Bug 589664 - Rewrite the JSON parser. r=njn, anticipating more review but getting it in-tree now for simplicity, even if more changes need to be made later

This commit is contained in:
Jeff Walden 2011-03-23 16:34:53 -07:00
Родитель d315235a27
Коммит 100a8553ea
16 изменённых файлов: 1172 добавлений и 18 удалений

Просмотреть файл

@ -157,6 +157,7 @@ CPPSRCS = \
jsnum.cpp \
jsobj.cpp \
json.cpp \
jsonparser.cpp \
jsopcode.cpp \
jsparse.cpp \
jsproxy.cpp \
@ -218,6 +219,7 @@ INSTALLED_HEADERS = \
jsobj.h \
jsobjinlines.h \
json.h \
jsonparser.h \
jsopcode.tbl \
jsopcode.h \
jsopcodeinlines.h \

Просмотреть файл

@ -307,7 +307,7 @@ MSG_DEF(JSMSG_BAD_OBJECT_INIT, 224, 0, JSEXN_SYNTAXERR, "invalid object i
MSG_DEF(JSMSG_CANT_SET_ARRAY_ATTRS, 225, 0, JSEXN_INTERNALERR, "can't set attributes on indexed array properties")
MSG_DEF(JSMSG_EVAL_ARITY, 226, 0, JSEXN_TYPEERR, "eval accepts only one parameter")
MSG_DEF(JSMSG_MISSING_FUN_ARG, 227, 2, JSEXN_TYPEERR, "missing argument {0} when calling function {1}")
MSG_DEF(JSMSG_JSON_BAD_PARSE, 228, 0, JSEXN_SYNTAXERR, "JSON.parse")
MSG_DEF(JSMSG_JSON_BAD_PARSE, 228, 1, JSEXN_SYNTAXERR, "JSON.parse: {0}")
MSG_DEF(JSMSG_JSON_BAD_STRINGIFY, 229, 0, JSEXN_ERR, "JSON.stringify")
MSG_DEF(JSMSG_XDR_CLOSURE_WRAPPER, 230, 1, JSEXN_INTERNALERR, "can't XDR closure wrapper for function {0}")
MSG_DEF(JSMSG_NOT_NONNULL_OBJECT, 231, 0, JSEXN_TYPEERR, "value is not a non-null object")

Просмотреть файл

@ -3216,6 +3216,7 @@ class AutoVectorRooter : protected AutoGCRooter
void infallibleAppend(const T &v) { vector.infallibleAppend(v); }
void popBack() { vector.popBack(); }
T popCopy() { return vector.popCopy(); }
bool growBy(size_t inc) {
size_t oldLength = vector.length();

Просмотреть файл

@ -66,6 +66,7 @@
#include "jslock.h"
#include "jsnum.h"
#include "jsobj.h"
#include "jsonparser.h"
#include "jsopcode.h"
#include "jsparse.h"
#include "jsproxy.h"
@ -1215,6 +1216,7 @@ EvalKernel(JSContext *cx, uintN argc, Value *vp, EvalType evalType, JSStackFrame
* will be lost.
*/
if (length > 2 && chars[0] == '(' && chars[length - 1] == ')') {
#if USE_OLD_AND_BUSTED_JSON_PARSER
JSONParser *jp = js_BeginJSONParse(cx, vp, /* suppressErrors = */true);
if (jp != NULL) {
/* Run JSON-parser on string inside ( and ). */
@ -1223,6 +1225,14 @@ EvalKernel(JSContext *cx, uintN argc, Value *vp, EvalType evalType, JSStackFrame
if (ok)
return true;
}
#else
JSONSourceParser parser(cx, chars + 1, length - 2, JSONSourceParser::StrictJSON,
JSONSourceParser::NoError);
if (!parser.parse(vp))
return false;
if (!vp->isUndefined())
return true;
#endif
}
/*

Просмотреть файл

@ -51,6 +51,7 @@
#include "jsiter.h"
#include "jsnum.h"
#include "jsobj.h"
#include "jsonparser.h"
#include "jsprf.h"
#include "jsscan.h"
#include "jsstr.h"
@ -115,25 +116,17 @@ js_json_parse(JSContext *cx, uintN argc, Value *vp)
{
JSString *s = NULL;
Value *argv = vp + 2;
AutoValueRooter reviver(cx);
Value reviver = UndefinedValue();
if (!JS_ConvertArguments(cx, argc, Jsvalify(argv), "S / v", &s, reviver.addr()))
if (!JS_ConvertArguments(cx, argc, Jsvalify(argv), "S / v", &s, &reviver))
return JS_FALSE;
JSLinearString *linearStr = s->ensureLinear(cx);
if (!linearStr)
return JS_FALSE;
JS::Anchor<JSString *> anchor(linearStr);
JSONParser *jp = js_BeginJSONParse(cx, vp);
JSBool ok = jp != NULL;
if (ok) {
const jschar *chars = linearStr->chars();
size_t length = linearStr->length();
ok = js_ConsumeJSONText(cx, jp, chars, length);
ok &= !!js_FinishJSONParse(cx, jp, reviver.value());
}
return ok;
return ParseJSONWithReviver(cx, linearStr->chars(), linearStr->length(), reviver, vp);
}
JSBool
@ -768,7 +761,7 @@ static JSBool
JSONParseError(JSONParser *jp, JSContext *cx)
{
if (!jp->suppressErrors)
JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_JSON_BAD_PARSE);
JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_JSON_BAD_PARSE, "syntax error");
return JS_FALSE;
}
@ -851,7 +844,7 @@ js_FinishJSONParse(JSContext *cx, JSONParser *jp, const Value &reviver)
ok = false;
} else if (!ok) {
JSONParseError(jp, cx);
} else if (reviver.isObject() && reviver.toObject().isCallable()) {
} else if (js_IsCallable(reviver)) {
ok = Revive(cx, reviver, vp);
}
@ -863,15 +856,27 @@ js_FinishJSONParse(JSContext *cx, JSONParser *jp, const Value &reviver)
namespace js {
JSBool
ParseJSONWithReviver(JSContext *cx, const jschar *chars, uint32 length, const Value &reviver,
ParseJSONWithReviver(JSContext *cx, const jschar *chars, size_t length, const Value &reviver,
Value *vp, DecodingMode decodingMode /* = STRICT */)
{
#if USE_OLD_AND_BUSTED_JSON_PARSER
JSONParser *jp = js_BeginJSONParse(cx, vp);
if (!jp)
return false;
JSBool ok = js_ConsumeJSONText(cx, jp, chars, length, decodingMode);
ok &= !!js_FinishJSONParse(cx, jp, reviver);
return ok;
#else
JSONSourceParser parser(cx, chars, length,
decodingMode == STRICT
? JSONSourceParser::StrictJSON
: JSONSourceParser::LegacyJSON);
if (!parser.parse(vp))
return false;
if (js_IsCallable(reviver))
return Revive(cx, reviver, vp);
return true;
#endif
}
} /* namespace js */

Просмотреть файл

@ -137,7 +137,7 @@ js_FinishJSONParse(JSContext *cx, JSONParser *jp, const js::Value &reviver);
namespace js {
extern JS_FRIEND_API(JSBool)
ParseJSONWithReviver(JSContext *cx, const jschar *chars, uint32 length, const Value &filter,
ParseJSONWithReviver(JSContext *cx, const jschar *chars, size_t length, const Value &filter,
Value *vp, DecodingMode decodingMode = STRICT);
} /* namespace js */

695
js/src/jsonparser.cpp Normal file
Просмотреть файл

@ -0,0 +1,695 @@
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
* vim: set ts=8 sw=4 et tw=99:
*
* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is SpiderMonkey JSON.
*
* The Initial Developer of the Original Code is
* the Mozilla Foundation.
* Portions created by the Initial Developer are Copyright (C) 2011
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Jeff Walden <jwalden+code@mit.edu> (original author)
*
* Alternatively, the contents of this file may be used under the terms of
* either of the GNU General Public License Version 2 or later (the "GPL"),
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#include "jsarray.h"
#include "jsnum.h"
#include "jsonparser.h"
#include "jsobjinlines.h"
#include "jsstrinlines.h"
using namespace js;
void
JSONSourceParser::error(const char *msg)
{
if (errorHandling == RaiseError)
JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_JSON_BAD_PARSE, msg);
}
bool
JSONSourceParser::errorReturn()
{
return errorHandling == NoError;
}
template<JSONSourceParser::StringType ST>
JSONSourceParser::Token
JSONSourceParser::readString()
{
JS_ASSERT(current < end);
JS_ASSERT(*current == '"');
/*
* JSONString:
* /^"([^\u0000-\u001F"\\]|\\(["/\\bfnrt]|u[0-9a-fA-F]{4}))*"$/
*/
if (++current == end) {
error("unterminated string literal");
return token(Error);
}
/*
* Optimization: if the source contains no escaped characters, create the
* string directly from the source text.
*/
RangeCheckedPointer<const jschar> start = current;
for (; current < end; current++) {
if (*current == '"') {
size_t length = current - start;
current++;
JSFlatString *str = (ST == JSONSourceParser::PropertyName)
? js_AtomizeChars(cx, start, length, 0)
: js_NewStringCopyN(cx, start, length);
if (!str)
return token(OOM);
return stringToken(str);
}
if (*current == '\\')
break;
if (*current <= 0x001F) {
error("bad control character in string literal");
return token(Error);
}
}
/*
* Slow case: string contains escaped characters. Copy a maximal sequence
* of unescaped characters into a temporary buffer, then an escaped
* character, and repeat until the entire string is consumed.
*/
StringBuffer buffer(cx);
do {
if (start < current && !buffer.append(start, current))
return token(OOM);
if (current >= end)
break;
jschar c = *current++;
if (c == '"') {
JSFlatString *str = (ST == JSONSourceParser::PropertyName)
? buffer.finishAtom()
: buffer.finishString();
if (!str)
return token(OOM);
return stringToken(str);
}
if (c != '\\') {
error("bad character in string literal");
return token(Error);
}
if (current >= end)
break;
switch (*current++) {
case '"': c = '"'; break;
case '/': c = '/'; break;
case '\\': c = '\\'; break;
case 'b': c = '\b'; break;
case 'f': c = '\f'; break;
case 'n': c = '\n'; break;
case 'r': c = '\r'; break;
case 't': c = '\t'; break;
case 'u':
if (end - current < 4) {
error("bad Unicode escape");
return token(Error);
}
if (JS7_ISHEX(current[0]) &&
JS7_ISHEX(current[1]) &&
JS7_ISHEX(current[2]) &&
JS7_ISHEX(current[3]))
{
c = (JS7_UNHEX(current[0]) << 12)
| (JS7_UNHEX(current[1]) << 8)
| (JS7_UNHEX(current[2]) << 4)
| (JS7_UNHEX(current[3]));
current += 4;
break;
}
/* FALL THROUGH */
default:
error("bad escaped character");
return token(Error);
}
if (!buffer.append(c))
return token(OOM);
start = current;
for (; current < end; current++) {
if (*current == '"' || *current == '\\' || *current <= 0x001F)
break;
}
} while (current < end);
error("unterminated string");
return token(Error);
}
JSONSourceParser::Token
JSONSourceParser::readNumber()
{
JS_ASSERT(current < end);
JS_ASSERT(JS7_ISDEC(*current) || *current == '-');
/*
* JSONNumber:
* /^-?(0|[1-9][0-9]+)(\.[0-9]+)?([eE][\+\-]?[0-9]+)?$/
*/
bool negative = *current == '-';
/* -? */
if (negative && ++current == end) {
error("no number after minus sign");
return token(Error);
}
const RangeCheckedPointer<const jschar> digitStart = current;
/* 0|[1-9][0-9]+ */
if (!JS7_ISDEC(*current)) {
error("unexpected non-digit");
return token(Error);
}
if (*current++ != '0') {
for (; current < end; current++) {
if (!JS7_ISDEC(*current))
break;
}
}
/* Fast path: no fractional or exponent part. */
if (current == end || (*current != '.' && *current != 'e' && *current != 'E')) {
const jschar *dummy;
jsdouble d;
if (!GetPrefixInteger(cx, digitStart, current, 10, &dummy, &d))
return token(OOM);
JS_ASSERT(current == dummy);
return numberToken(negative ? -d : d);
}
/* (\.[0-9]+)? */
if (current < end && *current == '.') {
if (++current == end) {
error("missing digits after decimal point");
return token(Error);
}
if (!JS7_ISDEC(*current)) {
error("unterminated fractional number");
return token(Error);
}
while (++current < end) {
if (!JS7_ISDEC(*current))
break;
}
}
/* ([eE][\+\-]?[0-9]+)? */
if (current < end && (*current == 'e' || *current == 'E')) {
if (++current == end) {
error("missing digits after exponent indicator");
return token(Error);
}
if (*current == '+' || *current == '-') {
if (++current == end) {
error("missing digits after exponent sign");
return token(Error);
}
}
if (!JS7_ISDEC(*current)) {
error("exponent part is missing a number");
return token(Error);
}
while (++current < end) {
if (!JS7_ISDEC(*current))
break;
}
}
jsdouble d;
const jschar *finish;
if (!js_strtod(cx, digitStart, current, &finish, &d))
return token(OOM);
JS_ASSERT(current == finish);
return numberToken(negative ? -d : d);
}
static inline bool
IsJSONWhitespace(jschar c)
{
return c == '\t' || c == '\r' || c == '\n' || c == ' ';
}
JSONSourceParser::Token
JSONSourceParser::advance()
{
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("unexpected end of data");
return token(Error);
}
switch (*current) {
case '"':
return readString<LiteralValue>();
case '-':
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7':
case '8':
case '9':
return readNumber();
case 't':
if (end - current < 4 || current[1] != 'r' || current[2] != 'u' || current[3] != 'e') {
error("unexpected keyword");
return token(Error);
}
current += 4;
return token(True);
case 'f':
if (end - current < 5 ||
current[1] != 'a' || current[2] != 'l' || current[3] != 's' || current[4] != 'e')
{
error("unexpected keyword");
return token(Error);
}
current += 5;
return token(False);
case 'n':
if (end - current < 4 || current[1] != 'u' || current[2] != 'l' || current[3] != 'l') {
error("unexpected keyword");
return token(Error);
}
current += 4;
return token(Null);
case '[':
current++;
return token(ArrayOpen);
case ']':
current++;
return token(ArrayClose);
case '{':
current++;
return token(ObjectOpen);
case '}':
current++;
return token(ObjectClose);
case ',':
current++;
return token(Comma);
case ':':
current++;
return token(Colon);
default:
error("unexpected character");
return token(Error);
}
}
JSONSourceParser::Token
JSONSourceParser::advanceAfterObjectOpen()
{
JS_ASSERT(current[-1] == '{');
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("end of data while reading object contents");
return token(Error);
}
if (*current == '"')
return readString<PropertyName>();
if (*current == '}') {
current++;
return token(ObjectClose);
}
error("expected property name or '}'");
return token(Error);
}
static inline void
AssertPastValue(const jschar *current)
{
/*
* We're past an arbitrary JSON value, so the previous character is
* *somewhat* constrained, even if this assertion is pretty broad. Don't
* knock it till you tried it: this assertion *did* catch a bug once.
*/
JS_ASSERT((current[-1] == 'l' &&
current[-2] == 'l' &&
current[-3] == 'u' &&
current[-4] == 'n') ||
(current[-1] == 'e' &&
current[-2] == 'u' &&
current[-3] == 'r' &&
current[-4] == 't') ||
(current[-1] == 'e' &&
current[-2] == 's' &&
current[-3] == 'l' &&
current[-4] == 'a' &&
current[-5] == 'f') ||
current[-1] == '}' ||
current[-1] == ']' ||
current[-1] == '"' ||
JS7_ISDEC(current[-1]));
}
JSONSourceParser::Token
JSONSourceParser::advanceAfterArrayElement()
{
AssertPastValue(current);
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("end of data when ',' or ']' was expected");
return token(Error);
}
if (*current == ',') {
current++;
return token(Comma);
}
if (*current == ']') {
current++;
return token(ArrayClose);
}
error("expected ',' or ']' after array element");
return token(Error);
}
JSONSourceParser::Token
JSONSourceParser::advancePropertyName()
{
JS_ASSERT(current[-1] == ',');
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("end of data when property name was expected");
return token(Error);
}
if (*current == '"')
return readString<PropertyName>();
if (parsingMode == LegacyJSON && *current == '}') {
/*
* Previous JSON parsing accepted trailing commas in non-empty object
* syntax, and some users depend on this. (Specifically, Places data
* serialization in versions of Firefox before 4.0. We can remove this
* mode when profile upgrades from 3.6 become unsupported.) Permit
* such trailing commas only when legacy parsing is specifically
* requested.
*/
current++;
return token(ObjectClose);
}
error("expected double-quoted property name");
return token(Error);
}
JSONSourceParser::Token
JSONSourceParser::advancePropertyColon()
{
JS_ASSERT(current[-1] == '"');
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("end of data after property name when ':' was expected");
return token(Error);
}
if (*current == ':') {
current++;
return token(Colon);
}
error("expected ':' after property name in object");
return token(Error);
}
JSONSourceParser::Token
JSONSourceParser::advanceAfterProperty()
{
AssertPastValue(current);
while (current < end && IsJSONWhitespace(*current))
current++;
if (current >= end) {
error("end of data after property value in object");
return token(Error);
}
if (*current == ',') {
current++;
return token(Comma);
}
if (*current == '}') {
current++;
return token(ObjectClose);
}
error("expected ',' or '}' after property value in object");
return token(Error);
}
/*
* This enum is local to JSONSourceParser::parse, below, but ISO C++98 doesn't
* allow templates to depend on local types. Boo-urns!
*/
enum ParserState { FinishArrayElement, FinishObjectMember, JSONValue };
bool
JSONSourceParser::parse(Value *vp)
{
Vector<ParserState> stateStack(cx);
AutoValueVector valueStack(cx);
*vp = UndefinedValue();
Token token;
ParserState state = JSONValue;
while (true) {
switch (state) {
case FinishObjectMember: {
Value v = valueStack.popCopy();
/*
* NB: Relies on js_DefineNativeProperty performing
* js_CheckForStringIndex.
*/
jsid propid = ATOM_TO_JSID(&valueStack.popCopy().toString()->asAtom());
if (!js_DefineNativeProperty(cx, &valueStack.back().toObject(), propid, v,
PropertyStub, StrictPropertyStub, JSPROP_ENUMERATE,
0, 0, NULL))
{
return false;
}
token = advanceAfterProperty();
if (token == ObjectClose)
break;
if (token != Comma) {
if (token == OOM)
return false;
if (token != Error)
error("expected ',' or '}' after property-value pair in object literal");
return errorReturn();
}
token = advancePropertyName();
/* FALL THROUGH */
}
JSONMember:
if (token == String) {
if (!valueStack.append(atomValue()))
return false;
token = advancePropertyColon();
if (token != Colon) {
JS_ASSERT(token == Error);
return errorReturn();
}
if (!stateStack.append(FinishObjectMember))
return false;
goto JSONValue;
}
if (token == ObjectClose) {
JS_ASSERT(state == FinishObjectMember);
JS_ASSERT(parsingMode == LegacyJSON);
break;
}
if (token == OOM)
return false;
if (token != Error)
error("property names must be double-quoted strings");
return errorReturn();
case FinishArrayElement: {
Value v = valueStack.popCopy();
if (!js_ArrayCompPush(cx, &valueStack.back().toObject(), v))
return false;
token = advanceAfterArrayElement();
if (token == Comma) {
if (!stateStack.append(FinishArrayElement))
return false;
goto JSONValue;
}
if (token == ArrayClose)
break;
JS_ASSERT(token == Error);
return errorReturn();
}
JSONValue:
case JSONValue:
token = advance();
JSONValueSwitch:
switch (token) {
case String:
case Number:
if (!valueStack.append(token == String ? stringValue() : numberValue()))
return false;
break;
case True:
if (!valueStack.append(BooleanValue(true)))
return false;
break;
case False:
if (!valueStack.append(BooleanValue(false)))
return false;
break;
case Null:
if (!valueStack.append(NullValue()))
return false;
break;
case ArrayOpen: {
JSObject *obj = NewDenseEmptyArray(cx);
if (!obj || !valueStack.append(ObjectValue(*obj)))
return false;
token = advance();
if (token == ArrayClose)
break;
if (!stateStack.append(FinishArrayElement))
return false;
goto JSONValueSwitch;
}
case ObjectOpen: {
JSObject *obj = NewBuiltinClassInstance(cx, &js_ObjectClass);
if (!obj || !valueStack.append(ObjectValue(*obj)))
return false;
token = advanceAfterObjectOpen();
if (token == ObjectClose)
break;
goto JSONMember;
}
case ArrayClose:
if (parsingMode == LegacyJSON &&
!stateStack.empty() &&
stateStack.back() == FinishArrayElement) {
/*
* Previous JSON parsing accepted trailing commas in
* non-empty array syntax, and some users depend on this.
* (Specifically, Places data serialization in versions of
* Firefox prior to 4.0. We can remove this mode when
* profile upgrades from 3.6 become unsupported.) Permit
* such trailing commas only when specifically
* instructed to do so.
*/
stateStack.popBack();
break;
}
/* FALL THROUGH */
case ObjectClose:
case Colon:
case Comma:
error("unexpected character");
return errorReturn();
case OOM:
return false;
case Error:
return errorReturn();
}
break;
}
if (stateStack.empty())
break;
state = stateStack.popCopy();
}
for (; current < end; current++) {
if (!IsJSONWhitespace(*current)) {
error("unexpected non-whitespace character after JSON data");
return errorReturn();
}
}
JS_ASSERT(end == current);
JS_ASSERT(valueStack.length() == 1);
*vp = valueStack[0];
return true;
}

178
js/src/jsonparser.h Normal file
Просмотреть файл

@ -0,0 +1,178 @@
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
* vim: set ts=8 sw=4 et tw=99:
*
* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is SpiderMonkey JSON.
*
* The Initial Developer of the Original Code is
* the Mozilla Foundation.
* Portions created by the Initial Developer are Copyright (C) 2011
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
* Jeff Walden <jwalden+code@mit.edu> (original author)
*
* Alternatively, the contents of this file may be used under the terms of
* either of the GNU General Public License Version 2 or later (the "GPL"),
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#ifndef jsonparser_h___
#define jsonparser_h___
#include "jscntxt.h"
#include "jsstr.h"
#include "jstl.h"
#include "jsvalue.h"
/*
* This class should be JSONParser, but the old JSON parser uses that name, so
* until we remove the old parser the class name will be overlong.
*
* NB: This class must only be used on the stack as it contains a js::Value.
*/
class JSONSourceParser
{
public:
enum ErrorHandling { RaiseError, NoError };
enum ParsingMode { StrictJSON, LegacyJSON };
private:
/* Data members */
JSContext * const cx;
js::RangeCheckedPointer<const jschar> current;
const js::RangeCheckedPointer<const jschar> end;
js::Value v;
const ParsingMode parsingMode;
const ErrorHandling errorHandling;
enum Token { String, Number, True, False, Null,
ArrayOpen, ArrayClose,
ObjectOpen, ObjectClose,
Colon, Comma,
OOM, Error };
#ifdef DEBUG
Token lastToken;
#endif
public:
/* Public API */
/*
* Create a parser for the provided JSON data. The parser will accept
* certain legacy, non-JSON syntax if decodingMode is LegacyJSON.
* Description of this syntax is deliberately omitted: new code should only
* use strict JSON parsing.
*/
JSONSourceParser(JSContext *cx, const jschar *data, size_t length,
ParsingMode parsingMode = StrictJSON,
ErrorHandling errorHandling = RaiseError)
: cx(cx),
current(data, data, length),
end(data + length, data, length),
parsingMode(parsingMode),
errorHandling(errorHandling)
#ifdef DEBUG
, lastToken(Error)
#endif
{
JS_ASSERT(current <= end);
}
/*
* Parse the JSON data specified at construction time. If it parses
* successfully, store the prescribed value in *vp and return true. If an
* internal error (e.g. OOM) occurs during parsing, return false.
* Otherwise, if invalid input was specifed but no internal error occurred,
* behavior depends upon the error handling specified at construction: if
* error handling is RaiseError then throw a SyntaxError and return false,
* otherwise return true and set *vp to |undefined|. (JSON syntax can't
* represent |undefined|, so the JSON data couldn't have specified it.)
*/
bool parse(js::Value *vp);
private:
js::Value numberValue() const {
JS_ASSERT(lastToken == Number);
JS_ASSERT(v.isNumber());
return v;
}
js::Value stringValue() const {
JS_ASSERT(lastToken == String);
JS_ASSERT(v.isString());
return v;
}
js::Value atomValue() const {
js::Value strval = stringValue();
JS_ASSERT(strval.toString()->isAtom());
return strval;
}
Token token(Token t) {
JS_ASSERT(t != String);
JS_ASSERT(t != Number);
#ifdef DEBUG
lastToken = t;
#endif
return t;
}
Token stringToken(JSString *str) {
this->v = js::StringValue(str);
#ifdef DEBUG
lastToken = String;
#endif
return String;
}
Token numberToken(jsdouble d) {
this->v = js::NumberValue(d);
#ifdef DEBUG
lastToken = Number;
#endif
return Number;
}
enum StringType { PropertyName, LiteralValue };
template<StringType ST> Token readString();
Token readNumber();
Token advance();
Token advancePropertyName();
Token advancePropertyColon();
Token advanceAfterProperty();
Token advanceAfterObjectOpen();
Token advanceAfterArrayElement();
void error(const char *msg);
bool errorReturn();
};
#endif /* jsonparser_h___ */

Просмотреть файл

@ -219,3 +219,13 @@
* support likely to be made opt-in at some future time.
*/
#define OLD_GETTER_SETTER_METHODS 1
/*
* Embedders: don't change this: it's a bake-until-ready hack only!
*
* NB: Changing this value requires adjusting the pass/fail state of a handful
* of tests in ecma_5/JSON/ which the old parser implemented incorrectly.
* Also make sure to rename JSONSourceParser to just JSONParser when the
* old parser is removed completely.
*/
#define USE_OLD_AND_BUSTED_JSON_PARSER 0

Просмотреть файл

@ -70,6 +70,7 @@
#include "jslock.h"
#include "jsnum.h"
#include "jsobj.h"
#include "json.h"
#include "jsparse.h"
#include "jsreflect.h"
#include "jsscope.h"
@ -4595,6 +4596,23 @@ NewGlobal(JSContext *cx, uintN argc, jsval *vp)
return true;
}
static JSBool
ParseLegacyJSON(JSContext *cx, uintN argc, jsval *vp)
{
if (argc != 1 || !JSVAL_IS_STRING(JS_ARGV(cx, vp)[0])) {
JS_ReportErrorNumber(cx, my_GetErrorMessage, NULL, JSSMSG_INVALID_ARGS, "parseLegacyJSON");
return false;
}
JSString *str = JSVAL_TO_STRING(JS_ARGV(cx, vp)[0]);
size_t length;
const jschar *chars = JS_GetStringCharsAndLength(cx, str, &length);
if (!chars)
return false;
return js::ParseJSONWithReviver(cx, chars, length, js::NullValue(), js::Valueify(vp), LEGACY);
}
static JSFunctionSpec shell_functions[] = {
JS_FN("version", Version, 0,0),
JS_FN("revertVersion", RevertVersion, 0,0),
@ -4694,6 +4712,7 @@ static JSFunctionSpec shell_functions[] = {
#endif
JS_FN("stringstats", StringStats, 0,0),
JS_FN("newGlobal", NewGlobal, 1,0),
JS_FN("parseLegacyJSON",ParseLegacyJSON,1,0),
JS_FS_END
};
@ -4829,6 +4848,8 @@ static const char *const shell_help_messages[] = {
"newGlobal(kind) Return a new global object, in the current\n"
" compartment if kind === 'same-compartment' or in a\n"
" new compartment if kind === 'new-compartment'",
"parseLegacyJSON(str) Parse str as legacy JSON, returning the result if the\n"
" parse succeeded and throwing a SyntaxError if not.",
/* Keep these last: see the static assertion below. */
#ifdef MOZ_PROFILING

Просмотреть файл

@ -5,7 +5,6 @@ script parse.js
script parse-crockford-01.js
script parse-primitives.js
script parse-reviver.js
fails script parse-octal-syntax-error.js # our JSON.parse wrongly accepts octal numbers
script parse-syntax-errors-01.js
script parse-syntax-errors-02.js
script stringify.js
@ -20,3 +19,11 @@ script stringify-replacer.js
script stringify-replacer-with-array-indexes.js
script stringify-toJSON-arguments.js
script trailing-comma.js
# These tests pass with the new parser but fail with the old one. If you're
# changing which parser is used, you'll need to change these test declarations
# accordingly. (And if you're removing the old parser, re-alphabetize these
# tests into the list above.)
script parse-number-syntax.js
script parse-octal-syntax-error.js
script parse-syntax-errors-03.js

Просмотреть файл

@ -0,0 +1,32 @@
// Any copyright is dedicated to the Public Domain.
// http://creativecommons.org/licenses/publicdomain/
testJSON('-', true);
testJSON('+', true);
testJSON('-f', true);
testJSON('+f', true);
testJSON('00', true);
testJSON('01', true);
testJSON('1.', true);
testJSON('1.0e', true);
testJSON('1.0e+', true);
testJSON('1.0e-', true);
testJSON('1.0e+z', true);
testJSON('1.0e-z', true);
testJSON('1.0ee', true);
testJSON('1.e1', true);
testJSON('1.e+1', true);
testJSON('1.e-1', true);
testJSON('.', true);
testJSON('.1', true);
testJSON('.1e', true);
testJSON('.1e1', true);
testJSON('.1e+1', true);
testJSON('.1e-1', true);
/******************************************************************************/
if (typeof reportCompare === "function")
reportCompare(true, true);
print("Tests complete");

Просмотреть файл

@ -0,0 +1,55 @@
// Any copyright is dedicated to the Public Domain.
// http://creativecommons.org/licenses/publicdomain/
testJSON('[', true);
testJSON('[1', true);
testJSON('[1,]', true);
testJSON('[1,{', true);
testJSON('[1,}', true);
testJSON('[1,{]', true);
testJSON('[1,}]', true);
testJSON('[1,{"', true);
testJSON('[1,}"', true);
testJSON('[1,{"\\', true);
testJSON('[1,}"\\', true);
testJSON('[1,"', true);
testJSON('[1,"\\', true);
testJSON('{', true);
testJSON('{1', true);
testJSON('{,', true);
testJSON('{"', true);
testJSON('{"\\', true);
testJSON('{"\\u', true);
testJSON('{"\\uG', true);
testJSON('{"\\u0', true);
testJSON('{"\\u01', true);
testJSON('{"\\u012', true);
testJSON('{"\\u0123', true);
testJSON('{"\\u0123"', true);
testJSON('{"a"', true);
testJSON('{"a"}', true);
testJSON('{"a":', true);
testJSON('{"a",}', true);
testJSON('{"a":}', true);
testJSON('{"a":,}', true);
testJSON('{"a":5,}', true);
testJSON('{"a":5,[', true);
testJSON('{"a":5,"', true);
testJSON('{"a":5,"', true);
testJSON('{"a":5,"\\', true);
testJSON("a[false ]".substring(1, 7) /* "[false" */, true);
testJSON('this', true);
testJSON('[1,{}]', false);
testJSON('{}', false);
testJSON('{"a":5}', false);
testJSON('{"\\u0123":5}', false);
/******************************************************************************/
if (typeof reportCompare === "function")
reportCompare(true, true);
print("Tests complete");

Просмотреть файл

@ -2,6 +2,11 @@ gTestsubsuite='JSON';
function testJSON(str, expectSyntaxError)
{
// Leading and trailing whitespace never affect parsing, so test the string
// multiple times with and without whitespace around it as it's easy and can
// potentially detect bugs.
// Try the provided string
try
{
JSON.parse(str);
@ -26,4 +31,82 @@ function testJSON(str, expectSyntaxError)
"have parsed as JSON, exception: " + e);
}
}
// Now try the provided string with trailing whitespace
try
{
JSON.parse(str + " ");
reportCompare(false, expectSyntaxError,
"string <" + str + " > " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON");
}
catch (e)
{
if (!(e instanceof SyntaxError))
{
reportCompare(true, false,
"parsing string <" + str + " > threw a non-SyntaxError " +
"exception: " + e);
}
else
{
reportCompare(true, expectSyntaxError,
"string <" + str + " > " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON, exception: " + e);
}
}
// Now try the provided string with leading whitespace
try
{
JSON.parse(" " + str);
reportCompare(false, expectSyntaxError,
"string < " + str + "> " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON");
}
catch (e)
{
if (!(e instanceof SyntaxError))
{
reportCompare(true, false,
"parsing string < " + str + "> threw a non-SyntaxError " +
"exception: " + e);
}
else
{
reportCompare(true, expectSyntaxError,
"string < " + str + "> " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON, exception: " + e);
}
}
// Now try the provided string with whitespace surrounding it
try
{
JSON.parse(" " + str + " ");
reportCompare(false, expectSyntaxError,
"string < " + str + " > " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON");
}
catch (e)
{
if (!(e instanceof SyntaxError))
{
reportCompare(true, false,
"parsing string < " + str + " > threw a non-SyntaxError " +
"exception: " + e);
}
else
{
reportCompare(true, expectSyntaxError,
"string < " + str + " > " +
"should" + (expectSyntaxError ? "n't" : "") + " " +
"have parsed as JSON, exception: " + e);
}
}
}

Просмотреть файл

@ -14,6 +14,7 @@ script eval-native-callback-is-indirect.js
script extension-methods-reject-null-undefined-this.js
skip-if(!xulRuntime.shell) script function-definition-with.js # needs evaluate()
script iterator-in-catch.js
skip-if(!xulRuntime.shell) script legacy-JSON.js # needs parseLegacyJSON
fails script nested-delete-name-in-evalcode.js # bug 604301, at a minimum
script proxy-strict.js
script regress-bug567606.js

Просмотреть файл

@ -0,0 +1,54 @@
// Any copyright is dedicated to the Public Domain.
// http://creativecommons.org/licenses/publicdomain/
try
{
parseLegacyJSON("[,]");
throw new Error("didn't throw");
}
catch (e)
{
assertEq(e instanceof SyntaxError, true, "didn't get syntax error, got: " + e);
}
try
{
parseLegacyJSON("{,}");
throw new Error("didn't throw");
}
catch (e)
{
assertEq(e instanceof SyntaxError, true, "didn't get syntax error, got: " + e);
}
assertEq(parseLegacyJSON("[1,]").length, 1);
assertEq(parseLegacyJSON("[1, ]").length, 1);
assertEq(parseLegacyJSON("[1 , ]").length, 1);
assertEq(parseLegacyJSON("[1 ,]").length, 1);
assertEq(parseLegacyJSON("[1,2,]").length, 2);
assertEq(parseLegacyJSON("[1,2, ]").length, 2);
assertEq(parseLegacyJSON("[1,2 , ]").length, 2);
assertEq(parseLegacyJSON("[1,2 ,]").length, 2);
assertEq(parseLegacyJSON('{"a": 2,}').a, 2);
assertEq(parseLegacyJSON('{"a": 2, }').a, 2);
assertEq(parseLegacyJSON('{"a": 2 , }').a, 2);
assertEq(parseLegacyJSON('{"a": 2 ,}').a, 2);
var obj;
obj = parseLegacyJSON('{"a": 2,"b": 3,}');
assertEq(obj.a + obj.b, 5);
obj = parseLegacyJSON('{"a": 2,"b": 3, }');
assertEq(obj.a + obj.b, 5);
obj = parseLegacyJSON('{"a": 2,"b": 3 , }');
assertEq(obj.a + obj.b, 5);
obj = parseLegacyJSON('{"a": 2,"b": 3 ,}');
assertEq(obj.a + obj.b, 5);
/******************************************************************************/
if (typeof reportCompare === "function")
reportCompare(true, true);
print("Tests complete");