• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1"""Define partial Python code Parser used by editor and hyperparser.
2
3Instances of ParseMap are used with str.translate.
4
5The following bound search and match functions are defined:
6_synchre - start of popular statement;
7_junkre - whitespace or comment line;
8_match_stringre: string, possibly without closer;
9_itemre - line that may have bracket structure start;
10_closere - line that must be followed by dedent.
11_chew_ordinaryre - non-special characters.
12"""
13import re
14
15# Reason last statement is continued (or C_NONE if it's not).
16(C_NONE, C_BACKSLASH, C_STRING_FIRST_LINE,
17 C_STRING_NEXT_LINES, C_BRACKET) = range(5)
18
19# Find what looks like the start of a popular statement.
20
21_synchre = re.compile(r"""
22    ^
23    [ \t]*
24    (?: while
25    |   else
26    |   def
27    |   return
28    |   assert
29    |   break
30    |   class
31    |   continue
32    |   elif
33    |   try
34    |   except
35    |   raise
36    |   import
37    |   yield
38    )
39    \b
40""", re.VERBOSE | re.MULTILINE).search
41
42# Match blank line or non-indenting comment line.
43
44_junkre = re.compile(r"""
45    [ \t]*
46    (?: \# \S .* )?
47    \n
48""", re.VERBOSE).match
49
50# Match any flavor of string; the terminating quote is optional
51# so that we're robust in the face of incomplete program text.
52
53_match_stringre = re.compile(r"""
54    \""" [^"\\]* (?:
55                     (?: \\. | "(?!"") )
56                     [^"\\]*
57                 )*
58    (?: \""" )?
59
60|   " [^"\\\n]* (?: \\. [^"\\\n]* )* "?
61
62|   ''' [^'\\]* (?:
63                   (?: \\. | '(?!'') )
64                   [^'\\]*
65                )*
66    (?: ''' )?
67
68|   ' [^'\\\n]* (?: \\. [^'\\\n]* )* '?
69""", re.VERBOSE | re.DOTALL).match
70
71# Match a line that starts with something interesting;
72# used to find the first item of a bracket structure.
73
74_itemre = re.compile(r"""
75    [ \t]*
76    [^\s#\\]    # if we match, m.end()-1 is the interesting char
77""", re.VERBOSE).match
78
79# Match start of statements that should be followed by a dedent.
80
81_closere = re.compile(r"""
82    \s*
83    (?: return
84    |   break
85    |   continue
86    |   raise
87    |   pass
88    )
89    \b
90""", re.VERBOSE).match
91
92# Chew up non-special chars as quickly as possible.  If match is
93# successful, m.end() less 1 is the index of the last boring char
94# matched.  If match is unsuccessful, the string starts with an
95# interesting char.
96
97_chew_ordinaryre = re.compile(r"""
98    [^[\](){}#'"\\]+
99""", re.VERBOSE).match
100
101
102class ParseMap(dict):
103    r"""Dict subclass that maps anything not in dict to 'x'.
104
105    This is designed to be used with str.translate in study1.
106    Anything not specifically mapped otherwise becomes 'x'.
107    Example: replace everything except whitespace with 'x'.
108
109    >>> keepwhite = ParseMap((ord(c), ord(c)) for c in ' \t\n\r')
110    >>> "a + b\tc\nd".translate(keepwhite)
111    'x x x\tx\nx'
112    """
113    # Calling this triples access time; see bpo-32940
114    def __missing__(self, key):
115        return 120  # ord('x')
116
117
118# Map all ascii to 120 to avoid __missing__ call, then replace some.
119trans = ParseMap.fromkeys(range(128), 120)
120trans.update((ord(c), ord('(')) for c in "({[")  # open brackets => '(';
121trans.update((ord(c), ord(')')) for c in ")}]")  # close brackets => ')'.
122trans.update((ord(c), ord(c)) for c in "\"'\\\n#")  # Keep these.
123
124
125class Parser:
126
127    def __init__(self, indentwidth, tabwidth):
128        self.indentwidth = indentwidth
129        self.tabwidth = tabwidth
130
131    def set_code(self, s):
132        assert len(s) == 0 or s[-1] == '\n'
133        self.code = s
134        self.study_level = 0
135
136    def find_good_parse_start(self, is_char_in_string):
137        """
138        Return index of a good place to begin parsing, as close to the
139        end of the string as possible.  This will be the start of some
140        popular stmt like "if" or "def".  Return None if none found:
141        the caller should pass more prior context then, if possible, or
142        if not (the entire program text up until the point of interest
143        has already been tried) pass 0 to set_lo().
144
145        This will be reliable iff given a reliable is_char_in_string()
146        function, meaning that when it says "no", it's absolutely
147        guaranteed that the char is not in a string.
148        """
149        code, pos = self.code, None
150
151        # Peek back from the end for a good place to start,
152        # but don't try too often; pos will be left None, or
153        # bumped to a legitimate synch point.
154        limit = len(code)
155        for tries in range(5):
156            i = code.rfind(":\n", 0, limit)
157            if i < 0:
158                break
159            i = code.rfind('\n', 0, i) + 1  # start of colon line (-1+1=0)
160            m = _synchre(code, i, limit)
161            if m and not is_char_in_string(m.start()):
162                pos = m.start()
163                break
164            limit = i
165        if pos is None:
166            # Nothing looks like a block-opener, or stuff does
167            # but is_char_in_string keeps returning true; most likely
168            # we're in or near a giant string, the colorizer hasn't
169            # caught up enough to be helpful, or there simply *aren't*
170            # any interesting stmts.  In any of these cases we're
171            # going to have to parse the whole thing to be sure, so
172            # give it one last try from the start, but stop wasting
173            # time here regardless of the outcome.
174            m = _synchre(code)
175            if m and not is_char_in_string(m.start()):
176                pos = m.start()
177            return pos
178
179        # Peeking back worked; look forward until _synchre no longer
180        # matches.
181        i = pos + 1
182        while m := _synchre(code, i):
183            s, i = m.span()
184            if not is_char_in_string(s):
185                pos = s
186        return pos
187
188    def set_lo(self, lo):
189        """ Throw away the start of the string.
190
191        Intended to be called with the result of find_good_parse_start().
192        """
193        assert lo == 0 or self.code[lo-1] == '\n'
194        if lo > 0:
195            self.code = self.code[lo:]
196
197    def _study1(self):
198        """Find the line numbers of non-continuation lines.
199
200        As quickly as humanly possible <wink>, find the line numbers (0-
201        based) of the non-continuation lines.
202        Creates self.{goodlines, continuation}.
203        """
204        if self.study_level >= 1:
205            return
206        self.study_level = 1
207
208        # Map all uninteresting characters to "x", all open brackets
209        # to "(", all close brackets to ")", then collapse runs of
210        # uninteresting characters.  This can cut the number of chars
211        # by a factor of 10-40, and so greatly speed the following loop.
212        code = self.code
213        code = code.translate(trans)
214        code = code.replace('xxxxxxxx', 'x')
215        code = code.replace('xxxx', 'x')
216        code = code.replace('xx', 'x')
217        code = code.replace('xx', 'x')
218        code = code.replace('\nx', '\n')
219        # Replacing x\n with \n would be incorrect because
220        # x may be preceded by a backslash.
221
222        # March over the squashed version of the program, accumulating
223        # the line numbers of non-continued stmts, and determining
224        # whether & why the last stmt is a continuation.
225        continuation = C_NONE
226        level = lno = 0     # level is nesting level; lno is line number
227        self.goodlines = goodlines = [0]
228        push_good = goodlines.append
229        i, n = 0, len(code)
230        while i < n:
231            ch = code[i]
232            i = i+1
233
234            # cases are checked in decreasing order of frequency
235            if ch == 'x':
236                continue
237
238            if ch == '\n':
239                lno = lno + 1
240                if level == 0:
241                    push_good(lno)
242                    # else we're in an unclosed bracket structure
243                continue
244
245            if ch == '(':
246                level = level + 1
247                continue
248
249            if ch == ')':
250                if level:
251                    level = level - 1
252                    # else the program is invalid, but we can't complain
253                continue
254
255            if ch == '"' or ch == "'":
256                # consume the string
257                quote = ch
258                if code[i-1:i+2] == quote * 3:
259                    quote = quote * 3
260                firstlno = lno
261                w = len(quote) - 1
262                i = i+w
263                while i < n:
264                    ch = code[i]
265                    i = i+1
266
267                    if ch == 'x':
268                        continue
269
270                    if code[i-1:i+w] == quote:
271                        i = i+w
272                        break
273
274                    if ch == '\n':
275                        lno = lno + 1
276                        if w == 0:
277                            # unterminated single-quoted string
278                            if level == 0:
279                                push_good(lno)
280                            break
281                        continue
282
283                    if ch == '\\':
284                        assert i < n
285                        if code[i] == '\n':
286                            lno = lno + 1
287                        i = i+1
288                        continue
289
290                    # else comment char or paren inside string
291
292                else:
293                    # didn't break out of the loop, so we're still
294                    # inside a string
295                    if (lno - 1) == firstlno:
296                        # before the previous \n in code, we were in the first
297                        # line of the string
298                        continuation = C_STRING_FIRST_LINE
299                    else:
300                        continuation = C_STRING_NEXT_LINES
301                continue    # with outer loop
302
303            if ch == '#':
304                # consume the comment
305                i = code.find('\n', i)
306                assert i >= 0
307                continue
308
309            assert ch == '\\'
310            assert i < n
311            if code[i] == '\n':
312                lno = lno + 1
313                if i+1 == n:
314                    continuation = C_BACKSLASH
315            i = i+1
316
317        # The last stmt may be continued for all 3 reasons.
318        # String continuation takes precedence over bracket
319        # continuation, which beats backslash continuation.
320        if (continuation != C_STRING_FIRST_LINE
321            and continuation != C_STRING_NEXT_LINES and level > 0):
322            continuation = C_BRACKET
323        self.continuation = continuation
324
325        # Push the final line number as a sentinel value, regardless of
326        # whether it's continued.
327        assert (continuation == C_NONE) == (goodlines[-1] == lno)
328        if goodlines[-1] != lno:
329            push_good(lno)
330
331    def get_continuation_type(self):
332        self._study1()
333        return self.continuation
334
335    def _study2(self):
336        """
337        study1 was sufficient to determine the continuation status,
338        but doing more requires looking at every character.  study2
339        does this for the last interesting statement in the block.
340        Creates:
341            self.stmt_start, stmt_end
342                slice indices of last interesting stmt
343            self.stmt_bracketing
344                the bracketing structure of the last interesting stmt; for
345                example, for the statement "say(boo) or die",
346                stmt_bracketing will be ((0, 0), (0, 1), (2, 0), (2, 1),
347                (4, 0)). Strings and comments are treated as brackets, for
348                the matter.
349            self.lastch
350                last interesting character before optional trailing comment
351            self.lastopenbracketpos
352                if continuation is C_BRACKET, index of last open bracket
353        """
354        if self.study_level >= 2:
355            return
356        self._study1()
357        self.study_level = 2
358
359        # Set p and q to slice indices of last interesting stmt.
360        code, goodlines = self.code, self.goodlines
361        i = len(goodlines) - 1  # Index of newest line.
362        p = len(code)  # End of goodlines[i]
363        while i:
364            assert p
365            # Make p be the index of the stmt at line number goodlines[i].
366            # Move p back to the stmt at line number goodlines[i-1].
367            q = p
368            for nothing in range(goodlines[i-1], goodlines[i]):
369                # tricky: sets p to 0 if no preceding newline
370                p = code.rfind('\n', 0, p-1) + 1
371            # The stmt code[p:q] isn't a continuation, but may be blank
372            # or a non-indenting comment line.
373            if  _junkre(code, p):
374                i = i-1
375            else:
376                break
377        if i == 0:
378            # nothing but junk!
379            assert p == 0
380            q = p
381        self.stmt_start, self.stmt_end = p, q
382
383        # Analyze this stmt, to find the last open bracket (if any)
384        # and last interesting character (if any).
385        lastch = ""
386        stack = []  # stack of open bracket indices
387        push_stack = stack.append
388        bracketing = [(p, 0)]
389        while p < q:
390            # suck up all except ()[]{}'"#\\
391            m = _chew_ordinaryre(code, p, q)
392            if m:
393                # we skipped at least one boring char
394                newp = m.end()
395                # back up over totally boring whitespace
396                i = newp - 1    # index of last boring char
397                while i >= p and code[i] in " \t\n":
398                    i = i-1
399                if i >= p:
400                    lastch = code[i]
401                p = newp
402                if p >= q:
403                    break
404
405            ch = code[p]
406
407            if ch in "([{":
408                push_stack(p)
409                bracketing.append((p, len(stack)))
410                lastch = ch
411                p = p+1
412                continue
413
414            if ch in ")]}":
415                if stack:
416                    del stack[-1]
417                lastch = ch
418                p = p+1
419                bracketing.append((p, len(stack)))
420                continue
421
422            if ch == '"' or ch == "'":
423                # consume string
424                # Note that study1 did this with a Python loop, but
425                # we use a regexp here; the reason is speed in both
426                # cases; the string may be huge, but study1 pre-squashed
427                # strings to a couple of characters per line.  study1
428                # also needed to keep track of newlines, and we don't
429                # have to.
430                bracketing.append((p, len(stack)+1))
431                lastch = ch
432                p = _match_stringre(code, p, q).end()
433                bracketing.append((p, len(stack)))
434                continue
435
436            if ch == '#':
437                # consume comment and trailing newline
438                bracketing.append((p, len(stack)+1))
439                p = code.find('\n', p, q) + 1
440                assert p > 0
441                bracketing.append((p, len(stack)))
442                continue
443
444            assert ch == '\\'
445            p = p+1     # beyond backslash
446            assert p < q
447            if code[p] != '\n':
448                # the program is invalid, but can't complain
449                lastch = ch + code[p]
450            p = p+1     # beyond escaped char
451
452        # end while p < q:
453
454        self.lastch = lastch
455        self.lastopenbracketpos = stack[-1] if stack else None
456        self.stmt_bracketing = tuple(bracketing)
457
458    def compute_bracket_indent(self):
459        """Return number of spaces the next line should be indented.
460
461        Line continuation must be C_BRACKET.
462        """
463        self._study2()
464        assert self.continuation == C_BRACKET
465        j = self.lastopenbracketpos
466        code = self.code
467        n = len(code)
468        origi = i = code.rfind('\n', 0, j) + 1
469        j = j+1     # one beyond open bracket
470        # find first list item; set i to start of its line
471        while j < n:
472            m = _itemre(code, j)
473            if m:
474                j = m.end() - 1     # index of first interesting char
475                extra = 0
476                break
477            else:
478                # this line is junk; advance to next line
479                i = j = code.find('\n', j) + 1
480        else:
481            # nothing interesting follows the bracket;
482            # reproduce the bracket line's indentation + a level
483            j = i = origi
484            while code[j] in " \t":
485                j = j+1
486            extra = self.indentwidth
487        return len(code[i:j].expandtabs(self.tabwidth)) + extra
488
489    def get_num_lines_in_stmt(self):
490        """Return number of physical lines in last stmt.
491
492        The statement doesn't have to be an interesting statement.  This is
493        intended to be called when continuation is C_BACKSLASH.
494        """
495        self._study1()
496        goodlines = self.goodlines
497        return goodlines[-1] - goodlines[-2]
498
499    def compute_backslash_indent(self):
500        """Return number of spaces the next line should be indented.
501
502        Line continuation must be C_BACKSLASH.  Also assume that the new
503        line is the first one following the initial line of the stmt.
504        """
505        self._study2()
506        assert self.continuation == C_BACKSLASH
507        code = self.code
508        i = self.stmt_start
509        while code[i] in " \t":
510            i = i+1
511        startpos = i
512
513        # See whether the initial line starts an assignment stmt; i.e.,
514        # look for an = operator
515        endpos = code.find('\n', startpos) + 1
516        found = level = 0
517        while i < endpos:
518            ch = code[i]
519            if ch in "([{":
520                level = level + 1
521                i = i+1
522            elif ch in ")]}":
523                if level:
524                    level = level - 1
525                i = i+1
526            elif ch == '"' or ch == "'":
527                i = _match_stringre(code, i, endpos).end()
528            elif ch == '#':
529                # This line is unreachable because the # makes a comment of
530                # everything after it.
531                break
532            elif level == 0 and ch == '=' and \
533                   (i == 0 or code[i-1] not in "=<>!") and \
534                   code[i+1] != '=':
535                found = 1
536                break
537            else:
538                i = i+1
539
540        if found:
541            # found a legit =, but it may be the last interesting
542            # thing on the line
543            i = i+1     # move beyond the =
544            found = re.match(r"\s*\\", code[i:endpos]) is None
545
546        if not found:
547            # oh well ... settle for moving beyond the first chunk
548            # of non-whitespace chars
549            i = startpos
550            while code[i] not in " \t\n":
551                i = i+1
552
553        return len(code[self.stmt_start:i].expandtabs(\
554                                     self.tabwidth)) + 1
555
556    def get_base_indent_string(self):
557        """Return the leading whitespace on the initial line of the last
558        interesting stmt.
559        """
560        self._study2()
561        i, n = self.stmt_start, self.stmt_end
562        j = i
563        code = self.code
564        while j < n and code[j] in " \t":
565            j = j + 1
566        return code[i:j]
567
568    def is_block_opener(self):
569        "Return True if the last interesting statement opens a block."
570        self._study2()
571        return self.lastch == ':'
572
573    def is_block_closer(self):
574        "Return True if the last interesting statement closes a block."
575        self._study2()
576        return _closere(self.code, self.stmt_start) is not None
577
578    def get_last_stmt_bracketing(self):
579        """Return bracketing structure of the last interesting statement.
580
581        The returned tuple is in the format defined in _study2().
582        """
583        self._study2()
584        return self.stmt_bracketing
585
586
587if __name__ == '__main__':
588    from unittest import main
589    main('idlelib.idle_test.test_pyparse', verbosity=2)
590