cpplint.py

    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'
      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'
          ' for more information.')

def CheckForNonConstReference(filename, clean_lines, linenum,
                              nesting_state, error):
  """Check for non-const references.

  Separate from CheckLanguage since it scans backwards from current
  line, instead of scanning forward.

  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.
  """
  # Do nothing if there is no '&' on current line.
  line = clean_lines.elided[linenum]
  if '&' not in line:
    return

  # Long type names may be broken across multiple lines, usually in one
  # of these forms:
  #   LongType
  #       ::LongTypeContinued &identifier
  #   LongType::
  #       LongTypeContinued &identifier
  #   LongType<
  #       ...>::LongTypeContinued &identifier
  #
  # If we detected a type split across two lines, join the previous
  # line to current line so that we can match const references
  # accordingly.
  #
  # Note that this only scans back one line, since scanning back
  # arbitrary number of lines would be expensive.  If you have a type
  # that spans more than 2 lines, please use a typedef.
  if linenum > 1:
    previous = None
    if Match(r'\s*::(?:[\w<>]|::)+\s*&\s*\S', line):
      # previous_line\n + ::current_line
      previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+[\w<>])\s*$',
                        clean_lines.elided[linenum - 1])
    elif Match(r'\s*[a-zA-Z_]([\w<>]|::)+\s*&\s*\S', line):
      # previous_line::\n + current_line
      previous = Search(r'\b((?:const\s*)?(?:[\w<>]|::)+::)\s*$',
                        clean_lines.elided[linenum - 1])
    if previous:
      line = previous.group(1) + line.lstrip()
    else:
      # Check for templated parameter that is split across multiple lines
      endpos = line.rfind('>')
      if endpos > -1:
        (_, startline, startpos) = ReverseCloseExpression(
            clean_lines, linenum, endpos)
        if startpos > -1 and startline < linenum:
          # Found the matching < on an earlier line, collect all
          # pieces up to current line.
          line = ''
          for i in xrange(startline, linenum + 1):
            line += clean_lines.elided[i].strip()

  # Check for non-const references in function parameters.  A single '&' may
  # found in the following places:
  #   inside expression: binary & for bitwise AND
  #   inside expression: unary & for taking the address of something
  #   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
  # 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:+>)?|'
                           r'operator\s*[<>][<>]|'
                           r'static_assert|COMPILE_ASSERT'
                           r')\s*\(')
  if Search(whitelisted_functions, line):
    check_params = False
  elif not Search(r'\S+\([^)]*$', line):
    # Don't see a whitelisted 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

  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))


def CheckCStyleCast(filename, linenum, line, raw_line, cast_type, pattern,
                    error):
  """Checks for a C-style cast by looking for the pattern.

  Args:
    filename: The name of the current 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.
    error: The function to call with any errors found.

  Returns:
    True if an error was emitted.
    False otherwise.
  """
  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

  # operator++(int) and operator--(int)
  if (line[0:match.start(1) - 1].endswith(' operator++') or
      line[0:match.start(1) - 1].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)>;
  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

  # At this point, all that should be left is actual casts.
  error(filename, linenum, 'readability/casting', 4,
        'Using C-style cast.  Use %s<%s>(...) instead' %
        (cast_type, match.group(1)))

  return True


_HEADERS_CONTAINING_TEMPLATES = (
    ('<deque>', ('deque',)),
    ('<functional>', ('unary_function', 'binary_function',
                      'plus', 'minus', 'multiplies', 'divides', 'modulus',
                      'negate',
                      'equal_to', 'not_equal_to', 'greater', 'less',
                      'greater_equal', 'less_equal',
                      'logical_and', 'logical_or', 'logical_not',
                      'unary_negate', 'not1', 'binary_negate', 'not2',
                      'bind1st', 'bind2nd',
                      'pointer_to_unary_function',
                      'pointer_to_binary_function',
                      'ptr_fun',
                      'mem_fun_t', 'mem_fun', 'mem_fun1_t', 'mem_fun1_ref_t',
                      'mem_fun_ref_t',
                      'const_mem_fun_t', 'const_mem_fun1_t',
                      'const_mem_fun_ref_t', 'const_mem_fun1_ref_t',
                      'mem_fun_ref',
                     )),
    ('<limits>', ('numeric_limits',)),
    ('<list>', ('list',)),
    ('<map>', ('map', 'multimap',)),
    ('<memory>', ('allocator',)),
    ('<queue>', ('queue', 'priority_queue',)),
    ('<set>', ('set', 'multiset',)),
    ('<stack>', ('stack',)),
    ('<string>', ('char_traits', 'basic_string',)),
    ('<utility>', ('pair',)),
    ('<vector>', ('vector',)),

    # gcc extensions.
    # Note: std::hash is their hash, ::hash is our hash
    ('<hash_map>', ('hash_map', 'hash_multimap',)),
    ('<hash_set>', ('hash_set', 'hash_multiset',)),
    ('<slist>', ('slist',)),
    )

_RE_PATTERN_STRING = re.compile(r'\bstring\b')

_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_templates = []
for _header, _templates in _HEADERS_CONTAINING_TEMPLATES:
  for _template in _templates:
    _re_pattern_templates.append(
        (re.compile(r'(\<|\b)' + _template + r'\s*\<'),
         _template + '<>',
         _header))


def FilesBelongToSameModule(filename_cc, filename_h):
  """Check if these two filenames belong to the same module.

  The concept of a 'module' here is a as follows:
  foo.h, foo-inl.h, foo.cc, foo_test.cc and foo_unittest.cc belong to the
  same 'module' if they are in the same directory.
  some/path/public/xyzzy and some/path/internal/xyzzy are also considered
  to belong to the same module here.

  If the filename_cc contains a longer path than the filename_h, for example,
  '/absolute/path/to/base/sysinfo.cc', and this file would include
  'base/sysinfo.h', this function also produces the prefix needed to open the
  header. This is used by the caller of this function to more robustly open the
  header file. We don't have access to the real include paths in this context,
  so we need this guesswork here.

  Known bugs: tools/base/bar.cc and base/bar.h belong to the same module
  according to this implementation. Because of this, this function gives
  some false positives. This should be sufficiently rare in practice.

  Args:
    filename_cc: is the path for the .cc file
    filename_h: is the path for the header path

  Returns:
    Tuple with a bool and a string:
    bool: True if filename_cc and filename_h belong to the same module.
    string: the additional prefix needed to open the header file.
  """

  if not filename_cc.endswith('.cc'):
    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.replace('/public/', '/')
  filename_cc = filename_cc.replace('/internal/', '/')

  if not filename_h.endswith('.h'):
    return (False, '')
  filename_h = filename_h[:-len('.h')]
  if filename_h.endswith('-inl'):
    filename_h = filename_h[:-len('-inl')]
  filename_h = filename_h.replace('/public/', '/')
  filename_h = filename_h.replace('/internal/', '/')

  files_belong_to_same_module = filename_cc.endswith(filename_h)
  common_path = ''
  if files_belong_to_same_module:
    common_path = filename_cc[:-len(filename_h)]
  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.

  Args:
    filename: the name of the header to read.
    include_state: an _IncludeState instance 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.
  """
  headerfile = None
  try:
    headerfile = io.open(filename, 'r', 'utf8', 'replace')
  except IOError:
    return False
  linenum = 0
  for line in headerfile:
    linenum += 1
    clean_line = CleanseComments(line)
    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))
  return True


def CheckForIncludeWhatYouUse(filename, clean_lines, include_state, error,
                              io=codecs):
  """Reports for missing stl includes.

  This function will output warnings to make sure you are including the headers
  necessary for the stl containers and functions that you use. We only give one
  reason to include a header. For example, if you use both equal_to<> and
  less<> in a .h file, only one (the latter in the file) of these will be
  reported as a reason to include the <functional>.

  Args:
    filename: The name of the current file.
    clean_lines: A CleansedLines instance containing the file.
    include_state: An _IncludeState instance.
    error: The function to call with any errors found.
    io: The IO factory to use to read the header file. Provided for unittest
        injection.
  """
  required = {}  # A map of header name to linenumber and the template entity.
                 # Example of required: { '<functional>': (1219, 'less<>') }

  for linenum in xrange(clean_lines.NumLines()):
    line = clean_lines.elided[linenum]
    if not line or line[0] == '#':
      continue

    # String is special -- it is a non-templatized type in STL.
    matched = _RE_PATTERN_STRING.search(line)
    if matched:
      # Don't warn about strings 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['<string>'] = (linenum, 'string')

    for pattern, template, header in _re_pattern_algorithm_header:
      if pattern.search(line):
        required[header] = (linenum, template)

    # The following function is just a speed up, no semantics are changed.
    if not '<' in line:  # Reduces the cpu time usage by skipping lines.
      continue

    for pattern, template, header in _re_pattern_templates:
      if pattern.search(line):
        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()

  # Did we find the header for this file (if any) and succesfully load it?
  header_found = False

  # Use the absolute path so that matching works properly.
  abs_filename = FileInfo(filename).FullName()

  # For Emacs's flymake.
  # If cpplint is invoked from Emacs's flymake, a temporary file is generated
  # by flymake and that file name might end with '_flymake.cc'. In that case,
  # restore original file name here so that the corresponding header file can be
  # found.
  # e.g. If the file name is 'foo_flymake.cc', we should search for 'foo.h'
  # 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
  # the keys.
  header_keys = include_state.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):
      header_found = True

  # If we can't find the header file for a .cc, assume it's because we don't
  # know where to look. In that case we'll give up as we're not sure they
  # didn't include it in the .h file.
  # TODO(unknown): Do a better job of finding .h files so we are confident that
  # not having the .h file means there isn't one.
  if filename.endswith('.cc') and not header_found:
    return

  # 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:
      error(filename, required[required_header_unstripped][0],
            'build/include_what_you_use', 4,
            'Add #include ' + required_header_unstripped + ' for ' + template)


_RE_PATTERN_EXPLICIT_MAKEPAIR = re.compile(r'\bmake_pair\s*<')


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
  specified explicitly, and such use isn't intended in any case.

  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 = _RE_PATTERN_EXPLICIT_MAKEPAIR.search(line)
  if match:
    error(filename, linenum, 'build/explicit_make_pair',
          4,  # 4 = high confidence
          'For C++11-compatibility, omit template arguments from make_pair'
          ' OR use pair directly OR if appropriate, construct a pair directly')


def ProcessLine(filename, file_extension, clean_lines, line,
                include_state, function_state, nesting_state, error,
                extra_check_functions=[]):
  """Processes a single line in the file.

  Args:
    filename: Filename of the file that is being processed.
    file_extension: The extension (dot not included) of the file.
    clean_lines: An array of strings, each representing a line of the file,
                 with comments stripped.
    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
                   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
    extra_check_functions: An array of additional check functions that will be
                           run on each source line. Each function takes 4
                           arguments: filename, clean_lines, line, error
  """
  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
  CheckForFunctionLengths(filename, clean_lines, line, function_state, error)
  CheckForMultilineCommentsAndStrings(filename, clean_lines, line, error)
  CheckStyle(filename, clean_lines, line, file_extension, nesting_state, error)
  CheckLanguage(filename, clean_lines, line, file_extension, include_state,
                nesting_state, error)
  CheckForNonConstReference(filename, clean_lines, line, nesting_state, error)
  CheckForNonStandardConstructs(filename, clean_lines, line,
                                nesting_state, error)
  CheckVlogArguments(filename, clean_lines, line, error)
  CheckPosixThreading(filename, clean_lines, line, error)
  CheckInvalidIncrement(filename, clean_lines, line, error)
  CheckMakePairUsesDeduction(filename, clean_lines, line, error)
  for check_fn in extra_check_functions:
    check_fn(filename, clean_lines, line, error)

def ProcessFileData(filename, file_extension, lines, error,
                    extra_check_functions=[]):
  """Performs lint checks and reports any errors to the given error function.

  Args:
    filename: Filename of the file that is being processed.
    file_extension: The extension (dot not included) of the file.
    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.
    error: A callable to which errors are reported, which takes 4 arguments:
           filename, line number, error level, and message
    extra_check_functions: An array of additional check functions that will be
                           run on each source line. Each function takes 4
                           arguments: filename, clean_lines, line, error
  """
  lines = (['// marker so line numbers and indices both start at 1'] + lines +
           ['// marker so line numbers end in a known way'])

  include_state = _IncludeState()
  function_state = _FunctionState()
  nesting_state = _NestingState()

  ResetNolintSuppressions()

  CheckForCopyright(filename, lines, error)

  if file_extension == 'h':
    CheckForHeaderGuard(filename, lines, error)

  RemoveMultiLineComments(filename, lines, error)
  clean_lines = CleansedLines(lines)
  for line in xrange(clean_lines.NumLines()):
    ProcessLine(filename, file_extension, clean_lines, line,
                include_state, function_state, nesting_state, error,
                extra_check_functions)
  nesting_state.CheckCompletedBlocks(filename, error)

  CheckForIncludeWhatYouUse(filename, clean_lines, 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 ProcessFile(filename, vlevel, extra_check_functions=[]):
  """Does google-lint on a single file.

  Args:
    filename: The name of the file to parse.

    vlevel: The level of errors to report.  Every error of confidence
    >= verbose_level will be reported.  0 is a good default.

    extra_check_functions: An array of additional check functions that will be
                           run on each source line. Each function takes 4
                           arguments: filename, clean_lines, line, error
  """

  _SetVerboseLevel(vlevel)

  try:
    # Support the UNIX convention of using "-" for stdin.  Note that
    # we are not opening the file with universal newline support
    # (which codecs doesn't support anyway), so the resulting lines do
    # 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.

    if filename == '-':
      lines = codecs.StreamReaderWriter(sys.stdin,
                                        codecs.getreader('utf8'),
                                        codecs.getwriter('utf8'),
                                        'replace').read().split('\n')
    else:
      lines = codecs.open(filename, 'r', 'utf8', 'replace').read().split('\n')

    carriage_return_found = False
    # Remove trailing '\r'.
    for linenum in range(len(lines)):
      if lines[linenum].endswith('\r'):
        lines[linenum] = lines[linenum].rstrip('\r')
        carriage_return_found = True

  except IOError:
    sys.stderr.write(
        "Skipping input '%s': Can't open for reading\n" % filename)
    return

  # Note, if no dot is found, this will give the entire filename as the ext.
  file_extension = filename[filename.rfind('.') + 1:]

  # When reading from stdin, the extension is unknown, so no cpplint tests
  # should rely on the extension.
  if filename != '-' and file_extension not in _valid_extensions:
    sys.stderr.write('Ignoring %s; not a valid file name '
                     '(%s)\n' % (filename, ', '.join(_valid_extensions)))
  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)


def PrintUsage(message):
  """Prints a brief usage string and exits, optionally with an error message.

  Args:
    message: The optional error message.
  """
  sys.stderr.write(_USAGE)
  if message:
    sys.exit('\nFATAL ERROR: ' + message)
  else:
    sys.exit(1)


def PrintCategories():
  """Prints a list of all the error-categories used by error messages.

  These are the categories used to filter messages via --filter.
  """
  sys.stderr.write(''.join('  %s\n' % cat for cat in _ERROR_CATEGORIES))
  sys.exit(0)


def ParseArguments(args):
  """Parses the command line arguments.

  This may set the output format and verbosity level as side-effects.

  Args:
    args: The command line arguments:

  Returns:
    The list of filenames to lint.
  """
  try:
    (opts, filenames) = getopt.getopt(args, '', ['help', 'output=', 'verbose=',
                                                 'counting=',
                                                 'filter=',
                                                 'root=',
                                                 'linelength=',
                                                 'extensions='])
  except getopt.GetoptError:
    PrintUsage('Invalid arguments.')

  verbosity = _VerboseLevel()
  output_format = _OutputFormat()
  filters = ''
  counting_style = ''

  for (opt, val) in opts:
    if opt == '--help':
      PrintUsage(None)
    elif opt == '--output':
      if val not in ('emacs', 'vs7', 'eclipse'):
        PrintUsage('The only allowed output formats are emacs, vs7 and eclipse.')
      output_format = val
    elif opt == '--verbose':
      verbosity = int(val)
    elif opt == '--filter':
      filters = val
      if not filters:
        PrintCategories()
    elif opt == '--counting':
      if val not in ('total', 'toplevel', 'detailed'):
        PrintUsage('Valid counting options are total, toplevel, and detailed')
      counting_style = val
    elif opt == '--root':
      global _root
      _root = val
    elif opt == '--linelength':
      global _line_length
      try:
          _line_length = int(val)
      except ValueError:
          PrintUsage('Line length must be digits.')
    elif opt == '--extensions':
      global _valid_extensions
      try:
          _valid_extensions = set(val.split(','))
      except ValueError:
          PrintUsage('Extensions must be comma seperated list.')

  if not filenames:
    PrintUsage('No files were specified.')

  _SetOutputFormat(output_format)
  _SetVerboseLevel(verbosity)
  _SetFilters(filters)
  _SetCountingStyle(counting_style)

  return filenames


def main():
  filenames = ParseArguments(sys.argv[1:])

  # Change stderr to write with replacement characters so we don't die
  # if we try to print something containing non-ASCII characters.
  sys.stderr = codecs.StreamReaderWriter(sys.stderr,
                                         codecs.getreader('utf8'),
                                         codecs.getwriter('utf8'),
                                         'replace')

  _cpplint_state.ResetErrorCounts()
  for filename in filenames:
    ProcessFile(filename, _cpplint_state.verbose_level)
  _cpplint_state.PrintErrorCounts()

  sys.exit(_cpplint_state.error_count > 0)


if __name__ == '__main__':
  main()