X-Git-Url: http://www.project-moonshot.org/gitweb/?a=blobdiff_plain;f=src%2Fload.c;h=f05230ba41e0d1f2867b3d314a8d47ccd99212a9;hb=5422a862de2354b3419ef628bac5a18c6ef522da;hp=ace963a91f64932a8ffbbcd9ee9b7bdefa743017;hpb=17a69c2d66a86757877c3d4159e999da3a5434bf;p=jansson.git

diff --git a/src/load.c b/src/load.c
index ace963a..f05230b 100644
--- a/src/load.c
+++ b/src/load.c
@@ -1,97 +1,306 @@
+/*
+ * Copyright (c) 2009, 2010 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 <ctype.h>
+#include <stdarg.h>
 #include <assert.h>
 
 #include <jansson.h>
+#include "jansson_private.h"
+#include "strbuffer.h"
+#include "utf.h"
+
+#define TOKEN_INVALID         -1
+#define TOKEN_EOF              0
+#define TOKEN_STRING         256
+#define TOKEN_INTEGER        257
+#define TOKEN_REAL           258
+#define TOKEN_TRUE           259
+#define TOKEN_FALSE          260
+#define TOKEN_NULL           261
 
+/* read one byte from stream, return EOF on end of file */
+typedef int (*get_func)(void *data);
+
+/* return non-zero if end of file has been reached */
+typedef int (*eof_func)(void *data);
+
+typedef struct {
+    get_func get;
+    eof_func eof;
+    void *data;
+    int stream_pos;
+    char buffer[5];
+    int buffer_pos;
+} stream_t;
 
-#define JSON_TOKEN_INVALID         -1
-#define JSON_TOKEN_EOF              0
-#define JSON_TOKEN_STRING         256
-#define JSON_TOKEN_NUMBER         257
-#define JSON_TOKEN_TRUE           258
-#define JSON_TOKEN_FALSE          259
-#define JSON_TOKEN_NULL           260
 
 typedef struct {
-    const char *input;
-    const char *start;
+    stream_t stream;
+    strbuffer_t saved_text;
     int token;
     int line, column;
     union {
         char *string;
-        double number;
+        json_int_t integer;
+        double real;
     } value;
-} json_lex;
+} lex_t;
 
 
 /*** error reporting ***/
 
-static __thread char *json_error_msg = NULL;
+static void error_set(json_error_t *error, const lex_t *lex,
+                      const char *msg, ...)
+{
+    va_list ap;
+    char msg_text[JSON_ERROR_TEXT_LENGTH];
+
+    int line = -1, col = -1;
+    const char *result = msg_text;
+
+    if(!error)
+        return;
+
+    va_start(ap, msg);
+    vsnprintf(msg_text, JSON_ERROR_TEXT_LENGTH, msg, ap);
+    va_end(ap);
+
+    if(lex)
+    {
+        const char *saved_text = strbuffer_value(&lex->saved_text);
+        char msg_with_context[JSON_ERROR_TEXT_LENGTH];
+
+        line = lex->line;
 
-static void json_set_error(const json_lex *lex, const char *msg)
+        if(saved_text && saved_text[0])
+        {
+            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(msg_with_context, JSON_ERROR_TEXT_LENGTH,
+                     "%s near end of file", msg_text);
+            result = msg_with_context;
+        }
+    }
+
+    jsonp_error_set(error, line, col, "%s", result);
+}
+
+
+/*** lexical analyzer ***/
+
+static void
+stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
 {
-    free(json_error_msg);
-    if(*lex->start)
-        asprintf(&json_error_msg, "%s near '%.*s' on line %d", msg,
-                 (int)(lex->input - lex->start), lex->start, lex->line);
-    else
-        asprintf(&json_error_msg, "%s near end of file", msg);
+    stream->get = get;
+    stream->eof = eof;
+    stream->data = data;
+    stream->stream_pos = 0;
+    stream->buffer[0] = '\0';
+    stream->buffer_pos = 0;
 }
 
-const char *json_get_error(void)
+static char stream_get(stream_t *stream, json_error_t *error)
 {
-    if(!json_error_msg)
-        json_error_msg = strdup("success");
-    return json_error_msg;
+    char c;
+
+    if(!stream->buffer[stream->buffer_pos])
+    {
+        stream->buffer[0] = stream->get(stream->data);
+        stream->buffer_pos = 0;
+
+        c = stream->buffer[0];
+
+        if((unsigned char)c >= 0x80 && c != (char)EOF)
+        {
+            /* 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->stream_pos += count;
+            stream->buffer[count] = '\0';
+        }
+        else {
+            stream->buffer[1] = '\0';
+            stream->stream_pos++;
+        }
+    }
+
+    return stream->buffer[stream->buffer_pos++];
+
+out:
+    error_set(error, NULL, "unable to decode byte 0x%x at position %d",
+              (unsigned char)c, stream->stream_pos);
+
+    stream->buffer[0] = EOF;
+    stream->buffer[1] = '\0';
+    stream->buffer_pos = 1;
+
+    return EOF;
+}
+
+static void stream_unget(stream_t *stream, char c)
+{
+    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 json_scan_string(json_lex *lex)
+static int lex_eof(lex_t *lex)
 {
-    /* skip the " */
-    const char *p = lex->input + 1;
+    return lex->stream.eof(lex->stream.data);
+}
+
+static void lex_save(lex_t *lex, char c)
+{
+    strbuffer_append_byte(&lex->saved_text, c);
+}
+
+static int lex_get_save(lex_t *lex, json_error_t *error)
+{
+    char c = stream_get(&lex->stream, error);
+    lex_save(lex, c);
+    return c;
+}
+
+static void lex_unget_unsave(lex_t *lex, char c)
+{
+    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++;
+    }
+}
+
+/* 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(isdigit(c))
+            value += c - '0';
+        else if(islower(c))
+            value += c - 'a' + 10;
+        else if(isupper(c))
+            value += c - 'A' + 10;
+        else
+            assert(0);
+    }
+
+    return value;
+}
+
+static void lex_scan_string(lex_t *lex, json_error_t *error)
+{
+    char c;
+    const char *p;
     char *t;
+    int i;
 
-    lex->token = JSON_TOKEN_INVALID;
+    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 == (char)EOF) {
+            lex_unget_unsave(lex, c);
+            if(lex_eof(lex))
+                error_set(error, lex, "premature end of input");
             goto out;
         }
 
-        if(0 <= *p && *p <= 31) {
+        else if((unsigned char)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(!isxdigit(c)) {
+                        lex_unget_unsave(lex, c);
+                        error_set(error, lex, "invalid escape");
                         goto out;
+                    }
+                    c = lex_get_save(lex, error);
                 }
             }
-            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 {
+                lex_unget_unsave(lex, c);
+                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 = malloc(lex->saved_text.length + 1);
     if(!lex->value.string) {
         /* this is not very nice, since TOKEN_INVALID is returned */
         goto out;
@@ -100,16 +309,65 @@ static void json_scan_string(json_lex *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;
@@ -120,186 +378,271 @@ static void json_scan_string(json_lex *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 = JSON_TOKEN_STRING;
+    lex->token = TOKEN_STRING;
+    return;
 
 out:
-    lex->input = p;
+    free(lex->value.string);
 }
 
-static void json_scan_number(json_lex *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, char c, json_error_t *error)
 {
-    const char *p = lex->input;
+    const char *saved_text;
     char *end;
+    double value;
 
-    lex->token = JSON_TOKEN_INVALID;
+    lex->token = TOKEN_INVALID;
 
-    if(*p == '-')
-        p++;
+    if(c == '-')
+        c = lex_get_save(lex, error);
 
-    if(*p == '0')
-        p++;
-    else /* *p != '0' */ {
-        p++;
-        while(isdigit(*p))
-            p++;
+    if(c == '0') {
+        c = lex_get_save(lex, error);
+        if(isdigit(c)) {
+            lex_unget_unsave(lex, c);
+            goto out;
+        }
+    }
+    else if(isdigit(c)) {
+        c = lex_get_save(lex, error);
+        while(isdigit(c))
+            c = lex_get_save(lex, error);
     }
+    else {
+      lex_unget_unsave(lex, c);
+      goto out;
+    }
+
+    if(c != '.' && c != 'E' && c != 'e') {
+        json_int_t value;
+
+        lex_unget_unsave(lex, c);
 
-    if(*p == '.') {
-        p++;
-        if(!isdigit(*(p++)))
+        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;
+        }
 
-        while(isdigit(*p))
-            p++;
+        assert(end == saved_text + lex->saved_text.length);
+
+        lex->token = TOKEN_INTEGER;
+        lex->value.integer = value;
+        return 0;
     }
 
-    if(*p == 'E' || *p == 'e') {
-        p++;
-        if(*p == '+' || *p == '-')
-            p++;
+    if(c == '.') {
+        c = lex_get(lex, error);
+        if(!isdigit(c))
+            goto out;
+        lex_save(lex, c);
+
+        c = lex_get_save(lex, error);
+        while(isdigit(c))
+            c = lex_get_save(lex, error);
+    }
 
-        if(!isdigit(*(p++)))
+    if(c == 'E' || c == 'e') {
+        c = lex_get_save(lex, error);
+        if(c == '+' || c == '-')
+            c = lex_get_save(lex, error);
+
+        if(!isdigit(c)) {
+            lex_unget_unsave(lex, c);
             goto out;
+        }
 
-        while(isdigit(*p))
-            p++;
+        c = lex_get_save(lex, error);
+        while(isdigit(c))
+            c = lex_get_save(lex, error);
     }
 
-    lex->token = JSON_TOKEN_NUMBER;
+    lex_unget_unsave(lex, c);
 
-    lex->value.number = strtod(lex->start, &end);
-    assert(end == p);
+    saved_text = strbuffer_value(&lex->saved_text);
+    value = strtod(saved_text, &end);
+    assert(end == saved_text + lex->saved_text.length);
+
+    if(errno == ERANGE && value != 0) {
+        error_set(error, lex, "real number overflow");
+        goto out;
+    }
+
+    lex->token = TOKEN_REAL;
+    lex->value.real = value;
+    return 0;
 
 out:
-    lex->input = p;
+    return -1;
 }
 
-static int json_lex_scan(json_lex *lex)
+static int lex_scan(lex_t *lex, json_error_t *error)
 {
     char c;
 
-    if(lex->token == JSON_TOKEN_STRING) {
-      free(lex->value.string);
-      lex->value.string = NULL;
+    strbuffer_clear(&lex->saved_text);
+
+    if(lex->token == TOKEN_STRING) {
+        free(lex->value.string);
+        lex->value.string = NULL;
     }
 
-    while(isspace(*lex->input)) {
-        if(*lex->input == '\n')
+    c = lex_get(lex, error);
+    while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
+    {
+        if(c == '\n')
             lex->line++;
 
-        lex->input++;
+        c = lex_get(lex, error);
     }
 
-    lex->start = lex->input;
-    c = *lex->input;
+    if(c == (char)EOF) {
+        if(lex_eof(lex))
+            lex->token = TOKEN_EOF;
+        else
+            lex->token = TOKEN_INVALID;
+        goto out;
+    }
 
-    if(c == '\0')
-        lex->token = JSON_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 == '"')
-        json_scan_string(lex);
+        lex_scan_string(lex, error);
 
-    else if(isdigit(c) || c == '-')
-        json_scan_number(lex);
+    else if(isdigit(c) || c == '-') {
+        if(lex_scan_number(lex, c, error))
+            goto out;
+    }
 
-    else if(isalpha(c)) {
+    else if(isupper(c) || islower(c)) {
         /* eat up the whole identifier for clearer error messages */
-        int len;
-
-        while(isalpha(*lex->input))
-            lex->input++;
-        len = lex->input - lex->start;
-
-        if(strncmp(lex->start, "true", len) == 0)
-            lex->token = JSON_TOKEN_TRUE;
-        else if(strncmp(lex->start, "false", len) == 0)
-            lex->token = JSON_TOKEN_FALSE;
-        else if(strncmp(lex->start, "null", len) == 0)
-            lex->token = JSON_TOKEN_NULL;
+        const char *saved_text;
+
+        c = lex_get_save(lex, error);
+        while(isupper(c) || islower(c))
+            c = lex_get_save(lex, error);
+        lex_unget_unsave(lex, c);
+
+        saved_text = strbuffer_value(&lex->saved_text);
+
+        if(strcmp(saved_text, "true") == 0)
+            lex->token = TOKEN_TRUE;
+        else if(strcmp(saved_text, "false") == 0)
+            lex->token = TOKEN_FALSE;
+        else if(strcmp(saved_text, "null") == 0)
+            lex->token = TOKEN_NULL;
         else
-            lex->token = JSON_TOKEN_INVALID;
+            lex->token = TOKEN_INVALID;
     }
 
     else {
-        lex->token = JSON_TOKEN_INVALID;
-        lex->input++;
+        /* 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;
     }
 
+out:
     return lex->token;
 }
 
-static int json_lex_init(json_lex *lex, const char *input)
+static char *lex_steal_string(lex_t *lex)
+{
+    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, eof_func eof, void *data)
 {
-    lex->input = input;
-    lex->token = JSON_TOKEN_INVALID;
+    stream_init(&lex->stream, get, eof, data);
+    if(strbuffer_init(&lex->saved_text))
+        return -1;
+
+    lex->token = TOKEN_INVALID;
     lex->line = 1;
 
-    json_lex_scan(lex);
     return 0;
 }
 
-static void json_lex_close(json_lex *lex)
+static void lex_close(lex_t *lex)
 {
-    if(lex->token == JSON_TOKEN_STRING)
+    if(lex->token == TOKEN_STRING)
         free(lex->value.string);
+    strbuffer_close(&lex->saved_text);
 }
 
 
 /*** parser ***/
 
-static json_t *json_parse(json_lex *lex);
+static json_t *parse_value(lex_t *lex, json_error_t *error);
 
-static json_t *json_parse_object(json_lex *lex)
+static json_t *parse_object(lex_t *lex, json_error_t *error)
 {
     json_t *object = json_object();
     if(!object)
         return NULL;
 
-    json_lex_scan(lex);
+    lex_scan(lex, error);
+    if(lex->token == '}')
+        return object;
+
     while(1) {
         char *key;
         json_t *value;
 
-        if(lex->token != JSON_TOKEN_STRING) {
-            json_set_error(lex, "string expected");
+        if(lex->token != TOKEN_STRING) {
+            error_set(error, lex, "string or '}' expected");
             goto error;
         }
 
-        key = strdup(lex->value.string);
+        key = lex_steal_string(lex);
         if(!key)
             return NULL;
 
-        json_lex_scan(lex);
+        lex_scan(lex, error);
         if(lex->token != ':') {
-            json_set_error(lex, "':' expected");
+            free(key);
+            error_set(error, lex, "':' expected");
             goto error;
         }
 
-        json_lex_scan(lex);
-
-        value = json_parse(lex);
-        if(!value)
+        lex_scan(lex, error);
+        value = parse_value(lex, error);
+        if(!value) {
+            free(key);
             goto error;
+        }
 
-        if(json_object_set(object, key, value)) {
+        if(json_object_set_nocheck(object, key, value)) {
+            free(key);
             json_decref(value);
             goto error;
         }
@@ -307,14 +650,15 @@ static json_t *json_parse_object(json_lex *lex)
         json_decref(value);
         free(key);
 
+        lex_scan(lex, error);
         if(lex->token != ',')
             break;
 
-        json_lex_scan(lex);
+        lex_scan(lex, error);
     }
 
     if(lex->token != '}') {
-        json_set_error(lex, "'}' expected");
+        error_set(error, lex, "'}' expected");
         goto error;
     }
 
@@ -325,34 +669,36 @@ error:
     return NULL;
 }
 
-static json_t *json_parse_array(json_lex *lex)
+static json_t *parse_array(lex_t *lex, json_error_t *error)
 {
     json_t *array = json_array();
     if(!array)
         return NULL;
 
-    json_lex_scan(lex);
-    if(lex->token != ']') {
-        while(1) {
-            json_t *elem = json_parse(lex);
-            if(!elem)
-                goto error;
-
-            if(json_array_append(array, elem)) {
-                json_decref(elem);
-                goto error;
-            }
-            json_decref(elem);
+    lex_scan(lex, error);
+    if(lex->token == ']')
+        return array;
 
-            if(lex->token != ',')
-                break;
+    while(lex->token) {
+        json_t *elem = parse_value(lex, error);
+        if(!elem)
+            goto error;
 
-            json_lex_scan(lex);
+        if(json_array_append(array, elem)) {
+            json_decref(elem);
+            goto error;
         }
+        json_decref(elem);
+
+        lex_scan(lex, error);
+        if(lex->token != ',')
+            break;
+
+        lex_scan(lex, error);
     }
 
     if(lex->token != ']') {
-        json_set_error(lex, "']' expected");
+        error_set(error, lex, "']' expected");
         goto error;
     }
 
@@ -363,81 +709,177 @@ error:
     return NULL;
 }
 
-static json_t *json_parse(json_lex *lex)
+static json_t *parse_value(lex_t *lex, json_error_t *error)
 {
     json_t *json;
 
     switch(lex->token) {
-        case JSON_TOKEN_STRING: {
-            json = json_string(lex->value.string);
+        case TOKEN_STRING: {
+            json = json_string_nocheck(lex->value.string);
             break;
         }
 
-        case JSON_TOKEN_NUMBER: {
-            json = json_number(lex->value.number);
+        case TOKEN_INTEGER: {
+            json = json_integer(lex->value.integer);
             break;
         }
 
-        case JSON_TOKEN_TRUE:
+        case TOKEN_REAL: {
+            json = json_real(lex->value.real);
+            break;
+        }
+
+        case TOKEN_TRUE:
             json = json_true();
             break;
 
-        case JSON_TOKEN_FALSE:
+        case TOKEN_FALSE:
             json = json_false();
             break;
 
-        case JSON_TOKEN_NULL:
+        case TOKEN_NULL:
             json = json_null();
             break;
 
         case '{':
-            json = json_parse_object(lex);
+            json = parse_object(lex, error);
             break;
 
         case '[':
-            json = json_parse_array(lex);
+            json = parse_array(lex, error);
             break;
 
-        case JSON_TOKEN_INVALID:
-            json_set_error(lex, "invalid token");
+        case TOKEN_INVALID:
+            error_set(error, lex, "invalid token");
             return NULL;
 
         default:
-            json_set_error(lex, "unexpected token");
+            error_set(error, lex, "unexpected token");
             return NULL;
     }
 
     if(!json)
         return NULL;
 
-    json_lex_scan(lex);
     return json;
 }
 
-json_t *json_loads(const char *string)
+static json_t *parse_json(lex_t *lex, json_error_t *error)
 {
-    json_lex lex;
-    json_t *result = NULL;
+    lex_scan(lex, error);
+    if(lex->token != '[' && lex->token != '{') {
+        error_set(error, lex, "'[' or '{' expected");
+        return NULL;
+    }
 
-    if(json_lex_init(&lex, string))
+    return parse_value(lex, error);
+}
+
+typedef struct
+{
+    const char *data;
+    int pos;
+} string_data_t;
+
+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
+    {
+        stream->pos++;
+        return c;
+    }
+}
+
+static int string_eof(void *data)
+{
+    string_data_t *stream = (string_data_t *)data;
+    return (stream->data[stream->pos] == '\0');
+}
+
+json_t *json_loads(const char *string, size_t flags, json_error_t *error)
+{
+    lex_t lex;
+    json_t *result;
+    (void)flags; /* unused */
+
+    string_data_t stream_data = {string, 0};
+
+    if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
         return NULL;
 
-    if(lex.token != '[' && lex.token != '{') {
-        json_set_error(&lex, "'[' or '{' expected");
+    jsonp_error_init(error, "<string>");
+
+    result = parse_json(&lex, error);
+    if(!result)
         goto out;
+
+    lex_scan(&lex, error);
+    if(lex.token != TOKEN_EOF) {
+        error_set(error, &lex, "end of file expected");
+        json_decref(result);
+        result = NULL;
     }
 
-    result = json_parse(&lex);
+out:
+    lex_close(&lex);
+    return result;
+}
+
+json_t *json_loadf(FILE *input, size_t flags, json_error_t *error)
+{
+    lex_t lex;
+    const char *source;
+    json_t *result;
+    (void)flags; /* unused */
+
+    if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
+        return NULL;
+
+    if(input == stdin)
+        source = "<stdin>";
+    else
+        source = "<stream>";
+
+    jsonp_error_init(error, source);
+
+    result = parse_json(&lex, error);
     if(!result)
         goto out;
 
-    if(lex.token != JSON_TOKEN_EOF) {
-        json_set_error(&lex, "end of file expected");
+    lex_scan(&lex, error);
+    if(lex.token != TOKEN_EOF) {
+        error_set(error, &lex, "end of file expected");
         json_decref(result);
         result = NULL;
     }
 
 out:
-    json_lex_close(&lex);
+    lex_close(&lex);
+    return result;
+}
+
+json_t *json_load_file(const char *path, size_t flags, json_error_t *error)
+{
+    json_t *result;
+    FILE *fp;
+
+    jsonp_error_init(error, path);
+
+    fp = fopen(path, "r");
+    if(!fp)
+    {
+        error_set(error, NULL, "unable to open %s: %s",
+                  path, strerror(errno));
+        return NULL;
+    }
+
+    result = json_loadf(fp, flags, error);
+
+    fclose(fp);
     return result;
 }