summaryrefslogtreecommitdiff
path: root/ext/json/generator
diff options
context:
space:
mode:
Diffstat (limited to 'ext/json/generator')
-rw-r--r--ext/json/generator/depend171
-rw-r--r--ext/json/generator/extconf.rb19
-rw-r--r--ext/json/generator/generator.c2472
-rw-r--r--ext/json/generator/generator.h171
4 files changed, 1697 insertions, 1136 deletions
diff --git a/ext/json/generator/depend b/ext/json/generator/depend
index 54e8ae3eb4..3ba4acfdd2 100644
--- a/ext/json/generator/depend
+++ b/ext/json/generator/depend
@@ -1,13 +1,174 @@
$(OBJS): $(ruby_headers)
-generator.o: generator.c generator.h $(srcdir)/../fbuffer/fbuffer.h
+generator.o: generator.c $(srcdir)/../fbuffer/fbuffer.h
# AUTOGENERATED DEPENDENCIES START
generator.o: $(RUBY_EXTCONF_H)
generator.o: $(arch_hdrdir)/ruby/config.h
+generator.o: $(hdrdir)/ruby.h
+generator.o: $(hdrdir)/ruby/assert.h
generator.o: $(hdrdir)/ruby/backward.h
+generator.o: $(hdrdir)/ruby/backward/2/assume.h
+generator.o: $(hdrdir)/ruby/backward/2/attributes.h
+generator.o: $(hdrdir)/ruby/backward/2/bool.h
+generator.o: $(hdrdir)/ruby/backward/2/inttypes.h
+generator.o: $(hdrdir)/ruby/backward/2/limits.h
+generator.o: $(hdrdir)/ruby/backward/2/long_long.h
+generator.o: $(hdrdir)/ruby/backward/2/stdalign.h
+generator.o: $(hdrdir)/ruby/backward/2/stdarg.h
generator.o: $(hdrdir)/ruby/defines.h
generator.o: $(hdrdir)/ruby/encoding.h
generator.o: $(hdrdir)/ruby/intern.h
+generator.o: $(hdrdir)/ruby/internal/abi.h
+generator.o: $(hdrdir)/ruby/internal/anyargs.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/char.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/double.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/fixnum.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/gid_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/int.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/intptr_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/long.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/long_long.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/mode_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/off_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/pid_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/short.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/size_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/st_data_t.h
+generator.o: $(hdrdir)/ruby/internal/arithmetic/uid_t.h
+generator.o: $(hdrdir)/ruby/internal/assume.h
+generator.o: $(hdrdir)/ruby/internal/attr/alloc_size.h
+generator.o: $(hdrdir)/ruby/internal/attr/artificial.h
+generator.o: $(hdrdir)/ruby/internal/attr/cold.h
+generator.o: $(hdrdir)/ruby/internal/attr/const.h
+generator.o: $(hdrdir)/ruby/internal/attr/constexpr.h
+generator.o: $(hdrdir)/ruby/internal/attr/deprecated.h
+generator.o: $(hdrdir)/ruby/internal/attr/diagnose_if.h
+generator.o: $(hdrdir)/ruby/internal/attr/enum_extensibility.h
+generator.o: $(hdrdir)/ruby/internal/attr/error.h
+generator.o: $(hdrdir)/ruby/internal/attr/flag_enum.h
+generator.o: $(hdrdir)/ruby/internal/attr/forceinline.h
+generator.o: $(hdrdir)/ruby/internal/attr/format.h
+generator.o: $(hdrdir)/ruby/internal/attr/maybe_unused.h
+generator.o: $(hdrdir)/ruby/internal/attr/noalias.h
+generator.o: $(hdrdir)/ruby/internal/attr/nodiscard.h
+generator.o: $(hdrdir)/ruby/internal/attr/noexcept.h
+generator.o: $(hdrdir)/ruby/internal/attr/noinline.h
+generator.o: $(hdrdir)/ruby/internal/attr/nonnull.h
+generator.o: $(hdrdir)/ruby/internal/attr/noreturn.h
+generator.o: $(hdrdir)/ruby/internal/attr/packed_struct.h
+generator.o: $(hdrdir)/ruby/internal/attr/pure.h
+generator.o: $(hdrdir)/ruby/internal/attr/restrict.h
+generator.o: $(hdrdir)/ruby/internal/attr/returns_nonnull.h
+generator.o: $(hdrdir)/ruby/internal/attr/warning.h
+generator.o: $(hdrdir)/ruby/internal/attr/weakref.h
+generator.o: $(hdrdir)/ruby/internal/cast.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/apple.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/clang.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/gcc.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/intel.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/msvc.h
+generator.o: $(hdrdir)/ruby/internal/compiler_is/sunpro.h
+generator.o: $(hdrdir)/ruby/internal/compiler_since.h
+generator.o: $(hdrdir)/ruby/internal/config.h
+generator.o: $(hdrdir)/ruby/internal/constant_p.h
+generator.o: $(hdrdir)/ruby/internal/core.h
+generator.o: $(hdrdir)/ruby/internal/core/rarray.h
+generator.o: $(hdrdir)/ruby/internal/core/rbasic.h
+generator.o: $(hdrdir)/ruby/internal/core/rbignum.h
+generator.o: $(hdrdir)/ruby/internal/core/rclass.h
+generator.o: $(hdrdir)/ruby/internal/core/rdata.h
+generator.o: $(hdrdir)/ruby/internal/core/rfile.h
+generator.o: $(hdrdir)/ruby/internal/core/rhash.h
+generator.o: $(hdrdir)/ruby/internal/core/rmatch.h
+generator.o: $(hdrdir)/ruby/internal/core/robject.h
+generator.o: $(hdrdir)/ruby/internal/core/rregexp.h
+generator.o: $(hdrdir)/ruby/internal/core/rstring.h
+generator.o: $(hdrdir)/ruby/internal/core/rstruct.h
+generator.o: $(hdrdir)/ruby/internal/core/rtypeddata.h
+generator.o: $(hdrdir)/ruby/internal/ctype.h
+generator.o: $(hdrdir)/ruby/internal/dllexport.h
+generator.o: $(hdrdir)/ruby/internal/dosish.h
+generator.o: $(hdrdir)/ruby/internal/encoding/coderange.h
+generator.o: $(hdrdir)/ruby/internal/encoding/ctype.h
+generator.o: $(hdrdir)/ruby/internal/encoding/encoding.h
+generator.o: $(hdrdir)/ruby/internal/encoding/pathname.h
+generator.o: $(hdrdir)/ruby/internal/encoding/re.h
+generator.o: $(hdrdir)/ruby/internal/encoding/sprintf.h
+generator.o: $(hdrdir)/ruby/internal/encoding/string.h
+generator.o: $(hdrdir)/ruby/internal/encoding/symbol.h
+generator.o: $(hdrdir)/ruby/internal/encoding/transcode.h
+generator.o: $(hdrdir)/ruby/internal/error.h
+generator.o: $(hdrdir)/ruby/internal/eval.h
+generator.o: $(hdrdir)/ruby/internal/event.h
+generator.o: $(hdrdir)/ruby/internal/fl_type.h
+generator.o: $(hdrdir)/ruby/internal/gc.h
+generator.o: $(hdrdir)/ruby/internal/glob.h
+generator.o: $(hdrdir)/ruby/internal/globals.h
+generator.o: $(hdrdir)/ruby/internal/has/attribute.h
+generator.o: $(hdrdir)/ruby/internal/has/builtin.h
+generator.o: $(hdrdir)/ruby/internal/has/c_attribute.h
+generator.o: $(hdrdir)/ruby/internal/has/cpp_attribute.h
+generator.o: $(hdrdir)/ruby/internal/has/declspec_attribute.h
+generator.o: $(hdrdir)/ruby/internal/has/extension.h
+generator.o: $(hdrdir)/ruby/internal/has/feature.h
+generator.o: $(hdrdir)/ruby/internal/has/warning.h
+generator.o: $(hdrdir)/ruby/internal/intern/array.h
+generator.o: $(hdrdir)/ruby/internal/intern/bignum.h
+generator.o: $(hdrdir)/ruby/internal/intern/class.h
+generator.o: $(hdrdir)/ruby/internal/intern/compar.h
+generator.o: $(hdrdir)/ruby/internal/intern/complex.h
+generator.o: $(hdrdir)/ruby/internal/intern/cont.h
+generator.o: $(hdrdir)/ruby/internal/intern/dir.h
+generator.o: $(hdrdir)/ruby/internal/intern/enum.h
+generator.o: $(hdrdir)/ruby/internal/intern/enumerator.h
+generator.o: $(hdrdir)/ruby/internal/intern/error.h
+generator.o: $(hdrdir)/ruby/internal/intern/eval.h
+generator.o: $(hdrdir)/ruby/internal/intern/file.h
+generator.o: $(hdrdir)/ruby/internal/intern/hash.h
+generator.o: $(hdrdir)/ruby/internal/intern/io.h
+generator.o: $(hdrdir)/ruby/internal/intern/load.h
+generator.o: $(hdrdir)/ruby/internal/intern/marshal.h
+generator.o: $(hdrdir)/ruby/internal/intern/numeric.h
+generator.o: $(hdrdir)/ruby/internal/intern/object.h
+generator.o: $(hdrdir)/ruby/internal/intern/parse.h
+generator.o: $(hdrdir)/ruby/internal/intern/proc.h
+generator.o: $(hdrdir)/ruby/internal/intern/process.h
+generator.o: $(hdrdir)/ruby/internal/intern/random.h
+generator.o: $(hdrdir)/ruby/internal/intern/range.h
+generator.o: $(hdrdir)/ruby/internal/intern/rational.h
+generator.o: $(hdrdir)/ruby/internal/intern/re.h
+generator.o: $(hdrdir)/ruby/internal/intern/ruby.h
+generator.o: $(hdrdir)/ruby/internal/intern/select.h
+generator.o: $(hdrdir)/ruby/internal/intern/select/largesize.h
+generator.o: $(hdrdir)/ruby/internal/intern/set.h
+generator.o: $(hdrdir)/ruby/internal/intern/signal.h
+generator.o: $(hdrdir)/ruby/internal/intern/sprintf.h
+generator.o: $(hdrdir)/ruby/internal/intern/string.h
+generator.o: $(hdrdir)/ruby/internal/intern/struct.h
+generator.o: $(hdrdir)/ruby/internal/intern/thread.h
+generator.o: $(hdrdir)/ruby/internal/intern/time.h
+generator.o: $(hdrdir)/ruby/internal/intern/variable.h
+generator.o: $(hdrdir)/ruby/internal/intern/vm.h
+generator.o: $(hdrdir)/ruby/internal/interpreter.h
+generator.o: $(hdrdir)/ruby/internal/iterator.h
+generator.o: $(hdrdir)/ruby/internal/memory.h
+generator.o: $(hdrdir)/ruby/internal/method.h
+generator.o: $(hdrdir)/ruby/internal/module.h
+generator.o: $(hdrdir)/ruby/internal/newobj.h
+generator.o: $(hdrdir)/ruby/internal/scan_args.h
+generator.o: $(hdrdir)/ruby/internal/special_consts.h
+generator.o: $(hdrdir)/ruby/internal/static_assert.h
+generator.o: $(hdrdir)/ruby/internal/stdalign.h
+generator.o: $(hdrdir)/ruby/internal/stdbool.h
+generator.o: $(hdrdir)/ruby/internal/stdckdint.h
+generator.o: $(hdrdir)/ruby/internal/symbol.h
+generator.o: $(hdrdir)/ruby/internal/value.h
+generator.o: $(hdrdir)/ruby/internal/value_type.h
+generator.o: $(hdrdir)/ruby/internal/variable.h
+generator.o: $(hdrdir)/ruby/internal/warning_push.h
+generator.o: $(hdrdir)/ruby/internal/xmalloc.h
generator.o: $(hdrdir)/ruby/missing.h
generator.o: $(hdrdir)/ruby/onigmo.h
generator.o: $(hdrdir)/ruby/oniguruma.h
@@ -16,8 +177,10 @@ generator.o: $(hdrdir)/ruby/regex.h
generator.o: $(hdrdir)/ruby/ruby.h
generator.o: $(hdrdir)/ruby/st.h
generator.o: $(hdrdir)/ruby/subst.h
-generator.o: $(top_srcdir)/ext/json/fbuffer/fbuffer.h
-generator.o: $(top_srcdir)/include/ruby.h
+generator.o: $(srcdir)/../fbuffer/fbuffer.h
+generator.o: $(srcdir)/../json.h
+generator.o: $(srcdir)/../simd/simd.h
+generator.o: $(srcdir)/../vendor/fpconv.c
+generator.o: $(srcdir)/../vendor/jeaiii-ltoa.h
generator.o: generator.c
-generator.o: generator.h
# AUTOGENERATED DEPENDENCIES END
diff --git a/ext/json/generator/extconf.rb b/ext/json/generator/extconf.rb
index 8627c5f4bd..33af03ea30 100644
--- a/ext/json/generator/extconf.rb
+++ b/ext/json/generator/extconf.rb
@@ -1,4 +1,19 @@
require 'mkmf'
-$defs << "-DJSON_GENERATOR"
-create_makefile 'json/ext/generator'
+if RUBY_ENGINE == 'truffleruby'
+ # The pure-Ruby generator is faster on TruffleRuby, so skip compiling the generator extension
+ File.write('Makefile', dummy_makefile("").join)
+else
+ append_cflags("-std=c99")
+ have_const("RUBY_TYPED_EMBEDDABLE", "ruby.h") # RUBY_VERSION >= 3.3
+ have_func("ruby_xfree_sized", "ruby.h") # RUBY_VERSION >= 4.1
+
+ $defs << "-DJSON_GENERATOR"
+ $defs << "-DJSON_DEBUG" if ENV.fetch("JSON_DEBUG", "0") != "0"
+
+ if enable_config('generator-use-simd', default=!ENV["JSON_DISABLE_SIMD"])
+ load __dir__ + "/../simd/conf.rb"
+ end
+
+ create_makefile 'json/ext/generator'
+end
diff --git a/ext/json/generator/generator.c b/ext/json/generator/generator.c
index 2bf8074562..110b5f6b32 100644
--- a/ext/json/generator/generator.c
+++ b/ext/json/generator/generator.c
@@ -1,986 +1,1353 @@
+#include "../json.h"
#include "../fbuffer/fbuffer.h"
-#include "generator.h"
+#include "../vendor/fpconv.c"
-#ifdef HAVE_RUBY_ENCODING_H
-static VALUE CEncoding_UTF_8;
-static ID i_encoding, i_encode;
+#include <math.h>
+#include <ctype.h>
+
+#include "../simd/simd.h"
+
+/* ruby api and some helpers */
+
+enum duplicate_key_action {
+ JSON_DEPRECATED = 0,
+ JSON_IGNORE,
+ JSON_RAISE,
+};
+
+typedef struct JSON_Generator_StateStruct {
+ VALUE indent;
+ VALUE space;
+ VALUE space_before;
+ VALUE object_nl;
+ VALUE array_nl;
+ VALUE as_json;
+
+ long max_nesting;
+ long depth;
+ long buffer_initial_length;
+
+ enum duplicate_key_action on_duplicate_key;
+
+ bool as_json_single_arg;
+ bool allow_nan;
+ bool ascii_only;
+ bool script_safe;
+ bool strict;
+} JSON_Generator_State;
+
+static VALUE mJSON, cState, cFragment, eGeneratorError, eNestingError, Encoding_UTF_8;
+
+static ID i_to_s, i_to_json, i_new, i_encode;
+static VALUE sym_indent, sym_space, sym_space_before, sym_object_nl, sym_array_nl, sym_max_nesting, sym_allow_nan, sym_allow_duplicate_key,
+ sym_ascii_only, sym_depth, sym_buffer_initial_length, sym_script_safe, sym_escape_slash, sym_strict, sym_as_json;
+
+
+#define GET_STATE_TO(self, state) \
+ TypedData_Get_Struct(self, JSON_Generator_State, &JSON_Generator_State_type, state)
+
+#define GET_STATE(self) \
+ JSON_Generator_State *state; \
+ GET_STATE_TO(self, state)
+
+struct generate_json_data;
+
+typedef void (*generator_func)(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+
+struct generate_json_data {
+ FBuffer *buffer;
+ VALUE vstate;
+ JSON_Generator_State *state;
+ VALUE obj;
+ generator_func func;
+ 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);
+static void generate_json_object(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_array(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_string(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_null(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_false(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_true(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_fixnum(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_bignum(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_float(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+static void generate_json_fragment(FBuffer *buffer, struct generate_json_data *data, VALUE obj);
+
+static int usascii_encindex, utf8_encindex, binary_encindex;
+
+NORETURN(static void) raise_generator_error_str(VALUE invalid_object, VALUE str)
+{
+ rb_enc_associate_index(str, utf8_encindex);
+ VALUE exc = rb_exc_new_str(eGeneratorError, str);
+ rb_ivar_set(exc, rb_intern("@invalid_object"), invalid_object);
+ rb_exc_raise(exc);
+}
+
+#ifdef RBIMPL_ATTR_FORMAT
+RBIMPL_ATTR_FORMAT(RBIMPL_PRINTF_FORMAT, 2, 3)
#endif
+NORETURN(static void) raise_generator_error(VALUE invalid_object, const char *fmt, ...)
+{
+ va_list args;
+ va_start(args, fmt);
+ VALUE str = rb_vsprintf(fmt, args);
+ va_end(args);
+ raise_generator_error_str(invalid_object, str);
+}
+
+// 0 - single byte char that don't need to be escaped.
+// (x | 8) - char that needs to be escaped.
+static const unsigned char CHAR_LENGTH_MASK = 7;
+static const unsigned char ESCAPE_MASK = 8;
+
+typedef struct _search_state {
+ const char *ptr;
+ const char *end;
+ const char *cursor;
+ FBuffer *buffer;
-static VALUE mJSON, mExt, mGenerator, cState, mGeneratorMethods, mObject,
- mHash, mArray,
-#ifdef RUBY_INTEGER_UNIFICATION
- mInteger,
+#ifdef HAVE_SIMD
+ const char *chunk_base;
+ const char *chunk_end;
+ bool has_matches;
+
+#if defined(HAVE_SIMD_NEON)
+ uint64_t matches_mask;
+#elif defined(HAVE_SIMD_SSE2)
+ int matches_mask;
#else
- mFixnum, mBignum,
-#endif
- mFloat, mString, mString_Extend,
- mTrueClass, mFalseClass, mNilClass, eGeneratorError,
- eNestingError, CRegexp_MULTILINE, CJSON_SAFE_STATE_PROTOTYPE,
- i_SAFE_STATE_PROTOTYPE;
+#error "Unknown SIMD Implementation."
+#endif /* HAVE_SIMD_NEON */
+#endif /* HAVE_SIMD */
+} search_state;
-static ID i_to_s, i_to_json, i_new, i_indent, i_space, i_space_before,
- i_object_nl, i_array_nl, i_max_nesting, i_allow_nan, i_ascii_only,
- i_pack, i_unpack, i_create_id, i_extend, i_key_p,
- i_aref, i_send, i_respond_to_p, i_match, i_keys, i_depth,
- i_buffer_initial_length, i_dup;
+ALWAYS_INLINE(static) void search_flush(search_state *search)
+{
+ // Do not remove this conditional without profiling, specifically escape-heavy text.
+ // escape_UTF8_char_basic will advance search->ptr and search->cursor (effectively a search_flush).
+ // For back-to-back characters that need to be escaped, specifically for the SIMD code paths, this method
+ // will be called just before calling escape_UTF8_char_basic. There will be no characters to append for the
+ // consecutive characters that need to be escaped. While the fbuffer_append is a no-op if
+ // nothing needs to be flushed, we can save a few memory references with this conditional.
+ if (search->ptr > search->cursor) {
+ fbuffer_append(search->buffer, search->cursor, search->ptr - search->cursor);
+ search->cursor = search->ptr;
+ }
+}
-/*
- * Copyright 2001-2004 Unicode, Inc.
+static const unsigned char escape_table_basic[256] = {
+ // ASCII Control Characters
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ // ASCII Characters
+ 0, 0, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // '"'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9, 0, 0, 0, // '\\'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+};
+
+static inline unsigned char search_escape_basic(search_state *search)
+{
+ while (search->ptr < search->end) {
+ if (RB_UNLIKELY(escape_table_basic[(const unsigned char)*search->ptr])) {
+ search_flush(search);
+ return 1;
+ } else {
+ search->ptr++;
+ }
+ }
+ search_flush(search);
+ return 0;
+}
+
+ALWAYS_INLINE(static) void escape_UTF8_char_basic(search_state *search)
+{
+ const unsigned char ch = (unsigned char)*search->ptr;
+ switch (ch) {
+ case '"': fbuffer_append(search->buffer, "\\\"", 2); break;
+ case '\\': fbuffer_append(search->buffer, "\\\\", 2); break;
+ case '/': fbuffer_append(search->buffer, "\\/", 2); break;
+ case '\b': fbuffer_append(search->buffer, "\\b", 2); break;
+ case '\f': fbuffer_append(search->buffer, "\\f", 2); break;
+ case '\n': fbuffer_append(search->buffer, "\\n", 2); break;
+ case '\r': fbuffer_append(search->buffer, "\\r", 2); break;
+ case '\t': fbuffer_append(search->buffer, "\\t", 2); break;
+ default: {
+ const char *hexdig = "0123456789abcdef";
+ char scratch[6] = { '\\', 'u', '0', '0', 0, 0 };
+ scratch[4] = hexdig[(ch >> 4) & 0xf];
+ scratch[5] = hexdig[ch & 0xf];
+ fbuffer_append(search->buffer, scratch, 6);
+ break;
+ }
+ }
+ search->ptr++;
+ search->cursor = search->ptr;
+}
+
+/* Converts in_string to a JSON string (without the wrapping '"'
+ * characters) in FBuffer out_buffer.
+ *
+ * Character are JSON-escaped according to:
*
- * Disclaimer
+ * - Always: ASCII control characters (0x00-0x1F), dquote, and
+ * backslash.
*
- * This source code is provided as is by Unicode, Inc. No claims are
- * made as to fitness for any particular purpose. No warranties of any
- * kind are expressed or implied. The recipient agrees to determine
- * applicability of information provided. If this file has been
- * purchased on magnetic or optical media from Unicode, Inc., the
- * sole remedy for any claim will be exchange of defective media
- * within 90 days of receipt.
+ * - If out_ascii_only: non-ASCII characters (>0x7F)
*
- * Limitations on Rights to Redistribute This Code
+ * - If script_safe: forwardslash (/), line separator (U+2028), and
+ * paragraph separator (U+2029)
*
- * Unicode, Inc. hereby grants the right to freely use the information
- * supplied in this file in the creation of products supporting the
- * Unicode Standard, and to make copies of this file in any form
- * for internal or external distribution as long as this notice
- * remains attached.
+ * Everything else (should be UTF-8) is just passed through and
+ * appended to the result.
*/
-/*
- * Index into the table below with the first byte of a UTF-8 sequence to
- * get the number of trailing bytes that are supposed to follow it.
- * Note that *legal* UTF-8 values can't have 4 or 5-bytes. The table is
- * left as-is for anyone who may want to do such conversion, which was
- * allowed in earlier algorithms.
- */
-static const char trailingBytesForUTF8[256] = {
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
- 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1, 1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,
- 2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2, 3,3,3,3,3,3,3,3,4,4,4,4,5,5,5,5
-};
-/*
- * Magic values subtracted from a buffer value during UTF8 conversion.
- * This table contains as many values as there might be trailing bytes
- * in a UTF-8 sequence.
- */
-static const UTF32 offsetsFromUTF8[6] = { 0x00000000UL, 0x00003080UL, 0x000E2080UL,
- 0x03C82080UL, 0xFA082080UL, 0x82082080UL };
+#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
-/*
- * Utility routine to tell whether a sequence of bytes is legal UTF-8.
- * This must be called with the length pre-determined by the first byte.
- * If not calling this from ConvertUTF8to*, then the length can be set by:
- * length = trailingBytesForUTF8[*source]+1;
- * and the sequence is illegal right away if there aren't that many bytes
- * available.
- * If presented with a length > 4, this returns 0. The Unicode
- * definition of UTF-8 goes up to 4-byte sequences.
- */
-static unsigned char isLegalUTF8(const UTF8 *source, unsigned long length)
-{
- UTF8 a;
- const UTF8 *srcptr = source+length;
- switch (length) {
- default: return 0;
- /* Everything else falls through when "1"... */
- case 4: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return 0;
- case 3: if ((a = (*--srcptr)) < 0x80 || a > 0xBF) return 0;
- case 2: if ((a = (*--srcptr)) > 0xBF) return 0;
-
- switch (*source) {
- /* no fall-through in this inner switch */
- case 0xE0: if (a < 0xA0) return 0; break;
- case 0xED: if (a > 0x9F) return 0; break;
- case 0xF0: if (a < 0x90) return 0; break;
- case 0xF4: if (a > 0x8F) return 0; break;
- default: if (a < 0x80) return 0;
- }
+static inline unsigned char search_escape_basic(search_state *search);
- case 1: if (*source >= 0x80 && *source < 0xC2) return 0;
+static inline void convert_UTF8_to_JSON(search_state *search)
+{
+#ifdef HAVE_SIMD
+#if defined(HAVE_SIMD_NEON)
+ while (search_escape_basic_neon(search)) {
+ escape_UTF8_char_basic(search);
}
- if (*source > 0xF4) return 0;
- return 1;
+#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 */
}
-/* Escapes the UTF16 character and stores the result in the buffer buf. */
-static void unicode_escape(char *buf, UTF16 character)
+static inline void escape_UTF8_char(search_state *search, unsigned char ch_len)
{
- const char *digits = "0123456789abcdef";
-
- buf[2] = digits[character >> 12];
- buf[3] = digits[(character >> 8) & 0xf];
- buf[4] = digits[(character >> 4) & 0xf];
- buf[5] = digits[character & 0xf];
+ const unsigned char ch = (unsigned char)*search->ptr;
+ switch (ch_len) {
+ case 1: {
+ switch (ch) {
+ case '"': fbuffer_append(search->buffer, "\\\"", 2); break;
+ case '\\': fbuffer_append(search->buffer, "\\\\", 2); break;
+ case '/': fbuffer_append(search->buffer, "\\/", 2); break;
+ case '\b': fbuffer_append(search->buffer, "\\b", 2); break;
+ case '\f': fbuffer_append(search->buffer, "\\f", 2); break;
+ case '\n': fbuffer_append(search->buffer, "\\n", 2); break;
+ case '\r': fbuffer_append(search->buffer, "\\r", 2); break;
+ case '\t': fbuffer_append(search->buffer, "\\t", 2); break;
+ default: {
+ const char *hexdig = "0123456789abcdef";
+ char scratch[6] = { '\\', 'u', '0', '0', 0, 0 };
+ scratch[4] = hexdig[(ch >> 4) & 0xf];
+ scratch[5] = hexdig[ch & 0xf];
+ fbuffer_append(search->buffer, scratch, 6);
+ break;
+ }
+ }
+ break;
+ }
+ case 3: {
+ if (search->ptr[2] & 1) {
+ fbuffer_append(search->buffer, "\\u2029", 6);
+ } else {
+ fbuffer_append(search->buffer, "\\u2028", 6);
+ }
+ break;
+ }
+ }
+ search->cursor = (search->ptr += ch_len);
}
-/* Escapes the UTF16 character and stores the result in the buffer buf, then
- * the buffer buf is appended to the FBuffer buffer. */
-static void unicode_escape_to_buffer(FBuffer *buffer, char buf[6], UTF16
- character)
+#ifdef HAVE_SIMD
+
+ALWAYS_INLINE(static) char *copy_remaining_bytes(search_state *search, unsigned long vec_len, unsigned long len)
{
- unicode_escape(buf, character);
- fbuffer_append(buffer, buf, 6);
+ RBIMPL_ASSERT_OR_ASSUME(len < vec_len);
+
+ // Flush the buffer so everything up until the last 'len' characters are unflushed.
+ search_flush(search);
+
+ FBuffer *buf = search->buffer;
+ fbuffer_inc_capa(buf, vec_len);
+
+ char *s = (buf->ptr + buf->len);
+
+ // Pad the buffer with dummy characters that won't need escaping.
+ // 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.
+ 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;
}
-/* Converts string to a JSON string in FBuffer buffer, where all but the ASCII
- * and control characters are JSON escaped. */
-static void convert_UTF8_to_JSON_ASCII(FBuffer *buffer, VALUE string)
+#ifdef HAVE_SIMD_NEON
+
+ALWAYS_INLINE(static) unsigned char neon_next_match(search_state *search)
{
- const UTF8 *source = (UTF8 *) RSTRING_PTR(string);
- const UTF8 *sourceEnd = source + RSTRING_LEN(string);
- char buf[6] = { '\\', 'u' };
+ uint64_t mask = search->matches_mask;
+ uint32_t index = trailing_zeros64(mask) >> 2;
+
+ // It is assumed escape_UTF8_char_basic will only ever increase search->ptr by at most one character.
+ // If we want to use a similar approach for full escaping we'll need to ensure:
+ // search->chunk_base + index >= search->ptr
+ // However, since we know escape_UTF8_char_basic only increases search->ptr by one, if the next match
+ // is one byte after the previous match then:
+ // search->chunk_base + index == search->ptr
+ search->ptr = search->chunk_base + index;
+ mask &= mask - 1;
+ search->matches_mask = mask;
+ search_flush(search);
+ return 1;
+}
- while (source < sourceEnd) {
- UTF32 ch = 0;
- unsigned short extraBytesToRead = trailingBytesForUTF8[*source];
- if (source + extraBytesToRead >= sourceEnd) {
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "partial character in source, but hit end");
- }
- if (!isLegalUTF8(source, extraBytesToRead+1)) {
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "source sequence is illegal/malformed utf-8");
+static inline unsigned char search_escape_basic_neon(search_state *search)
+{
+ if (RB_UNLIKELY(search->has_matches)) {
+ // There are more matches if search->matches_mask > 0.
+ if (search->matches_mask > 0) {
+ return neon_next_match(search);
+ } else {
+ // neon_next_match will only advance search->ptr up to the last matching character.
+ // Skip over any characters in the last chunk that occur after the last match.
+ search->has_matches = false;
+ search->ptr = search->chunk_end;
}
- /*
- * The cases all fall through. See "Note A" below.
- */
- switch (extraBytesToRead) {
- case 5: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
- case 4: ch += *source++; ch <<= 6; /* remember, illegal UTF-8 */
- case 3: ch += *source++; ch <<= 6;
- case 2: ch += *source++; ch <<= 6;
- case 1: ch += *source++; ch <<= 6;
- case 0: ch += *source++;
+ }
+
+ /*
+ * The code below implements an SIMD-based algorithm to determine if N bytes at a time
+ * need to be escaped.
+ *
+ * Assume the ptr = "Te\sting!" (the double quotes are included in the string)
+ *
+ * The explanation will be limited to the first 8 bytes of the string for simplicity. However
+ * the vector insructions may work on larger vectors.
+ *
+ * First, we load three constants 'lower_bound', 'backslash' and 'dblquote" in vector registers.
+ *
+ * lower_bound: [20 20 20 20 20 20 20 20]
+ * backslash: [5C 5C 5C 5C 5C 5C 5C 5C]
+ * dblquote: [22 22 22 22 22 22 22 22]
+ *
+ * Next we load the first chunk of the ptr:
+ * [22 54 65 5C 73 74 69 6E] (" T e \ s t i n)
+ *
+ * First we check if any byte in chunk is less than 32 (0x20). This returns the following vector
+ * as no bytes are less than 32 (0x20):
+ * [0 0 0 0 0 0 0 0]
+ *
+ * Next, we check if any byte in chunk is equal to a backslash:
+ * [0 0 0 FF 0 0 0 0]
+ *
+ * Finally we check if any byte in chunk is equal to a double quote:
+ * [FF 0 0 0 0 0 0 0]
+ *
+ * Now we have three vectors where each byte indicates if the corresponding byte in chunk
+ * needs to be escaped. We combine these vectors with a series of logical OR instructions.
+ * This is the needs_escape vector and it is equal to:
+ * [FF 0 0 FF 0 0 0 0]
+ *
+ * Next we compute the bitwise AND between each byte and 0x1 and compute the horizontal sum of
+ * the values in the vector. This computes how many bytes need to be escaped within this chunk.
+ *
+ * Finally we compute a mask that indicates which bytes need to be escaped. If the mask is 0 then,
+ * no bytes need to be escaped and we can continue to the next chunk. If the mask is not 0 then we
+ * have at least one byte that needs to be escaped.
+ */
+
+ if (string_scan_simd_neon(&search->ptr, search->end, &search->matches_mask)) {
+ search->has_matches = true;
+ search->chunk_base = search->ptr;
+ search->chunk_end = search->ptr + sizeof(uint8x16_t);
+ return neon_next_match(search);
+ }
+
+ // There are fewer than 16 bytes left.
+ unsigned long remaining = (search->end - search->ptr);
+ if (remaining >= SIMD_MINIMUM_THRESHOLD) {
+ char *s = copy_remaining_bytes(search, sizeof(uint8x16_t), remaining);
+
+ uint64_t mask = compute_chunk_mask_neon(s);
+
+ if (!mask) {
+ // Nothing to escape, ensure search_flush doesn't do anything by setting
+ // search->cursor to search->ptr.
+ fbuffer_consumed(search->buffer, remaining);
+ search->ptr = search->end;
+ search->cursor = search->end;
+ return 0;
}
- ch -= offsetsFromUTF8[extraBytesToRead];
-
- if (ch <= UNI_MAX_BMP) { /* Target is a character <= 0xFFFF */
- /* UTF-16 surrogate values are illegal in UTF-32 */
- if (ch >= UNI_SUR_HIGH_START && ch <= UNI_SUR_LOW_END) {
-#if UNI_STRICT_CONVERSION
- source -= (extraBytesToRead+1); /* return to the illegal value itself */
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "source sequence is illegal/malformed utf-8");
+
+ search->matches_mask = mask;
+ search->has_matches = true;
+ search->chunk_end = search->end;
+ search->chunk_base = search->ptr;
+ return neon_next_match(search);
+ }
+
+ if (search->ptr < search->end) {
+ return search_escape_basic(search);
+ }
+
+ search_flush(search);
+ return 0;
+}
+#endif /* HAVE_SIMD_NEON */
+
+#ifdef HAVE_SIMD_SSE2
+
+ALWAYS_INLINE(static) unsigned char sse2_next_match(search_state *search)
+{
+ int mask = search->matches_mask;
+ int index = trailing_zeros(mask);
+
+ // It is assumed escape_UTF8_char_basic will only ever increase search->ptr by at most one character.
+ // If we want to use a similar approach for full escaping we'll need to ensure:
+ // search->chunk_base + index >= search->ptr
+ // However, since we know escape_UTF8_char_basic only increases search->ptr by one, if the next match
+ // is one byte after the previous match then:
+ // search->chunk_base + index == search->ptr
+ search->ptr = search->chunk_base + index;
+ mask &= mask - 1;
+ search->matches_mask = mask;
+ search_flush(search);
+ return 1;
+}
+
+#if defined(__clang__) || defined(__GNUC__)
+#define TARGET_SSE2 __attribute__((target("sse2")))
#else
- unicode_escape_to_buffer(buffer, buf, UNI_REPLACEMENT_CHAR);
+#define TARGET_SSE2
#endif
+
+ALWAYS_INLINE(static) TARGET_SSE2 unsigned char search_escape_basic_sse2(search_state *search)
+{
+ if (RB_UNLIKELY(search->has_matches)) {
+ // There are more matches if search->matches_mask > 0.
+ if (search->matches_mask > 0) {
+ return sse2_next_match(search);
+ } else {
+ // sse2_next_match will only advance search->ptr up to the last matching character.
+ // Skip over any characters in the last chunk that occur after the last match.
+ search->has_matches = false;
+ if (RB_UNLIKELY(search->chunk_base + sizeof(__m128i) >= search->end)) {
+ search->ptr = search->end;
} else {
- /* normal case */
- if (ch >= 0x20 && ch <= 0x7f) {
- switch (ch) {
- case '\\':
- fbuffer_append(buffer, "\\\\", 2);
- break;
- case '"':
- fbuffer_append(buffer, "\\\"", 2);
- break;
- default:
- fbuffer_append_char(buffer, (char)ch);
- break;
- }
- } else {
- switch (ch) {
- case '\n':
- fbuffer_append(buffer, "\\n", 2);
- break;
- case '\r':
- fbuffer_append(buffer, "\\r", 2);
- break;
- case '\t':
- fbuffer_append(buffer, "\\t", 2);
- break;
- case '\f':
- fbuffer_append(buffer, "\\f", 2);
- break;
- case '\b':
- fbuffer_append(buffer, "\\b", 2);
- break;
- default:
- unicode_escape_to_buffer(buffer, buf, (UTF16) ch);
- break;
+ search->ptr = search->chunk_base + sizeof(__m128i);
+ }
+ }
+ }
+
+ if (string_scan_simd_sse2(&search->ptr, search->end, &search->matches_mask)) {
+ search->has_matches = true;
+ search->chunk_base = search->ptr;
+ search->chunk_end = search->ptr + sizeof(__m128i);
+ return sse2_next_match(search);
+ }
+
+ // There are fewer than 16 bytes left.
+ unsigned long remaining = (search->end - search->ptr);
+ if (remaining >= SIMD_MINIMUM_THRESHOLD) {
+ char *s = copy_remaining_bytes(search, sizeof(__m128i), remaining);
+
+ int needs_escape_mask = compute_chunk_mask_sse2(s);
+
+ if (needs_escape_mask == 0) {
+ // Nothing to escape, ensure search_flush doesn't do anything by setting
+ // search->cursor to search->ptr.
+ fbuffer_consumed(search->buffer, remaining);
+ search->ptr = search->end;
+ search->cursor = search->end;
+ return 0;
+ }
+
+ search->has_matches = true;
+ search->matches_mask = needs_escape_mask;
+ search->chunk_base = search->ptr;
+ return sse2_next_match(search);
+ }
+
+ if (search->ptr < search->end) {
+ return search_escape_basic(search);
+ }
+
+ search_flush(search);
+ return 0;
+}
+
+#endif /* HAVE_SIMD_SSE2 */
+
+#endif /* HAVE_SIMD */
+
+static const unsigned char script_safe_escape_table[256] = {
+ // ASCII Control Characters
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ // ASCII Characters
+ 0, 0, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9, // '"' and '/'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9, 0, 0, 0, // '\\'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ // Continuation byte
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ // First byte of a 2-byte code point
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+ // First byte of a 3-byte code point
+ 3, 3,11, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, // 0xE2 is the start of \u2028 and \u2029
+ //First byte of a 4+ byte code point
+ 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 9, 9,
+};
+
+static inline unsigned char search_script_safe_escape(search_state *search)
+{
+ while (search->ptr < search->end) {
+ unsigned char ch = (unsigned char)*search->ptr;
+ unsigned char ch_len = script_safe_escape_table[ch];
+
+ if (RB_UNLIKELY(ch_len)) {
+ if (ch_len & ESCAPE_MASK) {
+ if (RB_UNLIKELY(ch_len == 11)) {
+ const unsigned char *uptr = (const unsigned char *)search->ptr;
+ if (!(uptr[1] == 0x80 && (uptr[2] >> 1) == 0x54)) {
+ search->ptr += 3;
+ continue;
}
}
+ search_flush(search);
+ return ch_len & CHAR_LENGTH_MASK;
+ } else {
+ search->ptr += ch_len;
}
- } else if (ch > UNI_MAX_UTF16) {
-#if UNI_STRICT_CONVERSION
- source -= (extraBytesToRead+1); /* return to the start */
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "source sequence is illegal/malformed utf8");
-#else
- unicode_escape_to_buffer(buffer, buf, UNI_REPLACEMENT_CHAR);
-#endif
} else {
- /* target is a character in range 0xFFFF - 0x10FFFF. */
- ch -= halfBase;
- unicode_escape_to_buffer(buffer, buf, (UTF16)((ch >> halfShift) + UNI_SUR_HIGH_START));
- unicode_escape_to_buffer(buffer, buf, (UTF16)((ch & halfMask) + UNI_SUR_LOW_START));
+ search->ptr++;
}
}
- RB_GC_GUARD(string);
-}
-
-/* Converts string to a JSON string in FBuffer buffer, where only the
- * characters required by the JSON standard are JSON escaped. The remaining
- * characters (should be UTF8) are just passed through and appended to the
- * result. */
-static void convert_UTF8_to_JSON(FBuffer *buffer, VALUE string)
-{
- const char *ptr = RSTRING_PTR(string), *p;
- unsigned long len = RSTRING_LEN(string), start = 0, end = 0;
- const char *escape = NULL;
- int escape_len;
- unsigned char c;
- char buf[6] = { '\\', 'u' };
-
- for (start = 0, end = 0; end < len;) {
- p = ptr + end;
- c = (unsigned char) *p;
- if (c < 0x20) {
- switch (c) {
- case '\n':
- escape = "\\n";
- escape_len = 2;
- break;
- case '\r':
- escape = "\\r";
- escape_len = 2;
- break;
- case '\t':
- escape = "\\t";
- escape_len = 2;
- break;
- case '\f':
- escape = "\\f";
- escape_len = 2;
- break;
- case '\b':
- escape = "\\b";
- escape_len = 2;
- break;
- default:
- unicode_escape(buf, (UTF16) *p);
- escape = buf;
- escape_len = 6;
+ search_flush(search);
+ return 0;
+}
+
+static void convert_UTF8_to_script_safe_JSON(search_state *search)
+{
+ unsigned char ch_len;
+ while ((ch_len = search_script_safe_escape(search))) {
+ escape_UTF8_char(search, ch_len);
+ }
+}
+
+static const unsigned char ascii_only_escape_table[256] = {
+ // ASCII Control Characters
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+ // ASCII Characters
+ 0, 0, 9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, // '"'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 9, 0, 0, 0, // '\\'
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
+ // Continuation byte
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ // First byte of a 2-byte code point
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+ // First byte of a 3-byte code point
+ 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
+ //First byte of a 4+ byte code point
+ 4, 4, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 9, 9,
+};
+
+static inline unsigned char search_ascii_only_escape(search_state *search, const unsigned char escape_table[256])
+{
+ while (search->ptr < search->end) {
+ unsigned char ch = (unsigned char)*search->ptr;
+ unsigned char ch_len = escape_table[ch];
+
+ if (RB_UNLIKELY(ch_len)) {
+ search_flush(search);
+ return ch_len & CHAR_LENGTH_MASK;
+ } else {
+ search->ptr++;
+ }
+ }
+ search_flush(search);
+ return 0;
+}
+
+static inline void full_escape_UTF8_char(search_state *search, unsigned char ch_len)
+{
+ const unsigned char ch = (unsigned char)*search->ptr;
+ switch (ch_len) {
+ case 1: {
+ switch (ch) {
+ case '"': fbuffer_append(search->buffer, "\\\"", 2); break;
+ case '\\': fbuffer_append(search->buffer, "\\\\", 2); break;
+ case '/': fbuffer_append(search->buffer, "\\/", 2); break;
+ case '\b': fbuffer_append(search->buffer, "\\b", 2); break;
+ case '\f': fbuffer_append(search->buffer, "\\f", 2); break;
+ case '\n': fbuffer_append(search->buffer, "\\n", 2); break;
+ case '\r': fbuffer_append(search->buffer, "\\r", 2); break;
+ case '\t': fbuffer_append(search->buffer, "\\t", 2); break;
+ default: {
+ const char *hexdig = "0123456789abcdef";
+ char scratch[6] = { '\\', 'u', '0', '0', 0, 0 };
+ scratch[4] = hexdig[(ch >> 4) & 0xf];
+ scratch[5] = hexdig[ch & 0xf];
+ fbuffer_append(search->buffer, scratch, 6);
break;
+ }
}
- } else {
- switch (c) {
- case '\\':
- escape = "\\\\";
- escape_len = 2;
+ break;
+ }
+ default: {
+ const char *hexdig = "0123456789abcdef";
+ char scratch[12] = { '\\', 'u', 0, 0, 0, 0, '\\', 'u' };
+
+ uint32_t wchar = 0;
+
+ switch (ch_len) {
+ case 2:
+ wchar = ch & 0x1F;
break;
- case '"':
- escape = "\\\"";
- escape_len = 2;
+ case 3:
+ wchar = ch & 0x0F;
break;
- default:
- {
- unsigned short clen = trailingBytesForUTF8[c] + 1;
- if (end + clen > len) {
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "partial character in source, but hit end");
- }
- if (!isLegalUTF8((UTF8 *) p, clen)) {
- rb_raise(rb_path2class("JSON::GeneratorError"),
- "source sequence is illegal/malformed utf-8");
- }
- end += clen;
- }
- continue;
+ case 4:
+ wchar = ch & 0x07;
break;
}
+
+ for (short i = 1; i < ch_len; i++) {
+ wchar = (wchar << 6) | (search->ptr[i] & 0x3F);
+ }
+
+ if (wchar <= 0xFFFF) {
+ scratch[2] = hexdig[wchar >> 12];
+ scratch[3] = hexdig[(wchar >> 8) & 0xf];
+ scratch[4] = hexdig[(wchar >> 4) & 0xf];
+ scratch[5] = hexdig[wchar & 0xf];
+ fbuffer_append(search->buffer, scratch, 6);
+ } else {
+ uint16_t hi, lo;
+ wchar -= 0x10000;
+ hi = 0xD800 + (uint16_t)(wchar >> 10);
+ lo = 0xDC00 + (uint16_t)(wchar & 0x3FF);
+
+ scratch[2] = hexdig[hi >> 12];
+ scratch[3] = hexdig[(hi >> 8) & 0xf];
+ scratch[4] = hexdig[(hi >> 4) & 0xf];
+ scratch[5] = hexdig[hi & 0xf];
+
+ scratch[8] = hexdig[lo >> 12];
+ scratch[9] = hexdig[(lo >> 8) & 0xf];
+ scratch[10] = hexdig[(lo >> 4) & 0xf];
+ scratch[11] = hexdig[lo & 0xf];
+
+ fbuffer_append(search->buffer, scratch, 12);
+ }
+
+ break;
}
- fbuffer_append(buffer, ptr + start, end - start);
- fbuffer_append(buffer, escape, escape_len);
- start = ++end;
- escape = NULL;
}
- fbuffer_append(buffer, ptr + start, end - start);
+ search->cursor = (search->ptr += ch_len);
}
-static char *fstrndup(const char *ptr, unsigned long len) {
- char *result;
- if (len <= 0) return NULL;
- result = ALLOC_N(char, len);
- memcpy(result, ptr, len);
- return result;
+static void convert_UTF8_to_ASCII_only_JSON(search_state *search, const unsigned char escape_table[256])
+{
+ unsigned char ch_len;
+ while ((ch_len = search_ascii_only_escape(search, escape_table))) {
+ full_escape_UTF8_char(search, ch_len);
+ }
}
-/*
- * Document-module: JSON::Ext::Generator
- *
- * This is the JSON generator implemented as a C extension. It can be
- * configured to be used by setting
- *
- * JSON.generator = JSON::Ext::Generator
- *
- * with the method generator= in JSON.
- *
- */
-
-/*
- * call-seq: to_json(state = nil)
- *
- * Returns a JSON string containing a JSON object, that is generated from
- * this Hash instance.
- * _state_ is a JSON::State object, that can also be used to configure the
- * produced JSON string output further.
- */
-static VALUE mHash_to_json(int argc, VALUE *argv, VALUE self)
+static void State_mark(void *ptr)
{
- GENERATE_JSON(object);
+ JSON_Generator_State *state = ptr;
+ rb_gc_mark_movable(state->indent);
+ rb_gc_mark_movable(state->space);
+ rb_gc_mark_movable(state->space_before);
+ rb_gc_mark_movable(state->object_nl);
+ rb_gc_mark_movable(state->array_nl);
+ rb_gc_mark_movable(state->as_json);
}
-/*
- * call-seq: to_json(state = nil)
- *
- * Returns a JSON string containing a JSON array, that is generated from
- * this Array instance.
- * _state_ is a JSON::State object, that can also be used to configure the
- * produced JSON string output further.
- */
-static VALUE mArray_to_json(int argc, VALUE *argv, VALUE self) {
- GENERATE_JSON(array);
+static void State_compact(void *ptr)
+{
+ JSON_Generator_State *state = ptr;
+ state->indent = rb_gc_location(state->indent);
+ state->space = rb_gc_location(state->space);
+ state->space_before = rb_gc_location(state->space_before);
+ state->object_nl = rb_gc_location(state->object_nl);
+ state->array_nl = rb_gc_location(state->array_nl);
+ state->as_json = rb_gc_location(state->as_json);
}
-#ifdef RUBY_INTEGER_UNIFICATION
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string representation for this Integer number.
- */
-static VALUE mInteger_to_json(int argc, VALUE *argv, VALUE self)
+static size_t State_memsize(const void *ptr)
{
- GENERATE_JSON(integer);
+ return sizeof(JSON_Generator_State);
}
-#else
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string representation for this Integer number.
- */
-static VALUE mFixnum_to_json(int argc, VALUE *argv, VALUE self)
+static const rb_data_type_t JSON_Generator_State_type = {
+ .wrap_struct_name = "JSON/Generator/State",
+ .function = {
+ .dmark = State_mark,
+ .dfree = RUBY_DEFAULT_FREE,
+ .dsize = State_memsize,
+ .dcompact = State_compact,
+ },
+ .flags = RUBY_TYPED_WB_PROTECTED | RUBY_TYPED_FREE_IMMEDIATELY | RUBY_TYPED_FROZEN_SHAREABLE | RUBY_TYPED_EMBEDDABLE,
+};
+
+static void state_init(JSON_Generator_State *state)
{
- GENERATE_JSON(fixnum);
+ state->max_nesting = 100;
+ state->buffer_initial_length = FBUFFER_INITIAL_LENGTH_DEFAULT;
}
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string representation for this Integer number.
- */
-static VALUE mBignum_to_json(int argc, VALUE *argv, VALUE self)
+static VALUE cState_s_allocate(VALUE klass)
{
- GENERATE_JSON(bignum);
+ JSON_Generator_State *state;
+ VALUE obj = TypedData_Make_Struct(klass, JSON_Generator_State, &JSON_Generator_State_type, state);
+ state_init(state);
+ return obj;
}
-#endif
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string representation for this Float number.
- */
-static VALUE mFloat_to_json(int argc, VALUE *argv, VALUE self)
+static void vstate_spill(struct generate_json_data *data)
+{
+ VALUE vstate = cState_s_allocate(cState);
+ GET_STATE(vstate);
+ MEMCPY(state, data->state, JSON_Generator_State, 1);
+ data->state = state;
+ data->vstate = vstate;
+ RB_OBJ_WRITTEN(vstate, Qundef, state->indent);
+ RB_OBJ_WRITTEN(vstate, Qundef, state->space);
+ RB_OBJ_WRITTEN(vstate, Qundef, state->space_before);
+ RB_OBJ_WRITTEN(vstate, Qundef, state->object_nl);
+ RB_OBJ_WRITTEN(vstate, Qundef, state->array_nl);
+ RB_OBJ_WRITTEN(vstate, Qundef, state->as_json);
+}
+
+static inline VALUE json_call_to_json(struct generate_json_data *data, VALUE obj)
{
- GENERATE_JSON(float);
+ if (RB_UNLIKELY(!data->vstate)) {
+ vstate_spill(data);
+ }
+ GET_STATE(data->vstate);
+ state->depth = data->depth;
+ VALUE tmp = rb_funcall(obj, i_to_json, 1, data->vstate);
+ // no need to restore state->depth, vstate is just a temporary State
+ return tmp;
}
-/*
- * call-seq: String.included(modul)
- *
- * Extends _modul_ with the String::Extend module.
- */
-static VALUE mString_included_s(VALUE self, VALUE modul) {
- VALUE result = rb_funcall(modul, i_extend, 1, mString_Extend);
- return result;
+static VALUE
+json_call_as_json(JSON_Generator_State *state, VALUE object, VALUE is_key)
+{
+ VALUE proc_args[2] = {object, is_key};
+ return rb_proc_call_with_block(state->as_json, 2, proc_args, Qnil);
}
-/*
- * call-seq: to_json(*)
- *
- * This string should be encoded with UTF-8 A call to this method
- * returns a JSON string encoded with UTF16 big endian characters as
- * \u????.
- */
-static VALUE mString_to_json(int argc, VALUE *argv, VALUE self)
+static VALUE
+convert_string_subclass(VALUE key)
{
- GENERATE_JSON(string);
+ VALUE key_to_s = rb_funcall(key, i_to_s, 0);
+
+ if (RB_UNLIKELY(!RB_TYPE_P(key_to_s, T_STRING))) {
+ VALUE cname = rb_obj_class(key);
+ rb_raise(rb_eTypeError,
+ "can't convert %"PRIsVALUE" to %s (%"PRIsVALUE"#%s gives %"PRIsVALUE")",
+ cname, "String", cname, "to_s", rb_obj_class(key_to_s));
+ }
+
+ return key_to_s;
}
-/*
- * call-seq: to_json_raw_object()
- *
- * This method creates a raw object hash, that can be nested into
- * other data structures and will be generated as a raw string. This
- * method should be used, if you want to convert raw strings to JSON
- * instead of UTF-8 strings, e. g. binary data.
- */
-static VALUE mString_to_json_raw_object(VALUE self)
+static bool enc_utf8_compatible_p(int enc_idx)
{
- VALUE ary;
- 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);
- return result;
+ if (enc_idx == usascii_encindex) return true;
+ if (enc_idx == utf8_encindex) return true;
+ return false;
}
-/*
- * call-seq: to_json_raw(*args)
- *
- * This method creates a JSON text from the result of a call to
- * to_json_raw_object of this String.
- */
-static VALUE mString_to_json_raw(int argc, VALUE *argv, VALUE self)
+static VALUE encode_json_string_try(VALUE str)
{
- VALUE obj = mString_to_json_raw_object(self);
- Check_Type(obj, T_HASH);
- return mHash_to_json(argc, argv, obj);
+ return rb_funcall(str, i_encode, 1, Encoding_UTF_8);
}
-/*
- * call-seq: json_create(o)
- *
- * Raw Strings are JSON Objects (the raw bytes are stored in an array for the
- * key "raw"). The Ruby String can be created by this module method.
- */
-static VALUE mString_Extend_json_create(VALUE self, VALUE o)
+static VALUE encode_json_string_rescue(VALUE str, VALUE exception)
{
- VALUE ary;
- Check_Type(o, T_HASH);
- ary = rb_hash_aref(o, rb_str_new2("raw"));
- return rb_funcall(ary, i_pack, 1, rb_str_new2("C*"));
+ raise_generator_error_str(str, rb_funcall(exception, rb_intern("message"), 0));
+ return Qundef;
}
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string for true: 'true'.
- */
-static VALUE mTrueClass_to_json(int argc, VALUE *argv, VALUE self)
-{
- GENERATE_JSON(true);
+static inline int json_str_coderange(VALUE str) {
+ int coderange = RB_ENC_CODERANGE(str);
+ if (coderange == RUBY_ENC_CODERANGE_UNKNOWN) {
+ coderange = rb_enc_str_coderange(str);
+ }
+ return coderange;
}
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string for false: 'false'.
- */
-static VALUE mFalseClass_to_json(int argc, VALUE *argv, VALUE self)
+static inline bool valid_json_string_p(VALUE str)
{
- GENERATE_JSON(false);
+ int coderange = json_str_coderange(str);
+
+ if (RB_LIKELY(coderange == ENC_CODERANGE_7BIT)) {
+ return true;
+ }
+
+ if (RB_LIKELY(coderange == ENC_CODERANGE_VALID)) {
+ return enc_utf8_compatible_p(RB_ENCODING_GET_INLINED(str));
+ }
+
+ return false;
}
-/*
- * call-seq: to_json(*)
- *
- * Returns a JSON string for nil: 'null'.
- */
-static VALUE mNilClass_to_json(int argc, VALUE *argv, VALUE self)
+NOINLINE(static) VALUE convert_invalid_encoding(struct generate_json_data *data, VALUE str, bool as_json_called, bool is_key)
{
- GENERATE_JSON(null);
+ if (!as_json_called && data->state->strict && RTEST(data->state->as_json)) {
+ VALUE coerced_str = json_call_as_json(data->state, str, Qfalse);
+ if (coerced_str != str) {
+ if (RB_TYPE_P(coerced_str, T_STRING)) {
+ if (!valid_json_string_p(coerced_str)) {
+ raise_generator_error(str, "source sequence is illegal/malformed utf-8");
+ }
+ } else {
+ // as_json could return another type than T_STRING
+ if (is_key) {
+ raise_generator_error(coerced_str, "%"PRIsVALUE" not allowed as object key in JSON", CLASS_OF(coerced_str));
+ }
+ }
+
+ return coerced_str;
+ }
+ }
+
+ if (RB_ENCODING_GET_INLINED(str) == binary_encindex) {
+ VALUE 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;
+ }
+ }
+
+ return rb_rescue(encode_json_string_try, str, encode_json_string_rescue, str);
}
-/*
- * call-seq: to_json(*)
- *
- * Converts this object to a string (calling #to_s), converts
- * it to a JSON string, and returns the result. This is a fallback, if no
- * special method #to_json was defined for some object.
- */
-static VALUE mObject_to_json(int argc, VALUE *argv, VALUE self)
+ALWAYS_INLINE(static) VALUE ensure_valid_encoding(struct generate_json_data *data, VALUE str, bool as_json_called, bool is_key)
{
- VALUE state;
- VALUE string = rb_funcall(self, i_to_s, 0);
- rb_scan_args(argc, argv, "01", &state);
- Check_Type(string, T_STRING);
- state = cState_from_state_s(cState, state);
- return cState_partial_generate(state, string);
+ if (RB_LIKELY(valid_json_string_p(str))) {
+ return str;
+ }
+ else {
+ return convert_invalid_encoding(data, str, as_json_called, is_key);
+ }
}
-static void State_free(void *ptr)
+static void raw_generate_json_string(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- JSON_Generator_State *state = ptr;
- if (state->indent) ruby_xfree(state->indent);
- if (state->space) ruby_xfree(state->space);
- if (state->space_before) ruby_xfree(state->space_before);
- if (state->object_nl) ruby_xfree(state->object_nl);
- if (state->array_nl) ruby_xfree(state->array_nl);
- if (state->array_delim) fbuffer_free(state->array_delim);
- if (state->object_delim) fbuffer_free(state->object_delim);
- if (state->object_delim2) fbuffer_free(state->object_delim2);
- ruby_xfree(state);
+ fbuffer_append_char(buffer, '"');
+
+ long len;
+ search_state search;
+ search.buffer = buffer;
+ RSTRING_GETMEM(obj, search.ptr, len);
+ search.cursor = search.ptr;
+ search.end = search.ptr + len;
+
+#ifdef HAVE_SIMD
+ search.matches_mask = 0;
+ search.has_matches = false;
+ search.chunk_base = NULL;
+ search.chunk_end = NULL;
+#endif /* HAVE_SIMD */
+
+ switch (json_str_coderange(obj)) {
+ case ENC_CODERANGE_7BIT:
+ case ENC_CODERANGE_VALID:
+ if (RB_UNLIKELY(data->state->ascii_only)) {
+ convert_UTF8_to_ASCII_only_JSON(&search, data->state->script_safe ? script_safe_escape_table : ascii_only_escape_table);
+ } else if (RB_UNLIKELY(data->state->script_safe)) {
+ convert_UTF8_to_script_safe_JSON(&search);
+ } else {
+ convert_UTF8_to_JSON(&search);
+ }
+ break;
+ default:
+ raise_generator_error(obj, "source sequence is illegal/malformed utf-8");
+ break;
+ }
+ fbuffer_append_char(buffer, '"');
}
-static size_t State_memsize(const void *ptr)
+static void generate_json_string(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- const JSON_Generator_State *state = ptr;
- size_t size = sizeof(*state);
- if (state->indent) size += state->indent_len + 1;
- if (state->space) size += state->space_len + 1;
- if (state->space_before) size += state->space_before_len + 1;
- if (state->object_nl) size += state->object_nl_len + 1;
- if (state->array_nl) size += state->array_nl_len + 1;
- if (state->array_delim) size += FBUFFER_CAPA(state->array_delim);
- if (state->object_delim) size += FBUFFER_CAPA(state->object_delim);
- if (state->object_delim2) size += FBUFFER_CAPA(state->object_delim2);
- return size;
+ obj = ensure_valid_encoding(data, obj, false, false);
+ raw_generate_json_string(buffer, data, obj);
}
-#ifdef NEW_TYPEDDATA_WRAPPER
-static const rb_data_type_t JSON_Generator_State_type = {
- "JSON/Generator/State",
- {NULL, State_free, State_memsize,},
-#ifdef RUBY_TYPED_FREE_IMMEDIATELY
- 0, 0,
- RUBY_TYPED_FREE_IMMEDIATELY,
-#endif
+struct hash_foreach_arg {
+ VALUE hash;
+ struct generate_json_data *data;
+ int first_key_type;
+ bool first;
+ bool mixed_keys_encountered;
};
-#endif
-static VALUE cState_s_allocate(VALUE klass)
+NOINLINE(static) void
+json_inspect_hash_with_mixed_keys(struct hash_foreach_arg *arg)
{
- JSON_Generator_State *state;
- return TypedData_Make_Struct(klass, JSON_Generator_State,
- &JSON_Generator_State_type, state);
+ if (arg->mixed_keys_encountered) {
+ return;
+ }
+ arg->mixed_keys_encountered = true;
+
+ JSON_Generator_State *state = arg->data->state;
+ if (state->on_duplicate_key != JSON_IGNORE) {
+ VALUE do_raise = state->on_duplicate_key == JSON_RAISE ? Qtrue : Qfalse;
+ rb_funcall(mJSON, rb_intern("on_mixed_keys_hash"), 2, arg->hash, do_raise);
+ }
}
-/*
- * call-seq: configure(opts)
- *
- * Configure this State instance with the Hash _opts_, and return
- * itself.
- */
-static VALUE cState_configure(VALUE self, VALUE opts)
+static int
+json_object_i(VALUE key, VALUE val, VALUE _arg)
{
- VALUE tmp;
- GET_STATE(self);
- tmp = rb_check_convert_type(opts, T_HASH, "Hash", "to_hash");
- if (NIL_P(tmp)) tmp = rb_convert_type(opts, T_HASH, "Hash", "to_h");
- opts = tmp;
- tmp = rb_hash_aref(opts, ID2SYM(i_indent));
- if (RTEST(tmp)) {
- unsigned long len;
- Check_Type(tmp, T_STRING);
- len = RSTRING_LEN(tmp);
- state->indent = fstrndup(RSTRING_PTR(tmp), len + 1);
- state->indent_len = len;
- }
- tmp = rb_hash_aref(opts, ID2SYM(i_space));
- if (RTEST(tmp)) {
- unsigned long len;
- Check_Type(tmp, T_STRING);
- len = RSTRING_LEN(tmp);
- state->space = fstrndup(RSTRING_PTR(tmp), len + 1);
- state->space_len = len;
- }
- tmp = rb_hash_aref(opts, ID2SYM(i_space_before));
- if (RTEST(tmp)) {
- unsigned long len;
- Check_Type(tmp, T_STRING);
- len = RSTRING_LEN(tmp);
- state->space_before = fstrndup(RSTRING_PTR(tmp), len + 1);
- state->space_before_len = len;
- }
- tmp = rb_hash_aref(opts, ID2SYM(i_array_nl));
- if (RTEST(tmp)) {
- unsigned long len;
- Check_Type(tmp, T_STRING);
- len = RSTRING_LEN(tmp);
- state->array_nl = fstrndup(RSTRING_PTR(tmp), len + 1);
- state->array_nl_len = len;
+ struct hash_foreach_arg *arg = (struct hash_foreach_arg *)_arg;
+ struct generate_json_data *data = arg->data;
+
+ FBuffer *buffer = data->buffer;
+ JSON_Generator_State *state = data->state;
+
+ long depth = data->depth;
+ int key_type = rb_type(key);
+
+ if (arg->first) {
+ arg->first = false;
+ arg->first_key_type = key_type;
}
- tmp = rb_hash_aref(opts, ID2SYM(i_object_nl));
- if (RTEST(tmp)) {
- unsigned long len;
- Check_Type(tmp, T_STRING);
- len = RSTRING_LEN(tmp);
- state->object_nl = fstrndup(RSTRING_PTR(tmp), len + 1);
- state->object_nl_len = len;
+ else {
+ fbuffer_append_char(buffer, ',');
}
- tmp = ID2SYM(i_max_nesting);
- state->max_nesting = 100;
- if (option_given_p(opts, tmp)) {
- VALUE max_nesting = rb_hash_aref(opts, tmp);
- if (RTEST(max_nesting)) {
- Check_Type(max_nesting, T_FIXNUM);
- state->max_nesting = FIX2LONG(max_nesting);
- } else {
- state->max_nesting = 0;
- }
+
+ if (RB_UNLIKELY(data->state->object_nl)) {
+ fbuffer_append_str(buffer, data->state->object_nl);
}
- tmp = ID2SYM(i_depth);
- state->depth = 0;
- if (option_given_p(opts, tmp)) {
- VALUE depth = rb_hash_aref(opts, tmp);
- if (RTEST(depth)) {
- Check_Type(depth, T_FIXNUM);
- state->depth = FIX2LONG(depth);
- } else {
- state->depth = 0;
- }
+ if (RB_UNLIKELY(data->state->indent)) {
+ fbuffer_append_str_repeat(buffer, data->state->indent, depth);
}
- tmp = ID2SYM(i_buffer_initial_length);
- if (option_given_p(opts, tmp)) {
- VALUE buffer_initial_length = rb_hash_aref(opts, tmp);
- if (RTEST(buffer_initial_length)) {
- long initial_length;
- Check_Type(buffer_initial_length, T_FIXNUM);
- initial_length = FIX2LONG(buffer_initial_length);
- if (initial_length > 0) state->buffer_initial_length = initial_length;
- }
+
+ VALUE key_to_s;
+ bool as_json_called = false;
+
+ start:
+ switch (key_type) {
+ case T_STRING:
+ if (RB_UNLIKELY(arg->first_key_type != T_STRING)) {
+ json_inspect_hash_with_mixed_keys(arg);
+ }
+
+ if (RB_LIKELY(RBASIC_CLASS(key) == rb_cString)) {
+ key_to_s = key;
+ } else {
+ key_to_s = convert_string_subclass(key);
+ }
+ break;
+ case T_SYMBOL:
+ if (RB_UNLIKELY(arg->first_key_type != T_SYMBOL)) {
+ json_inspect_hash_with_mixed_keys(arg);
+ }
+
+ key_to_s = rb_sym2str(key);
+ break;
+ default:
+ if (data->state->strict) {
+ if (RTEST(data->state->as_json) && !as_json_called) {
+ key = json_call_as_json(data->state, key, Qtrue);
+ key_type = rb_type(key);
+ as_json_called = true;
+ goto start;
+ } else {
+ raise_generator_error(key, "%"PRIsVALUE" not allowed as object key in JSON", CLASS_OF(key));
+ }
+ }
+ key_to_s = rb_convert_type(key, T_STRING, "String", "to_s");
+ break;
}
- tmp = rb_hash_aref(opts, ID2SYM(i_allow_nan));
- state->allow_nan = RTEST(tmp);
- tmp = rb_hash_aref(opts, ID2SYM(i_ascii_only));
- state->ascii_only = RTEST(tmp);
- return self;
-}
-static void set_state_ivars(VALUE hash, VALUE state)
-{
- VALUE ivars = rb_obj_instance_variables(state);
- int i = 0;
- for (i = 0; i < RARRAY_LEN(ivars); i++) {
- VALUE key = rb_funcall(rb_ary_entry(ivars, i), i_to_s, 0);
- long key_len = RSTRING_LEN(key);
- VALUE value = rb_iv_get(state, StringValueCStr(key));
- rb_hash_aset(hash, rb_str_intern(rb_str_substr(key, 1, key_len - 1)), value);
+ key_to_s = ensure_valid_encoding(data, key_to_s, as_json_called, true);
+
+ if (RB_LIKELY(RBASIC_CLASS(key_to_s) == rb_cString)) {
+ raw_generate_json_string(buffer, data, key_to_s);
+ } else {
+ generate_json(buffer, data, key_to_s);
}
-}
+ if (RB_UNLIKELY(state->space_before)) fbuffer_append_str(buffer, data->state->space_before);
+ fbuffer_append_char(buffer, ':');
+ if (RB_UNLIKELY(state->space)) fbuffer_append_str(buffer, data->state->space);
+ generate_json(buffer, data, val);
-/*
- * call-seq: to_h
- *
- * Returns the configuration instance variables as a hash, that can be
- * passed to the configure method.
- */
-static VALUE cState_to_h(VALUE self)
-{
- VALUE result = rb_hash_new();
- GET_STATE(self);
- set_state_ivars(result, self);
- rb_hash_aset(result, ID2SYM(i_indent), rb_str_new(state->indent, state->indent_len));
- rb_hash_aset(result, ID2SYM(i_space), rb_str_new(state->space, state->space_len));
- rb_hash_aset(result, ID2SYM(i_space_before), rb_str_new(state->space_before, state->space_before_len));
- rb_hash_aset(result, ID2SYM(i_object_nl), rb_str_new(state->object_nl, state->object_nl_len));
- rb_hash_aset(result, ID2SYM(i_array_nl), rb_str_new(state->array_nl, state->array_nl_len));
- rb_hash_aset(result, ID2SYM(i_allow_nan), state->allow_nan ? Qtrue : Qfalse);
- rb_hash_aset(result, ID2SYM(i_ascii_only), state->ascii_only ? Qtrue : Qfalse);
- rb_hash_aset(result, ID2SYM(i_max_nesting), LONG2FIX(state->max_nesting));
- rb_hash_aset(result, ID2SYM(i_depth), LONG2FIX(state->depth));
- rb_hash_aset(result, ID2SYM(i_buffer_initial_length), LONG2FIX(state->buffer_initial_length));
- return result;
+ return ST_CONTINUE;
}
-/*
-* call-seq: [](name)
-*
-* Returns the value returned by method +name+.
-*/
-static VALUE cState_aref(VALUE self, VALUE name)
+static inline long increase_depth(struct generate_json_data *data)
{
- name = rb_funcall(name, i_to_s, 0);
- if (RTEST(rb_funcall(self, i_respond_to_p, 1, name))) {
- return rb_funcall(self, i_send, 1, name);
- } else {
- return rb_ivar_get(self, rb_intern_str(rb_str_concat(rb_str_new2("@"), name)));
+ JSON_Generator_State *state = data->state;
+ long depth = ++data->depth;
+ if (RB_UNLIKELY(depth > state->max_nesting && state->max_nesting)) {
+ rb_raise(eNestingError, "nesting of %ld is too deep. Did you try to serialize objects with circular references?", --data->depth);
}
+ return depth;
}
-/*
-* call-seq: []=(name, value)
-*
-* Sets the attribute name to value.
-*/
-static VALUE cState_aset(VALUE self, VALUE name, VALUE value)
+static void generate_json_object(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- VALUE name_writer;
+ long depth = increase_depth(data);
- name = rb_funcall(name, i_to_s, 0);
- name_writer = rb_str_cat2(rb_str_dup(name), "=");
- if (RTEST(rb_funcall(self, i_respond_to_p, 1, name_writer))) {
- return rb_funcall(self, i_send, 2, name_writer, value);
- } else {
- rb_ivar_set(self, rb_intern_str(rb_str_concat(rb_str_new2("@"), name)), value);
+ if (RHASH_SIZE(obj) == 0) {
+ fbuffer_append(buffer, "{}", 2);
+ --data->depth;
+ return;
}
- return Qnil;
-}
-static void generate_json_object(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
-{
- char *object_nl = state->object_nl;
- long object_nl_len = state->object_nl_len;
- char *indent = state->indent;
- long indent_len = state->indent_len;
- long max_nesting = state->max_nesting;
- char *delim = FBUFFER_PTR(state->object_delim);
- long delim_len = FBUFFER_LEN(state->object_delim);
- char *delim2 = FBUFFER_PTR(state->object_delim2);
- long delim2_len = FBUFFER_LEN(state->object_delim2);
- long depth = ++state->depth;
- int i, j;
- VALUE key, key_to_s, keys;
- if (max_nesting != 0 && depth > max_nesting) {
- fbuffer_free(buffer);
- rb_raise(eNestingError, "nesting of %ld is too deep", --state->depth);
- }
fbuffer_append_char(buffer, '{');
- keys = rb_funcall(obj, i_keys, 0);
- for(i = 0; i < RARRAY_LEN(keys); i++) {
- if (i > 0) fbuffer_append(buffer, delim, delim_len);
- if (object_nl) {
- fbuffer_append(buffer, object_nl, object_nl_len);
- }
- if (indent) {
- for (j = 0; j < depth; j++) {
- fbuffer_append(buffer, indent, indent_len);
- }
- }
- key = rb_ary_entry(keys, i);
- key_to_s = rb_funcall(key, i_to_s, 0);
- Check_Type(key_to_s, T_STRING);
- generate_json(buffer, Vstate, state, key_to_s);
- fbuffer_append(buffer, delim2, delim2_len);
- generate_json(buffer, Vstate, state, rb_hash_aref(obj, key));
- }
- depth = --state->depth;
- if (object_nl) {
- fbuffer_append(buffer, object_nl, object_nl_len);
- if (indent) {
- for (j = 0; j < depth; j++) {
- fbuffer_append(buffer, indent, indent_len);
- }
+
+ struct hash_foreach_arg arg = {
+ .hash = obj,
+ .data = data,
+ .first = true,
+ };
+ rb_hash_foreach(obj, json_object_i, (VALUE)&arg);
+
+ depth = --data->depth;
+ if (RB_UNLIKELY(data->state->object_nl)) {
+ fbuffer_append_str(buffer, data->state->object_nl);
+ if (RB_UNLIKELY(data->state->indent)) {
+ fbuffer_append_str_repeat(buffer, data->state->indent, depth);
}
}
fbuffer_append_char(buffer, '}');
}
-static void generate_json_array(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
-{
- char *array_nl = state->array_nl;
- long array_nl_len = state->array_nl_len;
- char *indent = state->indent;
- long indent_len = state->indent_len;
- long max_nesting = state->max_nesting;
- char *delim = FBUFFER_PTR(state->array_delim);
- long delim_len = FBUFFER_LEN(state->array_delim);
- long depth = ++state->depth;
- int i, j;
- if (max_nesting != 0 && depth > max_nesting) {
- fbuffer_free(buffer);
- rb_raise(eNestingError, "nesting of %ld is too deep", --state->depth);
+static void generate_json_array(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
+{
+ long depth = increase_depth(data);
+
+ if (RARRAY_LEN(obj) == 0) {
+ fbuffer_append(buffer, "[]", 2);
+ --data->depth;
+ return;
}
+
fbuffer_append_char(buffer, '[');
- if (array_nl) fbuffer_append(buffer, array_nl, array_nl_len);
- for(i = 0; i < RARRAY_LEN(obj); i++) {
- if (i > 0) fbuffer_append(buffer, delim, delim_len);
- if (indent) {
- for (j = 0; j < depth; j++) {
- fbuffer_append(buffer, indent, indent_len);
- }
+ if (RB_UNLIKELY(data->state->array_nl)) fbuffer_append_str(buffer, data->state->array_nl);
+ for (int i = 0; i < RARRAY_LEN(obj); i++) {
+ if (i > 0) {
+ fbuffer_append_char(buffer, ',');
+ if (RB_UNLIKELY(data->state->array_nl)) fbuffer_append_str(buffer, data->state->array_nl);
+ }
+ if (RB_UNLIKELY(data->state->indent)) {
+ fbuffer_append_str_repeat(buffer, data->state->indent, depth);
}
- generate_json(buffer, Vstate, state, rb_ary_entry(obj, i));
+ generate_json(buffer, data, RARRAY_AREF(obj, i));
}
- state->depth = --depth;
- if (array_nl) {
- fbuffer_append(buffer, array_nl, array_nl_len);
- if (indent) {
- for (j = 0; j < depth; j++) {
- fbuffer_append(buffer, indent, indent_len);
- }
+ data->depth = --depth;
+ if (RB_UNLIKELY(data->state->array_nl)) {
+ fbuffer_append_str(buffer, data->state->array_nl);
+ if (RB_UNLIKELY(data->state->indent)) {
+ fbuffer_append_str_repeat(buffer, data->state->indent, depth);
}
}
fbuffer_append_char(buffer, ']');
}
-static void generate_json_string(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_fallback(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- fbuffer_append_char(buffer, '"');
-#ifdef HAVE_RUBY_ENCODING_H
- obj = rb_funcall(obj, i_encode, 1, CEncoding_UTF_8);
-#endif
- if (state->ascii_only) {
- convert_UTF8_to_JSON_ASCII(buffer, obj);
+ VALUE tmp;
+ if (rb_respond_to(obj, i_to_json)) {
+ tmp = json_call_to_json(data, obj);
+ Check_Type(tmp, T_STRING);
+ fbuffer_append_str(buffer, tmp);
} else {
- convert_UTF8_to_JSON(buffer, obj);
+ tmp = rb_funcall(obj, i_to_s, 0);
+ Check_Type(tmp, T_STRING);
+ generate_json_string(buffer, data, tmp);
}
- fbuffer_append_char(buffer, '"');
}
-static void generate_json_null(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static inline void generate_json_symbol(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
+{
+ if (data->state->strict) {
+ generate_json_string(buffer, data, rb_sym2str(obj));
+ } else {
+ generate_json_fallback(buffer, data, obj);
+ }
+}
+
+static void generate_json_null(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
fbuffer_append(buffer, "null", 4);
}
-static void generate_json_false(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_false(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
fbuffer_append(buffer, "false", 5);
}
-static void generate_json_true(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_true(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
fbuffer_append(buffer, "true", 4);
}
-static void generate_json_fixnum(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_fixnum(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
fbuffer_append_long(buffer, FIX2LONG(obj));
}
-static void generate_json_bignum(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+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
-static void generate_json_integer(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
-{
- if (FIXNUM_P(obj))
- generate_json_fixnum(buffer, Vstate, state, obj);
- else
- generate_json_bignum(buffer, Vstate, state, obj);
-}
-#endif
-static void generate_json_float(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_float(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
double value = RFLOAT_VALUE(obj);
- char allow_nan = state->allow_nan;
- VALUE tmp = rb_funcall(obj, i_to_s, 0);
- if (!allow_nan) {
- if (isinf(value)) {
- fbuffer_free(buffer);
- rb_raise(eGeneratorError, "%u: %"PRIsVALUE" not allowed in JSON", __LINE__, RB_OBJ_STRING(tmp));
- } else if (isnan(value)) {
- fbuffer_free(buffer);
- rb_raise(eGeneratorError, "%u: %"PRIsVALUE" not allowed in JSON", __LINE__, RB_OBJ_STRING(tmp));
+ char allow_nan = data->state->allow_nan;
+ if (isinf(value) || isnan(value)) {
+ /* for NaN and Infinity values we either raise an error or rely on Float#to_s. */
+ if (!allow_nan) {
+ if (data->state->strict && data->state->as_json) {
+ VALUE casted_obj = json_call_as_json(data->state, obj, Qfalse);
+ if (casted_obj != obj) {
+ increase_depth(data);
+ generate_json(buffer, data, casted_obj);
+ data->depth--;
+ return;
+ }
+ }
+ raise_generator_error(obj, "%"PRIsVALUE" not allowed in JSON", rb_funcall(obj, i_to_s, 0));
}
+
+ VALUE tmp = rb_funcall(obj, i_to_s, 0);
+ fbuffer_append_str(buffer, tmp);
+ return;
}
- fbuffer_append_str(buffer, tmp);
+
+ /* This implementation writes directly into the buffer. We reserve
+ * the 32 characters that fpconv_dtoa states as its maximum.
+ */
+ fbuffer_inc_capa(buffer, 32);
+ char* d = buffer->ptr + buffer->len;
+ int len = fpconv_dtoa(value, d);
+ /* fpconv_dtoa converts a float to its shortest string representation,
+ * but it adds a ".0" if this is a plain integer.
+ */
+ fbuffer_consumed(buffer, len);
}
-static void generate_json(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj)
+static void generate_json_fragment(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- VALUE tmp;
- VALUE klass = CLASS_OF(obj);
- if (klass == rb_cHash) {
- generate_json_object(buffer, Vstate, state, obj);
- } else if (klass == rb_cArray) {
- generate_json_array(buffer, Vstate, state, obj);
- } else if (klass == rb_cString) {
- generate_json_string(buffer, Vstate, state, obj);
- } else if (obj == Qnil) {
- generate_json_null(buffer, Vstate, state, obj);
+ VALUE fragment = RSTRUCT_GET(obj, 0);
+ Check_Type(fragment, T_STRING);
+ fbuffer_append_str(buffer, fragment);
+}
+
+static inline void generate_json_general(FBuffer *buffer, struct generate_json_data *data, VALUE obj, bool fallback)
+{
+ bool as_json_called = false;
+start:
+ if (obj == Qnil) {
+ generate_json_null(buffer, data, obj);
} else if (obj == Qfalse) {
- generate_json_false(buffer, Vstate, state, obj);
+ generate_json_false(buffer, data, obj);
} else if (obj == Qtrue) {
- generate_json_true(buffer, Vstate, state, obj);
- } else if (FIXNUM_P(obj)) {
- generate_json_fixnum(buffer, Vstate, state, obj);
- } else if (RB_TYPE_P(obj, T_BIGNUM)) {
- generate_json_bignum(buffer, Vstate, state, obj);
- } else if (klass == rb_cFloat) {
- generate_json_float(buffer, Vstate, state, obj);
- } else if (rb_respond_to(obj, i_to_json)) {
- tmp = rb_funcall(obj, i_to_json, 1, Vstate);
- Check_Type(tmp, T_STRING);
- fbuffer_append_str(buffer, tmp);
+ generate_json_true(buffer, data, obj);
+ } else if (RB_SPECIAL_CONST_P(obj)) {
+ if (RB_FIXNUM_P(obj)) {
+ generate_json_fixnum(buffer, data, obj);
+ } else if (RB_FLONUM_P(obj)) {
+ generate_json_float(buffer, data, obj);
+ } else if (RB_STATIC_SYM_P(obj)) {
+ generate_json_symbol(buffer, data, obj);
+ } else {
+ goto general;
+ }
} else {
- tmp = rb_funcall(obj, i_to_s, 0);
- Check_Type(tmp, T_STRING);
- generate_json_string(buffer, Vstate, state, tmp);
+ VALUE klass = RBASIC_CLASS(obj);
+ switch (RB_BUILTIN_TYPE(obj)) {
+ case T_BIGNUM:
+ generate_json_bignum(buffer, data, obj);
+ break;
+ case T_HASH:
+ if (fallback && klass != rb_cHash) goto general;
+ generate_json_object(buffer, data, obj);
+ break;
+ case T_ARRAY:
+ if (fallback && klass != rb_cArray) goto general;
+ generate_json_array(buffer, data, obj);
+ break;
+ case T_STRING:
+ if (fallback && klass != rb_cString) goto general;
+
+ if (RB_LIKELY(valid_json_string_p(obj))) {
+ raw_generate_json_string(buffer, data, obj);
+ } else if (as_json_called) {
+ raise_generator_error(obj, "source sequence is illegal/malformed utf-8");
+ } else {
+ obj = ensure_valid_encoding(data, obj, false, false);
+ as_json_called = true;
+ goto start;
+ }
+ break;
+ case T_SYMBOL:
+ generate_json_symbol(buffer, data, obj);
+ break;
+ case T_FLOAT:
+ if (fallback && klass != rb_cFloat) goto general;
+ generate_json_float(buffer, data, obj);
+ break;
+ case T_STRUCT:
+ if (klass != cFragment) goto general;
+ generate_json_fragment(buffer, data, obj);
+ break;
+ default:
+ general:
+ if (data->state->strict) {
+ if (RTEST(data->state->as_json) && !as_json_called) {
+ obj = json_call_as_json(data->state, obj, Qfalse);
+ as_json_called = true;
+ goto start;
+ } else {
+ raise_generator_error(obj, "%"PRIsVALUE" not allowed in JSON", CLASS_OF(obj));
+ }
+ } else {
+ generate_json_fallback(buffer, data, obj);
+ }
+ }
}
}
-static FBuffer *cState_prepare_buffer(VALUE self)
+static void generate_json(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
{
- FBuffer *buffer;
- GET_STATE(self);
- buffer = fbuffer_alloc(state->buffer_initial_length);
+ generate_json_general(buffer, data, obj, true);
+}
- if (state->object_delim) {
- fbuffer_clear(state->object_delim);
- } else {
- state->object_delim = fbuffer_alloc(16);
- }
- fbuffer_append_char(state->object_delim, ',');
- if (state->object_delim2) {
- fbuffer_clear(state->object_delim2);
- } else {
- state->object_delim2 = fbuffer_alloc(16);
- }
- if (state->space_before) fbuffer_append(state->object_delim2, state->space_before, state->space_before_len);
- fbuffer_append_char(state->object_delim2, ':');
- if (state->space) fbuffer_append(state->object_delim2, state->space, state->space_len);
+static void generate_json_no_fallback(FBuffer *buffer, struct generate_json_data *data, VALUE obj)
+{
+ generate_json_general(buffer, data, obj, false);
+}
- if (state->array_delim) {
- fbuffer_clear(state->array_delim);
- } else {
- state->array_delim = fbuffer_alloc(16);
- }
- fbuffer_append_char(state->array_delim, ',');
- if (state->array_nl) fbuffer_append(state->array_delim, state->array_nl, state->array_nl_len);
- return buffer;
+static VALUE generate_json_try(VALUE d)
+{
+ struct generate_json_data *data = (struct generate_json_data *)d;
+
+ data->func(data->buffer, data, data->obj);
+
+ return fbuffer_finalize(data->buffer);
}
-static VALUE cState_partial_generate(VALUE self, VALUE obj)
+static VALUE generate_json_ensure(VALUE d)
{
- FBuffer *buffer = cState_prepare_buffer(self);
- GET_STATE(self);
- generate_json(buffer, self, state, obj);
- return fbuffer_to_s(buffer);
+ struct generate_json_data *data = (struct generate_json_data *)d;
+ fbuffer_free(data->buffer);
+
+ return Qundef;
}
-/*
- * call-seq: generate(obj)
+static inline VALUE cState_partial_generate(VALUE self, VALUE obj, generator_func func, VALUE io)
+{
+ GET_STATE(self);
+
+ char stack_buffer[FBUFFER_STACK_SIZE];
+ FBuffer buffer = {
+ .io = RTEST(io) ? io : Qfalse,
+ };
+ fbuffer_stack_init(&buffer, state->buffer_initial_length, stack_buffer, FBUFFER_STACK_SIZE);
+
+ struct generate_json_data data = {
+ .buffer = &buffer,
+ .vstate = Qfalse, // don't use self as it may be frozen and its depth is mutated when calling to_json
+ .state = state,
+ .depth = state->depth,
+ .obj = obj,
+ .func = func
+ };
+ return rb_ensure(generate_json_try, (VALUE)&data, generate_json_ensure, (VALUE)&data);
+}
+
+/* call-seq:
+ * generate(obj) -> String
+ * generate(obj, anIO) -> anIO
*
* Generates a valid JSON document from object +obj+ and returns the
* result. If no valid JSON document can be created this method raises a
* GeneratorError exception.
*/
-static VALUE cState_generate(VALUE self, VALUE obj)
+static VALUE cState_generate(int argc, VALUE *argv, VALUE self)
{
- VALUE result = cState_partial_generate(self, obj);
- GET_STATE(self);
- (void)state;
- return result;
+ rb_check_arity(argc, 1, 2);
+ VALUE obj = argv[0];
+ VALUE io = argc > 1 ? argv[1] : Qnil;
+ return cState_partial_generate(self, obj, generate_json, io);
+}
+
+/* :nodoc: */
+static VALUE cState_generate_no_fallback(int argc, VALUE *argv, VALUE self)
+{
+ rb_check_arity(argc, 1, 2);
+ VALUE obj = argv[0];
+ VALUE io = argc > 1 ? argv[1] : Qnil;
+ return cState_partial_generate(self, obj, generate_json_no_fallback, io);
}
-/*
- * call-seq: new(opts = {})
- *
- * Instantiates a new State object, configured by _opts_.
- *
- * _opts_ can have the following keys:
- *
- * * *indent*: a string used to indent levels (default: ''),
- * * *space*: a string that is put after, a : or , delimiter (default: ''),
- * * *space_before*: a string that is put before a : pair delimiter (default: ''),
- * * *object_nl*: a string that is put at the end of a JSON object (default: ''),
- * * *array_nl*: a string that is put at the end of a JSON array (default: ''),
- * * *allow_nan*: true if NaN, Infinity, and -Infinity should be
- * generated, otherwise an exception is thrown, if these values are
- * encountered. This options defaults to false.
- * * *buffer_initial_length*: sets the initial length of the generator's
- * internal buffer.
- */
static VALUE cState_initialize(int argc, VALUE *argv, VALUE self)
{
- VALUE opts;
- GET_STATE(self);
- state->max_nesting = 100;
- state->buffer_initial_length = FBUFFER_INITIAL_LENGTH_DEFAULT;
- rb_scan_args(argc, argv, "01", &opts);
- if (!NIL_P(opts)) cState_configure(self, opts);
+ rb_warn("The json gem extension was loaded with the stdlib ruby code. You should upgrade rubygems with `gem update --system`");
return self;
}
@@ -1000,14 +1367,14 @@ static VALUE cState_init_copy(VALUE obj, VALUE orig)
if (!objState) rb_raise(rb_eArgError, "unallocated JSON::State");
MEMCPY(objState, origState, JSON_Generator_State, 1);
- objState->indent = fstrndup(origState->indent, origState->indent_len);
- objState->space = fstrndup(origState->space, origState->space_len);
- objState->space_before = fstrndup(origState->space_before, origState->space_before_len);
- objState->object_nl = fstrndup(origState->object_nl, origState->object_nl_len);
- objState->array_nl = fstrndup(origState->array_nl, origState->array_nl_len);
- if (origState->array_delim) objState->array_delim = fbuffer_dup(origState->array_delim);
- if (origState->object_delim) objState->object_delim = fbuffer_dup(origState->object_delim);
- if (origState->object_delim2) objState->object_delim2 = fbuffer_dup(origState->object_delim2);
+
+ RB_OBJ_WRITTEN(obj, Qundef, objState->indent);
+ RB_OBJ_WRITTEN(obj, Qundef, objState->space);
+ RB_OBJ_WRITTEN(obj, Qundef, objState->space_before);
+ RB_OBJ_WRITTEN(obj, Qundef, objState->object_nl);
+ RB_OBJ_WRITTEN(obj, Qundef, objState->array_nl);
+ RB_OBJ_WRITTEN(obj, Qundef, objState->as_json);
+
return obj;
}
@@ -1025,10 +1392,7 @@ static VALUE cState_from_state_s(VALUE self, VALUE opts)
} else if (rb_obj_is_kind_of(opts, rb_cHash)) {
return rb_funcall(self, i_new, 1, opts);
} else {
- if (NIL_P(CJSON_SAFE_STATE_PROTOTYPE)) {
- CJSON_SAFE_STATE_PROTOTYPE = rb_const_get(mJSON, i_SAFE_STATE_PROTOTYPE);
- }
- return rb_funcall(CJSON_SAFE_STATE_PROTOTYPE, i_dup, 0);
+ return rb_class_new_instance(0, NULL, cState);
}
}
@@ -1040,7 +1404,18 @@ static VALUE cState_from_state_s(VALUE self, VALUE opts)
static VALUE cState_indent(VALUE self)
{
GET_STATE(self);
- return state->indent ? rb_str_new(state->indent, state->indent_len) : rb_str_new2("");
+ return state->indent ? state->indent : rb_str_freeze(rb_utf8_str_new("", 0));
+}
+
+static VALUE string_config(VALUE config)
+{
+ if (RTEST(config)) {
+ Check_Type(config, T_STRING);
+ if (RSTRING_LEN(config)) {
+ return rb_str_new_frozen(config);
+ }
+ }
+ return Qfalse;
}
/*
@@ -1050,21 +1425,9 @@ static VALUE cState_indent(VALUE self)
*/
static VALUE cState_indent_set(VALUE self, VALUE indent)
{
- unsigned long len;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(indent, T_STRING);
- len = RSTRING_LEN(indent);
- if (len == 0) {
- if (state->indent) {
- ruby_xfree(state->indent);
- state->indent = NULL;
- state->indent_len = 0;
- }
- } else {
- if (state->indent) ruby_xfree(state->indent);
- state->indent = fstrndup(RSTRING_PTR(indent), len);
- state->indent_len = len;
- }
+ RB_OBJ_WRITE(self, &state->indent, string_config(indent));
return Qnil;
}
@@ -1077,7 +1440,7 @@ static VALUE cState_indent_set(VALUE self, VALUE indent)
static VALUE cState_space(VALUE self)
{
GET_STATE(self);
- return state->space ? rb_str_new(state->space, state->space_len) : rb_str_new2("");
+ return state->space ? state->space : rb_str_freeze(rb_utf8_str_new("", 0));
}
/*
@@ -1088,21 +1451,9 @@ static VALUE cState_space(VALUE self)
*/
static VALUE cState_space_set(VALUE self, VALUE space)
{
- unsigned long len;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(space, T_STRING);
- len = RSTRING_LEN(space);
- if (len == 0) {
- if (state->space) {
- ruby_xfree(state->space);
- state->space = NULL;
- state->space_len = 0;
- }
- } else {
- if (state->space) ruby_xfree(state->space);
- state->space = fstrndup(RSTRING_PTR(space), len);
- state->space_len = len;
- }
+ RB_OBJ_WRITE(self, &state->space, string_config(space));
return Qnil;
}
@@ -1114,7 +1465,7 @@ static VALUE cState_space_set(VALUE self, VALUE space)
static VALUE cState_space_before(VALUE self)
{
GET_STATE(self);
- return state->space_before ? rb_str_new(state->space_before, state->space_before_len) : rb_str_new2("");
+ return state->space_before ? state->space_before : rb_str_freeze(rb_utf8_str_new("", 0));
}
/*
@@ -1124,21 +1475,9 @@ static VALUE cState_space_before(VALUE self)
*/
static VALUE cState_space_before_set(VALUE self, VALUE space_before)
{
- unsigned long len;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(space_before, T_STRING);
- len = RSTRING_LEN(space_before);
- if (len == 0) {
- if (state->space_before) {
- ruby_xfree(state->space_before);
- state->space_before = NULL;
- state->space_before_len = 0;
- }
- } else {
- if (state->space_before) ruby_xfree(state->space_before);
- state->space_before = fstrndup(RSTRING_PTR(space_before), len);
- state->space_before_len = len;
- }
+ RB_OBJ_WRITE(self, &state->space_before, string_config(space_before));
return Qnil;
}
@@ -1151,7 +1490,7 @@ static VALUE cState_space_before_set(VALUE self, VALUE space_before)
static VALUE cState_object_nl(VALUE self)
{
GET_STATE(self);
- return state->object_nl ? rb_str_new(state->object_nl, state->object_nl_len) : rb_str_new2("");
+ return state->object_nl ? state->object_nl : rb_str_freeze(rb_utf8_str_new("", 0));
}
/*
@@ -1162,20 +1501,9 @@ static VALUE cState_object_nl(VALUE self)
*/
static VALUE cState_object_nl_set(VALUE self, VALUE object_nl)
{
- unsigned long len;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(object_nl, T_STRING);
- len = RSTRING_LEN(object_nl);
- if (len == 0) {
- if (state->object_nl) {
- ruby_xfree(state->object_nl);
- state->object_nl = NULL;
- }
- } else {
- if (state->object_nl) ruby_xfree(state->object_nl);
- state->object_nl = fstrndup(RSTRING_PTR(object_nl), len);
- state->object_nl_len = len;
- }
+ RB_OBJ_WRITE(self, &state->object_nl, string_config(object_nl));
return Qnil;
}
@@ -1187,7 +1515,7 @@ static VALUE cState_object_nl_set(VALUE self, VALUE object_nl)
static VALUE cState_array_nl(VALUE self)
{
GET_STATE(self);
- return state->array_nl ? rb_str_new(state->array_nl, state->array_nl_len) : rb_str_new2("");
+ return state->array_nl ? state->array_nl : rb_str_freeze(rb_utf8_str_new("", 0));
}
/*
@@ -1197,23 +1525,35 @@ static VALUE cState_array_nl(VALUE self)
*/
static VALUE cState_array_nl_set(VALUE self, VALUE array_nl)
{
- unsigned long len;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(array_nl, T_STRING);
- len = RSTRING_LEN(array_nl);
- if (len == 0) {
- if (state->array_nl) {
- ruby_xfree(state->array_nl);
- state->array_nl = NULL;
- }
- } else {
- if (state->array_nl) ruby_xfree(state->array_nl);
- state->array_nl = fstrndup(RSTRING_PTR(array_nl), len);
- state->array_nl_len = len;
- }
+ RB_OBJ_WRITE(self, &state->array_nl, string_config(array_nl));
return Qnil;
}
+/*
+ * call-seq: as_json()
+ *
+ * This string is put at the end of a line that holds a JSON array.
+ */
+static VALUE cState_as_json(VALUE self)
+{
+ GET_STATE(self);
+ return state->as_json;
+}
+
+/*
+ * call-seq: as_json=(as_json)
+ *
+ * This string is put at the end of a line that holds a JSON array.
+ */
+static VALUE cState_as_json_set(VALUE self, VALUE as_json)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ RB_OBJ_WRITE(self, &state->as_json, rb_convert_type(as_json, T_DATA, "Proc", "to_proc"));
+ return Qnil;
+}
/*
* call-seq: check_circular?
@@ -1239,6 +1579,25 @@ static VALUE cState_max_nesting(VALUE self)
return LONG2FIX(state->max_nesting);
}
+static long long_config(VALUE num)
+{
+ return RTEST(num) ? NUM2LONG(num) : 0;
+}
+
+// depth must never be negative; reject early with a clear error.
+static long depth_config(VALUE num)
+{
+ if (!RTEST(num)) return 0;
+ long d = NUM2LONG(num);
+ if (RB_UNLIKELY(d < 0)) {
+ rb_raise(rb_eArgError, "depth must be >= 0 (got %ld)", d);
+ }
+ if (RB_UNLIKELY(d > INT_MAX)) {
+ rb_raise(rb_eArgError, "depth is too large (got %ld)", d);
+ }
+ return d;
+}
+
/*
* call-seq: max_nesting=(depth)
*
@@ -1247,9 +1606,68 @@ static VALUE cState_max_nesting(VALUE self)
*/
static VALUE cState_max_nesting_set(VALUE self, VALUE depth)
{
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(depth, T_FIXNUM);
- return state->max_nesting = FIX2LONG(depth);
+ state->max_nesting = long_config(depth);
+ return Qnil;
+}
+
+/*
+ * call-seq: script_safe
+ *
+ * If this boolean is true, the forward slashes will be escaped in
+ * the json output.
+ */
+static VALUE cState_script_safe(VALUE self)
+{
+ GET_STATE(self);
+ return state->script_safe ? Qtrue : Qfalse;
+}
+
+/*
+ * call-seq: script_safe=(enable)
+ *
+ * This sets whether or not the forward slashes will be escaped in
+ * the json output.
+ */
+static VALUE cState_script_safe_set(VALUE self, VALUE enable)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ state->script_safe = RTEST(enable);
+ return Qnil;
+}
+
+/*
+ * call-seq: strict
+ *
+ * If this boolean is false, types unsupported by the JSON format will
+ * be serialized as strings.
+ * If this boolean is true, types unsupported by the JSON format will
+ * raise a JSON::GeneratorError.
+ */
+static VALUE cState_strict(VALUE self)
+{
+ GET_STATE(self);
+ return state->strict ? Qtrue : Qfalse;
+}
+
+/*
+ * call-seq: strict=(enable)
+ *
+ * This sets whether or not to serialize types unsupported by the
+ * JSON format as strings.
+ * If this boolean is false, types unsupported by the JSON format will
+ * be serialized as strings.
+ * If this boolean is true, types unsupported by the JSON format will
+ * raise a JSON::GeneratorError.
+ */
+static VALUE cState_strict_set(VALUE self, VALUE enable)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ state->strict = RTEST(enable);
+ return Qnil;
}
/*
@@ -1265,9 +1683,22 @@ static VALUE cState_allow_nan_p(VALUE self)
}
/*
+ * call-seq: allow_nan=(enable)
+ *
+ * This sets whether or not to serialize NaN, Infinity, and -Infinity
+ */
+static VALUE cState_allow_nan_set(VALUE self, VALUE enable)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ state->allow_nan = RTEST(enable);
+ return Qnil;
+}
+
+/*
* call-seq: ascii_only?
*
- * Returns true, if NaN, Infinity, and -Infinity should be generated, otherwise
+ * Returns true, if only ASCII characters should be generated. Otherwise
* returns false.
*/
static VALUE cState_ascii_only_p(VALUE self)
@@ -1277,6 +1708,32 @@ static VALUE cState_ascii_only_p(VALUE self)
}
/*
+ * call-seq: ascii_only=(enable)
+ *
+ * This sets whether only ASCII characters should be generated.
+ */
+static VALUE cState_ascii_only_set(VALUE self, VALUE enable)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ state->ascii_only = RTEST(enable);
+ return Qnil;
+}
+
+static VALUE cState_allow_duplicate_key_p(VALUE self)
+{
+ GET_STATE(self);
+ switch (state->on_duplicate_key) {
+ case JSON_IGNORE:
+ return Qtrue;
+ case JSON_DEPRECATED:
+ return Qnil;
+ default:
+ return Qfalse;
+ }
+}
+
+/*
* call-seq: depth
*
* This integer returns the current depth of data structure nesting.
@@ -1295,9 +1752,9 @@ static VALUE cState_depth(VALUE self)
*/
static VALUE cState_depth_set(VALUE self, VALUE depth)
{
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(depth, T_FIXNUM);
- state->depth = FIX2LONG(depth);
+ state->depth = depth_config(depth);
return Qnil;
}
@@ -1312,6 +1769,15 @@ static VALUE cState_buffer_initial_length(VALUE self)
return LONG2FIX(state->buffer_initial_length);
}
+static void buffer_initial_length_set(JSON_Generator_State *state, VALUE buffer_initial_length)
+{
+ Check_Type(buffer_initial_length, T_FIXNUM);
+ long initial_length = FIX2LONG(buffer_initial_length);
+ if (initial_length > 0) {
+ state->buffer_initial_length = initial_length;
+ }
+}
+
/*
* call-seq: buffer_initial_length=(length)
*
@@ -1320,34 +1786,142 @@ static VALUE cState_buffer_initial_length(VALUE self)
*/
static VALUE cState_buffer_initial_length_set(VALUE self, VALUE buffer_initial_length)
{
- long initial_length;
+ rb_check_frozen(self);
GET_STATE(self);
- Check_Type(buffer_initial_length, T_FIXNUM);
- initial_length = FIX2LONG(buffer_initial_length);
- if (initial_length > 0) {
- state->buffer_initial_length = initial_length;
- }
+ buffer_initial_length_set(state, buffer_initial_length);
return Qnil;
}
-/*
- *
- */
+struct configure_state_data {
+ JSON_Generator_State *state;
+ VALUE vstate; // Ruby object that owns the state, or Qfalse if stack-allocated
+};
+
+static inline void state_write_value(struct configure_state_data *data, VALUE *field, VALUE value)
+{
+ if (RTEST(data->vstate)) {
+ RB_OBJ_WRITE(data->vstate, field, value);
+ } else {
+ *field = value;
+ }
+}
+
+static int configure_state_i(VALUE key, VALUE val, VALUE _arg)
+{
+ struct configure_state_data *data = (struct configure_state_data *)_arg;
+ JSON_Generator_State *state = data->state;
+
+ if (key == sym_indent) { state_write_value(data, &state->indent, string_config(val)); }
+ else if (key == sym_space) { state_write_value(data, &state->space, string_config(val)); }
+ else if (key == sym_space_before) { state_write_value(data, &state->space_before, string_config(val)); }
+ else if (key == sym_object_nl) { state_write_value(data, &state->object_nl, string_config(val)); }
+ else if (key == sym_array_nl) { state_write_value(data, &state->array_nl, string_config(val)); }
+ else if (key == sym_max_nesting) { state->max_nesting = long_config(val); }
+ else if (key == sym_allow_nan) { state->allow_nan = RTEST(val); }
+ else if (key == sym_ascii_only) { state->ascii_only = RTEST(val); }
+ else if (key == sym_depth) { state->depth = depth_config(val); }
+ else if (key == sym_buffer_initial_length) { buffer_initial_length_set(state, val); }
+ else if (key == sym_script_safe) { state->script_safe = RTEST(val); }
+ else if (key == sym_escape_slash) { state->script_safe = RTEST(val); }
+ else if (key == sym_strict) { state->strict = RTEST(val); }
+ else if (key == sym_allow_duplicate_key) { state->on_duplicate_key = RTEST(val) ? JSON_IGNORE : JSON_RAISE; }
+ else if (key == sym_as_json) {
+ VALUE proc = RTEST(val) ? rb_convert_type(val, T_DATA, "Proc", "to_proc") : Qfalse;
+ state->as_json_single_arg = proc && rb_proc_arity(proc) == 1;
+ state_write_value(data, &state->as_json, proc);
+ }
+ return ST_CONTINUE;
+}
+
+static void configure_state(JSON_Generator_State *state, VALUE vstate, VALUE config)
+{
+ if (!RTEST(config)) return;
+
+ Check_Type(config, T_HASH);
+
+ if (!RHASH_SIZE(config)) return;
+
+ struct configure_state_data data = {
+ .state = state,
+ .vstate = vstate
+ };
+
+ // We assume in most cases few keys are set so it's faster to go over
+ // the provided keys than to check all possible keys.
+ rb_hash_foreach(config, configure_state_i, (VALUE)&data);
+}
+
+static VALUE cState_configure(VALUE self, VALUE opts)
+{
+ rb_check_frozen(self);
+ GET_STATE(self);
+ configure_state(state, self, opts);
+ return self;
+}
+
+static VALUE cState_m_do_generate(VALUE klass, VALUE obj, VALUE opts, VALUE io, generator_func func)
+{
+ JSON_Generator_State state = {0};
+ state_init(&state);
+ configure_state(&state, Qfalse, opts);
+
+ char stack_buffer[FBUFFER_STACK_SIZE];
+ FBuffer buffer = {
+ .io = RTEST(io) ? io : Qfalse,
+ };
+ fbuffer_stack_init(&buffer, state.buffer_initial_length, stack_buffer, FBUFFER_STACK_SIZE);
+
+ struct generate_json_data data = {
+ .buffer = &buffer,
+ .vstate = Qfalse,
+ .state = &state,
+ .depth = state.depth,
+ .obj = obj,
+ .func = func,
+ };
+ return rb_ensure(generate_json_try, (VALUE)&data, generate_json_ensure, (VALUE)&data);
+}
+
+static VALUE cState_m_generate(VALUE klass, VALUE obj, VALUE opts, VALUE io)
+{
+ return cState_m_do_generate(klass, obj, opts, io, generate_json);
+}
+
+static VALUE cState_m_generate_no_fallback(VALUE klass, VALUE obj, VALUE opts, VALUE io)
+{
+ return cState_m_do_generate(klass, obj, opts, io, generate_json_no_fallback);
+}
+
void Init_generator(void)
{
+#ifdef HAVE_RB_EXT_RACTOR_SAFE
+ rb_ext_ractor_safe(true);
+#endif
+
+#undef rb_intern
rb_require("json/common");
mJSON = rb_define_module("JSON");
- mExt = rb_define_module_under(mJSON, "Ext");
- mGenerator = rb_define_module_under(mExt, "Generator");
+ rb_global_variable(&cFragment);
+ cFragment = rb_const_get(mJSON, rb_intern("Fragment"));
+
+ VALUE mExt = rb_define_module_under(mJSON, "Ext");
+ VALUE mGenerator = rb_define_module_under(mExt, "Generator");
+
+ rb_global_variable(&eGeneratorError);
eGeneratorError = rb_path2class("JSON::GeneratorError");
+
+ rb_global_variable(&eNestingError);
eNestingError = rb_path2class("JSON::NestingError");
cState = rb_define_class_under(mGenerator, "State", rb_cObject);
rb_define_alloc_func(cState, cState_s_allocate);
rb_define_singleton_method(cState, "from_state", cState_from_state_s, 1);
rb_define_method(cState, "initialize", cState_initialize, -1);
+ rb_define_alias(cState, "initialize", "initialize"); // avoid method redefinition warnings
+ rb_define_private_method(cState, "_configure", cState_configure, 1);
+
rb_define_method(cState, "initialize_copy", cState_init_copy, 1);
rb_define_method(cState, "indent", cState_indent, 0);
rb_define_method(cState, "indent=", cState_indent_set, 1);
@@ -1359,85 +1933,65 @@ void Init_generator(void)
rb_define_method(cState, "object_nl=", cState_object_nl_set, 1);
rb_define_method(cState, "array_nl", cState_array_nl, 0);
rb_define_method(cState, "array_nl=", cState_array_nl_set, 1);
+ rb_define_method(cState, "as_json", cState_as_json, 0);
+ rb_define_method(cState, "as_json=", cState_as_json_set, 1);
rb_define_method(cState, "max_nesting", cState_max_nesting, 0);
rb_define_method(cState, "max_nesting=", cState_max_nesting_set, 1);
+ rb_define_method(cState, "script_safe", cState_script_safe, 0);
+ rb_define_method(cState, "script_safe?", cState_script_safe, 0);
+ rb_define_method(cState, "script_safe=", cState_script_safe_set, 1);
+ rb_define_alias(cState, "escape_slash", "script_safe");
+ rb_define_alias(cState, "escape_slash?", "script_safe?");
+ rb_define_alias(cState, "escape_slash=", "script_safe=");
+ rb_define_method(cState, "strict", cState_strict, 0);
+ rb_define_method(cState, "strict?", cState_strict, 0);
+ rb_define_method(cState, "strict=", cState_strict_set, 1);
rb_define_method(cState, "check_circular?", cState_check_circular_p, 0);
rb_define_method(cState, "allow_nan?", cState_allow_nan_p, 0);
+ rb_define_method(cState, "allow_nan=", cState_allow_nan_set, 1);
rb_define_method(cState, "ascii_only?", cState_ascii_only_p, 0);
+ rb_define_method(cState, "ascii_only=", cState_ascii_only_set, 1);
rb_define_method(cState, "depth", cState_depth, 0);
rb_define_method(cState, "depth=", cState_depth_set, 1);
rb_define_method(cState, "buffer_initial_length", cState_buffer_initial_length, 0);
rb_define_method(cState, "buffer_initial_length=", cState_buffer_initial_length_set, 1);
- rb_define_method(cState, "configure", cState_configure, 1);
- rb_define_alias(cState, "merge", "configure");
- rb_define_method(cState, "to_h", cState_to_h, 0);
- rb_define_alias(cState, "to_hash", "to_h");
- rb_define_method(cState, "[]", cState_aref, 1);
- rb_define_method(cState, "[]=", cState_aset, 2);
- rb_define_method(cState, "generate", cState_generate, 1);
-
- mGeneratorMethods = rb_define_module_under(mGenerator, "GeneratorMethods");
- mObject = rb_define_module_under(mGeneratorMethods, "Object");
- rb_define_method(mObject, "to_json", mObject_to_json, -1);
- mHash = rb_define_module_under(mGeneratorMethods, "Hash");
- rb_define_method(mHash, "to_json", mHash_to_json, -1);
- mArray = rb_define_module_under(mGeneratorMethods, "Array");
- rb_define_method(mArray, "to_json", mArray_to_json, -1);
-#ifdef RUBY_INTEGER_UNIFICATION
- mInteger = rb_define_module_under(mGeneratorMethods, "Integer");
- rb_define_method(mInteger, "to_json", mInteger_to_json, -1);
-#else
- mFixnum = rb_define_module_under(mGeneratorMethods, "Fixnum");
- rb_define_method(mFixnum, "to_json", mFixnum_to_json, -1);
- mBignum = rb_define_module_under(mGeneratorMethods, "Bignum");
- rb_define_method(mBignum, "to_json", mBignum_to_json, -1);
-#endif
- mFloat = rb_define_module_under(mGeneratorMethods, "Float");
- rb_define_method(mFloat, "to_json", mFloat_to_json, -1);
- mString = rb_define_module_under(mGeneratorMethods, "String");
- rb_define_singleton_method(mString, "included", mString_included_s, 1);
- rb_define_method(mString, "to_json", mString_to_json, -1);
- rb_define_method(mString, "to_json_raw", mString_to_json_raw, -1);
- rb_define_method(mString, "to_json_raw_object", mString_to_json_raw_object, 0);
- mString_Extend = rb_define_module_under(mString, "Extend");
- rb_define_method(mString_Extend, "json_create", mString_Extend_json_create, 1);
- mTrueClass = rb_define_module_under(mGeneratorMethods, "TrueClass");
- rb_define_method(mTrueClass, "to_json", mTrueClass_to_json, -1);
- mFalseClass = rb_define_module_under(mGeneratorMethods, "FalseClass");
- rb_define_method(mFalseClass, "to_json", mFalseClass_to_json, -1);
- mNilClass = rb_define_module_under(mGeneratorMethods, "NilClass");
- rb_define_method(mNilClass, "to_json", mNilClass_to_json, -1);
-
- CRegexp_MULTILINE = rb_const_get(rb_cRegexp, rb_intern("MULTILINE"));
+ rb_define_method(cState, "generate", cState_generate, -1);
+ rb_define_method(cState, "_generate_no_fallback", cState_generate_no_fallback, -1);
+
+ rb_define_private_method(cState, "allow_duplicate_key?", cState_allow_duplicate_key_p, 0);
+
+ rb_define_singleton_method(cState, "generate", cState_m_generate, 3);
+ rb_define_singleton_method(cState, "_generate_no_fallback", cState_m_generate_no_fallback, 3);
+
+ rb_global_variable(&Encoding_UTF_8);
+ Encoding_UTF_8 = rb_const_get(rb_path2class("Encoding"), rb_intern("UTF_8"));
+
i_to_s = rb_intern("to_s");
i_to_json = rb_intern("to_json");
i_new = rb_intern("new");
- i_indent = rb_intern("indent");
- i_space = rb_intern("space");
- i_space_before = rb_intern("space_before");
- i_object_nl = rb_intern("object_nl");
- i_array_nl = rb_intern("array_nl");
- i_max_nesting = rb_intern("max_nesting");
- i_allow_nan = rb_intern("allow_nan");
- i_ascii_only = rb_intern("ascii_only");
- i_depth = rb_intern("depth");
- i_buffer_initial_length = rb_intern("buffer_initial_length");
- i_pack = rb_intern("pack");
- i_unpack = rb_intern("unpack");
- i_create_id = rb_intern("create_id");
- i_extend = rb_intern("extend");
- i_key_p = rb_intern("key?");
- i_aref = rb_intern("[]");
- i_send = rb_intern("__send__");
- i_respond_to_p = rb_intern("respond_to?");
- i_match = rb_intern("match");
- i_keys = rb_intern("keys");
- i_dup = rb_intern("dup");
-#ifdef HAVE_RUBY_ENCODING_H
- CEncoding_UTF_8 = rb_funcall(rb_path2class("Encoding"), rb_intern("find"), 1, rb_str_new2("utf-8"));
- i_encoding = rb_intern("encoding");
i_encode = rb_intern("encode");
-#endif
- i_SAFE_STATE_PROTOTYPE = rb_intern("SAFE_STATE_PROTOTYPE");
- CJSON_SAFE_STATE_PROTOTYPE = Qnil;
+
+ sym_indent = ID2SYM(rb_intern("indent"));
+ sym_space = ID2SYM(rb_intern("space"));
+ sym_space_before = ID2SYM(rb_intern("space_before"));
+ sym_object_nl = ID2SYM(rb_intern("object_nl"));
+ sym_array_nl = ID2SYM(rb_intern("array_nl"));
+ sym_max_nesting = ID2SYM(rb_intern("max_nesting"));
+ sym_allow_nan = ID2SYM(rb_intern("allow_nan"));
+ sym_ascii_only = ID2SYM(rb_intern("ascii_only"));
+ sym_depth = ID2SYM(rb_intern("depth"));
+ sym_buffer_initial_length = ID2SYM(rb_intern("buffer_initial_length"));
+ sym_script_safe = ID2SYM(rb_intern("script_safe"));
+ sym_escape_slash = ID2SYM(rb_intern("escape_slash"));
+ sym_strict = ID2SYM(rb_intern("strict"));
+ sym_as_json = ID2SYM(rb_intern("as_json"));
+ sym_allow_duplicate_key = ID2SYM(rb_intern("allow_duplicate_key"));
+
+ usascii_encindex = rb_usascii_encindex();
+ utf8_encindex = rb_utf8_encindex();
+ binary_encindex = rb_ascii8bit_encindex();
+
+ rb_require("json/ext/generator/state");
+
+ simd_impl = find_simd_implementation();
}
diff --git a/ext/json/generator/generator.h b/ext/json/generator/generator.h
deleted file mode 100644
index c367a6209a..0000000000
--- a/ext/json/generator/generator.h
+++ /dev/null
@@ -1,171 +0,0 @@
-#ifndef _GENERATOR_H_
-#define _GENERATOR_H_
-
-#include <math.h>
-#include <ctype.h>
-
-#include "ruby.h"
-
-#ifdef HAVE_RUBY_RE_H
-#include "ruby/re.h"
-#else
-#include "re.h"
-#endif
-
-#ifndef rb_intern_str
-#define rb_intern_str(string) SYM2ID(rb_str_intern(string))
-#endif
-
-#ifndef rb_obj_instance_variables
-#define rb_obj_instance_variables(object) rb_funcall(object, rb_intern("instance_variables"), 0)
-#endif
-
-#define option_given_p(opts, key) RTEST(rb_funcall(opts, i_key_p, 1, key))
-
-/* unicode definitions */
-
-#define UNI_STRICT_CONVERSION 1
-
-typedef unsigned long UTF32; /* at least 32 bits */
-typedef unsigned short UTF16; /* at least 16 bits */
-typedef unsigned char UTF8; /* typically 8 bits */
-
-#define UNI_REPLACEMENT_CHAR (UTF32)0x0000FFFD
-#define UNI_MAX_BMP (UTF32)0x0000FFFF
-#define UNI_MAX_UTF16 (UTF32)0x0010FFFF
-#define UNI_MAX_UTF32 (UTF32)0x7FFFFFFF
-#define UNI_MAX_LEGAL_UTF32 (UTF32)0x0010FFFF
-
-#define UNI_SUR_HIGH_START (UTF32)0xD800
-#define UNI_SUR_HIGH_END (UTF32)0xDBFF
-#define UNI_SUR_LOW_START (UTF32)0xDC00
-#define UNI_SUR_LOW_END (UTF32)0xDFFF
-
-static const int halfShift = 10; /* used for shifting by 10 bits */
-
-static const UTF32 halfBase = 0x0010000UL;
-static const UTF32 halfMask = 0x3FFUL;
-
-static unsigned char isLegalUTF8(const UTF8 *source, unsigned long length);
-static void unicode_escape(char *buf, UTF16 character);
-static void unicode_escape_to_buffer(FBuffer *buffer, char buf[6], UTF16 character);
-static void convert_UTF8_to_JSON_ASCII(FBuffer *buffer, VALUE string);
-static void convert_UTF8_to_JSON(FBuffer *buffer, VALUE string);
-static char *fstrndup(const char *ptr, unsigned long len);
-
-/* ruby api and some helpers */
-
-typedef struct JSON_Generator_StateStruct {
- char *indent;
- long indent_len;
- char *space;
- long space_len;
- char *space_before;
- long space_before_len;
- char *object_nl;
- long object_nl_len;
- char *array_nl;
- long array_nl_len;
- FBuffer *array_delim;
- FBuffer *object_delim;
- FBuffer *object_delim2;
- long max_nesting;
- char allow_nan;
- char ascii_only;
- long depth;
- long buffer_initial_length;
-} JSON_Generator_State;
-
-#define GET_STATE_TO(self, state) \
- TypedData_Get_Struct(self, JSON_Generator_State, &JSON_Generator_State_type, state)
-
-#define GET_STATE(self) \
- JSON_Generator_State *state; \
- GET_STATE_TO(self, state)
-
-#define GENERATE_JSON(type) \
- FBuffer *buffer; \
- VALUE Vstate; \
- JSON_Generator_State *state; \
- \
- rb_scan_args(argc, argv, "01", &Vstate); \
- Vstate = cState_from_state_s(cState, Vstate); \
- TypedData_Get_Struct(Vstate, JSON_Generator_State, &JSON_Generator_State_type, state); \
- buffer = cState_prepare_buffer(Vstate); \
- generate_json_##type(buffer, Vstate, state, self); \
- return fbuffer_to_s(buffer)
-
-static VALUE mHash_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mArray_to_json(int argc, VALUE *argv, VALUE self);
-#ifdef RUBY_INTEGER_UNIFICATION
-static VALUE mInteger_to_json(int argc, VALUE *argv, VALUE self);
-#else
-static VALUE mFixnum_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mBignum_to_json(int argc, VALUE *argv, VALUE self);
-#endif
-static VALUE mFloat_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mString_included_s(VALUE self, VALUE modul);
-static VALUE mString_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mString_to_json_raw_object(VALUE self);
-static VALUE mString_to_json_raw(int argc, VALUE *argv, VALUE self);
-static VALUE mString_Extend_json_create(VALUE self, VALUE o);
-static VALUE mTrueClass_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mFalseClass_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mNilClass_to_json(int argc, VALUE *argv, VALUE self);
-static VALUE mObject_to_json(int argc, VALUE *argv, VALUE self);
-static void State_free(void *state);
-static VALUE cState_s_allocate(VALUE klass);
-static VALUE cState_configure(VALUE self, VALUE opts);
-static VALUE cState_to_h(VALUE self);
-static void generate_json(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_object(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_array(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_string(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_null(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_false(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_true(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-#ifdef RUBY_INTEGER_UNIFICATION
-static void generate_json_integer(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-#endif
-static void generate_json_fixnum(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_bignum(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static void generate_json_float(FBuffer *buffer, VALUE Vstate, JSON_Generator_State *state, VALUE obj);
-static VALUE cState_partial_generate(VALUE self, VALUE obj);
-static VALUE cState_generate(VALUE self, VALUE obj);
-static VALUE cState_initialize(int argc, VALUE *argv, VALUE self);
-static VALUE cState_from_state_s(VALUE self, VALUE opts);
-static VALUE cState_indent(VALUE self);
-static VALUE cState_indent_set(VALUE self, VALUE indent);
-static VALUE cState_space(VALUE self);
-static VALUE cState_space_set(VALUE self, VALUE space);
-static VALUE cState_space_before(VALUE self);
-static VALUE cState_space_before_set(VALUE self, VALUE space_before);
-static VALUE cState_object_nl(VALUE self);
-static VALUE cState_object_nl_set(VALUE self, VALUE object_nl);
-static VALUE cState_array_nl(VALUE self);
-static VALUE cState_array_nl_set(VALUE self, VALUE array_nl);
-static VALUE cState_max_nesting(VALUE self);
-static VALUE cState_max_nesting_set(VALUE self, VALUE depth);
-static VALUE cState_allow_nan_p(VALUE self);
-static VALUE cState_ascii_only_p(VALUE self);
-static VALUE cState_depth(VALUE self);
-static VALUE cState_depth_set(VALUE self, VALUE depth);
-static FBuffer *cState_prepare_buffer(VALUE self);
-#ifndef ZALLOC
-#define ZALLOC(type) ((type *)ruby_zalloc(sizeof(type)))
-static inline void *ruby_zalloc(size_t n)
-{
- void *p = ruby_xmalloc(n);
- memset(p, 0, n);
- return p;
-}
-#endif
-#ifdef TypedData_Make_Struct
-static const rb_data_type_t JSON_Generator_State_type;
-#define NEW_TYPEDDATA_WRAPPER 1
-#else
-#define TypedData_Make_Struct(klass, type, ignore, json) Data_Make_Struct(klass, type, NULL, State_free, json)
-#define TypedData_Get_Struct(self, JSON_Generator_State, ignore, json) Data_Get_Struct(self, JSON_Generator_State, json)
-#endif
-
-#endif
generated by cgit v1.2.3 (git 2.25.1) at 2026-06-04 20:52:08 +0000