Merge "Use signed variables in the lookahead."
[aom.git] / tools / cpplint.py
blob25fbef73d8edac6526704db76dacd5b7f70b65df
1 #!/usr/bin/python
3 # Copyright (c) 2009 Google Inc. All rights reserved.
5 # Redistribution and use in source and binary forms, with or without
6 # modification, are permitted provided that the following conditions are
7 # met:
9 # * Redistributions of source code must retain the above copyright
10 # notice, this list of conditions and the following disclaimer.
11 # * Redistributions in binary form must reproduce the above
12 # copyright notice, this list of conditions and the following disclaimer
13 # in the documentation and/or other materials provided with the
14 # distribution.
15 # * Neither the name of Google Inc. nor the names of its
16 # contributors may be used to endorse or promote products derived from
17 # this software without specific prior written permission.
19 # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 # "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 # LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 # A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
23 # OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
24 # SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
25 # LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 # DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
29 # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
31 """Does google-lint on c++ files.
33 The goal of this script is to identify places in the code that *may*
34 be in non-compliance with google style. It does not attempt to fix
35 up these problems -- the point is to educate. It does also not
36 attempt to find all problems, or to ensure that everything it does
37 find is legitimately a problem.
39 In particular, we can get very confused by /* and // inside strings!
40 We do a small hack, which is to ignore //'s with "'s after them on the
41 same line, but it is far from perfect (in either direction).
42 """
44 import codecs
45 import copy
46 import getopt
47 import math # for log
48 import os
49 import re
50 import sre_compile
51 import string
52 import sys
53 import unicodedata
56 _USAGE = """
57 Syntax: cpplint.py [--verbose=#] [--output=vs7] [--filter=-x,+y,...]
58 [--counting=total|toplevel|detailed] [--root=subdir]
59 [--linelength=digits]
60 <file> [file] ...
62 The style guidelines this tries to follow are those in
63 http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml
65 Every problem is given a confidence score from 1-5, with 5 meaning we are
66 certain of the problem, and 1 meaning it could be a legitimate construct.
67 This will miss some errors, and is not a substitute for a code review.
69 To suppress false-positive errors of a certain category, add a
70 'NOLINT(category)' comment to the line. NOLINT or NOLINT(*)
71 suppresses errors of all categories on that line.
73 The files passed in will be linted; at least one file must be provided.
74 Default linted extensions are .cc, .cpp, .cu, .cuh and .h. Change the
75 extensions with the --extensions flag.
77 Flags:
79 output=vs7
80 By default, the output is formatted to ease emacs parsing. Visual Studio
81 compatible output (vs7) may also be used. Other formats are unsupported.
83 verbose=#
84 Specify a number 0-5 to restrict errors to certain verbosity levels.
86 filter=-x,+y,...
87 Specify a comma-separated list of category-filters to apply: only
88 error messages whose category names pass the filters will be printed.
89 (Category names are printed with the message and look like
90 "[whitespace/indent]".) Filters are evaluated left to right.
91 "-FOO" and "FOO" means "do not print categories that start with FOO".
92 "+FOO" means "do print categories that start with FOO".
94 Examples: --filter=-whitespace,+whitespace/braces
95 --filter=whitespace,runtime/printf,+runtime/printf_format
96 --filter=-,+build/include_what_you_use
98 To see a list of all the categories used in cpplint, pass no arg:
99 --filter=
101 counting=total|toplevel|detailed
102 The total number of errors found is always printed. If
103 'toplevel' is provided, then the count of errors in each of
104 the top-level categories like 'build' and 'whitespace' will
105 also be printed. If 'detailed' is provided, then a count
106 is provided for each category like 'build/class'.
108 root=subdir
109 The root directory used for deriving header guard CPP variable.
110 By default, the header guard CPP variable is calculated as the relative
111 path to the directory that contains .git, .hg, or .svn. When this flag
112 is specified, the relative path is calculated from the specified
113 directory. If the specified directory does not exist, this flag is
114 ignored.
116 Examples:
117 Assuing that src/.git exists, the header guard CPP variables for
118 src/chrome/browser/ui/browser.h are:
120 No flag => CHROME_BROWSER_UI_BROWSER_H_
121 --root=chrome => BROWSER_UI_BROWSER_H_
122 --root=chrome/browser => UI_BROWSER_H_
124 linelength=digits
125 This is the allowed line length for the project. The default value is
126 80 characters.
128 Examples:
129 --linelength=120
131 extensions=extension,extension,...
132 The allowed file extensions that cpplint will check
134 Examples:
135 --extensions=hpp,cpp
138 # We categorize each error message we print. Here are the categories.
139 # We want an explicit list so we can list them all in cpplint --filter=.
140 # If you add a new error message with a new category, add it to the list
141 # here! cpplint_unittest.py should tell you if you forget to do this.
142 _ERROR_CATEGORIES = [
143 'build/class',
144 'build/deprecated',
145 'build/endif_comment',
146 'build/explicit_make_pair',
147 'build/forward_decl',
148 'build/header_guard',
149 'build/include',
150 'build/include_alpha',
151 'build/include_order',
152 'build/include_what_you_use',
153 'build/namespaces',
154 'build/printf_format',
155 'build/storage_class',
156 'legal/copyright',
157 'readability/alt_tokens',
158 'readability/braces',
159 'readability/casting',
160 'readability/check',
161 'readability/constructors',
162 'readability/fn_size',
163 'readability/function',
164 'readability/multiline_comment',
165 'readability/multiline_string',
166 'readability/namespace',
167 'readability/nolint',
168 'readability/nul',
169 'readability/streams',
170 'readability/todo',
171 'readability/utf8',
172 'runtime/arrays',
173 'runtime/casting',
174 'runtime/explicit',
175 'runtime/int',
176 'runtime/init',
177 'runtime/invalid_increment',
178 'runtime/member_string_references',
179 'runtime/memset',
180 'runtime/operator',
181 'runtime/printf',
182 'runtime/printf_format',
183 'runtime/references',
184 'runtime/sizeof',
185 'runtime/string',
186 'runtime/threadsafe_fn',
187 'runtime/vlog',
188 'whitespace/blank_line',
189 'whitespace/braces',
190 'whitespace/comma',
191 'whitespace/comments',
192 'whitespace/empty_conditional_body',
193 'whitespace/empty_loop_body',
194 'whitespace/end_of_line',
195 'whitespace/ending_newline',
196 'whitespace/forcolon',
197 'whitespace/indent',
198 'whitespace/line_length',
199 'whitespace/newline',
200 'whitespace/operators',
201 'whitespace/parens',
202 'whitespace/semicolon',
203 'whitespace/tab',
204 'whitespace/todo'
207 # The default state of the category filter. This is overrided by the --filter=
208 # flag. By default all errors are on, so only add here categories that should be
209 # off by default (i.e., categories that must be enabled by the --filter= flags).
210 # All entries here should start with a '-' or '+', as in the --filter= flag.
211 _DEFAULT_FILTERS = ['-build/include_alpha']
213 # We used to check for high-bit characters, but after much discussion we
214 # decided those were OK, as long as they were in UTF-8 and didn't represent
215 # hard-coded international strings, which belong in a separate i18n file.
218 # C++ headers
219 _CPP_HEADERS = frozenset([
220 # Legacy
221 'algobase.h',
222 'algo.h',
223 'alloc.h',
224 'builtinbuf.h',
225 'bvector.h',
226 'complex.h',
227 'defalloc.h',
228 'deque.h',
229 'editbuf.h',
230 'fstream.h',
231 'function.h',
232 'hash_map',
233 'hash_map.h',
234 'hash_set',
235 'hash_set.h',
236 'hashtable.h',
237 'heap.h',
238 'indstream.h',
239 'iomanip.h',
240 'iostream.h',
241 'istream.h',
242 'iterator.h',
243 'list.h',
244 'map.h',
245 'multimap.h',
246 'multiset.h',
247 'ostream.h',
248 'pair.h',
249 'parsestream.h',
250 'pfstream.h',
251 'procbuf.h',
252 'pthread_alloc',
253 'pthread_alloc.h',
254 'rope',
255 'rope.h',
256 'ropeimpl.h',
257 'set.h',
258 'slist',
259 'slist.h',
260 'stack.h',
261 'stdiostream.h',
262 'stl_alloc.h',
263 'stl_relops.h',
264 'streambuf.h',
265 'stream.h',
266 'strfile.h',
267 'strstream.h',
268 'tempbuf.h',
269 'tree.h',
270 'type_traits.h',
271 'vector.h',
272 # 17.6.1.2 C++ library headers
273 'algorithm',
274 'array',
275 'atomic',
276 'bitset',
277 'chrono',
278 'codecvt',
279 'complex',
280 'condition_variable',
281 'deque',
282 'exception',
283 'forward_list',
284 'fstream',
285 'functional',
286 'future',
287 'initializer_list',
288 'iomanip',
289 'ios',
290 'iosfwd',
291 'iostream',
292 'istream',
293 'iterator',
294 'limits',
295 'list',
296 'locale',
297 'map',
298 'memory',
299 'mutex',
300 'new',
301 'numeric',
302 'ostream',
303 'queue',
304 'random',
305 'ratio',
306 'regex',
307 'set',
308 'sstream',
309 'stack',
310 'stdexcept',
311 'streambuf',
312 'string',
313 'strstream',
314 'system_error',
315 'thread',
316 'tuple',
317 'typeindex',
318 'typeinfo',
319 'type_traits',
320 'unordered_map',
321 'unordered_set',
322 'utility',
323 'valarray',
324 'vector',
325 # 17.6.1.2 C++ headers for C library facilities
326 'cassert',
327 'ccomplex',
328 'cctype',
329 'cerrno',
330 'cfenv',
331 'cfloat',
332 'cinttypes',
333 'ciso646',
334 'climits',
335 'clocale',
336 'cmath',
337 'csetjmp',
338 'csignal',
339 'cstdalign',
340 'cstdarg',
341 'cstdbool',
342 'cstddef',
343 'cstdint',
344 'cstdio',
345 'cstdlib',
346 'cstring',
347 'ctgmath',
348 'ctime',
349 'cuchar',
350 'cwchar',
351 'cwctype',
354 # Assertion macros. These are defined in base/logging.h and
355 # testing/base/gunit.h. Note that the _M versions need to come first
356 # for substring matching to work.
357 _CHECK_MACROS = [
358 'DCHECK', 'CHECK',
359 'EXPECT_TRUE_M', 'EXPECT_TRUE',
360 'ASSERT_TRUE_M', 'ASSERT_TRUE',
361 'EXPECT_FALSE_M', 'EXPECT_FALSE',
362 'ASSERT_FALSE_M', 'ASSERT_FALSE',
365 # Replacement macros for CHECK/DCHECK/EXPECT_TRUE/EXPECT_FALSE
366 _CHECK_REPLACEMENT = dict([(m, {}) for m in _CHECK_MACROS])
368 for op, replacement in [('==', 'EQ'), ('!=', 'NE'),
369 ('>=', 'GE'), ('>', 'GT'),
370 ('<=', 'LE'), ('<', 'LT')]:
371 _CHECK_REPLACEMENT['DCHECK'][op] = 'DCHECK_%s' % replacement
372 _CHECK_REPLACEMENT['CHECK'][op] = 'CHECK_%s' % replacement
373 _CHECK_REPLACEMENT['EXPECT_TRUE'][op] = 'EXPECT_%s' % replacement
374 _CHECK_REPLACEMENT['ASSERT_TRUE'][op] = 'ASSERT_%s' % replacement
375 _CHECK_REPLACEMENT['EXPECT_TRUE_M'][op] = 'EXPECT_%s_M' % replacement
376 _CHECK_REPLACEMENT['ASSERT_TRUE_M'][op] = 'ASSERT_%s_M' % replacement
378 for op, inv_replacement in [('==', 'NE'), ('!=', 'EQ'),
379 ('>=', 'LT'), ('>', 'LE'),
380 ('<=', 'GT'), ('<', 'GE')]:
381 _CHECK_REPLACEMENT['EXPECT_FALSE'][op] = 'EXPECT_%s' % inv_replacement
382 _CHECK_REPLACEMENT['ASSERT_FALSE'][op] = 'ASSERT_%s' % inv_replacement
383 _CHECK_REPLACEMENT['EXPECT_FALSE_M'][op] = 'EXPECT_%s_M' % inv_replacement
384 _CHECK_REPLACEMENT['ASSERT_FALSE_M'][op] = 'ASSERT_%s_M' % inv_replacement
386 # Alternative tokens and their replacements. For full list, see section 2.5
387 # Alternative tokens [lex.digraph] in the C++ standard.
389 # Digraphs (such as '%:') are not included here since it's a mess to
390 # match those on a word boundary.
391 _ALT_TOKEN_REPLACEMENT = {
392 'and': '&&',
393 'bitor': '|',
394 'or': '||',
395 'xor': '^',
396 'compl': '~',
397 'bitand': '&',
398 'and_eq': '&=',
399 'or_eq': '|=',
400 'xor_eq': '^=',
401 'not': '!',
402 'not_eq': '!='
405 # Compile regular expression that matches all the above keywords. The "[ =()]"
406 # bit is meant to avoid matching these keywords outside of boolean expressions.
408 # False positives include C-style multi-line comments and multi-line strings
409 # but those have always been troublesome for cpplint.
410 _ALT_TOKEN_REPLACEMENT_PATTERN = re.compile(
411 r'[ =()](' + ('|'.join(_ALT_TOKEN_REPLACEMENT.keys())) + r')(?=[ (]|$)')
414 # These constants define types of headers for use with
415 # _IncludeState.CheckNextIncludeOrder().
416 _C_SYS_HEADER = 1
417 _CPP_SYS_HEADER = 2
418 _LIKELY_MY_HEADER = 3
419 _POSSIBLE_MY_HEADER = 4
420 _OTHER_HEADER = 5
422 # These constants define the current inline assembly state
423 _NO_ASM = 0 # Outside of inline assembly block
424 _INSIDE_ASM = 1 # Inside inline assembly block
425 _END_ASM = 2 # Last line of inline assembly block
426 _BLOCK_ASM = 3 # The whole block is an inline assembly block
428 # Match start of assembly blocks
429 _MATCH_ASM = re.compile(r'^\s*(?:asm|_asm|__asm|__asm__)'
430 r'(?:\s+(volatile|__volatile__))?'
431 r'\s*[{(]')
434 _regexp_compile_cache = {}
436 # Finds occurrences of NOLINT or NOLINT(...).
437 _RE_SUPPRESSION = re.compile(r'\bNOLINT\b(\([^)]*\))?')
439 # {str, set(int)}: a map from error categories to sets of linenumbers
440 # on which those errors are expected and should be suppressed.
441 _error_suppressions = {}
443 # The root directory used for deriving header guard CPP variable.
444 # This is set by --root flag.
445 _root = None
447 # The allowed line length of files.
448 # This is set by --linelength flag.
449 _line_length = 80
451 # The allowed extensions for file names
452 # This is set by --extensions flag.
453 _valid_extensions = set(['cc', 'h', 'cpp', 'cu', 'cuh'])
455 def ParseNolintSuppressions(filename, raw_line, linenum, error):
456 """Updates the global list of error-suppressions.
458 Parses any NOLINT comments on the current line, updating the global
459 error_suppressions store. Reports an error if the NOLINT comment
460 was malformed.
462 Args:
463 filename: str, the name of the input file.
464 raw_line: str, the line of input text, with comments.
465 linenum: int, the number of the current line.
466 error: function, an error handler.
468 # FIXME(adonovan): "NOLINT(" is misparsed as NOLINT(*).
469 matched = _RE_SUPPRESSION.search(raw_line)
470 if matched:
471 category = matched.group(1)
472 if category in (None, '(*)'): # => "suppress all"
473 _error_suppressions.setdefault(None, set()).add(linenum)
474 else:
475 if category.startswith('(') and category.endswith(')'):
476 category = category[1:-1]
477 if category in _ERROR_CATEGORIES:
478 _error_suppressions.setdefault(category, set()).add(linenum)
479 else:
480 error(filename, linenum, 'readability/nolint', 5,
481 'Unknown NOLINT error category: %s' % category)
484 def ResetNolintSuppressions():
485 "Resets the set of NOLINT suppressions to empty."
486 _error_suppressions.clear()
489 def IsErrorSuppressedByNolint(category, linenum):
490 """Returns true if the specified error category is suppressed on this line.
492 Consults the global error_suppressions map populated by
493 ParseNolintSuppressions/ResetNolintSuppressions.
495 Args:
496 category: str, the category of the error.
497 linenum: int, the current line number.
498 Returns:
499 bool, True iff the error should be suppressed due to a NOLINT comment.
501 return (linenum in _error_suppressions.get(category, set()) or
502 linenum in _error_suppressions.get(None, set()))
504 def Match(pattern, s):
505 """Matches the string with the pattern, caching the compiled regexp."""
506 # The regexp compilation caching is inlined in both Match and Search for
507 # performance reasons; factoring it out into a separate function turns out
508 # to be noticeably expensive.
509 if pattern not in _regexp_compile_cache:
510 _regexp_compile_cache[pattern] = sre_compile.compile(pattern)
511 return _regexp_compile_cache[pattern].match(s)
514 def ReplaceAll(pattern, rep, s):
515 """Replaces instances of pattern in a string with a replacement.
517 The compiled regex is kept in a cache shared by Match and Search.
519 Args:
520 pattern: regex pattern
521 rep: replacement text
522 s: search string
524 Returns:
525 string with replacements made (or original string if no replacements)
527 if pattern not in _regexp_compile_cache:
528 _regexp_compile_cache[pattern] = sre_compile.compile(pattern)
529 return _regexp_compile_cache[pattern].sub(rep, s)
532 def Search(pattern, s):
533 """Searches the string for the pattern, caching the compiled regexp."""
534 if pattern not in _regexp_compile_cache:
535 _regexp_compile_cache[pattern] = sre_compile.compile(pattern)
536 return _regexp_compile_cache[pattern].search(s)
539 class _IncludeState(dict):
540 """Tracks line numbers for includes, and the order in which includes appear.
542 As a dict, an _IncludeState object serves as a mapping between include
543 filename and line number on which that file was included.
545 Call CheckNextIncludeOrder() once for each header in the file, passing
546 in the type constants defined above. Calls in an illegal order will
547 raise an _IncludeError with an appropriate error message.
550 # self._section will move monotonically through this set. If it ever
551 # needs to move backwards, CheckNextIncludeOrder will raise an error.
552 _INITIAL_SECTION = 0
553 _MY_H_SECTION = 1
554 _C_SECTION = 2
555 _CPP_SECTION = 3
556 _OTHER_H_SECTION = 4
558 _TYPE_NAMES = {
559 _C_SYS_HEADER: 'C system header',
560 _CPP_SYS_HEADER: 'C++ system header',
561 _LIKELY_MY_HEADER: 'header this file implements',
562 _POSSIBLE_MY_HEADER: 'header this file may implement',
563 _OTHER_HEADER: 'other header',
565 _SECTION_NAMES = {
566 _INITIAL_SECTION: "... nothing. (This can't be an error.)",
567 _MY_H_SECTION: 'a header this file implements',
568 _C_SECTION: 'C system header',
569 _CPP_SECTION: 'C++ system header',
570 _OTHER_H_SECTION: 'other header',
573 def __init__(self):
574 dict.__init__(self)
575 self.ResetSection()
577 def ResetSection(self):
578 # The name of the current section.
579 self._section = self._INITIAL_SECTION
580 # The path of last found header.
581 self._last_header = ''
583 def SetLastHeader(self, header_path):
584 self._last_header = header_path
586 def CanonicalizeAlphabeticalOrder(self, header_path):
587 """Returns a path canonicalized for alphabetical comparison.
589 - replaces "-" with "_" so they both cmp the same.
590 - removes '-inl' since we don't require them to be after the main header.
591 - lowercase everything, just in case.
593 Args:
594 header_path: Path to be canonicalized.
596 Returns:
597 Canonicalized path.
599 return header_path.replace('-inl.h', '.h').replace('-', '_').lower()
601 def IsInAlphabeticalOrder(self, clean_lines, linenum, header_path):
602 """Check if a header is in alphabetical order with the previous header.
604 Args:
605 clean_lines: A CleansedLines instance containing the file.
606 linenum: The number of the line to check.
607 header_path: Canonicalized header to be checked.
609 Returns:
610 Returns true if the header is in alphabetical order.
612 # If previous section is different from current section, _last_header will
613 # be reset to empty string, so it's always less than current header.
615 # If previous line was a blank line, assume that the headers are
616 # intentionally sorted the way they are.
617 if (self._last_header > header_path and
618 not Match(r'^\s*$', clean_lines.elided[linenum - 1])):
619 return False
620 return True
622 def CheckNextIncludeOrder(self, header_type):
623 """Returns a non-empty error message if the next header is out of order.
625 This function also updates the internal state to be ready to check
626 the next include.
628 Args:
629 header_type: One of the _XXX_HEADER constants defined above.
631 Returns:
632 The empty string if the header is in the right order, or an
633 error message describing what's wrong.
636 error_message = ('Found %s after %s' %
637 (self._TYPE_NAMES[header_type],
638 self._SECTION_NAMES[self._section]))
640 last_section = self._section
642 if header_type == _C_SYS_HEADER:
643 if self._section <= self._C_SECTION:
644 self._section = self._C_SECTION
645 else:
646 self._last_header = ''
647 return error_message
648 elif header_type == _CPP_SYS_HEADER:
649 if self._section <= self._CPP_SECTION:
650 self._section = self._CPP_SECTION
651 else:
652 self._last_header = ''
653 return error_message
654 elif header_type == _LIKELY_MY_HEADER:
655 if self._section <= self._MY_H_SECTION:
656 self._section = self._MY_H_SECTION
657 else:
658 self._section = self._OTHER_H_SECTION
659 elif header_type == _POSSIBLE_MY_HEADER:
660 if self._section <= self._MY_H_SECTION:
661 self._section = self._MY_H_SECTION
662 else:
663 # This will always be the fallback because we're not sure
664 # enough that the header is associated with this file.
665 self._section = self._OTHER_H_SECTION
666 else:
667 assert header_type == _OTHER_HEADER
668 self._section = self._OTHER_H_SECTION
670 if last_section != self._section:
671 self._last_header = ''
673 return ''
676 class _CppLintState(object):
677 """Maintains module-wide state.."""
679 def __init__(self):
680 self.verbose_level = 1 # global setting.
681 self.error_count = 0 # global count of reported errors
682 # filters to apply when emitting error messages
683 self.filters = _DEFAULT_FILTERS[:]
684 self.counting = 'total' # In what way are we counting errors?
685 self.errors_by_category = {} # string to int dict storing error counts
687 # output format:
688 # "emacs" - format that emacs can parse (default)
689 # "vs7" - format that Microsoft Visual Studio 7 can parse
690 self.output_format = 'emacs'
692 def SetOutputFormat(self, output_format):
693 """Sets the output format for errors."""
694 self.output_format = output_format
696 def SetVerboseLevel(self, level):
697 """Sets the module's verbosity, and returns the previous setting."""
698 last_verbose_level = self.verbose_level
699 self.verbose_level = level
700 return last_verbose_level
702 def SetCountingStyle(self, counting_style):
703 """Sets the module's counting options."""
704 self.counting = counting_style
706 def SetFilters(self, filters):
707 """Sets the error-message filters.
709 These filters are applied when deciding whether to emit a given
710 error message.
712 Args:
713 filters: A string of comma-separated filters (eg "+whitespace/indent").
714 Each filter should start with + or -; else we die.
716 Raises:
717 ValueError: The comma-separated filters did not all start with '+' or '-'.
718 E.g. "-,+whitespace,-whitespace/indent,whitespace/badfilter"
720 # Default filters always have less priority than the flag ones.
721 self.filters = _DEFAULT_FILTERS[:]
722 for filt in filters.split(','):
723 clean_filt = filt.strip()
724 if clean_filt:
725 self.filters.append(clean_filt)
726 for filt in self.filters:
727 if not (filt.startswith('+') or filt.startswith('-')):
728 raise ValueError('Every filter in --filters must start with + or -'
729 ' (%s does not)' % filt)
731 def ResetErrorCounts(self):
732 """Sets the module's error statistic back to zero."""
733 self.error_count = 0
734 self.errors_by_category = {}
736 def IncrementErrorCount(self, category):
737 """Bumps the module's error statistic."""
738 self.error_count += 1
739 if self.counting in ('toplevel', 'detailed'):
740 if self.counting != 'detailed':
741 category = category.split('/')[0]
742 if category not in self.errors_by_category:
743 self.errors_by_category[category] = 0
744 self.errors_by_category[category] += 1
746 def PrintErrorCounts(self):
747 """Print a summary of errors by category, and the total."""
748 for category, count in self.errors_by_category.iteritems():
749 sys.stderr.write('Category \'%s\' errors found: %d\n' %
750 (category, count))
751 sys.stderr.write('Total errors found: %d\n' % self.error_count)
753 _cpplint_state = _CppLintState()
756 def _OutputFormat():
757 """Gets the module's output format."""
758 return _cpplint_state.output_format
761 def _SetOutputFormat(output_format):
762 """Sets the module's output format."""
763 _cpplint_state.SetOutputFormat(output_format)
766 def _VerboseLevel():
767 """Returns the module's verbosity setting."""
768 return _cpplint_state.verbose_level
771 def _SetVerboseLevel(level):
772 """Sets the module's verbosity, and returns the previous setting."""
773 return _cpplint_state.SetVerboseLevel(level)
776 def _SetCountingStyle(level):
777 """Sets the module's counting options."""
778 _cpplint_state.SetCountingStyle(level)
781 def _Filters():
782 """Returns the module's list of output filters, as a list."""
783 return _cpplint_state.filters
786 def _SetFilters(filters):
787 """Sets the module's error-message filters.
789 These filters are applied when deciding whether to emit a given
790 error message.
792 Args:
793 filters: A string of comma-separated filters (eg "whitespace/indent").
794 Each filter should start with + or -; else we die.
796 _cpplint_state.SetFilters(filters)
799 class _FunctionState(object):
800 """Tracks current function name and the number of lines in its body."""
802 _NORMAL_TRIGGER = 250 # for --v=0, 500 for --v=1, etc.
803 _TEST_TRIGGER = 400 # about 50% more than _NORMAL_TRIGGER.
805 def __init__(self):
806 self.in_a_function = False
807 self.lines_in_function = 0
808 self.current_function = ''
810 def Begin(self, function_name):
811 """Start analyzing function body.
813 Args:
814 function_name: The name of the function being tracked.
816 self.in_a_function = True
817 self.lines_in_function = 0
818 self.current_function = function_name
820 def Count(self):
821 """Count line in current function body."""
822 if self.in_a_function:
823 self.lines_in_function += 1
825 def Check(self, error, filename, linenum):
826 """Report if too many lines in function body.
828 Args:
829 error: The function to call with any errors found.
830 filename: The name of the current file.
831 linenum: The number of the line to check.
833 if Match(r'T(EST|est)', self.current_function):
834 base_trigger = self._TEST_TRIGGER
835 else:
836 base_trigger = self._NORMAL_TRIGGER
837 trigger = base_trigger * 2**_VerboseLevel()
839 if self.lines_in_function > trigger:
840 error_level = int(math.log(self.lines_in_function / base_trigger, 2))
841 # 50 => 0, 100 => 1, 200 => 2, 400 => 3, 800 => 4, 1600 => 5, ...
842 if error_level > 5:
843 error_level = 5
844 error(filename, linenum, 'readability/fn_size', error_level,
845 'Small and focused functions are preferred:'
846 ' %s has %d non-comment lines'
847 ' (error triggered by exceeding %d lines).' % (
848 self.current_function, self.lines_in_function, trigger))
850 def End(self):
851 """Stop analyzing function body."""
852 self.in_a_function = False
855 class _IncludeError(Exception):
856 """Indicates a problem with the include order in a file."""
857 pass
860 class FileInfo:
861 """Provides utility functions for filenames.
863 FileInfo provides easy access to the components of a file's path
864 relative to the project root.
867 def __init__(self, filename):
868 self._filename = filename
870 def FullName(self):
871 """Make Windows paths like Unix."""
872 return os.path.abspath(self._filename).replace('\\', '/')
874 def RepositoryName(self):
875 """FullName after removing the local path to the repository.
877 If we have a real absolute path name here we can try to do something smart:
878 detecting the root of the checkout and truncating /path/to/checkout from
879 the name so that we get header guards that don't include things like
880 "C:\Documents and Settings\..." or "/home/username/..." in them and thus
881 people on different computers who have checked the source out to different
882 locations won't see bogus errors.
884 fullname = self.FullName()
886 if os.path.exists(fullname):
887 project_dir = os.path.dirname(fullname)
889 if os.path.exists(os.path.join(project_dir, ".svn")):
890 # If there's a .svn file in the current directory, we recursively look
891 # up the directory tree for the top of the SVN checkout
892 root_dir = project_dir
893 one_up_dir = os.path.dirname(root_dir)
894 while os.path.exists(os.path.join(one_up_dir, ".svn")):
895 root_dir = os.path.dirname(root_dir)
896 one_up_dir = os.path.dirname(one_up_dir)
898 prefix = os.path.commonprefix([root_dir, project_dir])
899 return fullname[len(prefix) + 1:]
901 # Not SVN <= 1.6? Try to find a git, hg, or svn top level directory by
902 # searching up from the current path.
903 root_dir = os.path.dirname(fullname)
904 while (root_dir != os.path.dirname(root_dir) and
905 not os.path.exists(os.path.join(root_dir, ".git")) and
906 not os.path.exists(os.path.join(root_dir, ".hg")) and
907 not os.path.exists(os.path.join(root_dir, ".svn"))):
908 root_dir = os.path.dirname(root_dir)
910 if (os.path.exists(os.path.join(root_dir, ".git")) or
911 os.path.exists(os.path.join(root_dir, ".hg")) or
912 os.path.exists(os.path.join(root_dir, ".svn"))):
913 prefix = os.path.commonprefix([root_dir, project_dir])
914 return fullname[len(prefix) + 1:]
916 # Don't know what to do; header guard warnings may be wrong...
917 return fullname
919 def Split(self):
920 """Splits the file into the directory, basename, and extension.
922 For 'chrome/browser/browser.cc', Split() would
923 return ('chrome/browser', 'browser', '.cc')
925 Returns:
926 A tuple of (directory, basename, extension).
929 googlename = self.RepositoryName()
930 project, rest = os.path.split(googlename)
931 return (project,) + os.path.splitext(rest)
933 def BaseName(self):
934 """File base name - text after the final slash, before the final period."""
935 return self.Split()[1]
937 def Extension(self):
938 """File extension - text following the final period."""
939 return self.Split()[2]
941 def NoExtension(self):
942 """File has no source file extension."""
943 return '/'.join(self.Split()[0:2])
945 def IsSource(self):
946 """File has a source file extension."""
947 return self.Extension()[1:] in ('c', 'cc', 'cpp', 'cxx')
950 def _ShouldPrintError(category, confidence, linenum):
951 """If confidence >= verbose, category passes filter and is not suppressed."""
953 # There are three ways we might decide not to print an error message:
954 # a "NOLINT(category)" comment appears in the source,
955 # the verbosity level isn't high enough, or the filters filter it out.
956 if IsErrorSuppressedByNolint(category, linenum):
957 return False
958 if confidence < _cpplint_state.verbose_level:
959 return False
961 is_filtered = False
962 for one_filter in _Filters():
963 if one_filter.startswith('-'):
964 if category.startswith(one_filter[1:]):
965 is_filtered = True
966 elif one_filter.startswith('+'):
967 if category.startswith(one_filter[1:]):
968 is_filtered = False
969 else:
970 assert False # should have been checked for in SetFilter.
971 if is_filtered:
972 return False
974 return True
977 def Error(filename, linenum, category, confidence, message):
978 """Logs the fact we've found a lint error.
980 We log where the error was found, and also our confidence in the error,
981 that is, how certain we are this is a legitimate style regression, and
982 not a misidentification or a use that's sometimes justified.
984 False positives can be suppressed by the use of
985 "cpplint(category)" comments on the offending line. These are
986 parsed into _error_suppressions.
988 Args:
989 filename: The name of the file containing the error.
990 linenum: The number of the line containing the error.
991 category: A string used to describe the "category" this bug
992 falls under: "whitespace", say, or "runtime". Categories
993 may have a hierarchy separated by slashes: "whitespace/indent".
994 confidence: A number from 1-5 representing a confidence score for
995 the error, with 5 meaning that we are certain of the problem,
996 and 1 meaning that it could be a legitimate construct.
997 message: The error message.
999 if _ShouldPrintError(category, confidence, linenum):
1000 _cpplint_state.IncrementErrorCount(category)
1001 if _cpplint_state.output_format == 'vs7':
1002 sys.stderr.write('%s(%s): %s [%s] [%d]\n' % (
1003 filename, linenum, message, category, confidence))
1004 elif _cpplint_state.output_format == 'eclipse':
1005 sys.stderr.write('%s:%s: warning: %s [%s] [%d]\n' % (
1006 filename, linenum, message, category, confidence))
1007 else:
1008 sys.stderr.write('%s:%s: %s [%s] [%d]\n' % (
1009 filename, linenum, message, category, confidence))
1012 # Matches standard C++ escape sequences per 2.13.2.3 of the C++ standard.
1013 _RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile(
1014 r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)')
1015 # Matches strings. Escape codes should already be removed by ESCAPES.
1016 _RE_PATTERN_CLEANSE_LINE_DOUBLE_QUOTES = re.compile(r'"[^"]*"')
1017 # Matches characters. Escape codes should already be removed by ESCAPES.
1018 _RE_PATTERN_CLEANSE_LINE_SINGLE_QUOTES = re.compile(r"'.'")
1019 # Matches multi-line C++ comments.
1020 # This RE is a little bit more complicated than one might expect, because we
1021 # have to take care of space removals tools so we can handle comments inside
1022 # statements better.
1023 # The current rule is: We only clear spaces from both sides when we're at the
1024 # end of the line. Otherwise, we try to remove spaces from the right side,
1025 # if this doesn't work we try on left side but only if there's a non-character
1026 # on the right.
1027 _RE_PATTERN_CLEANSE_LINE_C_COMMENTS = re.compile(
1028 r"""(\s*/\*.*\*/\s*$|
1029 /\*.*\*/\s+|
1030 \s+/\*.*\*/(?=\W)|
1031 /\*.*\*/)""", re.VERBOSE)
1034 def IsCppString(line):
1035 """Does line terminate so, that the next symbol is in string constant.
1037 This function does not consider single-line nor multi-line comments.
1039 Args:
1040 line: is a partial line of code starting from the 0..n.
1042 Returns:
1043 True, if next character appended to 'line' is inside a
1044 string constant.
1047 line = line.replace(r'\\', 'XX') # after this, \\" does not match to \"
1048 return ((line.count('"') - line.count(r'\"') - line.count("'\"'")) & 1) == 1
1051 def CleanseRawStrings(raw_lines):
1052 """Removes C++11 raw strings from lines.
1054 Before:
1055 static const char kData[] = R"(
1056 multi-line string
1059 After:
1060 static const char kData[] = ""
1061 (replaced by blank line)
1064 Args:
1065 raw_lines: list of raw lines.
1067 Returns:
1068 list of lines with C++11 raw strings replaced by empty strings.
1071 delimiter = None
1072 lines_without_raw_strings = []
1073 for line in raw_lines:
1074 if delimiter:
1075 # Inside a raw string, look for the end
1076 end = line.find(delimiter)
1077 if end >= 0:
1078 # Found the end of the string, match leading space for this
1079 # line and resume copying the original lines, and also insert
1080 # a "" on the last line.
1081 leading_space = Match(r'^(\s*)\S', line)
1082 line = leading_space.group(1) + '""' + line[end + len(delimiter):]
1083 delimiter = None
1084 else:
1085 # Haven't found the end yet, append a blank line.
1086 line = ''
1088 else:
1089 # Look for beginning of a raw string.
1090 # See 2.14.15 [lex.string] for syntax.
1091 matched = Match(r'^(.*)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line)
1092 if matched:
1093 delimiter = ')' + matched.group(2) + '"'
1095 end = matched.group(3).find(delimiter)
1096 if end >= 0:
1097 # Raw string ended on same line
1098 line = (matched.group(1) + '""' +
1099 matched.group(3)[end + len(delimiter):])
1100 delimiter = None
1101 else:
1102 # Start of a multi-line raw string
1103 line = matched.group(1) + '""'
1105 lines_without_raw_strings.append(line)
1107 # TODO(unknown): if delimiter is not None here, we might want to
1108 # emit a warning for unterminated string.
1109 return lines_without_raw_strings
1112 def FindNextMultiLineCommentStart(lines, lineix):
1113 """Find the beginning marker for a multiline comment."""
1114 while lineix < len(lines):
1115 if lines[lineix].strip().startswith('/*'):
1116 # Only return this marker if the comment goes beyond this line
1117 if lines[lineix].strip().find('*/', 2) < 0:
1118 return lineix
1119 lineix += 1
1120 return len(lines)
1123 def FindNextMultiLineCommentEnd(lines, lineix):
1124 """We are inside a comment, find the end marker."""
1125 while lineix < len(lines):
1126 if lines[lineix].strip().endswith('*/'):
1127 return lineix
1128 lineix += 1
1129 return len(lines)
1132 def RemoveMultiLineCommentsFromRange(lines, begin, end):
1133 """Clears a range of lines for multi-line comments."""
1134 # Having // dummy comments makes the lines non-empty, so we will not get
1135 # unnecessary blank line warnings later in the code.
1136 for i in range(begin, end):
1137 lines[i] = '// dummy'
1140 def RemoveMultiLineComments(filename, lines, error):
1141 """Removes multiline (c-style) comments from lines."""
1142 lineix = 0
1143 while lineix < len(lines):
1144 lineix_begin = FindNextMultiLineCommentStart(lines, lineix)
1145 if lineix_begin >= len(lines):
1146 return
1147 lineix_end = FindNextMultiLineCommentEnd(lines, lineix_begin)
1148 if lineix_end >= len(lines):
1149 error(filename, lineix_begin + 1, 'readability/multiline_comment', 5,
1150 'Could not find end of multi-line comment')
1151 return
1152 RemoveMultiLineCommentsFromRange(lines, lineix_begin, lineix_end + 1)
1153 lineix = lineix_end + 1
1156 def CleanseComments(line):
1157 """Removes //-comments and single-line C-style /* */ comments.
1159 Args:
1160 line: A line of C++ source.
1162 Returns:
1163 The line with single-line comments removed.
1165 commentpos = line.find('//')
1166 if commentpos != -1 and not IsCppString(line[:commentpos]):
1167 line = line[:commentpos].rstrip()
1168 # get rid of /* ... */
1169 return _RE_PATTERN_CLEANSE_LINE_C_COMMENTS.sub('', line)
1172 class CleansedLines(object):
1173 """Holds 3 copies of all lines with different preprocessing applied to them.
1175 1) elided member contains lines without strings and comments,
1176 2) lines member contains lines without comments, and
1177 3) raw_lines member contains all the lines without processing.
1178 All these three members are of <type 'list'>, and of the same length.
1181 def __init__(self, lines):
1182 self.elided = []
1183 self.lines = []
1184 self.raw_lines = lines
1185 self.num_lines = len(lines)
1186 self.lines_without_raw_strings = CleanseRawStrings(lines)
1187 for linenum in range(len(self.lines_without_raw_strings)):
1188 self.lines.append(CleanseComments(
1189 self.lines_without_raw_strings[linenum]))
1190 elided = self._CollapseStrings(self.lines_without_raw_strings[linenum])
1191 self.elided.append(CleanseComments(elided))
1193 def NumLines(self):
1194 """Returns the number of lines represented."""
1195 return self.num_lines
1197 @staticmethod
1198 def _CollapseStrings(elided):
1199 """Collapses strings and chars on a line to simple "" or '' blocks.
1201 We nix strings first so we're not fooled by text like '"http://"'
1203 Args:
1204 elided: The line being processed.
1206 Returns:
1207 The line with collapsed strings.
1209 if not _RE_PATTERN_INCLUDE.match(elided):
1210 # Remove escaped characters first to make quote/single quote collapsing
1211 # basic. Things that look like escaped characters shouldn't occur
1212 # outside of strings and chars.
1213 elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided)
1214 elided = _RE_PATTERN_CLEANSE_LINE_SINGLE_QUOTES.sub("''", elided)
1215 elided = _RE_PATTERN_CLEANSE_LINE_DOUBLE_QUOTES.sub('""', elided)
1216 return elided
1219 def FindEndOfExpressionInLine(line, startpos, depth, startchar, endchar):
1220 """Find the position just after the matching endchar.
1222 Args:
1223 line: a CleansedLines line.
1224 startpos: start searching at this position.
1225 depth: nesting level at startpos.
1226 startchar: expression opening character.
1227 endchar: expression closing character.
1229 Returns:
1230 On finding matching endchar: (index just after matching endchar, 0)
1231 Otherwise: (-1, new depth at end of this line)
1233 for i in xrange(startpos, len(line)):
1234 if line[i] == startchar:
1235 depth += 1
1236 elif line[i] == endchar:
1237 depth -= 1
1238 if depth == 0:
1239 return (i + 1, 0)
1240 return (-1, depth)
1243 def CloseExpression(clean_lines, linenum, pos):
1244 """If input points to ( or { or [ or <, finds the position that closes it.
1246 If lines[linenum][pos] points to a '(' or '{' or '[' or '<', finds the
1247 linenum/pos that correspond to the closing of the expression.
1249 Args:
1250 clean_lines: A CleansedLines instance containing the file.
1251 linenum: The number of the line to check.
1252 pos: A position on the line.
1254 Returns:
1255 A tuple (line, linenum, pos) pointer *past* the closing brace, or
1256 (line, len(lines), -1) if we never find a close. Note we ignore
1257 strings and comments when matching; and the line we return is the
1258 'cleansed' line at linenum.
1261 line = clean_lines.elided[linenum]
1262 startchar = line[pos]
1263 if startchar not in '({[<':
1264 return (line, clean_lines.NumLines(), -1)
1265 if startchar == '(': endchar = ')'
1266 if startchar == '[': endchar = ']'
1267 if startchar == '{': endchar = '}'
1268 if startchar == '<': endchar = '>'
1270 # Check first line
1271 (end_pos, num_open) = FindEndOfExpressionInLine(
1272 line, pos, 0, startchar, endchar)
1273 if end_pos > -1:
1274 return (line, linenum, end_pos)
1276 # Continue scanning forward
1277 while linenum < clean_lines.NumLines() - 1:
1278 linenum += 1
1279 line = clean_lines.elided[linenum]
1280 (end_pos, num_open) = FindEndOfExpressionInLine(
1281 line, 0, num_open, startchar, endchar)
1282 if end_pos > -1:
1283 return (line, linenum, end_pos)
1285 # Did not find endchar before end of file, give up
1286 return (line, clean_lines.NumLines(), -1)
1289 def FindStartOfExpressionInLine(line, endpos, depth, startchar, endchar):
1290 """Find position at the matching startchar.
1292 This is almost the reverse of FindEndOfExpressionInLine, but note
1293 that the input position and returned position differs by 1.
1295 Args:
1296 line: a CleansedLines line.
1297 endpos: start searching at this position.
1298 depth: nesting level at endpos.
1299 startchar: expression opening character.
1300 endchar: expression closing character.
1302 Returns:
1303 On finding matching startchar: (index at matching startchar, 0)
1304 Otherwise: (-1, new depth at beginning of this line)
1306 for i in xrange(endpos, -1, -1):
1307 if line[i] == endchar:
1308 depth += 1
1309 elif line[i] == startchar:
1310 depth -= 1
1311 if depth == 0:
1312 return (i, 0)
1313 return (-1, depth)
1316 def ReverseCloseExpression(clean_lines, linenum, pos):
1317 """If input points to ) or } or ] or >, finds the position that opens it.
1319 If lines[linenum][pos] points to a ')' or '}' or ']' or '>', finds the
1320 linenum/pos that correspond to the opening of the expression.
1322 Args:
1323 clean_lines: A CleansedLines instance containing the file.
1324 linenum: The number of the line to check.
1325 pos: A position on the line.
1327 Returns:
1328 A tuple (line, linenum, pos) pointer *at* the opening brace, or
1329 (line, 0, -1) if we never find the matching opening brace. Note
1330 we ignore strings and comments when matching; and the line we
1331 return is the 'cleansed' line at linenum.
1333 line = clean_lines.elided[linenum]
1334 endchar = line[pos]
1335 if endchar not in ')}]>':
1336 return (line, 0, -1)
1337 if endchar == ')': startchar = '('
1338 if endchar == ']': startchar = '['
1339 if endchar == '}': startchar = '{'
1340 if endchar == '>': startchar = '<'
1342 # Check last line
1343 (start_pos, num_open) = FindStartOfExpressionInLine(
1344 line, pos, 0, startchar, endchar)
1345 if start_pos > -1:
1346 return (line, linenum, start_pos)
1348 # Continue scanning backward
1349 while linenum > 0:
1350 linenum -= 1
1351 line = clean_lines.elided[linenum]
1352 (start_pos, num_open) = FindStartOfExpressionInLine(
1353 line, len(line) - 1, num_open, startchar, endchar)
1354 if start_pos > -1:
1355 return (line, linenum, start_pos)
1357 # Did not find startchar before beginning of file, give up
1358 return (line, 0, -1)
1361 def CheckForCopyright(filename, lines, error):
1362 """Logs an error if no Copyright message appears at the top of the file."""
1364 # We'll say it should occur by line 10. Don't forget there's a
1365 # dummy line at the front.
1366 for line in xrange(1, min(len(lines), 11)):
1367 if re.search(r'Copyright', lines[line], re.I): break
1368 else: # means no copyright line was found
1369 error(filename, 0, 'legal/copyright', 5,
1370 'No copyright message found. '
1371 'You should have a line: "Copyright [year] <Copyright Owner>"')
1374 def GetHeaderGuardCPPVariable(filename):
1375 """Returns the CPP variable that should be used as a header guard.
1377 Args:
1378 filename: The name of a C++ header file.
1380 Returns:
1381 The CPP variable that should be used as a header guard in the
1382 named file.
1386 # Restores original filename in case that cpplint is invoked from Emacs's
1387 # flymake.
1388 filename = re.sub(r'_flymake\.h$', '.h', filename)
1389 filename = re.sub(r'/\.flymake/([^/]*)$', r'/\1', filename)
1391 fileinfo = FileInfo(filename)
1392 file_path_from_root = fileinfo.RepositoryName()
1393 if _root:
1394 file_path_from_root = re.sub('^' + _root + os.sep, '', file_path_from_root)
1395 return re.sub(r'[-./\s]', '_', file_path_from_root).upper() + '_'
1398 def CheckForHeaderGuard(filename, lines, error):
1399 """Checks that the file contains a header guard.
1401 Logs an error if no #ifndef header guard is present. For other
1402 headers, checks that the full pathname is used.
1404 Args:
1405 filename: The name of the C++ header file.
1406 lines: An array of strings, each representing a line of the file.
1407 error: The function to call with any errors found.
1410 cppvar = GetHeaderGuardCPPVariable(filename)
1412 ifndef = None
1413 ifndef_linenum = 0
1414 define = None
1415 endif = None
1416 endif_linenum = 0
1417 for linenum, line in enumerate(lines):
1418 linesplit = line.split()
1419 if len(linesplit) >= 2:
1420 # find the first occurrence of #ifndef and #define, save arg
1421 if not ifndef and linesplit[0] == '#ifndef':
1422 # set ifndef to the header guard presented on the #ifndef line.
1423 ifndef = linesplit[1]
1424 ifndef_linenum = linenum
1425 if not define and linesplit[0] == '#define':
1426 define = linesplit[1]
1427 # find the last occurrence of #endif, save entire line
1428 if line.startswith('#endif'):
1429 endif = line
1430 endif_linenum = linenum
1432 if not ifndef:
1433 error(filename, 0, 'build/header_guard', 5,
1434 'No #ifndef header guard found, suggested CPP variable is: %s' %
1435 cppvar)
1436 return
1438 if not define:
1439 error(filename, 0, 'build/header_guard', 5,
1440 'No #define header guard found, suggested CPP variable is: %s' %
1441 cppvar)
1442 return
1444 # The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__
1445 # for backward compatibility.
1446 if ifndef != cppvar:
1447 error_level = 0
1448 if ifndef != cppvar + '_':
1449 error_level = 5
1451 ParseNolintSuppressions(filename, lines[ifndef_linenum], ifndef_linenum,
1452 error)
1453 error(filename, ifndef_linenum, 'build/header_guard', error_level,
1454 '#ifndef header guard has wrong style, please use: %s' % cppvar)
1456 if define != ifndef:
1457 error(filename, 0, 'build/header_guard', 5,
1458 '#ifndef and #define don\'t match, suggested CPP variable is: %s' %
1459 cppvar)
1460 return
1462 if endif != ('#endif // %s' % cppvar):
1463 error_level = 0
1464 if endif != ('#endif // %s' % (cppvar + '_')):
1465 error_level = 5
1467 ParseNolintSuppressions(filename, lines[endif_linenum], endif_linenum,
1468 error)
1469 error(filename, endif_linenum, 'build/header_guard', error_level,
1470 '#endif line should be "#endif // %s"' % cppvar)
1473 def CheckForBadCharacters(filename, lines, error):
1474 """Logs an error for each line containing bad characters.
1476 Two kinds of bad characters:
1478 1. Unicode replacement characters: These indicate that either the file
1479 contained invalid UTF-8 (likely) or Unicode replacement characters (which
1480 it shouldn't). Note that it's possible for this to throw off line
1481 numbering if the invalid UTF-8 occurred adjacent to a newline.
1483 2. NUL bytes. These are problematic for some tools.
1485 Args:
1486 filename: The name of the current file.
1487 lines: An array of strings, each representing a line of the file.
1488 error: The function to call with any errors found.
1490 for linenum, line in enumerate(lines):
1491 if u'\ufffd' in line:
1492 error(filename, linenum, 'readability/utf8', 5,
1493 'Line contains invalid UTF-8 (or Unicode replacement character).')
1494 if '\0' in line:
1495 error(filename, linenum, 'readability/nul', 5, 'Line contains NUL byte.')
1498 def CheckForNewlineAtEOF(filename, lines, error):
1499 """Logs an error if there is no newline char at the end of the file.
1501 Args:
1502 filename: The name of the current file.
1503 lines: An array of strings, each representing a line of the file.
1504 error: The function to call with any errors found.
1507 # The array lines() was created by adding two newlines to the
1508 # original file (go figure), then splitting on \n.
1509 # To verify that the file ends in \n, we just have to make sure the
1510 # last-but-two element of lines() exists and is empty.
1511 if len(lines) < 3 or lines[-2]:
1512 error(filename, len(lines) - 2, 'whitespace/ending_newline', 5,
1513 'Could not find a newline character at the end of the file.')
1516 def CheckForMultilineCommentsAndStrings(filename, clean_lines, linenum, error):
1517 """Logs an error if we see /* ... */ or "..." that extend past one line.
1519 /* ... */ comments are legit inside macros, for one line.
1520 Otherwise, we prefer // comments, so it's ok to warn about the
1521 other. Likewise, it's ok for strings to extend across multiple
1522 lines, as long as a line continuation character (backslash)
1523 terminates each line. Although not currently prohibited by the C++
1524 style guide, it's ugly and unnecessary. We don't do well with either
1525 in this lint program, so we warn about both.
1527 Args:
1528 filename: The name of the current file.
1529 clean_lines: A CleansedLines instance containing the file.
1530 linenum: The number of the line to check.
1531 error: The function to call with any errors found.
1533 line = clean_lines.elided[linenum]
1535 # Remove all \\ (escaped backslashes) from the line. They are OK, and the
1536 # second (escaped) slash may trigger later \" detection erroneously.
1537 line = line.replace('\\\\', '')
1539 if line.count('/*') > line.count('*/'):
1540 error(filename, linenum, 'readability/multiline_comment', 5,
1541 'Complex multi-line /*...*/-style comment found. '
1542 'Lint may give bogus warnings. '
1543 'Consider replacing these with //-style comments, '
1544 'with #if 0...#endif, '
1545 'or with more clearly structured multi-line comments.')
1547 if (line.count('"') - line.count('\\"')) % 2:
1548 error(filename, linenum, 'readability/multiline_string', 5,
1549 'Multi-line string ("...") found. This lint script doesn\'t '
1550 'do well with such strings, and may give bogus warnings. '
1551 'Use C++11 raw strings or concatenation instead.')
1554 threading_list = (
1555 ('asctime(', 'asctime_r('),
1556 ('ctime(', 'ctime_r('),
1557 ('getgrgid(', 'getgrgid_r('),
1558 ('getgrnam(', 'getgrnam_r('),
1559 ('getlogin(', 'getlogin_r('),
1560 ('getpwnam(', 'getpwnam_r('),
1561 ('getpwuid(', 'getpwuid_r('),
1562 ('gmtime(', 'gmtime_r('),
1563 ('localtime(', 'localtime_r('),
1564 ('rand(', 'rand_r('),
1565 ('strtok(', 'strtok_r('),
1566 ('ttyname(', 'ttyname_r('),
1570 def CheckPosixThreading(filename, clean_lines, linenum, error):
1571 """Checks for calls to thread-unsafe functions.
1573 Much code has been originally written without consideration of
1574 multi-threading. Also, engineers are relying on their old experience;
1575 they have learned posix before threading extensions were added. These
1576 tests guide the engineers to use thread-safe functions (when using
1577 posix directly).
1579 Args:
1580 filename: The name of the current file.
1581 clean_lines: A CleansedLines instance containing the file.
1582 linenum: The number of the line to check.
1583 error: The function to call with any errors found.
1585 line = clean_lines.elided[linenum]
1586 for single_thread_function, multithread_safe_function in threading_list:
1587 ix = line.find(single_thread_function)
1588 # Comparisons made explicit for clarity -- pylint: disable=g-explicit-bool-comparison
1589 if ix >= 0 and (ix == 0 or (not line[ix - 1].isalnum() and
1590 line[ix - 1] not in ('_', '.', '>'))):
1591 error(filename, linenum, 'runtime/threadsafe_fn', 2,
1592 'Consider using ' + multithread_safe_function +
1593 '...) instead of ' + single_thread_function +
1594 '...) for improved thread safety.')
1597 def CheckVlogArguments(filename, clean_lines, linenum, error):
1598 """Checks that VLOG() is only used for defining a logging level.
1600 For example, VLOG(2) is correct. VLOG(INFO), VLOG(WARNING), VLOG(ERROR), and
1601 VLOG(FATAL) are not.
1603 Args:
1604 filename: The name of the current file.
1605 clean_lines: A CleansedLines instance containing the file.
1606 linenum: The number of the line to check.
1607 error: The function to call with any errors found.
1609 line = clean_lines.elided[linenum]
1610 if Search(r'\bVLOG\((INFO|ERROR|WARNING|DFATAL|FATAL)\)', line):
1611 error(filename, linenum, 'runtime/vlog', 5,
1612 'VLOG() should be used with numeric verbosity level. '
1613 'Use LOG() if you want symbolic severity levels.')
1616 # Matches invalid increment: *count++, which moves pointer instead of
1617 # incrementing a value.
1618 _RE_PATTERN_INVALID_INCREMENT = re.compile(
1619 r'^\s*\*\w+(\+\+|--);')
1622 def CheckInvalidIncrement(filename, clean_lines, linenum, error):
1623 """Checks for invalid increment *count++.
1625 For example following function:
1626 void increment_counter(int* count) {
1627 *count++;
1629 is invalid, because it effectively does count++, moving pointer, and should
1630 be replaced with ++*count, (*count)++ or *count += 1.
1632 Args:
1633 filename: The name of the current file.
1634 clean_lines: A CleansedLines instance containing the file.
1635 linenum: The number of the line to check.
1636 error: The function to call with any errors found.
1638 line = clean_lines.elided[linenum]
1639 if _RE_PATTERN_INVALID_INCREMENT.match(line):
1640 error(filename, linenum, 'runtime/invalid_increment', 5,
1641 'Changing pointer instead of value (or unused value of operator*).')
1644 class _BlockInfo(object):
1645 """Stores information about a generic block of code."""
1647 def __init__(self, seen_open_brace):
1648 self.seen_open_brace = seen_open_brace
1649 self.open_parentheses = 0
1650 self.inline_asm = _NO_ASM
1652 def CheckBegin(self, filename, clean_lines, linenum, error):
1653 """Run checks that applies to text up to the opening brace.
1655 This is mostly for checking the text after the class identifier
1656 and the "{", usually where the base class is specified. For other
1657 blocks, there isn't much to check, so we always pass.
1659 Args:
1660 filename: The name of the current file.
1661 clean_lines: A CleansedLines instance containing the file.
1662 linenum: The number of the line to check.
1663 error: The function to call with any errors found.
1665 pass
1667 def CheckEnd(self, filename, clean_lines, linenum, error):
1668 """Run checks that applies to text after the closing brace.
1670 This is mostly used for checking end of namespace comments.
1672 Args:
1673 filename: The name of the current file.
1674 clean_lines: A CleansedLines instance containing the file.
1675 linenum: The number of the line to check.
1676 error: The function to call with any errors found.
1678 pass
1681 class _ClassInfo(_BlockInfo):
1682 """Stores information about a class."""
1684 def __init__(self, name, class_or_struct, clean_lines, linenum):
1685 _BlockInfo.__init__(self, False)
1686 self.name = name
1687 self.starting_linenum = linenum
1688 self.is_derived = False
1689 if class_or_struct == 'struct':
1690 self.access = 'public'
1691 self.is_struct = True
1692 else:
1693 self.access = 'private'
1694 self.is_struct = False
1696 # Remember initial indentation level for this class. Using raw_lines here
1697 # instead of elided to account for leading comments.
1698 initial_indent = Match(r'^( *)\S', clean_lines.raw_lines[linenum])
1699 if initial_indent:
1700 self.class_indent = len(initial_indent.group(1))
1701 else:
1702 self.class_indent = 0
1704 # Try to find the end of the class. This will be confused by things like:
1705 # class A {
1706 # } *x = { ...
1708 # But it's still good enough for CheckSectionSpacing.
1709 self.last_line = 0
1710 depth = 0
1711 for i in range(linenum, clean_lines.NumLines()):
1712 line = clean_lines.elided[i]
1713 depth += line.count('{') - line.count('}')
1714 if not depth:
1715 self.last_line = i
1716 break
1718 def CheckBegin(self, filename, clean_lines, linenum, error):
1719 # Look for a bare ':'
1720 if Search('(^|[^:]):($|[^:])', clean_lines.elided[linenum]):
1721 self.is_derived = True
1723 def CheckEnd(self, filename, clean_lines, linenum, error):
1724 # Check that closing brace is aligned with beginning of the class.
1725 # Only do this if the closing brace is indented by only whitespaces.
1726 # This means we will not check single-line class definitions.
1727 indent = Match(r'^( *)\}', clean_lines.elided[linenum])
1728 if indent and len(indent.group(1)) != self.class_indent:
1729 if self.is_struct:
1730 parent = 'struct ' + self.name
1731 else:
1732 parent = 'class ' + self.name
1733 error(filename, linenum, 'whitespace/indent', 3,
1734 'Closing brace should be aligned with beginning of %s' % parent)
1737 class _NamespaceInfo(_BlockInfo):
1738 """Stores information about a namespace."""
1740 def __init__(self, name, linenum):
1741 _BlockInfo.__init__(self, False)
1742 self.name = name or ''
1743 self.starting_linenum = linenum
1745 def CheckEnd(self, filename, clean_lines, linenum, error):
1746 """Check end of namespace comments."""
1747 line = clean_lines.raw_lines[linenum]
1749 # Check how many lines is enclosed in this namespace. Don't issue
1750 # warning for missing namespace comments if there aren't enough
1751 # lines. However, do apply checks if there is already an end of
1752 # namespace comment and it's incorrect.
1754 # TODO(unknown): We always want to check end of namespace comments
1755 # if a namespace is large, but sometimes we also want to apply the
1756 # check if a short namespace contained nontrivial things (something
1757 # other than forward declarations). There is currently no logic on
1758 # deciding what these nontrivial things are, so this check is
1759 # triggered by namespace size only, which works most of the time.
1760 if (linenum - self.starting_linenum < 10
1761 and not Match(r'};*\s*(//|/\*).*\bnamespace\b', line)):
1762 return
1764 # Look for matching comment at end of namespace.
1766 # Note that we accept C style "/* */" comments for terminating
1767 # namespaces, so that code that terminate namespaces inside
1768 # preprocessor macros can be cpplint clean.
1770 # We also accept stuff like "// end of namespace <name>." with the
1771 # period at the end.
1773 # Besides these, we don't accept anything else, otherwise we might
1774 # get false negatives when existing comment is a substring of the
1775 # expected namespace.
1776 if self.name:
1777 # Named namespace
1778 if not Match((r'};*\s*(//|/\*).*\bnamespace\s+' + re.escape(self.name) +
1779 r'[\*/\.\\\s]*$'),
1780 line):
1781 error(filename, linenum, 'readability/namespace', 5,
1782 'Namespace should be terminated with "// namespace %s"' %
1783 self.name)
1784 else:
1785 # Anonymous namespace
1786 if not Match(r'};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line):
1787 error(filename, linenum, 'readability/namespace', 5,
1788 'Namespace should be terminated with "// namespace"')
1791 class _PreprocessorInfo(object):
1792 """Stores checkpoints of nesting stacks when #if/#else is seen."""
1794 def __init__(self, stack_before_if):
1795 # The entire nesting stack before #if
1796 self.stack_before_if = stack_before_if
1798 # The entire nesting stack up to #else
1799 self.stack_before_else = []
1801 # Whether we have already seen #else or #elif
1802 self.seen_else = False
1805 class _NestingState(object):
1806 """Holds states related to parsing braces."""
1808 def __init__(self):
1809 # Stack for tracking all braces. An object is pushed whenever we
1810 # see a "{", and popped when we see a "}". Only 3 types of
1811 # objects are possible:
1812 # - _ClassInfo: a class or struct.
1813 # - _NamespaceInfo: a namespace.
1814 # - _BlockInfo: some other type of block.
1815 self.stack = []
1817 # Stack of _PreprocessorInfo objects.
1818 self.pp_stack = []
1820 def SeenOpenBrace(self):
1821 """Check if we have seen the opening brace for the innermost block.
1823 Returns:
1824 True if we have seen the opening brace, False if the innermost
1825 block is still expecting an opening brace.
1827 return (not self.stack) or self.stack[-1].seen_open_brace
1829 def InNamespaceBody(self):
1830 """Check if we are currently one level inside a namespace body.
1832 Returns:
1833 True if top of the stack is a namespace block, False otherwise.
1835 return self.stack and isinstance(self.stack[-1], _NamespaceInfo)
1837 def UpdatePreprocessor(self, line):
1838 """Update preprocessor stack.
1840 We need to handle preprocessors due to classes like this:
1841 #ifdef SWIG
1842 struct ResultDetailsPageElementExtensionPoint {
1843 #else
1844 struct ResultDetailsPageElementExtensionPoint : public Extension {
1845 #endif
1847 We make the following assumptions (good enough for most files):
1848 - Preprocessor condition evaluates to true from #if up to first
1849 #else/#elif/#endif.
1851 - Preprocessor condition evaluates to false from #else/#elif up
1852 to #endif. We still perform lint checks on these lines, but
1853 these do not affect nesting stack.
1855 Args:
1856 line: current line to check.
1858 if Match(r'^\s*#\s*(if|ifdef|ifndef)\b', line):
1859 # Beginning of #if block, save the nesting stack here. The saved
1860 # stack will allow us to restore the parsing state in the #else case.
1861 self.pp_stack.append(_PreprocessorInfo(copy.deepcopy(self.stack)))
1862 elif Match(r'^\s*#\s*(else|elif)\b', line):
1863 # Beginning of #else block
1864 if self.pp_stack:
1865 if not self.pp_stack[-1].seen_else:
1866 # This is the first #else or #elif block. Remember the
1867 # whole nesting stack up to this point. This is what we
1868 # keep after the #endif.
1869 self.pp_stack[-1].seen_else = True
1870 self.pp_stack[-1].stack_before_else = copy.deepcopy(self.stack)
1872 # Restore the stack to how it was before the #if
1873 self.stack = copy.deepcopy(self.pp_stack[-1].stack_before_if)
1874 else:
1875 # TODO(unknown): unexpected #else, issue warning?
1876 pass
1877 elif Match(r'^\s*#\s*endif\b', line):
1878 # End of #if or #else blocks.
1879 if self.pp_stack:
1880 # If we saw an #else, we will need to restore the nesting
1881 # stack to its former state before the #else, otherwise we
1882 # will just continue from where we left off.
1883 if self.pp_stack[-1].seen_else:
1884 # Here we can just use a shallow copy since we are the last
1885 # reference to it.
1886 self.stack = self.pp_stack[-1].stack_before_else
1887 # Drop the corresponding #if
1888 self.pp_stack.pop()
1889 else:
1890 # TODO(unknown): unexpected #endif, issue warning?
1891 pass
1893 def Update(self, filename, clean_lines, linenum, error):
1894 """Update nesting state with current line.
1896 Args:
1897 filename: The name of the current file.
1898 clean_lines: A CleansedLines instance containing the file.
1899 linenum: The number of the line to check.
1900 error: The function to call with any errors found.
1902 line = clean_lines.elided[linenum]
1904 # Update pp_stack first
1905 self.UpdatePreprocessor(line)
1907 # Count parentheses. This is to avoid adding struct arguments to
1908 # the nesting stack.
1909 if self.stack:
1910 inner_block = self.stack[-1]
1911 depth_change = line.count('(') - line.count(')')
1912 inner_block.open_parentheses += depth_change
1914 # Also check if we are starting or ending an inline assembly block.
1915 if inner_block.inline_asm in (_NO_ASM, _END_ASM):
1916 if (depth_change != 0 and
1917 inner_block.open_parentheses == 1 and
1918 _MATCH_ASM.match(line)):
1919 # Enter assembly block
1920 inner_block.inline_asm = _INSIDE_ASM
1921 else:
1922 # Not entering assembly block. If previous line was _END_ASM,
1923 # we will now shift to _NO_ASM state.
1924 inner_block.inline_asm = _NO_ASM
1925 elif (inner_block.inline_asm == _INSIDE_ASM and
1926 inner_block.open_parentheses == 0):
1927 # Exit assembly block
1928 inner_block.inline_asm = _END_ASM
1930 # Consume namespace declaration at the beginning of the line. Do
1931 # this in a loop so that we catch same line declarations like this:
1932 # namespace proto2 { namespace bridge { class MessageSet; } }
1933 while True:
1934 # Match start of namespace. The "\b\s*" below catches namespace
1935 # declarations even if it weren't followed by a whitespace, this
1936 # is so that we don't confuse our namespace checker. The
1937 # missing spaces will be flagged by CheckSpacing.
1938 namespace_decl_match = Match(r'^\s*namespace\b\s*([:\w]+)?(.*)$', line)
1939 if not namespace_decl_match:
1940 break
1942 new_namespace = _NamespaceInfo(namespace_decl_match.group(1), linenum)
1943 self.stack.append(new_namespace)
1945 line = namespace_decl_match.group(2)
1946 if line.find('{') != -1:
1947 new_namespace.seen_open_brace = True
1948 line = line[line.find('{') + 1:]
1950 # Look for a class declaration in whatever is left of the line
1951 # after parsing namespaces. The regexp accounts for decorated classes
1952 # such as in:
1953 # class LOCKABLE API Object {
1954 # };
1956 # Templates with class arguments may confuse the parser, for example:
1957 # template <class T
1958 # class Comparator = less<T>,
1959 # class Vector = vector<T> >
1960 # class HeapQueue {
1962 # Because this parser has no nesting state about templates, by the
1963 # time it saw "class Comparator", it may think that it's a new class.
1964 # Nested templates have a similar problem:
1965 # template <
1966 # typename ExportedType,
1967 # typename TupleType,
1968 # template <typename, typename> class ImplTemplate>
1970 # To avoid these cases, we ignore classes that are followed by '=' or '>'
1971 class_decl_match = Match(
1972 r'\s*(template\s*<[\w\s<>,:]*>\s*)?'
1973 r'(class|struct)\s+([A-Z_]+\s+)*(\w+(?:::\w+)*)'
1974 r'(([^=>]|<[^<>]*>|<[^<>]*<[^<>]*>\s*>)*)$', line)
1975 if (class_decl_match and
1976 (not self.stack or self.stack[-1].open_parentheses == 0)):
1977 self.stack.append(_ClassInfo(
1978 class_decl_match.group(4), class_decl_match.group(2),
1979 clean_lines, linenum))
1980 line = class_decl_match.group(5)
1982 # If we have not yet seen the opening brace for the innermost block,
1983 # run checks here.
1984 if not self.SeenOpenBrace():
1985 self.stack[-1].CheckBegin(filename, clean_lines, linenum, error)
1987 # Update access control if we are inside a class/struct
1988 if self.stack and isinstance(self.stack[-1], _ClassInfo):
1989 classinfo = self.stack[-1]
1990 access_match = Match(
1991 r'^(.*)\b(public|private|protected|signals)(\s+(?:slots\s*)?)?'
1992 r':(?:[^:]|$)',
1993 line)
1994 if access_match:
1995 classinfo.access = access_match.group(2)
1997 # Check that access keywords are indented +1 space. Skip this
1998 # check if the keywords are not preceded by whitespaces.
1999 indent = access_match.group(1)
2000 if (len(indent) != classinfo.class_indent + 1 and
2001 Match(r'^\s*$', indent)):
2002 if classinfo.is_struct:
2003 parent = 'struct ' + classinfo.name
2004 else:
2005 parent = 'class ' + classinfo.name
2006 slots = ''
2007 if access_match.group(3):
2008 slots = access_match.group(3)
2009 error(filename, linenum, 'whitespace/indent', 3,
2010 '%s%s: should be indented +1 space inside %s' % (
2011 access_match.group(2), slots, parent))
2013 # Consume braces or semicolons from what's left of the line
2014 while True:
2015 # Match first brace, semicolon, or closed parenthesis.
2016 matched = Match(r'^[^{;)}]*([{;)}])(.*)$', line)
2017 if not matched:
2018 break
2020 token = matched.group(1)
2021 if token == '{':
2022 # If namespace or class hasn't seen a opening brace yet, mark
2023 # namespace/class head as complete. Push a new block onto the
2024 # stack otherwise.
2025 if not self.SeenOpenBrace():
2026 self.stack[-1].seen_open_brace = True
2027 else:
2028 self.stack.append(_BlockInfo(True))
2029 if _MATCH_ASM.match(line):
2030 self.stack[-1].inline_asm = _BLOCK_ASM
2031 elif token == ';' or token == ')':
2032 # If we haven't seen an opening brace yet, but we already saw
2033 # a semicolon, this is probably a forward declaration. Pop
2034 # the stack for these.
2036 # Similarly, if we haven't seen an opening brace yet, but we
2037 # already saw a closing parenthesis, then these are probably
2038 # function arguments with extra "class" or "struct" keywords.
2039 # Also pop these stack for these.
2040 if not self.SeenOpenBrace():
2041 self.stack.pop()
2042 else: # token == '}'
2043 # Perform end of block checks and pop the stack.
2044 if self.stack:
2045 self.stack[-1].CheckEnd(filename, clean_lines, linenum, error)
2046 self.stack.pop()
2047 line = matched.group(2)
2049 def InnermostClass(self):
2050 """Get class info on the top of the stack.
2052 Returns:
2053 A _ClassInfo object if we are inside a class, or None otherwise.
2055 for i in range(len(self.stack), 0, -1):
2056 classinfo = self.stack[i - 1]
2057 if isinstance(classinfo, _ClassInfo):
2058 return classinfo
2059 return None
2061 def CheckCompletedBlocks(self, filename, error):
2062 """Checks that all classes and namespaces have been completely parsed.
2064 Call this when all lines in a file have been processed.
2065 Args:
2066 filename: The name of the current file.
2067 error: The function to call with any errors found.
2069 # Note: This test can result in false positives if #ifdef constructs
2070 # get in the way of brace matching. See the testBuildClass test in
2071 # cpplint_unittest.py for an example of this.
2072 for obj in self.stack:
2073 if isinstance(obj, _ClassInfo):
2074 error(filename, obj.starting_linenum, 'build/class', 5,
2075 'Failed to find complete declaration of class %s' %
2076 obj.name)
2077 elif isinstance(obj, _NamespaceInfo):
2078 error(filename, obj.starting_linenum, 'build/namespaces', 5,
2079 'Failed to find complete declaration of namespace %s' %
2080 obj.name)
2083 def CheckForNonStandardConstructs(filename, clean_lines, linenum,
2084 nesting_state, error):
2085 r"""Logs an error if we see certain non-ANSI constructs ignored by gcc-2.
2087 Complain about several constructs which gcc-2 accepts, but which are
2088 not standard C++. Warning about these in lint is one way to ease the
2089 transition to new compilers.
2090 - put storage class first (e.g. "static const" instead of "const static").
2091 - "%lld" instead of %qd" in printf-type functions.
2092 - "%1$d" is non-standard in printf-type functions.
2093 - "\%" is an undefined character escape sequence.
2094 - text after #endif is not allowed.
2095 - invalid inner-style forward declaration.
2096 - >? and <? operators, and their >?= and <?= cousins.
2098 Additionally, check for constructor/destructor style violations and reference
2099 members, as it is very convenient to do so while checking for
2100 gcc-2 compliance.
2102 Args:
2103 filename: The name of the current file.
2104 clean_lines: A CleansedLines instance containing the file.
2105 linenum: The number of the line to check.
2106 nesting_state: A _NestingState instance which maintains information about
2107 the current stack of nested blocks being parsed.
2108 error: A callable to which errors are reported, which takes 4 arguments:
2109 filename, line number, error level, and message
2112 # Remove comments from the line, but leave in strings for now.
2113 line = clean_lines.lines[linenum]
2115 if Search(r'printf\s*\(.*".*%[-+ ]?\d*q', line):
2116 error(filename, linenum, 'runtime/printf_format', 3,
2117 '%q in format strings is deprecated. Use %ll instead.')
2119 if Search(r'printf\s*\(.*".*%\d+\$', line):
2120 error(filename, linenum, 'runtime/printf_format', 2,
2121 '%N$ formats are unconventional. Try rewriting to avoid them.')
2123 # Remove escaped backslashes before looking for undefined escapes.
2124 line = line.replace('\\\\', '')
2126 if Search(r'("|\').*\\(%|\[|\(|{)', line):
2127 error(filename, linenum, 'build/printf_format', 3,
2128 '%, [, (, and { are undefined character escapes. Unescape them.')
2130 # For the rest, work with both comments and strings removed.
2131 line = clean_lines.elided[linenum]
2133 if Search(r'\b(const|volatile|void|char|short|int|long'
2134 r'|float|double|signed|unsigned'
2135 r'|schar|u?int8|u?int16|u?int32|u?int64)'
2136 r'\s+(register|static|extern|typedef)\b',
2137 line):
2138 error(filename, linenum, 'build/storage_class', 5,
2139 'Storage class (static, extern, typedef, etc) should be first.')
2141 if Match(r'\s*#\s*endif\s*[^/\s]+', line):
2142 error(filename, linenum, 'build/endif_comment', 5,
2143 'Uncommented text after #endif is non-standard. Use a comment.')
2145 if Match(r'\s*class\s+(\w+\s*::\s*)+\w+\s*;', line):
2146 error(filename, linenum, 'build/forward_decl', 5,
2147 'Inner-style forward declarations are invalid. Remove this line.')
2149 if Search(r'(\w+|[+-]?\d+(\.\d*)?)\s*(<|>)\?=?\s*(\w+|[+-]?\d+)(\.\d*)?',
2150 line):
2151 error(filename, linenum, 'build/deprecated', 3,
2152 '>? and <? (max and min) operators are non-standard and deprecated.')
2154 if Search(r'^\s*const\s*string\s*&\s*\w+\s*;', line):
2155 # TODO(unknown): Could it be expanded safely to arbitrary references,
2156 # without triggering too many false positives? The first
2157 # attempt triggered 5 warnings for mostly benign code in the regtest, hence
2158 # the restriction.
2159 # Here's the original regexp, for the reference:
2160 # type_name = r'\w+((\s*::\s*\w+)|(\s*<\s*\w+?\s*>))?'
2161 # r'\s*const\s*' + type_name + '\s*&\s*\w+\s*;'
2162 error(filename, linenum, 'runtime/member_string_references', 2,
2163 'const string& members are dangerous. It is much better to use '
2164 'alternatives, such as pointers or simple constants.')
2166 # Everything else in this function operates on class declarations.
2167 # Return early if the top of the nesting stack is not a class, or if
2168 # the class head is not completed yet.
2169 classinfo = nesting_state.InnermostClass()
2170 if not classinfo or not classinfo.seen_open_brace:
2171 return
2173 # The class may have been declared with namespace or classname qualifiers.
2174 # The constructor and destructor will not have those qualifiers.
2175 base_classname = classinfo.name.split('::')[-1]
2177 # Look for single-argument constructors that aren't marked explicit.
2178 # Technically a valid construct, but against style.
2179 args = Match(r'\s+(?:inline\s+)?%s\s*\(([^,()]+)\)'
2180 % re.escape(base_classname),
2181 line)
2182 if (args and
2183 args.group(1) != 'void' and
2184 not Match(r'(const\s+)?%s(\s+const)?\s*(?:<\w+>\s*)?&'
2185 % re.escape(base_classname), args.group(1).strip())):
2186 error(filename, linenum, 'runtime/explicit', 5,
2187 'Single-argument constructors should be marked explicit.')
2190 def CheckSpacingForFunctionCall(filename, line, linenum, error):
2191 """Checks for the correctness of various spacing around function calls.
2193 Args:
2194 filename: The name of the current file.
2195 line: The text of the line to check.
2196 linenum: The number of the line to check.
2197 error: The function to call with any errors found.
2200 # Since function calls often occur inside if/for/while/switch
2201 # expressions - which have their own, more liberal conventions - we
2202 # first see if we should be looking inside such an expression for a
2203 # function call, to which we can apply more strict standards.
2204 fncall = line # if there's no control flow construct, look at whole line
2205 for pattern in (r'\bif\s*\((.*)\)\s*{',
2206 r'\bfor\s*\((.*)\)\s*{',
2207 r'\bwhile\s*\((.*)\)\s*[{;]',
2208 r'\bswitch\s*\((.*)\)\s*{'):
2209 match = Search(pattern, line)
2210 if match:
2211 fncall = match.group(1) # look inside the parens for function calls
2212 break
2214 # Except in if/for/while/switch, there should never be space
2215 # immediately inside parens (eg "f( 3, 4 )"). We make an exception
2216 # for nested parens ( (a+b) + c ). Likewise, there should never be
2217 # a space before a ( when it's a function argument. I assume it's a
2218 # function argument when the char before the whitespace is legal in
2219 # a function name (alnum + _) and we're not starting a macro. Also ignore
2220 # pointers and references to arrays and functions coz they're too tricky:
2221 # we use a very simple way to recognize these:
2222 # " (something)(maybe-something)" or
2223 # " (something)(maybe-something," or
2224 # " (something)[something]"
2225 # Note that we assume the contents of [] to be short enough that
2226 # they'll never need to wrap.
2227 if ( # Ignore control structures.
2228 not Search(r'\b(if|for|while|switch|return|new|delete|catch|sizeof)\b',
2229 fncall) and
2230 # Ignore pointers/references to functions.
2231 not Search(r' \([^)]+\)\([^)]*(\)|,$)', fncall) and
2232 # Ignore pointers/references to arrays.
2233 not Search(r' \([^)]+\)\[[^\]]+\]', fncall)):
2234 if Search(r'\w\s*\(\s(?!\s*\\$)', fncall): # a ( used for a fn call
2235 error(filename, linenum, 'whitespace/parens', 4,
2236 'Extra space after ( in function call')
2237 elif Search(r'\(\s+(?!(\s*\\)|\()', fncall):
2238 error(filename, linenum, 'whitespace/parens', 2,
2239 'Extra space after (')
2240 if (Search(r'\w\s+\(', fncall) and
2241 not Search(r'#\s*define|typedef', fncall) and
2242 not Search(r'\w\s+\((\w+::)*\*\w+\)\(', fncall)):
2243 error(filename, linenum, 'whitespace/parens', 4,
2244 'Extra space before ( in function call')
2245 # If the ) is followed only by a newline or a { + newline, assume it's
2246 # part of a control statement (if/while/etc), and don't complain
2247 if Search(r'[^)]\s+\)\s*[^{\s]', fncall):
2248 # If the closing parenthesis is preceded by only whitespaces,
2249 # try to give a more descriptive error message.
2250 if Search(r'^\s+\)', fncall):
2251 error(filename, linenum, 'whitespace/parens', 2,
2252 'Closing ) should be moved to the previous line')
2253 else:
2254 error(filename, linenum, 'whitespace/parens', 2,
2255 'Extra space before )')
2258 def IsBlankLine(line):
2259 """Returns true if the given line is blank.
2261 We consider a line to be blank if the line is empty or consists of
2262 only white spaces.
2264 Args:
2265 line: A line of a string.
2267 Returns:
2268 True, if the given line is blank.
2270 return not line or line.isspace()
2273 def CheckForFunctionLengths(filename, clean_lines, linenum,
2274 function_state, error):
2275 """Reports for long function bodies.
2277 For an overview why this is done, see:
2278 http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions
2280 Uses a simplistic algorithm assuming other style guidelines
2281 (especially spacing) are followed.
2282 Only checks unindented functions, so class members are unchecked.
2283 Trivial bodies are unchecked, so constructors with huge initializer lists
2284 may be missed.
2285 Blank/comment lines are not counted so as to avoid encouraging the removal
2286 of vertical space and comments just to get through a lint check.
2287 NOLINT *on the last line of a function* disables this check.
2289 Args:
2290 filename: The name of the current file.
2291 clean_lines: A CleansedLines instance containing the file.
2292 linenum: The number of the line to check.
2293 function_state: Current function name and lines in body so far.
2294 error: The function to call with any errors found.
2296 lines = clean_lines.lines
2297 line = lines[linenum]
2298 raw = clean_lines.raw_lines
2299 raw_line = raw[linenum]
2300 joined_line = ''
2302 starting_func = False
2303 regexp = r'(\w(\w|::|\*|\&|\s)*)\(' # decls * & space::name( ...
2304 match_result = Match(regexp, line)
2305 if match_result:
2306 # If the name is all caps and underscores, figure it's a macro and
2307 # ignore it, unless it's TEST or TEST_F.
2308 function_name = match_result.group(1).split()[-1]
2309 if function_name == 'TEST' or function_name == 'TEST_F' or (
2310 not Match(r'[A-Z_]+$', function_name)):
2311 starting_func = True
2313 if starting_func:
2314 body_found = False
2315 for start_linenum in xrange(linenum, clean_lines.NumLines()):
2316 start_line = lines[start_linenum]
2317 joined_line += ' ' + start_line.lstrip()
2318 if Search(r'(;|})', start_line): # Declarations and trivial functions
2319 body_found = True
2320 break # ... ignore
2321 elif Search(r'{', start_line):
2322 body_found = True
2323 function = Search(r'((\w|:)*)\(', line).group(1)
2324 if Match(r'TEST', function): # Handle TEST... macros
2325 parameter_regexp = Search(r'(\(.*\))', joined_line)
2326 if parameter_regexp: # Ignore bad syntax
2327 function += parameter_regexp.group(1)
2328 else:
2329 function += '()'
2330 function_state.Begin(function)
2331 break
2332 if not body_found:
2333 # No body for the function (or evidence of a non-function) was found.
2334 error(filename, linenum, 'readability/fn_size', 5,
2335 'Lint failed to find start of function body.')
2336 elif Match(r'^\}\s*$', line): # function end
2337 function_state.Check(error, filename, linenum)
2338 function_state.End()
2339 elif not Match(r'^\s*$', line):
2340 function_state.Count() # Count non-blank/non-comment lines.
2343 _RE_PATTERN_TODO = re.compile(r'^//(\s*)TODO(\(.+?\))?:?(\s|$)?')
2346 def CheckComment(comment, filename, linenum, error):
2347 """Checks for common mistakes in TODO comments.
2349 Args:
2350 comment: The text of the comment from the line in question.
2351 filename: The name of the current file.
2352 linenum: The number of the line to check.
2353 error: The function to call with any errors found.
2355 match = _RE_PATTERN_TODO.match(comment)
2356 if match:
2357 # One whitespace is correct; zero whitespace is handled elsewhere.
2358 leading_whitespace = match.group(1)
2359 if len(leading_whitespace) > 1:
2360 error(filename, linenum, 'whitespace/todo', 2,
2361 'Too many spaces before TODO')
2363 username = match.group(2)
2364 if not username:
2365 error(filename, linenum, 'readability/todo', 2,
2366 'Missing username in TODO; it should look like '
2367 '"// TODO(my_username): Stuff."')
2369 middle_whitespace = match.group(3)
2370 # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison
2371 if middle_whitespace != ' ' and middle_whitespace != '':
2372 error(filename, linenum, 'whitespace/todo', 2,
2373 'TODO(my_username) should be followed by a space')
2375 def CheckAccess(filename, clean_lines, linenum, nesting_state, error):
2376 """Checks for improper use of DISALLOW* macros.
2378 Args:
2379 filename: The name of the current file.
2380 clean_lines: A CleansedLines instance containing the file.
2381 linenum: The number of the line to check.
2382 nesting_state: A _NestingState instance which maintains information about
2383 the current stack of nested blocks being parsed.
2384 error: The function to call with any errors found.
2386 line = clean_lines.elided[linenum] # get rid of comments and strings
2388 matched = Match((r'\s*(DISALLOW_COPY_AND_ASSIGN|'
2389 r'DISALLOW_EVIL_CONSTRUCTORS|'
2390 r'DISALLOW_IMPLICIT_CONSTRUCTORS)'), line)
2391 if not matched:
2392 return
2393 if nesting_state.stack and isinstance(nesting_state.stack[-1], _ClassInfo):
2394 if nesting_state.stack[-1].access != 'private':
2395 error(filename, linenum, 'readability/constructors', 3,
2396 '%s must be in the private: section' % matched.group(1))
2398 else:
2399 # Found DISALLOW* macro outside a class declaration, or perhaps it
2400 # was used inside a function when it should have been part of the
2401 # class declaration. We could issue a warning here, but it
2402 # probably resulted in a compiler error already.
2403 pass
2406 def FindNextMatchingAngleBracket(clean_lines, linenum, init_suffix):
2407 """Find the corresponding > to close a template.
2409 Args:
2410 clean_lines: A CleansedLines instance containing the file.
2411 linenum: Current line number.
2412 init_suffix: Remainder of the current line after the initial <.
2414 Returns:
2415 True if a matching bracket exists.
2417 line = init_suffix
2418 nesting_stack = ['<']
2419 while True:
2420 # Find the next operator that can tell us whether < is used as an
2421 # opening bracket or as a less-than operator. We only want to
2422 # warn on the latter case.
2424 # We could also check all other operators and terminate the search
2425 # early, e.g. if we got something like this "a<b+c", the "<" is
2426 # most likely a less-than operator, but then we will get false
2427 # positives for default arguments and other template expressions.
2428 match = Search(r'^[^<>(),;\[\]]*([<>(),;\[\]])(.*)$', line)
2429 if match:
2430 # Found an operator, update nesting stack
2431 operator = match.group(1)
2432 line = match.group(2)
2434 if nesting_stack[-1] == '<':
2435 # Expecting closing angle bracket
2436 if operator in ('<', '(', '['):
2437 nesting_stack.append(operator)
2438 elif operator == '>':
2439 nesting_stack.pop()
2440 if not nesting_stack:
2441 # Found matching angle bracket
2442 return True
2443 elif operator == ',':
2444 # Got a comma after a bracket, this is most likely a template
2445 # argument. We have not seen a closing angle bracket yet, but
2446 # it's probably a few lines later if we look for it, so just
2447 # return early here.
2448 return True
2449 else:
2450 # Got some other operator.
2451 return False
2453 else:
2454 # Expecting closing parenthesis or closing bracket
2455 if operator in ('<', '(', '['):
2456 nesting_stack.append(operator)
2457 elif operator in (')', ']'):
2458 # We don't bother checking for matching () or []. If we got
2459 # something like (] or [), it would have been a syntax error.
2460 nesting_stack.pop()
2462 else:
2463 # Scan the next line
2464 linenum += 1
2465 if linenum >= len(clean_lines.elided):
2466 break
2467 line = clean_lines.elided[linenum]
2469 # Exhausted all remaining lines and still no matching angle bracket.
2470 # Most likely the input was incomplete, otherwise we should have
2471 # seen a semicolon and returned early.
2472 return True
2475 def FindPreviousMatchingAngleBracket(clean_lines, linenum, init_prefix):
2476 """Find the corresponding < that started a template.
2478 Args:
2479 clean_lines: A CleansedLines instance containing the file.
2480 linenum: Current line number.
2481 init_prefix: Part of the current line before the initial >.
2483 Returns:
2484 True if a matching bracket exists.
2486 line = init_prefix
2487 nesting_stack = ['>']
2488 while True:
2489 # Find the previous operator
2490 match = Search(r'^(.*)([<>(),;\[\]])[^<>(),;\[\]]*$', line)
2491 if match:
2492 # Found an operator, update nesting stack
2493 operator = match.group(2)
2494 line = match.group(1)
2496 if nesting_stack[-1] == '>':
2497 # Expecting opening angle bracket
2498 if operator in ('>', ')', ']'):
2499 nesting_stack.append(operator)
2500 elif operator == '<':
2501 nesting_stack.pop()
2502 if not nesting_stack:
2503 # Found matching angle bracket
2504 return True
2505 elif operator == ',':
2506 # Got a comma before a bracket, this is most likely a
2507 # template argument. The opening angle bracket is probably
2508 # there if we look for it, so just return early here.
2509 return True
2510 else:
2511 # Got some other operator.
2512 return False
2514 else:
2515 # Expecting opening parenthesis or opening bracket
2516 if operator in ('>', ')', ']'):
2517 nesting_stack.append(operator)
2518 elif operator in ('(', '['):
2519 nesting_stack.pop()
2521 else:
2522 # Scan the previous line
2523 linenum -= 1
2524 if linenum < 0:
2525 break
2526 line = clean_lines.elided[linenum]
2528 # Exhausted all earlier lines and still no matching angle bracket.
2529 return False
2532 def CheckSpacing(filename, clean_lines, linenum, nesting_state, error):
2533 """Checks for the correctness of various spacing issues in the code.
2535 Things we check for: spaces around operators, spaces after
2536 if/for/while/switch, no spaces around parens in function calls, two
2537 spaces between code and comment, don't start a block with a blank
2538 line, don't end a function with a blank line, don't add a blank line
2539 after public/protected/private, don't have too many blank lines in a row.
2541 Args:
2542 filename: The name of the current file.
2543 clean_lines: A CleansedLines instance containing the file.
2544 linenum: The number of the line to check.
2545 nesting_state: A _NestingState instance which maintains information about
2546 the current stack of nested blocks being parsed.
2547 error: The function to call with any errors found.
2550 # Don't use "elided" lines here, otherwise we can't check commented lines.
2551 # Don't want to use "raw" either, because we don't want to check inside C++11
2552 # raw strings,
2553 raw = clean_lines.lines_without_raw_strings
2554 line = raw[linenum]
2556 # Before nixing comments, check if the line is blank for no good
2557 # reason. This includes the first line after a block is opened, and
2558 # blank lines at the end of a function (ie, right before a line like '}'
2560 # Skip all the blank line checks if we are immediately inside a
2561 # namespace body. In other words, don't issue blank line warnings
2562 # for this block:
2563 # namespace {
2567 # A warning about missing end of namespace comments will be issued instead.
2568 if IsBlankLine(line) and not nesting_state.InNamespaceBody():
2569 elided = clean_lines.elided
2570 prev_line = elided[linenum - 1]
2571 prevbrace = prev_line.rfind('{')
2572 # TODO(unknown): Don't complain if line before blank line, and line after,
2573 # both start with alnums and are indented the same amount.
2574 # This ignores whitespace at the start of a namespace block
2575 # because those are not usually indented.
2576 if prevbrace != -1 and prev_line[prevbrace:].find('}') == -1:
2577 # OK, we have a blank line at the start of a code block. Before we
2578 # complain, we check if it is an exception to the rule: The previous
2579 # non-empty line has the parameters of a function header that are indented
2580 # 4 spaces (because they did not fit in a 80 column line when placed on
2581 # the same line as the function name). We also check for the case where
2582 # the previous line is indented 6 spaces, which may happen when the
2583 # initializers of a constructor do not fit into a 80 column line.
2584 exception = False
2585 if Match(r' {6}\w', prev_line): # Initializer list?
2586 # We are looking for the opening column of initializer list, which
2587 # should be indented 4 spaces to cause 6 space indentation afterwards.
2588 search_position = linenum-2
2589 while (search_position >= 0
2590 and Match(r' {6}\w', elided[search_position])):
2591 search_position -= 1
2592 exception = (search_position >= 0
2593 and elided[search_position][:5] == ' :')
2594 else:
2595 # Search for the function arguments or an initializer list. We use a
2596 # simple heuristic here: If the line is indented 4 spaces; and we have a
2597 # closing paren, without the opening paren, followed by an opening brace
2598 # or colon (for initializer lists) we assume that it is the last line of
2599 # a function header. If we have a colon indented 4 spaces, it is an
2600 # initializer list.
2601 exception = (Match(r' {4}\w[^\(]*\)\s*(const\s*)?(\{\s*$|:)',
2602 prev_line)
2603 or Match(r' {4}:', prev_line))
2605 if not exception:
2606 error(filename, linenum, 'whitespace/blank_line', 2,
2607 'Redundant blank line at the start of a code block '
2608 'should be deleted.')
2609 # Ignore blank lines at the end of a block in a long if-else
2610 # chain, like this:
2611 # if (condition1) {
2612 # // Something followed by a blank line
2614 # } else if (condition2) {
2615 # // Something else
2617 if linenum + 1 < clean_lines.NumLines():
2618 next_line = raw[linenum + 1]
2619 if (next_line
2620 and Match(r'\s*}', next_line)
2621 and next_line.find('} else ') == -1):
2622 error(filename, linenum, 'whitespace/blank_line', 3,
2623 'Redundant blank line at the end of a code block '
2624 'should be deleted.')
2626 matched = Match(r'\s*(public|protected|private):', prev_line)
2627 if matched:
2628 error(filename, linenum, 'whitespace/blank_line', 3,
2629 'Do not leave a blank line after "%s:"' % matched.group(1))
2631 # Next, we complain if there's a comment too near the text
2632 commentpos = line.find('//')
2633 if commentpos != -1:
2634 # Check if the // may be in quotes. If so, ignore it
2635 # Comparisons made explicit for clarity -- pylint: disable=g-explicit-bool-comparison
2636 if (line.count('"', 0, commentpos) -
2637 line.count('\\"', 0, commentpos)) % 2 == 0: # not in quotes
2638 # Allow one space for new scopes, two spaces otherwise:
2639 if (not Match(r'^\s*{ //', line) and
2640 ((commentpos >= 1 and
2641 line[commentpos-1] not in string.whitespace) or
2642 (commentpos >= 2 and
2643 line[commentpos-2] not in string.whitespace))):
2644 error(filename, linenum, 'whitespace/comments', 2,
2645 'At least two spaces is best between code and comments')
2646 # There should always be a space between the // and the comment
2647 commentend = commentpos + 2
2648 if commentend < len(line) and not line[commentend] == ' ':
2649 # but some lines are exceptions -- e.g. if they're big
2650 # comment delimiters like:
2651 # //----------------------------------------------------------
2652 # or are an empty C++ style Doxygen comment, like:
2653 # ///
2654 # or C++ style Doxygen comments placed after the variable:
2655 # ///< Header comment
2656 # //!< Header comment
2657 # or they begin with multiple slashes followed by a space:
2658 # //////// Header comment
2659 match = (Search(r'[=/-]{4,}\s*$', line[commentend:]) or
2660 Search(r'^/$', line[commentend:]) or
2661 Search(r'^!< ', line[commentend:]) or
2662 Search(r'^/< ', line[commentend:]) or
2663 Search(r'^/+ ', line[commentend:]))
2664 if not match:
2665 error(filename, linenum, 'whitespace/comments', 4,
2666 'Should have a space between // and comment')
2667 CheckComment(line[commentpos:], filename, linenum, error)
2669 line = clean_lines.elided[linenum] # get rid of comments and strings
2671 # Don't try to do spacing checks for operator methods
2672 line = re.sub(r'operator(==|!=|<|<<|<=|>=|>>|>)\(', 'operator\(', line)
2674 # We allow no-spaces around = within an if: "if ( (a=Foo()) == 0 )".
2675 # Otherwise not. Note we only check for non-spaces on *both* sides;
2676 # sometimes people put non-spaces on one side when aligning ='s among
2677 # many lines (not that this is behavior that I approve of...)
2678 if Search(r'[\w.]=[\w.]', line) and not Search(r'\b(if|while) ', line):
2679 error(filename, linenum, 'whitespace/operators', 4,
2680 'Missing spaces around =')
2682 # It's ok not to have spaces around binary operators like + - * /, but if
2683 # there's too little whitespace, we get concerned. It's hard to tell,
2684 # though, so we punt on this one for now. TODO.
2686 # You should always have whitespace around binary operators.
2688 # Check <= and >= first to avoid false positives with < and >, then
2689 # check non-include lines for spacing around < and >.
2690 match = Search(r'[^<>=!\s](==|!=|<=|>=)[^<>=!\s]', line)
2691 if match:
2692 error(filename, linenum, 'whitespace/operators', 3,
2693 'Missing spaces around %s' % match.group(1))
2694 # We allow no-spaces around << when used like this: 10<<20, but
2695 # not otherwise (particularly, not when used as streams)
2696 # Also ignore using ns::operator<<;
2697 match = Search(r'(operator|\S)(?:L|UL|ULL|l|ul|ull)?<<(\S)', line)
2698 if (match and
2699 not (match.group(1).isdigit() and match.group(2).isdigit()) and
2700 not (match.group(1) == 'operator' and match.group(2) == ';')):
2701 error(filename, linenum, 'whitespace/operators', 3,
2702 'Missing spaces around <<')
2703 elif not Match(r'#.*include', line):
2704 # Avoid false positives on ->
2705 reduced_line = line.replace('->', '')
2707 # Look for < that is not surrounded by spaces. This is only
2708 # triggered if both sides are missing spaces, even though
2709 # technically should should flag if at least one side is missing a
2710 # space. This is done to avoid some false positives with shifts.
2711 match = Search(r'[^\s<]<([^\s=<].*)', reduced_line)
2712 if (match and
2713 not FindNextMatchingAngleBracket(clean_lines, linenum, match.group(1))):
2714 error(filename, linenum, 'whitespace/operators', 3,
2715 'Missing spaces around <')
2717 # Look for > that is not surrounded by spaces. Similar to the
2718 # above, we only trigger if both sides are missing spaces to avoid
2719 # false positives with shifts.
2720 match = Search(r'^(.*[^\s>])>[^\s=>]', reduced_line)
2721 if (match and
2722 not FindPreviousMatchingAngleBracket(clean_lines, linenum,
2723 match.group(1))):
2724 error(filename, linenum, 'whitespace/operators', 3,
2725 'Missing spaces around >')
2727 # We allow no-spaces around >> for almost anything. This is because
2728 # C++11 allows ">>" to close nested templates, which accounts for
2729 # most cases when ">>" is not followed by a space.
2731 # We still warn on ">>" followed by alpha character, because that is
2732 # likely due to ">>" being used for right shifts, e.g.:
2733 # value >> alpha
2735 # When ">>" is used to close templates, the alphanumeric letter that
2736 # follows would be part of an identifier, and there should still be
2737 # a space separating the template type and the identifier.
2738 # type<type<type>> alpha
2739 match = Search(r'>>[a-zA-Z_]', line)
2740 if match:
2741 error(filename, linenum, 'whitespace/operators', 3,
2742 'Missing spaces around >>')
2744 # There shouldn't be space around unary operators
2745 match = Search(r'(!\s|~\s|[\s]--[\s;]|[\s]\+\+[\s;])', line)
2746 if match:
2747 error(filename, linenum, 'whitespace/operators', 4,
2748 'Extra space for operator %s' % match.group(1))
2750 # A pet peeve of mine: no spaces after an if, while, switch, or for
2751 match = Search(r' (if\(|for\(|while\(|switch\()', line)
2752 if match:
2753 error(filename, linenum, 'whitespace/parens', 5,
2754 'Missing space before ( in %s' % match.group(1))
2756 # For if/for/while/switch, the left and right parens should be
2757 # consistent about how many spaces are inside the parens, and
2758 # there should either be zero or one spaces inside the parens.
2759 # We don't want: "if ( foo)" or "if ( foo )".
2760 # Exception: "for ( ; foo; bar)" and "for (foo; bar; )" are allowed.
2761 match = Search(r'\b(if|for|while|switch)\s*'
2762 r'\(([ ]*)(.).*[^ ]+([ ]*)\)\s*{\s*$',
2763 line)
2764 if match:
2765 if len(match.group(2)) != len(match.group(4)):
2766 if not (match.group(3) == ';' and
2767 len(match.group(2)) == 1 + len(match.group(4)) or
2768 not match.group(2) and Search(r'\bfor\s*\(.*; \)', line)):
2769 error(filename, linenum, 'whitespace/parens', 5,
2770 'Mismatching spaces inside () in %s' % match.group(1))
2771 if len(match.group(2)) not in [0, 1]:
2772 error(filename, linenum, 'whitespace/parens', 5,
2773 'Should have zero or one spaces inside ( and ) in %s' %
2774 match.group(1))
2776 # You should always have a space after a comma (either as fn arg or operator)
2778 # This does not apply when the non-space character following the
2779 # comma is another comma, since the only time when that happens is
2780 # for empty macro arguments.
2782 # We run this check in two passes: first pass on elided lines to
2783 # verify that lines contain missing whitespaces, second pass on raw
2784 # lines to confirm that those missing whitespaces are not due to
2785 # elided comments.
2786 if Search(r',[^,\s]', line) and Search(r',[^,\s]', raw[linenum]):
2787 error(filename, linenum, 'whitespace/comma', 3,
2788 'Missing space after ,')
2790 # You should always have a space after a semicolon
2791 # except for few corner cases
2792 # TODO(unknown): clarify if 'if (1) { return 1;}' is requires one more
2793 # space after ;
2794 if Search(r';[^\s};\\)/]', line):
2795 error(filename, linenum, 'whitespace/semicolon', 3,
2796 'Missing space after ;')
2798 # Next we will look for issues with function calls.
2799 CheckSpacingForFunctionCall(filename, line, linenum, error)
2801 # Except after an opening paren, or after another opening brace (in case of
2802 # an initializer list, for instance), you should have spaces before your
2803 # braces. And since you should never have braces at the beginning of a line,
2804 # this is an easy test.
2805 match = Match(r'^(.*[^ ({]){', line)
2806 if match:
2807 # Try a bit harder to check for brace initialization. This
2808 # happens in one of the following forms:
2809 # Constructor() : initializer_list_{} { ... }
2810 # Constructor{}.MemberFunction()
2811 # Type variable{};
2812 # FunctionCall(type{}, ...);
2813 # LastArgument(..., type{});
2814 # LOG(INFO) << type{} << " ...";
2815 # map_of_type[{...}] = ...;
2817 # We check for the character following the closing brace, and
2818 # silence the warning if it's one of those listed above, i.e.
2819 # "{.;,)<]".
2821 # To account for nested initializer list, we allow any number of
2822 # closing braces up to "{;,)<". We can't simply silence the
2823 # warning on first sight of closing brace, because that would
2824 # cause false negatives for things that are not initializer lists.
2825 # Silence this: But not this:
2826 # Outer{ if (...) {
2827 # Inner{...} if (...){ // Missing space before {
2828 # }; }
2830 # There is a false negative with this approach if people inserted
2831 # spurious semicolons, e.g. "if (cond){};", but we will catch the
2832 # spurious semicolon with a separate check.
2833 (endline, endlinenum, endpos) = CloseExpression(
2834 clean_lines, linenum, len(match.group(1)))
2835 trailing_text = ''
2836 if endpos > -1:
2837 trailing_text = endline[endpos:]
2838 for offset in xrange(endlinenum + 1,
2839 min(endlinenum + 3, clean_lines.NumLines() - 1)):
2840 trailing_text += clean_lines.elided[offset]
2841 if not Match(r'^[\s}]*[{.;,)<\]]', trailing_text):
2842 error(filename, linenum, 'whitespace/braces', 5,
2843 'Missing space before {')
2845 # Make sure '} else {' has spaces.
2846 if Search(r'}else', line):
2847 error(filename, linenum, 'whitespace/braces', 5,
2848 'Missing space before else')
2850 # You shouldn't have spaces before your brackets, except maybe after
2851 # 'delete []' or 'new char * []'.
2852 if Search(r'\w\s+\[', line) and not Search(r'delete\s+\[', line):
2853 error(filename, linenum, 'whitespace/braces', 5,
2854 'Extra space before [')
2856 # You shouldn't have a space before a semicolon at the end of the line.
2857 # There's a special case for "for" since the style guide allows space before
2858 # the semicolon there.
2859 if Search(r':\s*;\s*$', line):
2860 error(filename, linenum, 'whitespace/semicolon', 5,
2861 'Semicolon defining empty statement. Use {} instead.')
2862 elif Search(r'^\s*;\s*$', line):
2863 error(filename, linenum, 'whitespace/semicolon', 5,
2864 'Line contains only semicolon. If this should be an empty statement, '
2865 'use {} instead.')
2866 elif (Search(r'\s+;\s*$', line) and
2867 not Search(r'\bfor\b', line)):
2868 error(filename, linenum, 'whitespace/semicolon', 5,
2869 'Extra space before last semicolon. If this should be an empty '
2870 'statement, use {} instead.')
2872 # In range-based for, we wanted spaces before and after the colon, but
2873 # not around "::" tokens that might appear.
2874 if (Search('for *\(.*[^:]:[^: ]', line) or
2875 Search('for *\(.*[^: ]:[^:]', line)):
2876 error(filename, linenum, 'whitespace/forcolon', 2,
2877 'Missing space around colon in range-based for loop')
2880 def CheckSectionSpacing(filename, clean_lines, class_info, linenum, error):
2881 """Checks for additional blank line issues related to sections.
2883 Currently the only thing checked here is blank line before protected/private.
2885 Args:
2886 filename: The name of the current file.
2887 clean_lines: A CleansedLines instance containing the file.
2888 class_info: A _ClassInfo objects.
2889 linenum: The number of the line to check.
2890 error: The function to call with any errors found.
2892 # Skip checks if the class is small, where small means 25 lines or less.
2893 # 25 lines seems like a good cutoff since that's the usual height of
2894 # terminals, and any class that can't fit in one screen can't really
2895 # be considered "small".
2897 # Also skip checks if we are on the first line. This accounts for
2898 # classes that look like
2899 # class Foo { public: ... };
2901 # If we didn't find the end of the class, last_line would be zero,
2902 # and the check will be skipped by the first condition.
2903 if (class_info.last_line - class_info.starting_linenum <= 24 or
2904 linenum <= class_info.starting_linenum):
2905 return
2907 matched = Match(r'\s*(public|protected|private):', clean_lines.lines[linenum])
2908 if matched:
2909 # Issue warning if the line before public/protected/private was
2910 # not a blank line, but don't do this if the previous line contains
2911 # "class" or "struct". This can happen two ways:
2912 # - We are at the beginning of the class.
2913 # - We are forward-declaring an inner class that is semantically
2914 # private, but needed to be public for implementation reasons.
2915 # Also ignores cases where the previous line ends with a backslash as can be
2916 # common when defining classes in C macros.
2917 prev_line = clean_lines.lines[linenum - 1]
2918 if (not IsBlankLine(prev_line) and
2919 not Search(r'\b(class|struct)\b', prev_line) and
2920 not Search(r'\\$', prev_line)):
2921 # Try a bit harder to find the beginning of the class. This is to
2922 # account for multi-line base-specifier lists, e.g.:
2923 # class Derived
2924 # : public Base {
2925 end_class_head = class_info.starting_linenum
2926 for i in range(class_info.starting_linenum, linenum):
2927 if Search(r'\{\s*$', clean_lines.lines[i]):
2928 end_class_head = i
2929 break
2930 if end_class_head < linenum - 1:
2931 error(filename, linenum, 'whitespace/blank_line', 3,
2932 '"%s:" should be preceded by a blank line' % matched.group(1))
2935 def GetPreviousNonBlankLine(clean_lines, linenum):
2936 """Return the most recent non-blank line and its line number.
2938 Args:
2939 clean_lines: A CleansedLines instance containing the file contents.
2940 linenum: The number of the line to check.
2942 Returns:
2943 A tuple with two elements. The first element is the contents of the last
2944 non-blank line before the current line, or the empty string if this is the
2945 first non-blank line. The second is the line number of that line, or -1
2946 if this is the first non-blank line.
2949 prevlinenum = linenum - 1
2950 while prevlinenum >= 0:
2951 prevline = clean_lines.elided[prevlinenum]
2952 if not IsBlankLine(prevline): # if not a blank line...
2953 return (prevline, prevlinenum)
2954 prevlinenum -= 1
2955 return ('', -1)
2958 def CheckBraces(filename, clean_lines, linenum, error):
2959 """Looks for misplaced braces (e.g. at the end of line).
2961 Args:
2962 filename: The name of the current file.
2963 clean_lines: A CleansedLines instance containing the file.
2964 linenum: The number of the line to check.
2965 error: The function to call with any errors found.
2968 line = clean_lines.elided[linenum] # get rid of comments and strings
2970 if Match(r'\s*{\s*$', line):
2971 # We allow an open brace to start a line in the case where someone is using
2972 # braces in a block to explicitly create a new scope, which is commonly used
2973 # to control the lifetime of stack-allocated variables. Braces are also
2974 # used for brace initializers inside function calls. We don't detect this
2975 # perfectly: we just don't complain if the last non-whitespace character on
2976 # the previous non-blank line is ',', ';', ':', '(', '{', or '}', or if the
2977 # previous line starts a preprocessor block.
2978 prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0]
2979 if (not Search(r'[,;:}{(]\s*$', prevline) and
2980 not Match(r'\s*#', prevline)):
2981 error(filename, linenum, 'whitespace/braces', 4,
2982 '{ should almost always be at the end of the previous line')
2984 # An else clause should be on the same line as the preceding closing brace.
2985 if Match(r'\s*else\s*', line):
2986 prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0]
2987 if Match(r'\s*}\s*$', prevline):
2988 error(filename, linenum, 'whitespace/newline', 4,
2989 'An else should appear on the same line as the preceding }')
2991 # If braces come on one side of an else, they should be on both.
2992 # However, we have to worry about "else if" that spans multiple lines!
2993 if Search(r'}\s*else[^{]*$', line) or Match(r'[^}]*else\s*{', line):
2994 if Search(r'}\s*else if([^{]*)$', line): # could be multi-line if
2995 # find the ( after the if
2996 pos = line.find('else if')
2997 pos = line.find('(', pos)
2998 if pos > 0:
2999 (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos)
3000 if endline[endpos:].find('{') == -1: # must be brace after if
3001 error(filename, linenum, 'readability/braces', 5,
3002 'If an else has a brace on one side, it should have it on both')
3003 else: # common case: else not followed by a multi-line if
3004 error(filename, linenum, 'readability/braces', 5,
3005 'If an else has a brace on one side, it should have it on both')
3007 # Likewise, an else should never have the else clause on the same line
3008 if Search(r'\belse [^\s{]', line) and not Search(r'\belse if\b', line):
3009 error(filename, linenum, 'whitespace/newline', 4,
3010 'Else clause should never be on same line as else (use 2 lines)')
3012 # In the same way, a do/while should never be on one line
3013 if Match(r'\s*do [^\s{]', line):
3014 error(filename, linenum, 'whitespace/newline', 4,
3015 'do/while clauses should not be on a single line')
3017 # Block bodies should not be followed by a semicolon. Due to C++11
3018 # brace initialization, there are more places where semicolons are
3019 # required than not, so we use a whitelist approach to check these
3020 # rather than a blacklist. These are the places where "};" should
3021 # be replaced by just "}":
3022 # 1. Some flavor of block following closing parenthesis:
3023 # for (;;) {};
3024 # while (...) {};
3025 # switch (...) {};
3026 # Function(...) {};
3027 # if (...) {};
3028 # if (...) else if (...) {};
3030 # 2. else block:
3031 # if (...) else {};
3033 # 3. const member function:
3034 # Function(...) const {};
3036 # 4. Block following some statement:
3037 # x = 42;
3038 # {};
3040 # 5. Block at the beginning of a function:
3041 # Function(...) {
3042 # {};
3045 # Note that naively checking for the preceding "{" will also match
3046 # braces inside multi-dimensional arrays, but this is fine since
3047 # that expression will not contain semicolons.
3049 # 6. Block following another block:
3050 # while (true) {}
3051 # {};
3053 # 7. End of namespaces:
3054 # namespace {};
3056 # These semicolons seems far more common than other kinds of
3057 # redundant semicolons, possibly due to people converting classes
3058 # to namespaces. For now we do not warn for this case.
3060 # Try matching case 1 first.
3061 match = Match(r'^(.*\)\s*)\{', line)
3062 if match:
3063 # Matched closing parenthesis (case 1). Check the token before the
3064 # matching opening parenthesis, and don't warn if it looks like a
3065 # macro. This avoids these false positives:
3066 # - macro that defines a base class
3067 # - multi-line macro that defines a base class
3068 # - macro that defines the whole class-head
3070 # But we still issue warnings for macros that we know are safe to
3071 # warn, specifically:
3072 # - TEST, TEST_F, TEST_P, MATCHER, MATCHER_P
3073 # - TYPED_TEST
3074 # - INTERFACE_DEF
3075 # - EXCLUSIVE_LOCKS_REQUIRED, SHARED_LOCKS_REQUIRED, LOCKS_EXCLUDED:
3077 # We implement a whitelist of safe macros instead of a blacklist of
3078 # unsafe macros, even though the latter appears less frequently in
3079 # google code and would have been easier to implement. This is because
3080 # the downside for getting the whitelist wrong means some extra
3081 # semicolons, while the downside for getting the blacklist wrong
3082 # would result in compile errors.
3084 # In addition to macros, we also don't want to warn on compound
3085 # literals.
3086 closing_brace_pos = match.group(1).rfind(')')
3087 opening_parenthesis = ReverseCloseExpression(
3088 clean_lines, linenum, closing_brace_pos)
3089 if opening_parenthesis[2] > -1:
3090 line_prefix = opening_parenthesis[0][0:opening_parenthesis[2]]
3091 macro = Search(r'\b([A-Z_]+)\s*$', line_prefix)
3092 if ((macro and
3093 macro.group(1) not in (
3094 'TEST', 'TEST_F', 'MATCHER', 'MATCHER_P', 'TYPED_TEST',
3095 'EXCLUSIVE_LOCKS_REQUIRED', 'SHARED_LOCKS_REQUIRED',
3096 'LOCKS_EXCLUDED', 'INTERFACE_DEF')) or
3097 Search(r'\s+=\s*$', line_prefix)):
3098 match = None
3100 else:
3101 # Try matching cases 2-3.
3102 match = Match(r'^(.*(?:else|\)\s*const)\s*)\{', line)
3103 if not match:
3104 # Try matching cases 4-6. These are always matched on separate lines.
3106 # Note that we can't simply concatenate the previous line to the
3107 # current line and do a single match, otherwise we may output
3108 # duplicate warnings for the blank line case:
3109 # if (cond) {
3110 # // blank line
3112 prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0]
3113 if prevline and Search(r'[;{}]\s*$', prevline):
3114 match = Match(r'^(\s*)\{', line)
3116 # Check matching closing brace
3117 if match:
3118 (endline, endlinenum, endpos) = CloseExpression(
3119 clean_lines, linenum, len(match.group(1)))
3120 if endpos > -1 and Match(r'^\s*;', endline[endpos:]):
3121 # Current {} pair is eligible for semicolon check, and we have found
3122 # the redundant semicolon, output warning here.
3124 # Note: because we are scanning forward for opening braces, and
3125 # outputting warnings for the matching closing brace, if there are
3126 # nested blocks with trailing semicolons, we will get the error
3127 # messages in reversed order.
3128 error(filename, endlinenum, 'readability/braces', 4,
3129 "You don't need a ; after a }")
3132 def CheckEmptyBlockBody(filename, clean_lines, linenum, error):
3133 """Look for empty loop/conditional body with only a single semicolon.
3135 Args:
3136 filename: The name of the current file.
3137 clean_lines: A CleansedLines instance containing the file.
3138 linenum: The number of the line to check.
3139 error: The function to call with any errors found.
3142 # Search for loop keywords at the beginning of the line. Because only
3143 # whitespaces are allowed before the keywords, this will also ignore most
3144 # do-while-loops, since those lines should start with closing brace.
3146 # We also check "if" blocks here, since an empty conditional block
3147 # is likely an error.
3148 line = clean_lines.elided[linenum]
3149 matched = Match(r'\s*(for|while|if)\s*\(', line)
3150 if matched:
3151 # Find the end of the conditional expression
3152 (end_line, end_linenum, end_pos) = CloseExpression(
3153 clean_lines, linenum, line.find('('))
3155 # Output warning if what follows the condition expression is a semicolon.
3156 # No warning for all other cases, including whitespace or newline, since we
3157 # have a separate check for semicolons preceded by whitespace.
3158 if end_pos >= 0 and Match(r';', end_line[end_pos:]):
3159 if matched.group(1) == 'if':
3160 error(filename, end_linenum, 'whitespace/empty_conditional_body', 5,
3161 'Empty conditional bodies should use {}')
3162 else:
3163 error(filename, end_linenum, 'whitespace/empty_loop_body', 5,
3164 'Empty loop bodies should use {} or continue')
3167 def CheckCheck(filename, clean_lines, linenum, error):
3168 """Checks the use of CHECK and EXPECT macros.
3170 Args:
3171 filename: The name of the current file.
3172 clean_lines: A CleansedLines instance containing the file.
3173 linenum: The number of the line to check.
3174 error: The function to call with any errors found.
3177 # Decide the set of replacement macros that should be suggested
3178 lines = clean_lines.elided
3179 check_macro = None
3180 start_pos = -1
3181 for macro in _CHECK_MACROS:
3182 i = lines[linenum].find(macro)
3183 if i >= 0:
3184 check_macro = macro
3186 # Find opening parenthesis. Do a regular expression match here
3187 # to make sure that we are matching the expected CHECK macro, as
3188 # opposed to some other macro that happens to contain the CHECK
3189 # substring.
3190 matched = Match(r'^(.*\b' + check_macro + r'\s*)\(', lines[linenum])
3191 if not matched:
3192 continue
3193 start_pos = len(matched.group(1))
3194 break
3195 if not check_macro or start_pos < 0:
3196 # Don't waste time here if line doesn't contain 'CHECK' or 'EXPECT'
3197 return
3199 # Find end of the boolean expression by matching parentheses
3200 (last_line, end_line, end_pos) = CloseExpression(
3201 clean_lines, linenum, start_pos)
3202 if end_pos < 0:
3203 return
3204 if linenum == end_line:
3205 expression = lines[linenum][start_pos + 1:end_pos - 1]
3206 else:
3207 expression = lines[linenum][start_pos + 1:]
3208 for i in xrange(linenum + 1, end_line):
3209 expression += lines[i]
3210 expression += last_line[0:end_pos - 1]
3212 # Parse expression so that we can take parentheses into account.
3213 # This avoids false positives for inputs like "CHECK((a < 4) == b)",
3214 # which is not replaceable by CHECK_LE.
3215 lhs = ''
3216 rhs = ''
3217 operator = None
3218 while expression:
3219 matched = Match(r'^\s*(<<|<<=|>>|>>=|->\*|->|&&|\|\||'
3220 r'==|!=|>=|>|<=|<|\()(.*)$', expression)
3221 if matched:
3222 token = matched.group(1)
3223 if token == '(':
3224 # Parenthesized operand
3225 expression = matched.group(2)
3226 (end, _) = FindEndOfExpressionInLine(expression, 0, 1, '(', ')')
3227 if end < 0:
3228 return # Unmatched parenthesis
3229 lhs += '(' + expression[0:end]
3230 expression = expression[end:]
3231 elif token in ('&&', '||'):
3232 # Logical and/or operators. This means the expression
3233 # contains more than one term, for example:
3234 # CHECK(42 < a && a < b);
3236 # These are not replaceable with CHECK_LE, so bail out early.
3237 return
3238 elif token in ('<<', '<<=', '>>', '>>=', '->*', '->'):
3239 # Non-relational operator
3240 lhs += token
3241 expression = matched.group(2)
3242 else:
3243 # Relational operator
3244 operator = token
3245 rhs = matched.group(2)
3246 break
3247 else:
3248 # Unparenthesized operand. Instead of appending to lhs one character
3249 # at a time, we do another regular expression match to consume several
3250 # characters at once if possible. Trivial benchmark shows that this
3251 # is more efficient when the operands are longer than a single
3252 # character, which is generally the case.
3253 matched = Match(r'^([^-=!<>()&|]+)(.*)$', expression)
3254 if not matched:
3255 matched = Match(r'^(\s*\S)(.*)$', expression)
3256 if not matched:
3257 break
3258 lhs += matched.group(1)
3259 expression = matched.group(2)
3261 # Only apply checks if we got all parts of the boolean expression
3262 if not (lhs and operator and rhs):
3263 return
3265 # Check that rhs do not contain logical operators. We already know
3266 # that lhs is fine since the loop above parses out && and ||.
3267 if rhs.find('&&') > -1 or rhs.find('||') > -1:
3268 return
3270 # At least one of the operands must be a constant literal. This is
3271 # to avoid suggesting replacements for unprintable things like
3272 # CHECK(variable != iterator)
3274 # The following pattern matches decimal, hex integers, strings, and
3275 # characters (in that order).
3276 lhs = lhs.strip()
3277 rhs = rhs.strip()
3278 match_constant = r'^([-+]?(\d+|0[xX][0-9a-fA-F]+)[lLuU]{0,3}|".*"|\'.*\')$'
3279 if Match(match_constant, lhs) or Match(match_constant, rhs):
3280 # Note: since we know both lhs and rhs, we can provide a more
3281 # descriptive error message like:
3282 # Consider using CHECK_EQ(x, 42) instead of CHECK(x == 42)
3283 # Instead of:
3284 # Consider using CHECK_EQ instead of CHECK(a == b)
3286 # We are still keeping the less descriptive message because if lhs
3287 # or rhs gets long, the error message might become unreadable.
3288 error(filename, linenum, 'readability/check', 2,
3289 'Consider using %s instead of %s(a %s b)' % (
3290 _CHECK_REPLACEMENT[check_macro][operator],
3291 check_macro, operator))
3294 def CheckAltTokens(filename, clean_lines, linenum, error):
3295 """Check alternative keywords being used in boolean expressions.
3297 Args:
3298 filename: The name of the current file.
3299 clean_lines: A CleansedLines instance containing the file.
3300 linenum: The number of the line to check.
3301 error: The function to call with any errors found.
3303 line = clean_lines.elided[linenum]
3305 # Avoid preprocessor lines
3306 if Match(r'^\s*#', line):
3307 return
3309 # Last ditch effort to avoid multi-line comments. This will not help
3310 # if the comment started before the current line or ended after the
3311 # current line, but it catches most of the false positives. At least,
3312 # it provides a way to workaround this warning for people who use
3313 # multi-line comments in preprocessor macros.
3315 # TODO(unknown): remove this once cpplint has better support for
3316 # multi-line comments.
3317 if line.find('/*') >= 0 or line.find('*/') >= 0:
3318 return
3320 for match in _ALT_TOKEN_REPLACEMENT_PATTERN.finditer(line):
3321 error(filename, linenum, 'readability/alt_tokens', 2,
3322 'Use operator %s instead of %s' % (
3323 _ALT_TOKEN_REPLACEMENT[match.group(1)], match.group(1)))
3326 def GetLineWidth(line):
3327 """Determines the width of the line in column positions.
3329 Args:
3330 line: A string, which may be a Unicode string.
3332 Returns:
3333 The width of the line in column positions, accounting for Unicode
3334 combining characters and wide characters.
3336 if isinstance(line, unicode):
3337 width = 0
3338 for uc in unicodedata.normalize('NFC', line):
3339 if unicodedata.east_asian_width(uc) in ('W', 'F'):
3340 width += 2
3341 elif not unicodedata.combining(uc):
3342 width += 1
3343 return width
3344 else:
3345 return len(line)
3348 def CheckStyle(filename, clean_lines, linenum, file_extension, nesting_state,
3349 error):
3350 """Checks rules from the 'C++ style rules' section of cppguide.html.
3352 Most of these rules are hard to test (naming, comment style), but we
3353 do what we can. In particular we check for 2-space indents, line lengths,
3354 tab usage, spaces inside code, etc.
3356 Args:
3357 filename: The name of the current file.
3358 clean_lines: A CleansedLines instance containing the file.
3359 linenum: The number of the line to check.
3360 file_extension: The extension (without the dot) of the filename.
3361 nesting_state: A _NestingState instance which maintains information about
3362 the current stack of nested blocks being parsed.
3363 error: The function to call with any errors found.
3366 # Don't use "elided" lines here, otherwise we can't check commented lines.
3367 # Don't want to use "raw" either, because we don't want to check inside C++11
3368 # raw strings,
3369 raw_lines = clean_lines.lines_without_raw_strings
3370 line = raw_lines[linenum]
3372 if line.find('\t') != -1:
3373 error(filename, linenum, 'whitespace/tab', 1,
3374 'Tab found; better to use spaces')
3376 # One or three blank spaces at the beginning of the line is weird; it's
3377 # hard to reconcile that with 2-space indents.
3378 # NOTE: here are the conditions rob pike used for his tests. Mine aren't
3379 # as sophisticated, but it may be worth becoming so: RLENGTH==initial_spaces
3380 # if(RLENGTH > 20) complain = 0;
3381 # if(match($0, " +(error|private|public|protected):")) complain = 0;
3382 # if(match(prev, "&& *$")) complain = 0;
3383 # if(match(prev, "\\|\\| *$")) complain = 0;
3384 # if(match(prev, "[\",=><] *$")) complain = 0;
3385 # if(match($0, " <<")) complain = 0;
3386 # if(match(prev, " +for \\(")) complain = 0;
3387 # if(prevodd && match(prevprev, " +for \\(")) complain = 0;
3388 initial_spaces = 0
3389 cleansed_line = clean_lines.elided[linenum]
3390 while initial_spaces < len(line) and line[initial_spaces] == ' ':
3391 initial_spaces += 1
3392 if line and line[-1].isspace():
3393 error(filename, linenum, 'whitespace/end_of_line', 4,
3394 'Line ends in whitespace. Consider deleting these extra spaces.')
3395 # There are certain situations we allow one space, notably for section labels
3396 elif ((initial_spaces == 1 or initial_spaces == 3) and
3397 not Match(r'\s*\w+\s*:\s*$', cleansed_line)):
3398 error(filename, linenum, 'whitespace/indent', 3,
3399 'Weird number of spaces at line-start. '
3400 'Are you using a 2-space indent?')
3402 # Check if the line is a header guard.
3403 is_header_guard = False
3404 if file_extension == 'h':
3405 cppvar = GetHeaderGuardCPPVariable(filename)
3406 if (line.startswith('#ifndef %s' % cppvar) or
3407 line.startswith('#define %s' % cppvar) or
3408 line.startswith('#endif // %s' % cppvar)):
3409 is_header_guard = True
3410 # #include lines and header guards can be long, since there's no clean way to
3411 # split them.
3413 # URLs can be long too. It's possible to split these, but it makes them
3414 # harder to cut&paste.
3416 # The "$Id:...$" comment may also get very long without it being the
3417 # developers fault.
3418 if (not line.startswith('#include') and not is_header_guard and
3419 not Match(r'^\s*//.*http(s?)://\S*$', line) and
3420 not Match(r'^// \$Id:.*#[0-9]+ \$$', line)):
3421 line_width = GetLineWidth(line)
3422 extended_length = int((_line_length * 1.25))
3423 if line_width > extended_length:
3424 error(filename, linenum, 'whitespace/line_length', 4,
3425 'Lines should very rarely be longer than %i characters' %
3426 extended_length)
3427 elif line_width > _line_length:
3428 error(filename, linenum, 'whitespace/line_length', 2,
3429 'Lines should be <= %i characters long' % _line_length)
3431 if (cleansed_line.count(';') > 1 and
3432 # for loops are allowed two ;'s (and may run over two lines).
3433 cleansed_line.find('for') == -1 and
3434 (GetPreviousNonBlankLine(clean_lines, linenum)[0].find('for') == -1 or
3435 GetPreviousNonBlankLine(clean_lines, linenum)[0].find(';') != -1) and
3436 # It's ok to have many commands in a switch case that fits in 1 line
3437 not ((cleansed_line.find('case ') != -1 or
3438 cleansed_line.find('default:') != -1) and
3439 cleansed_line.find('break;') != -1)):
3440 error(filename, linenum, 'whitespace/newline', 0,
3441 'More than one command on the same line')
3443 # Some more style checks
3444 CheckBraces(filename, clean_lines, linenum, error)
3445 CheckEmptyBlockBody(filename, clean_lines, linenum, error)
3446 CheckAccess(filename, clean_lines, linenum, nesting_state, error)
3447 CheckSpacing(filename, clean_lines, linenum, nesting_state, error)
3448 CheckCheck(filename, clean_lines, linenum, error)
3449 CheckAltTokens(filename, clean_lines, linenum, error)
3450 classinfo = nesting_state.InnermostClass()
3451 if classinfo:
3452 CheckSectionSpacing(filename, clean_lines, classinfo, linenum, error)
3455 _RE_PATTERN_INCLUDE_NEW_STYLE = re.compile(r'#include +"[^/]+\.h"')
3456 _RE_PATTERN_INCLUDE = re.compile(r'^\s*#\s*include\s*([<"])([^>"]*)[>"].*$')
3457 # Matches the first component of a filename delimited by -s and _s. That is:
3458 # _RE_FIRST_COMPONENT.match('foo').group(0) == 'foo'
3459 # _RE_FIRST_COMPONENT.match('foo.cc').group(0) == 'foo'
3460 # _RE_FIRST_COMPONENT.match('foo-bar_baz.cc').group(0) == 'foo'
3461 # _RE_FIRST_COMPONENT.match('foo_bar-baz.cc').group(0) == 'foo'
3462 _RE_FIRST_COMPONENT = re.compile(r'^[^-_.]+')
3465 def _DropCommonSuffixes(filename):
3466 """Drops common suffixes like _test.cc or -inl.h from filename.
3468 For example:
3469 >>> _DropCommonSuffixes('foo/foo-inl.h')
3470 'foo/foo'
3471 >>> _DropCommonSuffixes('foo/bar/foo.cc')
3472 'foo/bar/foo'
3473 >>> _DropCommonSuffixes('foo/foo_internal.h')
3474 'foo/foo'
3475 >>> _DropCommonSuffixes('foo/foo_unusualinternal.h')
3476 'foo/foo_unusualinternal'
3478 Args:
3479 filename: The input filename.
3481 Returns:
3482 The filename with the common suffix removed.
3484 for suffix in ('test.cc', 'regtest.cc', 'unittest.cc',
3485 'inl.h', 'impl.h', 'internal.h'):
3486 if (filename.endswith(suffix) and len(filename) > len(suffix) and
3487 filename[-len(suffix) - 1] in ('-', '_')):
3488 return filename[:-len(suffix) - 1]
3489 return os.path.splitext(filename)[0]
3492 def _IsTestFilename(filename):
3493 """Determines if the given filename has a suffix that identifies it as a test.
3495 Args:
3496 filename: The input filename.
3498 Returns:
3499 True if 'filename' looks like a test, False otherwise.
3501 if (filename.endswith('_test.cc') or
3502 filename.endswith('_unittest.cc') or
3503 filename.endswith('_regtest.cc')):
3504 return True
3505 else:
3506 return False
3509 def _ClassifyInclude(fileinfo, include, is_system):
3510 """Figures out what kind of header 'include' is.
3512 Args:
3513 fileinfo: The current file cpplint is running over. A FileInfo instance.
3514 include: The path to a #included file.
3515 is_system: True if the #include used <> rather than "".
3517 Returns:
3518 One of the _XXX_HEADER constants.
3520 For example:
3521 >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'stdio.h', True)
3522 _C_SYS_HEADER
3523 >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'string', True)
3524 _CPP_SYS_HEADER
3525 >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/foo.h', False)
3526 _LIKELY_MY_HEADER
3527 >>> _ClassifyInclude(FileInfo('foo/foo_unknown_extension.cc'),
3528 ... 'bar/foo_other_ext.h', False)
3529 _POSSIBLE_MY_HEADER
3530 >>> _ClassifyInclude(FileInfo('foo/foo.cc'), 'foo/bar.h', False)
3531 _OTHER_HEADER
3533 # This is a list of all standard c++ header files, except
3534 # those already checked for above.
3535 is_cpp_h = include in _CPP_HEADERS
3537 if is_system:
3538 if is_cpp_h:
3539 return _CPP_SYS_HEADER
3540 else:
3541 return _C_SYS_HEADER
3543 # If the target file and the include we're checking share a
3544 # basename when we drop common extensions, and the include
3545 # lives in . , then it's likely to be owned by the target file.
3546 target_dir, target_base = (
3547 os.path.split(_DropCommonSuffixes(fileinfo.RepositoryName())))
3548 include_dir, include_base = os.path.split(_DropCommonSuffixes(include))
3549 if target_base == include_base and (
3550 include_dir == target_dir or
3551 include_dir == os.path.normpath(target_dir + '/../public')):
3552 return _LIKELY_MY_HEADER
3554 # If the target and include share some initial basename
3555 # component, it's possible the target is implementing the
3556 # include, so it's allowed to be first, but we'll never
3557 # complain if it's not there.
3558 target_first_component = _RE_FIRST_COMPONENT.match(target_base)
3559 include_first_component = _RE_FIRST_COMPONENT.match(include_base)
3560 if (target_first_component and include_first_component and
3561 target_first_component.group(0) ==
3562 include_first_component.group(0)):
3563 return _POSSIBLE_MY_HEADER
3565 return _OTHER_HEADER
3569 def CheckIncludeLine(filename, clean_lines, linenum, include_state, error):
3570 """Check rules that are applicable to #include lines.
3572 Strings on #include lines are NOT removed from elided line, to make
3573 certain tasks easier. However, to prevent false positives, checks
3574 applicable to #include lines in CheckLanguage must be put here.
3576 Args:
3577 filename: The name of the current file.
3578 clean_lines: A CleansedLines instance containing the file.
3579 linenum: The number of the line to check.
3580 include_state: An _IncludeState instance in which the headers are inserted.
3581 error: The function to call with any errors found.
3583 fileinfo = FileInfo(filename)
3585 line = clean_lines.lines[linenum]
3587 # "include" should use the new style "foo/bar.h" instead of just "bar.h"
3588 if _RE_PATTERN_INCLUDE_NEW_STYLE.search(line):
3589 error(filename, linenum, 'build/include', 4,
3590 'Include the directory when naming .h files')
3592 # we shouldn't include a file more than once. actually, there are a
3593 # handful of instances where doing so is okay, but in general it's
3594 # not.
3595 match = _RE_PATTERN_INCLUDE.search(line)
3596 if match:
3597 include = match.group(2)
3598 is_system = (match.group(1) == '<')
3599 if include in include_state:
3600 error(filename, linenum, 'build/include', 4,
3601 '"%s" already included at %s:%s' %
3602 (include, filename, include_state[include]))
3603 else:
3604 include_state[include] = linenum
3606 # We want to ensure that headers appear in the right order:
3607 # 1) for foo.cc, foo.h (preferred location)
3608 # 2) c system files
3609 # 3) cpp system files
3610 # 4) for foo.cc, foo.h (deprecated location)
3611 # 5) other google headers
3613 # We classify each include statement as one of those 5 types
3614 # using a number of techniques. The include_state object keeps
3615 # track of the highest type seen, and complains if we see a
3616 # lower type after that.
3617 error_message = include_state.CheckNextIncludeOrder(
3618 _ClassifyInclude(fileinfo, include, is_system))
3619 if error_message:
3620 error(filename, linenum, 'build/include_order', 4,
3621 '%s. Should be: %s.h, c system, c++ system, other.' %
3622 (error_message, fileinfo.BaseName()))
3623 canonical_include = include_state.CanonicalizeAlphabeticalOrder(include)
3624 if not include_state.IsInAlphabeticalOrder(
3625 clean_lines, linenum, canonical_include):
3626 error(filename, linenum, 'build/include_alpha', 4,
3627 'Include "%s" not in alphabetical order' % include)
3628 include_state.SetLastHeader(canonical_include)
3630 # Look for any of the stream classes that are part of standard C++.
3631 match = _RE_PATTERN_INCLUDE.match(line)
3632 if match:
3633 include = match.group(2)
3634 if Match(r'(f|ind|io|i|o|parse|pf|stdio|str|)?stream$', include):
3635 # Many unit tests use cout, so we exempt them.
3636 if not _IsTestFilename(filename):
3637 error(filename, linenum, 'readability/streams', 3,
3638 'Streams are highly discouraged.')
3641 def _GetTextInside(text, start_pattern):
3642 r"""Retrieves all the text between matching open and close parentheses.
3644 Given a string of lines and a regular expression string, retrieve all the text
3645 following the expression and between opening punctuation symbols like
3646 (, [, or {, and the matching close-punctuation symbol. This properly nested
3647 occurrences of the punctuations, so for the text like
3648 printf(a(), b(c()));
3649 a call to _GetTextInside(text, r'printf\(') will return 'a(), b(c())'.
3650 start_pattern must match string having an open punctuation symbol at the end.
3652 Args:
3653 text: The lines to extract text. Its comments and strings must be elided.
3654 It can be single line and can span multiple lines.
3655 start_pattern: The regexp string indicating where to start extracting
3656 the text.
3657 Returns:
3658 The extracted text.
3659 None if either the opening string or ending punctuation could not be found.
3661 # TODO(sugawarayu): Audit cpplint.py to see what places could be profitably
3662 # rewritten to use _GetTextInside (and use inferior regexp matching today).
3664 # Give opening punctuations to get the matching close-punctuations.
3665 matching_punctuation = {'(': ')', '{': '}', '[': ']'}
3666 closing_punctuation = set(matching_punctuation.itervalues())
3668 # Find the position to start extracting text.
3669 match = re.search(start_pattern, text, re.M)
3670 if not match: # start_pattern not found in text.
3671 return None
3672 start_position = match.end(0)
3674 assert start_position > 0, (
3675 'start_pattern must ends with an opening punctuation.')
3676 assert text[start_position - 1] in matching_punctuation, (
3677 'start_pattern must ends with an opening punctuation.')
3678 # Stack of closing punctuations we expect to have in text after position.
3679 punctuation_stack = [matching_punctuation[text[start_position - 1]]]
3680 position = start_position
3681 while punctuation_stack and position < len(text):
3682 if text[position] == punctuation_stack[-1]:
3683 punctuation_stack.pop()
3684 elif text[position] in closing_punctuation:
3685 # A closing punctuation without matching opening punctuations.
3686 return None
3687 elif text[position] in matching_punctuation:
3688 punctuation_stack.append(matching_punctuation[text[position]])
3689 position += 1
3690 if punctuation_stack:
3691 # Opening punctuations left without matching close-punctuations.
3692 return None
3693 # punctuations match.
3694 return text[start_position:position - 1]
3697 # Patterns for matching call-by-reference parameters.
3699 # Supports nested templates up to 2 levels deep using this messy pattern:
3700 # < (?: < (?: < [^<>]*
3702 # | [^<>] )*
3704 # | [^<>] )*
3706 _RE_PATTERN_IDENT = r'[_a-zA-Z]\w*' # =~ [[:alpha:]][[:alnum:]]*
3707 _RE_PATTERN_TYPE = (
3708 r'(?:const\s+)?(?:typename\s+|class\s+|struct\s+|union\s+|enum\s+)?'
3709 r'(?:\w|'
3710 r'\s*<(?:<(?:<[^<>]*>|[^<>])*>|[^<>])*>|'
3711 r'::)+')
3712 # A call-by-reference parameter ends with '& identifier'.
3713 _RE_PATTERN_REF_PARAM = re.compile(
3714 r'(' + _RE_PATTERN_TYPE + r'(?:\s*(?:\bconst\b|[*]))*\s*'
3715 r'&\s*' + _RE_PATTERN_IDENT + r')\s*(?:=[^,()]+)?[,)]')
3716 # A call-by-const-reference parameter either ends with 'const& identifier'
3717 # or looks like 'const type& identifier' when 'type' is atomic.
3718 _RE_PATTERN_CONST_REF_PARAM = (
3719 r'(?:.*\s*\bconst\s*&\s*' + _RE_PATTERN_IDENT +
3720 r'|const\s+' + _RE_PATTERN_TYPE + r'\s*&\s*' + _RE_PATTERN_IDENT + r')')
3723 def CheckLanguage(filename, clean_lines, linenum, file_extension,
3724 include_state, nesting_state, error):
3725 """Checks rules from the 'C++ language rules' section of cppguide.html.
3727 Some of these rules are hard to test (function overloading, using
3728 uint32 inappropriately), but we do the best we can.
3730 Args:
3731 filename: The name of the current file.
3732 clean_lines: A CleansedLines instance containing the file.
3733 linenum: The number of the line to check.
3734 file_extension: The extension (without the dot) of the filename.
3735 include_state: An _IncludeState instance in which the headers are inserted.
3736 nesting_state: A _NestingState instance which maintains information about
3737 the current stack of nested blocks being parsed.
3738 error: The function to call with any errors found.
3740 # If the line is empty or consists of entirely a comment, no need to
3741 # check it.
3742 line = clean_lines.elided[linenum]
3743 if not line:
3744 return
3746 match = _RE_PATTERN_INCLUDE.search(line)
3747 if match:
3748 CheckIncludeLine(filename, clean_lines, linenum, include_state, error)
3749 return
3751 # Reset include state across preprocessor directives. This is meant
3752 # to silence warnings for conditional includes.
3753 if Match(r'^\s*#\s*(?:ifdef|elif|else|endif)\b', line):
3754 include_state.ResetSection()
3756 # Make Windows paths like Unix.
3757 fullname = os.path.abspath(filename).replace('\\', '/')
3759 # TODO(unknown): figure out if they're using default arguments in fn proto.
3761 # Check to see if they're using an conversion function cast.
3762 # I just try to capture the most common basic types, though there are more.
3763 # Parameterless conversion functions, such as bool(), are allowed as they are
3764 # probably a member operator declaration or default constructor.
3765 match = Search(
3766 r'(\bnew\s+)?\b' # Grab 'new' operator, if it's there
3767 r'(int|float|double|bool|char|int32|uint32|int64|uint64)'
3768 r'(\([^)].*)', line)
3769 if match:
3770 matched_new = match.group(1)
3771 matched_type = match.group(2)
3772 matched_funcptr = match.group(3)
3774 # gMock methods are defined using some variant of MOCK_METHODx(name, type)
3775 # where type may be float(), int(string), etc. Without context they are
3776 # virtually indistinguishable from int(x) casts. Likewise, gMock's
3777 # MockCallback takes a template parameter of the form return_type(arg_type),
3778 # which looks much like the cast we're trying to detect.
3780 # std::function<> wrapper has a similar problem.
3782 # Return types for function pointers also look like casts if they
3783 # don't have an extra space.
3784 if (matched_new is None and # If new operator, then this isn't a cast
3785 not (Match(r'^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(', line) or
3786 Search(r'\bMockCallback<.*>', line) or
3787 Search(r'\bstd::function<.*>', line)) and
3788 not (matched_funcptr and
3789 Match(r'\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(',
3790 matched_funcptr))):
3791 # Try a bit harder to catch gmock lines: the only place where
3792 # something looks like an old-style cast is where we declare the
3793 # return type of the mocked method, and the only time when we
3794 # are missing context is if MOCK_METHOD was split across
3795 # multiple lines. The missing MOCK_METHOD is usually one or two
3796 # lines back, so scan back one or two lines.
3798 # It's not possible for gmock macros to appear in the first 2
3799 # lines, since the class head + section name takes up 2 lines.
3800 if (linenum < 2 or
3801 not (Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$',
3802 clean_lines.elided[linenum - 1]) or
3803 Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$',
3804 clean_lines.elided[linenum - 2]))):
3805 error(filename, linenum, 'readability/casting', 4,
3806 'Using deprecated casting style. '
3807 'Use static_cast<%s>(...) instead' %
3808 matched_type)
3810 CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
3811 'static_cast',
3812 r'\((int|float|double|bool|char|u?int(16|32|64))\)', error)
3814 # This doesn't catch all cases. Consider (const char * const)"hello".
3816 # (char *) "foo" should always be a const_cast (reinterpret_cast won't
3817 # compile).
3818 if CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
3819 'const_cast', r'\((char\s?\*+\s?)\)\s*"', error):
3820 pass
3821 else:
3822 # Check pointer casts for other than string constants
3823 CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
3824 'reinterpret_cast', r'\((\w+\s?\*+\s?)\)', error)
3826 # In addition, we look for people taking the address of a cast. This
3827 # is dangerous -- casts can assign to temporaries, so the pointer doesn't
3828 # point where you think.
3829 match = Search(
3830 r'(?:&\(([^)]+)\)[\w(])|'
3831 r'(?:&(static|dynamic|down|reinterpret)_cast\b)', line)
3832 if match and match.group(1) != '*':
3833 error(filename, linenum, 'runtime/casting', 4,
3834 ('Are you taking an address of a cast? '
3835 'This is dangerous: could be a temp var. '
3836 'Take the address before doing the cast, rather than after'))
3838 # Create an extended_line, which is the concatenation of the current and
3839 # next lines, for more effective checking of code that may span more than one
3840 # line.
3841 if linenum + 1 < clean_lines.NumLines():
3842 extended_line = line + clean_lines.elided[linenum + 1]
3843 else:
3844 extended_line = line
3846 # Check for people declaring static/global STL strings at the top level.
3847 # This is dangerous because the C++ language does not guarantee that
3848 # globals with constructors are initialized before the first access.
3849 match = Match(
3850 r'((?:|static +)(?:|const +))string +([a-zA-Z0-9_:]+)\b(.*)',
3851 line)
3852 # Make sure it's not a function.
3853 # Function template specialization looks like: "string foo<Type>(...".
3854 # Class template definitions look like: "string Foo<Type>::Method(...".
3856 # Also ignore things that look like operators. These are matched separately
3857 # because operator names cross non-word boundaries. If we change the pattern
3858 # above, we would decrease the accuracy of matching identifiers.
3859 if (match and
3860 not Search(r'\boperator\W', line) and
3861 not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)?\s*\(([^"]|$)', match.group(3))):
3862 error(filename, linenum, 'runtime/string', 4,
3863 'For a static/global string constant, use a C style string instead: '
3864 '"%schar %s[]".' %
3865 (match.group(1), match.group(2)))
3867 if Search(r'\b([A-Za-z0-9_]*_)\(\1\)', line):
3868 error(filename, linenum, 'runtime/init', 4,
3869 'You seem to be initializing a member variable with itself.')
3871 if file_extension == 'h':
3872 # TODO(unknown): check that 1-arg constructors are explicit.
3873 # How to tell it's a constructor?
3874 # (handled in CheckForNonStandardConstructs for now)
3875 # TODO(unknown): check that classes have DISALLOW_EVIL_CONSTRUCTORS
3876 # (level 1 error)
3877 pass
3879 # Check if people are using the verboten C basic types. The only exception
3880 # we regularly allow is "unsigned short port" for port.
3881 if Search(r'\bshort port\b', line):
3882 if not Search(r'\bunsigned short port\b', line):
3883 error(filename, linenum, 'runtime/int', 4,
3884 'Use "unsigned short" for ports, not "short"')
3885 else:
3886 match = Search(r'\b(short|long(?! +double)|long long)\b', line)
3887 if match:
3888 error(filename, linenum, 'runtime/int', 4,
3889 'Use int16/int64/etc, rather than the C type %s' % match.group(1))
3891 # When snprintf is used, the second argument shouldn't be a literal.
3892 match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line)
3893 if match and match.group(2) != '0':
3894 # If 2nd arg is zero, snprintf is used to calculate size.
3895 error(filename, linenum, 'runtime/printf', 3,
3896 'If you can, use sizeof(%s) instead of %s as the 2nd arg '
3897 'to snprintf.' % (match.group(1), match.group(2)))
3899 # Check if some verboten C functions are being used.
3900 if Search(r'\bsprintf\b', line):
3901 error(filename, linenum, 'runtime/printf', 5,
3902 'Never use sprintf. Use snprintf instead.')
3903 match = Search(r'\b(strcpy|strcat)\b', line)
3904 if match:
3905 error(filename, linenum, 'runtime/printf', 4,
3906 'Almost always, snprintf is better than %s' % match.group(1))
3908 # Check if some verboten operator overloading is going on
3909 # TODO(unknown): catch out-of-line unary operator&:
3910 # class X {};
3911 # int operator&(const X& x) { return 42; } // unary operator&
3912 # The trick is it's hard to tell apart from binary operator&:
3913 # class Y { int operator&(const Y& x) { return 23; } }; // binary operator&
3914 if Search(r'\boperator\s*&\s*\(\s*\)', line):
3915 error(filename, linenum, 'runtime/operator', 4,
3916 'Unary operator& is dangerous. Do not use it.')
3918 # Check for suspicious usage of "if" like
3919 # } if (a == b) {
3920 if Search(r'\}\s*if\s*\(', line):
3921 error(filename, linenum, 'readability/braces', 4,
3922 'Did you mean "else if"? If not, start a new line for "if".')
3924 # Check for potential format string bugs like printf(foo).
3925 # We constrain the pattern not to pick things like DocidForPrintf(foo).
3926 # Not perfect but it can catch printf(foo.c_str()) and printf(foo->c_str())
3927 # TODO(sugawarayu): Catch the following case. Need to change the calling
3928 # convention of the whole function to process multiple line to handle it.
3929 # printf(
3930 # boy_this_is_a_really_long_variable_that_cannot_fit_on_the_prev_line);
3931 printf_args = _GetTextInside(line, r'(?i)\b(string)?printf\s*\(')
3932 if printf_args:
3933 match = Match(r'([\w.\->()]+)$', printf_args)
3934 if match and match.group(1) != '__VA_ARGS__':
3935 function_name = re.search(r'\b((?:string)?printf)\s*\(',
3936 line, re.I).group(1)
3937 error(filename, linenum, 'runtime/printf', 4,
3938 'Potential format string bug. Do %s("%%s", %s) instead.'
3939 % (function_name, match.group(1)))
3941 # Check for potential memset bugs like memset(buf, sizeof(buf), 0).
3942 match = Search(r'memset\s*\(([^,]*),\s*([^,]*),\s*0\s*\)', line)
3943 if match and not Match(r"^''|-?[0-9]+|0x[0-9A-Fa-f]$", match.group(2)):
3944 error(filename, linenum, 'runtime/memset', 4,
3945 'Did you mean "memset(%s, 0, %s)"?'
3946 % (match.group(1), match.group(2)))
3948 if Search(r'\busing namespace\b', line):
3949 error(filename, linenum, 'build/namespaces', 5,
3950 'Do not use namespace using-directives. '
3951 'Use using-declarations instead.')
3953 # Detect variable-length arrays.
3954 match = Match(r'\s*(.+::)?(\w+) [a-z]\w*\[(.+)];', line)
3955 if (match and match.group(2) != 'return' and match.group(2) != 'delete' and
3956 match.group(3).find(']') == -1):
3957 # Split the size using space and arithmetic operators as delimiters.
3958 # If any of the resulting tokens are not compile time constants then
3959 # report the error.
3960 tokens = re.split(r'\s|\+|\-|\*|\/|<<|>>]', match.group(3))
3961 is_const = True
3962 skip_next = False
3963 for tok in tokens:
3964 if skip_next:
3965 skip_next = False
3966 continue
3968 if Search(r'sizeof\(.+\)', tok): continue
3969 if Search(r'arraysize\(\w+\)', tok): continue
3971 tok = tok.lstrip('(')
3972 tok = tok.rstrip(')')
3973 if not tok: continue
3974 if Match(r'\d+', tok): continue
3975 if Match(r'0[xX][0-9a-fA-F]+', tok): continue
3976 if Match(r'k[A-Z0-9]\w*', tok): continue
3977 if Match(r'(.+::)?k[A-Z0-9]\w*', tok): continue
3978 if Match(r'(.+::)?[A-Z][A-Z0-9_]*', tok): continue
3979 # A catch all for tricky sizeof cases, including 'sizeof expression',
3980 # 'sizeof(*type)', 'sizeof(const type)', 'sizeof(struct StructName)'
3981 # requires skipping the next token because we split on ' ' and '*'.
3982 if tok.startswith('sizeof'):
3983 skip_next = True
3984 continue
3985 is_const = False
3986 break
3987 if not is_const:
3988 error(filename, linenum, 'runtime/arrays', 1,
3989 'Do not use variable-length arrays. Use an appropriately named '
3990 "('k' followed by CamelCase) compile-time constant for the size.")
3992 # If DISALLOW_EVIL_CONSTRUCTORS, DISALLOW_COPY_AND_ASSIGN, or
3993 # DISALLOW_IMPLICIT_CONSTRUCTORS is present, then it should be the last thing
3994 # in the class declaration.
3995 match = Match(
3996 (r'\s*'
3997 r'(DISALLOW_(EVIL_CONSTRUCTORS|COPY_AND_ASSIGN|IMPLICIT_CONSTRUCTORS))'
3998 r'\(.*\);$'),
3999 line)
4000 if match and linenum + 1 < clean_lines.NumLines():
4001 next_line = clean_lines.elided[linenum + 1]
4002 # We allow some, but not all, declarations of variables to be present
4003 # in the statement that defines the class. The [\w\*,\s]* fragment of
4004 # the regular expression below allows users to declare instances of
4005 # the class or pointers to instances, but not less common types such
4006 # as function pointers or arrays. It's a tradeoff between allowing
4007 # reasonable code and avoiding trying to parse more C++ using regexps.
4008 if not Search(r'^\s*}[\w\*,\s]*;', next_line):
4009 error(filename, linenum, 'readability/constructors', 3,
4010 match.group(1) + ' should be the last thing in the class')
4012 # Check for use of unnamed namespaces in header files. Registration
4013 # macros are typically OK, so we allow use of "namespace {" on lines
4014 # that end with backslashes.
4015 if (file_extension == 'h'
4016 and Search(r'\bnamespace\s*{', line)
4017 and line[-1] != '\\'):
4018 error(filename, linenum, 'build/namespaces', 4,
4019 'Do not use unnamed namespaces in header files. See '
4020 'http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces'
4021 ' for more information.')
4023 def CheckForNonConstReference(filename, clean_lines, linenum,
4024 nesting_state, error):
4025 """Check for non-const references.
4027 Separate from CheckLanguage since it scans backwards from current
4028 line, instead of scanning forward.
4030 Args:
4031 filename: The name of the current file.
4032 clean_lines: A CleansedLines instance containing the file.
4033 linenum: The number of the line to check.
4034 nesting_state: A _NestingState instance which maintains information about
4035 the current stack of nested blocks being parsed.
4036 error: The function to call with any errors found.
4038 # Do nothing if there is no '&' on current line.
4039 line = clean_lines.elided[linenum]
4040 if '&' not in line:
4041 return
4043 # Long type names may be broken across multiple lines, usually in one
4044 # of these forms:
4045 # LongType
4046 # ::LongTypeContinued &identifier
4047 # LongType::
4048 # LongTypeContinued &identifier
4049 # LongType<
4050 # ...>::LongTypeContinued &identifier
4052 # If we detected a type split across two lines, join the previous
4053 # line to current line so that we can match const references
4054 # accordingly.
4056 # Note that this only scans back one line, since scanning back
4057 # arbitrary number of lines would be expensive. If you have a type
4058 # that spans more than 2 lines, please use a typedef.
4059 if linenum > 1:
4060 previous = None
4061 if Match(r'\s*::(?:[\w<>]|::)+\s*&\s*\S', line):
4062 # previous_line\n + ::current_line
4063 previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+[\w<>])\s*$',
4064 clean_lines.elided[linenum - 1])
4065 elif Match(r'\s*[a-zA-Z_]([\w<>]|::)+\s*&\s*\S', line):
4066 # previous_line::\n + current_line
4067 previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+::)\s*$',
4068 clean_lines.elided[linenum - 1])
4069 if previous:
4070 line = previous.group(1) + line.lstrip()
4071 else:
4072 # Check for templated parameter that is split across multiple lines
4073 endpos = line.rfind('>')
4074 if endpos > -1:
4075 (_, startline, startpos) = ReverseCloseExpression(
4076 clean_lines, linenum, endpos)
4077 if startpos > -1 and startline < linenum:
4078 # Found the matching < on an earlier line, collect all
4079 # pieces up to current line.
4080 line = ''
4081 for i in xrange(startline, linenum + 1):
4082 line += clean_lines.elided[i].strip()
4084 # Check for non-const references in function parameters. A single '&' may
4085 # found in the following places:
4086 # inside expression: binary & for bitwise AND
4087 # inside expression: unary & for taking the address of something
4088 # inside declarators: reference parameter
4089 # We will exclude the first two cases by checking that we are not inside a
4090 # function body, including one that was just introduced by a trailing '{'.
4091 # TODO(unknwon): Doesn't account for preprocessor directives.
4092 # TODO(unknown): Doesn't account for 'catch(Exception& e)' [rare].
4093 check_params = False
4094 if not nesting_state.stack:
4095 check_params = True # top level
4096 elif (isinstance(nesting_state.stack[-1], _ClassInfo) or
4097 isinstance(nesting_state.stack[-1], _NamespaceInfo)):
4098 check_params = True # within class or namespace
4099 elif Match(r'.*{\s*$', line):
4100 if (len(nesting_state.stack) == 1 or
4101 isinstance(nesting_state.stack[-2], _ClassInfo) or
4102 isinstance(nesting_state.stack[-2], _NamespaceInfo)):
4103 check_params = True # just opened global/class/namespace block
4104 # We allow non-const references in a few standard places, like functions
4105 # called "swap()" or iostream operators like "<<" or ">>". Do not check
4106 # those function parameters.
4108 # We also accept & in static_assert, which looks like a function but
4109 # it's actually a declaration expression.
4110 whitelisted_functions = (r'(?:[sS]wap(?:<\w:+>)?|'
4111 r'operator\s*[<>][<>]|'
4112 r'static_assert|COMPILE_ASSERT'
4113 r')\s*\(')
4114 if Search(whitelisted_functions, line):
4115 check_params = False
4116 elif not Search(r'\S+\([^)]*$', line):
4117 # Don't see a whitelisted function on this line. Actually we
4118 # didn't see any function name on this line, so this is likely a
4119 # multi-line parameter list. Try a bit harder to catch this case.
4120 for i in xrange(2):
4121 if (linenum > i and
4122 Search(whitelisted_functions, clean_lines.elided[linenum - i - 1])):
4123 check_params = False
4124 break
4126 if check_params:
4127 decls = ReplaceAll(r'{[^}]*}', ' ', line) # exclude function body
4128 for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls):
4129 if not Match(_RE_PATTERN_CONST_REF_PARAM, parameter):
4130 error(filename, linenum, 'runtime/references', 2,
4131 'Is this a non-const reference? '
4132 'If so, make const or use a pointer: ' +
4133 ReplaceAll(' *<', '<', parameter))
4136 def CheckCStyleCast(filename, linenum, line, raw_line, cast_type, pattern,
4137 error):
4138 """Checks for a C-style cast by looking for the pattern.
4140 Args:
4141 filename: The name of the current file.
4142 linenum: The number of the line to check.
4143 line: The line of code to check.
4144 raw_line: The raw line of code to check, with comments.
4145 cast_type: The string for the C++ cast to recommend. This is either
4146 reinterpret_cast, static_cast, or const_cast, depending.
4147 pattern: The regular expression used to find C-style casts.
4148 error: The function to call with any errors found.
4150 Returns:
4151 True if an error was emitted.
4152 False otherwise.
4154 match = Search(pattern, line)
4155 if not match:
4156 return False
4158 # e.g., sizeof(int)
4159 sizeof_match = Match(r'.*sizeof\s*$', line[0:match.start(1) - 1])
4160 if sizeof_match:
4161 error(filename, linenum, 'runtime/sizeof', 1,
4162 'Using sizeof(type). Use sizeof(varname) instead if possible')
4163 return True
4165 # operator++(int) and operator--(int)
4166 if (line[0:match.start(1) - 1].endswith(' operator++') or
4167 line[0:match.start(1) - 1].endswith(' operator--')):
4168 return False
4170 # A single unnamed argument for a function tends to look like old
4171 # style cast. If we see those, don't issue warnings for deprecated
4172 # casts, instead issue warnings for unnamed arguments where
4173 # appropriate.
4175 # These are things that we want warnings for, since the style guide
4176 # explicitly require all parameters to be named:
4177 # Function(int);
4178 # Function(int) {
4179 # ConstMember(int) const;
4180 # ConstMember(int) const {
4181 # ExceptionMember(int) throw (...);
4182 # ExceptionMember(int) throw (...) {
4183 # PureVirtual(int) = 0;
4185 # These are functions of some sort, where the compiler would be fine
4186 # if they had named parameters, but people often omit those
4187 # identifiers to reduce clutter:
4188 # (FunctionPointer)(int);
4189 # (FunctionPointer)(int) = value;
4190 # Function((function_pointer_arg)(int))
4191 # <TemplateArgument(int)>;
4192 # <(FunctionPointerTemplateArgument)(int)>;
4193 remainder = line[match.end(0):]
4194 if Match(r'^\s*(?:;|const\b|throw\b|=|>|\{|\))', remainder):
4195 # Looks like an unnamed parameter.
4197 # Don't warn on any kind of template arguments.
4198 if Match(r'^\s*>', remainder):
4199 return False
4201 # Don't warn on assignments to function pointers, but keep warnings for
4202 # unnamed parameters to pure virtual functions. Note that this pattern
4203 # will also pass on assignments of "0" to function pointers, but the
4204 # preferred values for those would be "nullptr" or "NULL".
4205 matched_zero = Match(r'^\s=\s*(\S+)\s*;', remainder)
4206 if matched_zero and matched_zero.group(1) != '0':
4207 return False
4209 # Don't warn on function pointer declarations. For this we need
4210 # to check what came before the "(type)" string.
4211 if Match(r'.*\)\s*$', line[0:match.start(0)]):
4212 return False
4214 # Don't warn if the parameter is named with block comments, e.g.:
4215 # Function(int /*unused_param*/);
4216 if '/*' in raw_line:
4217 return False
4219 # Passed all filters, issue warning here.
4220 error(filename, linenum, 'readability/function', 3,
4221 'All parameters should be named in a function')
4222 return True
4224 # At this point, all that should be left is actual casts.
4225 error(filename, linenum, 'readability/casting', 4,
4226 'Using C-style cast. Use %s<%s>(...) instead' %
4227 (cast_type, match.group(1)))
4229 return True
4232 _HEADERS_CONTAINING_TEMPLATES = (
4233 ('<deque>', ('deque',)),
4234 ('<functional>', ('unary_function', 'binary_function',
4235 'plus', 'minus', 'multiplies', 'divides', 'modulus',
4236 'negate',
4237 'equal_to', 'not_equal_to', 'greater', 'less',
4238 'greater_equal', 'less_equal',
4239 'logical_and', 'logical_or', 'logical_not',
4240 'unary_negate', 'not1', 'binary_negate', 'not2',
4241 'bind1st', 'bind2nd',
4242 'pointer_to_unary_function',
4243 'pointer_to_binary_function',
4244 'ptr_fun',
4245 'mem_fun_t', 'mem_fun', 'mem_fun1_t', 'mem_fun1_ref_t',
4246 'mem_fun_ref_t',
4247 'const_mem_fun_t', 'const_mem_fun1_t',
4248 'const_mem_fun_ref_t', 'const_mem_fun1_ref_t',
4249 'mem_fun_ref',
4251 ('<limits>', ('numeric_limits',)),
4252 ('<list>', ('list',)),
4253 ('<map>', ('map', 'multimap',)),
4254 ('<memory>', ('allocator',)),
4255 ('<queue>', ('queue', 'priority_queue',)),
4256 ('<set>', ('set', 'multiset',)),
4257 ('<stack>', ('stack',)),
4258 ('<string>', ('char_traits', 'basic_string',)),
4259 ('<utility>', ('pair',)),
4260 ('<vector>', ('vector',)),
4262 # gcc extensions.
4263 # Note: std::hash is their hash, ::hash is our hash
4264 ('<hash_map>', ('hash_map', 'hash_multimap',)),
4265 ('<hash_set>', ('hash_set', 'hash_multiset',)),
4266 ('<slist>', ('slist',)),
4269 _RE_PATTERN_STRING = re.compile(r'\bstring\b')
4271 _re_pattern_algorithm_header = []
4272 for _template in ('copy', 'max', 'min', 'min_element', 'sort', 'swap',
4273 'transform'):
4274 # Match max<type>(..., ...), max(..., ...), but not foo->max, foo.max or
4275 # type::max().
4276 _re_pattern_algorithm_header.append(
4277 (re.compile(r'[^>.]\b' + _template + r'(<.*?>)?\([^\)]'),
4278 _template,
4279 '<algorithm>'))
4281 _re_pattern_templates = []
4282 for _header, _templates in _HEADERS_CONTAINING_TEMPLATES:
4283 for _template in _templates:
4284 _re_pattern_templates.append(
4285 (re.compile(r'(\<|\b)' + _template + r'\s*\<'),
4286 _template + '<>',
4287 _header))
4290 def FilesBelongToSameModule(filename_cc, filename_h):
4291 """Check if these two filenames belong to the same module.
4293 The concept of a 'module' here is a as follows:
4294 foo.h, foo-inl.h, foo.cc, foo_test.cc and foo_unittest.cc belong to the
4295 same 'module' if they are in the same directory.
4296 some/path/public/xyzzy and some/path/internal/xyzzy are also considered
4297 to belong to the same module here.
4299 If the filename_cc contains a longer path than the filename_h, for example,
4300 '/absolute/path/to/base/sysinfo.cc', and this file would include
4301 'base/sysinfo.h', this function also produces the prefix needed to open the
4302 header. This is used by the caller of this function to more robustly open the
4303 header file. We don't have access to the real include paths in this context,
4304 so we need this guesswork here.
4306 Known bugs: tools/base/bar.cc and base/bar.h belong to the same module
4307 according to this implementation. Because of this, this function gives
4308 some false positives. This should be sufficiently rare in practice.
4310 Args:
4311 filename_cc: is the path for the .cc file
4312 filename_h: is the path for the header path
4314 Returns:
4315 Tuple with a bool and a string:
4316 bool: True if filename_cc and filename_h belong to the same module.
4317 string: the additional prefix needed to open the header file.
4320 if not filename_cc.endswith('.cc'):
4321 return (False, '')
4322 filename_cc = filename_cc[:-len('.cc')]
4323 if filename_cc.endswith('_unittest'):
4324 filename_cc = filename_cc[:-len('_unittest')]
4325 elif filename_cc.endswith('_test'):
4326 filename_cc = filename_cc[:-len('_test')]
4327 filename_cc = filename_cc.replace('/public/', '/')
4328 filename_cc = filename_cc.replace('/internal/', '/')
4330 if not filename_h.endswith('.h'):
4331 return (False, '')
4332 filename_h = filename_h[:-len('.h')]
4333 if filename_h.endswith('-inl'):
4334 filename_h = filename_h[:-len('-inl')]
4335 filename_h = filename_h.replace('/public/', '/')
4336 filename_h = filename_h.replace('/internal/', '/')
4338 files_belong_to_same_module = filename_cc.endswith(filename_h)
4339 common_path = ''
4340 if files_belong_to_same_module:
4341 common_path = filename_cc[:-len(filename_h)]
4342 return files_belong_to_same_module, common_path
4345 def UpdateIncludeState(filename, include_state, io=codecs):
4346 """Fill up the include_state with new includes found from the file.
4348 Args:
4349 filename: the name of the header to read.
4350 include_state: an _IncludeState instance in which the headers are inserted.
4351 io: The io factory to use to read the file. Provided for testability.
4353 Returns:
4354 True if a header was succesfully added. False otherwise.
4356 headerfile = None
4357 try:
4358 headerfile = io.open(filename, 'r', 'utf8', 'replace')
4359 except IOError:
4360 return False
4361 linenum = 0
4362 for line in headerfile:
4363 linenum += 1
4364 clean_line = CleanseComments(line)
4365 match = _RE_PATTERN_INCLUDE.search(clean_line)
4366 if match:
4367 include = match.group(2)
4368 # The value formatting is cute, but not really used right now.
4369 # What matters here is that the key is in include_state.
4370 include_state.setdefault(include, '%s:%d' % (filename, linenum))
4371 return True
4374 def CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error,
4375 io=codecs):
4376 """Reports for missing stl includes.
4378 This function will output warnings to make sure you are including the headers
4379 necessary for the stl containers and functions that you use. We only give one
4380 reason to include a header. For example, if you use both equal_to<> and
4381 less<> in a .h file, only one (the latter in the file) of these will be
4382 reported as a reason to include the <functional>.
4384 Args:
4385 filename: The name of the current file.
4386 clean_lines: A CleansedLines instance containing the file.
4387 include_state: An _IncludeState instance.
4388 error: The function to call with any errors found.
4389 io: The IO factory to use to read the header file. Provided for unittest
4390 injection.
4392 required = {} # A map of header name to linenumber and the template entity.
4393 # Example of required: { '<functional>': (1219, 'less<>') }
4395 for linenum in xrange(clean_lines.NumLines()):
4396 line = clean_lines.elided[linenum]
4397 if not line or line[0] == '#':
4398 continue
4400 # String is special -- it is a non-templatized type in STL.
4401 matched = _RE_PATTERN_STRING.search(line)
4402 if matched:
4403 # Don't warn about strings in non-STL namespaces:
4404 # (We check only the first match per line; good enough.)
4405 prefix = line[:matched.start()]
4406 if prefix.endswith('std::') or not prefix.endswith('::'):
4407 required['<string>'] = (linenum, 'string')
4409 for pattern, template, header in _re_pattern_algorithm_header:
4410 if pattern.search(line):
4411 required[header] = (linenum, template)
4413 # The following function is just a speed up, no semantics are changed.
4414 if not '<' in line: # Reduces the cpu time usage by skipping lines.
4415 continue
4417 for pattern, template, header in _re_pattern_templates:
4418 if pattern.search(line):
4419 required[header] = (linenum, template)
4421 # The policy is that if you #include something in foo.h you don't need to
4422 # include it again in foo.cc. Here, we will look at possible includes.
4423 # Let's copy the include_state so it is only messed up within this function.
4424 include_state = include_state.copy()
4426 # Did we find the header for this file (if any) and succesfully load it?
4427 header_found = False
4429 # Use the absolute path so that matching works properly.
4430 abs_filename = FileInfo(filename).FullName()
4432 # For Emacs's flymake.
4433 # If cpplint is invoked from Emacs's flymake, a temporary file is generated
4434 # by flymake and that file name might end with '_flymake.cc'. In that case,
4435 # restore original file name here so that the corresponding header file can be
4436 # found.
4437 # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h'
4438 # instead of 'foo_flymake.h'
4439 abs_filename = re.sub(r'_flymake\.cc$', '.cc', abs_filename)
4441 # include_state is modified during iteration, so we iterate over a copy of
4442 # the keys.
4443 header_keys = include_state.keys()
4444 for header in header_keys:
4445 (same_module, common_path) = FilesBelongToSameModule(abs_filename, header)
4446 fullpath = common_path + header
4447 if same_module and UpdateIncludeState(fullpath, include_state, io):
4448 header_found = True
4450 # If we can't find the header file for a .cc, assume it's because we don't
4451 # know where to look. In that case we'll give up as we're not sure they
4452 # didn't include it in the .h file.
4453 # TODO(unknown): Do a better job of finding .h files so we are confident that
4454 # not having the .h file means there isn't one.
4455 if filename.endswith('.cc') and not header_found:
4456 return
4458 # All the lines have been processed, report the errors found.
4459 for required_header_unstripped in required:
4460 template = required[required_header_unstripped][1]
4461 if required_header_unstripped.strip('<>"') not in include_state:
4462 error(filename, required[required_header_unstripped][0],
4463 'build/include_what_you_use', 4,
4464 'Add #include ' + required_header_unstripped + ' for ' + template)
4467 _RE_PATTERN_EXPLICIT_MAKEPAIR = re.compile(r'\bmake_pair\s*<')
4470 def CheckMakePairUsesDeduction(filename, clean_lines, linenum, error):
4471 """Check that make_pair's template arguments are deduced.
4473 G++ 4.6 in C++0x mode fails badly if make_pair's template arguments are
4474 specified explicitly, and such use isn't intended in any case.
4476 Args:
4477 filename: The name of the current file.
4478 clean_lines: A CleansedLines instance containing the file.
4479 linenum: The number of the line to check.
4480 error: The function to call with any errors found.
4482 line = clean_lines.elided[linenum]
4483 match = _RE_PATTERN_EXPLICIT_MAKEPAIR.search(line)
4484 if match:
4485 error(filename, linenum, 'build/explicit_make_pair',
4486 4, # 4 = high confidence
4487 'For C++11-compatibility, omit template arguments from make_pair'
4488 ' OR use pair directly OR if appropriate, construct a pair directly')
4491 def ProcessLine(filename, file_extension, clean_lines, line,
4492 include_state, function_state, nesting_state, error,
4493 extra_check_functions=[]):
4494 """Processes a single line in the file.
4496 Args:
4497 filename: Filename of the file that is being processed.
4498 file_extension: The extension (dot not included) of the file.
4499 clean_lines: An array of strings, each representing a line of the file,
4500 with comments stripped.
4501 line: Number of line being processed.
4502 include_state: An _IncludeState instance in which the headers are inserted.
4503 function_state: A _FunctionState instance which counts function lines, etc.
4504 nesting_state: A _NestingState instance which maintains information about
4505 the current stack of nested blocks being parsed.
4506 error: A callable to which errors are reported, which takes 4 arguments:
4507 filename, line number, error level, and message
4508 extra_check_functions: An array of additional check functions that will be
4509 run on each source line. Each function takes 4
4510 arguments: filename, clean_lines, line, error
4512 raw_lines = clean_lines.raw_lines
4513 ParseNolintSuppressions(filename, raw_lines[line], line, error)
4514 nesting_state.Update(filename, clean_lines, line, error)
4515 if nesting_state.stack and nesting_state.stack[-1].inline_asm != _NO_ASM:
4516 return
4517 CheckForFunctionLengths(filename, clean_lines, line, function_state, error)
4518 CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error)
4519 CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error)
4520 CheckLanguage(filename, clean_lines, line, file_extension, include_state,
4521 nesting_state, error)
4522 CheckForNonConstReference(filename, clean_lines, line, nesting_state, error)
4523 CheckForNonStandardConstructs(filename, clean_lines, line,
4524 nesting_state, error)
4525 CheckVlogArguments(filename, clean_lines, line, error)
4526 CheckPosixThreading(filename, clean_lines, line, error)
4527 CheckInvalidIncrement(filename, clean_lines, line, error)
4528 CheckMakePairUsesDeduction(filename, clean_lines, line, error)
4529 for check_fn in extra_check_functions:
4530 check_fn(filename, clean_lines, line, error)
4532 def ProcessFileData(filename, file_extension, lines, error,
4533 extra_check_functions=[]):
4534 """Performs lint checks and reports any errors to the given error function.
4536 Args:
4537 filename: Filename of the file that is being processed.
4538 file_extension: The extension (dot not included) of the file.
4539 lines: An array of strings, each representing a line of the file, with the
4540 last element being empty if the file is terminated with a newline.
4541 error: A callable to which errors are reported, which takes 4 arguments:
4542 filename, line number, error level, and message
4543 extra_check_functions: An array of additional check functions that will be
4544 run on each source line. Each function takes 4
4545 arguments: filename, clean_lines, line, error
4547 lines = (['// marker so line numbers and indices both start at 1'] + lines +
4548 ['// marker so line numbers end in a known way'])
4550 include_state = _IncludeState()
4551 function_state = _FunctionState()
4552 nesting_state = _NestingState()
4554 ResetNolintSuppressions()
4556 CheckForCopyright(filename, lines, error)
4558 if file_extension == 'h':
4559 CheckForHeaderGuard(filename, lines, error)
4561 RemoveMultiLineComments(filename, lines, error)
4562 clean_lines = CleansedLines(lines)
4563 for line in xrange(clean_lines.NumLines()):
4564 ProcessLine(filename, file_extension, clean_lines, line,
4565 include_state, function_state, nesting_state, error,
4566 extra_check_functions)
4567 nesting_state.CheckCompletedBlocks(filename, error)
4569 CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error)
4571 # We check here rather than inside ProcessLine so that we see raw
4572 # lines rather than "cleaned" lines.
4573 CheckForBadCharacters(filename, lines, error)
4575 CheckForNewlineAtEOF(filename, lines, error)
4577 def ProcessFile(filename, vlevel, extra_check_functions=[]):
4578 """Does google-lint on a single file.
4580 Args:
4581 filename: The name of the file to parse.
4583 vlevel: The level of errors to report. Every error of confidence
4584 >= verbose_level will be reported. 0 is a good default.
4586 extra_check_functions: An array of additional check functions that will be
4587 run on each source line. Each function takes 4
4588 arguments: filename, clean_lines, line, error
4591 _SetVerboseLevel(vlevel)
4593 try:
4594 # Support the UNIX convention of using "-" for stdin. Note that
4595 # we are not opening the file with universal newline support
4596 # (which codecs doesn't support anyway), so the resulting lines do
4597 # contain trailing '\r' characters if we are reading a file that
4598 # has CRLF endings.
4599 # If after the split a trailing '\r' is present, it is removed
4600 # below. If it is not expected to be present (i.e. os.linesep !=
4601 # '\r\n' as in Windows), a warning is issued below if this file
4602 # is processed.
4604 if filename == '-':
4605 lines = codecs.StreamReaderWriter(sys.stdin,
4606 codecs.getreader('utf8'),
4607 codecs.getwriter('utf8'),
4608 'replace').read().split('\n')
4609 else:
4610 lines = codecs.open(filename, 'r', 'utf8', 'replace').read().split('\n')
4612 carriage_return_found = False
4613 # Remove trailing '\r'.
4614 for linenum in range(len(lines)):
4615 if lines[linenum].endswith('\r'):
4616 lines[linenum] = lines[linenum].rstrip('\r')
4617 carriage_return_found = True
4619 except IOError:
4620 sys.stderr.write(
4621 "Skipping input '%s': Can't open for reading\n" % filename)
4622 return
4624 # Note, if no dot is found, this will give the entire filename as the ext.
4625 file_extension = filename[filename.rfind('.') + 1:]
4627 # When reading from stdin, the extension is unknown, so no cpplint tests
4628 # should rely on the extension.
4629 if filename != '-' and file_extension not in _valid_extensions:
4630 sys.stderr.write('Ignoring %s; not a valid file name '
4631 '(%s)\n' % (filename, ', '.join(_valid_extensions)))
4632 else:
4633 ProcessFileData(filename, file_extension, lines, Error,
4634 extra_check_functions)
4635 if carriage_return_found and os.linesep != '\r\n':
4636 # Use 0 for linenum since outputting only one error for potentially
4637 # several lines.
4638 Error(filename, 0, 'whitespace/newline', 1,
4639 'One or more unexpected \\r (^M) found;'
4640 'better to use only a \\n')
4642 sys.stderr.write('Done processing %s\n' % filename)
4645 def PrintUsage(message):
4646 """Prints a brief usage string and exits, optionally with an error message.
4648 Args:
4649 message: The optional error message.
4651 sys.stderr.write(_USAGE)
4652 if message:
4653 sys.exit('\nFATAL ERROR: ' + message)
4654 else:
4655 sys.exit(1)
4658 def PrintCategories():
4659 """Prints a list of all the error-categories used by error messages.
4661 These are the categories used to filter messages via --filter.
4663 sys.stderr.write(''.join(' %s\n' % cat for cat in _ERROR_CATEGORIES))
4664 sys.exit(0)
4667 def ParseArguments(args):
4668 """Parses the command line arguments.
4670 This may set the output format and verbosity level as side-effects.
4672 Args:
4673 args: The command line arguments:
4675 Returns:
4676 The list of filenames to lint.
4678 try:
4679 (opts, filenames) = getopt.getopt(args, '', ['help', 'output=', 'verbose=',
4680 'counting=',
4681 'filter=',
4682 'root=',
4683 'linelength=',
4684 'extensions='])
4685 except getopt.GetoptError:
4686 PrintUsage('Invalid arguments.')
4688 verbosity = _VerboseLevel()
4689 output_format = _OutputFormat()
4690 filters = ''
4691 counting_style = ''
4693 for (opt, val) in opts:
4694 if opt == '--help':
4695 PrintUsage(None)
4696 elif opt == '--output':
4697 if val not in ('emacs', 'vs7', 'eclipse'):
4698 PrintUsage('The only allowed output formats are emacs, vs7 and eclipse.')
4699 output_format = val
4700 elif opt == '--verbose':
4701 verbosity = int(val)
4702 elif opt == '--filter':
4703 filters = val
4704 if not filters:
4705 PrintCategories()
4706 elif opt == '--counting':
4707 if val not in ('total', 'toplevel', 'detailed'):
4708 PrintUsage('Valid counting options are total, toplevel, and detailed')
4709 counting_style = val
4710 elif opt == '--root':
4711 global _root
4712 _root = val
4713 elif opt == '--linelength':
4714 global _line_length
4715 try:
4716 _line_length = int(val)
4717 except ValueError:
4718 PrintUsage('Line length must be digits.')
4719 elif opt == '--extensions':
4720 global _valid_extensions
4721 try:
4722 _valid_extensions = set(val.split(','))
4723 except ValueError:
4724 PrintUsage('Extensions must be comma seperated list.')
4726 if not filenames:
4727 PrintUsage('No files were specified.')
4729 _SetOutputFormat(output_format)
4730 _SetVerboseLevel(verbosity)
4731 _SetFilters(filters)
4732 _SetCountingStyle(counting_style)
4734 return filenames
4737 def main():
4738 filenames = ParseArguments(sys.argv[1:])
4740 # Change stderr to write with replacement characters so we don't die
4741 # if we try to print something containing non-ASCII characters.
4742 sys.stderr = codecs.StreamReaderWriter(sys.stderr,
4743 codecs.getreader('utf8'),
4744 codecs.getwriter('utf8'),
4745 'replace')
4747 _cpplint_state.ResetErrorCounts()
4748 for filename in filenames:
4749 ProcessFile(filename, _cpplint_state.verbose_level)
4750 _cpplint_state.PrintErrorCounts()
4752 sys.exit(_cpplint_state.error_count > 0)
4755 if __name__ == '__main__':
4756 main()