X-Git-Url: http://www.project-moonshot.org/gitweb/?a=blobdiff_plain;f=src%2Fload.c;h=74e33fcc0ba7b220b8ab4ae18936a50db4c5767e;hb=cf9b384bcbd2fab0bfb44e0165368d3abc965f32;hp=1e2ad2a78e1a901c97370942209b605323c874c9;hpb=93c5892bc3a8138ba44a626d0172563a714e5b64;p=jansson.git

diff --git a/src/load.c b/src/load.c
index 1e2ad2a..74e33fc 100644
--- a/src/load.c
+++ b/src/load.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2009 Petri Lehtinen <petri@digip.org>
+ * 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.
@@ -8,11 +8,11 @@
 #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>
@@ -20,6 +20,10 @@
 #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
 #define TOKEN_STRING         256
@@ -29,115 +33,125 @@
 #define TOKEN_FALSE          260
 #define TOKEN_NULL           261
 
-/* read one byte from stream, return EOF on end of file */
+/* 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);
 
-/* 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;
+    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 ***/
 
-static void error_init(json_error_t *error)
-{
-    if(error)
-    {
-        error->text[0] = '\0';
-        error->line = -1;
-    }
-}
+/*** error reporting ***/
 
 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];
+
+    int line = -1, col = -1;
+    size_t pos = 0;
+    const char *result = msg_text;
 
-    if(!error || error->text[0] != '\0') {
-        /* error already set */
+    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)
     {
         const char *saved_text = strbuffer_value(&lex->saved_text);
-        error->line = lex->line;
+        char msg_with_context[JSON_ERROR_TEXT_LENGTH];
+
+        line = lex->stream.line;
+        col = lex->stream.column;
+        pos = lex->stream.position;
+
         if(saved_text && saved_text[0])
         {
             if(lex->saved_text.length <= 20) {
-                snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
-                         "%s near '%s'", text, saved_text);
+                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", text);
         }
         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
-    {
-        error->line = -1;
-        snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
-    }
+
+    jsonp_error_set(error, line, col, pos, "%s", result);
 }
 
 
 /*** lexical analyzer ***/
 
-void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
+static void
+stream_init(stream_t *stream, get_func get, void *data)
 {
     stream->get = get;
-    stream->eof = eof;
     stream->data = data;
-    stream->stream_pos = 0;
     stream->buffer[0] = '\0';
     stream->buffer_pos = 0;
+
+    stream->state = STREAM_STATE_OK;
+    stream->line = 1;
+    stream->column = 0;
+    stream->position = 0;
 }
 
-static char stream_get(stream_t *stream, json_error_t *error)
+static int stream_get(stream_t *stream, json_error_t *error)
 {
-    char c;
+    int c;
+
+    if(stream->state != STREAM_STATE_OK)
+        return stream->state;
 
     if(!stream->buffer[stream->buffer_pos])
     {
-        stream->buffer[0] = stream->get(stream->data);
-        stream->buffer_pos = 0;
-
-        c = stream->buffer[0];
+        c = stream->get(stream->data);
+        if(c == EOF) {
+            stream->state = STREAM_STATE_EOF;
+            return STREAM_STATE_EOF;
+        }
 
-        if(c == EOF && stream->eof(stream->data))
-            return EOF;
+        stream->buffer[0] = c;
+        stream->buffer_pos = 0;
 
-        if(c < 0)
+        if(0x80 <= c && c <= 0xFF)
         {
             /* multi-byte UTF-8 sequence */
             int i, count;
@@ -151,33 +165,50 @@ static char stream_get(stream_t *stream, json_error_t *error)
             for(i = 1; i < count; i++)
                 stream->buffer[i] = stream->get(stream->data);
 
-            if(!utf8_check_full(stream->buffer, count))
+            if(!utf8_check_full(stream->buffer, count, NULL))
                 goto out;
 
-            stream->stream_pos += count;
             stream->buffer[count] = '\0';
         }
-        else {
+        else
             stream->buffer[1] = '\0';
-            stream->stream_pos++;
-        }
     }
 
-    return stream->buffer[stream->buffer_pos++];
+    c = 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->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++;
+    }
 
-    stream->buffer[0] = EOF;
-    stream->buffer[1] = '\0';
-    stream->buffer_pos = 1;
+    return c;
 
-    return EOF;
+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, char c)
+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);
@@ -189,29 +220,32 @@ static int lex_get(lex_t *lex, json_error_t *error)
     return stream_get(&lex->stream, error);
 }
 
-static int lex_eof(lex_t *lex)
-{
-    return lex->stream.eof(lex->stream.data);
-}
-
-static void lex_save(lex_t *lex, char c)
+static void lex_save(lex_t *lex, int 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);
+    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_unsave(lex_t *lex, char c)
+static void lex_unget(lex_t *lex, int c)
 {
-    char d;
     stream_unget(&lex->stream, c);
-    d = strbuffer_pop(&lex->saved_text);
-    assert(c == d);
+}
+
+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)
@@ -220,14 +254,15 @@ static void lex_save_cached(lex_t *lex)
     {
         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 int decode_unicode_escape(const char *str)
+static int32_t decode_unicode_escape(const char *str)
 {
     int i;
-    int value = 0;
+    int32_t value = 0;
 
     assert(str[0] == 'u');
 
@@ -249,7 +284,7 @@ static int decode_unicode_escape(const char *str)
 
 static void lex_scan_string(lex_t *lex, json_error_t *error)
 {
-    char c;
+    int c;
     const char *p;
     char *t;
     int i;
@@ -257,13 +292,14 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
     lex->value.string = NULL;
     lex->token = TOKEN_INVALID;
 
-    /* skip the " */
     c = lex_get_save(lex, error);
 
     while(c != '"') {
-        if(c == EOF) {
-            if(lex_eof(lex))
-                error_set(error, lex, "premature end of input");
+        if(c == STREAM_STATE_ERROR)
+            goto out;
+
+        else if(c == STREAM_STATE_EOF) {
+            error_set(error, lex, "premature end of input");
             goto out;
         }
 
@@ -283,7 +319,6 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
                 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;
                     }
@@ -294,7 +329,6 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
                     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;
             }
@@ -310,7 +344,7 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
          - 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);
+    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;
@@ -328,7 +362,7 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
             if(*p == 'u') {
                 char buffer[4];
                 int length;
-                int value;
+                int32_t value;
 
                 value = decode_unicode_escape(p);
                 p += 5;
@@ -336,14 +370,15 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
                 if(0xD800 <= value && value <= 0xDBFF) {
                     /* surrogate pair */
                     if(*p == '\\' && *(p + 1) == 'u') {
-                        int value2 = decode_unicode_escape(++p);
+                        int32_t value2 = decode_unicode_escape(++p);
                         p += 5;
 
                         if(0xDC00 <= value2 && value2 <= 0xDFFF) {
                             /* valid second surrogate */
-                            value = ((value - 0xD800) << 10) +
-                                    (value2 - 0xDC00) +
-                                    0x10000;
+                            value =
+                                ((value - 0xD800) << 10) +
+                                (value2 - 0xDC00) +
+                                0x10000;
                         }
                         else {
                             /* invalid second surrogate */
@@ -399,13 +434,20 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
     return;
 
 out:
-    free(lex->value.string);
+    jsonp_free(lex->value.string);
 }
 
-static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
+#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 *saved_text;
     char *end;
+    double value;
 
     lex->token = TOKEN_INVALID;
 
@@ -419,27 +461,46 @@ static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
             goto out;
         }
     }
-    else /* c != '0' */ {
+    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);
-        lex->token = TOKEN_INTEGER;
 
         saved_text = strbuffer_value(&lex->saved_text);
-        lex->value.integer = strtol(saved_text, &end, 10);
+
+        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);
 
-        return;
+        lex->token = TOKEN_INTEGER;
+        lex->value.integer = value;
+        return 0;
     }
 
     if(c == '.') {
         c = lex_get(lex, error);
-        if(!isdigit(c))
+        if(!isdigit(c)) {
+            lex_unget(lex, c);
             goto out;
+        }
         lex_save(lex, c);
 
         c = lex_get_save(lex, error);
@@ -463,41 +524,46 @@ static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
     }
 
     lex_unget_unsave(lex, c);
-    lex->token = TOKEN_REAL;
 
     saved_text = strbuffer_value(&lex->saved_text);
-    lex->value.real = strtod(saved_text, &end);
+    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:
-    return;
+    return -1;
 }
 
 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_get(lex, error);
     while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
-    {
-        if(c == '\n')
-            lex->line++;
-
         c = lex_get(lex, error);
+
+    if(c == STREAM_STATE_EOF) {
+        lex->token = TOKEN_EOF;
+        goto out;
     }
 
-    if(c == EOF) {
-        if(lex_eof(lex))
-            lex->token = TOKEN_EOF;
-        else
-            lex->token = TOKEN_INVALID;
+    if(c == STREAM_STATE_ERROR) {
+        lex->token = TOKEN_INVALID;
         goto out;
     }
 
@@ -509,8 +575,10 @@ static int lex_scan(lex_t *lex, json_error_t *error)
     else if(c == '"')
         lex_scan_string(lex, error);
 
-    else if(isdigit(c) || c == '-')
-        lex_scan_number(lex, c, error);
+    else if(isdigit(c) || c == '-') {
+        if(lex_scan_number(lex, c, error))
+            goto out;
+    }
 
     else if(isupper(c) || islower(c)) {
         /* eat up the whole identifier for clearer error messages */
@@ -555,22 +623,20 @@ static char *lex_steal_string(lex_t *lex)
     return result;
 }
 
-static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
+static int lex_init(lex_t *lex, get_func get, void *data)
 {
-    stream_init(&lex->stream, get, eof, data);
+    stream_init(&lex->stream, get, data);
     if(strbuffer_init(&lex->saved_text))
         return -1;
 
     lex->token = TOKEN_INVALID;
-    lex->line = 1;
-
     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);
 }
 
@@ -604,7 +670,7 @@ static json_t *parse_object(lex_t *lex, json_error_t *error)
 
         lex_scan(lex, error);
         if(lex->token != ':') {
-            free(key);
+            jsonp_free(key);
             error_set(error, lex, "':' expected");
             goto error;
         }
@@ -612,18 +678,18 @@ static json_t *parse_object(lex_t *lex, json_error_t *error)
         lex_scan(lex, error);
         value = parse_value(lex, error);
         if(!value) {
-            free(key);
+            jsonp_free(key);
             goto error;
         }
 
         if(json_object_set_nocheck(object, key, value)) {
-            free(key);
+            jsonp_free(key);
             json_decref(value);
             goto error;
         }
 
         json_decref(value);
-        free(key);
+        jsonp_free(key);
 
         lex_scan(lex, error);
         if(lex->token != ',')
@@ -717,7 +783,7 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
             break;
 
         case '{':
-          json = parse_object(lex, error);
+            json = parse_object(lex, error);
             break;
 
         case '[':
@@ -739,10 +805,8 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
     return json;
 }
 
-json_t *parse_json(lex_t *lex, json_error_t *error)
+static json_t *parse_json(lex_t *lex, json_error_t *error)
 {
-    error_init(error);
-
     lex_scan(lex, error);
     if(lex->token != '[' && lex->token != '{') {
         error_set(error, lex, "'[' or '{' expected");
@@ -768,29 +832,23 @@ static int string_get(void *data)
     else
     {
         stream->pos++;
-        return c;
+        return (unsigned char)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, json_error_t *error)
+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 = {string, 0};
 
-    string_data_t stream_data = {
-        .data = string,
-        .pos = 0
-    };
+    (void)flags; /* unused */
 
-    if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
+    if(lex_init(&lex, string_get, (void *)&stream_data))
         return NULL;
 
+    jsonp_error_init(error, "<string>");
+
     result = parse_json(&lex, error);
     if(!result)
         goto out;
@@ -807,14 +865,23 @@ out:
     return result;
 }
 
-json_t *json_loadf(FILE *input, json_error_t *error)
+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))
+    if(lex_init(&lex, (get_func)fgetc, 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;
@@ -831,11 +898,13 @@ out:
     return result;
 }
 
-json_t *json_load_file(const char *path, json_error_t *error)
+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)
     {
@@ -844,7 +913,7 @@ json_t *json_load_file(const char *path, json_error_t *error)
         return NULL;
     }
 
-    result = json_loadf(fp, error);
+    result = json_loadf(fp, flags, error);
 
     fclose(fp);
     return result;