immediate-char-values

~/ruby char-immediate-value ruby-2.0.0-github λ cat x.rb
str = "hello"

p [str[0], str[0].object_id]
p [str[1], str[1].object_id]
p [str[2], str[2].object_id]

~/ruby char-immediate-value ruby-2.0.0-github λ ./miniruby x.rb
["h", 26652]
["e", 25884]
["l", 27676]

patch.patch

diff --git include/ruby/encoding.h include/ruby/encoding.h
index 1c5d8da..acddf7f 100644
--- include/ruby/encoding.h
+++ include/ruby/encoding.h
@@ -34,7 +34,10 @@ RUBY_SYMBOL_EXPORT_BEGIN
 } while (0)
 #define ENCODING_SET(obj,i) rb_enc_set_index((obj), (i))
 
-#define ENCODING_GET_INLINED(obj) (int)((RBASIC(obj)->flags & ENCODING_MASK)>>ENCODING_SHIFT)
+#define ENCODING_GET_INLINED(obj) ( \
+    IMMSTR_P(obj) ? 0 : \
+  (int)((RBASIC(obj)->flags & ENCODING_MASK)>>ENCODING_SHIFT))
+
 #define ENCODING_GET(obj) \
     (ENCODING_GET_INLINED(obj) != ENCODING_INLINE_MAX ? \
      ENCODING_GET_INLINED(obj) : \
@@ -49,7 +52,9 @@ RUBY_SYMBOL_EXPORT_BEGIN
 #define ENC_CODERANGE_7BIT	((int)FL_USER8)
 #define ENC_CODERANGE_VALID	((int)FL_USER9)
 #define ENC_CODERANGE_BROKEN	((int)(FL_USER8|FL_USER9))
-#define ENC_CODERANGE(obj) ((int)RBASIC(obj)->flags & ENC_CODERANGE_MASK)
+#define ENC_CODERANGE(obj) ( \
+    IMMSTR_P(obj) ? ENC_CODERANGE_7BIT : \
+	(int)RBASIC(obj)->flags & ENC_CODERANGE_MASK)
 #define ENC_CODERANGE_ASCIIONLY(obj) (ENC_CODERANGE(obj) == ENC_CODERANGE_7BIT)
 #define ENC_CODERANGE_SET(obj,cr) (RBASIC(obj)->flags = \
 				   (RBASIC(obj)->flags & ~ENC_CODERANGE_MASK) | (cr))
diff --git include/ruby/ruby.h include/ruby/ruby.h
index d0ba1c2..ec9dfc3 100644
--- include/ruby/ruby.h
+++ include/ruby/ruby.h
@@ -355,6 +355,10 @@ rb_long2int_inline(long n)
 #define ID2SYM(x) (((VALUE)(x)<<RUBY_SPECIAL_SHIFT)|SYMBOL_FLAG)
 #define SYM2ID(x) RSHIFT((unsigned long)(x),RUBY_SPECIAL_SHIFT)
 
+#define IMMSTR_P(x) (((VALUE)(x)&~((~(VALUE)0)<<RUBY_SPECIAL_SHIFT))==STRING_FLAG)
+#define CHAR2IMM(x) (((VALUE)(x)<<RUBY_SPECIAL_SHIFT)|STRING_FLAG)
+#define IMM2CHAR(x) (char)RSHIFT((unsigned long)(x),RUBY_SPECIAL_SHIFT)
+
 #ifndef USE_FLONUM
 #if SIZEOF_VALUE >= SIZEOF_DOUBLE
 #define USE_FLONUM 1
@@ -377,6 +381,7 @@ rb_long2int_inline(long n)
 -------------------------
 ...xxxx xxx1 Fixnum
 ...0000 1110 Symbol
+...0001 1110 String (single character)
 ...0000 0000 Qfalse
 ...0000 0010 Qtrue
 ...0000 0100 Qnil
@@ -387,6 +392,7 @@ USE_FLONUM
 ...xxxx xxx1 Fixnum
 ...xxxx xx10 Flonum
 ...0000 1100 Symbol
+...0001 1100 String (single character)
 ...0000 0000 Qfalse  0x00 =  0
 ...0000 1000 Qnil    0x08 =  8
 ...0001 0100 Qtrue   0x14 = 20
@@ -406,6 +412,7 @@ enum ruby_special_consts {
     RUBY_FLONUM_MASK    = 0x03,
     RUBY_FLONUM_FLAG    = 0x02,
     RUBY_SYMBOL_FLAG    = 0x0c,
+    RUBY_STRING_FLAG    = 0x1c,
     RUBY_SPECIAL_SHIFT  = 8
 #else
     RUBY_Qfalse = 0,
@@ -418,6 +425,7 @@ enum ruby_special_consts {
     RUBY_FLONUM_MASK    = 0x00,	/* any values ANDed with FLONUM_MASK cannot be FLONUM_FLAG */
     RUBY_FLONUM_FLAG    = 0x02,
     RUBY_SYMBOL_FLAG    = 0x0e,
+    RUBY_SYMBOL_FLAG    = 0x1e,
     RUBY_SPECIAL_SHIFT  = 8
 #endif
 };
@@ -433,6 +441,7 @@ enum ruby_special_consts {
 #define FLONUM_FLAG RUBY_FLONUM_FLAG
 #endif
 #define SYMBOL_FLAG RUBY_SYMBOL_FLAG
+#define STRING_FLAG RUBY_STRING_FLAG
 
 #define RTEST(v) !(((VALUE)(v) & ~Qnil) == 0)
 #define NIL_P(v) !((VALUE)(v) != Qnil)
@@ -836,22 +845,26 @@ struct RString {
      (long)((RBASIC(str)->flags >> RSTRING_EMBED_LEN_SHIFT) & \
             (RSTRING_EMBED_LEN_MASK >> RSTRING_EMBED_LEN_SHIFT))
 #define RSTRING_LEN(str) \
-    (!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
-     RSTRING_EMBED_LEN(str) : \
-     RSTRING(str)->as.heap.len)
+    (IMMSTR_P(str) ? \
+	1 : \
+	(!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
+	RSTRING_EMBED_LEN(str) : \
+	    RSTRING(str)->as.heap.len))
 #define RSTRING_PTR(str) \
-    (!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
-     RSTRING(str)->as.ary : \
-     RSTRING(str)->as.heap.ptr)
+    (IMMSTR_P(str) ? \
+	((char*)&(str) + 1) : \
+	(!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
+	RSTRING(str)->as.ary : \
+	    RSTRING(str)->as.heap.ptr))
 #define RSTRING_END(str) \
-    (!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
-     (RSTRING(str)->as.ary + RSTRING_EMBED_LEN(str)) : \
-     (RSTRING(str)->as.heap.ptr + RSTRING(str)->as.heap.len))
+    (IMMSTR_P(str) ? \
+	((char*)&(str) + 2) : \
+	(!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
+	(RSTRING(str)->as.ary + RSTRING_EMBED_LEN(str)) : \
+	    (RSTRING(str)->as.heap.ptr + RSTRING(str)->as.heap.len)))
 #define RSTRING_LENINT(str) rb_long2int(RSTRING_LEN(str))
 #define RSTRING_GETMEM(str, ptrvar, lenvar) \
-    (!(RBASIC(str)->flags & RSTRING_NOEMBED) ? \
-     ((ptrvar) = RSTRING(str)->as.ary, (lenvar) = RSTRING_EMBED_LEN(str)) : \
-     ((ptrvar) = RSTRING(str)->as.heap.ptr, (lenvar) = RSTRING(str)->as.heap.len))
+    ((ptrvar) = RSTRING_PTR(str), (lenvar) = RSTRING_LEN(str))
 
 #define RARRAY_EMBED_LEN_MAX 3
 struct RArray {
@@ -1170,7 +1183,7 @@ struct RBignum {
     RBASIC(x)->flags |= RBASIC(s)->flags & FL_TAINT; \
 } while (0)
 
-#define OBJ_FROZEN(x) (!!(FL_ABLE(x)?(RBASIC(x)->flags&(FL_FREEZE)):(FIXNUM_P(x)||FLONUM_P(x)||SYMBOL_P(x))))
+#define OBJ_FROZEN(x) (!!(FL_ABLE(x)?(RBASIC(x)->flags&(FL_FREEZE)):(FIXNUM_P(x)||FLONUM_P(x)||SYMBOL_P(x)||IMMSTR_P(x))))
 #define OBJ_FREEZE(x) FL_SET((x), FL_FREEZE)
 
 #if USE_RGENGC
@@ -1603,6 +1616,7 @@ rb_class_of(VALUE obj)
 	if (FLONUM_P(obj)) return rb_cFloat;
 	if (obj == Qtrue)  return rb_cTrueClass;
 	if (SYMBOL_P(obj)) return rb_cSymbol;
+	if (IMMSTR_P(obj)) return rb_cString;
     }
     else if (!RTEST(obj)) {
 	if (obj == Qnil)   return rb_cNilClass;
@@ -1619,6 +1633,7 @@ rb_type(VALUE obj)
         if (FLONUM_P(obj)) return T_FLOAT;
         if (obj == Qtrue)  return T_TRUE;
 	if (SYMBOL_P(obj)) return T_SYMBOL;
+	if (IMMSTR_P(obj)) return T_STRING;
 	if (obj == Qundef) return T_UNDEF;
     }
     else if (!RTEST(obj)) {
@@ -1637,6 +1652,7 @@ rb_type(VALUE obj)
 	((type) == T_NIL) ? ((obj) == Qnil) : \
 	((type) == T_UNDEF) ? ((obj) == Qundef) : \
 	((type) == T_SYMBOL) ? SYMBOL_P(obj) : \
+	((type) == T_STRING && IMMSTR_P(obj)) ? 1 : \
         ((type) == T_FLOAT) ? RB_FLOAT_TYPE_P(obj) : \
 	(!SPECIAL_CONST_P(obj) && BUILTIN_TYPE(obj) == (type)))
 
diff --git numeric.c numeric.c
index 6979b21..bb1ccee 100644
--- numeric.c
+++ numeric.c
@@ -3256,11 +3256,13 @@ bit_coerce(VALUE *x, VALUE *y, int err)
 	do_coerce(x, y, err);
 	if (!FIXNUM_P(*x) && !RB_TYPE_P(*x, T_BIGNUM)
 	    && !FIXNUM_P(*y) && !RB_TYPE_P(*y, T_BIGNUM)) {
+	    VALUE y_str;
 	    if (!err) return FALSE;
+	    y_str = rb_inspect(*y);
 	    rb_raise(rb_eTypeError,
 		     "%s can't be coerced into %s for bitwise arithmetic",
 		     rb_special_const_p(*y) ?
-			RSTRING_PTR(rb_inspect(*y)) : rb_obj_classname(*y),
+			RSTRING_PTR(y_str) : rb_obj_classname(*y),
 		     rb_obj_classname(*x));
 	}
     }
diff --git string.c string.c
index 7bf4006..bcb48ce 100644
--- string.c
+++ string.c
@@ -1879,6 +1879,9 @@ rb_str_substr(VALUE str, long beg, long len)
     char *p = rb_str_subpos(str, beg, &len);
 
     if (!p) return Qnil;
+    if (len == 1 && ENCODING_GET(str) == ENCINDEX_UTF_8 || ENCODING_GET(str) == ENCINDEX_ASCII) {
+	return CHAR2IMM(*p);
+    }
     if (len > RSTRING_EMBED_LEN_MAX && p + len == RSTRING_END(str)) {
 	str2 = rb_str_new4(str);
 	str2 = str_new3(rb_obj_class(str2), str2);