Add some support for the code-point code-unit distinction.

runtime/vm/object.cc

Index: runtime/vm/object.cc
diff --git a/runtime/vm/object.cc b/runtime/vm/object.cc
index 8e1ad649c9101ed396520087d9e4d5edd0e3d118..cea047f4d87ca273c2e84ec6596fd0932d576984 100644
--- a/runtime/vm/object.cc
+++ b/runtime/vm/object.cc
@@ -2187,7 +2187,7 @@ static bool MatchesAccessorName(const String& name,
     return false;
   }
   for (intptr_t i = 0; i < prefix_length; i++) {
-    if (name.CharAt(i) != prefix[i]) {
+    if (name.CharAt(i) != static_cast<int32_t>(prefix[i])) {
       return false;
     }
   }
@@ -9882,7 +9882,7 @@ intptr_t String::Hash(const uint16_t* characters, intptr_t len) {
 }
 
 
-intptr_t String::Hash(const uint32_t* characters, intptr_t len) {
+intptr_t String::Hash(const int32_t* characters, intptr_t len) {
   return HashImpl(characters, len);
 }
 
@@ -9892,16 +9892,16 @@ int32_t String::CharAt(intptr_t index) const {
   ASSERT(RawObject::IsStringClassId(class_id));
   NoGCScope no_gc;
   if (class_id == kOneByteStringCid) {
-      return *OneByteString::CharAddr(*this, index);
+      return OneByteString::CharAt(*this, index);

[cshapiro, 2012/11/17 02:25:27]

Why was this changed?  I think the code on the right is okay.

[erikcorry, 2012/11/19 12:40:41]

Reverted.

   }
   if (class_id == kTwoByteStringCid) {
-      return *TwoByteString::CharAddr(*this, index);
+      return TwoByteString::CharAt(*this, index);
   }
   if (class_id == kExternalOneByteStringCid) {
-      return *ExternalOneByteString::CharAddr(*this, index);
+      return ExternalOneByteString::CharAt(*this, index);
   }
   ASSERT(class_id == kExternalTwoByteStringCid);
-  return *ExternalTwoByteString::CharAddr(*this, index);
+  return ExternalTwoByteString::CharAt(*this, index);
 }
 
 
@@ -9957,8 +9957,13 @@ bool String::Equals(const char* str) const {
     intptr_t consumed = Utf8::Decode(reinterpret_cast<const uint8_t*>(str),
                                      len,
                                      &ch);
-    if (consumed == 0 || this->CharAt(i) != ch) {
-      return false;
+    if (consumed == 0) return false;
+
+    if (ch <= Utf16::kMaxCodeUnit) {
+      if (this->CharAt(i) != ch) return false;
+    } else {
+      if (Utf16::CodePointAt(*this, i) != ch) return false;
+      i++;
     }
     str += consumed;
     len -= consumed;
@@ -9997,16 +10002,18 @@ bool String::Equals(const uint16_t* characters, intptr_t len) const {
 }
 
 
-bool String::Equals(const uint32_t* characters, intptr_t len) const {
+bool String::Equals(const int32_t* characters, intptr_t len) const {
   if (len != this->Length()) {
     // Lengths don't match.
     return false;
   }
 
   for (intptr_t i = 0; i < len; i++) {
-    if (this->CharAt(i) != static_cast<int32_t>(characters[i])) {
+    int32_t c = this->CharAt(i);

[cshapiro, 2012/11/17 02:25:27]

This also looks like noise.

[erikcorry, 2012/11/19 12:40:41]

This function was an attempt to compare a UTF-32 string in a C array with a Dart
UTF-16 string.  It didn't work, and I have removed it and fixed the place it was
used (the scanner) to do its UTF-16 expansion itself.

+    if (c != characters[i]) {
       return false;
     }
+    if (c > Utf16::kMaxCodeUnit) i++;

[cshapiro, 2012/11/17 02:25:27]

I am confused about this line.  A comment would be helpful.

[erikcorry, 2012/11/19 12:40:41]

Ditto

   }

[siva, 2012/11/16 22:32:04]

I don't quite get this implementation.

[erikcorry, 2012/11/19 12:40:41]

Ditto.

   return true;
 }
@@ -10016,13 +10023,15 @@ intptr_t String::CompareTo(const String& other) const {
   const intptr_t this_len = this->Length();
   const intptr_t other_len = other.IsNull() ? 0 : other.Length();
   const intptr_t len = (this_len < other_len) ? this_len : other_len;
+  // UTF-16 has the high surrogate before the low surrogate so we can compare
+  // one code unit at a time for efficiency and still get the right ordering.
   for (intptr_t i = 0; i < len; i++) {
-    int32_t this_code_point = this->CharAt(i);
-    int32_t other_code_point = other.CharAt(i);
-    if (this_code_point < other_code_point) {
+    int32_t this_code_unit = this->CharAt(i);
+    int32_t other_code_unit = other.CharAt(i);
+    if (this_code_unit < other_code_unit) {
       return -1;
     }
-    if (this_code_point > other_code_point) {
+    if (this_code_unit > other_code_unit) {
       return 1;
     }
   }
@@ -10066,7 +10075,7 @@ RawString* String::New(const uint8_t* utf8_array,
                        intptr_t array_len,
                        Heap::Space space) {
   Utf8::Type type;
-  intptr_t len = Utf8::CodePointCount(utf8_array, array_len, &type);
+  intptr_t len = Utf8::CodeUnitCount(utf8_array, array_len, &type);
   if (type == Utf8::kAscii) {
     const String& strobj = String::Handle(OneByteString::New(len, space));
     if (len > 0) {
@@ -10102,7 +10111,7 @@ RawString* String::New(const uint16_t* utf16_array,
 }
 
 
-RawString* String::New(const uint32_t* utf32_array,
+RawString* String::New(const int32_t* utf32_array,
                        intptr_t array_len,
                        Heap::Space space) {
   bool is_one_byte_string = true;
@@ -10391,23 +10400,37 @@ RawString* String::Transform(int32_t (*mapping)(int32_t ch),
   bool has_mapping = false;
   int32_t dst_max = 0;
   intptr_t len = str.Length();
+  intptr_t out_len = 0;
   // TODO(cshapiro): assume a transform is required, rollback if not.
-  for (intptr_t i = 0; i < len; ++i) {
+  intptr_t i = 0;
+  for (; i < len; ++i) {
     int32_t src = str.CharAt(i);
+    if (Utf16::IsSurrogate(src)) break;
     int32_t dst = mapping(src);
     if (src != dst) {
       has_mapping = true;
     }
     dst_max = Utils::Maximum(dst_max, dst);
+    out_len += dst > Utf16::kMaxCodeUnit ? 2 : 1;
+  }
+  for (; i < len; ++i) {
+    int32_t src = Utf16::CodePointAt(str, i);
+    int32_t dst = mapping(src);
+    if (src != dst) {
+      has_mapping = true;
+    }
+    dst_max = Utils::Maximum(dst_max, dst);
+    out_len += dst > Utf16::kMaxCodeUnit ? 2 : 1;
+    if (src > Utf16::kMaxCodeUnit) ++i;
   }
   if (!has_mapping) {
     return str.raw();
   }
   if (dst_max <= 0x7F) {
-    return OneByteString::Transform(mapping, str, space);
+    return OneByteString::Transform(mapping, str, out_len, space);
   }
   ASSERT(dst_max > 0x7F);
-  return TwoByteString::Transform(mapping, str, space);
+  return TwoByteString::Transform(mapping, str, out_len, space);
 }
 
 
@@ -10561,7 +10584,7 @@ RawOneByteString* OneByteString::New(const uint16_t* characters,
 }
 
 
-RawOneByteString* OneByteString::New(const uint32_t* characters,
+RawOneByteString* OneByteString::New(const int32_t* characters,
                                      intptr_t len,
                                      Heap::Space space) {
   const String& result = String::Handle(OneByteString::New(len, space));
@@ -10614,14 +10637,18 @@ RawOneByteString* OneByteString::ConcatAll(const Array& strings,
 
 RawOneByteString* OneByteString::Transform(int32_t (*mapping)(int32_t ch),
                                            const String& str,
+                                           int out_length,
                                            Heap::Space space) {
   ASSERT(!str.IsNull());
   intptr_t len = str.Length();
-  const String& result = String::Handle(OneByteString::New(len, space));
-  for (intptr_t i = 0; i < len; ++i) {
-    int32_t ch = mapping(str.CharAt(i));
-    ASSERT(ch >= 0 && ch <= 0x7F);
-    *CharAddr(result, i) = ch;
+  const String& result =
+      String::Handle(OneByteString::New(out_length, space));
+  for (intptr_t i = 0, j = 0; i < len; ++i, j++) {
+    int32_t old_ch = str.CharAt(i);
+    if (old_ch > Utf16::kMaxCodeUnit) i++;
+    int32_t ch = mapping(old_ch);
+    ASSERT(ch <= 0x7F);
+    *CharAddr(result, j) = ch;
   }
   return OneByteString::raw(result);
 }
@@ -10696,7 +10723,7 @@ RawTwoByteString* TwoByteString::New(const uint16_t* utf16_array,
 
 
 RawTwoByteString* TwoByteString::New(intptr_t utf16_len,
-                                     const uint32_t* utf32_array,
+                                     const int32_t* utf32_array,
                                      intptr_t array_len,
                                      Heap::Space space) {
   ASSERT((array_len > 0) && (utf16_len >= array_len));
@@ -10705,9 +10732,11 @@ RawTwoByteString* TwoByteString::New(intptr_t utf16_len,
     NoGCScope no_gc;
     intptr_t j = 0;
     for (intptr_t i = 0; i < array_len; ++i) {
-      if (utf32_array[i] > 0xffff) {
+      int32_t code_point = utf32_array[i];
+      if (code_point > Utf16::kMaxCodeUnit) {
         ASSERT(j < (utf16_len - 1));
-        Utf8::ConvertUTF32ToUTF16(utf32_array[i], CharAddr(result, j));
+        *CharAddr(result, j) = Utf16::LeadFromCodePoint(code_point);
+        *CharAddr(result, j + 1) = Utf16::TrailFromCodePoint(code_point);
         j += 2;
       } else {
         ASSERT(j < utf16_len);
@@ -10761,14 +10790,25 @@ RawTwoByteString* TwoByteString::ConcatAll(const Array& strings,
 
 RawTwoByteString* TwoByteString::Transform(int32_t (*mapping)(int32_t ch),
                                            const String& str,
+                                           int out_length,
                                            Heap::Space space) {
   ASSERT(!str.IsNull());
   intptr_t len = str.Length();
-  const String& result = String::Handle(TwoByteString::New(len, space));
-  for (intptr_t i = 0; i < len; ++i) {
-    int32_t ch = mapping(str.CharAt(i));
-    ASSERT(ch >= 0 && ch <= 0xFFFF);
-    *CharAddr(result, i) = ch;
+  const String& result =
+      String::Handle(TwoByteString::New(out_length, space));
+  for (intptr_t i = 0, j = 0; i < len; ++i, j++) {
+    int32_t old_ch = Utf16::CodePointAt(str, i);
+    if (old_ch > Utf16::kMaxCodeUnit) i++;
+    int32_t ch = mapping(old_ch);
+    ASSERT(ch <= Utf16::kMaxCodePoint);

[siva, 2012/11/16 22:32:04]

We assert here that ch <= Utf16::kMaxCodePoint
but we have code below that deals with the case when
ch > Utf16::kMaxCodePoint

The ASSERT should be removed?

[erikcorry, 2012/11/19 12:40:41]

No, the assert checks against kMaxCodePoint (0x10ffff) and the 'if' checks
agains kMaxCodeUnit (0xffff).

+    if (ch <= Utf16::kMaxCodeUnit) {
+      *CharAddr(result, j) = ch;
+    } else {
+      *CharAddr(result, j) = Utf16::LeadFromCodePoint(ch);
+      *CharAddr(result, j + 1) = Utf16::TrailFromCodePoint(ch);
+      ++j;
+    }
+    ASSERT(j <= out_length);
   }
   return TwoByteString::raw(result);
 }