/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*- * vim: set ts=8 sts=4 et sw=4 tw=99: * This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ #include "jsonparser.h" #include "mozilla/RangedPtr.h" #include "jsarray.h" #include "jsnum.h" #include "vm/StringBuffer.h" #include "jsobjinlines.h" using namespace js; using mozilla::RangedPtr; JSONParser::~JSONParser() { for (size_t i = 0; i < stack.length(); i++) { if (stack[i].state == FinishArrayElement) js_delete(&stack[i].elements()); else js_delete(&stack[i].properties()); } for (size_t i = 0; i < freeElements.length(); i++) js_delete(freeElements[i]); for (size_t i = 0; i < freeProperties.length(); i++) js_delete(freeProperties[i]); } void JSONParser::trace(JSTracer *trc) { for (size_t i = 0; i < stack.length(); i++) { if (stack[i].state == FinishArrayElement) { ElementVector &elements = stack[i].elements(); for (size_t j = 0; j < elements.length(); j++) gc::MarkValueRoot(trc, &elements[j], "JSONParser element"); } else { PropertyVector &properties = stack[i].properties(); for (size_t j = 0; j < properties.length(); j++) { gc::MarkValueRoot(trc, &properties[j].value, "JSONParser property value"); gc::MarkIdRoot(trc, &properties[j].id, "JSONParser property id"); } } } } void JSONParser::error(const char *msg) { if (errorHandling == RaiseError) JS_ReportErrorNumber(cx, js_GetErrorMessage, NULL, JSMSG_JSON_BAD_PARSE, msg); } bool JSONParser::errorReturn() { return errorHandling == NoError; } template JSONParser::Token JSONParser::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. */ RangedPtr start = current; for (; current < end; current++) { if (*current == '"') { size_t length = current - start; current++; JSFlatString *str = (ST == JSONParser::PropertyName) ? AtomizeChars(cx, start.get(), length) : js_NewStringCopyN(cx, start.get(), 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.get(), current.get())) return token(OOM); if (current >= end) break; jschar c = *current++; if (c == '"') { JSFlatString *str = (ST == JSONParser::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); } JSONParser::Token JSONParser::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 RangedPtr 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; double d; if (!GetPrefixInteger(cx, digitStart.get(), current.get(), 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; } } double d; const jschar *finish; if (!js_strtod(cx, digitStart.get(), current.get(), &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 == ' '; } JSONParser::Token JSONParser::advance() { while (current < end && IsJSONWhitespace(*current)) current++; if (current >= end) { error("unexpected end of data"); return token(Error); } switch (*current) { case '"': return readString(); 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); } } JSONParser::Token JSONParser::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(); if (*current == '}') { current++; return token(ObjectClose); } error("expected property name or '}'"); return token(Error); } static inline void AssertPastValue(const RangedPtr 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])); } JSONParser::Token JSONParser::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); } JSONParser::Token JSONParser::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(); 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); } JSONParser::Token JSONParser::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); } JSONParser::Token JSONParser::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); } JSObject * JSONParser::createFinishedObject(PropertyVector &properties) { /* * Look for an existing cached type and shape for objects with this set of * properties. */ if (cx->typeInferenceEnabled()) { JSObject *obj = cx->compartment->types.newTypedObject(cx, properties.begin(), properties.length()); if (obj) return obj; } /* * Make a new object sized for the given number of properties and fill its * shape in manually. */ gc::AllocKind allocKind = gc::GetGCObjectKind(properties.length()); RootedObject obj(cx, NewBuiltinClassInstance(cx, &ObjectClass, allocKind)); if (!obj) return NULL; RootedId propid(cx); RootedValue value(cx); for (size_t i = 0; i < properties.length(); i++) { propid = properties[i].id; value = properties[i].value; if (!DefineNativeProperty(cx, obj, propid, value, JS_PropertyStub, JS_StrictPropertyStub, JSPROP_ENUMERATE, 0, 0)) { return NULL; } } /* * Try to assign a new type to the object with type information for its * properties, and update the initializer type object cache with this * object's final shape. */ if (cx->typeInferenceEnabled()) cx->compartment->types.fixObjectType(cx, obj); return obj; } inline bool JSONParser::finishObject(MutableHandleValue vp, PropertyVector &properties) { JS_ASSERT(&properties == &stack.back().properties()); JSObject *obj = createFinishedObject(properties); if (!obj) return false; vp.setObject(*obj); if (!freeProperties.append(&properties)) return false; stack.popBack(); return true; } inline bool JSONParser::finishArray(MutableHandleValue vp, ElementVector &elements) { JS_ASSERT(&elements == &stack.back().elements()); JSObject *obj = NewDenseCopiedArray(cx, elements.length(), elements.begin()); if (!obj) return false; /* Try to assign a new type to the array according to its elements. */ if (cx->typeInferenceEnabled()) cx->compartment->types.fixArrayType(cx, obj); vp.setObject(*obj); if (!freeElements.append(&elements)) return false; stack.popBack(); return true; } bool JSONParser::parse(MutableHandleValue vp) { RootedValue value(cx); JS_ASSERT(stack.empty()); vp.setUndefined(); Token token; ParserState state = JSONValue; while (true) { switch (state) { case FinishObjectMember: { PropertyVector &properties = stack.back().properties(); properties.back().value = value; token = advanceAfterProperty(); if (token == ObjectClose) { if (!finishObject(&value, properties)) return false; 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) { jsid id = AtomToId(atomValue()); PropertyVector &properties = stack.back().properties(); if (!properties.append(IdValuePair(id))) return false; token = advancePropertyColon(); if (token != Colon) { JS_ASSERT(token == Error); return errorReturn(); } goto JSONValue; } if (token == ObjectClose) { JS_ASSERT(state == FinishObjectMember); JS_ASSERT(parsingMode == LegacyJSON); if (!finishObject(&value, stack.back().properties())) return false; break; } if (token == OOM) return false; if (token != Error) error("property names must be double-quoted strings"); return errorReturn(); case FinishArrayElement: { ElementVector &elements = stack.back().elements(); if (!elements.append(value.get())) return false; token = advanceAfterArrayElement(); if (token == Comma) goto JSONValue; if (token == ArrayClose) { if (!finishArray(&value, elements)) return false; break; } JS_ASSERT(token == Error); return errorReturn(); } JSONValue: case JSONValue: token = advance(); JSONValueSwitch: switch (token) { case String: value = stringValue(); break; case Number: value = numberValue(); break; case True: value = BooleanValue(true); break; case False: value = BooleanValue(false); break; case Null: value = NullValue(); break; case ArrayOpen: { ElementVector *elements; if (!freeElements.empty()) { elements = freeElements.popCopy(); elements->clear(); } else { elements = cx->new_(cx); if (!elements) return false; } if (!stack.append(elements)) return false; token = advance(); if (token == ArrayClose) { if (!finishArray(&value, *elements)) return false; break; } goto JSONValueSwitch; } case ObjectOpen: { PropertyVector *properties; if (!freeProperties.empty()) { properties = freeProperties.popCopy(); properties->clear(); } else { properties = cx->new_(cx); if (!properties) return false; } if (!stack.append(properties)) return false; token = advanceAfterObjectOpen(); if (token == ObjectClose) { if (!finishObject(&value, *properties)) return false; break; } goto JSONMember; } case ArrayClose: if (parsingMode == LegacyJSON && !stack.empty() && stack.back().state == 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. */ if (!finishArray(&value, stack.back().elements())) return false; break; } /* FALL THROUGH */ case ObjectClose: case Colon: case Comma: error("unexpected character"); return errorReturn(); case OOM: return false; case Error: return errorReturn(); } break; } if (stack.empty()) break; state = stack.back().state; } for (; current < end; current++) { if (!IsJSONWhitespace(*current)) { error("unexpected non-whitespace character after JSON data"); return errorReturn(); } } JS_ASSERT(end == current); JS_ASSERT(stack.empty()); vp.set(value); return true; }