[Truffle] String#each_char is now multibyte-encoding aware.

Author: nirvdrum

spec/truffle/tags/core/string/chars_tags.txt

@@ -44,6 +44,7 @@
 import org.jruby.truffle.runtime.rubinius.RubiniusByteArray;
 import org.jruby.truffle.runtime.util.ArrayUtils;
 import org.jruby.util.ByteList;
+import org.jruby.util.CodeRangeSupport;
 import org.jruby.util.Pack;
 import org.jruby.util.StringSupport;
 import org.jruby.util.io.EncodingUtils;
@@ -847,20 +848,74 @@ public Object eachChar(VirtualFrame frame, RubyString string, @SuppressWarnings(
             return toEnumNode.call(frame, string, "to_enum", null, getContext().newSymbol("each_char"));
         }
 
-        @Specialization
+        @Specialization(guards = "isValidOr7BitEncoding")
         public RubyString eachChar(VirtualFrame frame, RubyString string, RubyProc block) {
-            notDesignedForCompilation();
+            ByteList strByteList = string.getByteList();
+            byte[] ptrBytes = strByteList.unsafeBytes();
+            int ptr = strByteList.begin();
+            int len = strByteList.getRealSize();
+            Encoding enc = string.getBytes().getEncoding();
 
-            // TODO (nirvdrum 04-Feb-15): This needs to support Ruby' encoding and code range semantics.  For now, this hack will suffice for very simple Strings.
-            final String javaString = string.toString();
+            final int stringLength = string.getBytes().length();
+            int n;
+
+            for (int i = 0; i < stringLength; i += n) {
+                n = StringSupport.encFastMBCLen(ptrBytes, ptr + i, ptr + len, enc);
 
-            for (int i = 0; i < javaString.length(); i++) {
-                yield(frame, block, getContext().makeString(javaString.charAt(i)));
+                yield(frame, block, substr(string, i, n));
             }
 
             return string;
         }
 
+        @Specialization(guards = "!isValidOr7BitEncoding")
+        public RubyString eachCharMultiByteEncoding(VirtualFrame frame, RubyString string, RubyProc block) {
+            ByteList strByteList = string.getByteList();
+            byte[] ptrBytes = strByteList.unsafeBytes();
+            int ptr = strByteList.begin();
+            int len = strByteList.getRealSize();
+            Encoding enc = string.getBytes().getEncoding();
+
+            final int stringLength = string.getBytes().length();
+            int n;
+
+            for (int i = 0; i < stringLength; i += n) {
+                n = multiByteStringLength(enc, ptrBytes, ptr + i, ptr + len);
+
+                yield(frame, block, substr(string, i, n));
+            }
+
+            return string;
+        }
+
+        public static boolean isValidOr7BitEncoding(RubyString string) {
+            return string.isCodeRangeValid() || CodeRangeSupport.isCodeRangeAsciiOnly(string);
+        }
+
+        @TruffleBoundary
+        private int multiByteStringLength(Encoding enc, byte[] bytes, int p, int end) {
+            return StringSupport.length(enc, bytes, p, end);
+        }
+
+        // TODO (nirvdrum 10-Mar-15): This was extracted from JRuby, but likely will need to become a Rubinius primitive.
+        private Object substr(RubyString string, int beg, int len) {
+            final ByteList bytes = string.getBytes();
+
+            int length = bytes.length();
+            if (len < 0 || beg > length) return getContext().getCoreLibrary().getNilObject();
+
+            if (beg < 0) {
+                beg += length;
+                if (beg < 0) getContext().getCoreLibrary().getNilObject();
+            }
+
+            int end = Math.min(length, beg + len);
+
+            final ByteList substringBytes = new ByteList(bytes, beg, end - beg);
+            substringBytes.setEncoding(bytes.getEncoding());
+
+            return getContext().makeString(string.getLogicalClass(), substringBytes);
+        }
     }
 
     @CoreMethod(names = "empty?")