| // This file is generated by Parser_cpp.template. |
| |
| // Copyright 2016 The Chromium Authors. All rights reserved. |
| // Use of this source code is governed by a BSD-style license that can be |
| // found in the LICENSE file. |
| |
| {% for namespace in config.protocol.namespace %} |
| namespace {{namespace}} { |
| {% endfor %} |
| |
| namespace { |
| |
| const int stackLimit = 1000; |
| |
| enum Token { |
| ObjectBegin, |
| ObjectEnd, |
| ArrayBegin, |
| ArrayEnd, |
| StringLiteral, |
| Number, |
| BoolTrue, |
| BoolFalse, |
| NullToken, |
| ListSeparator, |
| ObjectPairSeparator, |
| InvalidToken, |
| }; |
| |
| const char* const nullString = "null"; |
| const char* const trueString = "true"; |
| const char* const falseString = "false"; |
| |
| bool isASCII(uint16_t c) |
| { |
| return !(c & ~0x7F); |
| } |
| |
| bool isSpaceOrNewLine(uint16_t c) |
| { |
| return isASCII(c) && c <= ' ' && (c == ' ' || (c <= 0xD && c >= 0x9)); |
| } |
| |
| double charactersToDouble(const uint16_t* characters, size_t length, bool* ok) |
| { |
| std::vector<char> buffer; |
| buffer.reserve(length + 1); |
| for (size_t i = 0; i < length; ++i) { |
| if (!isASCII(characters[i])) { |
| *ok = false; |
| return 0; |
| } |
| buffer.push_back(static_cast<char>(characters[i])); |
| } |
| buffer.push_back('\0'); |
| return StringUtil::toDouble(buffer.data(), length, ok); |
| } |
| |
| double charactersToDouble(const uint8_t* characters, size_t length, bool* ok) |
| { |
| std::string buffer(reinterpret_cast<const char*>(characters), length); |
| return StringUtil::toDouble(buffer.data(), length, ok); |
| } |
| |
| template<typename Char> |
| bool parseConstToken(const Char* start, const Char* end, const Char** tokenEnd, const char* token) |
| { |
| while (start < end && *token != '\0' && *start++ == *token++) { } |
| if (*token != '\0') |
| return false; |
| *tokenEnd = start; |
| return true; |
| } |
| |
| template<typename Char> |
| bool readInt(const Char* start, const Char* end, const Char** tokenEnd, bool canHaveLeadingZeros) |
| { |
| if (start == end) |
| return false; |
| bool haveLeadingZero = '0' == *start; |
| int length = 0; |
| while (start < end && '0' <= *start && *start <= '9') { |
| ++start; |
| ++length; |
| } |
| if (!length) |
| return false; |
| if (!canHaveLeadingZeros && length > 1 && haveLeadingZero) |
| return false; |
| *tokenEnd = start; |
| return true; |
| } |
| |
| template<typename Char> |
| bool parseNumberToken(const Char* start, const Char* end, const Char** tokenEnd) |
| { |
| // We just grab the number here. We validate the size in DecodeNumber. |
| // According to RFC4627, a valid number is: [minus] int [frac] [exp] |
| if (start == end) |
| return false; |
| Char c = *start; |
| if ('-' == c) |
| ++start; |
| |
| if (!readInt(start, end, &start, false)) |
| return false; |
| if (start == end) { |
| *tokenEnd = start; |
| return true; |
| } |
| |
| // Optional fraction part |
| c = *start; |
| if ('.' == c) { |
| ++start; |
| if (!readInt(start, end, &start, true)) |
| return false; |
| if (start == end) { |
| *tokenEnd = start; |
| return true; |
| } |
| c = *start; |
| } |
| |
| // Optional exponent part |
| if ('e' == c || 'E' == c) { |
| ++start; |
| if (start == end) |
| return false; |
| c = *start; |
| if ('-' == c || '+' == c) { |
| ++start; |
| if (start == end) |
| return false; |
| } |
| if (!readInt(start, end, &start, true)) |
| return false; |
| } |
| |
| *tokenEnd = start; |
| return true; |
| } |
| |
| template<typename Char> |
| bool readHexDigits(const Char* start, const Char* end, const Char** tokenEnd, int digits) |
| { |
| if (end - start < digits) |
| return false; |
| for (int i = 0; i < digits; ++i) { |
| Char c = *start++; |
| if (!(('0' <= c && c <= '9') || ('a' <= c && c <= 'f') || ('A' <= c && c <= 'F'))) |
| return false; |
| } |
| *tokenEnd = start; |
| return true; |
| } |
| |
| template<typename Char> |
| bool parseStringToken(const Char* start, const Char* end, const Char** tokenEnd) |
| { |
| while (start < end) { |
| Char c = *start++; |
| if ('\\' == c) { |
| if (start == end) |
| return false; |
| c = *start++; |
| // Make sure the escaped char is valid. |
| switch (c) { |
| case 'x': |
| if (!readHexDigits(start, end, &start, 2)) |
| return false; |
| break; |
| case 'u': |
| if (!readHexDigits(start, end, &start, 4)) |
| return false; |
| break; |
| case '\\': |
| case '/': |
| case 'b': |
| case 'f': |
| case 'n': |
| case 'r': |
| case 't': |
| case 'v': |
| case '"': |
| break; |
| default: |
| return false; |
| } |
| } else if ('"' == c) { |
| *tokenEnd = start; |
| return true; |
| } |
| } |
| return false; |
| } |
| |
| template<typename Char> |
| bool skipComment(const Char* start, const Char* end, const Char** commentEnd) |
| { |
| if (start == end) |
| return false; |
| |
| if (*start != '/' || start + 1 >= end) |
| return false; |
| ++start; |
| |
| if (*start == '/') { |
| // Single line comment, read to newline. |
| for (++start; start < end; ++start) { |
| if (*start == '\n' || *start == '\r') { |
| *commentEnd = start + 1; |
| return true; |
| } |
| } |
| *commentEnd = end; |
| // Comment reaches end-of-input, which is fine. |
| return true; |
| } |
| |
| if (*start == '*') { |
| Char previous = '\0'; |
| // Block comment, read until end marker. |
| for (++start; start < end; previous = *start++) { |
| if (previous == '*' && *start == '/') { |
| *commentEnd = start + 1; |
| return true; |
| } |
| } |
| // Block comment must close before end-of-input. |
| return false; |
| } |
| |
| return false; |
| } |
| |
| template<typename Char> |
| void skipWhitespaceAndComments(const Char* start, const Char* end, const Char** whitespaceEnd) |
| { |
| while (start < end) { |
| if (isSpaceOrNewLine(*start)) { |
| ++start; |
| } else if (*start == '/') { |
| const Char* commentEnd; |
| if (!skipComment(start, end, &commentEnd)) |
| break; |
| start = commentEnd; |
| } else { |
| break; |
| } |
| } |
| *whitespaceEnd = start; |
| } |
| |
| template<typename Char> |
| Token parseToken(const Char* start, const Char* end, const Char** tokenStart, const Char** tokenEnd) |
| { |
| skipWhitespaceAndComments(start, end, tokenStart); |
| start = *tokenStart; |
| |
| if (start == end) |
| return InvalidToken; |
| |
| switch (*start) { |
| case 'n': |
| if (parseConstToken(start, end, tokenEnd, nullString)) |
| return NullToken; |
| break; |
| case 't': |
| if (parseConstToken(start, end, tokenEnd, trueString)) |
| return BoolTrue; |
| break; |
| case 'f': |
| if (parseConstToken(start, end, tokenEnd, falseString)) |
| return BoolFalse; |
| break; |
| case '[': |
| *tokenEnd = start + 1; |
| return ArrayBegin; |
| case ']': |
| *tokenEnd = start + 1; |
| return ArrayEnd; |
| case ',': |
| *tokenEnd = start + 1; |
| return ListSeparator; |
| case '{': |
| *tokenEnd = start + 1; |
| return ObjectBegin; |
| case '}': |
| *tokenEnd = start + 1; |
| return ObjectEnd; |
| case ':': |
| *tokenEnd = start + 1; |
| return ObjectPairSeparator; |
| case '0': |
| case '1': |
| case '2': |
| case '3': |
| case '4': |
| case '5': |
| case '6': |
| case '7': |
| case '8': |
| case '9': |
| case '-': |
| if (parseNumberToken(start, end, tokenEnd)) |
| return Number; |
| break; |
| case '"': |
| if (parseStringToken(start + 1, end, tokenEnd)) |
| return StringLiteral; |
| break; |
| } |
| return InvalidToken; |
| } |
| |
| template<typename Char> |
| int hexToInt(Char c) |
| { |
| if ('0' <= c && c <= '9') |
| return c - '0'; |
| if ('A' <= c && c <= 'F') |
| return c - 'A' + 10; |
| if ('a' <= c && c <= 'f') |
| return c - 'a' + 10; |
| DCHECK(false); |
| return 0; |
| } |
| |
| template<typename Char> |
| bool decodeString(const Char* start, const Char* end, StringBuilder* output) |
| { |
| while (start < end) { |
| uint16_t c = *start++; |
| if ('\\' != c) { |
| StringUtil::builderAppend(*output, c); |
| continue; |
| } |
| if (start == end) |
| return false; |
| c = *start++; |
| |
| if (c == 'x') { |
| // \x is not supported. |
| return false; |
| } |
| |
| switch (c) { |
| case '"': |
| case '/': |
| case '\\': |
| 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 'v': |
| c = '\v'; |
| break; |
| case 'u': |
| c = (hexToInt(*start) << 12) + |
| (hexToInt(*(start + 1)) << 8) + |
| (hexToInt(*(start + 2)) << 4) + |
| hexToInt(*(start + 3)); |
| start += 4; |
| break; |
| default: |
| return false; |
| } |
| StringUtil::builderAppend(*output, c); |
| } |
| return true; |
| } |
| |
| template<typename Char> |
| bool decodeString(const Char* start, const Char* end, String* output) |
| { |
| if (start == end) { |
| *output = ""; |
| return true; |
| } |
| if (start > end) |
| return false; |
| StringBuilder buffer; |
| StringUtil::builderReserve(buffer, end - start); |
| if (!decodeString(start, end, &buffer)) |
| return false; |
| *output = StringUtil::builderToString(buffer); |
| return true; |
| } |
| |
| template<typename Char> |
| std::unique_ptr<Value> buildValue(const Char* start, const Char* end, const Char** valueTokenEnd, int depth) |
| { |
| if (depth > stackLimit) |
| return nullptr; |
| |
| std::unique_ptr<Value> result; |
| const Char* tokenStart; |
| const Char* tokenEnd; |
| Token token = parseToken(start, end, &tokenStart, &tokenEnd); |
| switch (token) { |
| case InvalidToken: |
| return nullptr; |
| case NullToken: |
| result = Value::null(); |
| break; |
| case BoolTrue: |
| result = FundamentalValue::create(true); |
| break; |
| case BoolFalse: |
| result = FundamentalValue::create(false); |
| break; |
| case Number: { |
| bool ok; |
| double value = charactersToDouble(tokenStart, tokenEnd - tokenStart, &ok); |
| if (!ok) |
| return nullptr; |
| if (value >= INT_MIN && value <= INT_MAX && static_cast<int>(value) == value) |
| result = FundamentalValue::create(static_cast<int>(value)); |
| else |
| result = FundamentalValue::create(value); |
| break; |
| } |
| case StringLiteral: { |
| String value; |
| bool ok = decodeString(tokenStart + 1, tokenEnd - 1, &value); |
| if (!ok) |
| return nullptr; |
| result = StringValue::create(value); |
| break; |
| } |
| case ArrayBegin: { |
| std::unique_ptr<ListValue> array = ListValue::create(); |
| start = tokenEnd; |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| while (token != ArrayEnd) { |
| std::unique_ptr<Value> arrayNode = buildValue(start, end, &tokenEnd, depth + 1); |
| if (!arrayNode) |
| return nullptr; |
| array->pushValue(std::move(arrayNode)); |
| |
| // After a list value, we expect a comma or the end of the list. |
| start = tokenEnd; |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| if (token == ListSeparator) { |
| start = tokenEnd; |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| if (token == ArrayEnd) |
| return nullptr; |
| } else if (token != ArrayEnd) { |
| // Unexpected value after list value. Bail out. |
| return nullptr; |
| } |
| } |
| if (token != ArrayEnd) |
| return nullptr; |
| result = std::move(array); |
| break; |
| } |
| case ObjectBegin: { |
| std::unique_ptr<DictionaryValue> object = DictionaryValue::create(); |
| start = tokenEnd; |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| while (token != ObjectEnd) { |
| if (token != StringLiteral) |
| return nullptr; |
| String key; |
| if (!decodeString(tokenStart + 1, tokenEnd - 1, &key)) |
| return nullptr; |
| start = tokenEnd; |
| |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| if (token != ObjectPairSeparator) |
| return nullptr; |
| start = tokenEnd; |
| |
| std::unique_ptr<Value> value = buildValue(start, end, &tokenEnd, depth + 1); |
| if (!value) |
| return nullptr; |
| object->setValue(key, std::move(value)); |
| start = tokenEnd; |
| |
| // After a key/value pair, we expect a comma or the end of the |
| // object. |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| if (token == ListSeparator) { |
| start = tokenEnd; |
| token = parseToken(start, end, &tokenStart, &tokenEnd); |
| if (token == ObjectEnd) |
| return nullptr; |
| } else if (token != ObjectEnd) { |
| // Unexpected value after last object value. Bail out. |
| return nullptr; |
| } |
| } |
| if (token != ObjectEnd) |
| return nullptr; |
| result = std::move(object); |
| break; |
| } |
| |
| default: |
| // We got a token that's not a value. |
| return nullptr; |
| } |
| |
| skipWhitespaceAndComments(tokenEnd, end, valueTokenEnd); |
| return result; |
| } |
| |
| template<typename Char> |
| std::unique_ptr<Value> parseJSONInternal(const Char* start, unsigned length) |
| { |
| const Char* end = start + length; |
| const Char *tokenEnd; |
| std::unique_ptr<Value> value = buildValue(start, end, &tokenEnd, 0); |
| if (!value || tokenEnd != end) |
| return nullptr; |
| return value; |
| } |
| |
| } // anonymous namespace |
| |
| std::unique_ptr<Value> parseJSONCharacters(const uint16_t* characters, unsigned length) |
| { |
| return parseJSONInternal<uint16_t>(characters, length); |
| } |
| |
| std::unique_ptr<Value> parseJSONCharacters(const uint8_t* characters, unsigned length) |
| { |
| return parseJSONInternal<uint8_t>(characters, length); |
| } |
| |
| {% for namespace in config.protocol.namespace %} |
| } // namespace {{namespace}} |
| {% endfor %} |