oj 3.16.0 → 3.17.3

data/ext/oj/oj.c

@@ -18,6 +18,9 @@
 #include "odd.h"
 #include "parse.h"
 #include "rails.h"
+#include "simd.h"
+
+#define MAX_INDENT 16
 
 typedef struct _yesNoOpt {
     VALUE sym;
@@ -33,10 +36,10 @@ ID oj_array_append_id;
 ID oj_array_end_id;
 ID oj_array_start_id;
 ID oj_as_json_id;
-ID oj_at_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;
@@ -51,6 +54,7 @@ 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;
@@ -91,9 +95,7 @@ VALUE oj_array_class_sym;
 VALUE oj_create_additions_sym;
 VALUE oj_decimal_class_sym;
 VALUE oj_hash_class_sym;
-VALUE oj_in_sym;
 VALUE oj_indent_sym;
-VALUE oj_nanosecond_sym;
 VALUE oj_object_class_sym;
 VALUE oj_quirks_mode_sym;
 VALUE oj_safe_sym;
@@ -120,7 +122,9 @@ 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;
@@ -138,6 +142,7 @@ 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;
@@ -167,6 +172,8 @@ pthread_mutex_t oj_cache_mutex;
 VALUE oj_cache_mutex = Qnil;
 #endif
 
+SIMD_Implementation SIMD_Impl = SIMD_NONE;
+
 extern void oj_parser_init();
 
 const char oj_json_class[] = "json_class";
@@ -202,6 +209,7 @@ struct _options oj_default_options = {
     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
@@ -226,6 +234,8 @@ struct _options oj_default_options = {
         false,      // omit_nil
         false,      // omit_null_byte
         MAX_DEPTH,  // max_depth
+        NULL,       // only
+        NULL,       // except
     },
     {
         // str_rx
@@ -236,76 +246,109 @@ struct _options oj_default_options = {
     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
+ * - *: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
+ * - *: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.
+ * - *: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
+ * - *: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
+ * - *: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
+ * - *: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
+ * - *: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.
+ * - *: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)
+ * - *:cache_str* [_Fixnum_] maximum string value length to cache (strings less
+ *   than this are cached)
  * - *:integer_range* [_Range_] Dump integers outside range as strings.
- * - *: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)
+ * - *: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.
  */
@@ -414,6 +457,7 @@ static VALUE get_def_opts(VALUE self) {
     } 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;
@@ -471,6 +515,9 @@ static VALUE get_def_opts(VALUE self) {
     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 {
@@ -490,73 +537,72 @@ static VALUE get_def_opts(VALUE self) {
  *
  * 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.
+ *   - *: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.
+ *   - *: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.
+ *   - *: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.
+ *   - *: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).
+ *   - *: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.
+ *   - *: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.
+ *   - *: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.
+ *   - *: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.
  */
@@ -616,6 +662,95 @@ bool set_yesno_options(VALUE key, VALUE value, Options copts) {
     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;
@@ -633,7 +768,10 @@ static int parse_options_cb(VALUE k, VALUE v, VALUE opts) {
         case T_FIXNUM:
             copts->dump_opts.indent_size = 0;
             *copts->dump_opts.indent_str = '\0';
-            copts->indent                = FIX2INT(v);
+            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))) {
@@ -917,12 +1055,12 @@ static int parse_options_cb(VALUE k, VALUE v, VALUE opts) {
         OJ_R_FREE(copts->ignore);
         copts->ignore = NULL;
         if (Qnil != v) {
-            int cnt;
+            size_t cnt;
 
             rb_check_type(v, T_ARRAY);
-            cnt = (int)RARRAY_LEN(v);
+            cnt = RARRAY_LEN(v);
             if (0 < cnt) {
-                int i;
+                size_t i;
 
                 copts->ignore = OJ_R_ALLOC_N(VALUE, cnt + 1);
                 for (i = 0; i < cnt; i++) {
@@ -948,6 +1086,20 @@ static int parse_options_cb(VALUE k, VALUE v, VALUE opts) {
         } 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;
@@ -967,11 +1119,23 @@ static int parse_options_cb(VALUE k, VALUE v, VALUE opts) {
         }
     } else if (float_format_sym == k) {
         rb_check_type(v, T_STRING);
-        if (6 < (int)RSTRING_LEN(v)) {
+        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;
 }
@@ -1182,7 +1346,7 @@ static VALUE load_file(int argc, VALUE *argv, VALUE self) {
         OJ_FREE(wide_path);
     }
 #else
-    fd             = open(path, O_RDONLY);
+    fd = open(path, O_RDONLY);
 #endif
     if (0 == fd) {
         rb_raise(rb_eIOError, "%s", strerror(errno));
@@ -1271,8 +1435,7 @@ static VALUE dump_body(VALUE a) {
     if (0 == arg->out->buf) {
         rb_raise(rb_eNoMemError, "Not enough memory.");
     }
-    rstr = rb_str_new2(arg->out->buf);
-    rstr = oj_encode(rstr);
+    rstr = rb_utf8_str_new_cstr(arg->out->buf);
 
     return rstr;
 }
@@ -1329,18 +1492,16 @@ static VALUE dump(int argc, VALUE *argv, VALUE self) {
  * 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.
+ *   - *: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.
+ *   - *: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.
@@ -1374,8 +1535,7 @@ static VALUE to_json(int argc, VALUE *argv, VALUE self) {
     if (0 == out.buf) {
         rb_raise(rb_eNoMemError, "Not enough memory.");
     }
-    rstr = rb_str_new2(out.buf);
-    rstr = oj_encode(rstr);
+    rstr = rb_utf8_str_new_cstr(out.buf);
 
     oj_out_free(&out);
 
@@ -1435,10 +1595,10 @@ static VALUE to_stream(int argc, VALUE *argv, VALUE self) {
  *
  * - *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.
+ * - *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) {
@@ -1471,10 +1631,10 @@ static VALUE register_odd(int argc, VALUE *argv, VALUE self) {
  *
  * - *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.
+ * - *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) {
@@ -1700,8 +1860,8 @@ extern VALUE oj_define_mimic_json(int argc, VALUE *argv, VALUE self);
  *   - *: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.
+ *   - *: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.
  */
@@ -1758,8 +1918,8 @@ static VALUE mem_report(VALUE self) {
  * 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.
+ * - *: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.
@@ -1777,6 +1937,78 @@ static VALUE mem_report(VALUE self) {
  *
  * - *: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;
 
@@ -1843,10 +2075,10 @@ void Init_oj(void) {
     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_at_id           = rb_intern("at");
     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?");
@@ -1861,6 +2093,7 @@ void Init_oj(void) {
     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");
@@ -1949,6 +2182,8 @@ void Init_oj(void) {
     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"));
@@ -1993,14 +2228,10 @@ void Init_oj(void) {
     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_in_sym = ID2SYM(rb_intern("in"));
-    rb_gc_register_address(&oj_in_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_nanosecond_sym = ID2SYM(rb_intern("nanosecond"));
-    rb_gc_register_address(&oj_nanosecond_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"));
@@ -2059,6 +2290,10 @@ void Init_oj(void) {
     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);
@@ -2080,6 +2315,18 @@ void Init_oj(void) {
 #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 */
 }