RegExp: Add support for table-based character class

src/regexp-macro-assembler-irregexp.cc

Index: src/regexp-macro-assembler-irregexp.cc
===================================================================
--- src/regexp-macro-assembler-irregexp.cc	(revision 11134)
+++ src/regexp-macro-assembler-irregexp.cc	(working copy)
@@ -352,6 +352,43 @@
 }
 
 
+void RegExpMacroAssemblerIrregexp::CheckCharacterInRange(
+    uc16 from,
+    uc16 to,
+    Label* on_in_range) {
+  Emit(BC_CHECK_CHAR_IN_RANGE, 0);
+  Emit16(from);
+  Emit16(to);
+  EmitOrLink(on_in_range);
+}
+
+
+void RegExpMacroAssemblerIrregexp::CheckCharacterNotInRange(
+    uc16 from,
+    uc16 to,
+    Label* on_not_in_range) {
+  Emit(BC_CHECK_CHAR_NOT_IN_RANGE, 0);
+  Emit16(from);
+  Emit16(to);
+  EmitOrLink(on_not_in_range);
+}
+
+
+void RegExpMacroAssemblerIrregexp::CheckBitInTable(
+    Handle<ByteArray> table, Label* on_bit_set) {
+  Emit(BC_CHECK_BIT_IN_TABLE, 0);
+  EmitOrLink(on_bit_set);
+  for (int i = 0; i < kTableSize; i += kBitsPerByte) {
+    int byte = 0;
+    for (int j = 0; j < kBitsPerByte; j++) {
+      byte <<= 1;
+      byte |= (table->get(i + j) != 0) ? 1 : 0;

[ulan, 2012/03/27 11:02:43]

I think it should be (table->get(i * kBitsPerByte + j)).

Also, I would prefer not reversing the "conventional" bit ordering. Consider
this:

if (table->get(i*kBitsPerByte + j)) {
  byte |= 1 << j;
}

This would also simplify decoding in the interpreter.

[Erik Corry, 2012/03/28 09:40:26]

Done, but I kept the get(i + j) because I think it's clearer and I didn't
include your implicit boolean conversion.

+    }
+    Emit8(byte);
+  }
+}
+
+
 void RegExpMacroAssemblerIrregexp::CheckNotBackReference(int start_reg,
                                                          Label* on_not_equal) {
   ASSERT(start_reg >= 0);