Add harness for running Unicode.org Bidi tests.

Source/platform/text/BidiTestHarness.h

Index: Source/platform/text/BidiTestHarness.h
diff --git a/Source/platform/text/BidiTestHarness.h b/Source/platform/text/BidiTestHarness.h
new file mode 100644
index 0000000000000000000000000000000000000000..4cc8ed1cf50291705c713326ff95b3157361b479
--- /dev/null
+++ b/Source/platform/text/BidiTestHarness.h
@@ -0,0 +1,274 @@
+/*
+ * Copyright (C) 2013 Google Inc. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef BidiTestHarness_h
+#define BidiTestHarness_h
+
+#include <fstream>

[Jeffrey Yasskin, 2013/10/29 20:03:22]

I don't think you use this header: just #include <istream>?

[eseidel, 2013/10/29 20:49:44]

Done.

+#include <map>
+#include <string>
+
+// FIXME: We don't have any business owning this code. We should try to
+// upstream this to unicode.org if possible (for other implementations to use).
+// Unicode.org provides a reference implmentation, including parser:
+// http://www.unicode.org/Public/PROGRAMS/BidiReferenceC/6.3.0/source/brtest.c
+// But it, like the other implementations I've found, is rather tied to
+// the algorithms it is testing. This file seeks to only implement the parser bits.
+
+// Other C/C++ implementations of this parser:
+// https://github.com/googlei18n/fribidi-vs-unicode/blob/master/test.c
+// http://source.icu-project.org/repos/icu/icu/trunk/source/test/intltest/bidiconf.cpp
+// Both of those are too tied to their respective projects to be use to Blink.
+
+// There are non-C implmentations to parse BidiTest.txt as well, including:
+// https://github.com/twitter/twitter-cldr-rb/blob/master/spec/bidi/bidi_spec.rb
+
+namespace bidi_test {
+
+enum ParagraphDirection {
+    DirectionAutoLTR = 1,
+    DirectionLTR = 2,
+    DirectionRTL = 4,
+};
+const int kMaxParagraphDirection = DirectionAutoLTR | DirectionLTR | DirectionRTL;
+
+// For error printing:
+std::string nameFromParagraphDirection(ParagraphDirection paragraphDirection)
+{
+    switch (paragraphDirection) {
+    case bidi_test::DirectionAutoLTR:
+        return "Auto-LTR";
+    case bidi_test::DirectionLTR:
+        return "LTR";
+    case bidi_test::DirectionRTL:
+        return "RTL";
+    }
+}
+
+template<class Runner>
+class Harness {
+public:
+    Harness(Runner& runner)
+        : m_runner(runner)
+    {
+    }
+    void parse(std::istream& bidiTestFile);
+
+private:
+    Runner& m_runner;
+};
+
+// This trim() example comes from:
+// http://stackoverflow.com/questions/216823/whats-the-best-way-to-trim-stdstring
+// We could use boost::trim, but no other part of Blink uses boost yet.
+
+inline std::string& ltrim(std::string& s)

[Jeffrey Yasskin, 2013/10/29 20:03:22]

I'd either trim the argument in place or take the argument by const& and return
the trimmed string by value.

[eseidel, 2013/10/29 20:49:44]

Done.  I did the in-place way as it was simpler.  I'm sure we could write a
nicer trim...

+{
+    s.erase(s.begin(), std::find_if(s.begin(), s.end(), std::not1(std::ptr_fun<int, int>(std::isspace))));
+    return s;
+}
+
+inline std::string& rtrim(std::string& s)
+{
+    s.erase(std::find_if(s.rbegin(), s.rend(), std::not1(std::ptr_fun<int, int>(std::isspace))).base(), s.end());
+    return s;
+}
+
+inline std::string& trim(std::string& s)
+{
+    return ltrim(rtrim(s));
+}
+
+static bool parseStringList(std::string str, std::vector<std::string>& strings)
+{
+    std::string separators(" \t");
+    strings.clear();
+    size_t lastPos = str.find_first_not_of(separators); // skip leading spaces
+    size_t pos = str.find_first_of(separators, lastPos); // find next space
+
+    while (std::string::npos != pos || std::string::npos != lastPos) {
+        strings.push_back(str.substr(lastPos, pos - lastPos));
+        lastPos = str.find_first_not_of(separators, pos);
+        pos = str.find_first_of(separators, lastPos);
+    }
+    return true;

[Jeffrey Yasskin, 2013/10/29 20:03:22]

If this function can't return false, I'd return the std::vector<std::string>
instead of passing it as an out-parameter.

[eseidel, 2013/10/29 20:49:44]

Done.

+}
+
+static bool parseIntList(std::string str, std::vector<int>& ints)
+{
+    ints.clear();
+    std::vector<std::string> strings;
+    if (!parseStringList(str, strings))
+        return false;
+    for (int x = 0; x < strings.size(); x++) {
+        int i = atoi(strings[x].c_str());
+        ints.push_back(i);
+    }
+    return true;
+}
+
+static bool parseLevelRule(std::string line, std::vector<int>& levels)

[Jeffrey Yasskin, 2013/10/29 20:03:22]

Generally pass strings as const std::string& to avoid copies. The only time
you'd pass strings by value is when the function is going to make a copy as part
of its internal work anyway.

[eseidel, 2013/10/29 20:49:44]

Done.

+{
+    levels.clear();
+    std::vector<std::string> strings;
+    if (!parseStringList(line, strings))
+        return false;
+    for (int x = 0; x < strings.size(); x++) {
+        const std::string& levelString = strings[x];
+        int i;
+        if (levelString == "x")
+            i = -1;
+        else
+            i = atoi(levelString.c_str());
+        levels.push_back(i);
+    }
+    return true;}
+
+static bool parseReorderRule(std::string line, std::vector<int>& reorder)
+{
+    return parseIntList(line, reorder);
+}
+
+static bool parseTestString(std::string line, std::basic_string<UChar>& testString)
+{
+    testString.clear();
+    static std::map<std::string, UChar> charClassExamples;
+    if (charClassExamples.empty()) {

[Jeffrey Yasskin, 2013/10/29 20:03:22]

This isn't thread-safe, which you should comment. You can make it thread-safe
for everything except MSVC<2013 if you write a function that returns the map,
and initialize the local static with that.

[eseidel, 2013/10/29 20:49:44]

Done.

+        charClassExamples.insert({"L", 0x6c}); // 'l' for L
+        charClassExamples.insert({"R", 0x05D0}); // HEBREW ALEF
+        charClassExamples.insert({"EN", 0x33}); // '3' for EN
+        charClassExamples.insert({"ES", 0x2d}); // '-' for ES
+        charClassExamples.insert({"ET", 0x25}); // '%' for ET
+        charClassExamples.insert({"AN", 0x0660}); // arabic 0
+        charClassExamples.insert({"CS", 0x2c}); // ',' for CS
+        charClassExamples.insert({"B", 0x0A}); // <control-000A>
+        charClassExamples.insert({"S", 0x09}); // <control-0009>
+        charClassExamples.insert({"WS", 0x20}); // ' ' for WS
+        charClassExamples.insert({"ON", 0x3d}); // '=' for ON
+        charClassExamples.insert({"NSM", 0x05BF}); // HEBREW POINT RAFE
+        charClassExamples.insert({"AL", 0x0608}); // ARABIC RAY
+        charClassExamples.insert({"BN", 0x00AD}); // SOFT HYPHEN
+        charClassExamples.insert({"LRE", 0x202A});
+        charClassExamples.insert({"RLE", 0x202B});
+        charClassExamples.insert({"PDF", 0x202C});
+        charClassExamples.insert({"LRO", 0x202D});
+        charClassExamples.insert({"RLO", 0x202E});
+        charClassExamples.insert({"LRI", 0x2066});
+        charClassExamples.insert({"RLI", 0x2067});
+        charClassExamples.insert({"FSI", 0x2068});
+        charClassExamples.insert({"PDI", 0x2069});
+    }
+
+    std::vector<std::string> charClasses;
+    parseStringList(line, charClasses);
+    for (int i = 0; i < charClasses.size(); i++) {
+        testString.push_back(charClassExamples.find(charClasses[i])->second);
+    }
+    return true;
+}
+
+static bool parseParagraphDirectionMask(std::string line, int& modeMask)
+{
+
+    modeMask = atoi(line.c_str());

[Jeffrey Yasskin, 2013/10/29 20:03:22]

One cool thing you can do in C++ is use bitset<3> to store a set of 3 bits,
instead of using bits in an int:
http://en.cppreference.com/w/cpp/utility/bitset. You'd probably typedef it next
to the enum in order to avoid duplicating the "3".

enum ParagraphDirection {
  DirectionAutoLTR,
  DirectionLTR,
  DirectionRTL,
};
typedef std::bitset<3> ParagraphDirectionSet;

It has a constructor from a long long, so you'd use:
int modeMask = atoi(line.c_str());
out_param = ParagraphDirectionSet(modeMask);
return out_param.count() > 0 && modeMask < (1<<out_param.size());

You can use strtol() to get more error checking, since atoi() doesn't tell you
if the argument it bigger than an int or has trailing characters. But you may
not care.

[eseidel, 2013/10/29 20:49:44]

Although these are both awesome suggestions, I'm going to defer them to possible
later cleanup.  There is still another parser to write with this code. :)

+    return modeMask >= 1 && modeMask <= kMaxParagraphDirection;
+}
+
+static void parseError(const std::string& line, size_t lineNumber)
+{
+    // Use printf to avoid the expense of std::cout.
+    printf("Parse error, line %d : %s\n", lineNumber, line.c_str());

[Jeffrey Yasskin, 2013/10/29 20:03:22]

I think you're missing #include <stdio.h>

[eseidel, 2013/10/29 20:49:44]

Done.

+}
+
+template<class Runner>
+void Harness<Runner>::parse(std::istream& bidiTestFile)
+{
+    std::string levelsPrefix("@Levels");

[Jeffrey Yasskin, 2013/10/29 20:03:22]

Make these const if you're not planning to change them.

[eseidel, 2013/10/29 20:49:44]

Made them "static const".

+    std::string reorderPrefix("@Reorder");
+
+    // FIXME: UChar is an ICU type and cheating a bit to use here.
+    // uint16_t might be more portable.
+    std::basic_string<UChar> testString;
+    std::vector<int> levels;
+    std::vector<int> reorder;
+    int paragraphDirectionMask;
+
+    std::string originalLine;
+    std::string line;
+    size_t lineNumber = 0;
+    while (std::getline(bidiTestFile, originalLine)) {
+        lineNumber++;
+        line = originalLine;

[Jeffrey Yasskin, 2013/10/29 20:03:22]

I'd usually declare variables where they're initialized, when possible. You can
even use "std::getline(bidiTestFile, line) ... const std::string originalLine =
line;" to emphasize that 'originalLine isn't going to change.

[eseidel, 2013/10/29 20:49:44]

That makes an extra copy of line though, no?

[Jeffrey Yasskin, 2013/10/29 23:16:03]

It shouldn't make any more copies than you have in your current code:

std::string line;
while (std::getline(bidiTestFile, line)) {  // Copy from input.
  const std::string originalLine = line;  // 1 copy of contents, here or at the
first change to |line|.

+        size_t commentStart = originalLine.find_first_of('#');
+        if (commentStart != std::string::npos)
+            line = line.substr(0, commentStart);
+        line = trim(line);
+        if (line.empty())
+            continue;
+        if (line[0] == '@') {
+            if (!line.find(levelsPrefix)) {
+                if (!parseLevelRule(line.substr(levelsPrefix.length() + 1), levels))
+                    parseError(originalLine, lineNumber);
+                continue;
+            }
+            if (!line.find(reorderPrefix)) {
+                if (!parseReorderRule(line.substr(reorderPrefix.length() + 1), reorder))
+                    parseError(originalLine, lineNumber);
+                continue;
+            }
+        } else {
+            // Assume it's a data line.
+            size_t seperatorIndex = line.find_first_of(';');
+            if (seperatorIndex == std::string::npos) {
+                parseError(originalLine, lineNumber);
+                continue;
+            }
+            if (!parseTestString(line.substr(0, seperatorIndex), testString)) {

[Jeffrey Yasskin, 2013/10/29 20:03:22]

sp: seperator

[eseidel, 2013/10/29 20:49:44]

I'm considering taking 3rd grade again as a correspondence class. :(

+                parseError(originalLine, lineNumber);
+                continue;
+            }
+            if (!parseParagraphDirectionMask(line.substr(seperatorIndex + 1), paragraphDirectionMask)) {
+                parseError(originalLine, lineNumber);
+                continue;
+            }
+
+            if (paragraphDirectionMask & DirectionAutoLTR)

[Jeffrey Yasskin, 2013/10/29 20:03:22]

With a bitset, this becomes paragraphDirectionMask.test(DirectionAutoLTR) or
paragraphDirectionMask(DirectionAutoLTR). I tend to find that easier to read
than using the same type for bit names and sets of bits, but if you don't, feel
free to ignore the bitset comments.

+                m_runner.runTest(testString, reorder, levels, DirectionAutoLTR, originalLine, lineNumber);

[Jeffrey Yasskin, 2013/10/29 20:03:22]

It doesn't look like |levels| has a value coming into this function. Is the next
m_runner.runTest invocation using the resulting value? If not, I'd probably lean
toward making |levels| a local variable in runTest.

[eseidel, 2013/10/29 20:49:44]

It's confusing.  "levels" and "order" are both "global" variables with regards
to the parser.  They're set each time @Levels or @Reorder is encountered and
then all successive lines use those values until they're changed.

[Jeffrey Yasskin, 2013/10/29 23:16:03]

Oh, right, I just missed the assignment above because Rietveld wrapped the line
and broke my ctrl-f.

+            if (paragraphDirectionMask & DirectionLTR)
+                m_runner.runTest(testString, reorder, levels, DirectionLTR, originalLine, lineNumber);
+            if (paragraphDirectionMask & DirectionRTL)
+                m_runner.runTest(testString, reorder, levels, DirectionRTL, originalLine, lineNumber);
+        }
+    }
+}
+
+} // namespace bidi_test
+
+#endif // BidiTestHarness_h