Implement specced parsing algorithm for <area coords>

This implements the parsing algorithm for "list of floating point
numbers" and uses it to parse the 'coords' attribute on <area>.
The fractional part of numbers are no longer discarded.
The old parsing code in platform/Length.cpp is no longer used and is
removed.

BUG=578114

Review URL: https://codereview.chromium.org/1636333003

Cr-Commit-Position: refs/heads/master@{#371940}

third_party/WebKit/LayoutTests/fast/html/area-coords-expected.txt

-
-This is a testharness.js-based test.
-PASS COMMA: "2,2,10,10" (rect) 
-PASS SEMICOLON: "2;2;10;10" (rect) 
-PASS SPACE: "2 2 10 10" (rect) 
-PASS TAB: "2\t2\t10\t10" (rect) 
-PASS FORM FEED: "2\f2\f10\f10" (rect) 
-PASS LINE FEED: "2\n2\n10\n10" (rect) 
-PASS CARRIGAGE RETURN: "2\r2\r10\r10" (rect) 
-FAIL LINE TABULATION: "2\v2\v10\v10" (rect) assert_equals: elementFromPoint(3, 3) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="221010"></area>
-FAIL LINE NEXT: "2…2…10…10" (rect) assert_equals: elementFromPoint(3, 3) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2…2…10…10"></area>
-FAIL EN QUAD: "2 2 10 10" (rect) assert_equals: elementFromPoint(3, 3) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2 2 10 10"></area>
-FAIL abc between numbers: "2a2b20c20,2,10,10" (rect) assert_equals: elementFromPoint(11, 9) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2a2b20c20,2,10,10"><...
-FAIL COLON between numbers: "2:2:20:20,2,10,10" (rect) assert_equals: elementFromPoint(11, 9) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2:2:20:20,2,10,10"><...
-FAIL U+0000 between numbers: "2\02\020\020,2,10,10" (rect) assert_equals: elementFromPoint(11, 9) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2\02\020\020,2,10,10"><...
-PASS leading COMMA: ",2,2,10,10" (rect) 
-PASS leading SPACE: " 2,2,10,10" (rect) 
-PASS leading SEMICOLON: ";2,2,10,10" (rect) 
-PASS trailing COMMA: "2,2,10," (rect) 
-PASS trailing SPACE: "2,2,10 " (rect) 
-PASS trailing SEMICOLON: "2,2,10;" (rect) 
-PASS PERCENT: "2%,2%,10%,10%" (rect) 
-PASS CSS units: "2in,2in,10cm,10cm" (rect) 
-FAIL float: "1.4,1.4,10,10" (rect) assert_equals: elementFromPoint(1, 3) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="1.4,1.4,10,10"></area>
-FAIL number starting with PERIOD: ".4,.4,10,10" (rect) assert_equals: elementFromPoint(0, 0) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords=".4,.4,10,10"></area>
-FAIL sci-not: "2,2,1e1,1e1" (rect) assert_equals: elementFromPoint(3, 3) expected Element node <area id="area" shape="rect" coords="2,2,1e1,1e1"></area> but got Element node <img src="resources/images/blue-border.png" usemap="#x" i...
-PASS leading/trailing garbage: "='2,2,10,10' " (rect) 
-PASS non-ascii garbage: "“2,2,10,10\"" (rect) 
-FAIL URL garbage with number: "2,2,10ls/spain/holidays/regions/10/Canary+Islands/Canary+Islands.html" (rect) assert_equals: elementFromPoint(3, 3) expected Element node <img src="resources/images/blue-border.png" usemap="#x" i... but got Element node <area id="area" shape="rect" coords="2,2,10ls/spain/holid...
-PASS consecutive COMMAs: "2,,10,10" (rect) 
-PASS consecutive SPACEs: "2  10,10" (rect) 
-PASS consecutive SEMICOLONs: "2;;10,10" (rect) 
-PASS several consecutive separators: ",,2;,;2,;,10 \t\r\n10;;" (rect) 
-PASS one too many numbers, trailing COMMA: "100,100,120,100,100,120,300," (poly) 
-Harness: the test ran to completion.
-

third_party/WebKit/Source/core/html/HTMLAreaElement.cpp

@@ -25,6 +25,7 @@
 #include "core/dom/ElementTraversal.h"
 #include "core/html/HTMLImageElement.h"
 #include "core/html/HTMLMapElement.h"
+#include "core/html/parser/HTMLParserIdioms.h"
 #include "core/layout/HitTestResult.h"
 #include "core/layout/LayoutImage.h"
 #include "core/layout/LayoutView.h"
@@ -35,11 +36,10 @@ namespace blink {
 
 namespace {
 
-// Adapt a Length to the allowed range of a LayoutUnit.
-float clampCoordinate(const Length& length)
+// Adapt a double to the allowed range of a LayoutUnit and narrow it to float precision.
+float clampCoordinate(double value)
 {
-    ASSERT(length.isFixed());
-    return LayoutUnit(length.value()).toFloat();
+    return LayoutUnit(value).toFloat();
 }
 
 }
@@ -80,7 +80,7 @@ void HTMLAreaElement::parseAttribute(const QualifiedName& name, const AtomicStri
         }
         invalidateCachedRegion();
     } else if (name == coordsAttr) {
-        m_coords = parseHTMLAreaElementCoords(value.string());
+        m_coords = parseHTMLListOfFloatingPointNumbers(value.string());
         invalidateCachedRegion();
     } else if (name == altAttr || name == accesskeyAttr) {
         // Do nothing.
@@ -152,7 +152,7 @@ Path HTMLAreaElement::getRegion(const LayoutSize& size) const
         }
         break;
     case Circle:
-        if (m_coords.size() >= 3 && m_coords[2].value() > 0) {
+        if (m_coords.size() >= 3 && m_coords[2] > 0) {
             float r = clampCoordinate(m_coords[2]);
             path.addEllipse(FloatRect(clampCoordinate(m_coords[0]) - r, clampCoordinate(m_coords[1]) - r, 2 * r, 2 * r));
         }

third_party/WebKit/Source/core/html/HTMLAreaElement.h

@@ -64,7 +64,7 @@ private:
     void invalidateCachedRegion();
 
     OwnPtr<Path> m_region;
-    Vector<Length> m_coords;
+    Vector<double> m_coords;
     LayoutSize m_lastSize;
     Shape m_shape;
 };

third_party/WebKit/Source/core/html/parser/HTMLParserIdioms.cpp

@@ -25,13 +25,15 @@
 #include "core/html/parser/HTMLParserIdioms.h"
 
 #include "core/HTMLNames.h"
-#include <limits>
+#include "platform/ParsingUtilities.h"
 #include "wtf/MathExtras.h"
 #include "wtf/text/AtomicString.h"
 #include "wtf/text/StringBuilder.h"
 #include "wtf/text/StringHash.h"
 #include "wtf/text/TextEncoding.h"
 
+#include <limits>
+
 namespace blink {
 
 using namespace HTMLNames;
@@ -113,18 +115,8 @@ Decimal parseToDecimalForNumberType(const String& string, const Decimal& fallbac
     return value.isZero() ? Decimal(0) : value;
 }
 
-double parseToDoubleForNumberType(const String& string, double fallbackValue)
+static double checkDoubleValue(double value, bool valid, double fallbackValue)
 {
-    // http://www.whatwg.org/specs/web-apps/current-work/#floating-point-numbers
-    // String::toDouble() accepts leading + and whitespace characters, which are not valid here.
-    UChar firstCharacter = string[0];
-    if (firstCharacter != '-' && firstCharacter != '.' && !isASCIIDigit(firstCharacter))
-        return fallbackValue;
-    if (string.endsWith('.'))
-        return fallbackValue;
-
-    bool valid = false;
-    double value = string.toDouble(&valid);
     if (!valid)
         return fallbackValue;
 
@@ -140,6 +132,21 @@ double parseToDoubleForNumberType(const String& string, double fallbackValue)
     return value ? value : 0;
 }
 
+double parseToDoubleForNumberType(const String& string, double fallbackValue)
+{
+    // http://www.whatwg.org/specs/web-apps/current-work/#floating-point-numbers
+    // String::toDouble() accepts leading + and whitespace characters, which are not valid here.
+    UChar firstCharacter = string[0];
+    if (firstCharacter != '-' && firstCharacter != '.' && !isASCIIDigit(firstCharacter))
+        return fallbackValue;
+    if (string.endsWith('.'))
+        return fallbackValue;
+
+    bool valid = false;
+    double value = string.toDouble(&valid);
+    return checkDoubleValue(value, valid, fallbackValue);
+}
+
 template <typename CharacterType>
 static bool parseHTMLIntegerInternal(const CharacterType* position, const CharacterType* end, int& value)
 {
@@ -265,6 +272,48 @@ bool parseHTMLNonNegativeInteger(const String& input, unsigned& value)
     return parseHTMLNonNegativeIntegerInternal(start, start + length, value);
 }
 
+template<typename CharacterType>
+static bool isSpaceOrDelimiter(CharacterType c)
+{
+    return isHTMLSpace(c) || c == ',' || c == ';';
+}
+
+template<typename CharacterType>
+static bool isNotSpaceDelimiterOrNumberStart(CharacterType c)
+{
+    return !(isSpaceOrDelimiter(c) || isASCIIDigit(c) || c == '.' || c == '-');
+}
+
+template<typename CharacterType>
+static Vector<double> parseHTMLListOfFloatingPointNumbersInternal(
+    const CharacterType* position, const CharacterType* end)
+{
+    Vector<double> numbers;
+    skipWhile<CharacterType, isSpaceOrDelimiter>(position, end);
+
+    while (position < end) {
+        skipWhile<CharacterType, isNotSpaceDelimiterOrNumberStart>(position, end);
+
+        const CharacterType* unparsedNumberStart = position;
+        skipUntil<CharacterType, isSpaceOrDelimiter>(position, end);
+
+        size_t parsedLength = 0;
+        double number = charactersToDouble(unparsedNumberStart, position - unparsedNumberStart, parsedLength);
+        numbers.append(checkDoubleValue(number, parsedLength != 0, 0));
+
+        skipWhile<CharacterType, isSpaceOrDelimiter>(position, end);
+    }
+    return numbers;
+}
+
+// https://html.spec.whatwg.org/multipage/infrastructure.html#rules-for-parsing-a-list-of-floating-point-numbers
+Vector<double> parseHTMLListOfFloatingPointNumbers(const String& input)
+{
+    if (input.is8Bit())
+        return parseHTMLListOfFloatingPointNumbersInternal(input.characters8(), input.characters8() + input.length());
+    return parseHTMLListOfFloatingPointNumbersInternal(input.characters16(), input.characters16() + input.length());
+}
+
 static const char charsetString[] = "charset";
 static const size_t charsetLength = sizeof("charset") - 1;
 

third_party/WebKit/Source/core/html/parser/HTMLParserIdioms.h

@@ -60,6 +60,9 @@ bool parseHTMLInteger(const String&, int&);
 // http://www.whatwg.org/specs/web-apps/current-work/#rules-for-parsing-non-negative-integers
 bool parseHTMLNonNegativeInteger(const String&, unsigned&);
 
+// https://html.spec.whatwg.org/multipage/infrastructure.html#rules-for-parsing-a-list-of-floating-point-numbers
+Vector<double> parseHTMLListOfFloatingPointNumbers(const String&);
+
 typedef Vector<std::pair<String, String>> HTMLAttributeList;
 // The returned encoding might not be valid.
 WTF::TextEncoding encodingFromMetaAttributes(const HTMLAttributeList&);

third_party/WebKit/Source/platform/Length.cpp

@@ -26,85 +26,11 @@
 
 #include "platform/CalculationValue.h"
 #include "platform/animation/AnimationUtilities.h"
-#include "wtf/ASCIICType.h"
-#include "wtf/text/StringBuffer.h"
-#include "wtf/text/WTFString.h"
 
 using namespace WTF;
 
 namespace blink {
 
-template<typename CharType>
-static unsigned splitLength(const CharType* data, unsigned length, unsigned& intLength, unsigned& doubleLength)
-{
-    ASSERT(length);
-
-    unsigned i = 0;
-    while (i < length && isSpaceOrNewline(data[i]))
-        ++i;
-    if (i < length && (data[i] == '+' || data[i] == '-'))
-        ++i;
-    while (i < length && isASCIIDigit(data[i]))
-        ++i;
-    intLength = i;
-    while (i < length && (isASCIIDigit(data[i]) || data[i] == '.'))
-        ++i;
-    doubleLength = i;
-
-    // IE quirk: Skip whitespace between the number and the % character (20 % => 20%).
-    while (i < length && isSpaceOrNewline(data[i]))
-        ++i;
-
-    return i;
-}
-
-template<typename CharType>
-static Length parseHTMLAreaCoordinate(const CharType* data, unsigned length)
-{
-    unsigned intLength;
-    unsigned doubleLength;
-    splitLength(data, length, intLength, doubleLength);
-
-    return Length(charactersToIntStrict(data, intLength), Fixed);
-}
-
-// FIXME: Per HTML5, this should follow the "rules for parsing a list of integers".
-Vector<Length> parseHTMLAreaElementCoords(const String& string)
-{
-    unsigned length = string.length();
-    StringBuffer<LChar> spacified(length);
-    for (unsigned i = 0; i < length; i++) {
-        UChar cc = string[i];
-        if (cc > '9' || (cc < '0' && cc != '-' && cc != '.'))
-            spacified[i] = ' ';
-        else
-            spacified[i] = cc;
-    }
-    RefPtr<StringImpl> str = spacified.release();
-    str = str->simplifyWhiteSpace();
-    ASSERT(str->is8Bit());
-
-    if (!str->length())
-        return Vector<Length>();
-
-    unsigned len = str->count(' ') + 1;
-    Vector<Length> r(len);
-
-    unsigned i = 0;
-    unsigned pos = 0;
-    size_t pos2;
-
-    while ((pos2 = str->find(' ', pos)) != kNotFound) {
-        r[i++] = parseHTMLAreaCoordinate(str->characters8() + pos, pos2 - pos);
-        pos = pos2 + 1;
-    }
-    r[i] = parseHTMLAreaCoordinate(str->characters8() + pos, str->length() - pos);
-
-    ASSERT(i == len - 1);
-
-    return r;
-}
-
 class CalculationValueHandleMap {
     USING_FAST_MALLOC(CalculationValueHandleMap);
     WTF_MAKE_NONCOPYABLE(CalculationValueHandleMap);

third_party/WebKit/Source/platform/Length.h

@@ -322,8 +322,6 @@ private:
     bool m_isFloat;
 };
 
-PLATFORM_EXPORT Vector<Length> parseHTMLAreaElementCoords(const String&);
-
 } // namespace blink
 
 #endif // Length_h

third_party/WebKit/Source/wtf/text/WTFString.cpp

@@ -1139,6 +1139,16 @@ double charactersToDouble(const UChar* data, size_t length, bool* ok)
     return toDoubleType<UChar, DisallowTrailingJunk>(data, length, ok, parsedLength);
 }
 
+double charactersToDouble(const LChar* data, size_t length, size_t& parsedLength)
+{
+    return toDoubleType<LChar, AllowTrailingJunk>(data, length, nullptr, parsedLength);
+}
+
+double charactersToDouble(const UChar* data, size_t length, size_t& parsedLength)
+{
+    return toDoubleType<UChar, AllowTrailingJunk>(data, length, nullptr, parsedLength);
+}
+
 float charactersToFloat(const LChar* data, size_t length, bool* ok)
 {
     // FIXME: This will return ok even when the string fits into a double but

third_party/WebKit/Source/wtf/text/WTFString.h

@@ -70,6 +70,8 @@ WTF_EXPORT uint64_t charactersToUInt64(const UChar*, size_t, bool* ok = 0); // i
 // consistent with the above functions instead.
 WTF_EXPORT double charactersToDouble(const LChar*, size_t, bool* ok = 0);
 WTF_EXPORT double charactersToDouble(const UChar*, size_t, bool* ok = 0);
+WTF_EXPORT double charactersToDouble(const LChar*, size_t, size_t& parsedLength);
+WTF_EXPORT double charactersToDouble(const UChar*, size_t, size_t& parsedLength);
 WTF_EXPORT float charactersToFloat(const LChar*, size_t, bool* ok = 0);
 WTF_EXPORT float charactersToFloat(const UChar*, size_t, bool* ok = 0);
 WTF_EXPORT float charactersToFloat(const LChar*, size_t, size_t& parsedLength);