Make small-caps work correctly with tr locale

Source/wtf/text/StringImpl.cpp

Index: Source/wtf/text/StringImpl.cpp
diff --git a/Source/wtf/text/StringImpl.cpp b/Source/wtf/text/StringImpl.cpp
index 14fb36d6564cc40d8562c602b1b1fbdf9d4cea12..837e0d801e03d446f52e271040d153b439eb1ee7 100644
--- a/Source/wtf/text/StringImpl.cpp
+++ b/Source/wtf/text/StringImpl.cpp
@@ -530,7 +530,7 @@ PassRefPtr<StringImpl> StringImpl::lower()
 
         // Do a slower implementation for cases that include non-ASCII Latin-1 characters.
         for (int32_t i = 0; i < length; ++i)
-            data8[i] = static_cast<LChar>(Unicode::toLower(characters8()[i]));

[tkent, 2015/05/28 00:38:26]

I prefer keeping Unicode::.  We have Unicode::toLower, StringImpl::lower,
Unicode::toUpper, StringImpl::upper, StringImpl::toUpper.  They are confusing.

[rwlbuis, 2015/05/28 23:03:07]

Yes, I should not have included it. My idea was to make it consistent. Fixed.

+            data8[i] = static_cast<LChar>(toLower(characters8()[i]));
 
         return newImpl.release();
     }
@@ -564,12 +564,12 @@ PassRefPtr<StringImpl> StringImpl::lower()
     RefPtr<StringImpl> newImpl = createUninitialized(m_length, data16);
 
     bool error;
-    int32_t realLength = Unicode::toLower(data16, length, characters16(), m_length, &error);
+    int32_t realLength = toLower(data16, length, characters16(), m_length, &error);
     if (!error && realLength == length)
         return newImpl.release();
 
     newImpl = createUninitialized(realLength, data16);
-    Unicode::toLower(data16, realLength, characters16(), m_length, &error);
+    toLower(data16, realLength, characters16(), m_length, &error);
     if (error)
         return this;
     return newImpl.release();
@@ -608,7 +608,7 @@ PassRefPtr<StringImpl> StringImpl::upper()
             LChar c = characters8()[i];
             if (UNLIKELY(c == smallLetterSharpSCharacter))
                 ++numberSharpSCharacters;
-            UChar upper = static_cast<UChar>(Unicode::toUpper(c));
+            UChar upper = static_cast<UChar>(toUpper(c));
             if (UNLIKELY(upper > 0xff)) {
                 // Since this upper-cased character does not fit in an 8-bit string, we need to take the 16-bit path.
                 goto upconvert;
@@ -629,8 +629,9 @@ PassRefPtr<StringImpl> StringImpl::upper()
             if (c == smallLetterSharpSCharacter) {
                 *dest++ = 'S';
                 *dest++ = 'S';
-            } else
-                *dest++ = static_cast<LChar>(Unicode::toUpper(c));
+            } else {
+                *dest++ = static_cast<LChar>(toUpper(c));
+            }
         }
 
         return newImpl.release();
@@ -655,11 +656,11 @@ upconvert:
 
     // Do a slower implementation for cases that include non-ASCII characters.
     bool error;
-    int32_t realLength = Unicode::toUpper(data16, length, source16, m_length, &error);
+    int32_t realLength = toUpper(data16, length, source16, m_length, &error);
     if (!error && realLength == length)
         return newImpl;
     newImpl = createUninitialized(realLength, data16);
-    Unicode::toUpper(data16, realLength, source16, m_length, &error);
+    toUpper(data16, realLength, source16, m_length, &error);
     if (error)
         return this;
     return newImpl.release();
@@ -809,7 +810,7 @@ PassRefPtr<StringImpl> StringImpl::foldCase()
 
         // Do a slower implementation for cases that include non-ASCII Latin-1 characters.
         for (int32_t i = 0; i < length; ++i)
-            data[i] = static_cast<LChar>(Unicode::toLower(characters8()[i]));
+            data[i] = static_cast<LChar>(toLower(characters8()[i]));
 
         return newImpl.release();
     }
@@ -2094,4 +2095,20 @@ size_t StringImpl::sizeInBytes() const
     return size + sizeof(*this);
 }
 
+UChar32 toUpper(UChar32 c, const AtomicString& localeIdentifier)
+{
+    if (!localeIdentifier.isNull()) {
+        if (localeIdMatchesLang(localeIdentifier, "tr") || localeIdMatchesLang(localeIdentifier, "az")) {
+            if (c == 'i')
+                return 0x130; // Latin capital letter i with dot above

[tkent, 2015/05/28 00:38:26]

Please add latinCapitalLetterIWithDotAbove to wtf/unicode/CharacterNames.h.

[rwlbuis, 2015/05/28 23:03:08]

Done.

+            if (c == 0x131) // Latin small letter dotless i

[tkent, 2015/05/28 00:38:26]

Please add latinSmallLetterDotlessI to wtf/unicode/CharacterNames.h.

[rwlbuis, 2015/05/28 23:03:07]

Done.

+                return 'I';
+        } else if (localeIdMatchesLang(localeIdentifier, "lt")) {
+            // FIXME

[tkent, 2015/05/28 00:38:26]

FIXME -> TODO(rob.buis): blah blah

[rwlbuis, 2015/05/28 23:03:07]

In this case, I don't think I want to put my name there. I would not mind
working on it, but I don't know the language and don't know of a bug/testcase
for it. It is more a reminder that we'll have to do something based on
StringImpl::upper.

[tkent, 2015/05/29 05:49:50]

Please add TODO(rub.buis).
It doesn't mean you'll fix it.  Please refer to the discussion on blink-dev.

+        }
+    }
+
+    return toUpper(c);
+}
+
 } // namespace WTF