Revert of Extend big-disjunction optimization to case-independent regexps

src/jsregexp.cc

Index: src/jsregexp.cc
diff --git a/src/jsregexp.cc b/src/jsregexp.cc
index 49a2998fed5c7d7222a36c6b9da8ad553921ea90..e284e8cb15f3233eee2533bc9efda8b6735bd1e3 100644
--- a/src/jsregexp.cc
+++ b/src/jsregexp.cc
@@ -4828,34 +4828,6 @@
 }
 
 
-static unibrow::uchar Canonical(
-    unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize,
-    unibrow::uchar c) {
-  unibrow::uchar chars[unibrow::Ecma262Canonicalize::kMaxWidth];
-  int length = canonicalize->get(c, '\0', chars);
-  DCHECK_LE(length, 1);
-  unibrow::uchar canonical = c;
-  if (length == 1) canonical = chars[0];
-  return canonical;
-}
-
-
-int CompareFirstCharCaseIndependent(
-    unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize,
-    RegExpTree* const* a, RegExpTree* const* b) {
-  RegExpAtom* atom1 = (*a)->AsAtom();
-  RegExpAtom* atom2 = (*b)->AsAtom();
-  unibrow::uchar character1 = atom1->data().at(0);
-  unibrow::uchar character2 = atom2->data().at(0);
-  if (character1 == character2) return 0;
-  if (character1 >= 'a' || character2 >= 'a') {
-    character1 = Canonical(canonicalize, character1);
-    character2 = Canonical(canonicalize, character2);
-  }
-  return static_cast<int>(character1) - static_cast<int>(character2);
-}
-
-
 // We can stable sort runs of atoms, since the order does not matter if they
 // start with different characters.
 // Returns true if any consecutive atoms were found.
@@ -4879,23 +4851,15 @@
       i++;
     }
     // Sort atoms to get ones with common prefixes together.
-    // This step is more tricky if we are in a case-independent regexp,
+    // This step is not valid if we are in a case-independent regexp,
     // because it would change /is|I/ to /I|is/, and order matters when
     // the regexp parts don't match only disjoint starting points. To fix
-    // this we have a version of CompareFirstChar that uses case-
+    // this would need a version of CompareFirstChar that uses case-
     // independent character classes for comparison.
-    DCHECK_LT(first_atom, alternatives->length());
-    DCHECK_LE(i, alternatives->length());
-    DCHECK_LE(first_atom, i);
-    if (compiler->ignore_case()) {
-      unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize =
-          compiler->isolate()->regexp_macro_assembler_canonicalize();
-      auto compare_closure =
-          [canonicalize](RegExpTree* const* a, RegExpTree* const* b) {
-            return CompareFirstCharCaseIndependent(canonicalize, a, b);
-          };
-      alternatives->StableSort(compare_closure, first_atom, i - first_atom);
-    } else {
+    if (!compiler->ignore_case()) {
+      DCHECK_LT(first_atom, alternatives->length());
+      DCHECK_LE(i, alternatives->length());
+      DCHECK_LE(first_atom, i);
       alternatives->StableSort(CompareFirstChar, first_atom, i - first_atom);
     }
     if (i - first_atom > 1) found_consecutive_atoms = true;
@@ -4920,7 +4884,7 @@
       continue;
     }
     RegExpAtom* atom = alternative->AsAtom();
-    unibrow::uchar common_prefix = atom->data().at(0);
+    uc16 common_prefix = atom->data().at(0);
     int first_with_prefix = i;
     int prefix_length = atom->length();
     i++;
@@ -4928,15 +4892,7 @@
       alternative = alternatives->at(i);
       if (!alternative->IsAtom()) break;
       atom = alternative->AsAtom();
-      unibrow::uchar new_prefix = atom->data().at(0);
-      if (new_prefix != common_prefix) {
-        if (!compiler->ignore_case()) break;
-        unibrow::Mapping<unibrow::Ecma262Canonicalize>* canonicalize =
-            compiler->isolate()->regexp_macro_assembler_canonicalize();
-        new_prefix = Canonical(canonicalize, new_prefix);
-        common_prefix = Canonical(canonicalize, common_prefix);
-        if (new_prefix != common_prefix) break;
-      }
+      if (atom->data().at(0) != common_prefix) break;
       prefix_length = Min(prefix_length, atom->length());
       i++;
     }
@@ -4952,10 +4908,7 @@
         RegExpAtom* old_atom =
             alternatives->at(j + first_with_prefix)->AsAtom();
         for (int k = 1; k < prefix_length; k++) {
-          if (atom->data().at(k) != old_atom->data().at(k)) {
-            prefix_length = k;
-            break;
-          }
+          if (atom->data().at(k) != old_atom->data().at(k)) prefix_length = k;
         }
       }
       RegExpAtom* prefix =