+/*
+ * 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 <stdarg.h>
-#include <unistd.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_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 {
- const char *input;
- const char *start;
+ get_func get;
+ eof_func eof;
+ void *data;
+ int stream_pos;
+ char buffer[5];
+ int buffer_pos;
+} 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;
/*** 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);
if(lex)
{
+ const char *saved_text = strbuffer_value(&lex->saved_text);
error->line = lex->line;
- if(*lex->start)
+ 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(error->text, JSON_ERROR_TEXT_LENGTH,
+ "%s near '%s'", text, saved_text);
+ }
+ else
+ snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
}
else
{
/*** lexical analyzer ***/
-static void lex_scan_string(lex_t *lex)
+static 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, json_error_t *error)
+{
+ 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);
+}
+
+
+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)
+{
+ 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)
{
- /* skip the " */
- const char *p = lex->input + 1;
+ 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->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 <= 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(*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);
}
}
- 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 {
+ 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;
/* 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;
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;
+ 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, char 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(isdigit(c)) {
+ lex_unget_unsave(lex, c);
+ goto out;
+ }
}
- else /* *p != '0' */ {
- p++;
- while(isdigit(*p))
- p++;
+ 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(*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(!isdigit(c))
goto out;
+ lex_save(lex, c);
- p++;
- while(isdigit(*p))
- p++;
+ c = lex_get_save(lex, error);
+ while(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(!isdigit(c)) {
+ lex_unget_unsave(lex, c);
goto out;
+ }
- p++;
- while(isdigit(*p))
- p++;
+ c = lex_get_save(lex, error);
+ while(isdigit(c))
+ c = lex_get_save(lex, error);
}
- lex->token = TOKEN_REAL;
+ lex_unget_unsave(lex, c);
- lex->value.real = 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 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->input;
+ c = lex_get(lex, error);
while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
{
if(c == '\n')
lex->line++;
- lex->input++;
- c = *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 = 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(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 */
- int len;
+ const char *saved_text;
- while(isupper(*lex->input) || islower(*lex->input))
- lex->input++;
- len = lex->input - lex->start;
+ c = lex_get_save(lex, error);
+ while(isupper(c) || islower(c))
+ c = lex_get_save(lex, error);
+ lex_unget_unsave(lex, c);
- if(strncmp(lex->start, "true", len) == 0)
+ saved_text = strbuffer_value(&lex->saved_text);
+
+ 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)
+{
+ 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;
+ stream_init(&lex->stream, get, eof, data);
+ if(strbuffer_init(&lex->saved_text))
+ return -1;
+
lex->token = TOKEN_INVALID;
lex->line = 1;
- lex_scan(lex);
return 0;
}
{
if(lex->token == TOKEN_STRING)
free(lex->value.string);
+ strbuffer_close(&lex->saved_text);
}
/*** parser ***/
-static json_t *parse(lex_t *lex, json_error_t *error);
+static json_t *parse_value(lex_t *lex, json_error_t *error);
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;
- 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);
+ lex_scan(lex, error);
if(lex->token != ':') {
free(key);
error_set(error, lex, "':' expected");
goto error;
}
- lex_scan(lex);
-
- value = parse(lex, error);
+ 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;
json_decref(value);
free(key);
+ lex_scan(lex, error);
if(lex->token != ',')
break;
- lex_scan(lex);
+ lex_scan(lex, error);
}
if(lex->token != '}') {
if(!array)
return NULL;
- lex_scan(lex);
+ lex_scan(lex, error);
if(lex->token == ']')
return array;
while(lex->token) {
- json_t *elem = parse(lex, error);
+ json_t *elem = parse_value(lex, error);
if(!elem)
goto 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;
return NULL;
}
-static json_t *parse(lex_t *lex, json_error_t *error)
+static json_t *parse_value(lex_t *lex, 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;
}
break;
case '{':
- json = parse_object(lex, error);
+ json = parse_object(lex, error);
break;
case '[':
if(!json)
return NULL;
- lex_scan(lex);
return json;
}
-json_t *json_load(const char *path, json_error_t *error)
+static json_t *parse_json(lex_t *lex, json_error_t *error)
{
- json_t *result;
- FILE *fp;
+ error_init(error);
- fp = fopen(path, "r");
- if(!fp)
- {
- error_set(error, NULL, "unable to open %s: %s",
- path, strerror(errno));
+ lex_scan(lex, error);
+ if(lex->token != '[' && lex->token != '{') {
+ error_set(error, lex, "'[' or '{' expected");
return NULL;
}
- result = json_loadf(fp, error);
+ return parse_value(lex, error);
+}
- fclose(fp);
- return result;
+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, json_error_t *error)
+json_t *json_loads(const char *string, size_t flags, json_error_t *error)
{
lex_t lex;
- json_t *result = NULL;
+ json_t *result;
+ (void)flags; /* unused */
- if(lex_init(&lex, string))
- return NULL;
+ string_data_t stream_data = {string, 0};
- if(lex.token != '[' && lex.token != '{') {
- error_set(error, &lex, "'[' or '{' expected");
- goto out;
- }
+ if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
+ return NULL;
- result = parse(&lex, error);
+ 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);
return result;
}
-#define BUFFER_SIZE 4096
-
-json_t *json_loadf(FILE *input, json_error_t *error)
+json_t *json_loadf(FILE *input, size_t flags, json_error_t *error)
{
- strbuffer_t strbuff;
- char buffer[BUFFER_SIZE];
- size_t length;
- json_t *result = NULL;
+ lex_t lex;
+ json_t *result;
+ (void)flags; /* unused */
- if(strbuffer_init(&strbuff))
- return NULL;
+ if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
+ return NULL;
- while(1)
- {
- 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;
- }
+ result = parse_json(&lex, error);
+ if(!result)
+ goto out;
- result = json_loads(strbuffer_value(&strbuff), error);
+ lex_scan(&lex, error);
+ if(lex.token != TOKEN_EOF) {
+ error_set(error, &lex, "end of file expected");
+ json_decref(result);
+ result = NULL;
+ }
out:
- strbuffer_close(&strbuff);
+ lex_close(&lex);
return result;
}
-json_t *json_loadfd(int fd, json_error_t *error)
+json_t *json_load_file(const char *path, size_t flags, json_error_t *error)
{
- strbuffer_t strbuff;
- char buffer[BUFFER_SIZE];
- ssize_t length;
- json_t *result = NULL;
+ json_t *result;
+ FILE *fp;
- if(strbuffer_init(&strbuff))
- return NULL;
+ error_init(error);
- while(1)
+ fp = fopen(path, "r");
+ if(!fp)
{
- length = read(fd, buffer, BUFFER_SIZE);
- if(length == -1)
- {
- error_set(error, NULL, "read error: %s", strerror(errno));
- goto out;
- }
- else if(length == 0)
- break;
-
- if(strbuffer_append_bytes(&strbuff, buffer, length))
- {
- error_set(error, NULL, "error allocating memory");
- 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;
}
projects / jansson.git / blobdiff