Fredrik Lundh: here's the 96.6% version of SRE

6 files changed, 743 insertions, 303 deletions

diff --git a/Lib/sre.py b/Lib/sre.py
index 79878b3..455cd27 100644
--- a/Lib/sre.py
+++ b/Lib/sre.py
@@ -1,4 +1,3 @@
-# -*- Mode: Python; tab-width: 4 -*-
 #
 # Secret Labs' Regular Expression Engine
 # $Id$
@@ -7,39 +6,127 @@
 #
 # Copyright (c) 1998-2000 by Secret Labs AB.  All rights reserved.
 #
-# This code can only be used for 1.6 alpha testing.  All other use
-# require explicit permission from Secret Labs AB.
-#
 # Portions of this engine have been developed in cooperation with
 # CNRI.  Hewlett-Packard provided funding for 1.6 integration and
 # other compatibility work.
 #
 
-"""
-this is a long string
-"""
-
 import sre_compile
 
+# flags
+I = IGNORECASE = sre_compile.SRE_FLAG_IGNORECASE
+L = LOCALE = sre_compile.SRE_FLAG_LOCALE
+M = MULTILINE = sre_compile.SRE_FLAG_MULTILINE
+S = DOTALL = sre_compile.SRE_FLAG_DOTALL
+X = VERBOSE = sre_compile.SRE_FLAG_VERBOSE
+
 # --------------------------------------------------------------------
 # public interface
 
-def compile(pattern, flags=0):
-    return sre_compile.compile(pattern, _fixflags(flags))
+# FIXME: add docstrings
 
 def match(pattern, string, flags=0):
-    return compile(pattern, _fixflags(flags)).match(string)
+    return _compile(pattern, flags).match(string)
 
 def search(pattern, string, flags=0):
-    return compile(pattern, _fixflags(flags)).search(string)
+    return _compile(pattern, flags).search(string)
+
+def sub(pattern, repl, string, count=0):
+    return _compile(pattern).sub(repl, string, count)
+
+def subn(pattern, repl, string, count=0):
+    return _compile(pattern).subn(repl, string, count)
+
+def split(pattern, string, maxsplit=0):
+    return _compile(pattern).split(string, maxsplit)
 
-# FIXME: etc
+def findall(pattern, string, maxsplit=0):
+    return _compile(pattern).findall(string, maxsplit)
+
+def compile(pattern, flags=0):
+    return _compile(pattern, flags)
+
+def escape(pattern):
+    s = list(pattern)
+    for i in range(len(pattern)):
+        c = pattern[i]
+        if not ("a" <= c <= "z" or "A" <= c <= "Z" or "0" <= c <= "9"):
+            if c == "\000":
+                s[i] = "\\000"
+            else:
+                s[i] = "\\" + c
+    return pattern[:0].join(s)
 
 # --------------------------------------------------------------------
-# helpers
+# internals
+
+_cache = {}
+_MAXCACHE = 100
+
+def _compile(pattern, flags=0):
+    # internal: compile pattern
+    tp = type(pattern)
+    if tp not in (type(""), type(u"")):
+        return pattern
+    key = (tp, pattern, flags)
+    try:
+        return _cache[key]
+    except KeyError:
+        pass
+    p = sre_compile.compile(pattern, flags)
+    if len(_cache) >= _MAXCACHE:
+        _cache.clear()
+    _cache[key] = p
+    return p
+
+def _sub(pattern, template, string, count=0):
+    # internal: pattern.sub implementation hook
+    return _subn(pattern, template, string, count)[0]
+
+def _expand(match, template):
+    # internal: expand template
+    return template # FIXME
 
-def _fixflags(flags):
-    # convert flag bitmask to sequence
-    assert not flags
-    return ()
+def _subn(pattern, template, string, count=0):
+    # internal: pattern.subn implementation hook
+    if callable(template):
+        filter = callable
+    else:
+        # FIXME: prepare template
+        def filter(match, template=template):
+            return _expand(match, template)
+    n = i = 0
+    s = []
+    append = s.append
+    c = pattern.cursor(string)
+    while not count or n < count:
+        m = c.search()
+        if not m:
+            break
+        j = m.start()
+        if j > i:
+            append(string[i:j])
+        append(filter(m))
+        i = m.end()
+        n = n + 1
+    if i < len(string):
+        append(string[i:])
+    return string[:0].join(s), n
 
+def _split(pattern, string, maxsplit=0):
+    # internal: pattern.split implementation hook
+    n = i = 0
+    s = []
+    append = s.append
+    c = pattern.cursor(string)
+    while not maxsplit or n < maxsplit:
+        m = c.search()
+        if not m:
+            break
+        j = m.start()
+        append(string[i:j])
+        i = m.end()
+        n = n + 1
+    if i < len(string):
+        append(string[i:])
+    return s
diff --git a/Lib/sre_compile.py b/Lib/sre_compile.py
index 8738061..53da005 100644
--- a/Lib/sre_compile.py
+++ b/Lib/sre_compile.py
@@ -14,9 +14,6 @@
 # other compatibility work.
 #
 
-# FIXME: <fl> formalize (objectify?) and document the compiler code
-# format, so that other frontends can use this compiler
-
 import array, string, sys
 
 import _sre
@@ -45,64 +42,70 @@ class Code:
 	self.data.append(code)
     def todata(self):
 	# print self.data
-	return array.array(WORDSIZE, self.data).tostring()
-
-def _lower(literal):
-    # return _sre._lower(literal) # FIXME
-    return string.lower(literal)
+	try:
+	    return array.array(WORDSIZE, self.data).tostring()
+	except OverflowError:
+	    print self.data
+	    raise
 
-def _compile(code, pattern, flags):
+def _compile(code, pattern, flags, level=0):
     append = code.append
     for op, av in pattern:
 	if op is ANY:
-	    if "s" in flags:
-		append(CODES[op]) # any character at all!
+	    if flags & SRE_FLAG_DOTALL:
+		append(OPCODES[op]) # any character at all!
 	    else:
-		append(CODES[NOT_LITERAL])
-		append(10)
+		append(OPCODES[CATEGORY])
+		append(CHCODES[CATEGORY_NOT_LINEBREAK])
 	elif op in (SUCCESS, FAILURE):
-	    append(CODES[op])
+	    append(OPCODES[op])
 	elif op is AT:
-	    append(CODES[op])
-	    append(POSITIONS[av])
+	    append(OPCODES[op])
+	    if flags & SRE_FLAG_MULTILINE:
+		append(ATCODES[AT_MULTILINE[av]])
+	    else:
+		append(ATCODES[av])
 	elif op is BRANCH:
-	    append(CODES[op])
+	    append(OPCODES[op])
 	    tail = []
 	    for av in av[1]:
 		skip = len(code); append(0)
-		_compile(code, av, flags)
-		append(CODES[JUMP])
+		_compile(code, av, flags, level)
+		append(OPCODES[JUMP])
 		tail.append(len(code)); append(0)
 		code[skip] = len(code) - skip
 	    append(0) # end of branch
 	    for tail in tail:
 		code[tail] = len(code) - tail
 	elif op is CALL:
-	    append(CODES[op])
+	    append(OPCODES[op])
 	    skip = len(code); append(0)
-	    _compile(code, av, flags)
-	    append(CODES[SUCCESS])
+	    _compile(code, av, flags, level+1)
+	    append(OPCODES[SUCCESS])
 	    code[skip] = len(code) - skip
 	elif op is CATEGORY: # not used by current parser
-	    append(CODES[op])
-	    append(CATEGORIES[av])
+	    append(OPCODES[op])
+	    if flags & SRE_FLAG_LOCALE:
+		append(CH_LOCALE[CHCODES[av]])
+	    else:
+		append(CHCODES[av])
 	elif op is GROUP:
-	    if "i" in flags:
-		append(CODES[MAP_IGNORE[op]])
+	    if flags & SRE_FLAG_IGNORECASE:
+		append(OPCODES[OP_IGNORE[op]])
 	    else:
-		append(CODES[op])
-	    append(av)
+		append(OPCODES[op])
+	    append(av-1)
 	elif op is IN:
-	    if "i" in flags:
-		append(CODES[MAP_IGNORE[op]])
+	    if flags & SRE_FLAG_IGNORECASE:
+		append(OPCODES[OP_IGNORE[op]])
 		def fixup(literal):
-		    return ord(_lower(literal))
+		    return ord(literal.lower())
 	    else:
-		append(CODES[op])
+		append(OPCODES[op])
 		fixup = ord
 	    skip = len(code); append(0)
 	    for op, av in av:
-		append(CODES[op])
+		append(OPCODES[op])
 		if op is NEGATE:
 		    pass
 		elif op is LITERAL:
@@ -111,58 +114,60 @@ def _compile(code, pattern, flags):
 		    append(fixup(av[0]))
 		    append(fixup(av[1]))
 		elif op is CATEGORY:
-		    append(CATEGORIES[av])
+		    if flags & SRE_FLAG_LOCALE:
+			append(CH_LOCALE[CHCODES[av]])
+		    else:
+			append(CHCODES[av])
 		else:
 		    raise ValueError, "unsupported set operator"
-	    append(CODES[FAILURE])
+	    append(OPCODES[FAILURE])
 	    code[skip] = len(code) - skip
 	elif op in (LITERAL, NOT_LITERAL):
-	    if "i" in flags:
-		append(CODES[MAP_IGNORE[op]])
-		append(ord(_lower(av)))
+	    if flags & SRE_FLAG_IGNORECASE:
+		append(OPCODES[OP_IGNORE[op]])
+		append(ord(av.lower()))
 	    else:
-		append(CODES[op])
+		append(OPCODES[op])
 		append(ord(av))
 	elif op is MARK:
-	    append(CODES[op])
+	    append(OPCODES[op])
 	    append(av)
  	elif op in (REPEAT, MIN_REPEAT, MAX_REPEAT):
 	    lo, hi = av[2].getwidth()
  	    if lo == 0:
  		raise SyntaxError, "cannot repeat zero-width items"
 	    if lo == hi == 1 and op is MAX_REPEAT:
-		append(CODES[MAX_REPEAT_ONE])
+		append(OPCODES[MAX_REPEAT_ONE])
 		skip = len(code); append(0)
 		append(av[0])
 		append(av[1])
-		_compile(code, av[2], flags)
-		append(CODES[SUCCESS])
+		_compile(code, av[2], flags, level+1)
+		append(OPCODES[SUCCESS])
 		code[skip] = len(code) - skip
 	    else:
-		append(CODES[op])
+		append(OPCODES[op])
 		skip = len(code); append(0)
 		append(av[0])
 		append(av[1])
-		_compile(code, av[2], flags)
+		_compile(code, av[2], flags, level+1)
 		if op is MIN_REPEAT:
-		    append(CODES[MIN_UNTIL])
+		    append(OPCODES[MIN_UNTIL])
 		else:
-		    # FIXME: MAX_REPEAT PROBABLY DOESN'T WORK (?)
-		    append(CODES[MAX_UNTIL])
+		    append(OPCODES[MAX_UNTIL])
 		code[skip] = len(code) - skip
 	elif op is SUBPATTERN:
-## 	    group = av[0]
-## 	    if group:
-## 		append(CODES[MARK])
-## 		append((group-1)*2)
-	    _compile(code, av[1], flags)
-## 	    if group:
-## 		append(CODES[MARK])
-## 		append((group-1)*2+1)
+ 	    group = av[0]
+ 	    if group:
+ 		append(OPCODES[MARK])
+ 		append((group-1)*2)
+	    _compile(code, av[1], flags, level+1)
+ 	    if group:
+ 		append(OPCODES[MARK])
+ 		append((group-1)*2+1)
 	else:
 	    raise ValueError, ("unsupported operand type", op)
 
-def compile(p, flags=()):
+def compile(p, flags=0):
     # convert pattern list to internal format
     if type(p) in (type(""), type(u"")):
 	import sre_parse
@@ -170,12 +175,10 @@ def compile(p, flags=()):
 	p = sre_parse.parse(p)
     else:
 	pattern = None
-    # print p.getwidth()
-    # print p
+    flags = p.pattern.flags | flags
     code = Code()
-    _compile(code, p.data, p.pattern.flags)
-    code.append(CODES[SUCCESS])
-    # print list(code.data)
+    _compile(code, p.data, flags)
+    code.append(OPCODES[SUCCESS])
     data = code.todata()
     if 0: # debugging
 	print
@@ -183,5 +186,8 @@ def compile(p, flags=()):
 	import sre_disasm
 	sre_disasm.disasm(data)
 	print "-" * 68
-    # print len(data), p.pattern.groups, len(p.pattern.groupdict)
-    return _sre.compile(pattern, data, p.pattern.groups-1, p.pattern.groupdict)
+    return _sre.compile(
+	pattern, flags,
+	data,
+	p.pattern.groups-1, p.pattern.groupdict
+	)
diff --git a/Lib/sre_constants.py b/Lib/sre_constants.py
index af88309..531dc31 100644
--- a/Lib/sre_constants.py
+++ b/Lib/sre_constants.py
@@ -48,20 +48,31 @@ SUBPATTERN = "subpattern"
 
 # positions
 AT_BEGINNING = "at_beginning"
+AT_BEGINNING_LINE = "at_beginning_line"
 AT_BOUNDARY = "at_boundary"
 AT_NON_BOUNDARY = "at_non_boundary"
 AT_END = "at_end"
+AT_END_LINE = "at_end_line"
 
 # categories
-
 CATEGORY_DIGIT = "category_digit"
 CATEGORY_NOT_DIGIT = "category_not_digit"
 CATEGORY_SPACE = "category_space"
 CATEGORY_NOT_SPACE = "category_not_space"
 CATEGORY_WORD = "category_word"
 CATEGORY_NOT_WORD = "category_not_word"
+CATEGORY_LINEBREAK = "category_linebreak"
+CATEGORY_NOT_LINEBREAK = "category_not_linebreak"
+CATEGORY_LOC_DIGIT = "category_loc_digit"
+CATEGORY_LOC_NOT_DIGIT = "category_loc_not_digit"
+CATEGORY_LOC_SPACE = "category_loc_space"
+CATEGORY_LOC_NOT_SPACE = "category_loc_not_space"
+CATEGORY_LOC_WORD = "category_loc_word"
+CATEGORY_LOC_NOT_WORD = "category_loc_not_word"
+CATEGORY_LOC_LINEBREAK = "category_loc_linebreak"
+CATEGORY_LOC_NOT_LINEBREAK = "category_loc_not_linebreak"
 
-CODES = [
+OPCODES = [
 
     # failure=0 success=1 (just because it looks better that way :-)
     FAILURE, SUCCESS,
@@ -87,45 +98,75 @@ CODES = [
 
 ]
 
-# convert to dictionary
-c = {}
-i = 0
-for code in CODES:
-    c[code] = i
-    i = i + 1
-CODES = c
+ATCODES = [
+    AT_BEGINNING, AT_BEGINNING_LINE, AT_BOUNDARY,
+    AT_NON_BOUNDARY, AT_END, AT_END_LINE
+]
+
+CHCODES = [
+    CATEGORY_DIGIT, CATEGORY_NOT_DIGIT, CATEGORY_SPACE,
+    CATEGORY_NOT_SPACE, CATEGORY_WORD, CATEGORY_NOT_WORD,
+    CATEGORY_LINEBREAK, CATEGORY_NOT_LINEBREAK, CATEGORY_LOC_DIGIT,
+    CATEGORY_LOC_NOT_DIGIT, CATEGORY_LOC_SPACE,
+    CATEGORY_LOC_NOT_SPACE, CATEGORY_LOC_WORD, CATEGORY_LOC_NOT_WORD,
+    CATEGORY_LOC_LINEBREAK, CATEGORY_LOC_NOT_LINEBREAK
+]
+
+def makedict(list):
+    d = {}
+    i = 0
+    for item in list:
+	d[item] = i
+	i = i + 1
+    return d
+
+OPCODES = makedict(OPCODES)
+ATCODES = makedict(ATCODES)
+CHCODES = makedict(CHCODES)
 
 # replacement operations for "ignore case" mode
-MAP_IGNORE = {
+OP_IGNORE = {
     GROUP: GROUP_IGNORE,
     IN: IN_IGNORE,
     LITERAL: LITERAL_IGNORE,
     NOT_LITERAL: NOT_LITERAL_IGNORE
 }
 
-POSITIONS = {
-    AT_BEGINNING: ord("a"),
-    AT_BOUNDARY: ord("b"),
-    AT_NON_BOUNDARY: ord("B"),
-    AT_END: ord("z"),
+AT_MULTILINE = {
+    AT_BEGINNING: AT_BEGINNING_LINE,
+    AT_END: AT_END_LINE
 }
 
-CATEGORIES = {
-    CATEGORY_DIGIT: ord("d"),
-    CATEGORY_NOT_DIGIT: ord("D"),
-    CATEGORY_SPACE: ord("s"),
-    CATEGORY_NOT_SPACE: ord("S"),
-    CATEGORY_WORD: ord("w"),
-    CATEGORY_NOT_WORD: ord("W"),
+CH_LOCALE = {
+    CATEGORY_DIGIT: CATEGORY_LOC_DIGIT,
+    CATEGORY_NOT_DIGIT: CATEGORY_LOC_NOT_DIGIT,
+    CATEGORY_SPACE: CATEGORY_LOC_SPACE,
+    CATEGORY_NOT_SPACE: CATEGORY_LOC_NOT_SPACE,
+    CATEGORY_WORD: CATEGORY_LOC_WORD,
+    CATEGORY_NOT_WORD: CATEGORY_LOC_NOT_WORD,
+    CATEGORY_LINEBREAK: CATEGORY_LOC_LINEBREAK,
+    CATEGORY_NOT_LINEBREAK: CATEGORY_LOC_NOT_LINEBREAK
 }
 
+# flags
+SRE_FLAG_TEMPLATE = 1 # NYI
+SRE_FLAG_IGNORECASE = 2
+SRE_FLAG_LOCALE = 4
+SRE_FLAG_MULTILINE = 8
+SRE_FLAG_DOTALL = 16
+SRE_FLAG_VERBOSE = 32
+
 if __name__ == "__main__":
     import string
-    items = CODES.items()
-    items.sort(lambda a, b: cmp(a[1], b[1]))
+    def dump(f, d, prefix):
+	items = d.items()
+	items.sort(lambda a, b: cmp(a[1], b[1]))
+	for k, v in items:
+	    f.write("#define %s_%s %s\n" % (prefix, string.upper(k), v))
     f = open("sre_constants.h", "w")
-    f.write("/* generated by sre_constants.py */\n")
-    for k, v in items:
-	f.write("#define SRE_OP_" + string.upper(k) + " " + str(v) + "\n")
+    f.write("/* generated from sre_constants.py */\n")
+    dump(f, OPCODES, "SRE_OP")
+    dump(f, ATCODES, "SRE")
+    dump(f, CHCODES, "SRE")
     f.close()
     print "done"
diff --git a/Modules/_sre.c b/Modules/_sre.c
index 47b80c5..8d46844 100644
--- a/Modules/_sre.c
+++ b/Modules/_sre.c
@@ -6,13 +6,16 @@
  * simple regular expression matching engine
  *
  * partial history:
- * 99-10-24 fl	created (bits and pieces from the template matcher)
+ * 99-10-24 fl	created (based on the template matcher)
  * 99-11-13 fl	added categories, branching, and more (0.2)
  * 99-11-16 fl	some tweaks to compile on non-Windows platforms
  * 99-12-18 fl	non-literals, generic maximizing repeat (0.3)
  * 99-02-28 fl	tons of changes (not all to the better ;-) (0.4)
  * 99-03-06 fl	first alpha, sort of (0.5)
  * 99-03-14 fl	removed most compatibility stuff (0.6)
+ * 99-05-10 fl	towards third alpha (0.8.2)
+ * 99-05-13 fl	added experimental cursor stuff (0.8.3)
+ * 99-05-27 fl	final bug hunt (0.8.4)
  *
  * Copyright (c) 1997-2000 by Secret Labs AB.  All rights reserved.
  *
@@ -26,7 +29,7 @@
 
 #ifndef SRE_RECURSIVE
 
-char copyright[] = " SRE 0.6 Copyright (c) 1997-2000 by Secret Labs AB ";
+char copyright[] = " SRE 0.8.4 Copyright (c) 1997-2000 by Secret Labs AB ";
 
 #include "Python.h"
 
@@ -40,7 +43,7 @@ char copyright[] = " SRE 0.6 Copyright (c) 1997-2000 by Secret Labs AB ";
 #define INT_MAX 2147483647
 #endif
 
-#include <ctype.h> /* temporary hack */
+#include <ctype.h>
 
 /* defining this one enables tracing */
 #undef DEBUG
@@ -59,61 +62,69 @@ char copyright[] = " SRE 0.6 Copyright (c) 1997-2000 by Secret Labs AB ";
 
 #ifdef	DEBUG
 #define TRACE(v) printf v
-#define PTR(ptr) ((SRE_CHAR*) (ptr) - (SRE_CHAR*) state->beginning)
 #else
 #define TRACE(v)
 #endif
+#define PTR(ptr) ((SRE_CHAR*) (ptr) - (SRE_CHAR*) state->beginning)
 
 #define SRE_CODE unsigned short /* unsigned short or larger */
 
-typedef struct {
-	/* string pointers */
-	void* ptr; /* current position (also end of current slice) */
-	void* beginning; /* start of original string */
-	void* start; /* start of current slice */
-	void* end; /* end of original string */
-	/* character size */
-	int charsize;
-	/* registers */
-	int marks;
-	void* mark[64]; /* FIXME: <fl> should be dynamically allocated! */
-	/* FIXME */
-	/* backtracking stack */
-	void** stack;
-	int stacksize;
-	int stackbase;
-} SRE_STATE;
-
-#if 1 /* FIXME: <fl> fix this one! */
-#define SRE_TO_LOWER Py_UNICODE_TOLOWER
-#define SRE_IS_DIGIT Py_UNICODE_ISDIGIT
-#define SRE_IS_SPACE Py_UNICODE_ISSPACE
-#define SRE_IS_ALNUM(ch) ((ch) < 256 ? isalnum((ch)) : 0)
-#else
-#define SRE_TO_LOWER(ch) ((ch) < 256 ? tolower((ch)) : ch)
-#define SRE_IS_DIGIT(ch) ((ch) < 256 ? isdigit((ch)) : 0)
-#define SRE_IS_SPACE(ch) ((ch) < 256 ? isspace((ch)) : 0)
+/* -------------------------------------------------------------------- */
+/* search engine state */
+
+/* unicode character predicates */
+#define SRE_TO_LOWER(ch) Py_UNICODE_TOLOWER((Py_UNICODE)(ch))
+#define SRE_IS_DIGIT(ch) Py_UNICODE_ISDIGIT((Py_UNICODE)(ch))
+#define SRE_IS_SPACE(ch) Py_UNICODE_ISSPACE((Py_UNICODE)(ch))
+#define SRE_IS_LINEBREAK(ch) ((ch) == '\n')
+/* #define SRE_IS_LINEBREAK(ch) Py_UNICODE_ISLINEBREAK((Py_UNICODE)(ch)) */
 #define SRE_IS_ALNUM(ch) ((ch) < 256 ? isalnum((ch)) : 0)
-#endif
-
 #define SRE_IS_WORD(ch) (SRE_IS_ALNUM((ch)) || (ch) == '_')
 
+/* locale-specific character predicates */
+#define SRE_LOC_TO_LOWER(ch) ((ch) < 256 ? tolower((ch)) : ch)
+#define SRE_LOC_IS_DIGIT(ch) ((ch) < 256 ? isdigit((ch)) : 0)
+#define SRE_LOC_IS_SPACE(ch) ((ch) < 256 ? isspace((ch)) : 0)
+#define SRE_LOC_IS_LINEBREAK(ch) ((ch) == '\n')
+#define SRE_LOC_IS_ALNUM(ch) ((ch) < 256 ? isalnum((ch)) : 0)
+#define SRE_LOC_IS_WORD(ch) (SRE_LOC_IS_ALNUM((ch)) || (ch) == '_')
+
 LOCAL(int)
 sre_category(SRE_CODE category, unsigned int ch)
 {
 	switch (category) {
-	case 'd':
+	case SRE_CATEGORY_DIGIT:
 		return SRE_IS_DIGIT(ch);
-	case 'D':
+	case SRE_CATEGORY_NOT_DIGIT:
 		return !SRE_IS_DIGIT(ch);
-	case 's':
+	case SRE_CATEGORY_SPACE:
 		return SRE_IS_SPACE(ch);
-	case 'S':
+	case SRE_CATEGORY_NOT_SPACE:
 		return !SRE_IS_SPACE(ch);
-	case 'w':
+	case SRE_CATEGORY_WORD:
 		return SRE_IS_WORD(ch);
-	case 'W':
+	case SRE_CATEGORY_NOT_WORD:
 		return !SRE_IS_WORD(ch);
+	case SRE_CATEGORY_LINEBREAK:
+		return SRE_IS_LINEBREAK(ch);
+	case SRE_CATEGORY_NOT_LINEBREAK:
+		return !SRE_IS_LINEBREAK(ch);
+	case SRE_CATEGORY_LOC_DIGIT:
+		return SRE_LOC_IS_DIGIT(ch);
+	case SRE_CATEGORY_LOC_NOT_DIGIT:
+		return !SRE_LOC_IS_DIGIT(ch);
+	case SRE_CATEGORY_LOC_SPACE:
+		return SRE_LOC_IS_SPACE(ch);
+	case SRE_CATEGORY_LOC_NOT_SPACE:
+		return !SRE_LOC_IS_SPACE(ch);
+	case SRE_CATEGORY_LOC_WORD:
+		return SRE_LOC_IS_WORD(ch);
+	case SRE_CATEGORY_LOC_NOT_WORD:
+		return !SRE_LOC_IS_WORD(ch);
+	case SRE_CATEGORY_LOC_LINEBREAK:
+		return SRE_LOC_IS_LINEBREAK(ch);
+	case SRE_CATEGORY_LOC_NOT_LINEBREAK:
+		return !SRE_LOC_IS_LINEBREAK(ch);
 	}
 	return 0;
 }
@@ -174,7 +185,7 @@ _stack_extend(SRE_STATE* state, int lo, int hi)
 	return 0;
 }
 
-/* set things up for the 8-bit version */
+/* generate 8-bit version */
 
 #define SRE_CHAR unsigned char
 #define SRE_AT sre_at
@@ -192,7 +203,7 @@ _stack_extend(SRE_STATE* state, int lo, int hi)
 #undef SRE_AT
 #undef SRE_CHAR
 
-/* set things up for the 16-bit unicode version */
+/* generate 16-bit unicode version */
 
 #define SRE_CHAR Py_UNICODE
 #define SRE_AT sre_uat
@@ -211,20 +222,22 @@ _stack_extend(SRE_STATE* state, int lo, int hi)
 LOCAL(int)
 SRE_AT(SRE_STATE* state, SRE_CHAR* ptr, SRE_CODE at)
 {
-	/* check if pointer is at given position.  return 1 if so, 0
-	   otherwise */
+	/* check if pointer is at given position */
 
 	int this, that;
 
 	switch (at) {
-	case 'a':
-		/* beginning */
+	case SRE_AT_BEGINNING:
 		return ((void*) ptr == state->beginning);
-	case 'z':
-		/* end */
+	case SRE_AT_BEGINNING_LINE:
+		return ((void*) ptr == state->beginning ||
+                SRE_IS_LINEBREAK((int) ptr[-1]));
+	case SRE_AT_END:
 		return ((void*) ptr == state->end);
-	case 'b':
-		/* word boundary */
+	case SRE_AT_END_LINE:
+		return ((void*) ptr == state->end ||
+                SRE_IS_LINEBREAK((int) ptr[0]));
+	case SRE_AT_BOUNDARY:
 		if (state->beginning == state->end)
 			return 0;
 		that = ((void*) ptr > state->beginning) ?
@@ -232,8 +245,7 @@ SRE_AT(SRE_STATE* state, SRE_CHAR* ptr, SRE_CODE at)
 		this = ((void*) ptr < state->end) ?
 			SRE_IS_WORD((int) ptr[0]) : 0;
 		return this != that;
-	case 'B':
-		/* word non-boundary */
+	case SRE_AT_NON_BOUNDARY:
 		if (state->beginning == state->end)
 			return 0;
 		that = ((void*) ptr > state->beginning) ?
@@ -249,8 +261,7 @@ SRE_AT(SRE_STATE* state, SRE_CHAR* ptr, SRE_CODE at)
 LOCAL(int)
 SRE_MEMBER(SRE_CODE* set, SRE_CHAR ch)
 {
-	/* check if character is a member of the given set.	 return 1 if
-	   so, 0 otherwise */
+	/* check if character is a member of the given set */
 
 	int ok = 1;
 
@@ -301,29 +312,42 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 	int stackbase;
 	int i, count;
 
-	for (;;) {
+    /* FIXME: this is one ugly hack */
+    void* *mark = NULL;
+    void* mark_data[64];
 
-		TRACE(("[%p]\n", pattern));
+	for (;;) {
 
 		switch (*pattern++) {
 
 		case SRE_OP_FAILURE:
 			/* immediate failure */
 			TRACE(("%8d: failure\n", PTR(ptr)));
-			return 0;
+			goto failure;
 
 		case SRE_OP_SUCCESS:
 			/* end of pattern */
 			TRACE(("%8d: success\n", PTR(ptr)));
 			state->ptr = ptr;
-			return 1;
+			goto success;
 
 		case SRE_OP_AT:
 			/* match at given position */
+			/* args: <at> */
 			TRACE(("%8d: match at \\%c\n", PTR(ptr), *pattern));
 			if (!SRE_AT(state, ptr, *pattern))
-				return 0;
+				goto failure;
+			pattern++;
+			break;
+
+		case SRE_OP_CATEGORY:
+			/* match at given category */
+			/* args: <category> */
+			TRACE(("%8d: category match at \\%c\n", PTR(ptr), *pattern));
+			if (ptr >= end || !sre_category(pattern[0], ptr[0]))
+				goto failure;
 			pattern++;
+            ptr++;
 			break;
 
 		case SRE_OP_LITERAL:
@@ -331,7 +355,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			/* args: <code> */
 			TRACE(("%8d: literal %c\n", PTR(ptr), (SRE_CHAR) *pattern));
 			if (ptr >= end || *ptr != (SRE_CHAR) *pattern)
-				return 0;
+				goto failure;
 			pattern++;
 			ptr++;
 			break;
@@ -341,7 +365,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			/* args: <code> */
 			TRACE(("%8d: literal not %c\n", PTR(ptr), (SRE_CHAR) *pattern));
 			if (ptr >= end || *ptr == (SRE_CHAR) *pattern)
-				return 0;
+				goto failure;
 			pattern++;
 			ptr++;
 			break;
@@ -350,7 +374,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			/* match anything */
 			TRACE(("%8d: any\n", PTR(ptr)));
 			if (ptr >= end)
-				return 0;
+				goto failure;
 			ptr++;
 			break;
 
@@ -359,23 +383,47 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			/* args: <skip> <set> */
 			TRACE(("%8d: set %c\n", PTR(ptr), *ptr));
 			if (ptr >= end || !SRE_MEMBER(pattern + 1, *ptr))
-				return 0;
+				goto failure;
 			pattern += pattern[0];
 			ptr++;
 			break;
 
 		case SRE_OP_GROUP:
 			/* match backreference */
+			TRACE(("%8d: group %d\n", PTR(ptr), pattern[0]));
 			i = pattern[0];
 			{
-				/* FIXME: optimize size! */
+				/* FIXME: optimize! */
 				SRE_CHAR* p = (SRE_CHAR*) state->mark[i+i];
 				SRE_CHAR* e = (SRE_CHAR*) state->mark[i+i+1];
+                TRACE(("%8d: group %p %p\n", PTR(ptr), p, e));
 				if (!p || !e || e < p)
-					return 0;
+					goto failure;
 				while (p < e) {
+                    TRACE(("%8d: group test %c %c\n", PTR(ptr), *ptr, *p));
 					if (ptr >= end || *ptr != *p)
-						return 0;
+						goto failure;
+					p++; ptr++;
+				}
+			}
+			pattern++;
+			break;
+
+		case SRE_OP_GROUP_IGNORE:
+			/* match backreference */
+			TRACE(("%8d: group ignore %d\n", PTR(ptr), pattern[0]));
+			i = pattern[0];
+			{
+				/* FIXME: optimize! */
+				SRE_CHAR* p = (SRE_CHAR*) state->mark[i+i];
+				SRE_CHAR* e = (SRE_CHAR*) state->mark[i+i+1];
+                TRACE(("%8d: group %p %p\n", PTR(ptr), p, e));
+				if (!p || !e || e < p)
+					goto failure;
+				while (p < e) {
+                    TRACE(("%8d: group test %c %c\n", PTR(ptr), *ptr, *p));
+					if (ptr >= end || SRE_TO_LOWER(*ptr) != SRE_TO_LOWER(*p))
+						goto failure;
 					p++; ptr++;
 				}
 			}
@@ -385,7 +433,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 		case SRE_OP_LITERAL_IGNORE:
 			TRACE(("%8d: literal lower(%c)\n", PTR(ptr), (SRE_CHAR) *pattern));
 			if (ptr >= end || SRE_TO_LOWER(*ptr) != (SRE_CHAR) *pattern)
-				return 0;
+				goto failure;
 			pattern++;
 			ptr++;
 			break;
@@ -394,7 +442,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			TRACE(("%8d: literal not lower(%c)\n", PTR(ptr),
 				   (SRE_CHAR) *pattern));
 			if (ptr >= end || SRE_TO_LOWER(*ptr) == (SRE_CHAR) *pattern)
-				return 0;
+				goto failure;
 			pattern++;
 			ptr++;
 			break;
@@ -403,7 +451,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			TRACE(("%8d: set lower(%c)\n", PTR(ptr), *ptr));
 			if (ptr >= end
 				|| !SRE_MEMBER(pattern+1, (SRE_CHAR) SRE_TO_LOWER(*ptr)))
-				return 0;
+				goto failure;
 			pattern += pattern[0];
 			ptr++;
 			break;
@@ -412,7 +460,11 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			/* set mark */
 			/* args: <mark> */
 			TRACE(("%8d: set mark(%d)\n", PTR(ptr), pattern[0]));
-			state->mark[pattern[0]] = ptr;
+            if (!mark) {
+                mark = mark_data;
+                memcpy(mark, state->mark, sizeof(state->mark));
+            }
+            state->mark[pattern[0]] = ptr;
 			pattern++;
 			break;
 
@@ -429,21 +481,18 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			TRACE(("%8d: match subpattern\n", PTR(ptr)));
 			state->ptr = ptr;
 			if (!SRE_MATCH(state, pattern + 1))
-				return 0;
+				goto failure;
 			pattern += pattern[0];
 			ptr = state->ptr;
 			break;
 
 		case SRE_OP_MAX_REPEAT_ONE:
-
-			/* match repeated sequence (maximizing regexp).	 this
-			   variant only works if the repeated item is exactly one
-			   character wide, and we're not already collecting
-			   backtracking points.	 for other cases, use the
+			/* match repeated sequence (maximizing regexp) */
+            /* this variant only works if the repeated item is exactly
+			   one character wide, and we're not already collecting
+			   backtracking points.  for other cases, use the
 			   MAX_REPEAT operator instead */
-
 			/* args: <skip> <min> <max> <step> */
-
 			TRACE(("%8d: max repeat one {%d,%d}\n", PTR(ptr),
 				   pattern[1], pattern[2]));
 
@@ -454,7 +503,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				   string, and backtrack from there */
 				/* FIXME: must look for line endings */
 				if (ptr + pattern[1] > end)
-					return 0; /* cannot match */
+					goto failure; /* cannot match */
 				count = pattern[2];
 				if (count > end - ptr)
 					count = end - ptr;
@@ -515,7 +564,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				while (count < (int) pattern[2]) {
 					i = SRE_MATCH(state, pattern + 3);
 					if (i < 0)
-						return i;
+						goto failure;
 					if (i == 0)
 						break;
 					count++;
@@ -529,23 +578,20 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			   string.	check if the rest of the pattern matches, and
 			   backtrack if not. */
 
-			/* FIXME: <fl> this is a mess.	fix it! */
-
 			TRACE(("%8d: repeat %d found\n", PTR(ptr), count));
 
 			if (count < (int) pattern[1])
-				return 0;
+				goto failure;
 
 			if (pattern[pattern[0]] == SRE_OP_SUCCESS) {
 				/* tail is empty.  we're finished */
 				TRACE(("%8d: tail is empty\n", PTR(ptr)));
 				state->ptr = ptr;
-				return 1;
+				goto success;
 
 			} else if (pattern[pattern[0]] == SRE_OP_LITERAL) {
-				/* tail starts with a literal.	we can speed things up
-				   by skipping positions where the rest of the pattern
-				   cannot possibly match */
+				/* tail starts with a literal. skip positions where
+				   the rest of the pattern cannot possibly match */
 				SRE_CHAR chr = (SRE_CHAR) pattern[pattern[0]+1];
 				TRACE(("%8d: tail is literal %d\n", PTR(ptr), chr));
 				for (;;) {
@@ -562,7 +608,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 					i = SRE_MATCH(state, pattern + pattern[0]);
 					if (i > 0) {
 						TRACE(("%8d: repeat %d picked\n", PTR(ptr), count));
-						return 1;
+						goto success;
 					}
 					TRACE(("%8d: BACKTRACK\n", PTR(ptr)));
 					ptr--;
@@ -570,23 +616,21 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				}
 
 			} else {
+				/* general case */
 				TRACE(("%8d: tail is pattern\n", PTR(ptr)));
 				while (count >= (int) pattern[1]) {
 					state->ptr = ptr;
 					i = SRE_MATCH(state, pattern + pattern[0]);
 					if (i > 0) {
 						TRACE(("%8d: repeat %d picked\n", PTR(ptr), count));
-						return 1;
+						goto success;
 					}
 					TRACE(("%8d: BACKTRACK\n", PTR(ptr)));
 					ptr--;
 					count--;
 				}
 			}
-			return 0; /* failure! */
-
-/* ----------------------------------------------------------------------- */
-/* FIXME: the following section is just plain broken */
+			goto failure;
 
 		case SRE_OP_MAX_REPEAT:
 			/* match repeated sequence (maximizing regexp).	 repeated
@@ -611,7 +655,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 					i = _stack_extend(state, stackbase + count + 1,
 									  stackbase + pattern[2]);
 					if (i < 0)
-						return i;
+						goto failure;
 				}
 				state->stack[stackbase + count] = ptr;
 				/* check if we can match another item */
@@ -642,7 +686,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			   ptr points to the tail. */
 
 			if (count < (int) pattern[1])
-				return 0;
+				goto failure;
 
 			/* make sure that rest of the expression matches.  if it
 			   doesn't, backtrack */
@@ -659,7 +703,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			state->stackbase = stackbase;
 			if (i > 0) {
 				TRACE(("%8d: repeat %d picked\n", PTR(ptr), count));
-				return 1;
+				goto success;
 			}
 
 			/* backtrack! */
@@ -673,10 +717,10 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				state->stackbase = stackbase;
 				if (i > 0) {
 					TRACE(("%8d: repeat %d picked\n", PTR(ptr), count));
-					return 1;
+					goto success;
 				}
 			}
-			return 0; /* failure! */
+			goto failure;
 
 		case SRE_OP_MAX_UNTIL:
 			/* match repeated sequence (maximizing regexp).	 repeated
@@ -684,13 +728,11 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 
 			TRACE(("%8d: max until\n", PTR(ptr)));
 			state->ptr = ptr;
-			return 2; /* always succeeds, for now... */
-
-/* end of totally broken section */
-/* ----------------------------------------------------------------------- */
+			goto success; /* always succeeds, for now... */
 
 		case SRE_OP_MIN_REPEAT:
 			/* match repeated sequence (minimizing regexp) */
+            /* FIXME: HERE BE BUGS! */
 			TRACE(("%8d: min repeat %d %d\n", PTR(ptr),
 				   pattern[1], pattern[2]));
 			count = 0;
@@ -699,7 +741,7 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 			while (count < (int) pattern[1]) {
 				i = SRE_MATCH(state, pattern + 3);
 				if (i <= 0)
-					return i;
+					goto failure;
 				count++;
 			}
 			/* move forward until the tail matches. */
@@ -708,22 +750,22 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				i = SRE_MATCH(state, pattern + pattern[0]);
 				if (i > 0) {
 					TRACE(("%8d: repeat %d picked\n", PTR(ptr), count));
-					return 1;
+					goto success;
 				}
 				TRACE(("%8d: BACKTRACK\n", PTR(ptr)));
 				state->ptr = ptr; /* backtrack */
 				i = SRE_MATCH(state, pattern + 3);
 				if (i <= 0)
-					return i;
+					goto failure;
 				count++;
 			}
-			return 0; /* failure! */
+			goto failure;
 
 		case SRE_OP_MIN_UNTIL:
 			/* end of repeat group */
 			TRACE(("%8d: min until\n", PTR(ptr)));
 			state->ptr = ptr;
-			return 2; /* always succeeds, for now... */
+			goto success; /* always succeeds, for now... */
 
 		case SRE_OP_BRANCH:
 			/* match one of several subpatterns */
@@ -737,13 +779,13 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 					i = SRE_MATCH(state, pattern + 1);
 					if (i > 0) {
 						TRACE(("%8d: branch succeeded\n", PTR(ptr)));
-						return 1;
+						goto success;
 					}
 				}
 				pattern += *pattern;
 			}
 			TRACE(("%8d: branch failed\n", PTR(ptr)));
-			return 0; /* failure! */
+			goto failure;
 
 		case SRE_OP_REPEAT:
 			/* TEMPLATE: match repeated sequence (no backtracking) */
@@ -758,16 +800,24 @@ SRE_MATCH(SRE_STATE* state, SRE_CODE* pattern)
 				count++;
 			}
 			if (count <= (int) pattern[1])
-				return 0;
+				goto failure;
 			TRACE(("%8d: repeat %d matches\n", PTR(ptr), count));
 			pattern += pattern[0];
 			ptr = state->ptr;
 			break;
 
-		default:
+        default:
 			return SRE_ERROR_ILLEGAL;
 		}
 	}
+
+  failure:
+    if (mark)
+        memcpy(state->mark, mark, sizeof(state->mark));
+    return 0;
+
+  success:
+    return 1;
 }
 
 LOCAL(int)
@@ -832,6 +882,7 @@ SRE_SEARCH(SRE_STATE* state, SRE_CODE* pattern)
 
 staticforward PyTypeObject Pattern_Type;
 staticforward PyTypeObject Match_Type;
+staticforward PyTypeObject Cursor_Type;
 
 static PyObject *
 _compile(PyObject* self_, PyObject* args)
@@ -841,20 +892,25 @@ _compile(PyObject* self_, PyObject* args)
 	PatternObject* self;
 
 	PyObject* pattern;
+    int flags = 0;
 	PyObject* code;
 	int groups = 0;
 	PyObject* groupindex = NULL;
-	if (!PyArg_ParseTuple(args, "OO!|iO", &pattern,
-                          &PyString_Type, &code, &groups, &groupindex))
+	if (!PyArg_ParseTuple(args, "OiO!|iO", &pattern, &flags,
+                          &PyString_Type, &code,
+                          &groups, &groupindex))
 		return NULL;
 
-	self = PyObject_New(PatternObject, &Pattern_Type);
+	self = PyObject_NEW(PatternObject, &Pattern_Type);
 	if (self == NULL)
+
 		return NULL;
 
 	Py_INCREF(pattern);
 	self->pattern = pattern;
 
+    self->flags = flags;
+
 	Py_INCREF(code);
 	self->code = code;
 
@@ -872,6 +928,69 @@ _getcodesize(PyObject* self_, PyObject* args)
 	return Py_BuildValue("i", sizeof(SRE_CODE));
 }
 
+LOCAL(PyObject*)
+_setup(SRE_STATE* state, PyObject* args)
+{
+	/* prepare state object */
+
+	PyBufferProcs *buffer;
+	int i, count;
+	void* ptr;
+
+	PyObject* string;
+	int start = 0;
+	int end = INT_MAX;
+	if (!PyArg_ParseTuple(args, "O|ii", &string, &start, &end))
+		return NULL;
+
+	/* get pointer to string buffer */
+	buffer = string->ob_type->tp_as_buffer;
+	if (!buffer || !buffer->bf_getreadbuffer || !buffer->bf_getsegcount ||
+		buffer->bf_getsegcount(string, NULL) != 1) {
+		PyErr_SetString(PyExc_TypeError, "expected read-only buffer");
+		return NULL;
+	}
+
+	/* determine buffer size */
+	count = buffer->bf_getreadbuffer(string, 0, &ptr);
+	if (count < 0) {
+		/* sanity check */
+		PyErr_SetString(PyExc_TypeError, "buffer has negative size");
+		return NULL;
+	}
+
+	/* determine character size */
+	state->charsize = (PyUnicode_Check(string) ? sizeof(Py_UNICODE) : 1);
+
+	count /= state->charsize;
+
+	/* adjust boundaries */
+	if (start < 0)
+		start = 0;
+	else if (start > count)
+		start = count;
+
+	if (end < 0)
+		end = 0;
+	else if (end > count)
+		end = count;
+
+	state->beginning = ptr;
+
+	state->start = (void*) ((char*) ptr + start * state->charsize);
+	state->end = (void*) ((char*) ptr + end * state->charsize);
+
+	/* FIXME: dynamic! */
+	for (i = 0; i < 64; i++)
+		state->mark[i] = NULL;
+
+	state->stack = NULL;
+	state->stackbase = 0;
+	state->stacksize = 0;
+
+	return string;
+}
+
 static PyObject*
 _pattern_new_match(PatternObject* pattern, SRE_STATE* state,
                    PyObject* string, int status)
@@ -886,7 +1005,7 @@ _pattern_new_match(PatternObject* pattern, SRE_STATE* state,
 	if (status > 0) {
 
 		/* create match object (with room for extra group marks) */
-		match = PyObject_NewVar(MatchObject, &Match_Type, 2*pattern->groups);
+		match = PyObject_NEW_VAR(MatchObject, &Match_Type, 2*pattern->groups);
 		if (match == NULL)
 			return NULL;
 
@@ -930,70 +1049,32 @@ _pattern_new_match(PatternObject* pattern, SRE_STATE* state,
 	return Py_None;
 }
 
-/* -------------------------------------------------------------------- */
-/* pattern methods */
-
-LOCAL(PyObject*)
-_setup(SRE_STATE* state, PyObject* args)
+static PyObject*
+_pattern_cursor(PyObject* pattern, PyObject* args)
 {
-	/* prepare state object */
-
-	PyBufferProcs *buffer;
-	int i, count;
-	void* ptr;
-
-	PyObject* string;
-	int start = 0;
-	int end = INT_MAX;
-	if (!PyArg_ParseTuple(args, "O|ii", &string, &start, &end))
-		return NULL;
+	/* create search state object */
 
-	/* get pointer to string buffer */
-	buffer = string->ob_type->tp_as_buffer;
-	if (!buffer || !buffer->bf_getreadbuffer || !buffer->bf_getsegcount ||
-		buffer->bf_getsegcount(string, NULL) != 1) {
-		PyErr_SetString(PyExc_TypeError, "expected read-only buffer");
-		return NULL;
-	}
-
-	/* determine buffer size */
-	count = buffer->bf_getreadbuffer(string, 0, &ptr);
-	if (count < 0) {
-		/* sanity check */
-		PyErr_SetString(PyExc_TypeError, "buffer has negative size");
-		return NULL;
-	}
-
-	/* determine character size */
-	state->charsize = (PyUnicode_Check(string) ? sizeof(Py_UNICODE) : 1);
+	CursorObject* self;
+    PyObject* string;
 
-	count /= state->charsize;
+    /* create match object (with room for extra group marks) */
+    self = PyObject_NEW(CursorObject, &Cursor_Type);
+    if (self == NULL)
+        return NULL;
 
-	/* adjust boundaries */
-	if (start < 0)
-		start = 0;
-	else if (start > count)
-		start = count;
-
-	if (end < 0)
-		end = 0;
-	else if (end > count)
-		end = count;
-
-	state->beginning = ptr;
-
-	state->start = (void*) ((char*) ptr + start * state->charsize);
-	state->end = (void*) ((char*) ptr + end * state->charsize);
+    string = _setup(&self->state, args);
+    if (!string) {
+        /* FIXME: dealloc cursor object */
+        return NULL;
+    }
 
-	/* FIXME: dynamic! */
-	for (i = 0; i < 64; i++)
-		state->mark[i] = NULL;
+    Py_INCREF(pattern);
+    self->pattern = pattern;
 
-	state->stack = NULL;
-	state->stackbase = 0;
-	state->stacksize = 0;
+    Py_INCREF(string);
+    self->string = string;
 
-	return string;
+	return (PyObject*) self;
 }
 
 static void
@@ -1002,7 +1083,7 @@ _pattern_dealloc(PatternObject* self)
 	Py_XDECREF(self->code);
 	Py_XDECREF(self->pattern);
 	Py_XDECREF(self->groupindex);
-	PyObject_Del(self);
+	PyMem_DEL(self);
 }
 
 static PyObject*
@@ -1052,11 +1133,71 @@ _pattern_search(PatternObject* self, PyObject* args)
 }
 
 static PyObject*
-_pattern_findall(PatternObject* self, PyObject* args)
+call(char* function, PyObject* args)
+{
+    PyObject* name;
+    PyObject* module;
+    PyObject* func;
+    PyObject* result;
+
+    name = PyString_FromString("sre");
+    if (!name)
+        return NULL;
+    module = PyImport_Import(name);
+    Py_DECREF(name);
+    if (!module)
+        return NULL;
+    func = PyObject_GetAttrString(module, function);
+    Py_DECREF(module);
+    if (!func)
+        return NULL;
+    result = PyObject_CallObject(func, args);
+    Py_DECREF(func);
+    Py_DECREF(args);
+    return result;
+}
+
+static PyObject*
+_pattern_sub(PatternObject* self, PyObject* args)
 {
-	/* FIXME: not sure about the semantics here.  this is good enough
-	   for SXP, though... */
+	PyObject* template;
+	PyObject* string;
+    PyObject* count;
+	if (!PyArg_ParseTuple(args, "OOO", &template, &string, &count))
+		return NULL;
 
+    /* delegate to Python code */
+    return call("_sub", Py_BuildValue("OOOO", self, template, string, count));
+}
+
+static PyObject*
+_pattern_subn(PatternObject* self, PyObject* args)
+{
+	PyObject* template;
+	PyObject* string;
+    PyObject* count;
+	if (!PyArg_ParseTuple(args, "OOO", &template, &string, &count))
+		return NULL;
+
+    /* delegate to Python code */
+    return call("_subn", Py_BuildValue("OOOO", self, template, string, count));
+}
+
+static PyObject*
+_pattern_split(PatternObject* self, PyObject* args)
+{
+	PyObject* string;
+    PyObject* maxsplit;
+	if (!PyArg_ParseTuple(args, "OO", &string, &maxsplit))
+		return NULL;
+
+    /* delegate to Python code */
+    return call("_split", Py_BuildValue("OOO", self, string, maxsplit));
+}
+
+static PyObject*
+_pattern_findall(PatternObject* self, PyObject* args)
+{
 	SRE_STATE state;
 	PyObject* string;
 	PyObject* list;
@@ -1077,7 +1218,7 @@ _pattern_findall(PatternObject* self, PyObject* args)
 		if (state.charsize == 1) {
 			status = sre_match(&state, PatternObject_GetCode(self));
 		} else {
-			status = sre_umatch(&state,	PatternObject_GetCode(self));
+			status = sre_umatch(&state, PatternObject_GetCode(self));
 		}
 
 		if (status >= 0) {
@@ -1085,6 +1226,10 @@ _pattern_findall(PatternObject* self, PyObject* args)
 			if (status == 0)
 				state.ptr = (void*) ((char*) state.start + 1);
 
+            /* FIXME: if one group is defined, slice that group
+               instead.  if multiple groups are defined, add tuple
+               containing all slices */
+
 			item = PySequence_GetSlice(
 				string,
 				((char*) state.start - (char*) state.beginning),
@@ -1121,7 +1266,12 @@ error:
 static PyMethodDef _pattern_methods[] = {
 	{"match", (PyCFunction) _pattern_match, 1},
 	{"search", (PyCFunction) _pattern_search, 1},
+	{"sub", (PyCFunction) _pattern_sub, 1},
+	{"subn", (PyCFunction) _pattern_subn, 1},
+	{"split", (PyCFunction) _pattern_split, 1},
 	{"findall", (PyCFunction) _pattern_findall, 1},
+    /* experimental */
+	{"cursor", (PyCFunction) _pattern_cursor, 1},
 	{NULL, NULL}
 };
 
@@ -1142,7 +1292,15 @@ _pattern_getattr(PatternObject* self, char* name)
         Py_INCREF(self->pattern);
 		return self->pattern;
     }
-    
+
+    if (!strcmp(name, "flags"))
+		return Py_BuildValue("i", self->flags);
+
+	if (!strcmp(name, "groupindex") && self->groupindex) {
+        Py_INCREF(self->groupindex);
+		return self->groupindex;
+    }
+
 	PyErr_SetString(PyExc_AttributeError, name);
 	return NULL;
 }
@@ -1163,7 +1321,7 @@ _match_dealloc(MatchObject* self)
 {
 	Py_XDECREF(self->string);
 	Py_DECREF(self->pattern);
-	PyObject_Del(self);
+	PyMem_DEL(self);
 }
 
 static PyObject*
@@ -1244,6 +1402,8 @@ _match_groups(MatchObject* self, PyObject* args)
 	PyObject* result;
 	int index;
 
+    /* FIXME: <fl> handle default value! */
+
 	result = PyTuple_New(self->groups-1);
 	if (!result)
 		return NULL;
@@ -1269,6 +1429,8 @@ _match_groupdict(MatchObject* self, PyObject* args)
 	PyObject* keys;
 	int index;
 
+    /* FIXME: <fl> handle default value! */
+
 	result = PyDict_New();
 	if (!result)
 		return NULL;
@@ -1367,7 +1529,8 @@ _match_span(MatchObject* self, PyObject* args)
 
 	if (self->mark[index*2] < 0) {
 		Py_INCREF(Py_None);
-		return Py_None;
+		Py_INCREF(Py_None);
+        return Py_BuildValue("OO", Py_None, Py_None);
 	}
 
 	return Py_BuildValue("ii", self->mark[index*2], self->mark[index*2+1]);
@@ -1394,24 +1557,20 @@ _match_getattr(MatchObject* self, char* name)
 
 	PyErr_Clear();
 
-	/* attributes! */
+	/* attributes */
 	if (!strcmp(name, "string")) {
         Py_INCREF(self->string);
 		return self->string;
     }
-	if (!strcmp(name, "regs"))
-		/* FIXME: should return the whole list! */
-		return Py_BuildValue("((i,i))", self->mark[0], self->mark[1]);
+
 	if (!strcmp(name, "re")) {
         Py_INCREF(self->pattern);
 		return (PyObject*) self->pattern;
     }
-	if (!strcmp(name, "groupindex") && self->pattern->groupindex) {
-        Py_INCREF(self->pattern->groupindex);
-		return self->pattern->groupindex;
-    }
+
 	if (!strcmp(name, "pos"))
 		return Py_BuildValue("i", 0); /* FIXME */
+
 	if (!strcmp(name, "endpos"))
 		return Py_BuildValue("i", 0); /* FIXME */
 
@@ -1432,6 +1591,106 @@ statichere PyTypeObject Match_Type = {
 	(getattrfunc)_match_getattr, /*tp_getattr*/
 };
 
+/* -------------------------------------------------------------------- */
+/* cursor methods (experimental) */
+
+static void
+_cursor_dealloc(CursorObject* self)
+{
+	_stack_free(&self->state);
+    Py_DECREF(self->string);
+    Py_DECREF(self->pattern);
+	PyMem_DEL(self);
+}
+
+static PyObject*
+_cursor_match(CursorObject* self, PyObject* args)
+{
+    SRE_STATE* state = &self->state;
+    PyObject* match;
+    int status;
+
+    state->ptr = state->start;
+
+    if (state->charsize == 1) {
+        status = sre_match(state, PatternObject_GetCode(self->pattern));
+    } else {
+        status = sre_umatch(state, PatternObject_GetCode(self->pattern));
+    }
+
+    match = _pattern_new_match((PatternObject*) self->pattern,
+                               state, self->string, status);
+
+    if (status >= 0)
+        state->start = state->ptr;
+    else
+        state->start = (char*) state->ptr + state->charsize;
+
+    return match;
+}
+
+
+static PyObject*
+_cursor_search(CursorObject* self, PyObject* args)
+{
+    SRE_STATE* state = &self->state;
+    PyObject* match;
+    int status;
+
+    state->ptr = state->start;
+
+    if (state->charsize == 1) {
+        status = sre_search(state, PatternObject_GetCode(self->pattern));
+    } else {
+        status = sre_usearch(state, PatternObject_GetCode(self->pattern));
+    }
+
+    match = _pattern_new_match((PatternObject*) self->pattern,
+                               state, self->string, status);
+
+    if (status >= 0)
+        state->start = state->ptr;
+
+    return match;
+}
+
+static PyMethodDef _cursor_methods[] = {
+	{"match", (PyCFunction) _cursor_match, 0},
+	{"search", (PyCFunction) _cursor_search, 0},
+	{NULL, NULL}
+};
+
+static PyObject*  
+_cursor_getattr(CursorObject* self, char* name)
+{
+	PyObject* res;
+
+	res = Py_FindMethod(_cursor_methods, (PyObject*) self, name);
+	if (res)
+		return res;
+
+	PyErr_Clear();
+
+	/* attributes */
+	if (!strcmp(name, "pattern")) {
+        Py_INCREF(self->pattern);
+		return self->pattern;
+    }
+
+	PyErr_SetString(PyExc_AttributeError, name);
+	return NULL;
+}
+
+statichere PyTypeObject Cursor_Type = {
+	PyObject_HEAD_INIT(NULL)
+	0, "Cursor",
+	sizeof(CursorObject), /* size of basic object */
+	0,
+	(destructor)_cursor_dealloc, /*tp_dealloc*/
+	0, /*tp_print*/
+	(getattrfunc)_cursor_getattr, /*tp_getattr*/
+};
+
 static PyMethodDef _functions[] = {
 	{"compile", _compile, 1},
 	{"getcodesize", _getcodesize, 1},
@@ -1445,7 +1704,8 @@ __declspec(dllexport)
 init_sre()
 {
 	/* Patch object types */
-	Pattern_Type.ob_type = Match_Type.ob_type = &PyType_Type;
+	Pattern_Type.ob_type = Match_Type.ob_type =
+        Cursor_Type.ob_type = &PyType_Type;
 
 	Py_InitModule("_sre", _functions);
 }
diff --git a/Modules/sre.h b/Modules/sre.h
index 2936b05..3664c9d 100644
--- a/Modules/sre.h
+++ b/Modules/sre.h
@@ -14,17 +14,18 @@
 
 #include "sre_constants.h"
 
-/* Python objects */
-
 typedef struct {
     PyObject_HEAD
     PyObject* code; /* link to the code string object */
-    PyObject* pattern; /* link to the pattern source (or None) */
     int groups;
     PyObject* groupindex;
+    /* compatibility */
+    PyObject* pattern; /* pattern source (or None) */
+    int flags; /* flags used when compiling pattern source */
 } PatternObject;
 
-#define PatternObject_GetCode(o) ((void*) PyString_AS_STRING((o)->code))
+#define PatternObject_GetCode(o)\
+    ((void*) PyString_AS_STRING(((PatternObject*)(o))->code))
 
 typedef struct {
     PyObject_HEAD
@@ -34,5 +35,28 @@ typedef struct {
     int mark[2];
 } MatchObject;
 
-#endif
+typedef struct {
+    /* string pointers */
+    void* ptr; /* current position (also end of current slice) */
+    void* beginning; /* start of original string */
+    void* start; /* start of current slice */
+    void* end; /* end of original string */
+    /* character size */
+    int charsize;
+    /* registers */
+    int marks;
+    void* mark[64]; /* FIXME: <fl> should be dynamically allocated! */
+    /* backtracking stack */
+    void** stack;
+    int stacksize;
+    int stackbase;
+} SRE_STATE;
 
+typedef struct {
+    PyObject_HEAD
+    PyObject* pattern;
+    PyObject* string;
+    SRE_STATE state;
+} CursorObject;
+
+#endif
diff --git a/Modules/sre_constants.h b/Modules/sre_constants.h
index 6b940d3..c6b123e 100644
--- a/Modules/sre_constants.h
+++ b/Modules/sre_constants.h
@@ -1,4 +1,4 @@
-/* generated by sre_constants.py */
+/* generated from sre_constants.py */
 #define SRE_OP_FAILURE 0
 #define SRE_OP_SUCCESS 1
 #define SRE_OP_ANY 2
@@ -25,3 +25,25 @@
 #define SRE_OP_NEGATE 23
 #define SRE_OP_RANGE 24
 #define SRE_OP_REPEAT 25
+#define SRE_AT_BEGINNING 0
+#define SRE_AT_BEGINNING_LINE 1
+#define SRE_AT_BOUNDARY 2
+#define SRE_AT_NON_BOUNDARY 3
+#define SRE_AT_END 4
+#define SRE_AT_END_LINE 5
+#define SRE_CATEGORY_DIGIT 0
+#define SRE_CATEGORY_NOT_DIGIT 1
+#define SRE_CATEGORY_SPACE 2
+#define SRE_CATEGORY_NOT_SPACE 3
+#define SRE_CATEGORY_WORD 4
+#define SRE_CATEGORY_NOT_WORD 5
+#define SRE_CATEGORY_LINEBREAK 6
+#define SRE_CATEGORY_NOT_LINEBREAK 7
+#define SRE_CATEGORY_LOC_DIGIT 8
+#define SRE_CATEGORY_LOC_NOT_DIGIT 9
+#define SRE_CATEGORY_LOC_SPACE 10
+#define SRE_CATEGORY_LOC_NOT_SPACE 11
+#define SRE_CATEGORY_LOC_WORD 12
+#define SRE_CATEGORY_LOC_NOT_WORD 13
+#define SRE_CATEGORY_LOC_LINEBREAK 14
+#define SRE_CATEGORY_LOC_NOT_LINEBREAK 15