-#!/usr/bin/python
+#!/usr/bin/env python
#
# Copyright (c) 2009 Google Inc. All rights reserved.
#
import string
import sys
import unicodedata
+import sysconfig
+
+try:
+ xrange # Python 2
+except NameError:
+ xrange = range # Python 3
_USAGE = """
Syntax: cpplint.py [--verbose=#] [--output=vs7] [--filter=-x,+y,...]
[--counting=total|toplevel|detailed] [--root=subdir]
- [--linelength=digits]
+ [--linelength=digits] [--headers=x,y,...]
+ [--quiet]
<file> [file] ...
The style guidelines this tries to follow are those in
- http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml
+ https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml
Every problem is given a confidence score from 1-5, with 5 meaning we are
certain of the problem, and 1 meaning it could be a legitimate construct.
verbose=#
Specify a number 0-5 to restrict errors to certain verbosity levels.
+ quiet
+ Don't print anything if no errors are found.
+
filter=-x,+y,...
Specify a comma-separated list of category-filters to apply: only
error messages whose category names pass the filters will be printed.
ignored.
Examples:
- Assuing that src/.git exists, the header guard CPP variables for
- src/chrome/browser/ui/browser.h are:
+ Assuming that top/src/.git exists (and cwd=top/src), the header guard
+ CPP variables for top/src/chrome/browser/ui/browser.h are:
No flag => CHROME_BROWSER_UI_BROWSER_H_
--root=chrome => BROWSER_UI_BROWSER_H_
--root=chrome/browser => UI_BROWSER_H_
+ --root=.. => SRC_CHROME_BROWSER_UI_BROWSER_H_
linelength=digits
This is the allowed line length for the project. The default value is
Examples:
--extensions=hpp,cpp
+
+ headers=x,y,...
+ The header extensions that cpplint will treat as .h in checks. Values are
+ automatically added to --extensions list.
+
+ Examples:
+ --headers=hpp,hxx
+ --headers=hpp
+
+ cpplint.py supports per-directory configurations specified in CPPLINT.cfg
+ files. CPPLINT.cfg file can contain a number of key=value pairs.
+ Currently the following options are supported:
+
+ set noparent
+ filter=+filter1,-filter2,...
+ exclude_files=regex
+ linelength=80
+ root=subdir
+ headers=x,y,...
+
+ "set noparent" option prevents cpplint from traversing directory tree
+ upwards looking for more .cfg files in parent directories. This option
+ is usually placed in the top-level project directory.
+
+ The "filter" option is similar in function to --filter flag. It specifies
+ message filters in addition to the |_DEFAULT_FILTERS| and those specified
+ through --filter command-line flag.
+
+ "exclude_files" allows to specify a regular expression to be matched against
+ a file name. If the expression matches, the file is skipped and not run
+ through liner.
+
+ "linelength" allows to specify the allowed line length for the project.
+
+ The "root" option is similar in function to the --root flag (see example
+ above). Paths are relative to the directory of the CPPLINT.cfg.
+
+ The "headers" option is similar in function to the --headers flag
+ (see example above).
+
+ CPPLINT.cfg has an effect on files in the same directory and all
+ sub-directories, unless overridden by a nested configuration file.
+
+ Example file:
+ filter=-build/include_order,+build/include_alpha
+ exclude_files=.*\.cc
+
+ The above example disables build/include_order warning and enables
+ build/include_alpha as well as excludes all .cc from being
+ processed by linter, in the current directory (where the .cfg
+ file is located) and all sub-directories.
"""
# We categorize each error message we print. Here are the categories.
# If you add a new error message with a new category, add it to the list
# here! cpplint_unittest.py should tell you if you forget to do this.
_ERROR_CATEGORIES = [
- 'build/class',
- 'build/deprecated',
- 'build/endif_comment',
- 'build/explicit_make_pair',
- 'build/forward_decl',
- 'build/header_guard',
- 'build/include',
- 'build/include_alpha',
- 'build/include_order',
- 'build/include_what_you_use',
- 'build/namespaces',
- 'build/printf_format',
- 'build/storage_class',
- 'legal/copyright',
- 'readability/alt_tokens',
- 'readability/braces',
- 'readability/casting',
- 'readability/check',
- 'readability/constructors',
- 'readability/fn_size',
- 'readability/function',
- 'readability/multiline_comment',
- 'readability/multiline_string',
- 'readability/namespace',
- 'readability/nolint',
- 'readability/nul',
- 'readability/streams',
- 'readability/todo',
- 'readability/utf8',
- 'runtime/arrays',
- 'runtime/casting',
- 'runtime/explicit',
- 'runtime/int',
- 'runtime/init',
- 'runtime/invalid_increment',
- 'runtime/member_string_references',
- 'runtime/memset',
- 'runtime/operator',
- 'runtime/printf',
- 'runtime/printf_format',
- 'runtime/references',
- 'runtime/sizeof',
- 'runtime/string',
- 'runtime/threadsafe_fn',
- 'runtime/vlog',
- 'whitespace/blank_line',
- 'whitespace/braces',
- 'whitespace/comma',
- 'whitespace/comments',
- 'whitespace/empty_conditional_body',
- 'whitespace/empty_loop_body',
- 'whitespace/end_of_line',
- 'whitespace/ending_newline',
- 'whitespace/forcolon',
- 'whitespace/indent',
- 'whitespace/line_length',
- 'whitespace/newline',
- 'whitespace/operators',
- 'whitespace/parens',
- 'whitespace/semicolon',
- 'whitespace/tab',
- 'whitespace/todo'
- ]
-
-# The default state of the category filter. This is overrided by the --filter=
+ 'build/class',
+ 'build/c++11',
+ 'build/c++14',
+ 'build/c++tr1',
+ 'build/deprecated',
+ 'build/endif_comment',
+ 'build/explicit_make_pair',
+ 'build/forward_decl',
+ 'build/header_guard',
+ 'build/include',
+ 'build/include_alpha',
+ 'build/include_order',
+ 'build/include_what_you_use',
+ 'build/namespaces',
+ 'build/printf_format',
+ 'build/storage_class',
+ 'legal/copyright',
+ 'readability/alt_tokens',
+ 'readability/braces',
+ 'readability/casting',
+ 'readability/check',
+ 'readability/constructors',
+ 'readability/fn_size',
+ 'readability/inheritance',
+ 'readability/multiline_comment',
+ 'readability/multiline_string',
+ 'readability/namespace',
+ 'readability/nolint',
+ 'readability/nul',
+ 'readability/strings',
+ 'readability/todo',
+ 'readability/utf8',
+ 'runtime/arrays',
+ 'runtime/casting',
+ 'runtime/explicit',
+ 'runtime/int',
+ 'runtime/init',
+ 'runtime/invalid_increment',
+ 'runtime/member_string_references',
+ 'runtime/memset',
+ 'runtime/indentation_namespace',
+ 'runtime/operator',
+ 'runtime/printf',
+ 'runtime/printf_format',
+ 'runtime/references',
+ 'runtime/string',
+ 'runtime/threadsafe_fn',
+ 'runtime/vlog',
+ 'whitespace/blank_line',
+ 'whitespace/braces',
+ 'whitespace/comma',
+ 'whitespace/comments',
+ 'whitespace/empty_conditional_body',
+ 'whitespace/empty_if_body',
+ 'whitespace/empty_loop_body',
+ 'whitespace/end_of_line',
+ 'whitespace/ending_newline',
+ 'whitespace/forcolon',
+ 'whitespace/indent',
+ 'whitespace/line_length',
+ 'whitespace/newline',
+ 'whitespace/operators',
+ 'whitespace/parens',
+ 'whitespace/semicolon',
+ 'whitespace/tab',
+ 'whitespace/todo',
+ ]
+
+# These error categories are no longer enforced by cpplint, but for backwards-
+# compatibility they may still appear in NOLINT comments.
+_LEGACY_ERROR_CATEGORIES = [
+ 'readability/streams',
+ 'readability/function',
+ ]
+
+# The default state of the category filter. This is overridden by the --filter=
# flag. By default all errors are on, so only add here categories that should be
# off by default (i.e., categories that must be enabled by the --filter= flags).
# All entries here should start with a '-' or '+', as in the --filter= flag.
_DEFAULT_FILTERS = ['-build/include_alpha']
+# The default list of categories suppressed for C (not C++) files.
+_DEFAULT_C_SUPPRESSED_CATEGORIES = [
+ 'readability/casting',
+ ]
+
+# The default list of categories suppressed for Linux Kernel files.
+_DEFAULT_KERNEL_SUPPRESSED_CATEGORIES = [
+ 'whitespace/tab',
+ ]
+
# We used to check for high-bit characters, but after much discussion we
# decided those were OK, as long as they were in UTF-8 and didn't represent
# hard-coded international strings, which belong in a separate i18n file.
-
# C++ headers
_CPP_HEADERS = frozenset([
# Legacy
'random',
'ratio',
'regex',
+ 'scoped_allocator',
'set',
'sstream',
'stack',
'cwctype',
])
+# Type names
+_TYPES = re.compile(
+ r'^(?:'
+ # [dcl.type.simple]
+ r'(char(16_t|32_t)?)|wchar_t|'
+ r'bool|short|int|long|signed|unsigned|float|double|'
+ # [support.types]
+ r'(ptrdiff_t|size_t|max_align_t|nullptr_t)|'
+ # [cstdint.syn]
+ r'(u?int(_fast|_least)?(8|16|32|64)_t)|'
+ r'(u?int(max|ptr)_t)|'
+ r')$')
+
+
+# These headers are excluded from [build/include] and [build/include_order]
+# checks:
+# - Anything not following google file name conventions (containing an
+# uppercase character, such as Python.h or nsStringAPI.h, for example).
+# - Lua headers.
+_THIRD_PARTY_HEADERS_PATTERN = re.compile(
+ r'^(?:[^/]*[A-Z][^/]*\.h|lua\.h|lauxlib\.h|lualib\.h)$')
+
+# Pattern for matching FileInfo.BaseName() against test file name
+_TEST_FILE_SUFFIX = r'(_test|_unittest|_regtest)$'
+
+# Pattern that matches only complete whitespace, possibly across multiple lines.
+_EMPTY_CONDITIONAL_BODY_PATTERN = re.compile(r'^\s*$', re.DOTALL)
+
# Assertion macros. These are defined in base/logging.h and
-# testing/base/gunit.h. Note that the _M versions need to come first
-# for substring matching to work.
+# testing/base/public/gunit.h.
_CHECK_MACROS = [
'DCHECK', 'CHECK',
- 'EXPECT_TRUE_M', 'EXPECT_TRUE',
- 'ASSERT_TRUE_M', 'ASSERT_TRUE',
- 'EXPECT_FALSE_M', 'EXPECT_FALSE',
- 'ASSERT_FALSE_M', 'ASSERT_FALSE',
+ 'EXPECT_TRUE', 'ASSERT_TRUE',
+ 'EXPECT_FALSE', 'ASSERT_FALSE',
]
# Replacement macros for CHECK/DCHECK/EXPECT_TRUE/EXPECT_FALSE
_CHECK_REPLACEMENT['CHECK'][op] = 'CHECK_%s' % replacement
_CHECK_REPLACEMENT['EXPECT_TRUE'][op] = 'EXPECT_%s' % replacement
_CHECK_REPLACEMENT['ASSERT_TRUE'][op] = 'ASSERT_%s' % replacement
- _CHECK_REPLACEMENT['EXPECT_TRUE_M'][op] = 'EXPECT_%s_M' % replacement
- _CHECK_REPLACEMENT['ASSERT_TRUE_M'][op] = 'ASSERT_%s_M' % replacement
for op, inv_replacement in [('==', 'NE'), ('!=', 'EQ'),
('>=', 'LT'), ('>', 'LE'),
('<=', 'GT'), ('<', 'GE')]:
_CHECK_REPLACEMENT['EXPECT_FALSE'][op] = 'EXPECT_%s' % inv_replacement
_CHECK_REPLACEMENT['ASSERT_FALSE'][op] = 'ASSERT_%s' % inv_replacement
- _CHECK_REPLACEMENT['EXPECT_FALSE_M'][op] = 'EXPECT_%s_M' % inv_replacement
- _CHECK_REPLACEMENT['ASSERT_FALSE_M'][op] = 'ASSERT_%s_M' % inv_replacement
# Alternative tokens and their replacements. For full list, see section 2.5
# Alternative tokens [lex.digraph] in the C++ standard.
r'(?:\s+(volatile|__volatile__))?'
r'\s*[{(]')
+# Match strings that indicate we're working on a C (not C++) file.
+_SEARCH_C_FILE = re.compile(r'\b(?:LINT_C_FILE|'
+ r'vim?:\s*.*(\s*|:)filetype=c(\s*|:|$))')
-_regexp_compile_cache = {}
+# Match string that indicates we're working on a Linux Kernel file.
+_SEARCH_KERNEL_FILE = re.compile(r'\b(?:LINT_KERNEL_FILE)')
-# Finds occurrences of NOLINT or NOLINT(...).
-_RE_SUPPRESSION = re.compile(r'\bNOLINT\b(\([^)]*\))?')
+_regexp_compile_cache = {}
# {str, set(int)}: a map from error categories to sets of linenumbers
# on which those errors are expected and should be suppressed.
# The root directory used for deriving header guard CPP variable.
# This is set by --root flag.
_root = None
+_root_debug = False
# The allowed line length of files.
# This is set by --linelength flag.
# This is set by --extensions flag.
_valid_extensions = set(['cc', 'h', 'cpp', 'cu', 'cuh'])
+# Treat all headers starting with 'h' equally: .h, .hpp, .hxx etc.
+# This is set by --headers flag.
+_hpp_headers = set(['h'])
+
+# {str, bool}: a map from error categories to booleans which indicate if the
+# category should be suppressed for every line.
+_global_error_suppressions = {}
+
+def ProcessHppHeadersOption(val):
+ global _hpp_headers
+ try:
+ _hpp_headers = set(val.split(','))
+ # Automatically append to extensions list so it does not have to be set 2 times
+ _valid_extensions.update(_hpp_headers)
+ except ValueError:
+ PrintUsage('Header extensions must be comma separated list.')
+
+def IsHeaderExtension(file_extension):
+ return file_extension in _hpp_headers
+
def ParseNolintSuppressions(filename, raw_line, linenum, error):
- """Updates the global list of error-suppressions.
+ """Updates the global list of line error-suppressions.
Parses any NOLINT comments on the current line, updating the global
error_suppressions store. Reports an error if the NOLINT comment
linenum: int, the number of the current line.
error: function, an error handler.
"""
- # FIXME(adonovan): "NOLINT(" is misparsed as NOLINT(*).
- matched = _RE_SUPPRESSION.search(raw_line)
+ matched = Search(r'\bNOLINT(NEXTLINE)?\b(\([^)]+\))?', raw_line)
if matched:
- category = matched.group(1)
+ if matched.group(1):
+ suppressed_line = linenum + 1
+ else:
+ suppressed_line = linenum
+ category = matched.group(2)
if category in (None, '(*)'): # => "suppress all"
- _error_suppressions.setdefault(None, set()).add(linenum)
+ _error_suppressions.setdefault(None, set()).add(suppressed_line)
else:
if category.startswith('(') and category.endswith(')'):
category = category[1:-1]
if category in _ERROR_CATEGORIES:
- _error_suppressions.setdefault(category, set()).add(linenum)
- else:
+ _error_suppressions.setdefault(category, set()).add(suppressed_line)
+ elif category not in _LEGACY_ERROR_CATEGORIES:
error(filename, linenum, 'readability/nolint', 5,
'Unknown NOLINT error category: %s' % category)
+def ProcessGlobalSuppresions(lines):
+ """Updates the list of global error suppressions.
+
+ Parses any lint directives in the file that have global effect.
+
+ Args:
+ lines: An array of strings, each representing a line of the file, with the
+ last element being empty if the file is terminated with a newline.
+ """
+ for line in lines:
+ if _SEARCH_C_FILE.search(line):
+ for category in _DEFAULT_C_SUPPRESSED_CATEGORIES:
+ _global_error_suppressions[category] = True
+ if _SEARCH_KERNEL_FILE.search(line):
+ for category in _DEFAULT_KERNEL_SUPPRESSED_CATEGORIES:
+ _global_error_suppressions[category] = True
+
+
def ResetNolintSuppressions():
- "Resets the set of NOLINT suppressions to empty."
+ """Resets the set of NOLINT suppressions to empty."""
_error_suppressions.clear()
+ _global_error_suppressions.clear()
def IsErrorSuppressedByNolint(category, linenum):
"""Returns true if the specified error category is suppressed on this line.
Consults the global error_suppressions map populated by
- ParseNolintSuppressions/ResetNolintSuppressions.
+ ParseNolintSuppressions/ProcessGlobalSuppresions/ResetNolintSuppressions.
Args:
category: str, the category of the error.
linenum: int, the current line number.
Returns:
- bool, True iff the error should be suppressed due to a NOLINT comment.
+ bool, True iff the error should be suppressed due to a NOLINT comment or
+ global suppression.
"""
- return (linenum in _error_suppressions.get(category, set()) or
+ return (_global_error_suppressions.get(category, False) or
+ linenum in _error_suppressions.get(category, set()) or
linenum in _error_suppressions.get(None, set()))
+
def Match(pattern, s):
"""Matches the string with the pattern, caching the compiled regexp."""
# The regexp compilation caching is inlined in both Match and Search for
return _regexp_compile_cache[pattern].search(s)
-class _IncludeState(dict):
+def _IsSourceExtension(s):
+ """File extension (excluding dot) matches a source file extension."""
+ return s in ('c', 'cc', 'cpp', 'cxx')
+
+
+class _IncludeState(object):
"""Tracks line numbers for includes, and the order in which includes appear.
- As a dict, an _IncludeState object serves as a mapping between include
- filename and line number on which that file was included.
+ include_list contains list of lists of (header, line number) pairs.
+ It's a lists of lists rather than just one flat list to make it
+ easier to update across preprocessor boundaries.
Call CheckNextIncludeOrder() once for each header in the file, passing
in the type constants defined above. Calls in an illegal order will
}
def __init__(self):
- dict.__init__(self)
- self.ResetSection()
+ self.include_list = [[]]
+ self.ResetSection('')
+
+ def FindHeader(self, header):
+ """Check if a header has already been included.
+
+ Args:
+ header: header to check.
+ Returns:
+ Line number of previous occurrence, or -1 if the header has not
+ been seen before.
+ """
+ for section_list in self.include_list:
+ for f in section_list:
+ if f[0] == header:
+ return f[1]
+ return -1
+
+ def ResetSection(self, directive):
+ """Reset section checking for preprocessor directive.
- def ResetSection(self):
+ Args:
+ directive: preprocessor directive (e.g. "if", "else").
+ """
# The name of the current section.
self._section = self._INITIAL_SECTION
# The path of last found header.
self._last_header = ''
+ # Update list of includes. Note that we never pop from the
+ # include list.
+ if directive in ('if', 'ifdef', 'ifndef'):
+ self.include_list.append([])
+ elif directive in ('else', 'elif'):
+ self.include_list[-1] = []
+
def SetLastHeader(self, header_path):
self._last_header = header_path
# If previous line was a blank line, assume that the headers are
# intentionally sorted the way they are.
if (self._last_header > header_path and
- not Match(r'^\s*$', clean_lines.elided[linenum - 1])):
+ Match(r'^\s*#\s*include\b', clean_lines.elided[linenum - 1])):
return False
return True
self.error_count = 0 # global count of reported errors
# filters to apply when emitting error messages
self.filters = _DEFAULT_FILTERS[:]
+ # backup of filter list. Used to restore the state after each file.
+ self._filters_backup = self.filters[:]
self.counting = 'total' # In what way are we counting errors?
self.errors_by_category = {} # string to int dict storing error counts
+ self.quiet = False # Suppress non-error messagess?
# output format:
# "emacs" - format that emacs can parse (default)
"""Sets the output format for errors."""
self.output_format = output_format
+ def SetQuiet(self, quiet):
+ """Sets the module's quiet settings, and returns the previous setting."""
+ last_quiet = self.quiet
+ self.quiet = quiet
+ return last_quiet
+
def SetVerboseLevel(self, level):
"""Sets the module's verbosity, and returns the previous setting."""
last_verbose_level = self.verbose_level
"""
# Default filters always have less priority than the flag ones.
self.filters = _DEFAULT_FILTERS[:]
+ self.AddFilters(filters)
+
+ def AddFilters(self, filters):
+ """ Adds more filters to the existing list of error-message filters. """
for filt in filters.split(','):
clean_filt = filt.strip()
if clean_filt:
raise ValueError('Every filter in --filters must start with + or -'
' (%s does not)' % filt)
+ def BackupFilters(self):
+ """ Saves the current filter list to backup storage."""
+ self._filters_backup = self.filters[:]
+
+ def RestoreFilters(self):
+ """ Restores filters previously backed up."""
+ self.filters = self._filters_backup[:]
+
def ResetErrorCounts(self):
"""Sets the module's error statistic back to zero."""
self.error_count = 0
for category, count in self.errors_by_category.iteritems():
sys.stderr.write('Category \'%s\' errors found: %d\n' %
(category, count))
- sys.stderr.write('Total errors found: %d\n' % self.error_count)
+ sys.stdout.write('Total errors found: %d\n' % self.error_count)
_cpplint_state = _CppLintState()
"""Sets the module's output format."""
_cpplint_state.SetOutputFormat(output_format)
+def _Quiet():
+ """Return's the module's quiet setting."""
+ return _cpplint_state.quiet
+
+def _SetQuiet(quiet):
+ """Set the module's quiet status, and return previous setting."""
+ return _cpplint_state.SetQuiet(quiet)
+
def _VerboseLevel():
"""Returns the module's verbosity setting."""
"""
_cpplint_state.SetFilters(filters)
+def _AddFilters(filters):
+ """Adds more filter overrides.
+
+ Unlike _SetFilters, this function does not reset the current list of filters
+ available.
+
+ Args:
+ filters: A string of comma-separated filters (eg "whitespace/indent").
+ Each filter should start with + or -; else we die.
+ """
+ _cpplint_state.AddFilters(filters)
+
+def _BackupFilters():
+ """ Saves the current filter list to backup storage."""
+ _cpplint_state.BackupFilters()
+
+def _RestoreFilters():
+ """ Restores filters previously backed up."""
+ _cpplint_state.RestoreFilters()
class _FunctionState(object):
"""Tracks current function name and the number of lines in its body."""
filename: The name of the current file.
linenum: The number of the line to check.
"""
+ if not self.in_a_function:
+ return
+
if Match(r'T(EST|est)', self.current_function):
base_trigger = self._TEST_TRIGGER
else:
pass
-class FileInfo:
+class FileInfo(object):
"""Provides utility functions for filenames.
FileInfo provides easy access to the components of a file's path
# Not SVN <= 1.6? Try to find a git, hg, or svn top level directory by
# searching up from the current path.
- root_dir = os.path.dirname(fullname)
- while (root_dir != os.path.dirname(root_dir) and
- not os.path.exists(os.path.join(root_dir, ".git")) and
- not os.path.exists(os.path.join(root_dir, ".hg")) and
- not os.path.exists(os.path.join(root_dir, ".svn"))):
- root_dir = os.path.dirname(root_dir)
+ root_dir = current_dir = os.path.dirname(fullname)
+ while current_dir != os.path.dirname(current_dir):
+ if (os.path.exists(os.path.join(current_dir, ".git")) or
+ os.path.exists(os.path.join(current_dir, ".hg")) or
+ os.path.exists(os.path.join(current_dir, ".svn"))):
+ root_dir = current_dir
+ current_dir = os.path.dirname(current_dir)
if (os.path.exists(os.path.join(root_dir, ".git")) or
os.path.exists(os.path.join(root_dir, ".hg")) or
def IsSource(self):
"""File has a source file extension."""
- return self.Extension()[1:] in ('c', 'cc', 'cpp', 'cxx')
+ return _IsSourceExtension(self.Extension()[1:])
def _ShouldPrintError(category, confidence, linenum):
# the verbosity level isn't high enough, or the filters filter it out.
if IsErrorSuppressedByNolint(category, linenum):
return False
+
if confidence < _cpplint_state.verbose_level:
return False
if _ShouldPrintError(category, confidence, linenum):
_cpplint_state.IncrementErrorCount(category)
if _cpplint_state.output_format == 'vs7':
- sys.stderr.write('%s(%s): %s [%s] [%d]\n' % (
- filename, linenum, message, category, confidence))
+ sys.stderr.write('%s(%s): error cpplint: [%s] %s [%d]\n' % (
+ filename, linenum, category, message, confidence))
elif _cpplint_state.output_format == 'eclipse':
sys.stderr.write('%s:%s: warning: %s [%s] [%d]\n' % (
filename, linenum, message, category, confidence))
# Matches standard C++ escape sequences per 2.13.2.3 of the C++ standard.
_RE_PATTERN_CLEANSE_LINE_ESCAPES = re.compile(
r'\\([abfnrtv?"\\\']|\d+|x[0-9a-fA-F]+)')
-# Matches strings. Escape codes should already be removed by ESCAPES.
-_RE_PATTERN_CLEANSE_LINE_DOUBLE_QUOTES = re.compile(r'"[^"]*"')
-# Matches characters. Escape codes should already be removed by ESCAPES.
-_RE_PATTERN_CLEANSE_LINE_SINGLE_QUOTES = re.compile(r"'.'")
-# Matches multi-line C++ comments.
+# Match a single C style comment on the same line.
+_RE_PATTERN_C_COMMENTS = r'/\*(?:[^*]|\*(?!/))*\*/'
+# Matches multi-line C style comments.
# This RE is a little bit more complicated than one might expect, because we
# have to take care of space removals tools so we can handle comments inside
# statements better.
# if this doesn't work we try on left side but only if there's a non-character
# on the right.
_RE_PATTERN_CLEANSE_LINE_C_COMMENTS = re.compile(
- r"""(\s*/\*.*\*/\s*$|
- /\*.*\*/\s+|
- \s+/\*.*\*/(?=\W)|
- /\*.*\*/)""", re.VERBOSE)
+ r'(\s*' + _RE_PATTERN_C_COMMENTS + r'\s*$|' +
+ _RE_PATTERN_C_COMMENTS + r'\s+|' +
+ r'\s+' + _RE_PATTERN_C_COMMENTS + r'(?=\W)|' +
+ _RE_PATTERN_C_COMMENTS + r')')
def IsCppString(line):
delimiter = None
else:
# Haven't found the end yet, append a blank line.
- line = ''
+ line = '""'
- else:
+ # Look for beginning of a raw string, and replace them with
+ # empty strings. This is done in a loop to handle multiple raw
+ # strings on the same line.
+ while delimiter is None:
# Look for beginning of a raw string.
# See 2.14.15 [lex.string] for syntax.
- matched = Match(r'^(.*)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line)
- if matched:
+ #
+ # Once we have matched a raw string, we check the prefix of the
+ # line to make sure that the line is not part of a single line
+ # comment. It's done this way because we remove raw strings
+ # before removing comments as opposed to removing comments
+ # before removing raw strings. This is because there are some
+ # cpplint checks that requires the comments to be preserved, but
+ # we don't want to check comments that are inside raw strings.
+ matched = Match(r'^(.*?)\b(?:R|u8R|uR|UR|LR)"([^\s\\()]*)\((.*)$', line)
+ if (matched and
+ not Match(r'^([^\'"]|\'(\\.|[^\'])*\'|"(\\.|[^"])*")*//',
+ matched.group(1))):
delimiter = ')' + matched.group(2) + '"'
end = matched.group(3).find(delimiter)
else:
# Start of a multi-line raw string
line = matched.group(1) + '""'
+ else:
+ break
lines_without_raw_strings.append(line)
def RemoveMultiLineCommentsFromRange(lines, begin, end):
"""Clears a range of lines for multi-line comments."""
- # Having // dummy comments makes the lines non-empty, so we will not get
+ # Having // <empty> comments makes the lines non-empty, so we will not get
# unnecessary blank line warnings later in the code.
for i in range(begin, end):
- lines[i] = '// dummy'
+ lines[i] = '/**/'
def RemoveMultiLineComments(filename, lines, error):
class CleansedLines(object):
- """Holds 3 copies of all lines with different preprocessing applied to them.
+ """Holds 4 copies of all lines with different preprocessing applied to them.
- 1) elided member contains lines without strings and comments,
- 2) lines member contains lines without comments, and
+ 1) elided member contains lines without strings and comments.
+ 2) lines member contains lines without comments.
3) raw_lines member contains all the lines without processing.
- All these three members are of <type 'list'>, and of the same length.
+ 4) lines_without_raw_strings member is same as raw_lines, but with C++11 raw
+ strings removed.
+ All these members are of <type 'list'>, and of the same length.
"""
def __init__(self, lines):
Returns:
The line with collapsed strings.
"""
- if not _RE_PATTERN_INCLUDE.match(elided):
- # Remove escaped characters first to make quote/single quote collapsing
- # basic. Things that look like escaped characters shouldn't occur
- # outside of strings and chars.
- elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided)
- elided = _RE_PATTERN_CLEANSE_LINE_SINGLE_QUOTES.sub("''", elided)
- elided = _RE_PATTERN_CLEANSE_LINE_DOUBLE_QUOTES.sub('""', elided)
- return elided
+ if _RE_PATTERN_INCLUDE.match(elided):
+ return elided
+
+ # Remove escaped characters first to make quote/single quote collapsing
+ # basic. Things that look like escaped characters shouldn't occur
+ # outside of strings and chars.
+ elided = _RE_PATTERN_CLEANSE_LINE_ESCAPES.sub('', elided)
+
+ # Replace quoted strings and digit separators. Both single quotes
+ # and double quotes are processed in the same loop, otherwise
+ # nested quotes wouldn't work.
+ collapsed = ''
+ while True:
+ # Find the first quote character
+ match = Match(r'^([^\'"]*)([\'"])(.*)$', elided)
+ if not match:
+ collapsed += elided
+ break
+ head, quote, tail = match.groups()
+
+ if quote == '"':
+ # Collapse double quoted strings
+ second_quote = tail.find('"')
+ if second_quote >= 0:
+ collapsed += head + '""'
+ elided = tail[second_quote + 1:]
+ else:
+ # Unmatched double quote, don't bother processing the rest
+ # of the line since this is probably a multiline string.
+ collapsed += elided
+ break
+ else:
+ # Found single quote, check nearby text to eliminate digit separators.
+ #
+ # There is no special handling for floating point here, because
+ # the integer/fractional/exponent parts would all be parsed
+ # correctly as long as there are digits on both sides of the
+ # separator. So we are fine as long as we don't see something
+ # like "0.'3" (gcc 4.9.0 will not allow this literal).
+ if Search(r'\b(?:0[bBxX]?|[1-9])[0-9a-fA-F]*$', head):
+ match_literal = Match(r'^((?:\'?[0-9a-zA-Z_])*)(.*)$', "'" + tail)
+ collapsed += head + match_literal.group(1).replace("'", '')
+ elided = match_literal.group(2)
+ else:
+ second_quote = tail.find('\'')
+ if second_quote >= 0:
+ collapsed += head + "''"
+ elided = tail[second_quote + 1:]
+ else:
+ # Unmatched single quote
+ collapsed += elided
+ break
+ return collapsed
-def FindEndOfExpressionInLine(line, startpos, depth, startchar, endchar):
- """Find the position just after the matching endchar.
+
+def FindEndOfExpressionInLine(line, startpos, stack):
+ """Find the position just after the end of current parenthesized expression.
Args:
line: a CleansedLines line.
startpos: start searching at this position.
- depth: nesting level at startpos.
- startchar: expression opening character.
- endchar: expression closing character.
+ stack: nesting stack at startpos.
Returns:
- On finding matching endchar: (index just after matching endchar, 0)
- Otherwise: (-1, new depth at end of this line)
+ On finding matching end: (index just after matching end, None)
+ On finding an unclosed expression: (-1, None)
+ Otherwise: (-1, new stack at end of this line)
"""
for i in xrange(startpos, len(line)):
- if line[i] == startchar:
- depth += 1
- elif line[i] == endchar:
- depth -= 1
- if depth == 0:
- return (i + 1, 0)
- return (-1, depth)
+ char = line[i]
+ if char in '([{':
+ # Found start of parenthesized expression, push to expression stack
+ stack.append(char)
+ elif char == '<':
+ # Found potential start of template argument list
+ if i > 0 and line[i - 1] == '<':
+ # Left shift operator
+ if stack and stack[-1] == '<':
+ stack.pop()
+ if not stack:
+ return (-1, None)
+ elif i > 0 and Search(r'\boperator\s*$', line[0:i]):
+ # operator<, don't add to stack
+ continue
+ else:
+ # Tentative start of template argument list
+ stack.append('<')
+ elif char in ')]}':
+ # Found end of parenthesized expression.
+ #
+ # If we are currently expecting a matching '>', the pending '<'
+ # must have been an operator. Remove them from expression stack.
+ while stack and stack[-1] == '<':
+ stack.pop()
+ if not stack:
+ return (-1, None)
+ if ((stack[-1] == '(' and char == ')') or
+ (stack[-1] == '[' and char == ']') or
+ (stack[-1] == '{' and char == '}')):
+ stack.pop()
+ if not stack:
+ return (i + 1, None)
+ else:
+ # Mismatched parentheses
+ return (-1, None)
+ elif char == '>':
+ # Found potential end of template argument list.
+
+ # Ignore "->" and operator functions
+ if (i > 0 and
+ (line[i - 1] == '-' or Search(r'\boperator\s*$', line[0:i - 1]))):
+ continue
+
+ # Pop the stack if there is a matching '<'. Otherwise, ignore
+ # this '>' since it must be an operator.
+ if stack:
+ if stack[-1] == '<':
+ stack.pop()
+ if not stack:
+ return (i + 1, None)
+ elif char == ';':
+ # Found something that look like end of statements. If we are currently
+ # expecting a '>', the matching '<' must have been an operator, since
+ # template argument list should not contain statements.
+ while stack and stack[-1] == '<':
+ stack.pop()
+ if not stack:
+ return (-1, None)
+
+ # Did not find end of expression or unbalanced parentheses on this line
+ return (-1, stack)
def CloseExpression(clean_lines, linenum, pos):
If lines[linenum][pos] points to a '(' or '{' or '[' or '<', finds the
linenum/pos that correspond to the closing of the expression.
+ TODO(unknown): cpplint spends a fair bit of time matching parentheses.
+ Ideally we would want to index all opening and closing parentheses once
+ and have CloseExpression be just a simple lookup, but due to preprocessor
+ tricks, this is not so easy.
+
Args:
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
"""
line = clean_lines.elided[linenum]
- startchar = line[pos]
- if startchar not in '({[<':
+ if (line[pos] not in '({[<') or Match(r'<[<=]', line[pos:]):
return (line, clean_lines.NumLines(), -1)
- if startchar == '(': endchar = ')'
- if startchar == '[': endchar = ']'
- if startchar == '{': endchar = '}'
- if startchar == '<': endchar = '>'
# Check first line
- (end_pos, num_open) = FindEndOfExpressionInLine(
- line, pos, 0, startchar, endchar)
+ (end_pos, stack) = FindEndOfExpressionInLine(line, pos, [])
if end_pos > -1:
return (line, linenum, end_pos)
# Continue scanning forward
- while linenum < clean_lines.NumLines() - 1:
+ while stack and linenum < clean_lines.NumLines() - 1:
linenum += 1
line = clean_lines.elided[linenum]
- (end_pos, num_open) = FindEndOfExpressionInLine(
- line, 0, num_open, startchar, endchar)
+ (end_pos, stack) = FindEndOfExpressionInLine(line, 0, stack)
if end_pos > -1:
return (line, linenum, end_pos)
- # Did not find endchar before end of file, give up
+ # Did not find end of expression before end of file, give up
return (line, clean_lines.NumLines(), -1)
-def FindStartOfExpressionInLine(line, endpos, depth, startchar, endchar):
- """Find position at the matching startchar.
+def FindStartOfExpressionInLine(line, endpos, stack):
+ """Find position at the matching start of current expression.
This is almost the reverse of FindEndOfExpressionInLine, but note
that the input position and returned position differs by 1.
Args:
line: a CleansedLines line.
endpos: start searching at this position.
- depth: nesting level at endpos.
- startchar: expression opening character.
- endchar: expression closing character.
+ stack: nesting stack at endpos.
Returns:
- On finding matching startchar: (index at matching startchar, 0)
- Otherwise: (-1, new depth at beginning of this line)
+ On finding matching start: (index at matching start, None)
+ On finding an unclosed expression: (-1, None)
+ Otherwise: (-1, new stack at beginning of this line)
"""
- for i in xrange(endpos, -1, -1):
- if line[i] == endchar:
- depth += 1
- elif line[i] == startchar:
- depth -= 1
- if depth == 0:
- return (i, 0)
- return (-1, depth)
+ i = endpos
+ while i >= 0:
+ char = line[i]
+ if char in ')]}':
+ # Found end of expression, push to expression stack
+ stack.append(char)
+ elif char == '>':
+ # Found potential end of template argument list.
+ #
+ # Ignore it if it's a "->" or ">=" or "operator>"
+ if (i > 0 and
+ (line[i - 1] == '-' or
+ Match(r'\s>=\s', line[i - 1:]) or
+ Search(r'\boperator\s*$', line[0:i]))):
+ i -= 1
+ else:
+ stack.append('>')
+ elif char == '<':
+ # Found potential start of template argument list
+ if i > 0 and line[i - 1] == '<':
+ # Left shift operator
+ i -= 1
+ else:
+ # If there is a matching '>', we can pop the expression stack.
+ # Otherwise, ignore this '<' since it must be an operator.
+ if stack and stack[-1] == '>':
+ stack.pop()
+ if not stack:
+ return (i, None)
+ elif char in '([{':
+ # Found start of expression.
+ #
+ # If there are any unmatched '>' on the stack, they must be
+ # operators. Remove those.
+ while stack and stack[-1] == '>':
+ stack.pop()
+ if not stack:
+ return (-1, None)
+ if ((char == '(' and stack[-1] == ')') or
+ (char == '[' and stack[-1] == ']') or
+ (char == '{' and stack[-1] == '}')):
+ stack.pop()
+ if not stack:
+ return (i, None)
+ else:
+ # Mismatched parentheses
+ return (-1, None)
+ elif char == ';':
+ # Found something that look like end of statements. If we are currently
+ # expecting a '<', the matching '>' must have been an operator, since
+ # template argument list should not contain statements.
+ while stack and stack[-1] == '>':
+ stack.pop()
+ if not stack:
+ return (-1, None)
+
+ i -= 1
+
+ return (-1, stack)
def ReverseCloseExpression(clean_lines, linenum, pos):
return is the 'cleansed' line at linenum.
"""
line = clean_lines.elided[linenum]
- endchar = line[pos]
- if endchar not in ')}]>':
+ if line[pos] not in ')}]>':
return (line, 0, -1)
- if endchar == ')': startchar = '('
- if endchar == ']': startchar = '['
- if endchar == '}': startchar = '{'
- if endchar == '>': startchar = '<'
# Check last line
- (start_pos, num_open) = FindStartOfExpressionInLine(
- line, pos, 0, startchar, endchar)
+ (start_pos, stack) = FindStartOfExpressionInLine(line, pos, [])
if start_pos > -1:
return (line, linenum, start_pos)
# Continue scanning backward
- while linenum > 0:
+ while stack and linenum > 0:
linenum -= 1
line = clean_lines.elided[linenum]
- (start_pos, num_open) = FindStartOfExpressionInLine(
- line, len(line) - 1, num_open, startchar, endchar)
+ (start_pos, stack) = FindStartOfExpressionInLine(line, len(line) - 1, stack)
if start_pos > -1:
return (line, linenum, start_pos)
- # Did not find startchar before beginning of file, give up
+ # Did not find start of expression before beginning of file, give up
return (line, 0, -1)
"""Logs an error if no Copyright message appears at the top of the file."""
# We'll say it should occur by line 10. Don't forget there's a
- # dummy line at the front.
+ # placeholder line at the front.
for line in xrange(1, min(len(lines), 11)):
if re.search(r'Copyright', lines[line], re.I): break
else: # means no copyright line was found
'You should have a line: "Copyright [year] <Copyright Owner>"')
+def GetIndentLevel(line):
+ """Return the number of leading spaces in line.
+
+ Args:
+ line: A string to check.
+
+ Returns:
+ An integer count of leading spaces, possibly zero.
+ """
+ indent = Match(r'^( *)\S', line)
+ if indent:
+ return len(indent.group(1))
+ else:
+ return 0
+
+def PathSplitToList(path):
+ """Returns the path split into a list by the separator.
+
+ Args:
+ path: An absolute or relative path (e.g. '/a/b/c/' or '../a')
+
+ Returns:
+ A list of path components (e.g. ['a', 'b', 'c]).
+ """
+ lst = []
+ while True:
+ (head, tail) = os.path.split(path)
+ if head == path: # absolute paths end
+ lst.append(head)
+ break
+ if tail == path: # relative paths end
+ lst.append(tail)
+ break
+
+ path = head
+ lst.append(tail)
+
+ lst.reverse()
+ return lst
+
def GetHeaderGuardCPPVariable(filename):
"""Returns the CPP variable that should be used as a header guard.
# flymake.
filename = re.sub(r'_flymake\.h$', '.h', filename)
filename = re.sub(r'/\.flymake/([^/]*)$', r'/\1', filename)
+ # Replace 'c++' with 'cpp'.
+ filename = filename.replace('C++', 'cpp').replace('c++', 'cpp')
fileinfo = FileInfo(filename)
file_path_from_root = fileinfo.RepositoryName()
- if _root:
- file_path_from_root = re.sub('^' + _root + os.sep, '', file_path_from_root)
- return re.sub(r'[-./\s]', '_', file_path_from_root).upper() + '_'
+ def FixupPathFromRoot():
+ if _root_debug:
+ sys.stderr.write("\n_root fixup, _root = '%s', repository name = '%s'\n"
+ %(_root, fileinfo.RepositoryName()))
+
+ # Process the file path with the --root flag if it was set.
+ if not _root:
+ if _root_debug:
+ sys.stderr.write("_root unspecified\n")
+ return file_path_from_root
+
+ def StripListPrefix(lst, prefix):
+ # f(['x', 'y'], ['w, z']) -> None (not a valid prefix)
+ if lst[:len(prefix)] != prefix:
+ return None
+ # f(['a, 'b', 'c', 'd'], ['a', 'b']) -> ['c', 'd']
+ return lst[(len(prefix)):]
+
+ # root behavior:
+ # --root=subdir , lstrips subdir from the header guard
+ maybe_path = StripListPrefix(PathSplitToList(file_path_from_root),
+ PathSplitToList(_root))
+
+ if _root_debug:
+ sys.stderr.write(("_root lstrip (maybe_path=%s, file_path_from_root=%s," +
+ " _root=%s)\n") %(maybe_path, file_path_from_root, _root))
+
+ if maybe_path:
+ return os.path.join(*maybe_path)
+
+ # --root=.. , will prepend the outer directory to the header guard
+ full_path = fileinfo.FullName()
+ root_abspath = os.path.abspath(_root)
+
+ maybe_path = StripListPrefix(PathSplitToList(full_path),
+ PathSplitToList(root_abspath))
+
+ if _root_debug:
+ sys.stderr.write(("_root prepend (maybe_path=%s, full_path=%s, " +
+ "root_abspath=%s)\n") %(maybe_path, full_path, root_abspath))
+
+ if maybe_path:
+ return os.path.join(*maybe_path)
-def CheckForHeaderGuard(filename, lines, error):
+ if _root_debug:
+ sys.stderr.write("_root ignore, returning %s\n" %(file_path_from_root))
+
+ # --root=FAKE_DIR is ignored
+ return file_path_from_root
+
+ file_path_from_root = FixupPathFromRoot()
+ return re.sub(r'[^a-zA-Z0-9]', '_', file_path_from_root).upper() + '_'
+
+
+def CheckForHeaderGuard(filename, clean_lines, error):
"""Checks that the file contains a header guard.
Logs an error if no #ifndef header guard is present. For other
Args:
filename: The name of the C++ header file.
- lines: An array of strings, each representing a line of the file.
+ clean_lines: A CleansedLines instance containing the file.
error: The function to call with any errors found.
"""
+ # Don't check for header guards if there are error suppression
+ # comments somewhere in this file.
+ #
+ # Because this is silencing a warning for a nonexistent line, we
+ # only support the very specific NOLINT(build/header_guard) syntax,
+ # and not the general NOLINT or NOLINT(*) syntax.
+ raw_lines = clean_lines.lines_without_raw_strings
+ for i in raw_lines:
+ if Search(r'//\s*NOLINT\(build/header_guard\)', i):
+ return
+
cppvar = GetHeaderGuardCPPVariable(filename)
- ifndef = None
+ ifndef = ''
ifndef_linenum = 0
- define = None
- endif = None
+ define = ''
+ endif = ''
endif_linenum = 0
- for linenum, line in enumerate(lines):
+ for linenum, line in enumerate(raw_lines):
linesplit = line.split()
if len(linesplit) >= 2:
# find the first occurrence of #ifndef and #define, save arg
endif = line
endif_linenum = linenum
- if not ifndef:
+ if not ifndef or not define or ifndef != define:
error(filename, 0, 'build/header_guard', 5,
'No #ifndef header guard found, suggested CPP variable is: %s' %
cppvar)
return
- if not define:
- error(filename, 0, 'build/header_guard', 5,
- 'No #define header guard found, suggested CPP variable is: %s' %
- cppvar)
- return
-
# The guard should be PATH_FILE_H_, but we also allow PATH_FILE_H__
# for backward compatibility.
if ifndef != cppvar:
if ifndef != cppvar + '_':
error_level = 5
- ParseNolintSuppressions(filename, lines[ifndef_linenum], ifndef_linenum,
+ ParseNolintSuppressions(filename, raw_lines[ifndef_linenum], ifndef_linenum,
error)
error(filename, ifndef_linenum, 'build/header_guard', error_level,
'#ifndef header guard has wrong style, please use: %s' % cppvar)
- if define != ifndef:
- error(filename, 0, 'build/header_guard', 5,
- '#ifndef and #define don\'t match, suggested CPP variable is: %s' %
- cppvar)
+ # Check for "//" comments on endif line.
+ ParseNolintSuppressions(filename, raw_lines[endif_linenum], endif_linenum,
+ error)
+ match = Match(r'#endif\s*//\s*' + cppvar + r'(_)?\b', endif)
+ if match:
+ if match.group(1) == '_':
+ # Issue low severity warning for deprecated double trailing underscore
+ error(filename, endif_linenum, 'build/header_guard', 0,
+ '#endif line should be "#endif // %s"' % cppvar)
return
- if endif != ('#endif // %s' % cppvar):
- error_level = 0
- if endif != ('#endif // %s' % (cppvar + '_')):
- error_level = 5
+ # Didn't find the corresponding "//" comment. If this file does not
+ # contain any "//" comments at all, it could be that the compiler
+ # only wants "/**/" comments, look for those instead.
+ no_single_line_comments = True
+ for i in xrange(1, len(raw_lines) - 1):
+ line = raw_lines[i]
+ if Match(r'^(?:(?:\'(?:\.|[^\'])*\')|(?:"(?:\.|[^"])*")|[^\'"])*//', line):
+ no_single_line_comments = False
+ break
- ParseNolintSuppressions(filename, lines[endif_linenum], endif_linenum,
- error)
- error(filename, endif_linenum, 'build/header_guard', error_level,
- '#endif line should be "#endif // %s"' % cppvar)
+ if no_single_line_comments:
+ match = Match(r'#endif\s*/\*\s*' + cppvar + r'(_)?\s*\*/', endif)
+ if match:
+ if match.group(1) == '_':
+ # Low severity warning for double trailing underscore
+ error(filename, endif_linenum, 'build/header_guard', 0,
+ '#endif line should be "#endif /* %s */"' % cppvar)
+ return
+
+ # Didn't find anything
+ error(filename, endif_linenum, 'build/header_guard', 5,
+ '#endif line should be "#endif // %s"' % cppvar)
+
+
+def CheckHeaderFileIncluded(filename, include_state, error):
+ """Logs an error if a .cc file does not include its header."""
+
+ # Do not check test files
+ fileinfo = FileInfo(filename)
+ if Search(_TEST_FILE_SUFFIX, fileinfo.BaseName()):
+ return
+
+ headerfile = filename[0:len(filename) - len(fileinfo.Extension())] + '.h'
+ if not os.path.exists(headerfile):
+ return
+ headername = FileInfo(headerfile).RepositoryName()
+ first_include = 0
+ for section_list in include_state.include_list:
+ for f in section_list:
+ if headername in f[0] or f[0] in headername:
+ return
+ if not first_include:
+ first_include = f[1]
+
+ error(filename, first_include, 'build/include', 5,
+ '%s should include its header file %s' % (fileinfo.RepositoryName(),
+ headername))
def CheckForBadCharacters(filename, lines, error):
'Use C++11 raw strings or concatenation instead.')
-threading_list = (
- ('asctime(', 'asctime_r('),
- ('ctime(', 'ctime_r('),
- ('getgrgid(', 'getgrgid_r('),
- ('getgrnam(', 'getgrnam_r('),
- ('getlogin(', 'getlogin_r('),
- ('getpwnam(', 'getpwnam_r('),
- ('getpwuid(', 'getpwuid_r('),
- ('gmtime(', 'gmtime_r('),
- ('localtime(', 'localtime_r('),
- ('rand(', 'rand_r('),
- ('strtok(', 'strtok_r('),
- ('ttyname(', 'ttyname_r('),
+# (non-threadsafe name, thread-safe alternative, validation pattern)
+#
+# The validation pattern is used to eliminate false positives such as:
+# _rand(); // false positive due to substring match.
+# ->rand(); // some member function rand().
+# ACMRandom rand(seed); // some variable named rand.
+# ISAACRandom rand(); // another variable named rand.
+#
+# Basically we require the return value of these functions to be used
+# in some expression context on the same line by matching on some
+# operator before the function name. This eliminates constructors and
+# member function calls.
+_UNSAFE_FUNC_PREFIX = r'(?:[-+*/=%^&|(<]\s*|>\s+)'
+_THREADING_LIST = (
+ ('asctime(', 'asctime_r(', _UNSAFE_FUNC_PREFIX + r'asctime\([^)]+\)'),
+ ('ctime(', 'ctime_r(', _UNSAFE_FUNC_PREFIX + r'ctime\([^)]+\)'),
+ ('getgrgid(', 'getgrgid_r(', _UNSAFE_FUNC_PREFIX + r'getgrgid\([^)]+\)'),
+ ('getgrnam(', 'getgrnam_r(', _UNSAFE_FUNC_PREFIX + r'getgrnam\([^)]+\)'),
+ ('getlogin(', 'getlogin_r(', _UNSAFE_FUNC_PREFIX + r'getlogin\(\)'),
+ ('getpwnam(', 'getpwnam_r(', _UNSAFE_FUNC_PREFIX + r'getpwnam\([^)]+\)'),
+ ('getpwuid(', 'getpwuid_r(', _UNSAFE_FUNC_PREFIX + r'getpwuid\([^)]+\)'),
+ ('gmtime(', 'gmtime_r(', _UNSAFE_FUNC_PREFIX + r'gmtime\([^)]+\)'),
+ ('localtime(', 'localtime_r(', _UNSAFE_FUNC_PREFIX + r'localtime\([^)]+\)'),
+ ('rand(', 'rand_r(', _UNSAFE_FUNC_PREFIX + r'rand\(\)'),
+ ('strtok(', 'strtok_r(',
+ _UNSAFE_FUNC_PREFIX + r'strtok\([^)]+\)'),
+ ('ttyname(', 'ttyname_r(', _UNSAFE_FUNC_PREFIX + r'ttyname\([^)]+\)'),
)
error: The function to call with any errors found.
"""
line = clean_lines.elided[linenum]
- for single_thread_function, multithread_safe_function in threading_list:
- ix = line.find(single_thread_function)
- # Comparisons made explicit for clarity -- pylint: disable=g-explicit-bool-comparison
- if ix >= 0 and (ix == 0 or (not line[ix - 1].isalnum() and
- line[ix - 1] not in ('_', '.', '>'))):
+ for single_thread_func, multithread_safe_func, pattern in _THREADING_LIST:
+ # Additional pattern matching check to confirm that this is the
+ # function we are looking for
+ if Search(pattern, line):
error(filename, linenum, 'runtime/threadsafe_fn', 2,
- 'Consider using ' + multithread_safe_function +
- '...) instead of ' + single_thread_function +
+ 'Consider using ' + multithread_safe_func +
+ '...) instead of ' + single_thread_func +
'...) for improved thread safety.')
'VLOG() should be used with numeric verbosity level. '
'Use LOG() if you want symbolic severity levels.')
-
# Matches invalid increment: *count++, which moves pointer instead of
# incrementing a value.
_RE_PATTERN_INVALID_INCREMENT = re.compile(
'Changing pointer instead of value (or unused value of operator*).')
+def IsMacroDefinition(clean_lines, linenum):
+ if Search(r'^#define', clean_lines[linenum]):
+ return True
+
+ if linenum > 0 and Search(r'\\$', clean_lines[linenum - 1]):
+ return True
+
+ return False
+
+
+def IsForwardClassDeclaration(clean_lines, linenum):
+ return Match(r'^\s*(\btemplate\b)*.*class\s+\w+;\s*$', clean_lines[linenum])
+
+
class _BlockInfo(object):
"""Stores information about a generic block of code."""
- def __init__(self, seen_open_brace):
+ def __init__(self, linenum, seen_open_brace):
+ self.starting_linenum = linenum
self.seen_open_brace = seen_open_brace
self.open_parentheses = 0
self.inline_asm = _NO_ASM
+ self.check_namespace_indentation = False
def CheckBegin(self, filename, clean_lines, linenum, error):
"""Run checks that applies to text up to the opening brace.
"""
pass
+ def IsBlockInfo(self):
+ """Returns true if this block is a _BlockInfo.
+
+ This is convenient for verifying that an object is an instance of
+ a _BlockInfo, but not an instance of any of the derived classes.
+
+ Returns:
+ True for this class, False for derived classes.
+ """
+ return self.__class__ == _BlockInfo
+
+
+class _ExternCInfo(_BlockInfo):
+ """Stores information about an 'extern "C"' block."""
+
+ def __init__(self, linenum):
+ _BlockInfo.__init__(self, linenum, True)
+
class _ClassInfo(_BlockInfo):
"""Stores information about a class."""
def __init__(self, name, class_or_struct, clean_lines, linenum):
- _BlockInfo.__init__(self, False)
+ _BlockInfo.__init__(self, linenum, False)
self.name = name
- self.starting_linenum = linenum
self.is_derived = False
+ self.check_namespace_indentation = True
if class_or_struct == 'struct':
self.access = 'public'
self.is_struct = True
# Remember initial indentation level for this class. Using raw_lines here
# instead of elided to account for leading comments.
- initial_indent = Match(r'^( *)\S', clean_lines.raw_lines[linenum])
- if initial_indent:
- self.class_indent = len(initial_indent.group(1))
- else:
- self.class_indent = 0
+ self.class_indent = GetIndentLevel(clean_lines.raw_lines[linenum])
# Try to find the end of the class. This will be confused by things like:
# class A {
self.is_derived = True
def CheckEnd(self, filename, clean_lines, linenum, error):
+ # If there is a DISALLOW macro, it should appear near the end of
+ # the class.
+ seen_last_thing_in_class = False
+ for i in xrange(linenum - 1, self.starting_linenum, -1):
+ match = Search(
+ r'\b(DISALLOW_COPY_AND_ASSIGN|DISALLOW_IMPLICIT_CONSTRUCTORS)\(' +
+ self.name + r'\)',
+ clean_lines.elided[i])
+ if match:
+ if seen_last_thing_in_class:
+ error(filename, i, 'readability/constructors', 3,
+ match.group(1) + ' should be the last thing in the class')
+ break
+
+ if not Match(r'^\s*$', clean_lines.elided[i]):
+ seen_last_thing_in_class = True
+
# Check that closing brace is aligned with beginning of the class.
# Only do this if the closing brace is indented by only whitespaces.
# This means we will not check single-line class definitions.
"""Stores information about a namespace."""
def __init__(self, name, linenum):
- _BlockInfo.__init__(self, False)
+ _BlockInfo.__init__(self, linenum, False)
self.name = name or ''
- self.starting_linenum = linenum
+ self.check_namespace_indentation = True
def CheckEnd(self, filename, clean_lines, linenum, error):
"""Check end of namespace comments."""
# deciding what these nontrivial things are, so this check is
# triggered by namespace size only, which works most of the time.
if (linenum - self.starting_linenum < 10
- and not Match(r'};*\s*(//|/\*).*\bnamespace\b', line)):
+ and not Match(r'^\s*};*\s*(//|/\*).*\bnamespace\b', line)):
return
# Look for matching comment at end of namespace.
# expected namespace.
if self.name:
# Named namespace
- if not Match((r'};*\s*(//|/\*).*\bnamespace\s+' + re.escape(self.name) +
- r'[\*/\.\\\s]*$'),
+ if not Match((r'^\s*};*\s*(//|/\*).*\bnamespace\s+' +
+ re.escape(self.name) + r'[\*/\.\\\s]*$'),
line):
error(filename, linenum, 'readability/namespace', 5,
'Namespace should be terminated with "// namespace %s"' %
self.name)
else:
# Anonymous namespace
- if not Match(r'};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line):
- error(filename, linenum, 'readability/namespace', 5,
- 'Namespace should be terminated with "// namespace"')
+ if not Match(r'^\s*};*\s*(//|/\*).*\bnamespace[\*/\.\\\s]*$', line):
+ # If "// namespace anonymous" or "// anonymous namespace (more text)",
+ # mention "// anonymous namespace" as an acceptable form
+ if Match(r'^\s*}.*\b(namespace anonymous|anonymous namespace)\b', line):
+ error(filename, linenum, 'readability/namespace', 5,
+ 'Anonymous namespace should be terminated with "// namespace"'
+ ' or "// anonymous namespace"')
+ else:
+ error(filename, linenum, 'readability/namespace', 5,
+ 'Anonymous namespace should be terminated with "// namespace"')
class _PreprocessorInfo(object):
self.seen_else = False
-class _NestingState(object):
+class NestingState(object):
"""Holds states related to parsing braces."""
def __init__(self):
# - _BlockInfo: some other type of block.
self.stack = []
+ # Top of the previous stack before each Update().
+ #
+ # Because the nesting_stack is updated at the end of each line, we
+ # had to do some convoluted checks to find out what is the current
+ # scope at the beginning of the line. This check is simplified by
+ # saving the previous top of nesting stack.
+ #
+ # We could save the full stack, but we only need the top. Copying
+ # the full nesting stack would slow down cpplint by ~10%.
+ self.previous_stack_top = []
+
# Stack of _PreprocessorInfo objects.
self.pp_stack = []
"""
return self.stack and isinstance(self.stack[-1], _NamespaceInfo)
+ def InExternC(self):
+ """Check if we are currently one level inside an 'extern "C"' block.
+
+ Returns:
+ True if top of the stack is an extern block, False otherwise.
+ """
+ return self.stack and isinstance(self.stack[-1], _ExternCInfo)
+
+ def InClassDeclaration(self):
+ """Check if we are currently one level inside a class or struct declaration.
+
+ Returns:
+ True if top of the stack is a class/struct, False otherwise.
+ """
+ return self.stack and isinstance(self.stack[-1], _ClassInfo)
+
+ def InAsmBlock(self):
+ """Check if we are currently one level inside an inline ASM block.
+
+ Returns:
+ True if the top of the stack is a block containing inline ASM.
+ """
+ return self.stack and self.stack[-1].inline_asm != _NO_ASM
+
+ def InTemplateArgumentList(self, clean_lines, linenum, pos):
+ """Check if current position is inside template argument list.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ pos: position just after the suspected template argument.
+ Returns:
+ True if (linenum, pos) is inside template arguments.
+ """
+ while linenum < clean_lines.NumLines():
+ # Find the earliest character that might indicate a template argument
+ line = clean_lines.elided[linenum]
+ match = Match(r'^[^{};=\[\]\.<>]*(.)', line[pos:])
+ if not match:
+ linenum += 1
+ pos = 0
+ continue
+ token = match.group(1)
+ pos += len(match.group(0))
+
+ # These things do not look like template argument list:
+ # class Suspect {
+ # class Suspect x; }
+ if token in ('{', '}', ';'): return False
+
+ # These things look like template argument list:
+ # template <class Suspect>
+ # template <class Suspect = default_value>
+ # template <class Suspect[]>
+ # template <class Suspect...>
+ if token in ('>', '=', '[', ']', '.'): return True
+
+ # Check if token is an unmatched '<'.
+ # If not, move on to the next character.
+ if token != '<':
+ pos += 1
+ if pos >= len(line):
+ linenum += 1
+ pos = 0
+ continue
+
+ # We can't be sure if we just find a single '<', and need to
+ # find the matching '>'.
+ (_, end_line, end_pos) = CloseExpression(clean_lines, linenum, pos - 1)
+ if end_pos < 0:
+ # Not sure if template argument list or syntax error in file
+ return False
+ linenum = end_line
+ pos = end_pos
+ return False
+
def UpdatePreprocessor(self, line):
"""Update preprocessor stack.
# TODO(unknown): unexpected #endif, issue warning?
pass
+ # TODO(unknown): Update() is too long, but we will refactor later.
def Update(self, filename, clean_lines, linenum, error):
"""Update nesting state with current line.
"""
line = clean_lines.elided[linenum]
- # Update pp_stack first
+ # Remember top of the previous nesting stack.
+ #
+ # The stack is always pushed/popped and not modified in place, so
+ # we can just do a shallow copy instead of copy.deepcopy. Using
+ # deepcopy would slow down cpplint by ~28%.
+ if self.stack:
+ self.previous_stack_top = self.stack[-1]
+ else:
+ self.previous_stack_top = None
+
+ # Update pp_stack
self.UpdatePreprocessor(line)
# Count parentheses. This is to avoid adding struct arguments to
# such as in:
# class LOCKABLE API Object {
# };
- #
- # Templates with class arguments may confuse the parser, for example:
- # template <class T
- # class Comparator = less<T>,
- # class Vector = vector<T> >
- # class HeapQueue {
- #
- # Because this parser has no nesting state about templates, by the
- # time it saw "class Comparator", it may think that it's a new class.
- # Nested templates have a similar problem:
- # template <
- # typename ExportedType,
- # typename TupleType,
- # template <typename, typename> class ImplTemplate>
- #
- # To avoid these cases, we ignore classes that are followed by '=' or '>'
class_decl_match = Match(
- r'\s*(template\s*<[\w\s<>,:]*>\s*)?'
- r'(class|struct)\s+([A-Z_]+\s+)*(\w+(?:::\w+)*)'
- r'(([^=>]|<[^<>]*>|<[^<>]*<[^<>]*>\s*>)*)$', line)
+ r'^(\s*(?:template\s*<[\w\s<>,:]*>\s*)?'
+ r'(class|struct)\s+(?:[A-Z_]+\s+)*(\w+(?:::\w+)*))'
+ r'(.*)$', line)
if (class_decl_match and
(not self.stack or self.stack[-1].open_parentheses == 0)):
- self.stack.append(_ClassInfo(
- class_decl_match.group(4), class_decl_match.group(2),
- clean_lines, linenum))
- line = class_decl_match.group(5)
+ # We do not want to accept classes that are actually template arguments:
+ # template <class Ignore1,
+ # class Ignore2 = Default<Args>,
+ # template <Args> class Ignore3>
+ # void Function() {};
+ #
+ # To avoid template argument cases, we scan forward and look for
+ # an unmatched '>'. If we see one, assume we are inside a
+ # template argument list.
+ end_declaration = len(class_decl_match.group(1))
+ if not self.InTemplateArgumentList(clean_lines, linenum, end_declaration):
+ self.stack.append(_ClassInfo(
+ class_decl_match.group(3), class_decl_match.group(2),
+ clean_lines, linenum))
+ line = class_decl_match.group(4)
# If we have not yet seen the opening brace for the innermost block,
# run checks here.
# stack otherwise.
if not self.SeenOpenBrace():
self.stack[-1].seen_open_brace = True
+ elif Match(r'^extern\s*"[^"]*"\s*\{', line):
+ self.stack.append(_ExternCInfo(linenum))
else:
- self.stack.append(_BlockInfo(True))
+ self.stack.append(_BlockInfo(linenum, True))
if _MATCH_ASM.match(line):
self.stack[-1].inline_asm = _BLOCK_ASM
+
elif token == ';' or token == ')':
# If we haven't seen an opening brace yet, but we already saw
# a semicolon, this is probably a forward declaration. Pop
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: A callable to which errors are reported, which takes 4 arguments:
filename, line number, error level, and message
r'\s+(register|static|extern|typedef)\b',
line):
error(filename, linenum, 'build/storage_class', 5,
- 'Storage class (static, extern, typedef, etc) should be first.')
+ 'Storage-class specifier (static, extern, typedef, etc) should be '
+ 'at the beginning of the declaration.')
if Match(r'\s*#\s*endif\s*[^/\s]+', line):
error(filename, linenum, 'build/endif_comment', 5,
# Look for single-argument constructors that aren't marked explicit.
# Technically a valid construct, but against style.
- args = Match(r'\s+(?:inline\s+)?%s\s*\(([^,()]+)\)'
- % re.escape(base_classname),
- line)
- if (args and
- args.group(1) != 'void' and
- not Match(r'(const\s+)?%s(\s+const)?\s*(?:<\w+>\s*)?&'
- % re.escape(base_classname), args.group(1).strip())):
- error(filename, linenum, 'runtime/explicit', 5,
- 'Single-argument constructors should be marked explicit.')
-
-
-def CheckSpacingForFunctionCall(filename, line, linenum, error):
+ explicit_constructor_match = Match(
+ r'\s+(?:(?:inline|constexpr)\s+)*(explicit\s+)?'
+ r'(?:(?:inline|constexpr)\s+)*%s\s*'
+ r'\(((?:[^()]|\([^()]*\))*)\)'
+ % re.escape(base_classname),
+ line)
+
+ if explicit_constructor_match:
+ is_marked_explicit = explicit_constructor_match.group(1)
+
+ if not explicit_constructor_match.group(2):
+ constructor_args = []
+ else:
+ constructor_args = explicit_constructor_match.group(2).split(',')
+
+ # collapse arguments so that commas in template parameter lists and function
+ # argument parameter lists don't split arguments in two
+ i = 0
+ while i < len(constructor_args):
+ constructor_arg = constructor_args[i]
+ while (constructor_arg.count('<') > constructor_arg.count('>') or
+ constructor_arg.count('(') > constructor_arg.count(')')):
+ constructor_arg += ',' + constructor_args[i + 1]
+ del constructor_args[i + 1]
+ constructor_args[i] = constructor_arg
+ i += 1
+
+ defaulted_args = [arg for arg in constructor_args if '=' in arg]
+ noarg_constructor = (not constructor_args or # empty arg list
+ # 'void' arg specifier
+ (len(constructor_args) == 1 and
+ constructor_args[0].strip() == 'void'))
+ onearg_constructor = ((len(constructor_args) == 1 and # exactly one arg
+ not noarg_constructor) or
+ # all but at most one arg defaulted
+ (len(constructor_args) >= 1 and
+ not noarg_constructor and
+ len(defaulted_args) >= len(constructor_args) - 1))
+ initializer_list_constructor = bool(
+ onearg_constructor and
+ Search(r'\bstd\s*::\s*initializer_list\b', constructor_args[0]))
+ copy_constructor = bool(
+ onearg_constructor and
+ Match(r'(const\s+)?%s(\s*<[^>]*>)?(\s+const)?\s*(?:<\w+>\s*)?&'
+ % re.escape(base_classname), constructor_args[0].strip()))
+
+ if (not is_marked_explicit and
+ onearg_constructor and
+ not initializer_list_constructor and
+ not copy_constructor):
+ if defaulted_args:
+ error(filename, linenum, 'runtime/explicit', 5,
+ 'Constructors callable with one argument '
+ 'should be marked explicit.')
+ else:
+ error(filename, linenum, 'runtime/explicit', 5,
+ 'Single-parameter constructors should be marked explicit.')
+ elif is_marked_explicit and not onearg_constructor:
+ if noarg_constructor:
+ error(filename, linenum, 'runtime/explicit', 5,
+ 'Zero-parameter constructors should not be marked explicit.')
+
+
+def CheckSpacingForFunctionCall(filename, clean_lines, linenum, error):
"""Checks for the correctness of various spacing around function calls.
Args:
filename: The name of the current file.
- line: The text of the line to check.
+ clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
error: The function to call with any errors found.
"""
+ line = clean_lines.elided[linenum]
# Since function calls often occur inside if/for/while/switch
# expressions - which have their own, more liberal conventions - we
error(filename, linenum, 'whitespace/parens', 2,
'Extra space after (')
if (Search(r'\w\s+\(', fncall) and
- not Search(r'#\s*define|typedef', fncall) and
- not Search(r'\w\s+\((\w+::)*\*\w+\)\(', fncall)):
- error(filename, linenum, 'whitespace/parens', 4,
- 'Extra space before ( in function call')
+ not Search(r'_{0,2}asm_{0,2}\s+_{0,2}volatile_{0,2}\s+\(', fncall) and
+ not Search(r'#\s*define|typedef|using\s+\w+\s*=', fncall) and
+ not Search(r'\w\s+\((\w+::)*\*\w+\)\(', fncall) and
+ not Search(r'\bcase\s+\(', fncall)):
+ # TODO(unknown): Space after an operator function seem to be a common
+ # error, silence those for now by restricting them to highest verbosity.
+ if Search(r'\boperator_*\b', line):
+ error(filename, linenum, 'whitespace/parens', 0,
+ 'Extra space before ( in function call')
+ else:
+ error(filename, linenum, 'whitespace/parens', 4,
+ 'Extra space before ( in function call')
# If the ) is followed only by a newline or a { + newline, assume it's
# part of a control statement (if/while/etc), and don't complain
if Search(r'[^)]\s+\)\s*[^{\s]', fncall):
return not line or line.isspace()
+def CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line,
+ error):
+ is_namespace_indent_item = (
+ len(nesting_state.stack) > 1 and
+ nesting_state.stack[-1].check_namespace_indentation and
+ isinstance(nesting_state.previous_stack_top, _NamespaceInfo) and
+ nesting_state.previous_stack_top == nesting_state.stack[-2])
+
+ if ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item,
+ clean_lines.elided, line):
+ CheckItemIndentationInNamespace(filename, clean_lines.elided,
+ line, error)
+
+
def CheckForFunctionLengths(filename, clean_lines, linenum,
function_state, error):
"""Reports for long function bodies.
For an overview why this is done, see:
- http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions
+ https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Write_Short_Functions
Uses a simplistic algorithm assuming other style guidelines
(especially spacing) are followed.
"""
lines = clean_lines.lines
line = lines[linenum]
- raw = clean_lines.raw_lines
- raw_line = raw[linenum]
joined_line = ''
starting_func = False
_RE_PATTERN_TODO = re.compile(r'^//(\s*)TODO(\(.+?\))?:?(\s|$)?')
-def CheckComment(comment, filename, linenum, error):
- """Checks for common mistakes in TODO comments.
+def CheckComment(line, filename, linenum, next_line_start, error):
+ """Checks for common mistakes in comments.
Args:
- comment: The text of the comment from the line in question.
+ line: The line in question.
filename: The name of the current file.
linenum: The number of the line to check.
+ next_line_start: The first non-whitespace column of the next line.
error: The function to call with any errors found.
"""
- match = _RE_PATTERN_TODO.match(comment)
- if match:
- # One whitespace is correct; zero whitespace is handled elsewhere.
- leading_whitespace = match.group(1)
- if len(leading_whitespace) > 1:
- error(filename, linenum, 'whitespace/todo', 2,
- 'Too many spaces before TODO')
-
- username = match.group(2)
- if not username:
- error(filename, linenum, 'readability/todo', 2,
- 'Missing username in TODO; it should look like '
- '"// TODO(my_username): Stuff."')
-
- middle_whitespace = match.group(3)
- # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison
- if middle_whitespace != ' ' and middle_whitespace != '':
- error(filename, linenum, 'whitespace/todo', 2,
- 'TODO(my_username) should be followed by a space')
-
-def CheckAccess(filename, clean_lines, linenum, nesting_state, error):
- """Checks for improper use of DISALLOW* macros.
+ commentpos = line.find('//')
+ if commentpos != -1:
+ # Check if the // may be in quotes. If so, ignore it
+ if re.sub(r'\\.', '', line[0:commentpos]).count('"') % 2 == 0:
+ # Allow one space for new scopes, two spaces otherwise:
+ if (not (Match(r'^.*{ *//', line) and next_line_start == commentpos) and
+ ((commentpos >= 1 and
+ line[commentpos-1] not in string.whitespace) or
+ (commentpos >= 2 and
+ line[commentpos-2] not in string.whitespace))):
+ error(filename, linenum, 'whitespace/comments', 2,
+ 'At least two spaces is best between code and comments')
- Args:
- filename: The name of the current file.
- clean_lines: A CleansedLines instance containing the file.
- linenum: The number of the line to check.
- nesting_state: A _NestingState instance which maintains information about
- the current stack of nested blocks being parsed.
- error: The function to call with any errors found.
- """
- line = clean_lines.elided[linenum] # get rid of comments and strings
-
- matched = Match((r'\s*(DISALLOW_COPY_AND_ASSIGN|'
- r'DISALLOW_EVIL_CONSTRUCTORS|'
- r'DISALLOW_IMPLICIT_CONSTRUCTORS)'), line)
- if not matched:
- return
- if nesting_state.stack and isinstance(nesting_state.stack[-1], _ClassInfo):
- if nesting_state.stack[-1].access != 'private':
- error(filename, linenum, 'readability/constructors', 3,
- '%s must be in the private: section' % matched.group(1))
-
- else:
- # Found DISALLOW* macro outside a class declaration, or perhaps it
- # was used inside a function when it should have been part of the
- # class declaration. We could issue a warning here, but it
- # probably resulted in a compiler error already.
- pass
-
-
-def FindNextMatchingAngleBracket(clean_lines, linenum, init_suffix):
- """Find the corresponding > to close a template.
-
- Args:
- clean_lines: A CleansedLines instance containing the file.
- linenum: Current line number.
- init_suffix: Remainder of the current line after the initial <.
-
- Returns:
- True if a matching bracket exists.
- """
- line = init_suffix
- nesting_stack = ['<']
- while True:
- # Find the next operator that can tell us whether < is used as an
- # opening bracket or as a less-than operator. We only want to
- # warn on the latter case.
- #
- # We could also check all other operators and terminate the search
- # early, e.g. if we got something like this "a<b+c", the "<" is
- # most likely a less-than operator, but then we will get false
- # positives for default arguments and other template expressions.
- match = Search(r'^[^<>(),;\[\]]*([<>(),;\[\]])(.*)$', line)
- if match:
- # Found an operator, update nesting stack
- operator = match.group(1)
- line = match.group(2)
-
- if nesting_stack[-1] == '<':
- # Expecting closing angle bracket
- if operator in ('<', '(', '['):
- nesting_stack.append(operator)
- elif operator == '>':
- nesting_stack.pop()
- if not nesting_stack:
- # Found matching angle bracket
- return True
- elif operator == ',':
- # Got a comma after a bracket, this is most likely a template
- # argument. We have not seen a closing angle bracket yet, but
- # it's probably a few lines later if we look for it, so just
- # return early here.
- return True
- else:
- # Got some other operator.
- return False
-
- else:
- # Expecting closing parenthesis or closing bracket
- if operator in ('<', '(', '['):
- nesting_stack.append(operator)
- elif operator in (')', ']'):
- # We don't bother checking for matching () or []. If we got
- # something like (] or [), it would have been a syntax error.
- nesting_stack.pop()
-
- else:
- # Scan the next line
- linenum += 1
- if linenum >= len(clean_lines.elided):
- break
- line = clean_lines.elided[linenum]
-
- # Exhausted all remaining lines and still no matching angle bracket.
- # Most likely the input was incomplete, otherwise we should have
- # seen a semicolon and returned early.
- return True
-
-
-def FindPreviousMatchingAngleBracket(clean_lines, linenum, init_prefix):
- """Find the corresponding < that started a template.
-
- Args:
- clean_lines: A CleansedLines instance containing the file.
- linenum: Current line number.
- init_prefix: Part of the current line before the initial >.
-
- Returns:
- True if a matching bracket exists.
- """
- line = init_prefix
- nesting_stack = ['>']
- while True:
- # Find the previous operator
- match = Search(r'^(.*)([<>(),;\[\]])[^<>(),;\[\]]*$', line)
- if match:
- # Found an operator, update nesting stack
- operator = match.group(2)
- line = match.group(1)
-
- if nesting_stack[-1] == '>':
- # Expecting opening angle bracket
- if operator in ('>', ')', ']'):
- nesting_stack.append(operator)
- elif operator == '<':
- nesting_stack.pop()
- if not nesting_stack:
- # Found matching angle bracket
- return True
- elif operator == ',':
- # Got a comma before a bracket, this is most likely a
- # template argument. The opening angle bracket is probably
- # there if we look for it, so just return early here.
- return True
- else:
- # Got some other operator.
- return False
-
- else:
- # Expecting opening parenthesis or opening bracket
- if operator in ('>', ')', ']'):
- nesting_stack.append(operator)
- elif operator in ('(', '['):
- nesting_stack.pop()
-
- else:
- # Scan the previous line
- linenum -= 1
- if linenum < 0:
- break
- line = clean_lines.elided[linenum]
-
- # Exhausted all earlier lines and still no matching angle bracket.
- return False
+ # Checks for common mistakes in TODO comments.
+ comment = line[commentpos:]
+ match = _RE_PATTERN_TODO.match(comment)
+ if match:
+ # One whitespace is correct; zero whitespace is handled elsewhere.
+ leading_whitespace = match.group(1)
+ if len(leading_whitespace) > 1:
+ error(filename, linenum, 'whitespace/todo', 2,
+ 'Too many spaces before TODO')
+
+ username = match.group(2)
+ if not username:
+ error(filename, linenum, 'readability/todo', 2,
+ 'Missing username in TODO; it should look like '
+ '"// TODO(my_username): Stuff."')
+
+ middle_whitespace = match.group(3)
+ # Comparisons made explicit for correctness -- pylint: disable=g-explicit-bool-comparison
+ if middle_whitespace != ' ' and middle_whitespace != '':
+ error(filename, linenum, 'whitespace/todo', 2,
+ 'TODO(my_username) should be followed by a space')
+
+ # If the comment contains an alphanumeric character, there
+ # should be a space somewhere between it and the // unless
+ # it's a /// or //! Doxygen comment.
+ if (Match(r'//[^ ]*\w', comment) and
+ not Match(r'(///|//\!)(\s+|$)', comment)):
+ error(filename, linenum, 'whitespace/comments', 4,
+ 'Should have a space between // and comment')
def CheckSpacing(filename, clean_lines, linenum, nesting_state, error):
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: The function to call with any errors found.
"""
# }
#
# A warning about missing end of namespace comments will be issued instead.
- if IsBlankLine(line) and not nesting_state.InNamespaceBody():
+ #
+ # Also skip blank line checks for 'extern "C"' blocks, which are formatted
+ # like namespaces.
+ if (IsBlankLine(line) and
+ not nesting_state.InNamespaceBody() and
+ not nesting_state.InExternC()):
elided = clean_lines.elided
prev_line = elided[linenum - 1]
prevbrace = prev_line.rfind('{')
error(filename, linenum, 'whitespace/blank_line', 3,
'Do not leave a blank line after "%s:"' % matched.group(1))
- # Next, we complain if there's a comment too near the text
- commentpos = line.find('//')
- if commentpos != -1:
- # Check if the // may be in quotes. If so, ignore it
- # Comparisons made explicit for clarity -- pylint: disable=g-explicit-bool-comparison
- if (line.count('"', 0, commentpos) -
- line.count('\\"', 0, commentpos)) % 2 == 0: # not in quotes
- # Allow one space for new scopes, two spaces otherwise:
- if (not Match(r'^\s*{ //', line) and
- ((commentpos >= 1 and
- line[commentpos-1] not in string.whitespace) or
- (commentpos >= 2 and
- line[commentpos-2] not in string.whitespace))):
- error(filename, linenum, 'whitespace/comments', 2,
- 'At least two spaces is best between code and comments')
- # There should always be a space between the // and the comment
- commentend = commentpos + 2
- if commentend < len(line) and not line[commentend] == ' ':
- # but some lines are exceptions -- e.g. if they're big
- # comment delimiters like:
- # //----------------------------------------------------------
- # or are an empty C++ style Doxygen comment, like:
- # ///
- # or C++ style Doxygen comments placed after the variable:
- # ///< Header comment
- # //!< Header comment
- # or they begin with multiple slashes followed by a space:
- # //////// Header comment
- match = (Search(r'[=/-]{4,}\s*$', line[commentend:]) or
- Search(r'^/$', line[commentend:]) or
- Search(r'^!< ', line[commentend:]) or
- Search(r'^/< ', line[commentend:]) or
- Search(r'^/+ ', line[commentend:]))
- if not match:
- error(filename, linenum, 'whitespace/comments', 4,
- 'Should have a space between // and comment')
- CheckComment(line[commentpos:], filename, linenum, error)
-
- line = clean_lines.elided[linenum] # get rid of comments and strings
-
- # Don't try to do spacing checks for operator methods
- line = re.sub(r'operator(==|!=|<|<<|<=|>=|>>|>)\(', 'operator\(', line)
+ # Next, check comments
+ next_line_start = 0
+ if linenum + 1 < clean_lines.NumLines():
+ next_line = raw[linenum + 1]
+ next_line_start = len(next_line) - len(next_line.lstrip())
+ CheckComment(line, filename, linenum, next_line_start, error)
+
+ # get rid of comments and strings
+ line = clean_lines.elided[linenum]
+
+ # You shouldn't have spaces before your brackets, except maybe after
+ # 'delete []', 'return []() {};', or 'auto [abc, ...] = ...;'.
+ if Search(r'\w\s+\[', line) and not Search(r'(?:auto&?|delete|return)\s+\[', line):
+ error(filename, linenum, 'whitespace/braces', 5,
+ 'Extra space before [')
+
+ # In range-based for, we wanted spaces before and after the colon, but
+ # not around "::" tokens that might appear.
+ if (Search(r'for *\(.*[^:]:[^: ]', line) or
+ Search(r'for *\(.*[^: ]:[^:]', line)):
+ error(filename, linenum, 'whitespace/forcolon', 2,
+ 'Missing space around colon in range-based for loop')
+
+
+def CheckOperatorSpacing(filename, clean_lines, linenum, error):
+ """Checks for horizontal spacing around operators.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ # Don't try to do spacing checks for operator methods. Do this by
+ # replacing the troublesome characters with something else,
+ # preserving column position for all other characters.
+ #
+ # The replacement is done repeatedly to avoid false positives from
+ # operators that call operators.
+ while True:
+ match = Match(r'^(.*\boperator\b)(\S+)(\s*\(.*)$', line)
+ if match:
+ line = match.group(1) + ('_' * len(match.group(2))) + match.group(3)
+ else:
+ break
# We allow no-spaces around = within an if: "if ( (a=Foo()) == 0 )".
# Otherwise not. Note we only check for non-spaces on *both* sides;
# sometimes people put non-spaces on one side when aligning ='s among
# many lines (not that this is behavior that I approve of...)
- if Search(r'[\w.]=[\w.]', line) and not Search(r'\b(if|while) ', line):
+ if ((Search(r'[\w.]=', line) or
+ Search(r'=[\w.]', line))
+ and not Search(r'\b(if|while|for) ', line)
+ # Operators taken from [lex.operators] in C++11 standard.
+ and not Search(r'(>=|<=|==|!=|&=|\^=|\|=|\+=|\*=|\/=|\%=)', line)
+ and not Search(r'operator=', line)):
error(filename, linenum, 'whitespace/operators', 4,
'Missing spaces around =')
#
# Check <= and >= first to avoid false positives with < and >, then
# check non-include lines for spacing around < and >.
- match = Search(r'[^<>=!\s](==|!=|<=|>=)[^<>=!\s]', line)
+ #
+ # If the operator is followed by a comma, assume it's be used in a
+ # macro context and don't do any checks. This avoids false
+ # positives.
+ #
+ # Note that && is not included here. This is because there are too
+ # many false positives due to RValue references.
+ match = Search(r'[^<>=!\s](==|!=|<=|>=|\|\|)[^<>=!\s,;\)]', line)
if match:
error(filename, linenum, 'whitespace/operators', 3,
'Missing spaces around %s' % match.group(1))
- # We allow no-spaces around << when used like this: 10<<20, but
- # not otherwise (particularly, not when used as streams)
- # Also ignore using ns::operator<<;
- match = Search(r'(operator|\S)(?:L|UL|ULL|l|ul|ull)?<<(\S)', line)
- if (match and
- not (match.group(1).isdigit() and match.group(2).isdigit()) and
- not (match.group(1) == 'operator' and match.group(2) == ';')):
- error(filename, linenum, 'whitespace/operators', 3,
- 'Missing spaces around <<')
elif not Match(r'#.*include', line):
- # Avoid false positives on ->
- reduced_line = line.replace('->', '')
-
# Look for < that is not surrounded by spaces. This is only
# triggered if both sides are missing spaces, even though
# technically should should flag if at least one side is missing a
# space. This is done to avoid some false positives with shifts.
- match = Search(r'[^\s<]<([^\s=<].*)', reduced_line)
- if (match and
- not FindNextMatchingAngleBracket(clean_lines, linenum, match.group(1))):
- error(filename, linenum, 'whitespace/operators', 3,
- 'Missing spaces around <')
+ match = Match(r'^(.*[^\s<])<[^\s=<,]', line)
+ if match:
+ (_, _, end_pos) = CloseExpression(
+ clean_lines, linenum, len(match.group(1)))
+ if end_pos <= -1:
+ error(filename, linenum, 'whitespace/operators', 3,
+ 'Missing spaces around <')
# Look for > that is not surrounded by spaces. Similar to the
# above, we only trigger if both sides are missing spaces to avoid
# false positives with shifts.
- match = Search(r'^(.*[^\s>])>[^\s=>]', reduced_line)
- if (match and
- not FindPreviousMatchingAngleBracket(clean_lines, linenum,
- match.group(1))):
- error(filename, linenum, 'whitespace/operators', 3,
- 'Missing spaces around >')
+ match = Match(r'^(.*[^-\s>])>[^\s=>,]', line)
+ if match:
+ (_, _, start_pos) = ReverseCloseExpression(
+ clean_lines, linenum, len(match.group(1)))
+ if start_pos <= -1:
+ error(filename, linenum, 'whitespace/operators', 3,
+ 'Missing spaces around >')
+
+ # We allow no-spaces around << when used like this: 10<<20, but
+ # not otherwise (particularly, not when used as streams)
+ #
+ # We also allow operators following an opening parenthesis, since
+ # those tend to be macros that deal with operators.
+ match = Search(r'(operator|[^\s(<])(?:L|UL|LL|ULL|l|ul|ll|ull)?<<([^\s,=<])', line)
+ if (match and not (match.group(1).isdigit() and match.group(2).isdigit()) and
+ not (match.group(1) == 'operator' and match.group(2) == ';')):
+ error(filename, linenum, 'whitespace/operators', 3,
+ 'Missing spaces around <<')
# We allow no-spaces around >> for almost anything. This is because
# C++11 allows ">>" to close nested templates, which accounts for
error(filename, linenum, 'whitespace/operators', 4,
'Extra space for operator %s' % match.group(1))
- # A pet peeve of mine: no spaces after an if, while, switch, or for
+
+def CheckParenthesisSpacing(filename, clean_lines, linenum, error):
+ """Checks for horizontal spacing around parentheses.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ # No spaces after an if, while, switch, or for
match = Search(r' (if\(|for\(|while\(|switch\()', line)
if match:
error(filename, linenum, 'whitespace/parens', 5,
'Should have zero or one spaces inside ( and ) in %s' %
match.group(1))
+
+def CheckCommaSpacing(filename, clean_lines, linenum, error):
+ """Checks for horizontal spacing near commas and semicolons.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ raw = clean_lines.lines_without_raw_strings
+ line = clean_lines.elided[linenum]
+
# You should always have a space after a comma (either as fn arg or operator)
#
# This does not apply when the non-space character following the
# verify that lines contain missing whitespaces, second pass on raw
# lines to confirm that those missing whitespaces are not due to
# elided comments.
- if Search(r',[^,\s]', line) and Search(r',[^,\s]', raw[linenum]):
+ if (Search(r',[^,\s]', ReplaceAll(r'\boperator\s*,\s*\(', 'F(', line)) and
+ Search(r',[^,\s]', raw[linenum])):
error(filename, linenum, 'whitespace/comma', 3,
'Missing space after ,')
error(filename, linenum, 'whitespace/semicolon', 3,
'Missing space after ;')
- # Next we will look for issues with function calls.
- CheckSpacingForFunctionCall(filename, line, linenum, error)
+
+def _IsType(clean_lines, nesting_state, expr):
+ """Check if expression looks like a type name, returns true if so.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ nesting_state: A NestingState instance which maintains information about
+ the current stack of nested blocks being parsed.
+ expr: The expression to check.
+ Returns:
+ True, if token looks like a type.
+ """
+ # Keep only the last token in the expression
+ last_word = Match(r'^.*(\b\S+)$', expr)
+ if last_word:
+ token = last_word.group(1)
+ else:
+ token = expr
+
+ # Match native types and stdint types
+ if _TYPES.match(token):
+ return True
+
+ # Try a bit harder to match templated types. Walk up the nesting
+ # stack until we find something that resembles a typename
+ # declaration for what we are looking for.
+ typename_pattern = (r'\b(?:typename|class|struct)\s+' + re.escape(token) +
+ r'\b')
+ block_index = len(nesting_state.stack) - 1
+ while block_index >= 0:
+ if isinstance(nesting_state.stack[block_index], _NamespaceInfo):
+ return False
+
+ # Found where the opening brace is. We want to scan from this
+ # line up to the beginning of the function, minus a few lines.
+ # template <typename Type1, // stop scanning here
+ # ...>
+ # class C
+ # : public ... { // start scanning here
+ last_line = nesting_state.stack[block_index].starting_linenum
+
+ next_block_start = 0
+ if block_index > 0:
+ next_block_start = nesting_state.stack[block_index - 1].starting_linenum
+ first_line = last_line
+ while first_line >= next_block_start:
+ if clean_lines.elided[first_line].find('template') >= 0:
+ break
+ first_line -= 1
+ if first_line < next_block_start:
+ # Didn't find any "template" keyword before reaching the next block,
+ # there are probably no template things to check for this block
+ block_index -= 1
+ continue
+
+ # Look for typename in the specified range
+ for i in xrange(first_line, last_line + 1, 1):
+ if Search(typename_pattern, clean_lines.elided[i]):
+ return True
+ block_index -= 1
+
+ return False
+
+
+def CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error):
+ """Checks for horizontal spacing near commas.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ nesting_state: A NestingState instance which maintains information about
+ the current stack of nested blocks being parsed.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
# Except after an opening paren, or after another opening brace (in case of
# an initializer list, for instance), you should have spaces before your
- # braces. And since you should never have braces at the beginning of a line,
- # this is an easy test.
- match = Match(r'^(.*[^ ({]){', line)
+ # braces when they are delimiting blocks, classes, namespaces etc.
+ # And since you should never have braces at the beginning of a line,
+ # this is an easy test. Except that braces used for initialization don't
+ # follow the same rule; we often don't want spaces before those.
+ match = Match(r'^(.*[^ ({>]){', line)
+
if match:
# Try a bit harder to check for brace initialization. This
# happens in one of the following forms:
# LastArgument(..., type{});
# LOG(INFO) << type{} << " ...";
# map_of_type[{...}] = ...;
+ # ternary = expr ? new type{} : nullptr;
+ # OuterTemplate<InnerTemplateConstructor<Type>{}>
#
# We check for the character following the closing brace, and
# silence the warning if it's one of those listed above, i.e.
- # "{.;,)<]".
+ # "{.;,)<>]:".
#
# To account for nested initializer list, we allow any number of
# closing braces up to "{;,)<". We can't simply silence the
# There is a false negative with this approach if people inserted
# spurious semicolons, e.g. "if (cond){};", but we will catch the
# spurious semicolon with a separate check.
+ leading_text = match.group(1)
(endline, endlinenum, endpos) = CloseExpression(
clean_lines, linenum, len(match.group(1)))
trailing_text = ''
for offset in xrange(endlinenum + 1,
min(endlinenum + 3, clean_lines.NumLines() - 1)):
trailing_text += clean_lines.elided[offset]
- if not Match(r'^[\s}]*[{.;,)<\]]', trailing_text):
+ # We also suppress warnings for `uint64_t{expression}` etc., as the style
+ # guide recommends brace initialization for integral types to avoid
+ # overflow/truncation.
+ if (not Match(r'^[\s}]*[{.;,)<>\]:]', trailing_text)
+ and not _IsType(clean_lines, nesting_state, leading_text)):
error(filename, linenum, 'whitespace/braces', 5,
'Missing space before {')
error(filename, linenum, 'whitespace/braces', 5,
'Missing space before else')
- # You shouldn't have spaces before your brackets, except maybe after
- # 'delete []' or 'new char * []'.
- if Search(r'\w\s+\[', line) and not Search(r'delete\s+\[', line):
- error(filename, linenum, 'whitespace/braces', 5,
- 'Extra space before [')
-
# You shouldn't have a space before a semicolon at the end of the line.
# There's a special case for "for" since the style guide allows space before
# the semicolon there.
'Extra space before last semicolon. If this should be an empty '
'statement, use {} instead.')
- # In range-based for, we wanted spaces before and after the colon, but
- # not around "::" tokens that might appear.
- if (Search('for *\(.*[^:]:[^: ]', line) or
- Search('for *\(.*[^: ]:[^:]', line)):
- error(filename, linenum, 'whitespace/forcolon', 2,
- 'Missing space around colon in range-based for loop')
+
+def IsDecltype(clean_lines, linenum, column):
+ """Check if the token ending on (linenum, column) is decltype().
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: the number of the line to check.
+ column: end column of the token to check.
+ Returns:
+ True if this token is decltype() expression, False otherwise.
+ """
+ (text, _, start_col) = ReverseCloseExpression(clean_lines, linenum, column)
+ if start_col < 0:
+ return False
+ if Search(r'\bdecltype\s*$', text[0:start_col]):
+ return True
+ return False
def CheckSectionSpacing(filename, clean_lines, class_info, linenum, error):
# used for brace initializers inside function calls. We don't detect this
# perfectly: we just don't complain if the last non-whitespace character on
# the previous non-blank line is ',', ';', ':', '(', '{', or '}', or if the
- # previous line starts a preprocessor block.
+ # previous line starts a preprocessor block. We also allow a brace on the
+ # following line if it is part of an array initialization and would not fit
+ # within the 80 character limit of the preceding line.
prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0]
if (not Search(r'[,;:}{(]\s*$', prevline) and
- not Match(r'\s*#', prevline)):
+ not Match(r'\s*#', prevline) and
+ not (GetLineWidth(prevline) > _line_length - 2 and '[]' in prevline)):
error(filename, linenum, 'whitespace/braces', 4,
'{ should almost always be at the end of the previous line')
# An else clause should be on the same line as the preceding closing brace.
- if Match(r'\s*else\s*', line):
+ if Match(r'\s*else\b\s*(?:if\b|\{|$)', line):
prevline = GetPreviousNonBlankLine(clean_lines, linenum)[0]
if Match(r'\s*}\s*$', prevline):
error(filename, linenum, 'whitespace/newline', 4,
# If braces come on one side of an else, they should be on both.
# However, we have to worry about "else if" that spans multiple lines!
- if Search(r'}\s*else[^{]*$', line) or Match(r'[^}]*else\s*{', line):
- if Search(r'}\s*else if([^{]*)$', line): # could be multi-line if
- # find the ( after the if
- pos = line.find('else if')
- pos = line.find('(', pos)
- if pos > 0:
- (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos)
- if endline[endpos:].find('{') == -1: # must be brace after if
- error(filename, linenum, 'readability/braces', 5,
- 'If an else has a brace on one side, it should have it on both')
- else: # common case: else not followed by a multi-line if
- error(filename, linenum, 'readability/braces', 5,
- 'If an else has a brace on one side, it should have it on both')
+ if Search(r'else if\s*\(', line): # could be multi-line if
+ brace_on_left = bool(Search(r'}\s*else if\s*\(', line))
+ # find the ( after the if
+ pos = line.find('else if')
+ pos = line.find('(', pos)
+ if pos > 0:
+ (endline, _, endpos) = CloseExpression(clean_lines, linenum, pos)
+ brace_on_right = endline[endpos:].find('{') != -1
+ if brace_on_left != brace_on_right: # must be brace after if
+ error(filename, linenum, 'readability/braces', 5,
+ 'If an else has a brace on one side, it should have it on both')
+ elif Search(r'}\s*else[^{]*$', line) or Match(r'[^}]*else\s*{', line):
+ error(filename, linenum, 'readability/braces', 5,
+ 'If an else has a brace on one side, it should have it on both')
# Likewise, an else should never have the else clause on the same line
if Search(r'\belse [^\s{]', line) and not Search(r'\belse if\b', line):
error(filename, linenum, 'whitespace/newline', 4,
'do/while clauses should not be on a single line')
+ # Check single-line if/else bodies. The style guide says 'curly braces are not
+ # required for single-line statements'. We additionally allow multi-line,
+ # single statements, but we reject anything with more than one semicolon in
+ # it. This means that the first semicolon after the if should be at the end of
+ # its line, and the line after that should have an indent level equal to or
+ # lower than the if. We also check for ambiguous if/else nesting without
+ # braces.
+ if_else_match = Search(r'\b(if\s*\(|else\b)', line)
+ if if_else_match and not Match(r'\s*#', line):
+ if_indent = GetIndentLevel(line)
+ endline, endlinenum, endpos = line, linenum, if_else_match.end()
+ if_match = Search(r'\bif\s*\(', line)
+ if if_match:
+ # This could be a multiline if condition, so find the end first.
+ pos = if_match.end() - 1
+ (endline, endlinenum, endpos) = CloseExpression(clean_lines, linenum, pos)
+ # Check for an opening brace, either directly after the if or on the next
+ # line. If found, this isn't a single-statement conditional.
+ if (not Match(r'\s*{', endline[endpos:])
+ and not (Match(r'\s*$', endline[endpos:])
+ and endlinenum < (len(clean_lines.elided) - 1)
+ and Match(r'\s*{', clean_lines.elided[endlinenum + 1]))):
+ while (endlinenum < len(clean_lines.elided)
+ and ';' not in clean_lines.elided[endlinenum][endpos:]):
+ endlinenum += 1
+ endpos = 0
+ if endlinenum < len(clean_lines.elided):
+ endline = clean_lines.elided[endlinenum]
+ # We allow a mix of whitespace and closing braces (e.g. for one-liner
+ # methods) and a single \ after the semicolon (for macros)
+ endpos = endline.find(';')
+ if not Match(r';[\s}]*(\\?)$', endline[endpos:]):
+ # Semicolon isn't the last character, there's something trailing.
+ # Output a warning if the semicolon is not contained inside
+ # a lambda expression.
+ if not Match(r'^[^{};]*\[[^\[\]]*\][^{}]*\{[^{}]*\}\s*\)*[;,]\s*$',
+ endline):
+ error(filename, linenum, 'readability/braces', 4,
+ 'If/else bodies with multiple statements require braces')
+ elif endlinenum < len(clean_lines.elided) - 1:
+ # Make sure the next line is dedented
+ next_line = clean_lines.elided[endlinenum + 1]
+ next_indent = GetIndentLevel(next_line)
+ # With ambiguous nested if statements, this will error out on the
+ # if that *doesn't* match the else, regardless of whether it's the
+ # inner one or outer one.
+ if (if_match and Match(r'\s*else\b', next_line)
+ and next_indent != if_indent):
+ error(filename, linenum, 'readability/braces', 4,
+ 'Else clause should be indented at the same level as if. '
+ 'Ambiguous nested if/else chains require braces.')
+ elif next_indent > if_indent:
+ error(filename, linenum, 'readability/braces', 4,
+ 'If/else bodies with multiple statements require braces')
+
+
+def CheckTrailingSemicolon(filename, clean_lines, linenum, error):
+ """Looks for redundant trailing semicolon.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+
+ line = clean_lines.elided[linenum]
+
# Block bodies should not be followed by a semicolon. Due to C++11
# brace initialization, there are more places where semicolons are
- # required than not, so we use a whitelist approach to check these
- # rather than a blacklist. These are the places where "};" should
- # be replaced by just "}":
+ # required than not, so we explicitly list the allowed rules rather
+ # than listing the disallowed ones. These are the places where "};"
+ # should be replaced by just "}":
# 1. Some flavor of block following closing parenthesis:
# for (;;) {};
# while (...) {};
# - INTERFACE_DEF
# - EXCLUSIVE_LOCKS_REQUIRED, SHARED_LOCKS_REQUIRED, LOCKS_EXCLUDED:
#
- # We implement a whitelist of safe macros instead of a blacklist of
+ # We implement a list of safe macros instead of a list of
# unsafe macros, even though the latter appears less frequently in
# google code and would have been easier to implement. This is because
- # the downside for getting the whitelist wrong means some extra
- # semicolons, while the downside for getting the blacklist wrong
+ # the downside for getting the allowed checks wrong means some extra
+ # semicolons, while the downside for getting disallowed checks wrong
# would result in compile errors.
#
- # In addition to macros, we also don't want to warn on compound
- # literals.
+ # In addition to macros, we also don't want to warn on
+ # - Compound literals
+ # - Lambdas
+ # - alignas specifier with anonymous structs
+ # - decltype
closing_brace_pos = match.group(1).rfind(')')
opening_parenthesis = ReverseCloseExpression(
clean_lines, linenum, closing_brace_pos)
if opening_parenthesis[2] > -1:
line_prefix = opening_parenthesis[0][0:opening_parenthesis[2]]
- macro = Search(r'\b([A-Z_]+)\s*$', line_prefix)
+ macro = Search(r'\b([A-Z_][A-Z0-9_]*)\s*$', line_prefix)
+ func = Match(r'^(.*\])\s*$', line_prefix)
if ((macro and
macro.group(1) not in (
'TEST', 'TEST_F', 'MATCHER', 'MATCHER_P', 'TYPED_TEST',
'EXCLUSIVE_LOCKS_REQUIRED', 'SHARED_LOCKS_REQUIRED',
'LOCKS_EXCLUDED', 'INTERFACE_DEF')) or
+ (func and not Search(r'\boperator\s*\[\s*\]', func.group(1))) or
+ Search(r'\b(?:struct|union)\s+alignas\s*$', line_prefix) or
+ Search(r'\bdecltype$', line_prefix) or
Search(r'\s+=\s*$', line_prefix)):
match = None
+ if (match and
+ opening_parenthesis[1] > 1 and
+ Search(r'\]\s*$', clean_lines.elided[opening_parenthesis[1] - 1])):
+ # Multi-line lambda-expression
+ match = None
else:
# Try matching cases 2-3.
# outputting warnings for the matching closing brace, if there are
# nested blocks with trailing semicolons, we will get the error
# messages in reversed order.
+
+ # We need to check the line forward for NOLINT
+ raw_lines = clean_lines.raw_lines
+ ParseNolintSuppressions(filename, raw_lines[endlinenum-1], endlinenum-1,
+ error)
+ ParseNolintSuppressions(filename, raw_lines[endlinenum], endlinenum,
+ error)
+
error(filename, endlinenum, 'readability/braces', 4,
"You don't need a ; after a }")
line = clean_lines.elided[linenum]
matched = Match(r'\s*(for|while|if)\s*\(', line)
if matched:
- # Find the end of the conditional expression
+ # Find the end of the conditional expression.
(end_line, end_linenum, end_pos) = CloseExpression(
clean_lines, linenum, line.find('('))
error(filename, end_linenum, 'whitespace/empty_loop_body', 5,
'Empty loop bodies should use {} or continue')
+ # Check for if statements that have completely empty bodies (no comments)
+ # and no else clauses.
+ if end_pos >= 0 and matched.group(1) == 'if':
+ # Find the position of the opening { for the if statement.
+ # Return without logging an error if it has no brackets.
+ opening_linenum = end_linenum
+ opening_line_fragment = end_line[end_pos:]
+ # Loop until EOF or find anything that's not whitespace or opening {.
+ while not Search(r'^\s*\{', opening_line_fragment):
+ if Search(r'^(?!\s*$)', opening_line_fragment):
+ # Conditional has no brackets.
+ return
+ opening_linenum += 1
+ if opening_linenum == len(clean_lines.elided):
+ # Couldn't find conditional's opening { or any code before EOF.
+ return
+ opening_line_fragment = clean_lines.elided[opening_linenum]
+ # Set opening_line (opening_line_fragment may not be entire opening line).
+ opening_line = clean_lines.elided[opening_linenum]
+
+ # Find the position of the closing }.
+ opening_pos = opening_line_fragment.find('{')
+ if opening_linenum == end_linenum:
+ # We need to make opening_pos relative to the start of the entire line.
+ opening_pos += end_pos
+ (closing_line, closing_linenum, closing_pos) = CloseExpression(
+ clean_lines, opening_linenum, opening_pos)
+ if closing_pos < 0:
+ return
+
+ # Now construct the body of the conditional. This consists of the portion
+ # of the opening line after the {, all lines until the closing line,
+ # and the portion of the closing line before the }.
+ if (clean_lines.raw_lines[opening_linenum] !=
+ CleanseComments(clean_lines.raw_lines[opening_linenum])):
+ # Opening line ends with a comment, so conditional isn't empty.
+ return
+ if closing_linenum > opening_linenum:
+ # Opening line after the {. Ignore comments here since we checked above.
+ body = list(opening_line[opening_pos+1:])
+ # All lines until closing line, excluding closing line, with comments.
+ body.extend(clean_lines.raw_lines[opening_linenum+1:closing_linenum])
+ # Closing line before the }. Won't (and can't) have comments.
+ body.append(clean_lines.elided[closing_linenum][:closing_pos-1])
+ body = '\n'.join(body)
+ else:
+ # If statement has brackets and fits on a single line.
+ body = opening_line[opening_pos+1:closing_pos-1]
+
+ # Check if the body is empty
+ if not _EMPTY_CONDITIONAL_BODY_PATTERN.search(body):
+ return
+ # The body is empty. Now make sure there's not an else clause.
+ current_linenum = closing_linenum
+ current_line_fragment = closing_line[closing_pos:]
+ # Loop until EOF or find anything that's not whitespace or else clause.
+ while Search(r'^\s*$|^(?=\s*else)', current_line_fragment):
+ if Search(r'^(?=\s*else)', current_line_fragment):
+ # Found an else clause, so don't log an error.
+ return
+ current_linenum += 1
+ if current_linenum == len(clean_lines.elided):
+ break
+ current_line_fragment = clean_lines.elided[current_linenum]
+
+ # The body is empty and there's no else clause until EOF or other code.
+ error(filename, end_linenum, 'whitespace/empty_if_body', 4,
+ ('If statement had no body and no else clause'))
+
+
+def FindCheckMacro(line):
+ """Find a replaceable CHECK-like macro.
+
+ Args:
+ line: line to search on.
+ Returns:
+ (macro name, start position), or (None, -1) if no replaceable
+ macro is found.
+ """
+ for macro in _CHECK_MACROS:
+ i = line.find(macro)
+ if i >= 0:
+ # Find opening parenthesis. Do a regular expression match here
+ # to make sure that we are matching the expected CHECK macro, as
+ # opposed to some other macro that happens to contain the CHECK
+ # substring.
+ matched = Match(r'^(.*\b' + macro + r'\s*)\(', line)
+ if not matched:
+ continue
+ return (macro, len(matched.group(1)))
+ return (None, -1)
+
def CheckCheck(filename, clean_lines, linenum, error):
"""Checks the use of CHECK and EXPECT macros.
# Decide the set of replacement macros that should be suggested
lines = clean_lines.elided
- check_macro = None
- start_pos = -1
- for macro in _CHECK_MACROS:
- i = lines[linenum].find(macro)
- if i >= 0:
- check_macro = macro
-
- # Find opening parenthesis. Do a regular expression match here
- # to make sure that we are matching the expected CHECK macro, as
- # opposed to some other macro that happens to contain the CHECK
- # substring.
- matched = Match(r'^(.*\b' + check_macro + r'\s*)\(', lines[linenum])
- if not matched:
- continue
- start_pos = len(matched.group(1))
- break
- if not check_macro or start_pos < 0:
- # Don't waste time here if line doesn't contain 'CHECK' or 'EXPECT'
+ (check_macro, start_pos) = FindCheckMacro(lines[linenum])
+ if not check_macro:
return
# Find end of the boolean expression by matching parentheses
clean_lines, linenum, start_pos)
if end_pos < 0:
return
+
+ # If the check macro is followed by something other than a
+ # semicolon, assume users will log their own custom error messages
+ # and don't suggest any replacements.
+ if not Match(r'\s*;', last_line[end_pos:]):
+ return
+
if linenum == end_line:
expression = lines[linenum][start_pos + 1:end_pos - 1]
else:
if token == '(':
# Parenthesized operand
expression = matched.group(2)
- (end, _) = FindEndOfExpressionInLine(expression, 0, 1, '(', ')')
+ (end, _) = FindEndOfExpressionInLine(expression, 0, ['('])
if end < 0:
return # Unmatched parenthesis
lhs += '(' + expression[0:end]
if unicodedata.east_asian_width(uc) in ('W', 'F'):
width += 2
elif not unicodedata.combining(uc):
+ # Issue 337
+ # https://mail.python.org/pipermail/python-list/2012-August/628809.html
+ if (sys.version_info.major, sys.version_info.minor) <= (3, 2):
+ # https://github.com/python/cpython/blob/2.7/Include/unicodeobject.h#L81
+ is_wide_build = sysconfig.get_config_var("Py_UNICODE_SIZE") >= 4
+ # https://github.com/python/cpython/blob/2.7/Objects/unicodeobject.c#L564
+ is_low_surrogate = 0xDC00 <= ord(uc) <= 0xDFFF
+ if not is_wide_build and is_low_surrogate:
+ width -= 1
+
width += 1
return width
else:
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
file_extension: The extension (without the dot) of the filename.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: The function to call with any errors found.
"""
# raw strings,
raw_lines = clean_lines.lines_without_raw_strings
line = raw_lines[linenum]
+ prev = raw_lines[linenum - 1] if linenum > 0 else ''
if line.find('\t') != -1:
error(filename, linenum, 'whitespace/tab', 1,
# if(match($0, " <<")) complain = 0;
# if(match(prev, " +for \\(")) complain = 0;
# if(prevodd && match(prevprev, " +for \\(")) complain = 0;
+ scope_or_label_pattern = r'\s*\w+\s*:\s*\\?$'
+ classinfo = nesting_state.InnermostClass()
initial_spaces = 0
cleansed_line = clean_lines.elided[linenum]
while initial_spaces < len(line) and line[initial_spaces] == ' ':
initial_spaces += 1
- if line and line[-1].isspace():
- error(filename, linenum, 'whitespace/end_of_line', 4,
- 'Line ends in whitespace. Consider deleting these extra spaces.')
- # There are certain situations we allow one space, notably for section labels
- elif ((initial_spaces == 1 or initial_spaces == 3) and
- not Match(r'\s*\w+\s*:\s*$', cleansed_line)):
+ # There are certain situations we allow one space, notably for
+ # section labels, and also lines containing multi-line raw strings.
+ # We also don't check for lines that look like continuation lines
+ # (of lines ending in double quotes, commas, equals, or angle brackets)
+ # because the rules for how to indent those are non-trivial.
+ if (not Search(r'[",=><] *$', prev) and
+ (initial_spaces == 1 or initial_spaces == 3) and
+ not Match(scope_or_label_pattern, cleansed_line) and
+ not (clean_lines.raw_lines[linenum] != line and
+ Match(r'^\s*""', line))):
error(filename, linenum, 'whitespace/indent', 3,
'Weird number of spaces at line-start. '
'Are you using a 2-space indent?')
+ if line and line[-1].isspace():
+ error(filename, linenum, 'whitespace/end_of_line', 4,
+ 'Line ends in whitespace. Consider deleting these extra spaces.')
+
# Check if the line is a header guard.
is_header_guard = False
- if file_extension == 'h':
+ if IsHeaderExtension(file_extension):
cppvar = GetHeaderGuardCPPVariable(filename)
if (line.startswith('#ifndef %s' % cppvar) or
line.startswith('#define %s' % cppvar) or
# developers fault.
if (not line.startswith('#include') and not is_header_guard and
not Match(r'^\s*//.*http(s?)://\S*$', line) and
+ not Match(r'^\s*//\s*[^\s]*$', line) and
not Match(r'^// \$Id:.*#[0-9]+ \$$', line)):
line_width = GetLineWidth(line)
- extended_length = int((_line_length * 1.25))
- if line_width > extended_length:
- error(filename, linenum, 'whitespace/line_length', 4,
- 'Lines should very rarely be longer than %i characters' %
- extended_length)
- elif line_width > _line_length:
+ if line_width > _line_length:
error(filename, linenum, 'whitespace/line_length', 2,
'Lines should be <= %i characters long' % _line_length)
# Some more style checks
CheckBraces(filename, clean_lines, linenum, error)
+ CheckTrailingSemicolon(filename, clean_lines, linenum, error)
CheckEmptyBlockBody(filename, clean_lines, linenum, error)
- CheckAccess(filename, clean_lines, linenum, nesting_state, error)
CheckSpacing(filename, clean_lines, linenum, nesting_state, error)
+ CheckOperatorSpacing(filename, clean_lines, linenum, error)
+ CheckParenthesisSpacing(filename, clean_lines, linenum, error)
+ CheckCommaSpacing(filename, clean_lines, linenum, error)
+ CheckBracesSpacing(filename, clean_lines, linenum, nesting_state, error)
+ CheckSpacingForFunctionCall(filename, clean_lines, linenum, error)
CheckCheck(filename, clean_lines, linenum, error)
CheckAltTokens(filename, clean_lines, linenum, error)
classinfo = nesting_state.InnermostClass()
CheckSectionSpacing(filename, clean_lines, classinfo, linenum, error)
-_RE_PATTERN_INCLUDE_NEW_STYLE = re.compile(r'#include +"[^/]+\.h"')
_RE_PATTERN_INCLUDE = re.compile(r'^\s*#\s*include\s*([<"])([^>"]*)[>"].*$')
# Matches the first component of a filename delimited by -s and _s. That is:
# _RE_FIRST_COMPONENT.match('foo').group(0) == 'foo'
return os.path.splitext(filename)[0]
-def _IsTestFilename(filename):
- """Determines if the given filename has a suffix that identifies it as a test.
-
- Args:
- filename: The input filename.
-
- Returns:
- True if 'filename' looks like a test, False otherwise.
- """
- if (filename.endswith('_test.cc') or
- filename.endswith('_unittest.cc') or
- filename.endswith('_regtest.cc')):
- return True
- else:
- return False
-
-
def _ClassifyInclude(fileinfo, include, is_system):
"""Figures out what kind of header 'include' is.
error: The function to call with any errors found.
"""
fileinfo = FileInfo(filename)
-
line = clean_lines.lines[linenum]
# "include" should use the new style "foo/bar.h" instead of just "bar.h"
- if _RE_PATTERN_INCLUDE_NEW_STYLE.search(line):
+ # Only do this check if the included header follows google naming
+ # conventions. If not, assume that it's a 3rd party API that
+ # requires special include conventions.
+ #
+ # We also make an exception for Lua headers, which follow google
+ # naming convention but not the include convention.
+ match = Match(r'#include\s*"([^/]+\.h)"', line)
+ if match and not _THIRD_PARTY_HEADERS_PATTERN.match(match.group(1)):
error(filename, linenum, 'build/include', 4,
'Include the directory when naming .h files')
if match:
include = match.group(2)
is_system = (match.group(1) == '<')
- if include in include_state:
+ duplicate_line = include_state.FindHeader(include)
+ if duplicate_line >= 0:
error(filename, linenum, 'build/include', 4,
'"%s" already included at %s:%s' %
- (include, filename, include_state[include]))
- else:
- include_state[include] = linenum
+ (include, filename, duplicate_line))
+ elif (include.endswith('.cc') and
+ os.path.dirname(fileinfo.RepositoryName()) != os.path.dirname(include)):
+ error(filename, linenum, 'build/include', 4,
+ 'Do not include .cc files from other packages')
+ elif not _THIRD_PARTY_HEADERS_PATTERN.match(include):
+ include_state.include_list[-1].append((include, linenum))
# We want to ensure that headers appear in the right order:
# 1) for foo.cc, foo.h (preferred location)
'Include "%s" not in alphabetical order' % include)
include_state.SetLastHeader(canonical_include)
- # Look for any of the stream classes that are part of standard C++.
- match = _RE_PATTERN_INCLUDE.match(line)
- if match:
- include = match.group(2)
- if Match(r'(f|ind|io|i|o|parse|pf|stdio|str|)?stream$', include):
- # Many unit tests use cout, so we exempt them.
- if not _IsTestFilename(filename):
- error(filename, linenum, 'readability/streams', 3,
- 'Streams are highly discouraged.')
def _GetTextInside(text, start_pattern):
The extracted text.
None if either the opening string or ending punctuation could not be found.
"""
- # TODO(sugawarayu): Audit cpplint.py to see what places could be profitably
+ # TODO(unknown): Audit cpplint.py to see what places could be profitably
# rewritten to use _GetTextInside (and use inferior regexp matching today).
# Give opening punctuations to get the matching close-punctuations.
_RE_PATTERN_CONST_REF_PARAM = (
r'(?:.*\s*\bconst\s*&\s*' + _RE_PATTERN_IDENT +
r'|const\s+' + _RE_PATTERN_TYPE + r'\s*&\s*' + _RE_PATTERN_IDENT + r')')
+# Stream types.
+_RE_PATTERN_REF_STREAM_PARAM = (
+ r'(?:.*stream\s*&\s*' + _RE_PATTERN_IDENT + r')')
def CheckLanguage(filename, clean_lines, linenum, file_extension,
linenum: The number of the line to check.
file_extension: The extension (without the dot) of the filename.
include_state: An _IncludeState instance in which the headers are inserted.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: The function to call with any errors found.
"""
# Reset include state across preprocessor directives. This is meant
# to silence warnings for conditional includes.
- if Match(r'^\s*#\s*(?:ifdef|elif|else|endif)\b', line):
- include_state.ResetSection()
+ match = Match(r'^\s*#\s*(if|ifdef|ifndef|elif|else|endif)\b', line)
+ if match:
+ include_state.ResetSection(match.group(1))
# Make Windows paths like Unix.
fullname = os.path.abspath(filename).replace('\\', '/')
- # TODO(unknown): figure out if they're using default arguments in fn proto.
+ # Perform other checks now that we are sure that this is not an include line
+ CheckCasts(filename, clean_lines, linenum, error)
+ CheckGlobalStatic(filename, clean_lines, linenum, error)
+ CheckPrintf(filename, clean_lines, linenum, error)
- # Check to see if they're using an conversion function cast.
- # I just try to capture the most common basic types, though there are more.
- # Parameterless conversion functions, such as bool(), are allowed as they are
- # probably a member operator declaration or default constructor.
- match = Search(
- r'(\bnew\s+)?\b' # Grab 'new' operator, if it's there
- r'(int|float|double|bool|char|int32|uint32|int64|uint64)'
- r'(\([^)].*)', line)
- if match:
- matched_new = match.group(1)
- matched_type = match.group(2)
- matched_funcptr = match.group(3)
-
- # gMock methods are defined using some variant of MOCK_METHODx(name, type)
- # where type may be float(), int(string), etc. Without context they are
- # virtually indistinguishable from int(x) casts. Likewise, gMock's
- # MockCallback takes a template parameter of the form return_type(arg_type),
- # which looks much like the cast we're trying to detect.
- #
- # std::function<> wrapper has a similar problem.
- #
- # Return types for function pointers also look like casts if they
- # don't have an extra space.
- if (matched_new is None and # If new operator, then this isn't a cast
- not (Match(r'^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(', line) or
- Search(r'\bMockCallback<.*>', line) or
- Search(r'\bstd::function<.*>', line)) and
- not (matched_funcptr and
- Match(r'\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(',
- matched_funcptr))):
- # Try a bit harder to catch gmock lines: the only place where
- # something looks like an old-style cast is where we declare the
- # return type of the mocked method, and the only time when we
- # are missing context is if MOCK_METHOD was split across
- # multiple lines. The missing MOCK_METHOD is usually one or two
- # lines back, so scan back one or two lines.
- #
- # It's not possible for gmock macros to appear in the first 2
- # lines, since the class head + section name takes up 2 lines.
- if (linenum < 2 or
- not (Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$',
- clean_lines.elided[linenum - 1]) or
- Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$',
- clean_lines.elided[linenum - 2]))):
- error(filename, linenum, 'readability/casting', 4,
- 'Using deprecated casting style. '
- 'Use static_cast<%s>(...) instead' %
- matched_type)
-
- CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
- 'static_cast',
- r'\((int|float|double|bool|char|u?int(16|32|64))\)', error)
-
- # This doesn't catch all cases. Consider (const char * const)"hello".
- #
- # (char *) "foo" should always be a const_cast (reinterpret_cast won't
- # compile).
- if CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
- 'const_cast', r'\((char\s?\*+\s?)\)\s*"', error):
- pass
- else:
- # Check pointer casts for other than string constants
- CheckCStyleCast(filename, linenum, line, clean_lines.raw_lines[linenum],
- 'reinterpret_cast', r'\((\w+\s?\*+\s?)\)', error)
-
- # In addition, we look for people taking the address of a cast. This
- # is dangerous -- casts can assign to temporaries, so the pointer doesn't
- # point where you think.
- match = Search(
- r'(?:&\(([^)]+)\)[\w(])|'
- r'(?:&(static|dynamic|down|reinterpret)_cast\b)', line)
- if match and match.group(1) != '*':
- error(filename, linenum, 'runtime/casting', 4,
- ('Are you taking an address of a cast? '
- 'This is dangerous: could be a temp var. '
- 'Take the address before doing the cast, rather than after'))
-
- # Create an extended_line, which is the concatenation of the current and
- # next lines, for more effective checking of code that may span more than one
- # line.
- if linenum + 1 < clean_lines.NumLines():
- extended_line = line + clean_lines.elided[linenum + 1]
- else:
- extended_line = line
-
- # Check for people declaring static/global STL strings at the top level.
- # This is dangerous because the C++ language does not guarantee that
- # globals with constructors are initialized before the first access.
- match = Match(
- r'((?:|static +)(?:|const +))string +([a-zA-Z0-9_:]+)\b(.*)',
- line)
- # Make sure it's not a function.
- # Function template specialization looks like: "string foo<Type>(...".
- # Class template definitions look like: "string Foo<Type>::Method(...".
- #
- # Also ignore things that look like operators. These are matched separately
- # because operator names cross non-word boundaries. If we change the pattern
- # above, we would decrease the accuracy of matching identifiers.
- if (match and
- not Search(r'\boperator\W', line) and
- not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)?\s*\(([^"]|$)', match.group(3))):
- error(filename, linenum, 'runtime/string', 4,
- 'For a static/global string constant, use a C style string instead: '
- '"%schar %s[]".' %
- (match.group(1), match.group(2)))
-
- if Search(r'\b([A-Za-z0-9_]*_)\(\1\)', line):
- error(filename, linenum, 'runtime/init', 4,
- 'You seem to be initializing a member variable with itself.')
-
- if file_extension == 'h':
+ if IsHeaderExtension(file_extension):
# TODO(unknown): check that 1-arg constructors are explicit.
# How to tell it's a constructor?
# (handled in CheckForNonStandardConstructs for now)
- # TODO(unknown): check that classes have DISALLOW_EVIL_CONSTRUCTORS
+ # TODO(unknown): check that classes declare or disable copy/assign
# (level 1 error)
pass
error(filename, linenum, 'runtime/int', 4,
'Use int16/int64/etc, rather than the C type %s' % match.group(1))
- # When snprintf is used, the second argument shouldn't be a literal.
- match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line)
- if match and match.group(2) != '0':
- # If 2nd arg is zero, snprintf is used to calculate size.
- error(filename, linenum, 'runtime/printf', 3,
- 'If you can, use sizeof(%s) instead of %s as the 2nd arg '
- 'to snprintf.' % (match.group(1), match.group(2)))
-
- # Check if some verboten C functions are being used.
- if Search(r'\bsprintf\b', line):
- error(filename, linenum, 'runtime/printf', 5,
- 'Never use sprintf. Use snprintf instead.')
- match = Search(r'\b(strcpy|strcat)\b', line)
- if match:
- error(filename, linenum, 'runtime/printf', 4,
- 'Almost always, snprintf is better than %s' % match.group(1))
-
# Check if some verboten operator overloading is going on
# TODO(unknown): catch out-of-line unary operator&:
# class X {};
# Check for potential format string bugs like printf(foo).
# We constrain the pattern not to pick things like DocidForPrintf(foo).
# Not perfect but it can catch printf(foo.c_str()) and printf(foo->c_str())
- # TODO(sugawarayu): Catch the following case. Need to change the calling
+ # TODO(unknown): Catch the following case. Need to change the calling
# convention of the whole function to process multiple line to handle it.
# printf(
# boy_this_is_a_really_long_variable_that_cannot_fit_on_the_prev_line);
'Do not use variable-length arrays. Use an appropriately named '
"('k' followed by CamelCase) compile-time constant for the size.")
- # If DISALLOW_EVIL_CONSTRUCTORS, DISALLOW_COPY_AND_ASSIGN, or
- # DISALLOW_IMPLICIT_CONSTRUCTORS is present, then it should be the last thing
- # in the class declaration.
- match = Match(
- (r'\s*'
- r'(DISALLOW_(EVIL_CONSTRUCTORS|COPY_AND_ASSIGN|IMPLICIT_CONSTRUCTORS))'
- r'\(.*\);$'),
- line)
- if match and linenum + 1 < clean_lines.NumLines():
- next_line = clean_lines.elided[linenum + 1]
- # We allow some, but not all, declarations of variables to be present
- # in the statement that defines the class. The [\w\*,\s]* fragment of
- # the regular expression below allows users to declare instances of
- # the class or pointers to instances, but not less common types such
- # as function pointers or arrays. It's a tradeoff between allowing
- # reasonable code and avoiding trying to parse more C++ using regexps.
- if not Search(r'^\s*}[\w\*,\s]*;', next_line):
- error(filename, linenum, 'readability/constructors', 3,
- match.group(1) + ' should be the last thing in the class')
-
# Check for use of unnamed namespaces in header files. Registration
# macros are typically OK, so we allow use of "namespace {" on lines
# that end with backslashes.
- if (file_extension == 'h'
+ if (IsHeaderExtension(file_extension)
and Search(r'\bnamespace\s*{', line)
and line[-1] != '\\'):
error(filename, linenum, 'build/namespaces', 4,
'Do not use unnamed namespaces in header files. See '
- 'http://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces'
+ 'https://google-styleguide.googlecode.com/svn/trunk/cppguide.xml#Namespaces'
' for more information.')
+
+def CheckGlobalStatic(filename, clean_lines, linenum, error):
+ """Check for unsafe global or static objects.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ # Match two lines at a time to support multiline declarations
+ if linenum + 1 < clean_lines.NumLines() and not Search(r'[;({]', line):
+ line += clean_lines.elided[linenum + 1].strip()
+
+ # Check for people declaring static/global STL strings at the top level.
+ # This is dangerous because the C++ language does not guarantee that
+ # globals with constructors are initialized before the first access, and
+ # also because globals can be destroyed when some threads are still running.
+ # TODO(unknown): Generalize this to also find static unique_ptr instances.
+ # TODO(unknown): File bugs for clang-tidy to find these.
+ match = Match(
+ r'((?:|static +)(?:|const +))(?::*std::)?string( +const)? +'
+ r'([a-zA-Z0-9_:]+)\b(.*)',
+ line)
+
+ # Remove false positives:
+ # - String pointers (as opposed to values).
+ # string *pointer
+ # const string *pointer
+ # string const *pointer
+ # string *const pointer
+ #
+ # - Functions and template specializations.
+ # string Function<Type>(...
+ # string Class<Type>::Method(...
+ #
+ # - Operators. These are matched separately because operator names
+ # cross non-word boundaries, and trying to match both operators
+ # and functions at the same time would decrease accuracy of
+ # matching identifiers.
+ # string Class::operator*()
+ if (match and
+ not Search(r'\bstring\b(\s+const)?\s*[\*\&]\s*(const\s+)?\w', line) and
+ not Search(r'\boperator\W', line) and
+ not Match(r'\s*(<.*>)?(::[a-zA-Z0-9_]+)*\s*\(([^"]|$)', match.group(4))):
+ if Search(r'\bconst\b', line):
+ error(filename, linenum, 'runtime/string', 4,
+ 'For a static/global string constant, use a C style string '
+ 'instead: "%schar%s %s[]".' %
+ (match.group(1), match.group(2) or '', match.group(3)))
+ else:
+ error(filename, linenum, 'runtime/string', 4,
+ 'Static/global string variables are not permitted.')
+
+ if (Search(r'\b([A-Za-z0-9_]*_)\(\1\)', line) or
+ Search(r'\b([A-Za-z0-9_]*_)\(CHECK_NOTNULL\(\1\)\)', line)):
+ error(filename, linenum, 'runtime/init', 4,
+ 'You seem to be initializing a member variable with itself.')
+
+
+def CheckPrintf(filename, clean_lines, linenum, error):
+ """Check for printf related issues.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ # When snprintf is used, the second argument shouldn't be a literal.
+ match = Search(r'snprintf\s*\(([^,]*),\s*([0-9]*)\s*,', line)
+ if match and match.group(2) != '0':
+ # If 2nd arg is zero, snprintf is used to calculate size.
+ error(filename, linenum, 'runtime/printf', 3,
+ 'If you can, use sizeof(%s) instead of %s as the 2nd arg '
+ 'to snprintf.' % (match.group(1), match.group(2)))
+
+ # Check if some verboten C functions are being used.
+ if Search(r'\bsprintf\s*\(', line):
+ error(filename, linenum, 'runtime/printf', 5,
+ 'Never use sprintf. Use snprintf instead.')
+ match = Search(r'\b(strcpy|strcat)\s*\(', line)
+ if match:
+ error(filename, linenum, 'runtime/printf', 4,
+ 'Almost always, snprintf is better than %s' % match.group(1))
+
+
+def IsDerivedFunction(clean_lines, linenum):
+ """Check if current line contains an inherited function.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ Returns:
+ True if current line contains a function with "override"
+ virt-specifier.
+ """
+ # Scan back a few lines for start of current function
+ for i in xrange(linenum, max(-1, linenum - 10), -1):
+ match = Match(r'^([^()]*\w+)\(', clean_lines.elided[i])
+ if match:
+ # Look for "override" after the matching closing parenthesis
+ line, _, closing_paren = CloseExpression(
+ clean_lines, i, len(match.group(1)))
+ return (closing_paren >= 0 and
+ Search(r'\boverride\b', line[closing_paren:]))
+ return False
+
+
+def IsOutOfLineMethodDefinition(clean_lines, linenum):
+ """Check if current line contains an out-of-line method definition.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ Returns:
+ True if current line contains an out-of-line method definition.
+ """
+ # Scan back a few lines for start of current function
+ for i in xrange(linenum, max(-1, linenum - 10), -1):
+ if Match(r'^([^()]*\w+)\(', clean_lines.elided[i]):
+ return Match(r'^[^()]*\w+::\w+\(', clean_lines.elided[i]) is not None
+ return False
+
+
+def IsInitializerList(clean_lines, linenum):
+ """Check if current line is inside constructor initializer list.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ Returns:
+ True if current line appears to be inside constructor initializer
+ list, False otherwise.
+ """
+ for i in xrange(linenum, 1, -1):
+ line = clean_lines.elided[i]
+ if i == linenum:
+ remove_function_body = Match(r'^(.*)\{\s*$', line)
+ if remove_function_body:
+ line = remove_function_body.group(1)
+
+ if Search(r'\s:\s*\w+[({]', line):
+ # A lone colon tend to indicate the start of a constructor
+ # initializer list. It could also be a ternary operator, which
+ # also tend to appear in constructor initializer lists as
+ # opposed to parameter lists.
+ return True
+ if Search(r'\}\s*,\s*$', line):
+ # A closing brace followed by a comma is probably the end of a
+ # brace-initialized member in constructor initializer list.
+ return True
+ if Search(r'[{};]\s*$', line):
+ # Found one of the following:
+ # - A closing brace or semicolon, probably the end of the previous
+ # function.
+ # - An opening brace, probably the start of current class or namespace.
+ #
+ # Current line is probably not inside an initializer list since
+ # we saw one of those things without seeing the starting colon.
+ return False
+
+ # Got to the beginning of the file without seeing the start of
+ # constructor initializer list.
+ return False
+
+
def CheckForNonConstReference(filename, clean_lines, linenum,
nesting_state, error):
"""Check for non-const references.
filename: The name of the current file.
clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: The function to call with any errors found.
"""
if '&' not in line:
return
+ # If a function is inherited, current function doesn't have much of
+ # a choice, so any non-const references should not be blamed on
+ # derived function.
+ if IsDerivedFunction(clean_lines, linenum):
+ return
+
+ # Don't warn on out-of-line method definitions, as we would warn on the
+ # in-line declaration, if it isn't marked with 'override'.
+ if IsOutOfLineMethodDefinition(clean_lines, linenum):
+ return
+
# Long type names may be broken across multiple lines, usually in one
# of these forms:
# LongType
# inside declarators: reference parameter
# We will exclude the first two cases by checking that we are not inside a
# function body, including one that was just introduced by a trailing '{'.
- # TODO(unknwon): Doesn't account for preprocessor directives.
# TODO(unknown): Doesn't account for 'catch(Exception& e)' [rare].
- check_params = False
- if not nesting_state.stack:
- check_params = True # top level
- elif (isinstance(nesting_state.stack[-1], _ClassInfo) or
- isinstance(nesting_state.stack[-1], _NamespaceInfo)):
- check_params = True # within class or namespace
- elif Match(r'.*{\s*$', line):
- if (len(nesting_state.stack) == 1 or
- isinstance(nesting_state.stack[-2], _ClassInfo) or
- isinstance(nesting_state.stack[-2], _NamespaceInfo)):
- check_params = True # just opened global/class/namespace block
+ if (nesting_state.previous_stack_top and
+ not (isinstance(nesting_state.previous_stack_top, _ClassInfo) or
+ isinstance(nesting_state.previous_stack_top, _NamespaceInfo))):
+ # Not at toplevel, not within a class, and not within a namespace
+ return
+
+ # Avoid initializer lists. We only need to scan back from the
+ # current line for something that starts with ':'.
+ #
+ # We don't need to check the current line, since the '&' would
+ # appear inside the second set of parentheses on the current line as
+ # opposed to the first set.
+ if linenum > 0:
+ for i in xrange(linenum - 1, max(0, linenum - 10), -1):
+ previous_line = clean_lines.elided[i]
+ if not Search(r'[),]\s*$', previous_line):
+ break
+ if Match(r'^\s*:\s+\S', previous_line):
+ return
+
+ # Avoid preprocessors
+ if Search(r'\\\s*$', line):
+ return
+
+ # Avoid constructor initializer lists
+ if IsInitializerList(clean_lines, linenum):
+ return
+
# We allow non-const references in a few standard places, like functions
# called "swap()" or iostream operators like "<<" or ">>". Do not check
# those function parameters.
#
# We also accept & in static_assert, which looks like a function but
# it's actually a declaration expression.
- whitelisted_functions = (r'(?:[sS]wap(?:<\w:+>)?|'
+ allowed_functions = (r'(?:[sS]wap(?:<\w:+>)?|'
r'operator\s*[<>][<>]|'
r'static_assert|COMPILE_ASSERT'
r')\s*\(')
- if Search(whitelisted_functions, line):
- check_params = False
+ if Search(allowed_functions, line):
+ return
elif not Search(r'\S+\([^)]*$', line):
- # Don't see a whitelisted function on this line. Actually we
+ # Don't see an allowed function on this line. Actually we
# didn't see any function name on this line, so this is likely a
# multi-line parameter list. Try a bit harder to catch this case.
for i in xrange(2):
if (linenum > i and
- Search(whitelisted_functions, clean_lines.elided[linenum - i - 1])):
- check_params = False
- break
+ Search(allowed_functions, clean_lines.elided[linenum - i - 1])):
+ return
- if check_params:
- decls = ReplaceAll(r'{[^}]*}', ' ', line) # exclude function body
- for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls):
- if not Match(_RE_PATTERN_CONST_REF_PARAM, parameter):
- error(filename, linenum, 'runtime/references', 2,
- 'Is this a non-const reference? '
- 'If so, make const or use a pointer: ' +
- ReplaceAll(' *<', '<', parameter))
+ decls = ReplaceAll(r'{[^}]*}', ' ', line) # exclude function body
+ for parameter in re.findall(_RE_PATTERN_REF_PARAM, decls):
+ if (not Match(_RE_PATTERN_CONST_REF_PARAM, parameter) and
+ not Match(_RE_PATTERN_REF_STREAM_PARAM, parameter)):
+ error(filename, linenum, 'runtime/references', 2,
+ 'Is this a non-const reference? '
+ 'If so, make const or use a pointer: ' +
+ ReplaceAll(' *<', '<', parameter))
-def CheckCStyleCast(filename, linenum, line, raw_line, cast_type, pattern,
- error):
+def CheckCasts(filename, clean_lines, linenum, error):
+ """Various cast related checks.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ # Check to see if they're using an conversion function cast.
+ # I just try to capture the most common basic types, though there are more.
+ # Parameterless conversion functions, such as bool(), are allowed as they are
+ # probably a member operator declaration or default constructor.
+ match = Search(
+ r'(\bnew\s+(?:const\s+)?|\S<\s*(?:const\s+)?)?\b'
+ r'(int|float|double|bool|char|int32|uint32|int64|uint64)'
+ r'(\([^)].*)', line)
+ expecting_function = ExpectingFunctionArgs(clean_lines, linenum)
+ if match and not expecting_function:
+ matched_type = match.group(2)
+
+ # matched_new_or_template is used to silence two false positives:
+ # - New operators
+ # - Template arguments with function types
+ #
+ # For template arguments, we match on types immediately following
+ # an opening bracket without any spaces. This is a fast way to
+ # silence the common case where the function type is the first
+ # template argument. False negative with less-than comparison is
+ # avoided because those operators are usually followed by a space.
+ #
+ # function<double(double)> // bracket + no space = false positive
+ # value < double(42) // bracket + space = true positive
+ matched_new_or_template = match.group(1)
+
+ # Avoid arrays by looking for brackets that come after the closing
+ # parenthesis.
+ if Match(r'\([^()]+\)\s*\[', match.group(3)):
+ return
+
+ # Other things to ignore:
+ # - Function pointers
+ # - Casts to pointer types
+ # - Placement new
+ # - Alias declarations
+ matched_funcptr = match.group(3)
+ if (matched_new_or_template is None and
+ not (matched_funcptr and
+ (Match(r'\((?:[^() ]+::\s*\*\s*)?[^() ]+\)\s*\(',
+ matched_funcptr) or
+ matched_funcptr.startswith('(*)'))) and
+ not Match(r'\s*using\s+\S+\s*=\s*' + matched_type, line) and
+ not Search(r'new\(\S+\)\s*' + matched_type, line)):
+ error(filename, linenum, 'readability/casting', 4,
+ 'Using deprecated casting style. '
+ 'Use static_cast<%s>(...) instead' %
+ matched_type)
+
+ if not expecting_function:
+ CheckCStyleCast(filename, clean_lines, linenum, 'static_cast',
+ r'\((int|float|double|bool|char|u?int(16|32|64))\)', error)
+
+ # This doesn't catch all cases. Consider (const char * const)"hello".
+ #
+ # (char *) "foo" should always be a const_cast (reinterpret_cast won't
+ # compile).
+ if CheckCStyleCast(filename, clean_lines, linenum, 'const_cast',
+ r'\((char\s?\*+\s?)\)\s*"', error):
+ pass
+ else:
+ # Check pointer casts for other than string constants
+ CheckCStyleCast(filename, clean_lines, linenum, 'reinterpret_cast',
+ r'\((\w+\s?\*+\s?)\)', error)
+
+ # In addition, we look for people taking the address of a cast. This
+ # is dangerous -- casts can assign to temporaries, so the pointer doesn't
+ # point where you think.
+ #
+ # Some non-identifier character is required before the '&' for the
+ # expression to be recognized as a cast. These are casts:
+ # expression = &static_cast<int*>(temporary());
+ # function(&(int*)(temporary()));
+ #
+ # This is not a cast:
+ # reference_type&(int* function_param);
+ match = Search(
+ r'(?:[^\w]&\(([^)*][^)]*)\)[\w(])|'
+ r'(?:[^\w]&(static|dynamic|down|reinterpret)_cast\b)', line)
+ if match:
+ # Try a better error message when the & is bound to something
+ # dereferenced by the casted pointer, as opposed to the casted
+ # pointer itself.
+ parenthesis_error = False
+ match = Match(r'^(.*&(?:static|dynamic|down|reinterpret)_cast\b)<', line)
+ if match:
+ _, y1, x1 = CloseExpression(clean_lines, linenum, len(match.group(1)))
+ if x1 >= 0 and clean_lines.elided[y1][x1] == '(':
+ _, y2, x2 = CloseExpression(clean_lines, y1, x1)
+ if x2 >= 0:
+ extended_line = clean_lines.elided[y2][x2:]
+ if y2 < clean_lines.NumLines() - 1:
+ extended_line += clean_lines.elided[y2 + 1]
+ if Match(r'\s*(?:->|\[)', extended_line):
+ parenthesis_error = True
+
+ if parenthesis_error:
+ error(filename, linenum, 'readability/casting', 4,
+ ('Are you taking an address of something dereferenced '
+ 'from a cast? Wrapping the dereferenced expression in '
+ 'parentheses will make the binding more obvious'))
+ else:
+ error(filename, linenum, 'runtime/casting', 4,
+ ('Are you taking an address of a cast? '
+ 'This is dangerous: could be a temp var. '
+ 'Take the address before doing the cast, rather than after'))
+
+
+def CheckCStyleCast(filename, clean_lines, linenum, cast_type, pattern, error):
"""Checks for a C-style cast by looking for the pattern.
Args:
filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
linenum: The number of the line to check.
- line: The line of code to check.
- raw_line: The raw line of code to check, with comments.
cast_type: The string for the C++ cast to recommend. This is either
reinterpret_cast, static_cast, or const_cast, depending.
pattern: The regular expression used to find C-style casts.
True if an error was emitted.
False otherwise.
"""
+ line = clean_lines.elided[linenum]
match = Search(pattern, line)
if not match:
return False
- # e.g., sizeof(int)
- sizeof_match = Match(r'.*sizeof\s*$', line[0:match.start(1) - 1])
- if sizeof_match:
- error(filename, linenum, 'runtime/sizeof', 1,
- 'Using sizeof(type). Use sizeof(varname) instead if possible')
- return True
+ # Exclude lines with keywords that tend to look like casts
+ context = line[0:match.start(1) - 1]
+ if Match(r'.*\b(?:sizeof|alignof|alignas|[_A-Z][_A-Z0-9]*)\s*$', context):
+ return False
+
+ # Try expanding current context to see if we one level of
+ # parentheses inside a macro.
+ if linenum > 0:
+ for i in xrange(linenum - 1, max(0, linenum - 5), -1):
+ context = clean_lines.elided[i] + context
+ if Match(r'.*\b[_A-Z][_A-Z0-9]*\s*\((?:\([^()]*\)|[^()])*$', context):
+ return False
# operator++(int) and operator--(int)
- if (line[0:match.start(1) - 1].endswith(' operator++') or
- line[0:match.start(1) - 1].endswith(' operator--')):
+ if context.endswith(' operator++') or context.endswith(' operator--'):
return False
- # A single unnamed argument for a function tends to look like old
- # style cast. If we see those, don't issue warnings for deprecated
- # casts, instead issue warnings for unnamed arguments where
- # appropriate.
- #
- # These are things that we want warnings for, since the style guide
- # explicitly require all parameters to be named:
- # Function(int);
- # Function(int) {
- # ConstMember(int) const;
- # ConstMember(int) const {
- # ExceptionMember(int) throw (...);
- # ExceptionMember(int) throw (...) {
- # PureVirtual(int) = 0;
- #
- # These are functions of some sort, where the compiler would be fine
- # if they had named parameters, but people often omit those
- # identifiers to reduce clutter:
- # (FunctionPointer)(int);
- # (FunctionPointer)(int) = value;
- # Function((function_pointer_arg)(int))
- # <TemplateArgument(int)>;
- # <(FunctionPointerTemplateArgument)(int)>;
+ # A single unnamed argument for a function tends to look like old style cast.
+ # If we see those, don't issue warnings for deprecated casts.
remainder = line[match.end(0):]
- if Match(r'^\s*(?:;|const\b|throw\b|=|>|\{|\))', remainder):
- # Looks like an unnamed parameter.
-
- # Don't warn on any kind of template arguments.
- if Match(r'^\s*>', remainder):
- return False
-
- # Don't warn on assignments to function pointers, but keep warnings for
- # unnamed parameters to pure virtual functions. Note that this pattern
- # will also pass on assignments of "0" to function pointers, but the
- # preferred values for those would be "nullptr" or "NULL".
- matched_zero = Match(r'^\s=\s*(\S+)\s*;', remainder)
- if matched_zero and matched_zero.group(1) != '0':
- return False
-
- # Don't warn on function pointer declarations. For this we need
- # to check what came before the "(type)" string.
- if Match(r'.*\)\s*$', line[0:match.start(0)]):
- return False
-
- # Don't warn if the parameter is named with block comments, e.g.:
- # Function(int /*unused_param*/);
- if '/*' in raw_line:
- return False
-
- # Passed all filters, issue warning here.
- error(filename, linenum, 'readability/function', 3,
- 'All parameters should be named in a function')
- return True
+ if Match(r'^\s*(?:;|const\b|throw\b|final\b|override\b|[=>{),]|->)',
+ remainder):
+ return False
# At this point, all that should be left is actual casts.
error(filename, linenum, 'readability/casting', 4,
return True
+def ExpectingFunctionArgs(clean_lines, linenum):
+ """Checks whether where function type arguments are expected.
+
+ Args:
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+
+ Returns:
+ True if the line at 'linenum' is inside something that expects arguments
+ of function types.
+ """
+ line = clean_lines.elided[linenum]
+ return (Match(r'^\s*MOCK_(CONST_)?METHOD\d+(_T)?\(', line) or
+ (linenum >= 2 and
+ (Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\((?:\S+,)?\s*$',
+ clean_lines.elided[linenum - 1]) or
+ Match(r'^\s*MOCK_(?:CONST_)?METHOD\d+(?:_T)?\(\s*$',
+ clean_lines.elided[linenum - 2]) or
+ Search(r'\bstd::m?function\s*\<\s*$',
+ clean_lines.elided[linenum - 1]))))
+
+
_HEADERS_CONTAINING_TEMPLATES = (
('<deque>', ('deque',)),
('<functional>', ('unary_function', 'binary_function',
('<limits>', ('numeric_limits',)),
('<list>', ('list',)),
('<map>', ('map', 'multimap',)),
- ('<memory>', ('allocator',)),
+ ('<memory>', ('allocator', 'make_shared', 'make_unique', 'shared_ptr',
+ 'unique_ptr', 'weak_ptr')),
('<queue>', ('queue', 'priority_queue',)),
('<set>', ('set', 'multiset',)),
('<stack>', ('stack',)),
('<string>', ('char_traits', 'basic_string',)),
+ ('<tuple>', ('tuple',)),
+ ('<unordered_map>', ('unordered_map', 'unordered_multimap')),
+ ('<unordered_set>', ('unordered_set', 'unordered_multiset')),
('<utility>', ('pair',)),
('<vector>', ('vector',)),
('<slist>', ('slist',)),
)
-_RE_PATTERN_STRING = re.compile(r'\bstring\b')
+_HEADERS_MAYBE_TEMPLATES = (
+ ('<algorithm>', ('copy', 'max', 'min', 'min_element', 'sort',
+ 'transform',
+ )),
+ ('<utility>', ('forward', 'make_pair', 'move', 'swap')),
+ )
-_re_pattern_algorithm_header = []
-for _template in ('copy', 'max', 'min', 'min_element', 'sort', 'swap',
- 'transform'):
- # Match max<type>(..., ...), max(..., ...), but not foo->max, foo.max or
- # type::max().
- _re_pattern_algorithm_header.append(
- (re.compile(r'[^>.]\b' + _template + r'(<.*?>)?\([^\)]'),
- _template,
- '<algorithm>'))
+_RE_PATTERN_STRING = re.compile(r'\bstring\b')
+_re_pattern_headers_maybe_templates = []
+for _header, _templates in _HEADERS_MAYBE_TEMPLATES:
+ for _template in _templates:
+ # Match max<type>(..., ...), max(..., ...), but not foo->max, foo.max or
+ # type::max().
+ _re_pattern_headers_maybe_templates.append(
+ (re.compile(r'[^>.]\b' + _template + r'(<.*?>)?\([^\)]'),
+ _template,
+ _header))
+
+# Other scripts may reach in and modify this pattern.
_re_pattern_templates = []
for _header, _templates in _HEADERS_CONTAINING_TEMPLATES:
for _template in _templates:
string: the additional prefix needed to open the header file.
"""
- if not filename_cc.endswith('.cc'):
+ fileinfo = FileInfo(filename_cc)
+ if not fileinfo.IsSource():
return (False, '')
- filename_cc = filename_cc[:-len('.cc')]
- if filename_cc.endswith('_unittest'):
- filename_cc = filename_cc[:-len('_unittest')]
- elif filename_cc.endswith('_test'):
- filename_cc = filename_cc[:-len('_test')]
+ filename_cc = filename_cc[:-len(fileinfo.Extension())]
+ matched_test_suffix = Search(_TEST_FILE_SUFFIX, fileinfo.BaseName())
+ if matched_test_suffix:
+ filename_cc = filename_cc[:-len(matched_test_suffix.group(1))]
filename_cc = filename_cc.replace('/public/', '/')
filename_cc = filename_cc.replace('/internal/', '/')
return files_belong_to_same_module, common_path
-def UpdateIncludeState(filename, include_state, io=codecs):
- """Fill up the include_state with new includes found from the file.
+def UpdateIncludeState(filename, include_dict, io=codecs):
+ """Fill up the include_dict with new includes found from the file.
Args:
filename: the name of the header to read.
- include_state: an _IncludeState instance in which the headers are inserted.
+ include_dict: a dictionary in which the headers are inserted.
io: The io factory to use to read the file. Provided for testability.
Returns:
- True if a header was succesfully added. False otherwise.
+ True if a header was successfully added. False otherwise.
"""
headerfile = None
try:
match = _RE_PATTERN_INCLUDE.search(clean_line)
if match:
include = match.group(2)
- # The value formatting is cute, but not really used right now.
- # What matters here is that the key is in include_state.
- include_state.setdefault(include, '%s:%d' % (filename, linenum))
+ include_dict.setdefault(include, linenum)
return True
if prefix.endswith('std::') or not prefix.endswith('::'):
required['<string>'] = (linenum, 'string')
- for pattern, template, header in _re_pattern_algorithm_header:
+ for pattern, template, header in _re_pattern_headers_maybe_templates:
if pattern.search(line):
required[header] = (linenum, template)
continue
for pattern, template, header in _re_pattern_templates:
- if pattern.search(line):
- required[header] = (linenum, template)
+ matched = pattern.search(line)
+ if matched:
+ # Don't warn about IWYU in non-STL namespaces:
+ # (We check only the first match per line; good enough.)
+ prefix = line[:matched.start()]
+ if prefix.endswith('std::') or not prefix.endswith('::'):
+ required[header] = (linenum, template)
# The policy is that if you #include something in foo.h you don't need to
# include it again in foo.cc. Here, we will look at possible includes.
- # Let's copy the include_state so it is only messed up within this function.
- include_state = include_state.copy()
+ # Let's flatten the include_state include_list and copy it into a dictionary.
+ include_dict = dict([item for sublist in include_state.include_list
+ for item in sublist])
- # Did we find the header for this file (if any) and succesfully load it?
+ # Did we find the header for this file (if any) and successfully load it?
header_found = False
# Use the absolute path so that matching works properly.
# instead of 'foo_flymake.h'
abs_filename = re.sub(r'_flymake\.cc$', '.cc', abs_filename)
- # include_state is modified during iteration, so we iterate over a copy of
+ # include_dict is modified during iteration, so we iterate over a copy of
# the keys.
- header_keys = include_state.keys()
+ header_keys = include_dict.keys()
for header in header_keys:
(same_module, common_path) = FilesBelongToSameModule(abs_filename, header)
fullpath = common_path + header
- if same_module and UpdateIncludeState(fullpath, include_state, io):
+ if same_module and UpdateIncludeState(fullpath, include_dict, io):
header_found = True
# If we can't find the header file for a .cc, assume it's because we don't
# All the lines have been processed, report the errors found.
for required_header_unstripped in required:
template = required[required_header_unstripped][1]
- if required_header_unstripped.strip('<>"') not in include_state:
+ if required_header_unstripped.strip('<>"') not in include_dict:
error(filename, required[required_header_unstripped][0],
'build/include_what_you_use', 4,
'Add #include ' + required_header_unstripped + ' for ' + template)
def CheckMakePairUsesDeduction(filename, clean_lines, linenum, error):
"""Check that make_pair's template arguments are deduced.
- G++ 4.6 in C++0x mode fails badly if make_pair's template arguments are
+ G++ 4.6 in C++11 mode fails badly if make_pair's template arguments are
specified explicitly, and such use isn't intended in any case.
Args:
' OR use pair directly OR if appropriate, construct a pair directly')
+def CheckRedundantVirtual(filename, clean_lines, linenum, error):
+ """Check if line contains a redundant "virtual" function-specifier.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ # Look for "virtual" on current line.
+ line = clean_lines.elided[linenum]
+ virtual = Match(r'^(.*)(\bvirtual\b)(.*)$', line)
+ if not virtual: return
+
+ # Ignore "virtual" keywords that are near access-specifiers. These
+ # are only used in class base-specifier and do not apply to member
+ # functions.
+ if (Search(r'\b(public|protected|private)\s+$', virtual.group(1)) or
+ Match(r'^\s+(public|protected|private)\b', virtual.group(3))):
+ return
+
+ # Ignore the "virtual" keyword from virtual base classes. Usually
+ # there is a column on the same line in these cases (virtual base
+ # classes are rare in google3 because multiple inheritance is rare).
+ if Match(r'^.*[^:]:[^:].*$', line): return
+
+ # Look for the next opening parenthesis. This is the start of the
+ # parameter list (possibly on the next line shortly after virtual).
+ # TODO(unknown): doesn't work if there are virtual functions with
+ # decltype() or other things that use parentheses, but csearch suggests
+ # that this is rare.
+ end_col = -1
+ end_line = -1
+ start_col = len(virtual.group(2))
+ for start_line in xrange(linenum, min(linenum + 3, clean_lines.NumLines())):
+ line = clean_lines.elided[start_line][start_col:]
+ parameter_list = Match(r'^([^(]*)\(', line)
+ if parameter_list:
+ # Match parentheses to find the end of the parameter list
+ (_, end_line, end_col) = CloseExpression(
+ clean_lines, start_line, start_col + len(parameter_list.group(1)))
+ break
+ start_col = 0
+
+ if end_col < 0:
+ return # Couldn't find end of parameter list, give up
+
+ # Look for "override" or "final" after the parameter list
+ # (possibly on the next few lines).
+ for i in xrange(end_line, min(end_line + 3, clean_lines.NumLines())):
+ line = clean_lines.elided[i][end_col:]
+ match = Search(r'\b(override|final)\b', line)
+ if match:
+ error(filename, linenum, 'readability/inheritance', 4,
+ ('"virtual" is redundant since function is '
+ 'already declared as "%s"' % match.group(1)))
+
+ # Set end_col to check whole lines after we are done with the
+ # first line.
+ end_col = 0
+ if Search(r'[^\w]\s*$', line):
+ break
+
+
+def CheckRedundantOverrideOrFinal(filename, clean_lines, linenum, error):
+ """Check if line contains a redundant "override" or "final" virt-specifier.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ # Look for closing parenthesis nearby. We need one to confirm where
+ # the declarator ends and where the virt-specifier starts to avoid
+ # false positives.
+ line = clean_lines.elided[linenum]
+ declarator_end = line.rfind(')')
+ if declarator_end >= 0:
+ fragment = line[declarator_end:]
+ else:
+ if linenum > 1 and clean_lines.elided[linenum - 1].rfind(')') >= 0:
+ fragment = line
+ else:
+ return
+
+ # Check that at most one of "override" or "final" is present, not both
+ if Search(r'\boverride\b', fragment) and Search(r'\bfinal\b', fragment):
+ error(filename, linenum, 'readability/inheritance', 4,
+ ('"override" is redundant since function is '
+ 'already declared as "final"'))
+
+
+
+
+# Returns true if we are at a new block, and it is directly
+# inside of a namespace.
+def IsBlockInNameSpace(nesting_state, is_forward_declaration):
+ """Checks that the new block is directly in a namespace.
+
+ Args:
+ nesting_state: The _NestingState object that contains info about our state.
+ is_forward_declaration: If the class is a forward declared class.
+ Returns:
+ Whether or not the new block is directly in a namespace.
+ """
+ if is_forward_declaration:
+ if len(nesting_state.stack) >= 1 and (
+ isinstance(nesting_state.stack[-1], _NamespaceInfo)):
+ return True
+ else:
+ return False
+
+ return (len(nesting_state.stack) > 1 and
+ nesting_state.stack[-1].check_namespace_indentation and
+ isinstance(nesting_state.stack[-2], _NamespaceInfo))
+
+
+def ShouldCheckNamespaceIndentation(nesting_state, is_namespace_indent_item,
+ raw_lines_no_comments, linenum):
+ """This method determines if we should apply our namespace indentation check.
+
+ Args:
+ nesting_state: The current nesting state.
+ is_namespace_indent_item: If we just put a new class on the stack, True.
+ If the top of the stack is not a class, or we did not recently
+ add the class, False.
+ raw_lines_no_comments: The lines without the comments.
+ linenum: The current line number we are processing.
+
+ Returns:
+ True if we should apply our namespace indentation check. Currently, it
+ only works for classes and namespaces inside of a namespace.
+ """
+
+ is_forward_declaration = IsForwardClassDeclaration(raw_lines_no_comments,
+ linenum)
+
+ if not (is_namespace_indent_item or is_forward_declaration):
+ return False
+
+ # If we are in a macro, we do not want to check the namespace indentation.
+ if IsMacroDefinition(raw_lines_no_comments, linenum):
+ return False
+
+ return IsBlockInNameSpace(nesting_state, is_forward_declaration)
+
+
+# Call this method if the line is directly inside of a namespace.
+# If the line above is blank (excluding comments) or the start of
+# an inner namespace, it cannot be indented.
+def CheckItemIndentationInNamespace(filename, raw_lines_no_comments, linenum,
+ error):
+ line = raw_lines_no_comments[linenum]
+ if Match(r'^\s+', line):
+ error(filename, linenum, 'runtime/indentation_namespace', 4,
+ 'Do not indent within a namespace')
+
+
def ProcessLine(filename, file_extension, clean_lines, line,
include_state, function_state, nesting_state, error,
extra_check_functions=[]):
line: Number of line being processed.
include_state: An _IncludeState instance in which the headers are inserted.
function_state: A _FunctionState instance which counts function lines, etc.
- nesting_state: A _NestingState instance which maintains information about
+ nesting_state: A NestingState instance which maintains information about
the current stack of nested blocks being parsed.
error: A callable to which errors are reported, which takes 4 arguments:
filename, line number, error level, and message
raw_lines = clean_lines.raw_lines
ParseNolintSuppressions(filename, raw_lines[line], line, error)
nesting_state.Update(filename, clean_lines, line, error)
- if nesting_state.stack and nesting_state.stack[-1].inline_asm != _NO_ASM:
- return
+ CheckForNamespaceIndentation(filename, nesting_state, clean_lines, line,
+ error)
+ if nesting_state.InAsmBlock(): return
CheckForFunctionLengths(filename, clean_lines, line, function_state, error)
CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error)
CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error)
CheckPosixThreading(filename, clean_lines, line, error)
CheckInvalidIncrement(filename, clean_lines, line, error)
CheckMakePairUsesDeduction(filename, clean_lines, line, error)
+ CheckRedundantVirtual(filename, clean_lines, line, error)
+ CheckRedundantOverrideOrFinal(filename, clean_lines, line, error)
for check_fn in extra_check_functions:
check_fn(filename, clean_lines, line, error)
+def FlagCxx11Features(filename, clean_lines, linenum, error):
+ """Flag those c++11 features that we only allow in certain places.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line)
+
+ # Flag unapproved C++ TR1 headers.
+ if include and include.group(1).startswith('tr1/'):
+ error(filename, linenum, 'build/c++tr1', 5,
+ ('C++ TR1 headers such as <%s> are unapproved.') % include.group(1))
+
+ # Flag unapproved C++11 headers.
+ if include and include.group(1) in ('cfenv',
+ 'condition_variable',
+ 'fenv.h',
+ 'future',
+ 'mutex',
+ 'thread',
+ 'chrono',
+ 'ratio',
+ 'regex',
+ 'system_error',
+ ):
+ error(filename, linenum, 'build/c++11', 5,
+ ('<%s> is an unapproved C++11 header.') % include.group(1))
+
+ # The only place where we need to worry about C++11 keywords and library
+ # features in preprocessor directives is in macro definitions.
+ if Match(r'\s*#', line) and not Match(r'\s*#\s*define\b', line): return
+
+ # These are classes and free functions. The classes are always
+ # mentioned as std::*, but we only catch the free functions if
+ # they're not found by ADL. They're alphabetical by header.
+ for top_name in (
+ # type_traits
+ 'alignment_of',
+ 'aligned_union',
+ ):
+ if Search(r'\bstd::%s\b' % top_name, line):
+ error(filename, linenum, 'build/c++11', 5,
+ ('std::%s is an unapproved C++11 class or function. Send c-style '
+ 'an example of where it would make your code more readable, and '
+ 'they may let you use it.') % top_name)
+
+
+def FlagCxx14Features(filename, clean_lines, linenum, error):
+ """Flag those C++14 features that we restrict.
+
+ Args:
+ filename: The name of the current file.
+ clean_lines: A CleansedLines instance containing the file.
+ linenum: The number of the line to check.
+ error: The function to call with any errors found.
+ """
+ line = clean_lines.elided[linenum]
+
+ include = Match(r'\s*#\s*include\s+[<"]([^<"]+)[">]', line)
+
+ # Flag unapproved C++14 headers.
+ if include and include.group(1) in ('scoped_allocator', 'shared_mutex'):
+ error(filename, linenum, 'build/c++14', 5,
+ ('<%s> is an unapproved C++14 header.') % include.group(1))
+
+
def ProcessFileData(filename, file_extension, lines, error,
extra_check_functions=[]):
"""Performs lint checks and reports any errors to the given error function.
include_state = _IncludeState()
function_state = _FunctionState()
- nesting_state = _NestingState()
+ nesting_state = NestingState()
ResetNolintSuppressions()
CheckForCopyright(filename, lines, error)
-
- if file_extension == 'h':
- CheckForHeaderGuard(filename, lines, error)
-
+ ProcessGlobalSuppresions(lines)
RemoveMultiLineComments(filename, lines, error)
clean_lines = CleansedLines(lines)
+
+ if IsHeaderExtension(file_extension):
+ CheckForHeaderGuard(filename, clean_lines, error)
+
for line in xrange(clean_lines.NumLines()):
ProcessLine(filename, file_extension, clean_lines, line,
include_state, function_state, nesting_state, error,
extra_check_functions)
+ FlagCxx11Features(filename, clean_lines, line, error)
nesting_state.CheckCompletedBlocks(filename, error)
CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error)
+ # Check that the .cc file has included its header if it exists.
+ if _IsSourceExtension(file_extension):
+ CheckHeaderFileIncluded(filename, include_state, error)
+
# We check here rather than inside ProcessLine so that we see raw
# lines rather than "cleaned" lines.
CheckForBadCharacters(filename, lines, error)
CheckForNewlineAtEOF(filename, lines, error)
+def ProcessConfigOverrides(filename):
+ """ Loads the configuration files and processes the config overrides.
+
+ Args:
+ filename: The name of the file being processed by the linter.
+
+ Returns:
+ False if the current |filename| should not be processed further.
+ """
+
+ abs_filename = os.path.abspath(filename)
+ cfg_filters = []
+ keep_looking = True
+ while keep_looking:
+ abs_path, base_name = os.path.split(abs_filename)
+ if not base_name:
+ break # Reached the root directory.
+
+ cfg_file = os.path.join(abs_path, "CPPLINT.cfg")
+ abs_filename = abs_path
+ if not os.path.isfile(cfg_file):
+ continue
+
+ try:
+ with open(cfg_file) as file_handle:
+ for line in file_handle:
+ line, _, _ = line.partition('#') # Remove comments.
+ if not line.strip():
+ continue
+
+ name, _, val = line.partition('=')
+ name = name.strip()
+ val = val.strip()
+ if name == 'set noparent':
+ keep_looking = False
+ elif name == 'filter':
+ cfg_filters.append(val)
+ elif name == 'exclude_files':
+ # When matching exclude_files pattern, use the base_name of
+ # the current file name or the directory name we are processing.
+ # For example, if we are checking for lint errors in /foo/bar/baz.cc
+ # and we found the .cfg file at /foo/CPPLINT.cfg, then the config
+ # file's "exclude_files" filter is meant to be checked against "bar"
+ # and not "baz" nor "bar/baz.cc".
+ if base_name:
+ pattern = re.compile(val)
+ if pattern.match(base_name):
+ if _cpplint_state.quiet:
+ # Suppress "Ignoring file" warning when using --quiet.
+ return False
+ sys.stderr.write('Ignoring "%s": file excluded by "%s". '
+ 'File path component "%s" matches '
+ 'pattern "%s"\n' %
+ (filename, cfg_file, base_name, val))
+ return False
+ elif name == 'linelength':
+ global _line_length
+ try:
+ _line_length = int(val)
+ except ValueError:
+ sys.stderr.write('Line length must be numeric.')
+ elif name == 'root':
+ global _root
+ # root directories are specified relative to CPPLINT.cfg dir.
+ _root = os.path.join(os.path.dirname(cfg_file), val)
+ elif name == 'headers':
+ ProcessHppHeadersOption(val)
+ else:
+ sys.stderr.write(
+ 'Invalid configuration option (%s) in file %s\n' %
+ (name, cfg_file))
+
+ except IOError:
+ sys.stderr.write(
+ "Skipping config file '%s': Can't open for reading\n" % cfg_file)
+ keep_looking = False
+
+ # Apply all the accumulated filters in reverse order (top-level directory
+ # config options having the least priority).
+ for filter in reversed(cfg_filters):
+ _AddFilters(filter)
+
+ return True
+
+
def ProcessFile(filename, vlevel, extra_check_functions=[]):
"""Does google-lint on a single file.
"""
_SetVerboseLevel(vlevel)
+ _BackupFilters()
+ old_errors = _cpplint_state.error_count
+
+ if not ProcessConfigOverrides(filename):
+ _RestoreFilters()
+ return
+ lf_lines = []
+ crlf_lines = []
try:
# Support the UNIX convention of using "-" for stdin. Note that
# we are not opening the file with universal newline support
# contain trailing '\r' characters if we are reading a file that
# has CRLF endings.
# If after the split a trailing '\r' is present, it is removed
- # below. If it is not expected to be present (i.e. os.linesep !=
- # '\r\n' as in Windows), a warning is issued below if this file
- # is processed.
-
+ # below.
if filename == '-':
lines = codecs.StreamReaderWriter(sys.stdin,
codecs.getreader('utf8'),
else:
lines = codecs.open(filename, 'r', 'utf8', 'replace').read().split('\n')
- carriage_return_found = False
# Remove trailing '\r'.
- for linenum in range(len(lines)):
+ # The -1 accounts for the extra trailing blank line we get from split()
+ for linenum in range(len(lines) - 1):
if lines[linenum].endswith('\r'):
lines[linenum] = lines[linenum].rstrip('\r')
- carriage_return_found = True
+ crlf_lines.append(linenum + 1)
+ else:
+ lf_lines.append(linenum + 1)
except IOError:
sys.stderr.write(
"Skipping input '%s': Can't open for reading\n" % filename)
+ _RestoreFilters()
return
# Note, if no dot is found, this will give the entire filename as the ext.
else:
ProcessFileData(filename, file_extension, lines, Error,
extra_check_functions)
- if carriage_return_found and os.linesep != '\r\n':
- # Use 0 for linenum since outputting only one error for potentially
- # several lines.
- Error(filename, 0, 'whitespace/newline', 1,
- 'One or more unexpected \\r (^M) found;'
- 'better to use only a \\n')
- sys.stderr.write('Done processing %s\n' % filename)
+ # If end-of-line sequences are a mix of LF and CR-LF, issue
+ # warnings on the lines with CR.
+ #
+ # Don't issue any warnings if all lines are uniformly LF or CR-LF,
+ # since critique can handle these just fine, and the style guide
+ # doesn't dictate a particular end of line sequence.
+ #
+ # We can't depend on os.linesep to determine what the desired
+ # end-of-line sequence should be, since that will return the
+ # server-side end-of-line sequence.
+ if lf_lines and crlf_lines:
+ # Warn on every line with CR. An alternative approach might be to
+ # check whether the file is mostly CRLF or just LF, and warn on the
+ # minority, we bias toward LF here since most tools prefer LF.
+ for linenum in crlf_lines:
+ Error(filename, linenum, 'whitespace/newline', 1,
+ 'Unexpected \\r (^M) found; better to use only \\n')
+
+ # Suppress printing anything if --quiet was passed unless the error
+ # count has increased after processing this file.
+ if not _cpplint_state.quiet or old_errors != _cpplint_state.error_count:
+ sys.stdout.write('Done processing %s\n' % filename)
+ _RestoreFilters()
def PrintUsage(message):
'filter=',
'root=',
'linelength=',
- 'extensions='])
+ 'extensions=',
+ 'headers=',
+ 'quiet'])
except getopt.GetoptError:
PrintUsage('Invalid arguments.')
verbosity = _VerboseLevel()
output_format = _OutputFormat()
filters = ''
+ quiet = _Quiet()
counting_style = ''
for (opt, val) in opts:
if val not in ('emacs', 'vs7', 'eclipse'):
PrintUsage('The only allowed output formats are emacs, vs7 and eclipse.')
output_format = val
+ elif opt == '--quiet':
+ quiet = True
elif opt == '--verbose':
verbosity = int(val)
elif opt == '--filter':
try:
_valid_extensions = set(val.split(','))
except ValueError:
- PrintUsage('Extensions must be comma seperated list.')
+ PrintUsage('Extensions must be comma separated list.')
+ elif opt == '--headers':
+ ProcessHppHeadersOption(val)
if not filenames:
PrintUsage('No files were specified.')
_SetOutputFormat(output_format)
+ _SetQuiet(quiet)
_SetVerboseLevel(verbosity)
_SetFilters(filters)
_SetCountingStyle(counting_style)
_cpplint_state.ResetErrorCounts()
for filename in filenames:
ProcessFile(filename, _cpplint_state.verbose_level)
- _cpplint_state.PrintErrorCounts()
+ # If --quiet is passed, suppress printing error count unless there are errors.
+ if not _cpplint_state.quiet or _cpplint_state.error_count > 0:
+ _cpplint_state.PrintErrorCounts()
sys.exit(_cpplint_state.error_count > 0)
projects / libvpx / commitdiff