[ruby/json] Emit warnings when dumping binary strings

Because of it's Ruby 1.8 heritage, the C extension doesn't care
much about strings encoding. We should get stricter over time.

https://github.com/ruby/json/commit/42402fc13f
This commit is contained in:
Jean Boussier 2024-10-24 16:04:50 +02:00 committed by Hiroshi SHIBATA
parent f2b8829df0
commit 3782600f0f
5 changed files with 35 additions and 23 deletions

View File

@ -46,6 +46,8 @@ static void generate_json_fixnum(FBuffer *buffer, struct generate_json_data *dat
static void generate_json_bignum(FBuffer *buffer, struct generate_json_data *data, JSON_Generator_State *state, VALUE obj);
static void generate_json_float(FBuffer *buffer, struct generate_json_data *data, JSON_Generator_State *state, VALUE obj);
static int usascii_encindex, utf8_encindex, binary_encindex;
/* Converts in_string to a JSON string (without the wrapping '"'
* characters) in FBuffer out_buffer.
*
@ -535,7 +537,7 @@ static VALUE mString_to_json_raw_object(VALUE self)
VALUE result = rb_hash_new();
rb_hash_aset(result, rb_funcall(mJSON, i_create_id, 0), rb_class_name(rb_obj_class(self)));
ary = rb_funcall(self, i_unpack, 1, rb_str_new2("C*"));
rb_hash_aset(result, rb_str_new2("raw"), ary);
rb_hash_aset(result, rb_utf8_str_new_lit("raw"), ary);
return result;
}
@ -822,8 +824,6 @@ static void generate_json_array(FBuffer *buffer, struct generate_json_data *data
fbuffer_append_char(buffer, ']');
}
static int usascii_encindex, utf8_encindex, binary_encindex;
static inline int enc_utf8_compatible_p(int enc_idx)
{
if (enc_idx == usascii_encindex) return 1;
@ -837,13 +837,14 @@ static inline VALUE ensure_valid_encoding(VALUE str)
VALUE utf8_string;
if (RB_UNLIKELY(!enc_utf8_compatible_p(encindex))) {
if (encindex == binary_encindex) {
// For historical reason, we silently reinterpret binary strings as UTF-8 if it would work.
// TODO: Deprecate in 2.8.0
// TODO: Remove in 3.0.0
utf8_string = rb_enc_associate_index(rb_str_dup(str), utf8_encindex);
switch (rb_enc_str_coderange(utf8_string)) {
case ENC_CODERANGE_7BIT:
return utf8_string;
case ENC_CODERANGE_VALID:
// For historical reason, we silently reinterpret binary strings as UTF-8 if it would work.
// TODO: Raise in 3.0.0
rb_warn("JSON.generate: UTF-8 string passed as BINARY, this will raise an encoding error in json 3.0");
return utf8_string;
break;
}

View File

@ -35,7 +35,7 @@ class BigDecimal
def as_json(*)
{
JSON.create_id => self.class.name,
'b' => _dump,
'b' => _dump.force_encoding(Encoding::UTF_8),
}
end

View File

@ -1798,9 +1798,12 @@ static VALUE convert_encoding(VALUE source)
if (encindex == binary_encindex) {
// For historical reason, we silently reinterpret binary strings as UTF-8 if it would work.
// TODO: Deprecate in 2.8.0
// TODO: Remove in 3.0.0
return rb_enc_associate_index(rb_str_dup(source), utf8_encindex);
VALUE utf8_string = rb_enc_associate_index(rb_str_dup(source), utf8_encindex);
switch (rb_enc_str_coderange(utf8_string)) {
case ENC_CODERANGE_7BIT:
case ENC_CODERANGE_VALID:
return utf8_string;
}
}
return rb_str_conv_enc(source, rb_enc_from_index(encindex), rb_utf8_encoding());
@ -1955,7 +1958,7 @@ static VALUE cParser_initialize(int argc, VALUE *argv, VALUE self)
}
#line 1959 "parser.c"
#line 1962 "parser.c"
enum {JSON_start = 1};
enum {JSON_first_final = 10};
enum {JSON_error = 0};
@ -1963,7 +1966,7 @@ enum {JSON_error = 0};
enum {JSON_en_main = 1};
#line 867 "parser.rl"
#line 870 "parser.rl"
/*
@ -1981,16 +1984,16 @@ static VALUE cParser_parse(VALUE self)
GET_PARSER;
#line 1985 "parser.c"
#line 1988 "parser.c"
{
cs = JSON_start;
}
#line 884 "parser.rl"
#line 887 "parser.rl"
p = json->source;
pe = p + json->len;
#line 1994 "parser.c"
#line 1997 "parser.c"
{
if ( p == pe )
goto _test_eof;
@ -2024,7 +2027,7 @@ st0:
cs = 0;
goto _out;
tr2:
#line 859 "parser.rl"
#line 862 "parser.rl"
{
char *np = JSON_parse_value(json, p, pe, &result, 0);
if (np == NULL) { p--; {p++; cs = 10; goto _out;} } else {p = (( np))-1;}
@ -2034,7 +2037,7 @@ st10:
if ( ++p == pe )
goto _test_eof10;
case 10:
#line 2038 "parser.c"
#line 2041 "parser.c"
switch( (*p) ) {
case 13: goto st10;
case 32: goto st10;
@ -2123,7 +2126,7 @@ case 9:
_out: {}
}
#line 887 "parser.rl"
#line 890 "parser.rl"
if (cs >= JSON_first_final && p == pe) {
return result;

View File

@ -693,9 +693,12 @@ static VALUE convert_encoding(VALUE source)
if (encindex == binary_encindex) {
// For historical reason, we silently reinterpret binary strings as UTF-8 if it would work.
// TODO: Deprecate in 2.8.0
// TODO: Remove in 3.0.0
return rb_enc_associate_index(rb_str_dup(source), utf8_encindex);
VALUE utf8_string = rb_enc_associate_index(rb_str_dup(source), utf8_encindex);
switch (rb_enc_str_coderange(utf8_string)) {
case ENC_CODERANGE_7BIT:
case ENC_CODERANGE_VALID:
return utf8_string;
}
}
return rb_str_conv_enc(source, rb_enc_from_index(encindex), rb_utf8_encoding());

View File

@ -507,8 +507,13 @@ class JSONGeneratorTest < Test::Unit::TestCase
wrong_encoding_string = utf8_string.b
# This behavior is historical. Not necessary desirable. We should deprecated it.
# The pure and java version of the gem already don't behave this way.
assert_equal utf8_string.to_json, wrong_encoding_string.to_json
assert_equal JSON.dump(utf8_string), JSON.dump(wrong_encoding_string)
assert_warning(/UTF-8 string passed as BINARY, this will raise an encoding error in json 3.0/) do
assert_equal utf8_string.to_json, wrong_encoding_string.to_json
end
assert_warning(/UTF-8 string passed as BINARY, this will raise an encoding error in json 3.0/) do
assert_equal JSON.dump(utf8_string), JSON.dump(wrong_encoding_string)
end
end
def test_string_ext_included_calls_super