sre_compile.py revision fdb73ed4863c31a3d807f9f75ef3d5d6f4d8d4e8 
1#
2# Secret Labs' Regular Expression Engine
3#
4# convert template to internal format
5#
6# Copyright (c) 1997-2001 by Secret Labs AB.  All rights reserved.
7#
8# See the sre.py file for information on usage and redistribution.
9#
10
11"""Internal support module for sre"""
12
13import _sre, sys
14import sre_parse
15from sre_constants import *
16
17assert _sre.MAGIC == MAGIC, "SRE module mismatch"
18
19if _sre.CODESIZE == 2:
20    MAXCODE = 65535
21else:
22    MAXCODE = 0xFFFFFFFFL
23
24def _identityfunction(x):
25    return x
26
27_LITERAL_CODES = set([LITERAL, NOT_LITERAL])
28_REPEATING_CODES = set([REPEAT, MIN_REPEAT, MAX_REPEAT])
29_SUCCESS_CODES = set([SUCCESS, FAILURE])
30_ASSERT_CODES = set([ASSERT, ASSERT_NOT])
31
32def _compile(code, pattern, flags):
33    # internal: compile a (sub)pattern
34    emit = code.append
35    _len = len
36    LITERAL_CODES = _LITERAL_CODES
37    REPEATING_CODES = _REPEATING_CODES
38    SUCCESS_CODES = _SUCCESS_CODES
39    ASSERT_CODES = _ASSERT_CODES
40    for op, av in pattern:
41        if op in LITERAL_CODES:
42            if flags & SRE_FLAG_IGNORECASE:
43                emit(OPCODES[OP_IGNORE[op]])
44                emit(_sre.getlower(av, flags))
45            else:
46                emit(OPCODES[op])
47                emit(av)
48        elif op is IN:
49            if flags & SRE_FLAG_IGNORECASE:
50                emit(OPCODES[OP_IGNORE[op]])
51                def fixup(literal, flags=flags):
52                    return _sre.getlower(literal, flags)
53            else:
54                emit(OPCODES[op])
55                fixup = _identityfunction
56            skip = _len(code); emit(0)
57            _compile_charset(av, flags, code, fixup)
58            code[skip] = _len(code) - skip
59        elif op is ANY:
60            if flags & SRE_FLAG_DOTALL:
61                emit(OPCODES[ANY_ALL])
62            else:
63                emit(OPCODES[ANY])
64        elif op in REPEATING_CODES:
65            if flags & SRE_FLAG_TEMPLATE:
66                raise error, "internal: unsupported template operator"
67                emit(OPCODES[REPEAT])
68                skip = _len(code); emit(0)
69                emit(av[0])
70                emit(av[1])
71                _compile(code, av[2], flags)
72                emit(OPCODES[SUCCESS])
73                code[skip] = _len(code) - skip
74            elif _simple(av) and op is not REPEAT:
75                if op is MAX_REPEAT:
76                    emit(OPCODES[REPEAT_ONE])
77                else:
78                    emit(OPCODES[MIN_REPEAT_ONE])
79                skip = _len(code); emit(0)
80                emit(av[0])
81                emit(av[1])
82                _compile(code, av[2], flags)
83                emit(OPCODES[SUCCESS])
84                code[skip] = _len(code) - skip
85            else:
86                emit(OPCODES[REPEAT])
87                skip = _len(code); emit(0)
88                emit(av[0])
89                emit(av[1])
90                _compile(code, av[2], flags)
91                code[skip] = _len(code) - skip
92                if op is MAX_REPEAT:
93                    emit(OPCODES[MAX_UNTIL])
94                else:
95                    emit(OPCODES[MIN_UNTIL])
96        elif op is SUBPATTERN:
97            if av[0]:
98                emit(OPCODES[MARK])
99                emit((av[0]-1)*2)
100            # _compile_info(code, av[1], flags)
101            _compile(code, av[1], flags)
102            if av[0]:
103                emit(OPCODES[MARK])
104                emit((av[0]-1)*2+1)
105        elif op in SUCCESS_CODES:
106            emit(OPCODES[op])
107        elif op in ASSERT_CODES:
108            emit(OPCODES[op])
109            skip = _len(code); emit(0)
110            if av[0] >= 0:
111                emit(0) # look ahead
112            else:
113                lo, hi = av[1].getwidth()
114                if lo != hi:
115                    raise error, "look-behind requires fixed-width pattern"
116                emit(lo) # look behind
117            _compile(code, av[1], flags)
118            emit(OPCODES[SUCCESS])
119            code[skip] = _len(code) - skip
120        elif op is CALL:
121            emit(OPCODES[op])
122            skip = _len(code); emit(0)
123            _compile(code, av, flags)
124            emit(OPCODES[SUCCESS])
125            code[skip] = _len(code) - skip
126        elif op is AT:
127            emit(OPCODES[op])
128            if flags & SRE_FLAG_MULTILINE:
129                av = AT_MULTILINE.get(av, av)
130            if flags & SRE_FLAG_LOCALE:
131                av = AT_LOCALE.get(av, av)
132            elif flags & SRE_FLAG_UNICODE:
133                av = AT_UNICODE.get(av, av)
134            emit(ATCODES[av])
135        elif op is BRANCH:
136            emit(OPCODES[op])
137            tail = []
138            tailappend = tail.append
139            for av in av[1]:
140                skip = _len(code); emit(0)
141                # _compile_info(code, av, flags)
142                _compile(code, av, flags)
143                emit(OPCODES[JUMP])
144                tailappend(_len(code)); emit(0)
145                code[skip] = _len(code) - skip
146            emit(0) # end of branch
147            for tail in tail:
148                code[tail] = _len(code) - tail
149        elif op is CATEGORY:
150            emit(OPCODES[op])
151            if flags & SRE_FLAG_LOCALE:
152                av = CH_LOCALE[av]
153            elif flags & SRE_FLAG_UNICODE:
154                av = CH_UNICODE[av]
155            emit(CHCODES[av])
156        elif op is GROUPREF:
157            if flags & SRE_FLAG_IGNORECASE:
158                emit(OPCODES[OP_IGNORE[op]])
159            else:
160                emit(OPCODES[op])
161            emit(av-1)
162        elif op is GROUPREF_EXISTS:
163            emit(OPCODES[op])
164            emit(av[0]-1)
165            skipyes = _len(code); emit(0)
166            _compile(code, av[1], flags)
167            if av[2]:
168                emit(OPCODES[JUMP])
169                skipno = _len(code); emit(0)
170                code[skipyes] = _len(code) - skipyes + 1
171                _compile(code, av[2], flags)
172                code[skipno] = _len(code) - skipno
173            else:
174                code[skipyes] = _len(code) - skipyes + 1
175        else:
176            raise ValueError, ("unsupported operand type", op)
177
178def _compile_charset(charset, flags, code, fixup=None):
179    # compile charset subprogram
180    emit = code.append
181    if fixup is None:
182        fixup = _identityfunction
183    for op, av in _optimize_charset(charset, fixup):
184        emit(OPCODES[op])
185        if op is NEGATE:
186            pass
187        elif op is LITERAL:
188            emit(fixup(av))
189        elif op is RANGE:
190            emit(fixup(av[0]))
191            emit(fixup(av[1]))
192        elif op is CHARSET:
193            code.extend(av)
194        elif op is BIGCHARSET:
195            code.extend(av)
196        elif op is CATEGORY:
197            if flags & SRE_FLAG_LOCALE:
198                emit(CHCODES[CH_LOCALE[av]])
199            elif flags & SRE_FLAG_UNICODE:
200                emit(CHCODES[CH_UNICODE[av]])
201            else:
202                emit(CHCODES[av])
203        else:
204            raise error, "internal: unsupported set operator"
205    emit(OPCODES[FAILURE])
206
207def _optimize_charset(charset, fixup):
208    # internal: optimize character set
209    out = []
210    outappend = out.append
211    charmap = [0]*256
212    try:
213        for op, av in charset:
214            if op is NEGATE:
215                outappend((op, av))
216            elif op is LITERAL:
217                charmap[fixup(av)] = 1
218            elif op is RANGE:
219                for i in range(fixup(av[0]), fixup(av[1])+1):
220                    charmap[i] = 1
221            elif op is CATEGORY:
222                # XXX: could append to charmap tail
223                return charset # cannot compress
224    except IndexError:
225        # character set contains unicode characters
226        return _optimize_unicode(charset, fixup)
227    # compress character map
228    i = p = n = 0
229    runs = []
230    runsappend = runs.append
231    for c in charmap:
232        if c:
233            if n == 0:
234                p = i
235            n = n + 1
236        elif n:
237            runsappend((p, n))
238            n = 0
239        i = i + 1
240    if n:
241        runsappend((p, n))
242    if len(runs) <= 2:
243        # use literal/range
244        for p, n in runs:
245            if n == 1:
246                outappend((LITERAL, p))
247            else:
248                outappend((RANGE, (p, p+n-1)))
249        if len(out) < len(charset):
250            return out
251    else:
252        # use bitmap
253        data = _mk_bitmap(charmap)
254        outappend((CHARSET, data))
255        return out
256    return charset
257
258def _mk_bitmap(bits):
259    data = []
260    dataappend = data.append
261    if _sre.CODESIZE == 2:
262        start = (1, 0)
263    else:
264        start = (1L, 0L)
265    m, v = start
266    for c in bits:
267        if c:
268            v = v + m
269        m = m + m
270        if m > MAXCODE:
271            dataappend(v)
272            m, v = start
273    return data
274
275# To represent a big charset, first a bitmap of all characters in the
276# set is constructed. Then, this bitmap is sliced into chunks of 256
277# characters, duplicate chunks are eliminated, and each chunk is
278# given a number. In the compiled expression, the charset is
279# represented by a 32-bit word sequence, consisting of one word for
280# the number of different chunks, a sequence of 256 bytes (64 words)
281# of chunk numbers indexed by their original chunk position, and a
282# sequence of 256-bit chunks (8 words each).
283
284# Compression is normally good: in a typical charset, large ranges of
285# Unicode will be either completely excluded (e.g. if only cyrillic
286# letters are to be matched), or completely included (e.g. if large
287# subranges of Kanji match). These ranges will be represented by
288# chunks of all one-bits or all zero-bits.
289
290# Matching can be also done efficiently: the more significant byte of
291# the Unicode character is an index into the chunk number, and the
292# less significant byte is a bit index in the chunk (just like the
293# CHARSET matching).
294
295# In UCS-4 mode, the BIGCHARSET opcode still supports only subsets
296# of the basic multilingual plane; an efficient representation
297# for all of Unicode has not yet been developed. This means,
298# in particular, that negated charsets cannot be represented as
299# bigcharsets.
300
301def _optimize_unicode(charset, fixup):
302    try:
303        import array
304    except ImportError:
305        return charset
306    charmap = [0]*65536
307    negate = 0
308    try:
309        for op, av in charset:
310            if op is NEGATE:
311                negate = 1
312            elif op is LITERAL:
313                charmap[fixup(av)] = 1
314            elif op is RANGE:
315                for i in xrange(fixup(av[0]), fixup(av[1])+1):
316                    charmap[i] = 1
317            elif op is CATEGORY:
318                # XXX: could expand category
319                return charset # cannot compress
320    except IndexError:
321        # non-BMP characters
322        return charset
323    if negate:
324        if sys.maxunicode != 65535:
325            # XXX: negation does not work with big charsets
326            return charset
327        for i in xrange(65536):
328            charmap[i] = not charmap[i]
329    comps = {}
330    mapping = [0]*256
331    block = 0
332    data = []
333    for i in xrange(256):
334        chunk = tuple(charmap[i*256:(i+1)*256])
335        new = comps.setdefault(chunk, block)
336        mapping[i] = new
337        if new == block:
338            block = block + 1
339            data = data + _mk_bitmap(chunk)
340    header = [block]
341    if _sre.CODESIZE == 2:
342        code = 'H'
343    else:
344        code = 'I'
345    # Convert block indices to byte array of 256 bytes
346    mapping = array.array('B', mapping).tostring()
347    # Convert byte array to word array
348    mapping = array.array(code, mapping)
349    assert mapping.itemsize == _sre.CODESIZE
350    header = header + mapping.tolist()
351    data[0:0] = header
352    return [(BIGCHARSET, data)]
353
354def _simple(av):
355    # check if av is a "simple" operator
356    lo, hi = av[2].getwidth()
357    return lo == hi == 1 and av[2][0][0] != SUBPATTERN
358
359def _compile_info(code, pattern, flags):
360    # internal: compile an info block.  in the current version,
361    # this contains min/max pattern width, and an optional literal
362    # prefix or a character map
363    lo, hi = pattern.getwidth()
364    if lo == 0:
365        return # not worth it
366    # look for a literal prefix
367    prefix = []
368    prefixappend = prefix.append
369    prefix_skip = 0
370    charset = [] # not used
371    charsetappend = charset.append
372    if not (flags & SRE_FLAG_IGNORECASE):
373        # look for literal prefix
374        for op, av in pattern.data:
375            if op is LITERAL:
376                if len(prefix) == prefix_skip:
377                    prefix_skip = prefix_skip + 1
378                prefixappend(av)
379            elif op is SUBPATTERN and len(av[1]) == 1:
380                op, av = av[1][0]
381                if op is LITERAL:
382                    prefixappend(av)
383                else:
384                    break
385            else:
386                break
387        # if no prefix, look for charset prefix
388        if not prefix and pattern.data:
389            op, av = pattern.data[0]
390            if op is SUBPATTERN and av[1]:
391                op, av = av[1][0]
392                if op is LITERAL:
393                    charsetappend((op, av))
394                elif op is BRANCH:
395                    c = []
396                    cappend = c.append
397                    for p in av[1]:
398                        if not p:
399                            break
400                        op, av = p[0]
401                        if op is LITERAL:
402                            cappend((op, av))
403                        else:
404                            break
405                    else:
406                        charset = c
407            elif op is BRANCH:
408                c = []
409                cappend = c.append
410                for p in av[1]:
411                    if not p:
412                        break
413                    op, av = p[0]
414                    if op is LITERAL:
415                        cappend((op, av))
416                    else:
417                        break
418                else:
419                    charset = c
420            elif op is IN:
421                charset = av
422##     if prefix:
423##         print "*** PREFIX", prefix, prefix_skip
424##     if charset:
425##         print "*** CHARSET", charset
426    # add an info block
427    emit = code.append
428    emit(OPCODES[INFO])
429    skip = len(code); emit(0)
430    # literal flag
431    mask = 0
432    if prefix:
433        mask = SRE_INFO_PREFIX
434        if len(prefix) == prefix_skip == len(pattern.data):
435            mask = mask + SRE_INFO_LITERAL
436    elif charset:
437        mask = mask + SRE_INFO_CHARSET
438    emit(mask)
439    # pattern length
440    if lo < MAXCODE:
441        emit(lo)
442    else:
443        emit(MAXCODE)
444        prefix = prefix[:MAXCODE]
445    if hi < MAXCODE:
446        emit(hi)
447    else:
448        emit(0)
449    # add literal prefix
450    if prefix:
451        emit(len(prefix)) # length
452        emit(prefix_skip) # skip
453        code.extend(prefix)
454        # generate overlap table
455        table = [-1] + ([0]*len(prefix))
456        for i in xrange(len(prefix)):
457            table[i+1] = table[i]+1
458            while table[i+1] > 0 and prefix[i] != prefix[table[i+1]-1]:
459                table[i+1] = table[table[i+1]-1]+1
460        code.extend(table[1:]) # don't store first entry
461    elif charset:
462        _compile_charset(charset, flags, code)
463    code[skip] = len(code) - skip
464
465try:
466    unicode
467except NameError:
468    STRING_TYPES = (type(""),)
469else:
470    STRING_TYPES = (type(""), type(unicode("")))
471
472def isstring(obj):
473    for tp in STRING_TYPES:
474        if isinstance(obj, tp):
475            return 1
476    return 0
477
478def _code(p, flags):
479
480    flags = p.pattern.flags | flags
481    code = []
482
483    # compile info block
484    _compile_info(code, p, flags)
485
486    # compile the pattern
487    _compile(code, p.data, flags)
488
489    code.append(OPCODES[SUCCESS])
490
491    return code
492
493def compile(p, flags=0):
494    # internal: convert pattern list to internal format
495
496    if isstring(p):
497        pattern = p
498        p = sre_parse.parse(p, flags)
499    else:
500        pattern = None
501
502    code = _code(p, flags)
503
504    # print code
505
506    # XXX: <fl> get rid of this limitation!
507    if p.pattern.groups > 100:
508        raise AssertionError(
509            "sorry, but this version only supports 100 named groups"
510            )
511
512    # map in either direction
513    groupindex = p.pattern.groupdict
514    indexgroup = [None] * p.pattern.groups
515    for k, i in groupindex.items():
516        indexgroup[i] = k
517
518    return _sre.compile(
519        pattern, flags | p.pattern.flags, code,
520        p.pattern.groups-1,
521        groupindex, indexgroup
522        )
523