PyParse.py 19 KB
Newer Older
David Scherer's avatar
David Scherer committed
1 2 3 4
import re
import sys

# Reason last stmt is continued (or C_NONE if it's not).
5 6
(C_NONE, C_BACKSLASH, C_STRING_FIRST_LINE,
 C_STRING_NEXT_LINES, C_BRACKET) = range(5)
David Scherer's avatar
David Scherer committed
7 8 9

if 0:   # for throwaway debugging output
    def dump(*stuff):
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
10
        sys.__stdout__.write(" ".join(map(str, stuff)) + "\n")
David Scherer's avatar
David Scherer committed
11 12 13 14 15 16

# Find what looks like the start of a popular stmt.

_synchre = re.compile(r"""
    ^
    [ \t]*
17
    (?: while
David Scherer's avatar
David Scherer committed
18 19 20 21 22 23 24 25 26 27 28 29
    |   else
    |   def
    |   return
    |   assert
    |   break
    |   class
    |   continue
    |   elif
    |   try
    |   except
    |   raise
    |   import
30
    |   yield
David Scherer's avatar
David Scherer committed
31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
    )
    \b
""", re.VERBOSE | re.MULTILINE).search

# Match blank line or non-indenting comment line.

_junkre = re.compile(r"""
    [ \t]*
    (?: \# \S .* )?
    \n
""", re.VERBOSE).match

# Match any flavor of string; the terminating quote is optional
# so that we're robust in the face of incomplete program text.

_match_stringre = re.compile(r"""
    \""" [^"\\]* (?:
                     (?: \\. | "(?!"") )
                     [^"\\]*
                 )*
    (?: \""" )?

|   " [^"\\\n]* (?: \\. [^"\\\n]* )* "?

|   ''' [^'\\]* (?:
                   (?: \\. | '(?!'') )
                   [^'\\]*
                )*
    (?: ''' )?

|   ' [^'\\\n]* (?: \\. [^'\\\n]* )* '?
""", re.VERBOSE | re.DOTALL).match

# Match a line that starts with something interesting;
# used to find the first item of a bracket structure.

_itemre = re.compile(r"""
    [ \t]*
    [^\s#\\]    # if we match, m.end()-1 is the interesting char
""", re.VERBOSE).match

# Match start of stmts that should be followed by a dedent.

_closere = re.compile(r"""
    \s*
    (?: return
    |   break
    |   continue
    |   raise
    |   pass
    )
    \b
""", re.VERBOSE).match

# Chew up non-special chars as quickly as possible.  If match is
# successful, m.end() less 1 is the index of the last boring char
# matched.  If match is unsuccessful, the string starts with an
# interesting char.

_chew_ordinaryre = re.compile(r"""
    [^[\](){}#'"\\]+
""", re.VERBOSE).match

# Build translation table to map uninteresting chars to "x", open
# brackets to "(", and close brackets to ")".

97 98 99
_tran = {}
for i in range(256):
    _tran[i] = 'x'
David Scherer's avatar
David Scherer committed
100 101 102 103 104 105
for ch in "({[":
    _tran[ord(ch)] = '('
for ch in ")}]":
    _tran[ord(ch)] = ')'
for ch in "\"'\\\n#":
    _tran[ord(ch)] = ch
106
del i, ch
David Scherer's avatar
David Scherer committed
107 108 109 110 111 112 113

class Parser:

    def __init__(self, indentwidth, tabwidth):
        self.indentwidth = indentwidth
        self.tabwidth = tabwidth

114 115 116
    def set_str(self, s):
        assert len(s) == 0 or s[-1] == '\n'
        if isinstance(s, str):
117 118 119 120 121 122
            # The parse functions have no idea what to do with Unicode, so
            # replace all Unicode characters with "x".  This is "safe"
            # so long as the only characters germane to parsing the structure
            # of Python are 7-bit ASCII.  It's *necessary* because Unicode
            # strings don't have a .translate() method that supports
            # deletechars.
123
            uniphooey = s
Martin v. Löwis's avatar
Martin v. Löwis committed
124
            s = []
125
            push = s.append
126 127
            for raw in map(ord, uniphooey):
                push(raw < 127 and chr(raw) or "x")
128 129
            s = "".join(s)
        self.str = s
David Scherer's avatar
David Scherer committed
130 131 132 133 134 135 136 137 138 139 140 141 142
        self.study_level = 0

    # Return index of a good place to begin parsing, as close to the
    # end of the string as possible.  This will be the start of some
    # popular stmt like "if" or "def".  Return None if none found:
    # the caller should pass more prior context then, if possible, or
    # if not (the entire program text up until the point of interest
    # has already been tried) pass 0 to set_lo.
    #
    # This will be reliable iff given a reliable is_char_in_string
    # function, meaning that when it says "no", it's absolutely
    # guaranteed that the char is not in a string.

143
    def find_good_parse_start(self, is_char_in_string=None,
David Scherer's avatar
David Scherer committed
144 145 146 147 148 149 150 151 152 153 154 155
                              _synchre=_synchre):
        str, pos = self.str, None

        if not is_char_in_string:
            # no clue -- make the caller pass everything
            return None

        # Peek back from the end for a good place to start,
        # but don't try too often; pos will be left None, or
        # bumped to a legitimate synch point.
        limit = len(str)
        for tries in range(5):
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
156
            i = str.rfind(":\n", 0, limit)
David Scherer's avatar
David Scherer committed
157 158
            if i < 0:
                break
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
159
            i = str.rfind('\n', 0, i) + 1  # start of colon line
David Scherer's avatar
David Scherer committed
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203
            m = _synchre(str, i, limit)
            if m and not is_char_in_string(m.start()):
                pos = m.start()
                break
            limit = i
        if pos is None:
            # Nothing looks like a block-opener, or stuff does
            # but is_char_in_string keeps returning true; most likely
            # we're in or near a giant string, the colorizer hasn't
            # caught up enough to be helpful, or there simply *aren't*
            # any interesting stmts.  In any of these cases we're
            # going to have to parse the whole thing to be sure, so
            # give it one last try from the start, but stop wasting
            # time here regardless of the outcome.
            m = _synchre(str)
            if m and not is_char_in_string(m.start()):
                pos = m.start()
            return pos

        # Peeking back worked; look forward until _synchre no longer
        # matches.
        i = pos + 1
        while 1:
            m = _synchre(str, i)
            if m:
                s, i = m.span()
                if not is_char_in_string(s):
                    pos = s
            else:
                break
        return pos

    # Throw away the start of the string.  Intended to be called with
    # find_good_parse_start's result.

    def set_lo(self, lo):
        assert lo == 0 or self.str[lo-1] == '\n'
        if lo > 0:
            self.str = self.str[lo:]

    # As quickly as humanly possible <wink>, find the line numbers (0-
    # based) of the non-continuation lines.
    # Creates self.{goodlines, continuation}.

Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
204
    def _study1(self):
David Scherer's avatar
David Scherer committed
205 206 207 208 209 210 211 212 213
        if self.study_level >= 1:
            return
        self.study_level = 1

        # Map all uninteresting characters to "x", all open brackets
        # to "(", all close brackets to ")", then collapse runs of
        # uninteresting characters.  This can cut the number of chars
        # by a factor of 10-40, and so greatly speed the following loop.
        str = self.str
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
214 215 216 217 218 219
        str = str.translate(_tran)
        str = str.replace('xxxxxxxx', 'x')
        str = str.replace('xxxx', 'x')
        str = str.replace('xx', 'x')
        str = str.replace('xx', 'x')
        str = str.replace('\nx', '\n')
David Scherer's avatar
David Scherer committed
220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
        # note that replacing x\n with \n would be incorrect, because
        # x may be preceded by a backslash

        # March over the squashed version of the program, accumulating
        # the line numbers of non-continued stmts, and determining
        # whether & why the last stmt is a continuation.
        continuation = C_NONE
        level = lno = 0     # level is nesting level; lno is line number
        self.goodlines = goodlines = [0]
        push_good = goodlines.append
        i, n = 0, len(str)
        while i < n:
            ch = str[i]
            i = i+1

            # cases are checked in decreasing order of frequency
            if ch == 'x':
                continue

            if ch == '\n':
                lno = lno + 1
                if level == 0:
                    push_good(lno)
                    # else we're in an unclosed bracket structure
                continue

            if ch == '(':
                level = level + 1
                continue

            if ch == ')':
                if level:
                    level = level - 1
                    # else the program is invalid, but we can't complain
                continue

            if ch == '"' or ch == "'":
                # consume the string
                quote = ch
                if str[i-1:i+2] == quote * 3:
                    quote = quote * 3
261
                firstlno = lno
David Scherer's avatar
David Scherer committed
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295
                w = len(quote) - 1
                i = i+w
                while i < n:
                    ch = str[i]
                    i = i+1

                    if ch == 'x':
                        continue

                    if str[i-1:i+w] == quote:
                        i = i+w
                        break

                    if ch == '\n':
                        lno = lno + 1
                        if w == 0:
                            # unterminated single-quoted string
                            if level == 0:
                                push_good(lno)
                            break
                        continue

                    if ch == '\\':
                        assert i < n
                        if str[i] == '\n':
                            lno = lno + 1
                        i = i+1
                        continue

                    # else comment char or paren inside string

                else:
                    # didn't break out of the loop, so we're still
                    # inside a string
296 297 298 299 300 301
                    if (lno - 1) == firstlno:
                        # before the previous \n in str, we were in the first
                        # line of the string
                        continuation = C_STRING_FIRST_LINE
                    else:
                        continuation = C_STRING_NEXT_LINES
David Scherer's avatar
David Scherer committed
302 303 304 305
                continue    # with outer loop

            if ch == '#':
                # consume the comment
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
306
                i = str.find('\n', i)
David Scherer's avatar
David Scherer committed
307 308 309 310 311 312 313 314 315 316 317 318 319 320
                assert i >= 0
                continue

            assert ch == '\\'
            assert i < n
            if str[i] == '\n':
                lno = lno + 1
                if i+1 == n:
                    continuation = C_BACKSLASH
            i = i+1

        # The last stmt may be continued for all 3 reasons.
        # String continuation takes precedence over bracket
        # continuation, which beats backslash continuation.
321 322
        if (continuation != C_STRING_FIRST_LINE
            and continuation != C_STRING_NEXT_LINES and level > 0):
David Scherer's avatar
David Scherer committed
323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341
            continuation = C_BRACKET
        self.continuation = continuation

        # Push the final line number as a sentinel value, regardless of
        # whether it's continued.
        assert (continuation == C_NONE) == (goodlines[-1] == lno)
        if goodlines[-1] != lno:
            push_good(lno)

    def get_continuation_type(self):
        self._study1()
        return self.continuation

    # study1 was sufficient to determine the continuation status,
    # but doing more requires looking at every character.  study2
    # does this for the last interesting statement in the block.
    # Creates:
    #     self.stmt_start, stmt_end
    #         slice indices of last interesting stmt
342 343 344 345 346
    #     self.stmt_bracketing
    #         the bracketing structure of the last interesting stmt;
    #         for example, for the statement "say(boo) or die", stmt_bracketing
    #         will be [(0, 0), (3, 1), (8, 0)]. Strings and comments are
    #         treated as brackets, for the matter.
David Scherer's avatar
David Scherer committed
347 348 349 350 351 352
    #     self.lastch
    #         last non-whitespace character before optional trailing
    #         comment
    #     self.lastopenbracketpos
    #         if continuation is C_BRACKET, index of last open bracket

Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
353
    def _study2(self):
David Scherer's avatar
David Scherer committed
354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369
        if self.study_level >= 2:
            return
        self._study1()
        self.study_level = 2

        # Set p and q to slice indices of last interesting stmt.
        str, goodlines = self.str, self.goodlines
        i = len(goodlines) - 1
        p = len(str)    # index of newest line
        while i:
            assert p
            # p is the index of the stmt at line number goodlines[i].
            # Move p back to the stmt at line number goodlines[i-1].
            q = p
            for nothing in range(goodlines[i-1], goodlines[i]):
                # tricky: sets p to 0 if no preceding newline
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
370
                p = str.rfind('\n', 0, p-1) + 1
David Scherer's avatar
David Scherer committed
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
            # The stmt str[p:q] isn't a continuation, but may be blank
            # or a non-indenting comment line.
            if  _junkre(str, p):
                i = i-1
            else:
                break
        if i == 0:
            # nothing but junk!
            assert p == 0
            q = p
        self.stmt_start, self.stmt_end = p, q

        # Analyze this stmt, to find the last open bracket (if any)
        # and last interesting character (if any).
        lastch = ""
        stack = []  # stack of open bracket indices
        push_stack = stack.append
388
        bracketing = [(p, 0)]
David Scherer's avatar
David Scherer committed
389 390 391 392 393
        while p < q:
            # suck up all except ()[]{}'"#\\
            m = _chew_ordinaryre(str, p, q)
            if m:
                # we skipped at least one boring char
394
                newp = m.end()
David Scherer's avatar
David Scherer committed
395
                # back up over totally boring whitespace
396 397
                i = newp - 1    # index of last boring char
                while i >= p and str[i] in " \t\n":
David Scherer's avatar
David Scherer committed
398
                    i = i-1
399
                if i >= p:
David Scherer's avatar
David Scherer committed
400
                    lastch = str[i]
401
                p = newp
David Scherer's avatar
David Scherer committed
402 403 404 405 406 407 408
                if p >= q:
                    break

            ch = str[p]

            if ch in "([{":
                push_stack(p)
409
                bracketing.append((p, len(stack)))
David Scherer's avatar
David Scherer committed
410 411 412 413 414 415 416 417 418
                lastch = ch
                p = p+1
                continue

            if ch in ")]}":
                if stack:
                    del stack[-1]
                lastch = ch
                p = p+1
419
                bracketing.append((p, len(stack)))
David Scherer's avatar
David Scherer committed
420 421 422 423 424 425 426 427 428 429
                continue

            if ch == '"' or ch == "'":
                # consume string
                # Note that study1 did this with a Python loop, but
                # we use a regexp here; the reason is speed in both
                # cases; the string may be huge, but study1 pre-squashed
                # strings to a couple of characters per line.  study1
                # also needed to keep track of newlines, and we don't
                # have to.
430
                bracketing.append((p, len(stack)+1))
David Scherer's avatar
David Scherer committed
431 432
                lastch = ch
                p = _match_stringre(str, p, q).end()
433
                bracketing.append((p, len(stack)))
David Scherer's avatar
David Scherer committed
434 435 436 437
                continue

            if ch == '#':
                # consume comment and trailing newline
438
                bracketing.append((p, len(stack)+1))
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
439
                p = str.find('\n', p, q) + 1
David Scherer's avatar
David Scherer committed
440
                assert p > 0
441
                bracketing.append((p, len(stack)))
David Scherer's avatar
David Scherer committed
442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
                continue

            assert ch == '\\'
            p = p+1     # beyond backslash
            assert p < q
            if str[p] != '\n':
                # the program is invalid, but can't complain
                lastch = ch + str[p]
            p = p+1     # beyond escaped char

        # end while p < q:

        self.lastch = lastch
        if stack:
            self.lastopenbracketpos = stack[-1]
457
        self.stmt_bracketing = tuple(bracketing)
David Scherer's avatar
David Scherer committed
458 459 460 461

    # Assuming continuation is C_BRACKET, return the number
    # of spaces the next line should be indented.

Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
462
    def compute_bracket_indent(self):
David Scherer's avatar
David Scherer committed
463 464 465 466 467
        self._study2()
        assert self.continuation == C_BRACKET
        j = self.lastopenbracketpos
        str = self.str
        n = len(str)
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
468
        origi = i = str.rfind('\n', 0, j) + 1
David Scherer's avatar
David Scherer committed
469 470 471 472 473 474 475 476 477 478
        j = j+1     # one beyond open bracket
        # find first list item; set i to start of its line
        while j < n:
            m = _itemre(str, j)
            if m:
                j = m.end() - 1     # index of first interesting char
                extra = 0
                break
            else:
                # this line is junk; advance to next line
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
479
                i = j = str.find('\n', j) + 1
David Scherer's avatar
David Scherer committed
480 481 482 483 484 485 486
        else:
            # nothing interesting follows the bracket;
            # reproduce the bracket line's indentation + a level
            j = i = origi
            while str[j] in " \t":
                j = j+1
            extra = self.indentwidth
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
487
        return len(str[i:j].expandtabs(self.tabwidth)) + extra
David Scherer's avatar
David Scherer committed
488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512

    # Return number of physical lines in last stmt (whether or not
    # it's an interesting stmt!  this is intended to be called when
    # continuation is C_BACKSLASH).

    def get_num_lines_in_stmt(self):
        self._study1()
        goodlines = self.goodlines
        return goodlines[-1] - goodlines[-2]

    # Assuming continuation is C_BACKSLASH, return the number of spaces
    # the next line should be indented.  Also assuming the new line is
    # the first one following the initial line of the stmt.

    def compute_backslash_indent(self):
        self._study2()
        assert self.continuation == C_BACKSLASH
        str = self.str
        i = self.stmt_start
        while str[i] in " \t":
            i = i+1
        startpos = i

        # See whether the initial line starts an assignment stmt; i.e.,
        # look for an = operator
Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
513
        endpos = str.find('\n', startpos) + 1
David Scherer's avatar
David Scherer committed
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
        found = level = 0
        while i < endpos:
            ch = str[i]
            if ch in "([{":
                level = level + 1
                i = i+1
            elif ch in ")]}":
                if level:
                    level = level - 1
                i = i+1
            elif ch == '"' or ch == "'":
                i = _match_stringre(str, i, endpos).end()
            elif ch == '#':
                break
            elif level == 0 and ch == '=' and \
                   (i == 0 or str[i-1] not in "=<>!") and \
                   str[i+1] != '=':
                found = 1
                break
            else:
                i = i+1

        if found:
            # found a legit =, but it may be the last interesting
            # thing on the line
            i = i+1     # move beyond the =
            found = re.match(r"\s*\\", str[i:endpos]) is None

        if not found:
            # oh well ... settle for moving beyond the first chunk
            # of non-whitespace chars
            i = startpos
            while str[i] not in " \t\n":
                i = i+1

Kurt B. Kaiser's avatar
Kurt B. Kaiser committed
549
        return len(str[self.stmt_start:i].expandtabs(\
David Scherer's avatar
David Scherer committed
550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581
                                     self.tabwidth)) + 1

    # Return the leading whitespace on the initial line of the last
    # interesting stmt.

    def get_base_indent_string(self):
        self._study2()
        i, n = self.stmt_start, self.stmt_end
        j = i
        str = self.str
        while j < n and str[j] in " \t":
            j = j + 1
        return str[i:j]

    # Did the last interesting stmt open a block?

    def is_block_opener(self):
        self._study2()
        return self.lastch == ':'

    # Did the last interesting stmt close a block?

    def is_block_closer(self):
        self._study2()
        return _closere(self.str, self.stmt_start) is not None

    # index of last open bracket ({[, or None if none
    lastopenbracketpos = None

    def get_last_open_bracket_pos(self):
        self._study2()
        return self.lastopenbracketpos
582 583 584 585 586 587 588 589 590

    # the structure of the bracketing of the last interesting statement,
    # in the format defined in _study2, or None if the text didn't contain
    # anything
    stmt_bracketing = None

    def get_last_stmt_bracketing(self):
        self._study2()
        return self.stmt_bracketing