj    33.000 i/100ms
           oj strict    37.000 i/100ms
          Oj::Parser    43.000 i/100ms
           rapidjson    38.000 i/100ms
Calculating -------------------------------------
                json    303.853 (± 3.6%) i/s    (3.29 ms/i) -      1.540k in   5.076079s
                  oj    348.009 (± 2.0%) i/s    (2.87 ms/i) -      1.749k in   5.027738s
           oj strict    396.679 (± 3.3%) i/s    (2.52 ms/i) -      1.998k in   5.042271s
          Oj::Parser    406.699 (± 2.2%) i/s    (2.46 ms/i) -      2.064k in   5.077587s
           rapidjson    393.463 (± 3.3%) i/s    (2.54 ms/i) -      1.976k in   5.028501s

Comparison:
                json:      303.9 i/s
          Oj::Parser:      406.7 i/s - 1.34x  faster
           oj strict:      396.7 i/s - 1.31x  faster
           rapidjson:      393.5 i/s - 1.29x  faster
                  oj:      348.0 i/s - 1.15x  faster
```

After:

```
== Parsing twitter.json (567916 bytes)
ruby 3.3.4 (2024-07-09 revision be1089c8ec) +YJIT [arm64-darwin23]
Warming up --------------------------------------
                json    56.000 i/100ms
                  oj    62.000 i/100ms
           oj strict    72.000 i/100ms
          Oj::Parser    77.000 i/100ms
           rapidjson    55.000 i/100ms
Calculating -------------------------------------
                json    568.025 (± 2.1%) i/s    (1.76 ms/i) -      2.856k in   5.030272s
                  oj    630.936 (± 1.4%) i/s    (1.58 ms/i) -      3.162k in   5.012630s
           oj strict    705.784 (±11.2%) i/s    (1.42 ms/i) -      3.456k in   5.006706s
          Oj::Parser    783.989 (± 1.7%) i/s    (1.28 ms/i) -      3.927k in   5.010343s
           rapidjson    557.630 (± 2.0%) i/s    (1.79 ms/i) -      2.805k in   5.032388s

Comparison:
                json:      568.0 i/s
          Oj::Parser:      784.0 i/s - 1.38x  faster
           oj strict:      705.8 i/s - 1.24x  faster
                  oj:      630.9 i/s - 1.11x  faster
           rapidjson:      557.6 i/s - same-ish: difference falls within error

== Parsing citm_catalog.json (1727030 bytes)
ruby 3.3.4 (2024-07-09 revision be1089c8ec) +YJIT [arm64-darwin23]
Warming up --------------------------------------
                json    29.000 i/100ms
                  oj    33.000 i/100ms
           oj strict    38.000 i/100ms
          Oj::Parser    43.000 i/100ms
           rapidjson    37.000 i/100ms
Calculating -------------------------------------
                json    319.271 (± 3.1%) i/s    (3.13 ms/i) -      1.595k in   5.001128s
                  oj    347.946 (± 1.7%) i/s    (2.87 ms/i) -      1.749k in   5.028395s
           oj strict    396.914 (± 3.0%) i/s    (2.52 ms/i) -      2.014k in   5.079645s
          Oj::Parser    409.311 (± 2.7%) i/s    (2.44 ms/i) -      2.064k in   5.046626s
           rapidjson    394.752 (± 1.5%) i/s    (2.53 ms/i) -      1.998k in   5.062776s

Comparison:
                json:      319.3 i/s
          Oj::Parser:      409.3 i/s - 1.28x  faster
           oj strict:      396.9 i/s - 1.24x  faster
           rapidjson:      394.8 i/s - 1.24x  faster
                  oj:      347.9 i/s - 1.09x  faster
```

7e0f66546a
This commit is contained in:
Jean Boussier 2024-10-31 14:47:17 +01:00 • committed by Hiroshi SHIBATA
parent 081689b9e2
commit 165cc6cf40
3 changed files with 57 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

17
ext/json/parser/parser.rl
View file

@ -461,7 +461,7 @@ static char *JSON_parse_array(JSON_Parser *json, char *p, char *pe, VALUE *resul
}
}
static inline VALUE build_string(const char *buffer, const char *bufferStart, bool intern, bool symbolize)
static inline VALUE build_string(const char *start, const char *end, bool intern, bool symbolize)
{
if (symbolize) {
intern = true;
@ -469,12 +469,12 @@ static inline VALUE build_string(const char *buffer, const char *bufferStart, bo
VALUE result;
# ifdef HAVE_RB_ENC_INTERNED_STR
if (intern) {
result = rb_enc_interned_str(bufferStart, (long)(buffer - bufferStart), rb_utf8_encoding());
result = rb_enc_interned_str(start, (long)(end - start), rb_utf8_encoding());
} else {
result = rb_utf8_str_new(bufferStart, (long)(buffer - bufferStart));
result = rb_utf8_str_new(start, (long)(end - start));
}
# else
result = rb_utf8_str_new(bufferStart, (long)(buffer - bufferStart));
result = rb_utf8_str_new(start, (long)(end - start));
if (intern) {
# if STR_UMINUS_DEDUPE_FROZEN
// Starting from MRI 3.0 it is preferable to freeze the string
@ -499,7 +499,7 @@ static inline VALUE build_string(const char *buffer, const char *bufferStart, bo
}
static const size_t MAX_STACK_BUFFER_SIZE = 128;
static VALUE json_string_unescape(char *string, char *stringEnd, int intern, int symbolize)
static VALUE json_string_unescape(char *string, char *stringEnd, bool intern, bool symbolize)
{
VALUE result = Qnil;
size_t bufferSize = stringEnd - string;
@ -507,6 +507,11 @@ static VALUE json_string_unescape(char *string, char *stringEnd, int intern, int
int unescape_len;
char buf[4];
pe = memchr(p, '\\', bufferSize);
if (RB_LIKELY(pe == NULL)) {
return build_string(string, stringEnd, intern, symbolize);
}
if (bufferSize > MAX_STACK_BUFFER_SIZE) {
# ifdef HAVE_RB_ENC_INTERNED_STR
bufferStart = buffer = ALLOC_N(char, bufferSize ? bufferSize : 1);
@ -609,7 +614,7 @@ static VALUE json_string_unescape(char *string, char *stringEnd, int intern, int
buffer += pe - p;
}
result = build_string(buffer, bufferStart, intern, symbolize);
result = build_string(bufferStart, buffer, intern, symbolize);
if (bufferSize > MAX_STACK_BUFFER_SIZE) {
ruby_xfree(bufferStart);