X-Git-Url: http://www.project-moonshot.org/gitweb/?a=blobdiff_plain;f=src%2Fload.c;h=f004525a6073546062565e4e126edc182fa839c4;hb=86dc1d629b3ac74e5e9b65cbcb5ca53e1d15b698;hp=cae8c5300ddb9c084e8b134d730a87107630efd6;hpb=902bcdaa5e0d0273b1ba576c2fd676a5565b6d5e;p=jansson.git

diff --git a/src/load.c b/src/load.c
index cae8c53..f004525 100644
--- a/src/load.c
+++ b/src/load.c
@@ -1,11 +1,18 @@
+/*
+ * Copyright (c) 2009 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>
@@ -32,6 +39,7 @@ typedef struct {
     get_func get;
     eof_func eof;
     void *data;
+    int stream_pos;
     char buffer[5];
     int buffer_pos;
 } stream_t;
@@ -52,14 +60,25 @@ typedef struct {
 
 /*** error reporting ***/
 
+static void error_init(json_error_t *error)
+{
+    if(error)
+    {
+        error->text[0] = '\0';
+        error->line = -1;
+    }
+}
+
 static void error_set(json_error_t *error, const lex_t *lex,
                       const char *msg, ...)
 {
     va_list ap;
     char text[JSON_ERROR_TEXT_LENGTH];
 
-    if(!error)
+    if(!error || error->text[0] != '\0') {
+        /* error already set */
         return;
+    }
 
     va_start(ap, msg);
     vsnprintf(text, JSON_ERROR_TEXT_LENGTH, msg, ap);
@@ -71,8 +90,12 @@ static void error_set(json_error_t *error, const lex_t *lex,
         error->line = lex->line;
         if(saved_text && saved_text[0])
         {
-            snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
-                     "%s near '%s'", text, saved_text);
+            if(lex->saved_text.length <= 20) {
+                snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
+                         "%s near '%s'", text, saved_text);
+            }
+            else
+                snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
         }
         else
         {
@@ -95,32 +118,30 @@ void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
     stream->get = get;
     stream->eof = eof;
     stream->data = data;
+    stream->stream_pos = 0;
     stream->buffer[0] = '\0';
     stream->buffer_pos = 0;
 }
 
-static char stream_get(stream_t *stream)
+static char stream_get(stream_t *stream, json_error_t *error)
 {
+    char c;
+
     if(!stream->buffer[stream->buffer_pos])
     {
-        char c;
-
         stream->buffer[0] = stream->get(stream->data);
         stream->buffer_pos = 0;
 
         c = stream->buffer[0];
 
-        if(c == EOF && stream->eof(stream->data))
-            return EOF;
-
-        if(c < 0)
+        if((unsigned char)c >= 0x80 && c != (char)EOF)
         {
             /* multi-byte UTF-8 sequence */
             int i, count;
 
             count = utf8_check_first(c);
             if(!count)
-                return 0;
+                goto out;
 
             assert(count >= 2);
 
@@ -128,28 +149,41 @@ static char stream_get(stream_t *stream)
                 stream->buffer[i] = stream->get(stream->data);
 
             if(!utf8_check_full(stream->buffer, count))
-                return 0;
+                goto out;
 
+            stream->stream_pos += count;
             stream->buffer[count] = '\0';
         }
-        else
+        else {
             stream->buffer[1] = '\0';
+            stream->stream_pos++;
+        }
     }
 
-    return (char)stream->buffer[stream->buffer_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] == (unsigned char)c);
+    assert(stream->buffer[stream->buffer_pos] == c);
 }
 
 
-static int lex_get(lex_t *lex)
+static int lex_get(lex_t *lex, json_error_t *error)
 {
-    return stream_get(&lex->stream);
+    return stream_get(&lex->stream, error);
 }
 
 static int lex_eof(lex_t *lex)
@@ -162,9 +196,9 @@ static void lex_save(lex_t *lex, char c)
     strbuffer_append_byte(&lex->saved_text, c);
 }
 
-static int lex_get_save(lex_t *lex)
+static int lex_get_save(lex_t *lex, json_error_t *error)
 {
-    char c = stream_get(&lex->stream);
+    char c = stream_get(&lex->stream, error);
     lex_save(lex, c);
     return c;
 }
@@ -177,49 +211,93 @@ static void lex_unget_unsave(lex_t *lex, char c)
     assert(c == d);
 }
 
-static void lex_scan_string(lex_t *lex)
+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 int decode_unicode_escape(const char *str)
+{
+    int i;
+    int 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->value.string = NULL;
     lex->token = TOKEN_INVALID;
 
-    /* skip the " */
-    c = lex_get_save(lex);
+    c = lex_get_save(lex, error);
 
     while(c != '"') {
-        if(c == EOF && lex_eof(lex))
+        if(c == (char)EOF) {
+            lex_unget_unsave(lex, c);
+            if(lex_eof(lex))
+                error_set(error, lex, "premature end of input");
             goto out;
+        }
 
-        else if(0 <= c && c <= 0x1F) {
+        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(c == '\\') {
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
             if(c == 'u') {
-                c = lex_get_save(lex);
-                for(int i = 0; i < 4; i++) {
+                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);
+                    c = lex_get_save(lex, error);
                 }
             }
             else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
                     c == 'f' || c == 'n' || c == 'r' || c == 't')
-                c = lex_get_save(lex);
+                c = lex_get_save(lex, error);
             else {
                 lex_unget_unsave(lex, c);
+                error_set(error, lex, "invalid escape");
                 goto out;
             }
         }
         else
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
     }
 
     /* the actual value is at most of the same length as the source
@@ -245,11 +323,58 @@ static void lex_scan_string(lex_t *lex)
         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;
+                int value;
+
+                value = decode_unicode_escape(p);
+                p += 5;
+
+                if(0xD800 <= value && value <= 0xDBFF) {
+                    /* surrogate pair */
+                    if(*p == '\\' && *(p + 1) == 'u') {
+                        int 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;
@@ -260,114 +385,145 @@ 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++);
     }
     *t = '\0';
     lex->token = TOKEN_STRING;
+    return;
 
 out:
-    return;
+    free(lex->value.string);
 }
 
-static void lex_scan_number(lex_t *lex, char c)
+static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
 {
     const char *saved_text;
     char *end;
+    double value;
 
     lex->token = TOKEN_INVALID;
 
     if(c == '-')
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
 
     if(c == '0') {
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         if(isdigit(c)) {
             lex_unget_unsave(lex, c);
             goto out;
         }
     }
     else /* c != '0' */ {
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         while(isdigit(c))
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
     }
 
     if(c != '.' && c != 'E' && c != 'e') {
+        long 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);
+        value = strtol(saved_text, &end, 10);
         assert(end == saved_text + lex->saved_text.length);
 
-        return;
+        if((value == LONG_MAX && errno == ERANGE) || value > INT_MAX) {
+            error_set(error, lex, "too big integer");
+            goto out;
+        }
+        else if((value == LONG_MIN && errno == ERANGE) || value < INT_MIN) {
+            error_set(error, lex, "too big negative integer");
+            goto out;
+        }
+
+        lex->token = TOKEN_INTEGER;
+        lex->value.integer = (int)value;
+        return 0;
     }
 
     if(c == '.') {
-        c = lex_get(lex);
+        c = lex_get(lex, error);
         if(!isdigit(c))
             goto out;
         lex_save(lex, c);
 
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         while(isdigit(c))
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
     }
 
     if(c == 'E' || c == 'e') {
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         if(c == '+' || c == '-')
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
 
         if(!isdigit(c)) {
             lex_unget_unsave(lex, c);
             goto out;
         }
 
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         while(isdigit(c))
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, 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(value == 0 && errno == ERANGE) {
+        error_set(error, lex, "real number underflow");
+        goto out;
+    }
+
+    /* Cannot test for +/-HUGE_VAL because the HUGE_VAL constant is
+       only defined in C99 mode. So let's trust in sole errno. */
+    else if(errno == ERANGE) {
+        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)
+static int lex_scan(lex_t *lex, json_error_t *error)
 {
     char c;
 
     strbuffer_clear(&lex->saved_text);
 
     if(lex->token == TOKEN_STRING) {
-      free(lex->value.string);
-      lex->value.string = NULL;
+        free(lex->value.string);
+        lex->value.string = NULL;
     }
 
-    c = lex_get(lex);
+    c = lex_get(lex, error);
     while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
     {
         if(c == '\n')
             lex->line++;
 
-        c = lex_get(lex);
+        c = lex_get(lex, error);
     }
 
-    if(c == EOF && lex_eof(lex)) {
-        lex->token = TOKEN_EOF;
+    if(c == (char)EOF) {
+        if(lex_eof(lex))
+            lex->token = TOKEN_EOF;
+        else
+            lex->token = TOKEN_INVALID;
         goto out;
     }
 
@@ -377,18 +533,20 @@ static int lex_scan(lex_t *lex)
         lex->token = c;
 
     else if(c == '"')
-        lex_scan_string(lex);
+        lex_scan_string(lex, error);
 
-    else if(isdigit(c) || c == '-')
-        lex_scan_number(lex, c);
+    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 */
         const char *saved_text;
 
-        c = lex_get_save(lex);
+        c = lex_get_save(lex, error);
         while(isupper(c) || islower(c))
-            c = lex_get_save(lex);
+            c = lex_get_save(lex, error);
         lex_unget_unsave(lex, c);
 
         saved_text = strbuffer_value(&lex->saved_text);
@@ -403,13 +561,28 @@ static int lex_scan(lex_t *lex)
             lex->token = TOKEN_INVALID;
     }
 
-    else
+    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;
+    }
 
 out:
     return lex->token;
 }
 
+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)
 {
     stream_init(&lex->stream, get, eof, data);
@@ -426,6 +599,7 @@ static void lex_close(lex_t *lex)
 {
     if(lex->token == TOKEN_STRING)
         free(lex->value.string);
+    strbuffer_close(&lex->saved_text);
 }
 
 
@@ -439,7 +613,7 @@ static json_t *parse_object(lex_t *lex, json_error_t *error)
     if(!object)
         return NULL;
 
-    lex_scan(lex);
+    lex_scan(lex, error);
     if(lex->token == '}')
         return object;
 
@@ -452,18 +626,18 @@ static json_t *parse_object(lex_t *lex, json_error_t *error)
             goto error;
         }
 
-        key = strdup(lex->value.string);
+        key = lex_steal_string(lex);
         if(!key)
             return NULL;
 
-        lex_scan(lex);
+        lex_scan(lex, error);
         if(lex->token != ':') {
             free(key);
             error_set(error, lex, "':' expected");
             goto error;
         }
 
-        lex_scan(lex);
+        lex_scan(lex, error);
         value = parse_value(lex, error);
         if(!value) {
             free(key);
@@ -479,11 +653,11 @@ static json_t *parse_object(lex_t *lex, json_error_t *error)
         json_decref(value);
         free(key);
 
-        lex_scan(lex);
+        lex_scan(lex, error);
         if(lex->token != ',')
             break;
 
-        lex_scan(lex);
+        lex_scan(lex, error);
     }
 
     if(lex->token != '}') {
@@ -504,7 +678,7 @@ static json_t *parse_array(lex_t *lex, json_error_t *error)
     if(!array)
         return NULL;
 
-    lex_scan(lex);
+    lex_scan(lex, error);
     if(lex->token == ']')
         return array;
 
@@ -519,11 +693,11 @@ static json_t *parse_array(lex_t *lex, json_error_t *error)
         }
         json_decref(elem);
 
-        lex_scan(lex);
+        lex_scan(lex, error);
         if(lex->token != ',')
             break;
 
-        lex_scan(lex);
+        lex_scan(lex, error);
     }
 
     if(lex->token != ']') {
@@ -571,7 +745,7 @@ 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 '[':
@@ -595,8 +769,9 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
 
 json_t *parse_json(lex_t *lex, json_error_t *error)
 {
-    lex_scan(lex);
+    error_init(error);
 
+    lex_scan(lex, error);
     if(lex->token != '[' && lex->token != '{') {
         error_set(error, lex, "'[' or '{' expected");
         return NULL;
@@ -605,25 +780,6 @@ json_t *parse_json(lex_t *lex, json_error_t *error)
     return parse_value(lex, error);
 }
 
-json_t *json_load(const char *path, json_error_t *error)
-{
-    json_t *result;
-    FILE *fp;
-
-    fp = fopen(path, "r");
-    if(!fp)
-    {
-        error_set(error, NULL, "unable to open %s: %s",
-                  path, strerror(errno));
-        return NULL;
-    }
-
-    result = json_loadf(fp, error);
-
-    fclose(fp);
-    return result;
-}
-
 typedef struct
 {
     const char *data;
@@ -634,11 +790,14 @@ static int string_get(void *data)
 {
     char c;
     string_data_t *stream = (string_data_t *)data;
-    c = stream->data[stream->pos++];
+    c = stream->data[stream->pos];
     if(c == '\0')
         return EOF;
     else
+    {
+        stream->pos++;
         return c;
+    }
 }
 
 static int string_eof(void *data)
@@ -664,7 +823,7 @@ json_t *json_loads(const char *string, json_error_t *error)
     if(!result)
         goto out;
 
-    lex_scan(&lex);
+    lex_scan(&lex, error);
     if(lex.token != TOKEN_EOF) {
         error_set(error, &lex, "end of file expected");
         json_decref(result);
@@ -685,7 +844,36 @@ json_t *json_loadf(FILE *input, json_error_t *error)
         return NULL;
 
     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;
+    }
 
+out:
     lex_close(&lex);
     return result;
 }
+
+json_t *json_load_file(const char *path, json_error_t *error)
+{
+    json_t *result;
+    FILE *fp;
+
+    fp = fopen(path, "r");
+    if(!fp)
+    {
+        error_set(error, NULL, "unable to open %s: %s",
+                  path, strerror(errno));
+        return NULL;
+    }
+
+    result = json_loadf(fp, error);
+
+    fclose(fp);
+    return result;
+}