nano-safe-gem 0.0.1

data/oj-3.17.3/ext/oj/oj.c

@@ -0,0 +1,2332 @@
+// Copyright (c) 2012 Peter Ohler. All rights reserved.
+// Licensed under the MIT License. See LICENSE file in the project root for license details.
+
+#include "oj.h"
+
+#include <errno.h>
+#include <fcntl.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "dump.h"
+#include "encode.h"
+#include "intern.h"
+#include "mem.h"
+#include "odd.h"
+#include "parse.h"
+#include "rails.h"
+#include "simd.h"
+
+#define MAX_INDENT 16
+
+typedef struct _yesNoOpt {
+    VALUE sym;
+    char *attr;
+} *YesNoOpt;
+
+void Init_oj();
+
+VALUE Oj = Qnil;
+
+ID oj_add_value_id;
+ID oj_array_append_id;
+ID oj_array_end_id;
+ID oj_array_start_id;
+ID oj_as_json_id;
+ID oj_begin_id;
+ID oj_bigdecimal_id;
+ID oj_end_id;
+ID oj_eofq_id;
+ID oj_exclude_end_id;
+ID oj_error_id;
+ID oj_file_id;
+ID oj_fileno_id;
+ID oj_ftype_id;
+ID oj_hash_end_id;
+ID oj_hash_key_id;
+ID oj_hash_set_id;
+ID oj_hash_start_id;
+ID oj_iconv_id;
+ID oj_json_create_id;
+ID oj_length_id;
+ID oj_new_id;
+ID oj_parse_id;
+ID oj_plus_id;
+ID oj_pos_id;
+ID oj_raw_json_id;
+ID oj_read_id;
+ID oj_readpartial_id;
+ID oj_replace_id;
+ID oj_stat_id;
+ID oj_string_id;
+ID oj_to_h_id;
+ID oj_to_hash_id;
+ID oj_to_json_id;
+ID oj_to_s_id;
+ID oj_to_sym_id;
+ID oj_to_time_id;
+ID oj_tv_nsec_id;
+ID oj_tv_sec_id;
+ID oj_tv_usec_id;
+ID oj_utc_id;
+ID oj_utc_offset_id;
+ID oj_utcq_id;
+ID oj_write_id;
+
+VALUE oj_bag_class;
+VALUE oj_bigdecimal_class;
+VALUE oj_cstack_class;
+VALUE oj_date_class;
+VALUE oj_datetime_class;
+VALUE oj_enumerable_class;
+VALUE oj_parse_error_class;
+VALUE oj_stream_writer_class;
+VALUE oj_string_writer_class;
+VALUE oj_stringio_class;
+VALUE oj_struct_class;
+
+VALUE oj_slash_string;
+
+VALUE oj_allow_nan_sym;
+VALUE oj_array_class_sym;
+VALUE oj_create_additions_sym;
+VALUE oj_decimal_class_sym;
+VALUE oj_hash_class_sym;
+VALUE oj_indent_sym;
+VALUE oj_object_class_sym;
+VALUE oj_quirks_mode_sym;
+VALUE oj_safe_sym;
+VALUE oj_symbolize_names_sym;
+VALUE oj_trace_sym;
+
+static VALUE allow_blank_sym;
+static VALUE allow_gc_sym;
+static VALUE allow_invalid_unicode_sym;
+static VALUE ascii_sym;
+static VALUE auto_define_sym;
+static VALUE auto_sym;
+static VALUE bigdecimal_as_decimal_sym;
+static VALUE bigdecimal_load_sym;
+static VALUE bigdecimal_sym;
+static VALUE cache_keys_sym;
+static VALUE cache_str_sym;
+static VALUE cache_string_sym;
+static VALUE circular_sym;
+static VALUE class_cache_sym;
+static VALUE compat_bigdecimal_sym;
+static VALUE compat_sym;
+static VALUE create_id_sym;
+static VALUE custom_sym;
+static VALUE empty_string_sym;
+static VALUE escape_mode_sym;
+static VALUE except_sym;
+static VALUE integer_range_sym;
+static VALUE max_integer_digits_sym;
+static VALUE fast_sym;
+static VALUE float_prec_sym;
+static VALUE float_format_sym;
+static VALUE float_sym;
+static VALUE huge_sym;
+static VALUE ignore_sym;
+static VALUE ignore_under_sym;
+static VALUE json_sym;
+static VALUE match_string_sym;
+static VALUE mode_sym;
+static VALUE nan_sym;
+static VALUE newline_sym;
+static VALUE nilnil_sym;
+static VALUE null_sym;
+static VALUE object_sym;
+static VALUE omit_null_byte_sym;
+static VALUE omit_nil_sym;
+static VALUE only_sym;
+static VALUE rails_sym;
+static VALUE raise_sym;
+static VALUE ruby_sym;
+static VALUE sec_prec_sym;
+static VALUE slash_sym;
+static VALUE strict_sym;
+static VALUE symbol_keys_sym;
+static VALUE time_format_sym;
+static VALUE unicode_xss_sym;
+static VALUE unix_sym;
+static VALUE unix_zone_sym;
+static VALUE use_as_json_sym;
+static VALUE use_raw_json_sym;
+static VALUE use_to_hash_sym;
+static VALUE use_to_json_sym;
+static VALUE wab_sym;
+static VALUE word_sym;
+static VALUE xmlschema_sym;
+static VALUE xss_safe_sym;
+
+rb_encoding *oj_utf8_encoding       = 0;
+int          oj_utf8_encoding_index = 0;
+
+#ifdef HAVE_PTHREAD_MUTEX_INIT
+pthread_mutex_t oj_cache_mutex;
+#else
+VALUE oj_cache_mutex = Qnil;
+#endif
+
+SIMD_Implementation SIMD_Impl = SIMD_NONE;
+
+extern void oj_parser_init();
+
+const char oj_json_class[] = "json_class";
+
+struct _options oj_default_options = {
+    0,              // indent
+    No,             // circular
+    No,             // auto_define
+    No,             // sym_key
+    JSONEsc,        // escape_mode
+    ObjectMode,     // mode
+    Yes,            // class_cache
+    UnixTime,       // time_format
+    NotSet,         // bigdec_as_num
+    AutoDec,        // bigdec_load
+    false,          // compat_bigdec
+    No,             // to_hash
+    No,             // to_json
+    No,             // as_json
+    No,             // raw_json
+    No,             // nilnil
+    Yes,            // empty_string
+    Yes,            // allow_gc
+    Yes,            // quirks_mode
+    No,             // allow_invalid
+    No,             // create_ok
+    Yes,            // allow_nan
+    No,             // trace
+    No,             // safe
+    false,          // sec_prec_set
+    No,             // ignore_under
+    Yes,            // cache_keys
+    0,              // cache_str
+    0,              // int_range_min
+    0,              // int_range_max
+    0,              // max_integer_digits
+    oj_json_class,  // create_id
+    10,             // create_id_len
+    9,              // sec_prec
+    16,             // float_prec
+    "%0.15g",       // float_fmt
+    Qnil,           // hash_class
+    Qnil,           // array_class
+    {
+        // dump_opts
+        false,      // use
+        "",         // indent
+        "",         // before_sep
+        "",         // after_sep
+        "",         // hash_nl
+        "",         // array_nl
+        0,          // indent_size
+        0,          // before_size
+        0,          // after_size
+        0,          // hash_size
+        0,          // array_size
+        AutoNan,    // nan_dump
+        false,      // omit_nil
+        false,      // omit_null_byte
+        MAX_DEPTH,  // max_depth
+        NULL,       // only
+        NULL,       // except
+    },
+    {
+        // str_rx
+        NULL,    // head
+        NULL,    // tail
+        {'\0'},  // err
+    },
+    NULL,
+};
+
+static VALUE only_array_from_string(const char *str) {
+    volatile VALUE a = Qnil;
+
+    if (NULL != str && 2 < strlen(str)) {
+        str++;
+        char *cp;
+
+        a = rb_ary_new();
+        while (NULL != (cp = strchr(str, ':'))) {
+            rb_ary_push(a, rb_id2sym(rb_intern2(str, cp - str)));
+            str = cp + 1;
+        }
+    }
+    return a;
+}
+
+/* Document-method: default_options()
+ *	call-seq: default_options()
+ *
+ * Returns the default load and dump options as a Hash. The options are
+ * - *:indent* [_Fixnum_|_String_|_nil_] number of spaces to indent each element
+ *   in an JSON document, zero or nil is no newline between JSON elements,
+ *   negative indicates no newline between top level JSON elements in a stream,
+ *   a String indicates the string should be used for indentation
+ * - *:circular* [_Boolean_|_nil_] support circular references while dumping as
+ *   well as shared references
+ * - *:auto_define* [_Boolean_|_nil_] automatically define classes if they do not exist
+ * - *:symbol_keys* [_Boolean_|_nil_] use symbols instead of strings for hash keys
+ * - *:escape_mode* [_:newline_|_:json_|_:slash_|_:xss_safe_|_:ascii_|_:unicode_xss_|_nil_]
+ *   determines the characters to escape
+ * - *:class_cache* [_Boolean_|_nil_] cache classes for faster parsing (if dynamically
+ *   modifying classes or reloading classes then don't use this)
+ * - *:mode* [_:object_|_:strict_|_:compat_|_:null_|_:custom_|_:rails_|_:wab_] load and
+ *   dump modes to use for JSON
+ * - *:time_format* [_:unix_|_:unix_zone_|_:xmlschema_|_:ruby_] time format when dumping
+ * - *:bigdecimal_as_decimal* [_Boolean_|_nil_] dump BigDecimal as a decimal number or
+ *   as a String
+ * - *:bigdecimal_load* [_:bigdecimal_|_:float_|_:auto_|_:fast_|_:ruby_] load decimals
+ *   as BigDecimal instead of as a Float. :auto pick the most precise for the number
+ *   of digits. :float should be the same as ruby. :fast may require rounding but is
+ *   must faster.
+ * - *:compat_bigdecimal* [_true_|_false_] load decimals as BigDecimal instead of as
+ *   a Float when in compat or rails mode.
+ * - *:create_id* [_String_|_nil_] create id for json compatible object encoding,
+ *   default is 'json_class'
+ * - *:create_additions* [_Boolean_|_nil_] if true allow creation of instances using
+ *   create_id on load.
+ * - *:second_precision* [_Fixnum_|_nil_] number of digits after the decimal when
+ *   dumping the seconds portion of time
+ * - *:float_precision* [_Fixnum_|_nil_] number of digits of precision when dumping
+ *   floats, 0 indicates use Ruby
+ * - *:float_format* [_String_] the C printf format string for printing floats.
+ *   Default follows the float_precision and will be changed if float_precision is
+ *   changed. The string can be no more than 6 bytes.
+ * - *:use_to_json* [_Boolean_|_nil_] call to_json() methods on dump, default is false
+ * - *:use_as_json* [_Boolean_|_nil_] call as_json() methods on dump, default is false
+ * - *:use_raw_json* [_Boolean_|_nil_] call raw_json() methods on dump, default is false
+ * - *:nilnil* [_Boolean_|_nil_] if true a nil input to load will return nil and
+ *   not raise an Exception
+ * - *:empty_string* [_Boolean_|_nil_] if true an empty input will not raise an Exception
+ * - *:allow_gc* [_Boolean_|_nil_] allow or prohibit GC during parsing, default is true (allow)
+ * - *:quirks_mode* [_true,_|_false_|_nil_] Allow single JSON values instead of
+ *   documents, default is true (allow)
+ * - *:allow_invalid_unicode* [_true,_|_false_|_nil_] Allow invalid unicode,
+ *   default is false (don't allow)
+ * - *:allow_nan* [_true,_|_false_|_nil_] Allow Nan, Infinity, and -Infinity to
+ *   be parsed, default is true (allow)
+ * - *:indent_str* [_String_|_nil_] String to use for indentation, overriding the
+ *   indent option is not nil
+ * - *:space* [_String_|_nil_] String to use for the space after the colon in JSON
+ *   object fields
+ * - *:space_before* [_String_|_nil_] String to use before the colon separator in
+ *   JSON object fields
+ * - *:object_nl* [_String_|_nil_] String to use after a JSON object field value
+ * - *:array_nl* [_String_|_nil_] String to use after a JSON array value
+ * - *:nan* [_:null_|_:huge_|_:word_|_:raise_|_:auto_] how to dump Infinity and
+ *   NaN. :null places a null, :huge places a huge number, :word places Infinity
+ *   or NaN, :raise raises and exception, :auto uses default for each mode.
+ * - *:hash_class* [_Class_|_nil_] Class to use instead of Hash on load,
+ *   :object_class can also be used
+ * - *:array_class* [_Class_|_nil_] Class to use instead of Array on load
+ * - *:omit_nil* [_true_|_false_] if true Hash and Object attributes with nil
+ *   values are omitted
+ * - *:omit_null_byte* [_true_|_false_] if true null bytes in strings will be
+ *   omitted when dumping
+ * - *:ignore* [_nil_|_Array_] either nil or an Array of classes to ignore when dumping
+ * - *:ignore_under* [_Boolean_] if true then attributes that start with _ are
+ *   ignored when dumping in object or custom mode.
+ * - *:cache_keys* [_Boolean_] if true then hash keys are cached if less than 35 bytes.
+ * - *:cache_str* [_Fixnum_] maximum string value length to cache (strings less
+ *   than this are cached)
+ * - *:integer_range* [_Range_] Dump integers outside range as strings.
+ * - *:max_integer_digits* [_Fixnum_] Maximum number of decimal digits allowed in a
+ *   parsed integer. When the limit is exceeded a parse error is raised. 0 (the
+ *   default) disables the limit. Setting a reasonable limit is recommended when
+ *   parsing untrusted input to mitigate CPU-DoS attacks. Only applies to the
+ *   legacy parsers (Oj.load, Oj::Doc, JSON.parse mimic); Oj::Parser is unaffected.
+ * - *:trace* [_true,_|_false_] Trace all load and dump calls, default is false
+ *   (trace is off)
+ * - *:safe* [_true,_|_false_] Safe mimic breaks JSON mimic to be safer, default
+ *   is false (safe is off)
+ * - *:only* [_nil,_|_Array_] A list of the fields to encode. All others are skipped.
+ * - *:except* [_nil,_|_Array_] A list of the fields to not encode. All others are encoded.
+ *
+ * Return [_Hash_] all current option settings.
+ */
+static VALUE get_def_opts(VALUE self) {
+    VALUE opts = rb_hash_new();
+
+    if (0 == oj_default_options.dump_opts.indent_size) {
+        rb_hash_aset(opts, oj_indent_sym, INT2FIX(oj_default_options.indent));
+    } else {
+        rb_hash_aset(opts, oj_indent_sym, rb_str_new2(oj_default_options.dump_opts.indent_str));
+    }
+    rb_hash_aset(opts, sec_prec_sym, INT2FIX(oj_default_options.sec_prec));
+    rb_hash_aset(opts,
+                 circular_sym,
+                 (Yes == oj_default_options.circular) ? Qtrue : ((No == oj_default_options.circular) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        class_cache_sym,
+        (Yes == oj_default_options.class_cache) ? Qtrue : ((No == oj_default_options.class_cache) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        auto_define_sym,
+        (Yes == oj_default_options.auto_define) ? Qtrue : ((No == oj_default_options.auto_define) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 symbol_keys_sym,
+                 (Yes == oj_default_options.sym_key) ? Qtrue : ((No == oj_default_options.sym_key) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        bigdecimal_as_decimal_sym,
+        (Yes == oj_default_options.bigdec_as_num) ? Qtrue : ((No == oj_default_options.bigdec_as_num) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        oj_create_additions_sym,
+        (Yes == oj_default_options.create_ok) ? Qtrue : ((No == oj_default_options.create_ok) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 use_to_json_sym,
+                 (Yes == oj_default_options.to_json) ? Qtrue : ((No == oj_default_options.to_json) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 use_to_hash_sym,
+                 (Yes == oj_default_options.to_hash) ? Qtrue : ((No == oj_default_options.to_hash) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 use_as_json_sym,
+                 (Yes == oj_default_options.as_json) ? Qtrue : ((No == oj_default_options.as_json) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 use_raw_json_sym,
+                 (Yes == oj_default_options.raw_json) ? Qtrue : ((No == oj_default_options.raw_json) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 nilnil_sym,
+                 (Yes == oj_default_options.nilnil) ? Qtrue : ((No == oj_default_options.nilnil) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        empty_string_sym,
+        (Yes == oj_default_options.empty_string) ? Qtrue : ((No == oj_default_options.empty_string) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 allow_gc_sym,
+                 (Yes == oj_default_options.allow_gc) ? Qtrue : ((No == oj_default_options.allow_gc) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        oj_quirks_mode_sym,
+        (Yes == oj_default_options.quirks_mode) ? Qtrue : ((No == oj_default_options.quirks_mode) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        allow_invalid_unicode_sym,
+        (Yes == oj_default_options.allow_invalid) ? Qtrue : ((No == oj_default_options.allow_invalid) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        oj_allow_nan_sym,
+        (Yes == oj_default_options.allow_nan) ? Qtrue : ((No == oj_default_options.allow_nan) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 oj_trace_sym,
+                 (Yes == oj_default_options.trace) ? Qtrue : ((No == oj_default_options.trace) ? Qfalse : Qnil));
+    rb_hash_aset(opts,
+                 oj_safe_sym,
+                 (Yes == oj_default_options.safe) ? Qtrue : ((No == oj_default_options.safe) ? Qfalse : Qnil));
+    rb_hash_aset(opts, float_prec_sym, INT2FIX(oj_default_options.float_prec));
+    rb_hash_aset(opts, float_format_sym, rb_str_new_cstr(oj_default_options.float_fmt));
+    rb_hash_aset(opts, cache_str_sym, INT2FIX(oj_default_options.cache_str));
+    rb_hash_aset(
+        opts,
+        ignore_under_sym,
+        (Yes == oj_default_options.ignore_under) ? Qtrue : ((No == oj_default_options.ignore_under) ? Qfalse : Qnil));
+    rb_hash_aset(
+        opts,
+        cache_keys_sym,
+        (Yes == oj_default_options.cache_keys) ? Qtrue : ((No == oj_default_options.cache_keys) ? Qfalse : Qnil));
+
+    switch (oj_default_options.mode) {
+    case StrictMode: rb_hash_aset(opts, mode_sym, strict_sym); break;
+    case CompatMode: rb_hash_aset(opts, mode_sym, compat_sym); break;
+    case NullMode: rb_hash_aset(opts, mode_sym, null_sym); break;
+    case ObjectMode: rb_hash_aset(opts, mode_sym, object_sym); break;
+    case CustomMode: rb_hash_aset(opts, mode_sym, custom_sym); break;
+    case RailsMode: rb_hash_aset(opts, mode_sym, rails_sym); break;
+    case WabMode: rb_hash_aset(opts, mode_sym, wab_sym); break;
+    default: rb_hash_aset(opts, mode_sym, object_sym); break;
+    }
+
+    if (oj_default_options.int_range_max != 0 || oj_default_options.int_range_min != 0) {
+        VALUE range = rb_obj_alloc(rb_cRange);
+        VALUE min   = LONG2FIX(oj_default_options.int_range_min);
+        VALUE max   = LONG2FIX(oj_default_options.int_range_max);
+
+        rb_ivar_set(range, oj_begin_id, min);
+        rb_ivar_set(range, oj_end_id, max);
+        rb_hash_aset(opts, integer_range_sym, range);
+    } else {
+        rb_hash_aset(opts, integer_range_sym, Qnil);
+    }
+    rb_hash_aset(opts, max_integer_digits_sym, LONG2NUM((long)oj_default_options.max_integer_digits));
+    switch (oj_default_options.escape_mode) {
+    case NLEsc: rb_hash_aset(opts, escape_mode_sym, newline_sym); break;
+    case JSONEsc: rb_hash_aset(opts, escape_mode_sym, json_sym); break;
+    case SlashEsc: rb_hash_aset(opts, escape_mode_sym, slash_sym); break;
+    case XSSEsc: rb_hash_aset(opts, escape_mode_sym, xss_safe_sym); break;
+    case ASCIIEsc: rb_hash_aset(opts, escape_mode_sym, ascii_sym); break;
+    case JXEsc: rb_hash_aset(opts, escape_mode_sym, unicode_xss_sym); break;
+    default: rb_hash_aset(opts, escape_mode_sym, json_sym); break;
+    }
+    switch (oj_default_options.time_format) {
+    case XmlTime: rb_hash_aset(opts, time_format_sym, xmlschema_sym); break;
+    case RubyTime: rb_hash_aset(opts, time_format_sym, ruby_sym); break;
+    case UnixZTime: rb_hash_aset(opts, time_format_sym, unix_zone_sym); break;
+    case UnixTime:
+    default: rb_hash_aset(opts, time_format_sym, unix_sym); break;
+    }
+    switch (oj_default_options.bigdec_load) {
+    case BigDec: rb_hash_aset(opts, bigdecimal_load_sym, bigdecimal_sym); break;
+    case FloatDec: rb_hash_aset(opts, bigdecimal_load_sym, float_sym); break;
+    case FastDec: rb_hash_aset(opts, bigdecimal_load_sym, fast_sym); break;
+    case AutoDec:
+    default: rb_hash_aset(opts, bigdecimal_load_sym, auto_sym); break;
+    }
+    rb_hash_aset(opts, compat_bigdecimal_sym, oj_default_options.compat_bigdec ? Qtrue : Qfalse);
+    rb_hash_aset(opts,
+                 create_id_sym,
+                 (NULL == oj_default_options.create_id) ? Qnil : rb_str_new2(oj_default_options.create_id));
+    rb_hash_aset(
+        opts,
+        oj_space_sym,
+        (0 == oj_default_options.dump_opts.after_size) ? Qnil : rb_str_new2(oj_default_options.dump_opts.after_sep));
+    rb_hash_aset(
+        opts,
+        oj_space_before_sym,
+        (0 == oj_default_options.dump_opts.before_size) ? Qnil : rb_str_new2(oj_default_options.dump_opts.before_sep));
+    rb_hash_aset(
+        opts,
+        oj_object_nl_sym,
+        (0 == oj_default_options.dump_opts.hash_size) ? Qnil : rb_str_new2(oj_default_options.dump_opts.hash_nl));
+    rb_hash_aset(
+        opts,
+        oj_array_nl_sym,
+        (0 == oj_default_options.dump_opts.array_size) ? Qnil : rb_str_new2(oj_default_options.dump_opts.array_nl));
+
+    switch (oj_default_options.dump_opts.nan_dump) {
+    case NullNan: rb_hash_aset(opts, nan_sym, null_sym); break;
+    case RaiseNan: rb_hash_aset(opts, nan_sym, raise_sym); break;
+    case WordNan: rb_hash_aset(opts, nan_sym, word_sym); break;
+    case HugeNan: rb_hash_aset(opts, nan_sym, huge_sym); break;
+    case AutoNan:
+    default: rb_hash_aset(opts, nan_sym, auto_sym); break;
+    }
+    rb_hash_aset(opts, omit_nil_sym, oj_default_options.dump_opts.omit_nil ? Qtrue : Qfalse);
+    rb_hash_aset(opts, omit_null_byte_sym, oj_default_options.dump_opts.omit_null_byte ? Qtrue : Qfalse);
+    rb_hash_aset(opts, oj_hash_class_sym, oj_default_options.hash_class);
+    rb_hash_aset(opts, oj_array_class_sym, oj_default_options.array_class);
+
+    rb_hash_aset(opts, only_sym, only_array_from_string(oj_default_options.dump_opts.only));
+    rb_hash_aset(opts, except_sym, only_array_from_string(oj_default_options.dump_opts.except));
+
+    if (NULL == oj_default_options.ignore) {
+        rb_hash_aset(opts, ignore_sym, Qnil);
+    } else {
+        VALUE         *vp;
+        volatile VALUE a = rb_ary_new();
+
+        for (vp = oj_default_options.ignore; Qnil != *vp; vp++) {
+            rb_ary_push(a, *vp);
+        }
+        rb_hash_aset(opts, ignore_sym, a);
+    }
+    return opts;
+}
+
+/* Document-method: default_options=
+ *	call-seq: default_options=(opts)
+ *
+ * Sets the default options for load and dump.
+ * - *opts* [_Hash_] options to change
+ *   - *:indent* [_Fixnum_|_String_|_nil_] number of spaces to indent each element
+ *     in a JSON document or the String to use for indentation.
+ *   - :circular [_Boolean_|_nil_] support circular references while dumping.
+ *   - *:auto_define* [_Boolean_|_nil_] automatically define classes if they do not exist.
+ *   - *:symbol_keys* [_Boolean_|_nil_] convert hash keys to symbols.
+ *   - *:class_cache* [_Boolean_|_nil_] cache classes for faster parsing.
+ *   - *:escape* [_:newline_|_:json_|_:xss_safe_|_:ascii_|_unicode_xss_|_nil_]
+ *     mode encodes all high-bit characters as escaped sequences if :ascii, :json
+ *     is standand UTF-8 JSON encoding, :newline is the same as :json but newlines
+ *     are not escaped, :unicode_xss allows unicode but escapes &, <, and >, and
+ *     any \u20xx characters along with some others, and :xss_safe escapes &, <,
+ *     and >, and some others.
+ *   - *:bigdecimal_as_decimal* [_Boolean_|_nil_] dump BigDecimal as a decimal
+ *     number or as a String.
+ *   - *:bigdecimal_load* [_:bigdecimal_|_:float_|_:auto_|_nil_] load decimals as
+ *     BigDecimal instead of as a Float. :auto pick the most precise for the number of digits.
+ *   - *:compat_bigdecimal* [_true_|_false_] load decimals as BigDecimal instead
+ *     of as a Float in compat mode.
+ *   - *:mode* [_:object_|_:strict_|_:compat_|_:null_|_:custom_|_:rails_|_:wab_] load
+ *     and dump mode to use for JSON :strict raises an exception when a non-supported
+ *     Object is encountered. :compat attempts to extract variable values from an
+ *     Object using to_json() or to_hash() then it walks the Object's variables if
+ *     neither is found. The :object mode ignores to_hash() and to_json() methods
+ *     and encodes variables using code internal to the Oj gem. The :null mode
+ *     ignores non-supported Objects and replaces them with a null. The :custom
+ *     mode honors all dump options. The :rails more mimics rails and Active behavior.
+ *   - *:time_format* [_:unix_|_:xmlschema_|_:ruby_] time format when dumping in :compat
+ *     mode :unix decimal number denoting the number of seconds since 1/1/1970,
+ *     :unix_zone decimal number denoting the number of seconds since 1/1/1970
+ *     plus the utc_offset in the exponent, :xmlschema date-time format taken
+ *     from XML Schema as a String, :ruby Time.to_s formatted String.
+ *   - *:create_id* [_String_|_nil_] create id for json compatible object encoding
+ *   - *:create_additions* [_Boolean_|_nil_] if true allow creation of instances using create_id on load.
+ *   - *:second_precision* [_Fixnum_|_nil_] number of digits after the decimal
+ *     when dumping the seconds portion of time.
+ *   - *:float_format* [_String_] the C printf format string for printing floats.
+ *     Default follows the float_precision and will be changed if float_precision
+ *     is changed. The string can be no more than 6 bytes.
+ *   - *:float_precision* [_Fixnum_|_nil_] number of digits of precision when dumping floats, 0 indicates use Ruby.
+ *   - *:use_to_json* [_Boolean_|_nil_] call to_json() methods on dump, default is false.
+ *   - *:use_as_json* [_Boolean_|_nil_] call as_json() methods on dump, default is false.
+ *   - *:use_to_hash* [_Boolean_|_nil_] call to_hash() methods on dump, default is false.
+ *   - *:use_raw_json* [_Boolean_|_nil_] call raw_json() methods on dump, default is false.
+ *   - *:nilnil* [_Boolean_|_nil_] if true a nil input to load will return nil and not raise an Exception.
+ *   - *:allow_gc* [_Boolean_|_nil_] allow or prohibit GC during parsing, default is true (allow).
+ *   - *:quirks_mode* [_Boolean_|_nil_] allow single JSON values instead of documents, default is true (allow).
+ *   - *:allow_invalid_unicode* [_Boolean_|_nil_] allow invalid unicode, default is false (don't allow).
+ *   - *:allow_nan* [_Boolean_|_nil_] allow Nan, Infinity, and -Infinity, default is true (allow).
+ *   - *:space* [_String_|_nil_] String to use for the space after the colon in JSON object fields.
+ *   - *:space_before* [_String_|_nil_] String to use before the colon separator in JSON object fields.
+ *   - *:object_nl* [_String_|_nil_] String to use after a JSON object field value.
+ *   - *:array_nl* [_String_|_nil_] String to use after a JSON array value
+ *   - *:nan* [_:null_|_:huge_|_:word_|_:raise_] how to dump Infinity and NaN in null,
+ *     strict, and compat mode. :null places a null, :huge places a huge number, :word
+ *     places Infinity or NaN, :raise raises and exception, :auto uses default for each mode.
+ *   - *:hash_class* [_Class_|_nil_] Class to use instead of Hash on load, :object_class can also be used.
+ *   - *:array_class* [_Class_|_nil_] Class to use instead of Array on load.
+ *   - *:omit_nil* [_true_|_false_] if true Hash and Object attributes with nil values are omitted.
+ *   - *:ignore* [_nil_|Array] either nil or an Array of classes to ignore when dumping
+ *   - *:ignore_under* [_Boolean_] if true then attributes that start with _ are
+ *     ignored when dumping in object or custom mode.
+ *   - *:cache_keys* [_Boolean_] if true then hash keys are cached
+ *   - *:cache_str* [_Fixnum_] maximum string value length to cache (strings less than this are cached)
+ *   - *:integer_range* [_Range_] Dump integers outside range as strings.
+ *   - *:max_integer_digits* [_Fixnum_] Maximum decimal digits in a parsed integer
+ *     (0 = unlimited). Use to mitigate CPU-DoS via huge integer values in JSON.
+ *   - *:trace* [_Boolean_] turn trace on or off.
+ *   - *:safe* [_Boolean_] turn safe mimic on or off.
+ */
+static VALUE set_def_opts(VALUE self, VALUE opts) {
+    Check_Type(opts, T_HASH);
+    oj_parse_options(opts, &oj_default_options);
+
+    return Qnil;
+}
+
+bool oj_hash_has_key(VALUE hash, VALUE key) {
+    if (Qundef == rb_hash_lookup2(hash, key, Qundef)) {
+        return false;
+    }
+    return true;
+}
+
+bool set_yesno_options(VALUE key, VALUE value, Options copts) {
+    struct _yesNoOpt ynos[] = {{circular_sym, &copts->circular},
+                               {auto_define_sym, &copts->auto_define},
+                               {symbol_keys_sym, &copts->sym_key},
+                               {class_cache_sym, &copts->class_cache},
+                               {bigdecimal_as_decimal_sym, &copts->bigdec_as_num},
+                               {use_to_hash_sym, &copts->to_hash},
+                               {use_to_json_sym, &copts->to_json},
+                               {use_as_json_sym, &copts->as_json},
+                               {use_raw_json_sym, &copts->raw_json},
+                               {nilnil_sym, &copts->nilnil},
+                               {allow_blank_sym, &copts->nilnil},  // same as nilnil
+                               {empty_string_sym, &copts->empty_string},
+                               {allow_gc_sym, &copts->allow_gc},
+                               {oj_quirks_mode_sym, &copts->quirks_mode},
+                               {allow_invalid_unicode_sym, &copts->allow_invalid},
+                               {oj_allow_nan_sym, &copts->allow_nan},
+                               {oj_trace_sym, &copts->trace},
+                               {oj_safe_sym, &copts->safe},
+                               {ignore_under_sym, &copts->ignore_under},
+                               {oj_create_additions_sym, &copts->create_ok},
+                               {cache_keys_sym, &copts->cache_keys},
+                               {Qnil, 0}};
+    YesNoOpt         o;
+
+    for (o = ynos; 0 != o->attr; o++) {
+        if (key == o->sym) {
+            if (Qnil == value) {
+                *o->attr = NotSet;
+            } else if (Qtrue == value) {
+                *o->attr = Yes;
+            } else if (Qfalse == value) {
+                *o->attr = No;
+            } else {
+                rb_raise(rb_eArgError, "%s must be true, false, or nil.", rb_id2name(key));
+            }
+            return true;
+        }
+    }
+    return false;
+}
+
+static const char *make_only_value(VALUE v) {
+    switch (rb_type(v)) {
+    case RUBY_T_NIL:
+    case RUBY_T_NONE: return NULL;
+    case RUBY_T_ARRAY: {
+        long  len = rb_array_len(v);
+        long  i;
+        long  size = 0;
+        char *buf;
+        char *bp;
+
+        for (i = 0; i < len; i++) {
+            VALUE x = rb_ary_entry(v, i);
+
+            switch (rb_type(x)) {
+            case RUBY_T_SYMBOL:
+                size += strlen(rb_id2name(rb_sym2id(x)));
+                size++;
+                break;
+            case RUBY_T_STRING:
+                size += strlen(StringValueCStr(x));
+                size++;
+                break;
+            default: rb_raise(rb_eArgError, ":only and :except must be nil, symbol, string, or array."); break;
+            }
+        }
+        if (0 == size) {
+            return NULL;
+        }
+        buf   = OJ_R_ALLOC_N(char, size + 2);
+        bp    = buf;
+        *bp++ = ':';
+        for (i = 0; i < len; i++) {
+            VALUE       x = rb_ary_entry(v, i);
+            const char *str;
+
+            switch (rb_type(x)) {
+            case RUBY_T_SYMBOL:
+                str  = rb_id2name(rb_sym2id(x));
+                size = strlen(str);
+                memcpy(bp, str, size);
+                bp += size;
+                *bp++ = ':';
+                break;
+            case RUBY_T_STRING:
+                str  = StringValueCStr(x);
+                size = strlen(str);
+                memcpy(bp, str, size);
+                bp += size;
+                *bp++ = ':';
+                break;
+            default:
+                // ignore
+                break;
+            }
+        }
+        *bp = '\0';
+
+        return buf;
+    }
+    case RUBY_T_STRING: {
+        const char *str  = StringValueCStr(v);
+        size_t      size = strlen(str);
+        char       *buf  = OJ_R_ALLOC_N(char, size + 3);
+
+        buf[0] = ':';
+        strcpy(buf + 1, str);
+        buf[size + 1] = ':';
+        buf[size + 2] = '\0';
+
+        return buf;
+    }
+    case RUBY_T_SYMBOL: {
+        const char *str  = rb_id2name(rb_sym2id(v));
+        size_t      size = strlen(str);
+        char       *buf  = OJ_R_ALLOC_N(char, size + 3);
+
+        buf[0] = ':';
+        strcpy(buf + 1, str);
+        buf[size + 1] = ':';
+        buf[size + 2] = '\0';
+
+        return buf;
+    }
+    default: rb_raise(rb_eArgError, ":only and zzz :except must be nil, symbol, string, or array."); break;
+    }
+    return NULL;
+}
+
+static int parse_options_cb(VALUE k, VALUE v, VALUE opts) {
+    Options copts = (Options)opts;
+    size_t  len;
+
+    if (set_yesno_options(k, v, copts)) {
+        return ST_CONTINUE;
+    }
+    if (oj_indent_sym == k) {
+        switch (rb_type(v)) {
+        case T_NIL:
+            copts->dump_opts.indent_size = 0;
+            *copts->dump_opts.indent_str = '\0';
+            copts->indent                = 0;
+            break;
+        case T_FIXNUM:
+            copts->dump_opts.indent_size = 0;
+            *copts->dump_opts.indent_str = '\0';
+            if (MAX_INDENT < FIX2INT(v)) {
+                rb_raise(rb_eArgError, "indent is limited to %d characters.", MAX_INDENT);
+            }
+            copts->indent = FIX2INT(v);
+            break;
+        case T_STRING:
+            if (sizeof(copts->dump_opts.indent_str) <= (len = RSTRING_LEN(v))) {
+                rb_raise(rb_eArgError,
+                         "indent string is limited to %lu characters.",
+                         (unsigned long)sizeof(copts->dump_opts.indent_str));
+            }
+            strcpy(copts->dump_opts.indent_str, StringValuePtr(v));
+            copts->dump_opts.indent_size = (uint8_t)len;
+            copts->indent                = 0;
+            break;
+        default: rb_raise(rb_eTypeError, "indent must be a Fixnum, String, or nil."); break;
+        }
+    } else if (float_prec_sym == k) {
+        int n;
+
+        if (rb_cInteger != rb_obj_class(v)) {
+            rb_raise(rb_eArgError, ":float_precision must be a Integer.");
+        }
+        n = FIX2INT(v);
+        if (0 >= n) {
+            *copts->float_fmt = '\0';
+            copts->float_prec = 0;
+        } else {
+            if (20 < n) {
+                n = 20;
+            }
+            sprintf(copts->float_fmt, "%%0.%dg", n);
+            copts->float_prec = n;
+        }
+    } else if (cache_str_sym == k || cache_string_sym == k) {
+        int n;
+
+        if (rb_cInteger != rb_obj_class(v)) {
+            rb_raise(rb_eArgError, ":cache_str must be a Integer.");
+        }
+        n = FIX2INT(v);
+        if (0 >= n) {
+            copts->cache_str = 0;
+        } else {
+            if (32 < n) {
+                n = 32;
+            }
+            copts->cache_str = (char)n;
+        }
+    } else if (sec_prec_sym == k) {
+        int n;
+
+        if (rb_cInteger != rb_obj_class(v)) {
+            rb_raise(rb_eArgError, ":second_precision must be a Integer.");
+        }
+        n = NUM2INT(v);
+        if (0 > n) {
+            n                   = 0;
+            copts->sec_prec_set = false;
+        } else if (9 < n) {
+            n                   = 9;
+            copts->sec_prec_set = true;
+        } else {
+            copts->sec_prec_set = true;
+        }
+        copts->sec_prec = n;
+    } else if (mode_sym == k) {
+        if (wab_sym == v) {
+            copts->mode = WabMode;
+        } else if (object_sym == v) {
+            copts->mode = ObjectMode;
+        } else if (strict_sym == v) {
+            copts->mode = StrictMode;
+        } else if (compat_sym == v || json_sym == v) {
+            copts->mode = CompatMode;
+        } else if (null_sym == v) {
+            copts->mode = NullMode;
+        } else if (custom_sym == v) {
+            copts->mode = CustomMode;
+        } else if (rails_sym == v) {
+            copts->mode = RailsMode;
+        } else {
+            rb_raise(rb_eArgError, ":mode must be :object, :strict, :compat, :null, :custom, :rails, or :wab.");
+        }
+    } else if (time_format_sym == k) {
+        if (unix_sym == v) {
+            copts->time_format = UnixTime;
+        } else if (unix_zone_sym == v) {
+            copts->time_format = UnixZTime;
+        } else if (xmlschema_sym == v) {
+            copts->time_format = XmlTime;
+        } else if (ruby_sym == v) {
+            copts->time_format = RubyTime;
+        } else {
+            rb_raise(rb_eArgError, ":time_format must be :unix, :unix_zone, :xmlschema, or :ruby.");
+        }
+    } else if (escape_mode_sym == k) {
+        if (newline_sym == v) {
+            copts->escape_mode = NLEsc;
+        } else if (json_sym == v) {
+            copts->escape_mode = JSONEsc;
+        } else if (slash_sym == v) {
+            copts->escape_mode = SlashEsc;
+        } else if (xss_safe_sym == v) {
+            copts->escape_mode = XSSEsc;
+        } else if (ascii_sym == v) {
+            copts->escape_mode = ASCIIEsc;
+        } else if (unicode_xss_sym == v) {
+            copts->escape_mode = JXEsc;
+        } else {
+            rb_raise(rb_eArgError, ":encoding must be :newline, :json, :xss_safe, :unicode_xss, or :ascii.");
+        }
+    } else if (bigdecimal_load_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+
+        if (bigdecimal_sym == v || Qtrue == v) {
+            copts->bigdec_load = BigDec;
+        } else if (float_sym == v) {
+            copts->bigdec_load = FloatDec;
+        } else if (fast_sym == v) {
+            copts->bigdec_load = FastDec;
+        } else if (auto_sym == v || Qfalse == v) {
+            copts->bigdec_load = AutoDec;
+        } else {
+            rb_raise(rb_eArgError, ":bigdecimal_load must be :bigdecimal, :float, or :auto.");
+        }
+    } else if (compat_bigdecimal_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        copts->compat_bigdec = (Qtrue == v);
+    } else if (oj_decimal_class_sym == k) {
+        if (rb_cFloat == v) {
+            copts->compat_bigdec = false;
+        } else if (oj_bigdecimal_class == v) {
+            copts->compat_bigdec = true;
+        } else {
+            rb_raise(rb_eArgError, ":decimal_class must be BigDecimal or Float.");
+        }
+    } else if (create_id_sym == k) {
+        if (Qnil == v) {
+            if (oj_json_class != oj_default_options.create_id && NULL != copts->create_id) {
+                OJ_R_FREE((char *)oj_default_options.create_id);
+            }
+            copts->create_id     = NULL;
+            copts->create_id_len = 0;
+        } else if (T_STRING == rb_type(v)) {
+            const char *str = StringValuePtr(v);
+
+            len = RSTRING_LEN(v);
+            if (len != copts->create_id_len || 0 != strcmp(copts->create_id, str)) {
+                copts->create_id = OJ_R_ALLOC_N(char, len + 1);
+                strcpy((char *)copts->create_id, str);
+                copts->create_id_len = len;
+            }
+        } else {
+            rb_raise(rb_eArgError, ":create_id must be string.");
+        }
+    } else if (oj_space_sym == k) {
+        if (Qnil == v) {
+            copts->dump_opts.after_size = 0;
+            *copts->dump_opts.after_sep = '\0';
+        } else {
+            rb_check_type(v, T_STRING);
+            if (sizeof(copts->dump_opts.after_sep) <= (len = RSTRING_LEN(v))) {
+                rb_raise(rb_eArgError,
+                         "space string is limited to %lu characters.",
+                         (unsigned long)sizeof(copts->dump_opts.after_sep));
+            }
+            strcpy(copts->dump_opts.after_sep, StringValuePtr(v));
+            copts->dump_opts.after_size = (uint8_t)len;
+        }
+    } else if (oj_space_before_sym == k) {
+        if (Qnil == v) {
+            copts->dump_opts.before_size = 0;
+            *copts->dump_opts.before_sep = '\0';
+        } else {
+            rb_check_type(v, T_STRING);
+            if (sizeof(copts->dump_opts.before_sep) <= (len = RSTRING_LEN(v))) {
+                rb_raise(rb_eArgError,
+                         "sapce_before string is limited to %lu characters.",
+                         (unsigned long)sizeof(copts->dump_opts.before_sep));
+            }
+            strcpy(copts->dump_opts.before_sep, StringValuePtr(v));
+            copts->dump_opts.before_size = (uint8_t)len;
+        }
+    } else if (oj_object_nl_sym == k) {
+        if (Qnil == v) {
+            copts->dump_opts.hash_size = 0;
+            *copts->dump_opts.hash_nl  = '\0';
+        } else {
+            rb_check_type(v, T_STRING);
+            if (sizeof(copts->dump_opts.hash_nl) <= (len = RSTRING_LEN(v))) {
+                rb_raise(rb_eArgError,
+                         "object_nl string is limited to %lu characters.",
+                         (unsigned long)sizeof(copts->dump_opts.hash_nl));
+            }
+            strcpy(copts->dump_opts.hash_nl, StringValuePtr(v));
+            copts->dump_opts.hash_size = (uint8_t)len;
+        }
+    } else if (oj_array_nl_sym == k) {
+        if (Qnil == v) {
+            copts->dump_opts.array_size = 0;
+            *copts->dump_opts.array_nl  = '\0';
+        } else {
+            rb_check_type(v, T_STRING);
+            if (sizeof(copts->dump_opts.array_nl) <= (len = RSTRING_LEN(v))) {
+                rb_raise(rb_eArgError,
+                         "array_nl string is limited to %lu characters.",
+                         (unsigned long)sizeof(copts->dump_opts.array_nl));
+            }
+            strcpy(copts->dump_opts.array_nl, StringValuePtr(v));
+            copts->dump_opts.array_size = (uint8_t)len;
+        }
+    } else if (nan_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        if (null_sym == v) {
+            copts->dump_opts.nan_dump = NullNan;
+        } else if (huge_sym == v) {
+            copts->dump_opts.nan_dump = HugeNan;
+        } else if (word_sym == v) {
+            copts->dump_opts.nan_dump = WordNan;
+        } else if (raise_sym == v) {
+            copts->dump_opts.nan_dump = RaiseNan;
+        } else if (auto_sym == v) {
+            copts->dump_opts.nan_dump = AutoNan;
+        } else {
+            rb_raise(rb_eArgError, ":nan must be :null, :huge, :word, :raise, or :auto.");
+        }
+    } else if (omit_nil_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        if (Qtrue == v) {
+            copts->dump_opts.omit_nil = true;
+        } else if (Qfalse == v) {
+            copts->dump_opts.omit_nil = false;
+        } else {
+            rb_raise(rb_eArgError, ":omit_nil must be true or false.");
+        }
+    } else if (omit_null_byte_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        if (Qtrue == v) {
+            copts->dump_opts.omit_null_byte = true;
+        } else if (Qfalse == v) {
+            copts->dump_opts.omit_null_byte = false;
+        } else {
+            rb_raise(rb_eArgError, ":omit_null_byte must be true or false.");
+        }
+    } else if (oj_ascii_only_sym == k) {
+        // This is here only for backwards compatibility with the original Oj.
+        if (Qtrue == v) {
+            copts->escape_mode = ASCIIEsc;
+        } else if (Qfalse == v) {
+            copts->escape_mode = JSONEsc;
+        }
+    } else if (oj_hash_class_sym == k) {
+        if (Qnil == v) {
+            copts->hash_class = Qnil;
+        } else {
+            rb_check_type(v, T_CLASS);
+            copts->hash_class = v;
+        }
+    } else if (oj_object_class_sym == k) {
+        if (Qnil == v) {
+            copts->hash_class = Qnil;
+        } else {
+            rb_check_type(v, T_CLASS);
+            copts->hash_class = v;
+        }
+    } else if (oj_array_class_sym == k) {
+        if (Qnil == v) {
+            copts->array_class = Qnil;
+        } else {
+            rb_check_type(v, T_CLASS);
+            copts->array_class = v;
+        }
+    } else if (ignore_sym == k) {
+        OJ_R_FREE(copts->ignore);
+        copts->ignore = NULL;
+        if (Qnil != v) {
+            size_t cnt;
+
+            rb_check_type(v, T_ARRAY);
+            cnt = RARRAY_LEN(v);
+            if (0 < cnt) {
+                size_t i;
+
+                copts->ignore = OJ_R_ALLOC_N(VALUE, cnt + 1);
+                for (i = 0; i < cnt; i++) {
+                    copts->ignore[i] = RARRAY_AREF(v, i);
+                }
+                copts->ignore[i] = Qnil;
+            }
+        }
+    } else if (integer_range_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        if (rb_obj_class(v) == rb_cRange) {
+            VALUE min = rb_funcall(v, oj_begin_id, 0);
+            VALUE max = rb_funcall(v, oj_end_id, 0);
+
+            if (TYPE(min) != T_FIXNUM || TYPE(max) != T_FIXNUM) {
+                rb_raise(rb_eArgError, ":integer_range range bounds is not Fixnum.");
+            }
+
+            copts->int_range_min = FIX2LONG(min);
+            copts->int_range_max = FIX2LONG(max);
+        } else if (Qfalse != v) {
+            rb_raise(rb_eArgError, ":integer_range must be a range of Fixnum.");
+        }
+    } else if (max_integer_digits_sym == k) {
+        if (Qnil == v || Qfalse == v) {
+            copts->max_integer_digits = 0;
+        } else if (T_FIXNUM == rb_type(v)) {
+            long n = FIX2LONG(v);
+
+            if (n < 0) {
+                rb_raise(rb_eArgError, ":max_integer_digits must be >= 0.");
+            }
+
+            copts->max_integer_digits = (size_t)n;
+        } else {
+            rb_raise(rb_eArgError, ":max_integer_digits must be a non-negative Integer.");
+        }
+    } else if (symbol_keys_sym == k || oj_symbolize_names_sym == k) {
+        if (Qnil == v) {
+            return ST_CONTINUE;
+        }
+        copts->sym_key = (Qtrue == v) ? Yes : No;
+
+    } else if (oj_max_nesting_sym == k) {
+        if (Qtrue == v) {
+            copts->dump_opts.max_depth = 100;
+        } else if (Qfalse == v || Qnil == v) {
+            copts->dump_opts.max_depth = MAX_DEPTH;
+        } else if (T_FIXNUM == rb_type(v)) {
+            copts->dump_opts.max_depth = NUM2INT(v);
+            if (0 >= copts->dump_opts.max_depth) {
+                copts->dump_opts.max_depth = MAX_DEPTH;
+            }
+        }
+    } else if (float_format_sym == k) {
+        rb_check_type(v, T_STRING);
+        if (6 < RSTRING_LEN(v)) {
+            rb_raise(rb_eArgError, ":float_format must be 6 bytes or less.");
+        }
+        strncpy(copts->float_fmt, RSTRING_PTR(v), (size_t)RSTRING_LEN(v));
+        copts->float_fmt[RSTRING_LEN(v)] = '\0';
+    } else if (only_sym == k) {
+        if (NULL != copts->dump_opts.only) {
+            OJ_R_FREE((void *)copts->dump_opts.only);
+            copts->dump_opts.only = NULL;
+        }
+        copts->dump_opts.only = make_only_value(v);
+    } else if (except_sym == k) {
+        if (NULL != copts->dump_opts.except) {
+            OJ_R_FREE((void *)copts->dump_opts.except);
+            copts->dump_opts.except = NULL;
+        }
+        copts->dump_opts.except = make_only_value(v);
+    }
+    return ST_CONTINUE;
+}
+
+void oj_parse_options(VALUE ropts, Options copts) {
+    if (T_HASH != rb_type(ropts)) {
+        return;
+    }
+    rb_hash_foreach(ropts, parse_options_cb, (VALUE)copts);
+    oj_parse_opt_match_string(&copts->str_rx, ropts);
+
+    copts->dump_opts.use = (0 < copts->dump_opts.indent_size || 0 < copts->dump_opts.after_size ||
+                            0 < copts->dump_opts.before_size || 0 < copts->dump_opts.hash_size ||
+                            0 < copts->dump_opts.array_size);
+    return;
+}
+
+static int match_string_cb(VALUE key, VALUE value, VALUE rx) {
+    RxClass rc = (RxClass)rx;
+
+    if (T_CLASS != rb_type(value)) {
+        rb_raise(rb_eArgError, "for :match_string, the hash values must be a Class.");
+    }
+    switch (rb_type(key)) {
+    case T_REGEXP: oj_rxclass_rappend(rc, key, value); break;
+    case T_STRING:
+        if (0 != oj_rxclass_append(rc, StringValuePtr(key), value)) {
+            rb_raise(rb_eArgError, "%s", rc->err);
+        }
+        break;
+    default: rb_raise(rb_eArgError, "for :match_string, keys must either a String or RegExp."); break;
+    }
+    return ST_CONTINUE;
+}
+
+void oj_parse_opt_match_string(RxClass rc, VALUE ropts) {
+    VALUE v;
+
+    if (Qnil != (v = rb_hash_lookup(ropts, match_string_sym))) {
+        rb_check_type(v, T_HASH);
+        // Zero out rc. Pattern are not appended but override.
+        rc->head = NULL;
+        rc->tail = NULL;
+        *rc->err = '\0';
+        rb_hash_foreach(v, match_string_cb, (VALUE)rc);
+    }
+}
+
+/* Document-method: load
+ * call-seq: load(json, options={}) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into a Object, Hash, Array, String, Fixnum,
+ * Float, true, false, or nil according to the default mode or the mode
+ * specified. Raises an exception if the JSON is malformed or the classes
+ * specified are not valid. If the string input is not a valid JSON document (an
+ * empty string is not a valid JSON document) an exception is raised.
+ *
+ * When used with a document that has multiple JSON elements the block, if
+ * any, will be yielded to. If no block then the last element read will be
+ * returned.
+ *
+ * This parser operates on string and will attempt to load files into memory if
+ * a file object is passed as the first argument. A stream input will be parsed
+ * using a stream parser but others use the slightly faster string parser.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *json* [_String_|_IO_] JSON String or an Object that responds to read()
+ * - *options* [_Hash_] load options (same as default_options)
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+static VALUE load(int argc, VALUE *argv, VALUE self) {
+    Mode mode = oj_default_options.mode;
+
+    if (1 > argc) {
+        rb_raise(rb_eArgError, "Wrong number of arguments to load().");
+    }
+    if (2 <= argc) {
+        VALUE ropts = argv[1];
+        VALUE v;
+
+        if (Qnil != ropts || CompatMode != mode) {
+            Check_Type(ropts, T_HASH);
+            if (Qnil != (v = rb_hash_lookup(ropts, mode_sym))) {
+                if (object_sym == v) {
+                    mode = ObjectMode;
+                } else if (strict_sym == v) {
+                    mode = StrictMode;
+                } else if (compat_sym == v || json_sym == v) {
+                    mode = CompatMode;
+                } else if (null_sym == v) {
+                    mode = NullMode;
+                } else if (custom_sym == v) {
+                    mode = CustomMode;
+                } else if (rails_sym == v) {
+                    mode = RailsMode;
+                } else if (wab_sym == v) {
+                    mode = WabMode;
+                } else {
+                    rb_raise(rb_eArgError,
+                             ":mode must be :object, :strict, :compat, :null, :custom, :rails, or "
+                             ":wab.");
+                }
+            }
+        }
+    }
+    switch (mode) {
+    case StrictMode:
+    case NullMode: return oj_strict_parse(argc, argv, self);
+    case CompatMode:
+    case RailsMode: return oj_compat_parse(argc, argv, self);
+    case CustomMode: return oj_custom_parse(argc, argv, self);
+    case WabMode: return oj_wab_parse(argc, argv, self);
+    case ObjectMode:
+    default: break;
+    }
+    return oj_object_parse(argc, argv, self);
+}
+
+/* Document-method: load_file
+ * call-seq: load_file(path, options={}) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into a Object, Hash, Array, String, Fixnum,
+ * Float, true, false, or nil according to the default mode or the mode
+ * specified. Raises an exception if the JSON is malformed or the classes
+ * specified are not valid. If the string input is not a valid JSON document (an
+ * empty string is not a valid JSON document) an exception is raised.
+ *
+ * When used with a document that has multiple JSON elements the block, if
+ * any, will be yielded to. If no block then the last element read will be
+ * returned.
+ *
+ * If the input file is not a valid JSON document (an empty file is not a valid
+ * JSON document) an exception is raised.
+ *
+ * This is a stream based parser which allows a large or huge file to be loaded
+ * without pulling the whole file into memory.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *path* [_String_] to a file containing a JSON document
+ * - *options* [_Hash_] load options (same as default_options)
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Object_|_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+static VALUE load_file(int argc, VALUE *argv, VALUE self) {
+    char             *path;
+    int               fd;
+    Mode              mode = oj_default_options.mode;
+    struct _parseInfo pi;
+
+    if (1 > argc) {
+        rb_raise(rb_eArgError, "Wrong number of arguments to load().");
+    }
+    path = StringValuePtr(*argv);
+    parse_info_init(&pi);
+    pi.options   = oj_default_options;
+    pi.handler   = Qnil;
+    pi.err_class = Qnil;
+    pi.max_depth = 0;
+    if (2 <= argc) {
+        VALUE ropts = argv[1];
+        VALUE v;
+
+        Check_Type(ropts, T_HASH);
+        if (Qnil != (v = rb_hash_lookup(ropts, mode_sym))) {
+            if (object_sym == v) {
+                mode = ObjectMode;
+            } else if (strict_sym == v) {
+                mode = StrictMode;
+            } else if (compat_sym == v || json_sym == v) {
+                mode = CompatMode;
+            } else if (null_sym == v) {
+                mode = NullMode;
+            } else if (custom_sym == v) {
+                mode = CustomMode;
+            } else if (rails_sym == v) {
+                mode = RailsMode;
+            } else if (wab_sym == v) {
+                mode = WabMode;
+            } else {
+                rb_raise(rb_eArgError, ":mode must be :object, :strict, :compat, :null, :custom, :rails, or :wab.");
+            }
+        }
+    }
+#ifdef _WIN32
+    {
+        WCHAR *wide_path;
+        wide_path = rb_w32_mbstr_to_wstr(CP_UTF8, path, -1, NULL);
+        fd        = rb_w32_wopen(wide_path, O_RDONLY);
+        OJ_FREE(wide_path);
+    }
+#else
+    fd = open(path, O_RDONLY);
+#endif
+    if (0 == fd) {
+        rb_raise(rb_eIOError, "%s", strerror(errno));
+    }
+    switch (mode) {
+    case StrictMode:
+    case NullMode: oj_set_strict_callbacks(&pi); return oj_pi_sparse(argc, argv, &pi, fd);
+    case CustomMode: oj_set_custom_callbacks(&pi); return oj_pi_sparse(argc, argv, &pi, fd);
+    case CompatMode:
+    case RailsMode: oj_set_compat_callbacks(&pi); return oj_pi_sparse(argc, argv, &pi, fd);
+    case WabMode: oj_set_wab_callbacks(&pi); return oj_pi_sparse(argc, argv, &pi, fd);
+    case ObjectMode:
+    default: break;
+    }
+    oj_set_object_callbacks(&pi);
+
+    return oj_pi_sparse(argc, argv, &pi, fd);
+}
+
+/* Document-method: safe_load
+ * call-seq: safe_load(doc)
+ *
+ * Loads a JSON document in strict mode with :auto_define and :symbol_keys
+ * turned off. This function should be safe to use with JSON received on an
+ * unprotected public interface.
+ *
+ * - *doc* [_String__|_IO_] JSON String or IO to load.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Bignum_|_BigDecimal_|_nil_|_True_|_False_]
+ */
+static VALUE safe_load(VALUE self, VALUE doc) {
+    struct _parseInfo pi;
+    VALUE             args[1];
+
+    parse_info_init(&pi);
+    pi.err_class           = Qnil;
+    pi.max_depth           = 0;
+    pi.options             = oj_default_options;
+    pi.options.auto_define = No;
+    pi.options.sym_key     = No;
+    pi.options.mode        = StrictMode;
+    oj_set_strict_callbacks(&pi);
+    *args = doc;
+
+    return oj_pi_parse(1, args, &pi, 0, 0, 1);
+}
+
+/* Document-method: saj_parse
+ * call-seq: saj_parse(handler, io)
+ *
+ * Parses an IO stream or file containing a JSON document. Raises an exception
+ * if the JSON is malformed. This is a callback parser that calls the methods in
+ * the handler if they exist. A sample is the Oj::Saj class which can be used as
+ * a base class for the handler.
+ *
+ * - *handler* [_Oj::Saj_] responds to Oj::Saj methods
+ * - *io* [_IO_|_String_] IO Object to read from
+ */
+
+/* Document-method: sc_parse
+ * call-seq: sc_parse(handler, io)
+ *
+ * Parses an IO stream or file containing a JSON document. Raises an exception
+ * if the JSON is malformed. This is a callback parser (Simple Callback Parser)
+ * that calls the methods in the handler if they exist. A sample is the
+ * Oj::ScHandler class which can be used as a base class for the handler. This
+ * callback parser is slightly more efficient than the Saj callback parser and
+ * requires less argument checking.
+ *
+ * - *handler* [_Oj_::ScHandler_] responds to Oj::ScHandler methods
+ * - *io* [_IO__|_String_] IO Object to read from
+ */
+
+struct dump_arg {
+    struct _out     *out;
+    struct _options *copts;
+    int              argc;
+    VALUE           *argv;
+};
+
+static VALUE dump_body(VALUE a) {
+    volatile struct dump_arg *arg = (void *)a;
+    VALUE                     rstr;
+
+    oj_dump_obj_to_json_using_params(*arg->argv, arg->copts, arg->out, arg->argc - 1, arg->argv + 1);
+    if (0 == arg->out->buf) {
+        rb_raise(rb_eNoMemError, "Not enough memory.");
+    }
+    rstr = rb_utf8_str_new_cstr(arg->out->buf);
+
+    return rstr;
+}
+
+static VALUE dump_ensure(VALUE a) {
+    volatile struct dump_arg *arg = (void *)a;
+
+    oj_out_free(arg->out);
+
+    return Qnil;
+}
+
+/* Document-method: dump
+ * call-seq: dump(obj, options={})
+ *
+ * Dumps an Object (obj) to a string.
+ * - *obj* [_Object_] Object to serialize as an JSON document String
+ * - *options* [_Hash_] same as default_options
+ */
+static VALUE dump(int argc, VALUE *argv, VALUE self) {
+    struct dump_arg arg;
+    struct _out     out;
+    struct _options copts = oj_default_options;
+
+    if (1 > argc) {
+        rb_raise(rb_eArgError, "wrong number of arguments (0 for 1).");
+    }
+    if (CompatMode == copts.mode) {
+        copts.dump_opts.nan_dump = WordNan;
+    }
+    if (2 == argc) {
+        oj_parse_options(argv[1], &copts);
+    }
+    if (CompatMode == copts.mode && copts.escape_mode != ASCIIEsc) {
+        copts.escape_mode = JSONEsc;
+    }
+    arg.out   = &out;
+    arg.copts = &copts;
+    arg.argc  = argc;
+    arg.argv  = argv;
+
+    oj_out_init(arg.out);
+
+    arg.out->omit_nil       = copts.dump_opts.omit_nil;
+    arg.out->omit_null_byte = copts.dump_opts.omit_null_byte;
+
+    return rb_ensure(dump_body, (VALUE)&arg, dump_ensure, (VALUE)&arg);
+}
+
+/* Document-method: to_json
+ * call-seq: to_json(obj, options)
+ *
+ * Dumps an Object (obj) to a string. If the object has a to_json method that
+ * will be called. The mode is set to :compat.
+ * - *obj* [_Object_] Object to serialize as an JSON document String
+ * - *options* [_Hash_]
+ *   - *:max_nesting* [_Fixnum_|_boolean_] It true nesting is limited to 100.
+ *     If a Fixnum nesting is set to the provided value. The option to detect
+ *     circular references is available but is not compatible with the json gem.,
+ *     default is false or unlimited.
+ *   - *:allow_nan* [_boolean_] If true non JSON compliant words such as Nan and
+ *     Infinity will be used as appropriate, default is true.
+ *   - *:quirks_mode* [_boolean_] Allow single JSON values instead of documents, default is true (allow).
+ *   - *:indent* [_String_|_nil_] String to use for indentation, overriding the indent option if not nil.
+ *   - *:space* [_String_|_nil_] String to use for the space after the colon in JSON object fields.
+ *   - *:space_before* [_String_|_nil_] String to use before the colon separator in JSON object fields.
+ *   - *:object_nl* [_String_|_nil_] String to use after a JSON object field value.
+ *   - *:array_nl* [_String_|_nil_] String to use after a JSON array value.
+ *   - *:trace* [_Boolean_] If true trace is turned on.
+ *
+ * Returns [_String_] the encoded JSON.
+ */
+static VALUE to_json(int argc, VALUE *argv, VALUE self) {
+    struct _out     out;
+    struct _options copts = oj_default_options;
+    VALUE           rstr;
+
+    if (1 > argc) {
+        rb_raise(rb_eArgError, "wrong number of arguments (0 for 1).");
+    }
+    copts.escape_mode        = JXEsc;
+    copts.dump_opts.nan_dump = RaiseNan;
+    if (2 == argc) {
+        oj_parse_mimic_dump_options(argv[1], &copts);
+    }
+    copts.mode    = CompatMode;
+    copts.to_json = Yes;
+
+    oj_out_init(&out);
+
+    out.omit_nil       = copts.dump_opts.omit_nil;
+    out.omit_null_byte = copts.dump_opts.omit_null_byte;
+    // For obj.to_json or generate nan is not allowed but if called from dump
+    // it is.
+    oj_dump_obj_to_json_using_params(*argv, &copts, &out, argc - 1, argv + 1);
+
+    if (0 == out.buf) {
+        rb_raise(rb_eNoMemError, "Not enough memory.");
+    }
+    rstr = rb_utf8_str_new_cstr(out.buf);
+
+    oj_out_free(&out);
+
+    return rstr;
+}
+
+/* Document-method: to_file
+ * call-seq: to_file(file_path, obj, options={})
+ *
+ * Dumps an Object to the specified file.
+ * - *file* [_String_] _path file path to write the JSON document to
+ * - *obj* [_Object_] Object to serialize as an JSON document String
+ * - *options* [_Hash_] formatting options
+ *   - *:indent* [_Fixnum_] format expected
+ *   - *:circular* [_Boolean_] allow circular references, default: false
+ */
+static VALUE to_file(int argc, VALUE *argv, VALUE self) {
+    struct _options copts = oj_default_options;
+
+    if (3 == argc) {
+        oj_parse_options(argv[2], &copts);
+    }
+    oj_write_obj_to_file(argv[1], StringValuePtr(*argv), &copts);
+
+    return Qnil;
+}
+
+/* Document-method: to_stream
+ * call-seq: to_stream(io, obj, options={})
+ *
+ * Dumps an Object to the specified IO stream.
+ * - *io* [_IO_] IO stream to write the JSON document to
+ * - *obj* [_Object_] Object to serialize as an JSON document String
+ * - *options* [_Hash_] formatting options
+ *   - *:indent* [_Fixnum_] format expected
+ *   - *:circular* [_Boolean_] allow circular references, default: false
+ */
+static VALUE to_stream(int argc, VALUE *argv, VALUE self) {
+    struct _options copts = oj_default_options;
+
+    if (3 == argc) {
+        oj_parse_options(argv[2], &copts);
+    }
+    oj_write_obj_to_stream(argv[1], *argv, &copts);
+
+    return Qnil;
+}
+
+/* Document-method: register_odd
+ * call-seq: register_odd(clas, create_object, create_method, *members)
+ *
+ * Registers a class as special. This is useful for working around subclasses of
+ * primitive types as is done with ActiveSupport classes. The use of this
+ * function should be limited to just classes that can not be handled in the
+ * normal way. It is not intended as a hook for changing the output of all
+ * classes as it is not optimized for large numbers of classes.
+ *
+ * - *clas* [_Class__|_Module_] Class or Module to be made special
+ * - *create_object* [_Object_]  object to call the create method on
+ * - *create_method* [_Symbol_] method on the clas that will create a new instance
+ *   of the clas when given all the member values in the order specified.
+ * - *members* [_Symbol__|_String_] methods used to get the member values from
+ *   instances of the clas.
+ */
+static VALUE register_odd(int argc, VALUE *argv, VALUE self) {
+    if (3 > argc) {
+        rb_raise(rb_eArgError, "incorrect number of arguments.");
+    }
+    switch (rb_type(*argv)) {
+    case T_CLASS:
+    case T_MODULE: break;
+    default: rb_raise(rb_eTypeError, "expected a class or module."); break;
+    }
+    Check_Type(argv[2], T_SYMBOL);
+    if (MAX_ODD_ARGS < argc - 2) {
+        rb_raise(rb_eArgError, "too many members.");
+    }
+    oj_reg_odd(argv[0], argv[1], argv[2], argc - 3, argv + 3, false);
+
+    return Qnil;
+}
+
+/* Document-method: register_odd_raw
+ *	call-seq: register_odd_raw(clas, create_object, create_method, dump_method)
+ *
+ * Registers a class as special and expect the output to be a string that can be
+ * included in the dumped JSON directly. This is useful for working around
+ * subclasses of primitive types as is done with ActiveSupport classes. The use
+ * of this function should be limited to just classes that can not be handled in
+ * the normal way. It is not intended as a hook for changing the output of all
+ * classes as it is not optimized for large numbers of classes. Be careful with
+ * this option as the JSON may be incorrect if invalid JSON is returned.
+ *
+ * - *clas* [_Class_|_Module_] Class or Module to be made special
+ * - *create_object* [_Object_] object to call the create method on
+ * - *create_method* [_Symbol_] method on the clas that will create a new instance
+ *   of the clas when given all the member values in the order specified.
+ * - *dump_method* [_Symbol_|_String_] method to call on the object being
+ *   serialized to generate the raw JSON.
+ */
+static VALUE register_odd_raw(int argc, VALUE *argv, VALUE self) {
+    if (3 > argc) {
+        rb_raise(rb_eArgError, "incorrect number of arguments.");
+    }
+    switch (rb_type(*argv)) {
+    case T_CLASS:
+    case T_MODULE: break;
+    default: rb_raise(rb_eTypeError, "expected a class or module."); break;
+    }
+    Check_Type(argv[2], T_SYMBOL);
+    if (MAX_ODD_ARGS < argc - 2) {
+        rb_raise(rb_eArgError, "too many members.");
+    }
+    oj_reg_odd(argv[0], argv[1], argv[2], 1, argv + 3, true);
+
+    return Qnil;
+}
+
+////////////////////////////////////////////////////////////////////////////////
+// RDoc entries must be in the same file as the rb_define_method and must be
+// directly above the C method function. The extern declaration is enough to
+// get it to work.
+////////////////////////////////////////////////////////////////////////////////
+
+/* Document-method: strict_load
+ * call-seq: strict_load(json, options) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into an Hash, Array, String, Fixnum, Float,
+ * true, false, or nil. It parses using a mode that is strict in that it maps
+ * each primitive JSON type to a similar Ruby type. The :create_id is not
+ * honored in this mode. Note that a Ruby Hash is used to represent the JSON
+ * Object type. These two are not the same since the JSON Object type can have
+ * repeating entries with the same key and Ruby Hash can not.
+ *
+ * When used with a document that has multiple JSON elements the block, if
+ * any, will be yielded to. If no block then the last element read will be
+ * returned.
+ *
+ * Raises an exception if the JSON is malformed or the classes specified are not
+ * valid. If the input is not a valid JSON document (an empty string is not a
+ * valid JSON document) an exception is raised.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *json* [_String_|_IO_] JSON String or an Object that responds to read().
+ * - *options* [_Hash_] load options (same as default_options).
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+extern VALUE oj_strict_parse(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: compat_load
+ * call-seq: compat_load(json, options) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into an Object, Hash, Array, String, Fixnum,
+ * Float, true, false, or nil. It parses using a mode that is generally
+ * compatible with other Ruby JSON parsers in that it will create objects based
+ * on the :create_id value. It is not compatible in every way to every other
+ * parser though as each parser has it's own variations.
+ *
+ * When used with a document that has multiple JSON elements the block, if
+ * any, will be yielded to. If no block then the last element read will be
+ * returned.
+ *
+ * Raises an exception if the JSON is malformed or the classes specified are not
+ * valid. If the input is not a valid JSON document (an empty string is not a
+ * valid JSON document) an exception is raised.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *json* [_String_|_IO_] JSON String or an Object that responds to read().
+ * - *options* [_Hash_] load options (same as default_options).
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+extern VALUE oj_compat_parse(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: object_load
+ * call-seq: object_load(json, options) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into an Object, Hash, Array, String, Fixnum,
+ * Float, true, false, or nil. In the :object mode the JSON should have been
+ * generated by Oj.dump(). The parser will reconstitute the original marshalled
+ * or dumped Object. The :auto_define and :circular options have meaning with
+ * this parsing mode.
+ *
+ * Raises an exception if the JSON is malformed or the classes specified are not
+ * valid. If the input is not a valid JSON document (an empty string is not a
+ * valid JSON document) an exception is raised.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *json* [_String_|_IO_] JSON String or an Object that responds to read().
+ * - *options* [_Hash_] load options (same as default_options).
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+extern VALUE oj_object_parse(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: wab_load
+ * call-seq: wab_load(json, options) { _|_obj, start, len_|_ }
+ *
+ * Parses a JSON document String into an Hash, Array, String, Fixnum, Float,
+ * true, false, or nil. It parses using a mode that is :wab in that it maps
+ * each primitive JSON type to a similar Ruby type. The :create_id is not
+ * honored in this mode. Note that a Ruby Hash is used to represent the JSON
+ * Object type. These two are not the same since the JSON Object type can have
+ * repeating entries with the same key and Ruby Hash can not.
+ *
+ * When used with a document that has multiple JSON elements the block, if
+ * any, will be yielded to. If no block then the last element read will be
+ * returned.
+ *
+ * Raises an exception if the JSON is malformed or the classes specified are not
+ * valid. If the input is not a valid JSON document (an empty string is not a
+ * valid JSON document) an exception is raised.
+ *
+ * A block can be provided with a single argument. That argument will be the
+ * parsed JSON document. This is useful when parsing a string that includes
+ * multiple JSON documents. The block can take up to 3 arguments, the parsed
+ * object, the position in the string or stream of the start of the JSON for
+ * that object, and the length of the JSON for that object plus trailing
+ * whitespace.
+ *
+ * - *json* [_String_|_IO_] JSON String or an Object that responds to read().
+ * - *options* [_Hash_] load options (same as default_options).
+ * - *obj* [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_] parsed object.
+ * - *start* [_optional, _Integer_] start position of parsed JSON for obj.
+ * - *len* [_optional, _Integer_] length of parsed JSON for obj.
+ *
+ * Returns [_Hash_|_Array_|_String_|_Fixnum_|_Float_|_Boolean_|_nil_]
+ */
+extern VALUE oj_wab_parse(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: add_to_json
+ * call-seq: add_to_json(*args)
+ *
+ * Override simple to_s dump behavior in :compat mode to instead use an
+ * optimized dump that includes the classname and attributes so that the
+ * object can be re-created on load. The format is the same as the json gem
+ * but does not use the ruby methods for encoding.
+ *
+ * The classes supported for optimization are: Array, BigDecimal, Complex,
+ * Date, DateTime, Exception, Hash, Integer, OpenStruct, Range, Rational,
+ * Regexp, Struct, and Time. Providing no classes will result in all those
+ * classes being optimized.q
+ *
+ * - *args( [_Class_] zero or more classes to optimize.
+ */
+extern VALUE oj_add_to_json(int argc, VALUE *argv, VALUE self);
+
+/* @!method remove_to_json(*args)
+ *
+ * Reverts back to the to_s dump behavior in :compat mode to instead use an
+ * optimized dump that includes the classname and attributes so that the
+ * object can be re-created on load. The format is the same as the json gem
+ * but does not use the ruby methods for encoding.
+ *
+ * The classes supported for optimization are: Array, BigDecimal, Complex,
+ * Date, DateTime, Exception, Hash, Integer, OpenStruct, Range, Rational,
+ * Regexp, Struct, and Time. Providing no classes will result in all those
+ * classes being reverted from the optimized mode.
+ *
+ * - *args* [_Class_] zero or more classes to optimize.
+ */
+extern VALUE oj_remove_to_json(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: mimic_JSON
+ * call-seq: mimic_JSON()
+ *
+ * Creates the JSON module with methods and classes to mimic the JSON gem. After
+ * this method is invoked calls that expect the JSON module will use Oj instead
+ * and be faster than the original JSON. Most options that could be passed to
+ * the JSON methods are supported. The calls to set parser or generator will not
+ * raise an Exception but will not have any effect. The method can also be
+ * called after the json gem is loaded. The necessary methods on the json gem
+ * will be replaced with Oj methods.
+ *
+ * Note that this also sets the default options of :mode to :compat and
+ * :encoding to :unicode_xss.
+ *
+ * Returns [_Module_] the JSON module.
+ */
+extern VALUE oj_define_mimic_json(int argc, VALUE *argv, VALUE self);
+
+/* Document-method: generate
+ * call-seq: generate(obj, opts=nil)
+ *
+ * Encode obj as a JSON String. The obj argument must be a Hash, Array, or
+ * respond to to_h or to_json. Options other than those listed such as
+ * +:allow_nan+ or +:max_nesting+ are ignored. Calling this method will call
+ * Oj.mimic_JSON if it is not already called.
+ *
+ * - *obj* [_Object__|_Hash_|_Array_] object to convert to a JSON String
+ * - *opts* [_Hash_] options
+ *   - *:indent* [_String_] String to use for indentation.
+ *   - *:space* [_String_] String placed after a , or : delimiter
+ *   - *:space_before* [_String_] String placed before a : delimiter
+ *   - *:object_nl* [_String_] String placed after a JSON object
+ *   - *:array_nl* [_String_] String placed after a JSON array
+ *   - *:ascii_only* [_Boolean_] if not nil or false then use only ascii characters
+ *     in the output. Note JSON.generate does support this even if it is not documented.
+ *
+ * Returns [_String_]generated JSON.
+ */
+extern VALUE oj_mimic_generate(int argc, VALUE *argv, VALUE self);
+
+/* Document-module: Oj.optimize_rails()
+ *
+ * Sets the Oj as the Rails encoder and decoder. Oj::Rails.optimize is also
+ * called.
+ */
+extern VALUE oj_optimize_rails(VALUE self);
+
+/*
+extern void	oj_hash_test();
+static VALUE
+hash_test(VALUE self) {
+    oj_hash_test();
+    return Qnil;
+}
+*/
+/*
+extern void oj_hash_sizes();
+static VALUE
+hash_test(VALUE self) {
+    oj_hash_sizes();
+    return Qnil;
+}
+*/
+
+static VALUE protect_require(VALUE x) {
+    rb_require("time");
+    rb_require("bigdecimal");
+    return Qnil;
+}
+
+extern void print_all_odds(const char *label);
+
+static VALUE debug_odd(VALUE self, VALUE label) {
+    print_all_odds(RSTRING_PTR(label));
+    return Qnil;
+}
+
+static VALUE mem_report(VALUE self) {
+    oj_mem_report();
+    return Qnil;
+}
+
+/* Document-module: Oj
+ *
+ * Optimized JSON (Oj), as the name implies was written to provide speed
+ * optimized JSON handling.
+ *
+ * Oj uses modes to control how object are encoded and decoded. In addition
+ * global and options to methods allow additional behavior modifications. The
+ * modes are:
+ *
+ * - *:strict* mode will only allow the 7 basic JSON types to be serialized.
+ *   Any other Object will raise an Exception.
+ *
+ * - *:null* mode is similar to the :strict mode except any Object that is not
+ *   one of the JSON base types is replaced by a JSON null.
+ *
+ * - *:object* mode will dump any Object as a JSON Object with keys that match
+ *   the Ruby Object's variable names without the '@' character. This is the
+ *   highest performance mode.
+ *
+ * - *:compat* or *:json* mode is the compatible mode for the json gem. It mimics
+ *   the json gem including the options, defaults, and restrictions.
+ *
+ * - *:rails* is the compatibility mode for Rails or Active support.
+ *
+ * - *:custom* is the most configurable mode.
+ *
+ * - *:wab* specifically for WAB data exchange.
+ */
+
+// =============================================================================
+// Runtime SIMD CPU detection
+// Cross-platform support for Windows (MSVC), Linux, and macOS (GCC/Clang)
+// =============================================================================
+SIMD_Implementation oj_get_simd_implementation(void) {
+#ifdef HAVE_SIMD_X86
+    // x86/x86_64 runtime detection
+
+#if defined(_MSC_VER)
+    // MSVC: Use __cpuid intrinsic
+    int cpu_info[4];
+    __cpuid(cpu_info, 1);
+
+    // Check for SSE4.2 (bit 20 of ECX)
+    if (cpu_info[2] & (1 << 20)) {
+        return SIMD_SSE42;
+    }
+    // Check for SSE2 (bit 26 of EDX)
+    if (cpu_info[3] & (1 << 26)) {
+        return SIMD_SSE2;
+    }
+
+#elif defined(__GNUC__) || defined(__clang__)
+    // GCC/Clang: Use __builtin_cpu_supports if available
+#if defined(__has_builtin)
+#if __has_builtin(__builtin_cpu_supports)
+#define OJ_HAS_BUILTIN_CPU_SUPPORTS 1
+#endif
+#elif defined(__GNUC__) && (__GNUC__ > 4 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))
+    // GCC 4.8+ has __builtin_cpu_supports
+#define OJ_HAS_BUILTIN_CPU_SUPPORTS 1
+#endif
+
+#ifdef OJ_HAS_BUILTIN_CPU_SUPPORTS
+#ifdef HAVE_SIMD_SSE4_2
+    if (__builtin_cpu_supports("sse4.2")) {
+        return SIMD_SSE42;
+    }
+#endif
+#ifdef HAVE_SIMD_SSE2
+    if (__builtin_cpu_supports("sse2")) {
+        return SIMD_SSE2;
+    }
+#endif
+#else
+    // Fallback: Use CPUID instruction directly
+    unsigned int eax, ebx, ecx, edx;
+    if (__get_cpuid(1, &eax, &ebx, &ecx, &edx)) {
+        // Check for SSE4.2 (bit 20 of ECX)
+        if (ecx & (1 << 20)) {
+            return SIMD_SSE42;
+        }
+        // Check for SSE2 (bit 26 of EDX)
+        if (edx & (1 << 26)) {
+            return SIMD_SSE2;
+        }
+    }
+#endif  // OJ_HAS_BUILTIN_CPU_SUPPORTS
+
+#endif  // _MSC_VER vs GCC/Clang
+
+#endif  // HAVE_SIMD_X86
+
+#ifdef HAVE_SIMD_NEON
+    // ARM NEON is always available on ARM64 and detected at compile time
+    return SIMD_NEON;
+#endif
+
+    return SIMD_NONE;
+}
+
+void Init_oj(void) {
+    int err = 0;
+
+#if HAVE_RB_EXT_RACTOR_SAFE
+    rb_ext_ractor_safe(true);
+#endif
+    Oj = rb_define_module("Oj");
+    rb_gc_register_address(&Oj);
+
+    oj_cstack_class = rb_define_class_under(Oj, "CStack", rb_cObject);
+    rb_gc_register_address(&oj_cstack_class);
+
+    rb_undef_alloc_func(oj_cstack_class);
+
+    oj_string_writer_init();
+    oj_stream_writer_init();
+
+    rb_require("date");
+    // On Rubinius the require fails but can be done from a ruby file.
+    rb_protect(protect_require, Qnil, &err);
+    rb_require("stringio");
+    oj_utf8_encoding_index = rb_enc_find_index("UTF-8");
+    oj_utf8_encoding       = rb_enc_from_index(oj_utf8_encoding_index);
+
+    // rb_define_module_function(Oj, "hash_test", hash_test, 0);
+    rb_define_module_function(Oj, "debug_odd", debug_odd, 1);
+
+    rb_define_module_function(Oj, "default_options", get_def_opts, 0);
+    rb_define_module_function(Oj, "default_options=", set_def_opts, 1);
+
+    rb_define_module_function(Oj, "mimic_JSON", oj_define_mimic_json, -1);
+    rb_define_module_function(Oj, "load", load, -1);
+    rb_define_module_function(Oj, "load_file", load_file, -1);
+    rb_define_module_function(Oj, "safe_load", safe_load, 1);
+    rb_define_module_function(Oj, "strict_load", oj_strict_parse, -1);
+    rb_define_module_function(Oj, "compat_load", oj_compat_parse, -1);
+    rb_define_module_function(Oj, "object_load", oj_object_parse, -1);
+    rb_define_module_function(Oj, "wab_load", oj_wab_parse, -1);
+
+    rb_define_module_function(Oj, "dump", dump, -1);
+
+    rb_define_module_function(Oj, "to_file", to_file, -1);
+    rb_define_module_function(Oj, "to_stream", to_stream, -1);
+    // JSON gem compatibility
+    rb_define_module_function(Oj, "to_json", to_json, -1);
+    rb_define_module_function(Oj, "generate", oj_mimic_generate, -1);
+    rb_define_module_function(Oj, "fast_generate", oj_mimic_generate, -1);
+
+    rb_define_module_function(Oj, "add_to_json", oj_add_to_json, -1);
+    rb_define_module_function(Oj, "remove_to_json", oj_remove_to_json, -1);
+
+    rb_define_module_function(Oj, "register_odd", register_odd, -1);
+    rb_define_module_function(Oj, "register_odd_raw", register_odd_raw, -1);
+
+    rb_define_module_function(Oj, "saj_parse", oj_saj_parse, -1);
+    rb_define_module_function(Oj, "sc_parse", oj_sc_parse, -1);
+
+    rb_define_module_function(Oj, "optimize_rails", oj_optimize_rails, 0);
+
+    rb_define_module_function(Oj, "mem_report", mem_report, 0);
+
+    oj_add_value_id    = rb_intern("add_value");
+    oj_array_append_id = rb_intern("array_append");
+    oj_array_end_id    = rb_intern("array_end");
+    oj_array_start_id  = rb_intern("array_start");
+    oj_as_json_id      = rb_intern("as_json");
+    oj_begin_id        = rb_intern("begin");
+    oj_bigdecimal_id   = rb_intern("BigDecimal");
+    oj_end_id          = rb_intern("end");
+    oj_eofq_id         = rb_intern("eof?");
+    oj_error_id        = rb_intern("error");
+    oj_exclude_end_id  = rb_intern("exclude_end?");
+    oj_file_id         = rb_intern("file?");
+    oj_fileno_id       = rb_intern("fileno");
+    oj_ftype_id        = rb_intern("ftype");
+    oj_hash_end_id     = rb_intern("hash_end");
+    oj_hash_key_id     = rb_intern("hash_key");
+    oj_hash_set_id     = rb_intern("hash_set");
+    oj_hash_start_id   = rb_intern("hash_start");
+    oj_iconv_id        = rb_intern("iconv");
+    oj_json_create_id  = rb_intern("json_create");
+    oj_length_id       = rb_intern("length");
+    oj_new_id          = rb_intern("new");
+    oj_parse_id        = rb_intern("parse");
+    oj_plus_id         = rb_intern("+");
+    oj_pos_id          = rb_intern("pos");
+    oj_raw_json_id     = rb_intern("raw_json");
+    oj_read_id         = rb_intern("read");
+    oj_readpartial_id  = rb_intern("readpartial");
+    oj_replace_id      = rb_intern("replace");
+    oj_stat_id         = rb_intern("stat");
+    oj_string_id       = rb_intern("string");
+    oj_to_h_id         = rb_intern("to_h");
+    oj_to_hash_id      = rb_intern("to_hash");
+    oj_to_json_id      = rb_intern("to_json");
+    oj_to_s_id         = rb_intern("to_s");
+    oj_to_sym_id       = rb_intern("to_sym");
+    oj_to_time_id      = rb_intern("to_time");
+    oj_tv_nsec_id      = rb_intern("tv_nsec");
+    oj_tv_sec_id       = rb_intern("tv_sec");
+    oj_tv_usec_id      = rb_intern("tv_usec");
+    oj_utc_id          = rb_intern("utc");
+    oj_utc_offset_id   = rb_intern("utc_offset");
+    oj_utcq_id         = rb_intern("utc?");
+    oj_write_id        = rb_intern("write");
+
+    rb_require("oj/bag");
+    rb_require("oj/error");
+    rb_require("oj/mimic");
+    rb_require("oj/saj");
+    rb_require("oj/schandler");
+
+    oj_bag_class = rb_const_get_at(Oj, rb_intern("Bag"));
+    rb_gc_register_mark_object(oj_bag_class);
+    oj_bigdecimal_class = rb_const_get(rb_cObject, rb_intern("BigDecimal"));
+    rb_gc_register_mark_object(oj_bigdecimal_class);
+    oj_date_class = rb_const_get(rb_cObject, rb_intern("Date"));
+    rb_gc_register_mark_object(oj_date_class);
+    oj_datetime_class = rb_const_get(rb_cObject, rb_intern("DateTime"));
+    rb_gc_register_mark_object(oj_datetime_class);
+    oj_enumerable_class = rb_const_get(rb_cObject, rb_intern("Enumerable"));
+    rb_gc_register_mark_object(oj_enumerable_class);
+    oj_parse_error_class = rb_const_get_at(Oj, rb_intern("ParseError"));
+    rb_gc_register_mark_object(oj_parse_error_class);
+    oj_stringio_class = rb_const_get(rb_cObject, rb_intern("StringIO"));
+    rb_gc_register_mark_object(oj_stringio_class);
+    oj_struct_class = rb_const_get(rb_cObject, rb_intern("Struct"));
+    rb_gc_register_mark_object(oj_struct_class);
+    oj_json_parser_error_class    = rb_eEncodingError;  // replaced if mimic is called
+    oj_json_generator_error_class = rb_eEncodingError;  // replaced if mimic is called
+
+    allow_blank_sym = ID2SYM(rb_intern("allow_blank"));
+    rb_gc_register_address(&allow_blank_sym);
+    allow_gc_sym = ID2SYM(rb_intern("allow_gc"));
+    rb_gc_register_address(&allow_gc_sym);
+    allow_invalid_unicode_sym = ID2SYM(rb_intern("allow_invalid_unicode"));
+    rb_gc_register_address(&allow_invalid_unicode_sym);
+    ascii_sym = ID2SYM(rb_intern("ascii"));
+    rb_gc_register_address(&ascii_sym);
+    auto_define_sym = ID2SYM(rb_intern("auto_define"));
+    rb_gc_register_address(&auto_define_sym);
+    auto_sym = ID2SYM(rb_intern("auto"));
+    rb_gc_register_address(&auto_sym);
+    bigdecimal_as_decimal_sym = ID2SYM(rb_intern("bigdecimal_as_decimal"));
+    rb_gc_register_address(&bigdecimal_as_decimal_sym);
+    bigdecimal_load_sym = ID2SYM(rb_intern("bigdecimal_load"));
+    rb_gc_register_address(&bigdecimal_load_sym);
+    bigdecimal_sym = ID2SYM(rb_intern("bigdecimal"));
+    rb_gc_register_address(&bigdecimal_sym);
+    cache_keys_sym = ID2SYM(rb_intern("cache_keys"));
+    rb_gc_register_address(&cache_keys_sym);
+    cache_str_sym = ID2SYM(rb_intern("cache_str"));
+    rb_gc_register_address(&cache_str_sym);
+    cache_string_sym = ID2SYM(rb_intern("cache_string"));
+    rb_gc_register_address(&cache_string_sym);
+    circular_sym = ID2SYM(rb_intern("circular"));
+    rb_gc_register_address(&circular_sym);
+    class_cache_sym = ID2SYM(rb_intern("class_cache"));
+    rb_gc_register_address(&class_cache_sym);
+    compat_bigdecimal_sym = ID2SYM(rb_intern("compat_bigdecimal"));
+    rb_gc_register_address(&compat_bigdecimal_sym);
+    compat_sym = ID2SYM(rb_intern("compat"));
+    rb_gc_register_address(&compat_sym);
+    create_id_sym = ID2SYM(rb_intern("create_id"));
+    rb_gc_register_address(&create_id_sym);
+    custom_sym = ID2SYM(rb_intern("custom"));
+    rb_gc_register_address(&custom_sym);
+    empty_string_sym = ID2SYM(rb_intern("empty_string"));
+    rb_gc_register_address(&empty_string_sym);
+    escape_mode_sym = ID2SYM(rb_intern("escape_mode"));
+    rb_gc_register_address(&escape_mode_sym);
+    integer_range_sym = ID2SYM(rb_intern("integer_range"));
+    rb_gc_register_address(&integer_range_sym);
+    max_integer_digits_sym = ID2SYM(rb_intern("max_integer_digits"));
+    rb_gc_register_address(&max_integer_digits_sym);
+    fast_sym = ID2SYM(rb_intern("fast"));
+    rb_gc_register_address(&fast_sym);
+    float_format_sym = ID2SYM(rb_intern("float_format"));
+    rb_gc_register_address(&float_format_sym);
+    float_prec_sym = ID2SYM(rb_intern("float_precision"));
+    rb_gc_register_address(&float_prec_sym);
+    float_sym = ID2SYM(rb_intern("float"));
+    rb_gc_register_address(&float_sym);
+    huge_sym = ID2SYM(rb_intern("huge"));
+    rb_gc_register_address(&huge_sym);
+    ignore_sym = ID2SYM(rb_intern("ignore"));
+    rb_gc_register_address(&ignore_sym);
+    ignore_under_sym = ID2SYM(rb_intern("ignore_under"));
+    rb_gc_register_address(&ignore_under_sym);
+    json_sym = ID2SYM(rb_intern("json"));
+    rb_gc_register_address(&json_sym);
+    match_string_sym = ID2SYM(rb_intern("match_string"));
+    rb_gc_register_address(&match_string_sym);
+    mode_sym = ID2SYM(rb_intern("mode"));
+    rb_gc_register_address(&mode_sym);
+    nan_sym = ID2SYM(rb_intern("nan"));
+    rb_gc_register_address(&nan_sym);
+    newline_sym = ID2SYM(rb_intern("newline"));
+    rb_gc_register_address(&newline_sym);
+    nilnil_sym = ID2SYM(rb_intern("nilnil"));
+    rb_gc_register_address(&nilnil_sym);
+    null_sym = ID2SYM(rb_intern("null"));
+    rb_gc_register_address(&null_sym);
+    object_sym = ID2SYM(rb_intern("object"));
+    rb_gc_register_address(&object_sym);
+    oj_allow_nan_sym = ID2SYM(rb_intern("allow_nan"));
+    rb_gc_register_address(&oj_allow_nan_sym);
+    oj_array_class_sym = ID2SYM(rb_intern("array_class"));
+    rb_gc_register_address(&oj_array_class_sym);
+    oj_array_nl_sym = ID2SYM(rb_intern("array_nl"));
+    rb_gc_register_address(&oj_array_nl_sym);
+    oj_ascii_only_sym = ID2SYM(rb_intern("ascii_only"));
+    rb_gc_register_address(&oj_ascii_only_sym);
+    oj_create_additions_sym = ID2SYM(rb_intern("create_additions"));
+    rb_gc_register_address(&oj_create_additions_sym);
+    oj_decimal_class_sym = ID2SYM(rb_intern("decimal_class"));
+    rb_gc_register_address(&oj_decimal_class_sym);
+    oj_hash_class_sym = ID2SYM(rb_intern("hash_class"));
+    rb_gc_register_address(&oj_hash_class_sym);
+    oj_indent_sym = ID2SYM(rb_intern("indent"));
+    rb_gc_register_address(&oj_indent_sym);
+    oj_max_nesting_sym = ID2SYM(rb_intern("max_nesting"));
+    rb_gc_register_address(&oj_max_nesting_sym);
+    oj_object_class_sym = ID2SYM(rb_intern("object_class"));
+    rb_gc_register_address(&oj_object_class_sym);
+    oj_object_nl_sym = ID2SYM(rb_intern("object_nl"));
+    rb_gc_register_address(&oj_object_nl_sym);
+    oj_quirks_mode_sym = ID2SYM(rb_intern("quirks_mode"));
+    rb_gc_register_address(&oj_quirks_mode_sym);
+    oj_safe_sym = ID2SYM(rb_intern("safe"));
+    rb_gc_register_address(&oj_safe_sym);
+    omit_null_byte_sym = ID2SYM(rb_intern("omit_null_byte"));
+    rb_gc_register_address(&omit_null_byte_sym);
+    oj_space_before_sym = ID2SYM(rb_intern("space_before"));
+    rb_gc_register_address(&oj_space_before_sym);
+    oj_space_sym = ID2SYM(rb_intern("space"));
+    rb_gc_register_address(&oj_space_sym);
+    oj_trace_sym = ID2SYM(rb_intern("trace"));
+    rb_gc_register_address(&oj_trace_sym);
+    omit_nil_sym = ID2SYM(rb_intern("omit_nil"));
+    rb_gc_register_address(&omit_nil_sym);
+    rails_sym = ID2SYM(rb_intern("rails"));
+    rb_gc_register_address(&rails_sym);
+    raise_sym = ID2SYM(rb_intern("raise"));
+    rb_gc_register_address(&raise_sym);
+    ruby_sym = ID2SYM(rb_intern("ruby"));
+    rb_gc_register_address(&ruby_sym);
+    sec_prec_sym = ID2SYM(rb_intern("second_precision"));
+    rb_gc_register_address(&sec_prec_sym);
+    slash_sym = ID2SYM(rb_intern("slash"));
+    rb_gc_register_address(&slash_sym);
+    strict_sym = ID2SYM(rb_intern("strict"));
+    rb_gc_register_address(&strict_sym);
+    symbol_keys_sym = ID2SYM(rb_intern("symbol_keys"));
+    rb_gc_register_address(&symbol_keys_sym);
+    oj_symbolize_names_sym = ID2SYM(rb_intern("symbolize_names"));
+    rb_gc_register_address(&oj_symbolize_names_sym);
+    time_format_sym = ID2SYM(rb_intern("time_format"));
+    rb_gc_register_address(&time_format_sym);
+    unicode_xss_sym = ID2SYM(rb_intern("unicode_xss"));
+    rb_gc_register_address(&unicode_xss_sym);
+    unix_sym = ID2SYM(rb_intern("unix"));
+    rb_gc_register_address(&unix_sym);
+    unix_zone_sym = ID2SYM(rb_intern("unix_zone"));
+    rb_gc_register_address(&unix_zone_sym);
+    use_as_json_sym = ID2SYM(rb_intern("use_as_json"));
+    rb_gc_register_address(&use_as_json_sym);
+    use_raw_json_sym = ID2SYM(rb_intern("use_raw_json"));
+    rb_gc_register_address(&use_raw_json_sym);
+    use_to_hash_sym = ID2SYM(rb_intern("use_to_hash"));
+    rb_gc_register_address(&use_to_hash_sym);
+    use_to_json_sym = ID2SYM(rb_intern("use_to_json"));
+    rb_gc_register_address(&use_to_json_sym);
+    wab_sym = ID2SYM(rb_intern("wab"));
+    rb_gc_register_address(&wab_sym);
+    word_sym = ID2SYM(rb_intern("word"));
+    rb_gc_register_address(&word_sym);
+    xmlschema_sym = ID2SYM(rb_intern("xmlschema"));
+    rb_gc_register_address(&xmlschema_sym);
+    xss_safe_sym = ID2SYM(rb_intern("xss_safe"));
+    rb_gc_register_address(&xss_safe_sym);
+    only_sym = ID2SYM(rb_intern("only"));
+    rb_gc_register_address(&only_sym);
+    except_sym = ID2SYM(rb_intern("except"));
+    rb_gc_register_address(&except_sym);
+
+    oj_slash_string = rb_str_new2("/");
+    rb_gc_register_address(&oj_slash_string);
+    OBJ_FREEZE(oj_slash_string);
+
+    oj_default_options.mode = ObjectMode;
+
+    oj_hash_init();
+    oj_odd_init();
+    oj_mimic_rails_init();
+
+#ifdef HAVE_PTHREAD_MUTEX_INIT
+    if (0 != (err = pthread_mutex_init(&oj_cache_mutex, 0))) {
+        rb_raise(rb_eException, "failed to initialize a mutex. %s", strerror(err));
+    }
+#else
+    oj_cache_mutex = rb_mutex_new();
+    rb_gc_register_address(&oj_cache_mutex);
+#endif
+    oj_init_doc();
+
+    SIMD_Impl = oj_get_simd_implementation();
+
+    oj_parser_init();
+    oj_scanner_init();
+
+#ifdef HAVE_SIMD_NEON
+    initialize_neon();
+#endif /* HAVE_SIMD_NEON */
+
+#ifdef HAVE_SIMD_SSE4_2
+    if (SIMD_Impl == SIMD_SSE42) {
+        initialize_sse42();
+    }
+#endif /* HAVE_SIMD_SSE4_2 */
+}