json 2.18.0 → 2.18.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 51eab66896e862b679d424133f11e1367d5d8e71add943e67cf0673d0d562fcd
-  data.tar.gz: 7b69d4a42137897fe9a45bd60a21b759d133c112cb4ba16020099f27074ac2fd
+  metadata.gz: cb2890db4c527125d27bc7c21fc64d3ac532ffbec8080f89a678daf48c36e09e
+  data.tar.gz: c4b37d085d05d3c43df97b3c24898dc6be61c76ba64c749b5a8a86bf4fc1198d
 SHA512:
-  metadata.gz: ea3b026c8ccd6cb477858bf06f07f8b5adc5bcf7b52a175487c19dc2835ef63db2e4f87074a00ec2fe2c70e588c205f679116536da40f15e767f35351a52fc5c
-  data.tar.gz: f58144a5329ad95128e00bbc5670280f6a699e04cf05060bdfc7acaf112e62eb44c14c36328a9683d86e138c6eb2f3599f5ec35347867ac99ab9f0e16813c4df
+  metadata.gz: fb55ef5a0aa6961ef0fe3bb30f398834820357045ad27a8fdb7e53eaba3af7c4d356ef26c0e73b7a87d2d9d51e500eae7193d7d1ae3aa1058c7973bcc462674b
+  data.tar.gz: bfb499789bbcee7f5f8d67e32ded664dc62c632ae39fe80bf4bff3d6aec16eee3730a7c4883216a393326f2ab33e94e5f9c58da4c5b31627347108c36c2b211c

data/CHANGES.md

@@ -2,6 +2,11 @@
 
 ### Unreleased
 
+### 2026-02-03 (2.18.1)
+
+* Fix a potential crash in very specific circumstance if GC triggers during a call to `to_json`
+  without first invoking a user defined `#to_json` method.
+
 ### 2025-12-11 (2.18.0)
 
 * Add `:allow_control_characters` parser options, to allow JSON strings containing unescaped ASCII control characters (e.g. newlines).
@@ -66,7 +71,7 @@
 * Fix `JSON.generate` `strict: true` mode to also restrict hash keys.
 * Fix `JSON::Coder` to also invoke block for hash keys that aren't strings nor symbols.
 * Fix `JSON.unsafe_load` usage with proc
-* Fix the parser to more consistently reject invalid UTF-16 surogate pairs. 
+* Fix the parser to more consistently reject invalid UTF-16 surogate pairs.
 * Stop defining `String.json_create`, `String#to_json_raw`, `String#to_json_raw_object` when `json/add` isn't loaded.
 
 ### 2025-07-28 (2.13.2)

data/ext/json/ext/fbuffer/fbuffer.h

@@ -161,23 +161,25 @@ static inline void fbuffer_append_reserved_char(FBuffer *fb, char chr)
 
 static void fbuffer_append_str(FBuffer *fb, VALUE str)
 {
-    const char *newstr = StringValuePtr(str);
-    unsigned long len = RSTRING_LEN(str);
+    const char *ptr;
+    unsigned long len;
+    RSTRING_GETMEM(str, ptr, len);
 
-    fbuffer_append(fb, newstr, len);
+    fbuffer_append(fb, ptr, len);
 }
 
 static void fbuffer_append_str_repeat(FBuffer *fb, VALUE str, size_t repeat)
 {
-    const char *newstr = StringValuePtr(str);
-    unsigned long len = RSTRING_LEN(str);
+    const char *ptr;
+    unsigned long len;
+    RSTRING_GETMEM(str, ptr, len);
 
     fbuffer_inc_capa(fb, repeat * len);
     while (repeat) {
 #if JSON_DEBUG
         fb->requested = len;
 #endif
-        fbuffer_append_reserved(fb, newstr, len);
+        fbuffer_append_reserved(fb, ptr, len);
         repeat--;
     }
 }

data/ext/json/ext/generator/generator.c

@@ -63,6 +63,8 @@ struct generate_json_data {
     long depth;
 };
 
+static SIMD_Implementation simd_impl;
+
 static VALUE cState_from_state_s(VALUE self, VALUE opts);
 static VALUE cState_partial_generate(VALUE self, VALUE obj, generator_func, VALUE io);
 static void generate_json(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
@@ -155,8 +157,6 @@ static const unsigned char escape_table_basic[256] = {
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
 };
 
-static unsigned char (*search_escape_basic_impl)(search_state *);
-
 static inline unsigned char search_escape_basic(search_state *search)
 {
     while (search->ptr < search->end) {
@@ -212,11 +212,39 @@ ALWAYS_INLINE(static) void escape_UTF8_char_basic(search_state *search)
  * Everything else (should be UTF-8) is just passed through and
  * appended to the result.
  */
+
+
+#if defined(HAVE_SIMD_NEON)
+static inline unsigned char search_escape_basic_neon(search_state *search);
+#elif defined(HAVE_SIMD_SSE2)
+static inline unsigned char search_escape_basic_sse2(search_state *search);
+#endif
+
+static inline unsigned char search_escape_basic(search_state *search);
+
 static inline void convert_UTF8_to_JSON(search_state *search)
 {
-    while (search_escape_basic_impl(search)) {
+#ifdef HAVE_SIMD
+#if defined(HAVE_SIMD_NEON)
+    while (search_escape_basic_neon(search)) {
+        escape_UTF8_char_basic(search);
+    }
+#elif defined(HAVE_SIMD_SSE2)
+    if (simd_impl == SIMD_SSE2) {
+        while (search_escape_basic_sse2(search)) {
+            escape_UTF8_char_basic(search);
+        }
+        return;
+    }
+    while (search_escape_basic(search)) {
+        escape_UTF8_char_basic(search);
+    }
+#endif
+#else
+    while (search_escape_basic(search)) {
         escape_UTF8_char_basic(search);
     }
+#endif /* HAVE_SIMD */
 }
 
 static inline void escape_UTF8_char(search_state *search, unsigned char ch_len)
@@ -260,6 +288,8 @@ static inline void escape_UTF8_char(search_state *search, unsigned char ch_len)
 
 ALWAYS_INLINE(static) char *copy_remaining_bytes(search_state *search, unsigned long vec_len, unsigned long len)
 {
+    RBIMPL_ASSERT_OR_ASSUME(len < vec_len);
+
     // Flush the buffer so everything up until the last 'len' characters are unflushed.
     search_flush(search);
 
@@ -269,12 +299,18 @@ ALWAYS_INLINE(static) char *copy_remaining_bytes(search_state *search, unsigned
     char *s = (buf->ptr + buf->len);
 
     // Pad the buffer with dummy characters that won't need escaping.
-    // This seem wateful at first sight, but memset of vector length is very fast.
-    memset(s, 'X', vec_len);
+    // This seem wasteful at first sight, but memset of vector length is very fast.
+    // This is a space as it can be directly represented as an immediate on AArch64.
+    memset(s, ' ', vec_len);
 
     // Optimistically copy the remaining 'len' characters to the output FBuffer. If there are no characters
     // to escape, then everything ends up in the correct spot. Otherwise it was convenient temporary storage.
-    MEMCPY(s, search->ptr, char, len);
+    if (vec_len == 16) {
+        RBIMPL_ASSERT_OR_ASSUME(len >= SIMD_MINIMUM_THRESHOLD);
+        json_fast_memcpy16(s, search->ptr, len);
+    } else {
+        MEMCPY(s, search->ptr, char, len);
+    }
 
     return s;
 }
@@ -1091,6 +1127,7 @@ static void raw_generate_json_string(FBuffer *buffer, struct generate_json_data
     search.matches_mask = 0;
     search.has_matches = false;
     search.chunk_base = NULL;
+    search.chunk_end = NULL;
 #endif /* HAVE_SIMD */
 
     switch (rb_enc_str_coderange(obj)) {
@@ -1337,7 +1374,7 @@ static void generate_json_fixnum(FBuffer *buffer, struct generate_json_data *dat
 static void generate_json_bignum(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
 {
     VALUE tmp = rb_funcall(obj, i_to_s, 0);
-    fbuffer_append_str(buffer, tmp);
+    fbuffer_append_str(buffer, StringValue(tmp));
 }
 
 #ifdef RUBY_INTEGER_UNIFICATION
@@ -1503,7 +1540,9 @@ static VALUE cState_partial_generate(VALUE self, VALUE obj, generator_func func,
         .obj = obj,
         .func = func
     };
-    return rb_ensure(generate_json_try, (VALUE)&data, generate_json_ensure, (VALUE)&data);
+    VALUE result = rb_ensure(generate_json_try, (VALUE)&data, generate_json_ensure, (VALUE)&data);
+    RB_GC_GUARD(self);
+    return result;
 }
 
 /* call-seq:
@@ -2181,22 +2220,5 @@ void Init_generator(void)
 
     rb_require("json/ext/generator/state");
 
-
-    switch (find_simd_implementation()) {
-#ifdef HAVE_SIMD
-#ifdef HAVE_SIMD_NEON
-        case SIMD_NEON:
-            search_escape_basic_impl = search_escape_basic_neon;
-            break;
-#endif /* HAVE_SIMD_NEON */
-#ifdef HAVE_SIMD_SSE2
-        case SIMD_SSE2:
-            search_escape_basic_impl = search_escape_basic_sse2;
-            break;
-#endif /* HAVE_SIMD_SSE2 */
-#endif /* HAVE_SIMD */
-        default:
-            search_escape_basic_impl = search_escape_basic;
-            break;
-    }
+    simd_impl = find_simd_implementation();
 }

data/ext/json/ext/json.h

@@ -5,6 +5,10 @@
 #include "ruby/encoding.h"
 #include <stdint.h>
 
+#ifndef RBIMPL_ASSERT_OR_ASSUME
+# define RBIMPL_ASSERT_OR_ASSUME(x)
+#endif
+
 #if defined(RUBY_DEBUG) && RUBY_DEBUG
 # define JSON_ASSERT RUBY_ASSERT
 #else

data/ext/json/ext/parser/parser.c

@@ -477,23 +477,24 @@ static const signed char digit_values[256] = {
     -1, -1, -1, -1, -1, -1, -1
 };
 
-static uint32_t unescape_unicode(JSON_ParserState *state, const unsigned char *p)
+static uint32_t unescape_unicode(JSON_ParserState *state, const char *sp, const char *spe)
 {
-    signed char b;
-    uint32_t result = 0;
-    b = digit_values[p[0]];
-    if (b < 0) raise_parse_error_at("incomplete unicode character escape sequence at %s", state, (char *)p - 2);
-    result = (result << 4) | (unsigned char)b;
-    b = digit_values[p[1]];
-    if (b < 0) raise_parse_error_at("incomplete unicode character escape sequence at %s", state, (char *)p - 2);
-    result = (result << 4) | (unsigned char)b;
-    b = digit_values[p[2]];
-    if (b < 0) raise_parse_error_at("incomplete unicode character escape sequence at %s", state, (char *)p - 2);
-    result = (result << 4) | (unsigned char)b;
-    b = digit_values[p[3]];
-    if (b < 0) raise_parse_error_at("incomplete unicode character escape sequence at %s", state, (char *)p - 2);
-    result = (result << 4) | (unsigned char)b;
-    return result;
+    if (RB_UNLIKELY(sp > spe - 4)) {
+        raise_parse_error_at("incomplete unicode character escape sequence at %s", state, sp - 2);
+    }
+
+    const unsigned char *p = (const unsigned char *)sp;
+
+    const signed char b0 = digit_values[p[0]];
+    const signed char b1 = digit_values[p[1]];
+    const signed char b2 = digit_values[p[2]];
+    const signed char b3 = digit_values[p[3]];
+
+    if (RB_UNLIKELY((signed char)(b0 | b1 | b2 | b3) < 0)) {
+        raise_parse_error_at("incomplete unicode character escape sequence at %s", state, sp - 2);
+    }
+
+    return ((uint32_t)b0 << 12) | ((uint32_t)b1 << 8) | ((uint32_t)b2 << 4) | (uint32_t)b3;
 }
 
 #define GET_PARSER_CONFIG                          \
@@ -643,7 +644,7 @@ static inline VALUE json_string_fastpath(JSON_ParserState *state, JSON_ParserCon
 typedef struct _json_unescape_positions {
     long size;
     const char **positions;
-    bool has_more;
+    unsigned long additional_backslashes;
 } JSON_UnescapePositions;
 
 static inline const char *json_next_backslash(const char *pe, const char *stringEnd, JSON_UnescapePositions *positions)
@@ -657,7 +658,8 @@ static inline const char *json_next_backslash(const char *pe, const char *string
         }
     }
 
-    if (positions->has_more) {
+    if (positions->additional_backslashes) {
+        positions->additional_backslashes--;
         return memchr(pe, '\\', stringEnd - pe);
     }
 
@@ -707,50 +709,41 @@ NOINLINE(static) VALUE json_string_unescape(JSON_ParserState *state, JSON_Parser
             case 'f':
                 APPEND_CHAR('\f');
                 break;
-            case 'u':
-                if (pe > stringEnd - 5) {
-                    raise_parse_error_at("incomplete unicode character escape sequence at %s", state, p);
-                } else {
-                    uint32_t ch = unescape_unicode(state, (unsigned char *) ++pe);
-                    pe += 3;
-                    /* To handle values above U+FFFF, we take a sequence of
-                     * \uXXXX escapes in the U+D800..U+DBFF then
-                     * U+DC00..U+DFFF ranges, take the low 10 bits from each
-                     * to make a 20-bit number, then add 0x10000 to get the
-                     * final codepoint.
-                     *
-                     * See Unicode 15: 3.8 "Surrogates", 5.3 "Handling
-                     * Surrogate Pairs in UTF-16", and 23.6 "Surrogates
-                     * Area".
-                     */
-                    if ((ch & 0xFC00) == 0xD800) {
-                        pe++;
-                        if (pe > stringEnd - 6) {
-                            raise_parse_error_at("incomplete surrogate pair at %s", state, p);
-                        }
-                        if (pe[0] == '\\' && pe[1] == 'u') {
-                            uint32_t sur = unescape_unicode(state, (unsigned char *) pe + 2);
-
-                            if ((sur & 0xFC00) != 0xDC00) {
-                                raise_parse_error_at("invalid surrogate pair at %s", state, p);
-                            }
-
-                            ch = (((ch & 0x3F) << 10) | ((((ch >> 6) & 0xF) + 1) << 16)
-                                    | (sur & 0x3FF));
-                            pe += 5;
-                        } else {
-                            raise_parse_error_at("incomplete surrogate pair at %s", state, p);
-                            break;
+            case 'u': {
+                uint32_t ch = unescape_unicode(state, ++pe, stringEnd);
+                pe += 3;
+                /* To handle values above U+FFFF, we take a sequence of
+                 * \uXXXX escapes in the U+D800..U+DBFF then
+                 * U+DC00..U+DFFF ranges, take the low 10 bits from each
+                 * to make a 20-bit number, then add 0x10000 to get the
+                 * final codepoint.
+                 *
+                 * See Unicode 15: 3.8 "Surrogates", 5.3 "Handling
+                 * Surrogate Pairs in UTF-16", and 23.6 "Surrogates
+                 * Area".
+                 */
+                if ((ch & 0xFC00) == 0xD800) {
+                    pe++;
+                    if (RB_LIKELY((pe <= stringEnd - 6) && memcmp(pe, "\\u", 2) == 0)) {
+                        uint32_t sur = unescape_unicode(state, pe + 2, stringEnd);
+
+                        if (RB_UNLIKELY((sur & 0xFC00) != 0xDC00)) {
+                            raise_parse_error_at("invalid surrogate pair at %s", state, p);
                         }
-                    }
 
-                    char buf[4];
-                    int unescape_len = convert_UTF32_to_UTF8(buf, ch);
-                    MEMCPY(buffer, buf, char, unescape_len);
-                    buffer += unescape_len;
-                    p = ++pe;
+                        ch = (((ch & 0x3F) << 10) | ((((ch >> 6) & 0xF) + 1) << 16) | (sur & 0x3FF));
+                        pe += 5;
+                    } else {
+                        raise_parse_error_at("incomplete surrogate pair at %s", state, p);
+                        break;
+                    }
                 }
+
+                int unescape_len = convert_UTF32_to_UTF8(buffer, ch);
+                buffer += unescape_len;
+                p = ++pe;
                 break;
+            }
             default:
                 if ((unsigned char)*pe < 0x20) {
                     if (!config->allow_control_characters) {
@@ -992,7 +985,7 @@ static VALUE json_parse_escaped_string(JSON_ParserState *state, JSON_ParserConfi
     JSON_UnescapePositions positions = {
         .size = 0,
         .positions = backslashes,
-        .has_more = false,
+        .additional_backslashes = 0,
     };
 
     do {
@@ -1007,7 +1000,7 @@ static VALUE json_parse_escaped_string(JSON_ParserState *state, JSON_ParserConfi
                     backslashes[positions.size] = state->cursor;
                     positions.size++;
                 } else {
-                    positions.has_more = true;
+                    positions.additional_backslashes++;
                 }
                 state->cursor++;
                 break;

data/ext/json/ext/simd/simd.h

@@ -58,7 +58,34 @@ static inline int trailing_zeros(int input)
 
 #ifdef JSON_ENABLE_SIMD
 
-#define SIMD_MINIMUM_THRESHOLD 6
+#define SIMD_MINIMUM_THRESHOLD 4
+
+ALWAYS_INLINE(static) void json_fast_memcpy16(char *dst, const char *src, size_t len)
+{
+    RBIMPL_ASSERT_OR_ASSUME(len < 16);
+    RBIMPL_ASSERT_OR_ASSUME(len >= SIMD_MINIMUM_THRESHOLD); // 4
+#if defined(__has_builtin) && __has_builtin(__builtin_memcpy)
+    // If __builtin_memcpy is available, use it to copy between SIMD_MINIMUM_THRESHOLD (4) and vec_len-1 (15) bytes.
+    // These copies overlap. The first copy will copy the first 8 (or 4) bytes. The second copy will copy
+    // the last 8 (or 4) bytes but overlap with the first copy. The overlapping bytes will be in the correct
+    // position in both copies.
+
+    // Please do not attempt to replace __builtin_memcpy with memcpy without profiling and/or looking at the
+    // generated assembly. On clang-specifically (tested on Apple clang version 17.0.0 (clang-1700.0.13.3)),
+    // when using memcpy, the compiler will notice the only difference is a 4 or 8 and generate a conditional
+    // select instruction instead of direct loads and stores with a branch. This ends up slower than the branch
+    // plus two loads and stores generated when using __builtin_memcpy.
+    if (len >= 8) {
+        __builtin_memcpy(dst, src, 8);
+        __builtin_memcpy(dst + len - 8, src + len - 8, 8);
+    } else {
+        __builtin_memcpy(dst, src, 4);
+        __builtin_memcpy(dst + len - 4, src + len - 4, 4);
+    }
+#else
+    MEMCPY(dst, src, char, len);
+#endif
+}
 
 #if defined(__ARM_NEON) || defined(__ARM_NEON__) || defined(__aarch64__) || defined(_M_ARM64)
 #include <arm_neon.h>

data/ext/json/ext/vendor/fpconv.c

@@ -449,7 +449,7 @@ static int filter_special(double fp, char* dest)
  * }
  *
  */
-static int fpconv_dtoa(double d, char dest[28])
+static int fpconv_dtoa(double d, char dest[32])
 {
     char digits[18];
 

data/lib/json/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module JSON
-  VERSION = '2.18.0'
+  VERSION = '2.18.1'
 end

data/lib/json.rb

@@ -6,6 +6,15 @@ require 'json/common'
 #
 # \JSON is a lightweight data-interchange format.
 #
+# \JSON is easy for us humans to read and write,
+# and equally simple for machines to read (parse) and write (generate).
+#
+# \JSON is language-independent, making it an ideal interchange format
+# for applications in differing programming languages
+# and on differing operating systems.
+#
+# == \JSON Values
+#
 # A \JSON value is one of the following:
 # - Double-quoted text:  <tt>"foo"</tt>.
 # - Number:  +1+, +1.0+, +2.0e2+.
@@ -173,6 +182,18 @@ require 'json/common'
 # When enabled:
 #   JSON.parse('[1,]', allow_trailing_comma: true) # => [1]
 #
+# ---
+#
+# Option +allow_control_characters+ (boolean) specifies whether to allow
+# unescaped ASCII control characters, such as newlines, in strings;
+# defaults to +false+.
+#
+# With the default, +false+:
+#   JSON.parse(%{"Hello\nWorld"}) # invalid ASCII control character in string (JSON::ParserError)
+#
+# When enabled:
+#   JSON.parse(%{"Hello\nWorld"}, allow_control_characters: true) # => "Hello\nWorld"
+#
 # ====== Output Options
 #
 # Option +freeze+ (boolean) specifies whether the returned objects will be frozen;

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: json
 version: !ruby/object:Gem::Version
-  version: 2.18.0
+  version: 2.18.1
 platform: ruby
 authors:
 - Florian Frank
@@ -84,7 +84,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.6.9
+rubygems_version: 4.1.0.dev
 specification_version: 4
 summary: JSON Implementation for Ruby
 test_files: []