| 1 | // Copyright 2014 The Chromium Authors. All rights reserved. |
| 2 | // Copyright (C) 2016 Apple Inc. All rights reserved. |
| 3 | // |
| 4 | // Redistribution and use in source and binary forms, with or without |
| 5 | // modification, are permitted provided that the following conditions are |
| 6 | // met: |
| 7 | // |
| 8 | // * Redistributions of source code must retain the above copyright |
| 9 | // notice, this list of conditions and the following disclaimer. |
| 10 | // * Redistributions in binary form must reproduce the above |
| 11 | // copyright notice, this list of conditions and the following disclaimer |
| 12 | // in the documentation and/or other materials provided with the |
| 13 | // distribution. |
| 14 | // * Neither the name of Google Inc. nor the names of its |
| 15 | // contributors may be used to endorse or promote products derived from |
| 16 | // this software without specific prior written permission. |
| 17 | // |
| 18 | // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 19 | // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 20 | // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 21 | // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 22 | // OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 23 | // SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 24 | // LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 25 | // DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 26 | // THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 27 | // (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 28 | // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 29 | |
| 30 | #include "config.h" |
| 31 | #include "SizesCalcParser.h" |
| 32 | |
| 33 | #include "CSSParserToken.h" |
| 34 | #include "RenderView.h" |
| 35 | #include "SizesAttributeParser.h" |
| 36 | |
| 37 | namespace WebCore { |
| 38 | |
| 39 | SizesCalcParser::SizesCalcParser(CSSParserTokenRange range, const Document& document) |
| 40 | : m_result(0) |
| 41 | , m_document(document) |
| 42 | { |
| 43 | m_isValid = calcToReversePolishNotation(range) && calculate(); |
| 44 | } |
| 45 | |
| 46 | float SizesCalcParser::result() const |
| 47 | { |
| 48 | ASSERT(m_isValid); |
| 49 | return m_result; |
| 50 | } |
| 51 | |
| 52 | static bool operatorPriority(UChar cc, bool& highPriority) |
| 53 | { |
| 54 | if (cc == '+' || cc == '-') |
| 55 | highPriority = false; |
| 56 | else if (cc == '*' || cc == '/') |
| 57 | highPriority = true; |
| 58 | else |
| 59 | return false; |
| 60 | return true; |
| 61 | } |
| 62 | |
| 63 | bool SizesCalcParser::handleOperator(Vector<CSSParserToken>& stack, const CSSParserToken& token) |
| 64 | { |
| 65 | // If the token is an operator, o1, then: |
| 66 | // while there is an operator token, o2, at the top of the stack, and |
| 67 | // either o1 is left-associative and its precedence is equal to that of o2, |
| 68 | // or o1 has precedence less than that of o2, |
| 69 | // pop o2 off the stack, onto the output queue; |
| 70 | // push o1 onto the stack. |
| 71 | bool stackOperatorPriority; |
| 72 | bool incomingOperatorPriority; |
| 73 | |
| 74 | if (!operatorPriority(token.delimiter(), incomingOperatorPriority)) |
| 75 | return false; |
| 76 | if (!stack.isEmpty() && stack.last().type() == DelimiterToken) { |
| 77 | if (!operatorPriority(stack.last().delimiter(), stackOperatorPriority)) |
| 78 | return false; |
| 79 | if (!incomingOperatorPriority || stackOperatorPriority) { |
| 80 | appendOperator(stack.last()); |
| 81 | stack.removeLast(); |
| 82 | } |
| 83 | } |
| 84 | stack.append(token); |
| 85 | return true; |
| 86 | } |
| 87 | |
| 88 | void SizesCalcParser::appendNumber(const CSSParserToken& token) |
| 89 | { |
| 90 | SizesCalcValue value; |
| 91 | value.value = token.numericValue(); |
| 92 | m_valueList.append(value); |
| 93 | } |
| 94 | |
| 95 | bool SizesCalcParser::appendLength(const CSSParserToken& token) |
| 96 | { |
| 97 | SizesCalcValue value; |
| 98 | double result = SizesAttributeParser::computeLength(token.numericValue(), token.unitType(), m_document); |
| 99 | value.value = result; |
| 100 | value.isLength = true; |
| 101 | m_valueList.append(value); |
| 102 | return true; |
| 103 | } |
| 104 | |
| 105 | void SizesCalcParser::appendOperator(const CSSParserToken& token) |
| 106 | { |
| 107 | SizesCalcValue value; |
| 108 | value.operation = token.delimiter(); |
| 109 | m_valueList.append(value); |
| 110 | } |
| 111 | |
| 112 | bool SizesCalcParser::calcToReversePolishNotation(CSSParserTokenRange range) |
| 113 | { |
| 114 | // This method implements the shunting yard algorithm, to turn the calc syntax into a reverse polish notation. |
| 115 | // http://en.wikipedia.org/wiki/Shunting-yard_algorithm |
| 116 | |
| 117 | Vector<CSSParserToken> stack; |
| 118 | while (!range.atEnd()) { |
| 119 | const CSSParserToken& token = range.consume(); |
| 120 | switch (token.type()) { |
| 121 | case NumberToken: |
| 122 | appendNumber(token); |
| 123 | break; |
| 124 | case DimensionToken: |
| 125 | if (!CSSPrimitiveValue::isLength(token.unitType()) || !appendLength(token)) |
| 126 | return false; |
| 127 | break; |
| 128 | case DelimiterToken: |
| 129 | if (!handleOperator(stack, token)) |
| 130 | return false; |
| 131 | break; |
| 132 | case FunctionToken: |
| 133 | if (!equalIgnoringASCIICase(token.value(), "calc" )) |
| 134 | return false; |
| 135 | // "calc(" is the same as "(" |
| 136 | FALLTHROUGH; |
| 137 | case LeftParenthesisToken: |
| 138 | // If the token is a left parenthesis, then push it onto the stack. |
| 139 | stack.append(token); |
| 140 | break; |
| 141 | case RightParenthesisToken: |
| 142 | // If the token is a right parenthesis: |
| 143 | // Until the token at the top of the stack is a left parenthesis, pop operators off the stack onto the output queue. |
| 144 | while (!stack.isEmpty() && stack.last().type() != LeftParenthesisToken && stack.last().type() != FunctionToken) { |
| 145 | appendOperator(stack.last()); |
| 146 | stack.removeLast(); |
| 147 | } |
| 148 | // If the stack runs out without finding a left parenthesis, then there are mismatched parentheses. |
| 149 | if (stack.isEmpty()) |
| 150 | return false; |
| 151 | // Pop the left parenthesis from the stack, but not onto the output queue. |
| 152 | stack.removeLast(); |
| 153 | break; |
| 154 | case WhitespaceToken: |
| 155 | case EOFToken: |
| 156 | break; |
| 157 | case CDOToken: |
| 158 | case CDCToken: |
| 159 | case AtKeywordToken: |
| 160 | case HashToken: |
| 161 | case UrlToken: |
| 162 | case BadUrlToken: |
| 163 | case PercentageToken: |
| 164 | case IncludeMatchToken: |
| 165 | case DashMatchToken: |
| 166 | case PrefixMatchToken: |
| 167 | case SuffixMatchToken: |
| 168 | case SubstringMatchToken: |
| 169 | case ColumnToken: |
| 170 | case UnicodeRangeToken: |
| 171 | case IdentToken: |
| 172 | case CommaToken: |
| 173 | case ColonToken: |
| 174 | case SemicolonToken: |
| 175 | case LeftBraceToken: |
| 176 | case LeftBracketToken: |
| 177 | case RightBraceToken: |
| 178 | case RightBracketToken: |
| 179 | case StringToken: |
| 180 | case BadStringToken: |
| 181 | return false; |
| 182 | case CommentToken: |
| 183 | ASSERT_NOT_REACHED(); |
| 184 | return false; |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | // When there are no more tokens to read: |
| 189 | // While there are still operator tokens in the stack: |
| 190 | while (!stack.isEmpty()) { |
| 191 | // If the operator token on the top of the stack is a parenthesis, then there are mismatched parentheses. |
| 192 | CSSParserTokenType type = stack.last().type(); |
| 193 | if (type == LeftParenthesisToken || type == FunctionToken) |
| 194 | return false; |
| 195 | // Pop the operator onto the output queue. |
| 196 | appendOperator(stack.last()); |
| 197 | stack.removeLast(); |
| 198 | } |
| 199 | return true; |
| 200 | } |
| 201 | |
| 202 | static bool operateOnStack(Vector<SizesCalcValue>& stack, UChar operation) |
| 203 | { |
| 204 | if (stack.size() < 2) |
| 205 | return false; |
| 206 | SizesCalcValue rightOperand = stack.last(); |
| 207 | stack.removeLast(); |
| 208 | SizesCalcValue leftOperand = stack.last(); |
| 209 | stack.removeLast(); |
| 210 | bool isLength; |
| 211 | switch (operation) { |
| 212 | case '+': |
| 213 | if (rightOperand.isLength != leftOperand.isLength) |
| 214 | return false; |
| 215 | isLength = (rightOperand.isLength && leftOperand.isLength); |
| 216 | stack.append(SizesCalcValue(leftOperand.value + rightOperand.value, isLength)); |
| 217 | break; |
| 218 | case '-': |
| 219 | if (rightOperand.isLength != leftOperand.isLength) |
| 220 | return false; |
| 221 | isLength = (rightOperand.isLength && leftOperand.isLength); |
| 222 | stack.append(SizesCalcValue(leftOperand.value - rightOperand.value, isLength)); |
| 223 | break; |
| 224 | case '*': |
| 225 | if (rightOperand.isLength && leftOperand.isLength) |
| 226 | return false; |
| 227 | isLength = (rightOperand.isLength || leftOperand.isLength); |
| 228 | stack.append(SizesCalcValue(leftOperand.value * rightOperand.value, isLength)); |
| 229 | break; |
| 230 | case '/': |
| 231 | if (rightOperand.isLength || !rightOperand.value) |
| 232 | return false; |
| 233 | stack.append(SizesCalcValue(leftOperand.value / rightOperand.value, leftOperand.isLength)); |
| 234 | break; |
| 235 | default: |
| 236 | return false; |
| 237 | } |
| 238 | return true; |
| 239 | } |
| 240 | |
| 241 | bool SizesCalcParser::calculate() |
| 242 | { |
| 243 | Vector<SizesCalcValue> stack; |
| 244 | for (const auto& value : m_valueList) { |
| 245 | if (!value.operation) |
| 246 | stack.append(value); |
| 247 | else { |
| 248 | if (!operateOnStack(stack, value.operation)) |
| 249 | return false; |
| 250 | } |
| 251 | } |
| 252 | if (stack.size() == 1 && stack.last().isLength) { |
| 253 | m_result = std::max(clampTo<float>(stack.last().value), (float)0.0); |
| 254 | return true; |
| 255 | } |
| 256 | return false; |
| 257 | } |
| 258 | |
| 259 | } // namespace WebCore |
| 260 | |