Tentative implementation of string slices (hidden under the flag --string-slices).

src/objects.h

Index: src/objects.h
diff --git a/src/objects.h b/src/objects.h
index 9ef14dbdab6868ed3f2e5461b02be1351dc8f597..8acf9636703e9a4ecbe0d32a63b546838dc0aa69 100644
--- a/src/objects.h
+++ b/src/objects.h
@@ -86,6 +86,7 @@
 //         - SeqString
 //           - SeqAsciiString
 //           - SeqTwoByteString
+//         - SlicedString
 //         - ConsString
 //         - ExternalString
 //           - ExternalAsciiString
@@ -280,6 +281,7 @@ static const int kVariableSizeSentinel = 0;
   V(ASCII_STRING_TYPE)                                                         \
   V(CONS_STRING_TYPE)                                                          \
   V(CONS_ASCII_STRING_TYPE)                                                    \
+  V(SLICED_STRING_TYPE)                                                        \
   V(EXTERNAL_STRING_TYPE)                                                      \
   V(EXTERNAL_STRING_WITH_ASCII_DATA_TYPE)                                      \
   V(EXTERNAL_ASCII_STRING_TYPE)                                                \
@@ -396,6 +398,14 @@ static const int kVariableSizeSentinel = 0;
     ConsString::kSize,                                                         \
     cons_ascii_string,                                                         \
     ConsAsciiString)                                                           \
+  V(SLICED_STRING_TYPE,                                                        \
+    SlicedString::kSize,                                                       \
+    sliced_string,                                                             \
+    SlicedString)                                                              \
+  V(SLICED_ASCII_STRING_TYPE,                                                  \
+    SlicedString::kSize,                                                       \
+    sliced_ascii_string,                                                       \
+    SlicedAsciiString)                                                         \
   V(EXTERNAL_STRING_TYPE,                                                      \
     ExternalTwoByteString::kSize,                                              \
     external_string,                                                           \
@@ -469,9 +479,15 @@ const uint32_t kStringRepresentationMask = 0x03;
 enum StringRepresentationTag {
   kSeqStringTag = 0x0,
   kConsStringTag = 0x1,
-  kExternalStringTag = 0x2
+  kExternalStringTag = 0x2,
+  kSlicedStringTag = 0x3
 };
-const uint32_t kIsConsStringMask = 0x1;
+const uint32_t kIsIndirectStringMask = 0x1;
+const uint32_t kIsIndirectStringTag = 0x1;
+STATIC_ASSERT(
+    (kConsStringTag & kIsIndirectStringMask) == kIsIndirectStringTag);
+STATIC_ASSERT(
+    (kSlicedStringTag & kIsIndirectStringMask) == kIsIndirectStringTag);
 
 // If bit 7 is clear, then bit 3 indicates whether this two-byte
 // string actually contains ascii data.
@@ -506,6 +522,8 @@ enum InstanceType {
   ASCII_STRING_TYPE = kAsciiStringTag | kSeqStringTag,
   CONS_STRING_TYPE = kTwoByteStringTag | kConsStringTag,
   CONS_ASCII_STRING_TYPE = kAsciiStringTag | kConsStringTag,
+  SLICED_STRING_TYPE = kTwoByteStringTag | kSlicedStringTag,
+  SLICED_ASCII_STRING_TYPE = kAsciiStringTag | kSlicedStringTag,
   EXTERNAL_STRING_TYPE = kTwoByteStringTag | kExternalStringTag,
   EXTERNAL_STRING_WITH_ASCII_DATA_TYPE =
       kTwoByteStringTag | kExternalStringTag | kAsciiDataHintTag,
@@ -709,6 +727,7 @@ class MaybeObject BASE_EMBEDDED {
   V(SeqString)                                 \
   V(ExternalString)                            \
   V(ConsString)                                \
+  V(SlicedString)                              \
   V(ExternalTwoByteString)                     \
   V(ExternalAsciiString)                       \
   V(SeqTwoByteString)                          \
@@ -5709,6 +5728,8 @@ class StringShape BASE_EMBEDDED {
   inline bool IsSequential();
   inline bool IsExternal();
   inline bool IsCons();
+  inline bool IsSliced();
+  inline bool IsIndirect();
   inline bool IsExternalAscii();
   inline bool IsExternalTwoByte();
   inline bool IsSequentialAscii();
@@ -5794,6 +5815,18 @@ class String: public HeapObject {
   // string.
   inline String* TryFlattenGetString(PretenureFlag pretenure = NOT_TENURED);
 
+  // Equivalent to Flatten for string slices. A new parent string is created
+  // with exactly the required characters. Offset is set to 0.
+  inline MaybeObject* TryTruncate(PretenureFlag pretenure = NOT_TENURED);
+  inline String* TryTruncateGetString(PretenureFlag pretenure = NOT_TENURED);
+
+  // Combines flatten and truncate.
+  inline MaybeObject* TryFlattenOrTruncate(PretenureFlag pretenure
+                                           = NOT_TENURED);
+
+  // Returns the parent of a sliced string or first part of a cons string.
+  inline String* GetIndirect();
+
   Vector<const char> ToAsciiVector();
   Vector<const uc16> ToUC16Vector();
 
@@ -5881,6 +5914,8 @@ class String: public HeapObject {
   void StringVerify();
 #endif
   inline bool IsFlat();
+  inline bool IsTruncated();
+  inline bool IsFlatAndTruncated();
 
   // Layout description.
   static const int kLengthOffset = HeapObject::kHeaderSize;
@@ -6043,6 +6078,11 @@ class String: public HeapObject {
   // mutates the ConsString and might return a failure.
   MUST_USE_RESULT MaybeObject* SlowTryFlatten(PretenureFlag pretenure);
 
+  // Try to create a new sequential string as parent for the slice that is
+  // as short as possible. This is a no-op unless the string is a SlicedString.
+  // Truncating mutates the SlicedString and might return a failure.
+  MUST_USE_RESULT MaybeObject* SlowTryTruncate(PretenureFlag pretenure);
+
   static inline bool IsHashFieldComputed(uint32_t field);
 
   // Slow case of String::Equals.  This implementation works on any strings
@@ -6223,11 +6263,57 @@ class ConsString: public String {
   typedef FixedBodyDescriptor<kFirstOffset, kSecondOffset + kPointerSize, kSize>
           BodyDescriptor;
 
+#ifdef DEBUG
+  void ConsStringVerify();
+#endif
+
  private:
   DISALLOW_IMPLICIT_CONSTRUCTORS(ConsString);
 };
 
 
+class SlicedString: public String {

[Vitaly Repeshko, 2011/08/05 12:14:14]

Please document classes. Explain that only slices of flat strings are allowed
and that nested slices are automatically simplified.

+ public:
+
+  inline String* parent();
+  inline void set_parent(String* parent);
+  inline int offset();
+  inline void set_offset(int offset);
+
+  // Dispatched behavior.
+  uint16_t SlicedStringGet(int index);
+
+  // Casting.
+  static inline SlicedString* cast(Object* obj);
+
+  // Layout description.
+  static const int kParentOffset = POINTER_SIZE_ALIGN(String::kSize);
+  static const int kOffsetOffset = kParentOffset + kPointerSize;
+  static const int kSize = kOffsetOffset + kPointerSize;
+
+  // Support for StringInputBuffer
+  inline const unibrow::byte* SlicedStringReadBlock(ReadBlockBuffer* buffer,
+                                                    unsigned* offset_ptr,
+                                                    unsigned chars);
+  inline void SlicedStringReadBlockIntoBuffer(ReadBlockBuffer* buffer,
+                                              unsigned* offset_ptr,
+                                              unsigned chars);
+  // Minimum length for a sliced string.
+  static const int kMinLength = 13;
+
+  typedef FixedBodyDescriptor<kParentOffset,
+                              kOffsetOffset + kPointerSize, kSize>
+          BodyDescriptor;
+
+#ifdef DEBUG
+  void SlicedStringVerify();
+#endif
+
+ private:
+  DISALLOW_IMPLICIT_CONSTRUCTORS(SlicedString);
+};
+
+
 // The ExternalString class describes string values that are backed by
 // a string resource that lies outside the V8 heap.  ExternalStrings
 // consist of the length field common to all strings, a pointer to the