X-Git-Url: http://review.tizen.org/git/?a=blobdiff_plain;f=src%2Fload.c;h=f5cbf685900403806ab79b511be3e34f6a89ea46;hb=bef87fc2581bf2ca6bdacecc0673badb749dcd7f;hp=bd6bac6a673fa2a753f3eb23f733a69894958709;hpb=4c414bdd6d1038195bb6a125fc86eb68d67857e3;p=profile%2Fivi%2Fjansson.git

diff --git a/src/load.c b/src/load.c
index bd6bac6..f5cbf68 100644
--- a/src/load.c
+++ b/src/load.c
@@ -1,15 +1,26 @@
+/*
+ * Copyright (c) 2009-2011 Petri Lehtinen <petri@digip.org>
+ *
+ * Jansson is free software; you can redistribute it and/or modify
+ * it under the terms of the MIT license. See LICENSE for details.
+ */
+
 #define _GNU_SOURCE
-#include <ctype.h>
 #include <errno.h>
+#include <limits.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
-#include <stdarg.h>
-#include <unistd.h>
 #include <assert.h>
 
 #include <jansson.h>
+#include "jansson_private.h"
 #include "strbuffer.h"
+#include "utf.h"
+
+#define STREAM_STATE_OK        0
+#define STREAM_STATE_EOF      -1
+#define STREAM_STATE_ERROR    -2
 
 #define TOKEN_INVALID         -1
 #define TOKEN_EOF              0
@@ -20,18 +31,43 @@
 #define TOKEN_FALSE          260
 #define TOKEN_NULL           261
 
+/* Locale independent versions of isxxx() functions */
+#define l_isupper(c)  ('A' <= c && c <= 'Z')
+#define l_islower(c)  ('a' <= c && c <= 'z')
+#define l_isalpha(c)  (l_isupper(c) || l_islower(c))
+#define l_isdigit(c)  ('0' <= c && c <= '9')
+#define l_isxdigit(c) \
+    (l_isdigit(c) || 'A' <= c || c <= 'F' || 'a' <= c || c <= 'f')
+
+/* Read one byte from stream, convert to unsigned char, then int, and
+   return. return EOF on end of file. This corresponds to the
+   behaviour of fgetc(). */
+typedef int (*get_func)(void *data);
+
 typedef struct {
-    const char *input;
-    const char *start;
+    get_func get;
+    void *data;
+    char buffer[5];
+    int buffer_pos;
+    int state;
+    int line;
+    int column, last_column;
+    size_t position;
+} stream_t;
+
+typedef struct {
+    stream_t stream;
+    strbuffer_t saved_text;
     int token;
-    int line, column;
     union {
         char *string;
-        int integer;
+        json_int_t integer;
         double real;
     } value;
 } lex_t;
 
+#define stream_to_lex(stream) container_of(stream, lex_t, stream)
+
 
 /*** error reporting ***/
 
@@ -39,80 +75,282 @@ static void error_set(json_error_t *error, const lex_t *lex,
                       const char *msg, ...)
 {
     va_list ap;
-    char text[JSON_ERROR_TEXT_LENGTH];
+    char msg_text[JSON_ERROR_TEXT_LENGTH];
+    char msg_with_context[JSON_ERROR_TEXT_LENGTH];
+
+    int line = -1, col = -1;
+    size_t pos = 0;
+    const char *result = msg_text;
 
     if(!error)
         return;
 
     va_start(ap, msg);
-    vsnprintf(text, JSON_ERROR_TEXT_LENGTH, msg, ap);
+    vsnprintf(msg_text, JSON_ERROR_TEXT_LENGTH, msg, ap);
     va_end(ap);
 
     if(lex)
     {
-        error->line = lex->line;
-        if(*lex->start)
+        const char *saved_text = strbuffer_value(&lex->saved_text);
+
+        line = lex->stream.line;
+        col = lex->stream.column;
+        pos = lex->stream.position;
+
+        if(saved_text && saved_text[0])
         {
-            int n = (int)(lex->input - lex->start);
-            snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
-                     "%s near '%.*s'", text, n, lex->start);
+            if(lex->saved_text.length <= 20) {
+                snprintf(msg_with_context, JSON_ERROR_TEXT_LENGTH,
+                         "%s near '%s'", msg_text, saved_text);
+                result = msg_with_context;
+            }
         }
         else
         {
-            snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
-                     "%s near end of file", text);
+            if(lex->stream.state == STREAM_STATE_ERROR) {
+                /* No context for UTF-8 decoding errors */
+                result = msg_text;
+            }
+            else {
+                snprintf(msg_with_context, JSON_ERROR_TEXT_LENGTH,
+                         "%s near end of file", msg_text);
+                result = msg_with_context;
+            }
         }
     }
-    else
+
+    jsonp_error_set(error, line, col, pos, "%s", result);
+}
+
+
+/*** lexical analyzer ***/
+
+static void
+stream_init(stream_t *stream, get_func get, void *data)
+{
+    stream->get = get;
+    stream->data = data;
+    stream->buffer[0] = '\0';
+    stream->buffer_pos = 0;
+
+    stream->state = STREAM_STATE_OK;
+    stream->line = 1;
+    stream->column = 0;
+    stream->position = 0;
+}
+
+static int stream_get(stream_t *stream, json_error_t *error)
+{
+    int c;
+
+    if(stream->state != STREAM_STATE_OK)
+        return stream->state;
+
+    if(!stream->buffer[stream->buffer_pos])
     {
-        error->line = -1;
-        snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
+        c = stream->get(stream->data);
+        if(c == EOF) {
+            stream->state = STREAM_STATE_EOF;
+            return STREAM_STATE_EOF;
+        }
+
+        stream->buffer[0] = c;
+        stream->buffer_pos = 0;
+
+        if(0x80 <= c && c <= 0xFF)
+        {
+            /* multi-byte UTF-8 sequence */
+            int i, count;
+
+            count = utf8_check_first(c);
+            if(!count)
+                goto out;
+
+            assert(count >= 2);
+
+            for(i = 1; i < count; i++)
+                stream->buffer[i] = stream->get(stream->data);
+
+            if(!utf8_check_full(stream->buffer, count, NULL))
+                goto out;
+
+            stream->buffer[count] = '\0';
+        }
+        else
+            stream->buffer[1] = '\0';
+    }
+
+    c = stream->buffer[stream->buffer_pos++];
+
+    stream->position++;
+    if(c == '\n') {
+        stream->line++;
+        stream->last_column = stream->column;
+        stream->column = 0;
+    }
+    else if(utf8_check_first(c)) {
+        /* track the Unicode character column, so increment only if
+           this is the first character of a UTF-8 sequence */
+        stream->column++;
+    }
+
+    return c;
+
+out:
+    stream->state = STREAM_STATE_ERROR;
+    error_set(error, stream_to_lex(stream), "unable to decode byte 0x%x", c);
+    return STREAM_STATE_ERROR;
+}
+
+static void stream_unget(stream_t *stream, int c)
+{
+    if(c == STREAM_STATE_EOF || c == STREAM_STATE_ERROR)
+        return;
+
+    stream->position--;
+    if(c == '\n') {
+        stream->line--;
+        stream->column = stream->last_column;
     }
+    else if(utf8_check_first(c))
+        stream->column--;
+
+    assert(stream->buffer_pos > 0);
+    stream->buffer_pos--;
+    assert(stream->buffer[stream->buffer_pos] == c);
 }
 
 
-/*** lexical analyzer ***/
+static int lex_get(lex_t *lex, json_error_t *error)
+{
+    return stream_get(&lex->stream, error);
+}
 
-static void lex_scan_string(lex_t *lex)
+static void lex_save(lex_t *lex, int c)
 {
-    /* skip the " */
-    const char *p = lex->input + 1;
+    strbuffer_append_byte(&lex->saved_text, c);
+}
+
+static int lex_get_save(lex_t *lex, json_error_t *error)
+{
+    int c = stream_get(&lex->stream, error);
+    if(c != STREAM_STATE_EOF && c != STREAM_STATE_ERROR)
+        lex_save(lex, c);
+    return c;
+}
+
+static void lex_unget(lex_t *lex, int c)
+{
+    stream_unget(&lex->stream, c);
+}
+
+static void lex_unget_unsave(lex_t *lex, int c)
+{
+    if(c != STREAM_STATE_EOF && c != STREAM_STATE_ERROR) {
+        char d;
+        stream_unget(&lex->stream, c);
+        d = strbuffer_pop(&lex->saved_text);
+        assert(c == d);
+    }
+}
+
+static void lex_save_cached(lex_t *lex)
+{
+    while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
+    {
+        lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
+        lex->stream.buffer_pos++;
+        lex->stream.position++;
+    }
+}
+
+/* assumes that str points to 'u' plus at least 4 valid hex digits */
+static int32_t decode_unicode_escape(const char *str)
+{
+    int i;
+    int32_t value = 0;
+
+    assert(str[0] == 'u');
+
+    for(i = 1; i <= 4; i++) {
+        char c = str[i];
+        value <<= 4;
+        if(l_isdigit(c))
+            value += c - '0';
+        else if(l_islower(c))
+            value += c - 'a' + 10;
+        else if(l_isupper(c))
+            value += c - 'A' + 10;
+        else
+            assert(0);
+    }
+
+    return value;
+}
+
+static void lex_scan_string(lex_t *lex, json_error_t *error)
+{
+    int c;
+    const char *p;
     char *t;
+    int i;
 
+    lex->value.string = NULL;
     lex->token = TOKEN_INVALID;
 
-    while(*p != '"') {
-        if(*p == '\0') {
-            /* unterminated string literal */
+    c = lex_get_save(lex, error);
+
+    while(c != '"') {
+        if(c == STREAM_STATE_ERROR)
+            goto out;
+
+        else if(c == STREAM_STATE_EOF) {
+            error_set(error, lex, "premature end of input");
             goto out;
         }
 
-        if(0 <= *p && *p <= 0x1F) {
+        else if(0 <= c && c <= 0x1F) {
             /* control character */
+            lex_unget_unsave(lex, c);
+            if(c == '\n')
+                error_set(error, lex, "unexpected newline", c);
+            else
+                error_set(error, lex, "control character 0x%x", c);
             goto out;
         }
-        else if(*p == '\\') {
-            p++;
-            if(*p == 'u') {
-                p++;
-                for(int i = 0; i < 4; i++, p++) {
-                    if(!isxdigit(*p))
+
+        else if(c == '\\') {
+            c = lex_get_save(lex, error);
+            if(c == 'u') {
+                c = lex_get_save(lex, error);
+                for(i = 0; i < 4; i++) {
+                    if(!l_isxdigit(c)) {
+                        error_set(error, lex, "invalid escape");
                         goto out;
+                    }
+                    c = lex_get_save(lex, error);
                 }
             }
-            else if(*p == '"' || *p == '\\' || *p == '/' || *p == 'b' ||
-                    *p == 'f' || *p == 'n' || *p == 'r' || *p == 't')
-                p++;
-            else
+            else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
+                    c == 'f' || c == 'n' || c == 'r' || c == 't')
+                c = lex_get_save(lex, error);
+            else {
+                error_set(error, lex, "invalid escape");
                 goto out;
+            }
         }
         else
-            p++;
+            c = lex_get_save(lex, error);
     }
 
     /* the actual value is at most of the same length as the source
-       string */
-    lex->value.string = malloc(p - lex->start);
+       string, because:
+         - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
+         - a single \uXXXX escape (length 6) is converted to at most 3 bytes
+         - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
+           are converted to 4 bytes
+    */
+    lex->value.string = jsonp_malloc(lex->saved_text.length + 1);
     if(!lex->value.string) {
         /* this is not very nice, since TOKEN_INVALID is returned */
         goto out;
@@ -121,16 +359,65 @@ static void lex_scan_string(lex_t *lex)
     /* the target */
     t = lex->value.string;
 
-    p = lex->input + 1;
+    /* + 1 to skip the " */
+    p = strbuffer_value(&lex->saved_text) + 1;
+
     while(*p != '"') {
         if(*p == '\\') {
             p++;
             if(*p == 'u') {
-                /* TODO: \uXXXX not supported yet */
-                free(lex->value.string);
-                lex->value.string = NULL;
-                goto out;
-            } else {
+                char buffer[4];
+                int length;
+                int32_t value;
+
+                value = decode_unicode_escape(p);
+                p += 5;
+
+                if(0xD800 <= value && value <= 0xDBFF) {
+                    /* surrogate pair */
+                    if(*p == '\\' && *(p + 1) == 'u') {
+                        int32_t value2 = decode_unicode_escape(++p);
+                        p += 5;
+
+                        if(0xDC00 <= value2 && value2 <= 0xDFFF) {
+                            /* valid second surrogate */
+                            value =
+                                ((value - 0xD800) << 10) +
+                                (value2 - 0xDC00) +
+                                0x10000;
+                        }
+                        else {
+                            /* invalid second surrogate */
+                            error_set(error, lex,
+                                      "invalid Unicode '\\u%04X\\u%04X'",
+                                      value, value2);
+                            goto out;
+                        }
+                    }
+                    else {
+                        /* no second surrogate */
+                        error_set(error, lex, "invalid Unicode '\\u%04X'",
+                                  value);
+                        goto out;
+                    }
+                }
+                else if(0xDC00 <= value && value <= 0xDFFF) {
+                    error_set(error, lex, "invalid Unicode '\\u%04X'", value);
+                    goto out;
+                }
+                else if(value == 0)
+                {
+                    error_set(error, lex, "\\u0000 is not allowed");
+                    goto out;
+                }
+
+                if(utf8_encode(value, buffer, &length))
+                    assert(0);
+
+                memcpy(t, buffer, length);
+                t += length;
+            }
+            else {
                 switch(*p) {
                     case '"': case '\\': case '/':
                         *t = *p; break;
@@ -141,222 +428,284 @@ static void lex_scan_string(lex_t *lex)
                     case 't': *t = '\t'; break;
                     default: assert(0);
                 }
+                t++;
+                p++;
             }
         }
         else
-            *t = *p;
-
-        t++;
-        p++;
+            *(t++) = *(p++);
     }
-    /* skip the " */
-    p++;
-
     *t = '\0';
     lex->token = TOKEN_STRING;
+    return;
 
 out:
-    lex->input = p;
+    jsonp_free(lex->value.string);
 }
 
-static void lex_scan_number(lex_t *lex)
+#if JSON_INTEGER_IS_LONG_LONG
+#define json_strtoint     strtoll
+#else
+#define json_strtoint     strtol
+#endif
+
+static int lex_scan_number(lex_t *lex, int c, json_error_t *error)
 {
-    const char *p = lex->input;
+    const char *saved_text;
     char *end;
+    double value;
 
     lex->token = TOKEN_INVALID;
 
-    if(*p == '-')
-        p++;
+    if(c == '-')
+        c = lex_get_save(lex, error);
 
-    if(*p == '0') {
-        p++;
-        if(isdigit(*p))
-          goto out;
+    if(c == '0') {
+        c = lex_get_save(lex, error);
+        if(l_isdigit(c)) {
+            lex_unget_unsave(lex, c);
+            goto out;
+        }
     }
-    else /* *p != '0' */ {
-        p++;
-        while(isdigit(*p))
-            p++;
+    else if(l_isdigit(c)) {
+        c = lex_get_save(lex, error);
+        while(l_isdigit(c))
+            c = lex_get_save(lex, error);
+    }
+    else {
+        lex_unget_unsave(lex, c);
+        goto out;
     }
 
-    if(*p != '.' && *p != 'E' && *p != 'e') {
-        lex->token = TOKEN_INTEGER;
+    if(c != '.' && c != 'E' && c != 'e') {
+        json_int_t value;
 
-        lex->value.integer = strtol(lex->start, &end, 10);
-        assert(end == p);
+        lex_unget_unsave(lex, c);
 
-        goto out;
+        saved_text = strbuffer_value(&lex->saved_text);
+
+        errno = 0;
+        value = json_strtoint(saved_text, &end, 10);
+        if(errno == ERANGE) {
+            if(value < 0)
+                error_set(error, lex, "too big negative integer");
+            else
+                error_set(error, lex, "too big integer");
+            goto out;
+        }
+
+        assert(end == saved_text + lex->saved_text.length);
+
+        lex->token = TOKEN_INTEGER;
+        lex->value.integer = value;
+        return 0;
     }
 
-    if(*p == '.') {
-        p++;
-        if(!isdigit(*p))
+    if(c == '.') {
+        c = lex_get(lex, error);
+        if(!l_isdigit(c)) {
+            lex_unget(lex, c);
             goto out;
+        }
+        lex_save(lex, c);
 
-        p++;
-        while(isdigit(*p))
-            p++;
+        c = lex_get_save(lex, error);
+        while(l_isdigit(c))
+            c = lex_get_save(lex, error);
     }
 
-    if(*p == 'E' || *p == 'e') {
-        p++;
-        if(*p == '+' || *p == '-')
-            p++;
+    if(c == 'E' || c == 'e') {
+        c = lex_get_save(lex, error);
+        if(c == '+' || c == '-')
+            c = lex_get_save(lex, error);
 
-        if(!isdigit(*p))
+        if(!l_isdigit(c)) {
+            lex_unget_unsave(lex, c);
             goto out;
+        }
 
-        p++;
-        while(isdigit(*p))
-            p++;
+        c = lex_get_save(lex, error);
+        while(l_isdigit(c))
+            c = lex_get_save(lex, error);
     }
 
-    lex->token = TOKEN_REAL;
+    lex_unget_unsave(lex, c);
+
+    if(jsonp_strtod(&lex->saved_text, &value)) {
+        error_set(error, lex, "real number overflow");
+        goto out;
+    }
 
-    lex->value.real = strtod(lex->start, &end);
-    assert(end == p);
+    lex->token = TOKEN_REAL;
+    lex->value.real = value;
+    return 0;
 
 out:
-    lex->input = p;
+    return -1;
 }
 
-static int lex_scan(lex_t *lex)
+static int lex_scan(lex_t *lex, json_error_t *error)
 {
-    char c;
+    int c;
+
+    strbuffer_clear(&lex->saved_text);
 
     if(lex->token == TOKEN_STRING) {
-      free(lex->value.string);
-      lex->value.string = NULL;
+        jsonp_free(lex->value.string);
+        lex->value.string = NULL;
     }
 
-    c = *lex->input;
+    c = lex_get(lex, error);
     while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
-    {
-        if(c == '\n')
-            lex->line++;
+        c = lex_get(lex, error);
 
-        lex->input++;
-        c = *lex->input;
+    if(c == STREAM_STATE_EOF) {
+        lex->token = TOKEN_EOF;
+        goto out;
     }
 
-    lex->start = lex->input;
-    c = *lex->input;
+    if(c == STREAM_STATE_ERROR) {
+        lex->token = TOKEN_INVALID;
+        goto out;
+    }
 
-    if(c == '\0')
-        lex->token = TOKEN_EOF;
+    lex_save(lex, c);
 
-    else if(c == '{' || c == '}' || c == '[' || c == ']' ||
-            c == ':' || c == ',')
-    {
+    if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
         lex->token = c;
-        lex->input++;
-    }
 
     else if(c == '"')
-        lex_scan_string(lex);
+        lex_scan_string(lex, error);
 
-    else if(isdigit(c) || c == '-')
-        lex_scan_number(lex);
+    else if(l_isdigit(c) || c == '-') {
+        if(lex_scan_number(lex, c, error))
+            goto out;
+    }
 
-    else if(isupper(c) || islower(c)) {
+    else if(l_isalpha(c)) {
         /* eat up the whole identifier for clearer error messages */
-        int len;
+        const char *saved_text;
+
+        c = lex_get_save(lex, error);
+        while(l_isalpha(c))
+            c = lex_get_save(lex, error);
+        lex_unget_unsave(lex, c);
 
-        while(isupper(*lex->input) || islower(*lex->input))
-            lex->input++;
-        len = lex->input - lex->start;
+        saved_text = strbuffer_value(&lex->saved_text);
 
-        if(strncmp(lex->start, "true", len) == 0)
+        if(strcmp(saved_text, "true") == 0)
             lex->token = TOKEN_TRUE;
-        else if(strncmp(lex->start, "false", len) == 0)
+        else if(strcmp(saved_text, "false") == 0)
             lex->token = TOKEN_FALSE;
-        else if(strncmp(lex->start, "null", len) == 0)
+        else if(strcmp(saved_text, "null") == 0)
             lex->token = TOKEN_NULL;
         else
             lex->token = TOKEN_INVALID;
     }
 
     else {
+        /* save the rest of the input UTF-8 sequence to get an error
+           message of valid UTF-8 */
+        lex_save_cached(lex);
         lex->token = TOKEN_INVALID;
-        lex->input++;
     }
 
+out:
     return lex->token;
 }
 
-static int lex_init(lex_t *lex, const char *input)
+static char *lex_steal_string(lex_t *lex)
 {
-    lex->input = input;
-    lex->token = TOKEN_INVALID;
-    lex->line = 1;
+    char *result = NULL;
+    if(lex->token == TOKEN_STRING)
+    {
+        result = lex->value.string;
+        lex->value.string = NULL;
+    }
+    return result;
+}
+
+static int lex_init(lex_t *lex, get_func get, void *data)
+{
+    stream_init(&lex->stream, get, data);
+    if(strbuffer_init(&lex->saved_text))
+        return -1;
 
-    lex_scan(lex);
+    lex->token = TOKEN_INVALID;
     return 0;
 }
 
 static void lex_close(lex_t *lex)
 {
     if(lex->token == TOKEN_STRING)
-        free(lex->value.string);
+        jsonp_free(lex->value.string);
+    strbuffer_close(&lex->saved_text);
 }
 
 
 /*** parser ***/
 
-static json_t *parse_value(lex_t *lex, json_error_t *error);
+static json_t *parse_value(lex_t *lex, size_t flags, json_error_t *error);
 
-static json_t *parse_object(lex_t *lex, json_error_t *error)
+static json_t *parse_object(lex_t *lex, size_t flags, json_error_t *error)
 {
     json_t *object = json_object();
     if(!object)
         return NULL;
 
-    lex_scan(lex);
+    lex_scan(lex, error);
     if(lex->token == '}')
         return object;
 
-    while(lex->token) {
+    while(1) {
         char *key;
         json_t *value;
 
         if(lex->token != TOKEN_STRING) {
-            error_set(error, lex, "string expected");
+            error_set(error, lex, "string or '}' expected");
             goto error;
         }
 
-        key = strdup(lex->value.string);
+        key = lex_steal_string(lex);
         if(!key)
             return NULL;
 
-        lex_scan(lex);
+        if(flags & JSON_REJECT_DUPLICATES) {
+            if(json_object_get(object, key)) {
+                jsonp_free(key);
+                error_set(error, lex, "duplicate object key");
+                goto error;
+            }
+        }
+
+        lex_scan(lex, error);
         if(lex->token != ':') {
-            free(key);
+            jsonp_free(key);
             error_set(error, lex, "':' expected");
             goto error;
         }
 
-        lex_scan(lex);
-
-        value = parse_value(lex, error);
+        lex_scan(lex, error);
+        value = parse_value(lex, flags, error);
         if(!value) {
-            free(key);
+            jsonp_free(key);
             goto error;
         }
 
-        if(json_object_set(object, key, value)) {
-            free(key);
+        if(json_object_set_nocheck(object, key, value)) {
+            jsonp_free(key);
             json_decref(value);
             goto error;
         }
 
         json_decref(value);
-        free(key);
+        jsonp_free(key);
 
+        lex_scan(lex, error);
         if(lex->token != ',')
             break;
 
-        lex_scan(lex);
+        lex_scan(lex, error);
     }
 
     if(lex->token != '}') {
@@ -371,18 +720,18 @@ error:
     return NULL;
 }
 
-static json_t *parse_array(lex_t *lex, json_error_t *error)
+static json_t *parse_array(lex_t *lex, size_t flags, json_error_t *error)
 {
     json_t *array = json_array();
     if(!array)
         return NULL;
 
-    lex_scan(lex);
+    lex_scan(lex, error);
     if(lex->token == ']')
         return array;
 
     while(lex->token) {
-        json_t *elem = parse_value(lex, error);
+        json_t *elem = parse_value(lex, flags, error);
         if(!elem)
             goto error;
 
@@ -392,13 +741,13 @@ static json_t *parse_array(lex_t *lex, json_error_t *error)
         }
         json_decref(elem);
 
+        lex_scan(lex, error);
         if(lex->token != ',')
             break;
 
-        lex_scan(lex);
+        lex_scan(lex, error);
     }
 
-
     if(lex->token != ']') {
         error_set(error, lex, "']' expected");
         goto error;
@@ -411,13 +760,13 @@ error:
     return NULL;
 }
 
-static json_t *parse_value(lex_t *lex, json_error_t *error)
+static json_t *parse_value(lex_t *lex, size_t flags, json_error_t *error)
 {
     json_t *json;
 
     switch(lex->token) {
         case TOKEN_STRING: {
-            json = json_string(lex->value.string);
+            json = json_string_nocheck(lex->value.string);
             break;
         }
 
@@ -444,11 +793,11 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
             break;
 
         case '{':
-          json = parse_object(lex, error);
+            json = parse_object(lex, flags, error);
             break;
 
         case '[':
-            json = parse_array(lex, error);
+            json = parse_array(lex, flags, error);
             break;
 
         case TOKEN_INVALID:
@@ -463,93 +812,176 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
     if(!json)
         return NULL;
 
-    lex_scan(lex);
     return json;
 }
 
-json_t *parse_json(lex_t *lex, json_error_t *error)
+static json_t *parse_json(lex_t *lex, size_t flags, json_error_t *error)
 {
-    if(lex->token != '[' && lex->token != '{') {
-        error_set(error, lex, "'[' or '{' expected");
+    json_t *result;
+
+    lex_scan(lex, error);
+    if(!(flags & JSON_DECODE_ANY)) {
+        if(lex->token != '[' && lex->token != '{') {
+            error_set(error, lex, "'[' or '{' expected");
+            return NULL;
+        }
+    }
+
+    result = parse_value(lex, flags, error);
+    if(!result)
         return NULL;
+
+    if(!(flags & JSON_DISABLE_EOF_CHECK)) {
+        lex_scan(lex, error);
+        if(lex->token != TOKEN_EOF) {
+            error_set(error, lex, "end of file expected");
+            json_decref(result);
+            result = NULL;
+        }
     }
 
-    return parse_value(lex, error);
+    return result;
 }
 
-json_t *json_load(const char *path, json_error_t *error)
+typedef struct
 {
-    json_t *result;
-    FILE *fp;
+    const char *data;
+    int pos;
+} string_data_t;
 
-    fp = fopen(path, "r");
-    if(!fp)
+static int string_get(void *data)
+{
+    char c;
+    string_data_t *stream = (string_data_t *)data;
+    c = stream->data[stream->pos];
+    if(c == '\0')
+        return EOF;
+    else
     {
-        error_set(error, NULL, "unable to open %s: %s",
-                       path, strerror(errno));
+        stream->pos++;
+        return (unsigned char)c;
+    }
+}
+
+json_t *json_loads(const char *string, size_t flags, json_error_t *error)
+{
+    lex_t lex;
+    json_t *result;
+    string_data_t stream_data;
+
+    jsonp_error_init(error, "<string>");
+
+    if (string == NULL) {
+        error_set(error, NULL, "wrong arguments");
         return NULL;
     }
 
-    result = json_loadf(fp, error);
+    stream_data.data = string;
+    stream_data.pos = 0;
 
-    fclose(fp);
+    if(lex_init(&lex, string_get, (void *)&stream_data))
+        return NULL;
+
+    result = parse_json(&lex, flags, error);
+
+    lex_close(&lex);
     return result;
 }
 
-json_t *json_loads(const char *string, json_error_t *error)
+typedef struct
+{
+    const char *data;
+    size_t len;
+    size_t pos;
+} buffer_data_t;
+
+static int buffer_get(void *data)
+{
+    char c;
+    buffer_data_t *stream = data;
+    if(stream->pos >= stream->len)
+      return EOF;
+
+    c = stream->data[stream->pos];
+    stream->pos++;
+    return (unsigned char)c;
+}
+
+json_t *json_loadb(const char *buffer, size_t buflen, size_t flags, json_error_t *error)
 {
     lex_t lex;
-    json_t *result = NULL;
+    json_t *result;
+    buffer_data_t stream_data;
+
+    jsonp_error_init(error, "<buffer>");
 
-    if(lex_init(&lex, string))
+    if (buffer == NULL) {
+        error_set(error, NULL, "wrong arguments");
         return NULL;
+    }
 
-    result = parse_json(&lex, error);
-    if(!result)
-        goto out;
+    stream_data.data = buffer;
+    stream_data.pos = 0;
+    stream_data.len = buflen;
 
-    if(lex.token != TOKEN_EOF) {
-        error_set(error, &lex, "end of file expected");
-        json_decref(result);
-        result = NULL;
-    }
+    if(lex_init(&lex, buffer_get, (void *)&stream_data))
+        return NULL;
+
+    result = parse_json(&lex, flags, error);
 
-out:
     lex_close(&lex);
     return result;
 }
 
-#define BUFFER_SIZE 4096
+json_t *json_loadf(FILE *input, size_t flags, json_error_t *error)
+{
+    lex_t lex;
+    const char *source;
+    json_t *result;
+
+    if(input == stdin)
+        source = "<stdin>";
+    else
+        source = "<stream>";
+
+    jsonp_error_init(error, source);
 
-json_t *json_loadf(FILE *input, json_error_t *error)
+    if (input == NULL) {
+        error_set(error, NULL, "wrong arguments");
+        return NULL;
+    }
+
+    if(lex_init(&lex, (get_func)fgetc, input))
+        return NULL;
+
+    result = parse_json(&lex, flags, error);
+
+    lex_close(&lex);
+    return result;
+}
+
+json_t *json_load_file(const char *path, size_t flags, json_error_t *error)
 {
-    strbuffer_t strbuff;
-    char buffer[BUFFER_SIZE];
-    size_t length;
-    json_t *result = NULL;
+    json_t *result;
+    FILE *fp;
+
+    jsonp_error_init(error, path);
 
-    if(strbuffer_init(&strbuff))
-      return NULL;
+    if (path == NULL) {
+        error_set(error, NULL, "wrong arguments");
+        return NULL;
+    }
 
-    while(1)
+    fp = fopen(path, "rb");
+    if(!fp)
     {
-        length = fread(buffer, 1, BUFFER_SIZE, input);
-        if(length == 0)
-        {
-            if(ferror(input))
-            {
-                error_set(error, NULL, "read error");
-                goto out;
-            }
-            break;
-        }
-        if(strbuffer_append_bytes(&strbuff, buffer, length))
-            goto out;
+        error_set(error, NULL, "unable to open %s: %s",
+                  path, strerror(errno));
+        return NULL;
     }
 
-    result = json_loads(strbuffer_value(&strbuff), error);
+    result = json_loadf(fp, flags, error);
 
-out:
-    strbuffer_close(&strbuff);
+    fclose(fp);
     return result;
 }