Source code for flake8.checker

"""Checker Manager and Checker classes."""
import argparse
import collections
import errno
import itertools
import logging
import multiprocessing.pool
import signal
import tokenize
from typing import Any
from typing import Dict
from typing import List
from typing import Optional
from typing import Tuple

from flake8 import defaults
from flake8 import exceptions
from flake8 import processor
from flake8 import utils
from flake8.discover_files import expand_paths
from flake8.plugins.finder import Checkers
from flake8.plugins.finder import LoadedPlugin
from flake8.style_guide import StyleGuideManager

Results = List[Tuple[str, int, int, str, Optional[str]]]

LOG = logging.getLogger(__name__)

SERIAL_RETRY_ERRNOS = {
    # ENOSPC: Added by sigmavirus24
    # > On some operating systems (OSX), multiprocessing may cause an
    # > ENOSPC error while trying to create a Semaphore.
    # > In those cases, we should replace the customized Queue Report
    # > class with pep8's StandardReport class to ensure users don't run
    # > into this problem.
    # > (See also: https://github.com/pycqa/flake8/issues/117)
    errno.ENOSPC,
    # NOTE(sigmavirus24): When adding to this list, include the reasoning
    # on the lines before the error code and always append your error
    # code. Further, please always add a trailing `,` to reduce the visual
    # noise in diffs.
}


[docs]class Manager: """Manage the parallelism and checker instances for each plugin and file. This class will be responsible for the following: - Determining the parallelism of Flake8, e.g.: * Do we use :mod:`multiprocessing` or is it unavailable? * Do we automatically decide on the number of jobs to use or did the user provide that? - Falling back to a serial way of processing files if we run into an OSError related to :mod:`multiprocessing` - Organizing the results of each checker so we can group the output together and make our output deterministic. """ def __init__( self, style_guide: StyleGuideManager, plugins: Checkers, ) -> None: """Initialize our Manager instance.""" self.style_guide = style_guide self.options = style_guide.options self.plugins = plugins self.jobs = self._job_count() self._all_checkers: List[FileChecker] = [] self.checkers: List[FileChecker] = [] self.statistics = { "files": 0, "logical lines": 0, "physical lines": 0, "tokens": 0, } self.exclude = tuple( itertools.chain(self.options.exclude, self.options.extend_exclude) ) def _process_statistics(self) -> None: for checker in self.checkers: for statistic in defaults.STATISTIC_NAMES: self.statistics[statistic] += checker.statistics[statistic] self.statistics["files"] += len(self.checkers) def _job_count(self) -> int: # First we walk through all of our error cases: # - multiprocessing library is not present # - we're running on windows in which case we know we have significant # implementation issues # - the user provided stdin and that's not something we can handle # well # - we're processing a diff, which again does not work well with # multiprocessing and which really shouldn't require multiprocessing # - the user provided some awful input # class state is only preserved when using the `fork` strategy. if multiprocessing.get_start_method() != "fork": LOG.warning( "The multiprocessing module is not available. " "Ignoring --jobs arguments." ) return 0 if utils.is_using_stdin(self.options.filenames): LOG.warning( "The --jobs option is not compatible with supplying " "input using - . Ignoring --jobs arguments." ) return 0 if self.options.diff: LOG.warning( "The --diff option was specified with --jobs but " "they are not compatible. Ignoring --jobs arguments." ) return 0 jobs = self.options.jobs # If the value is "auto", we want to let the multiprocessing library # decide the number based on the number of CPUs. However, if that # function is not implemented for this particular value of Python we # default to 1 if jobs.is_auto: try: return multiprocessing.cpu_count() except NotImplementedError: return 0 # Otherwise, we know jobs should be an integer and we can just convert # it to an integer return jobs.n_jobs def _handle_results(self, filename: str, results: Results) -> int: style_guide = self.style_guide reported_results_count = 0 for (error_code, line_number, column, text, physical_line) in results: reported_results_count += style_guide.handle_error( code=error_code, filename=filename, line_number=line_number, column_number=column, text=text, physical_line=physical_line, ) return reported_results_count
[docs] def make_checkers(self, paths: Optional[List[str]] = None) -> None: """Create checkers for each file.""" if paths is None: paths = self.options.filenames self._all_checkers = [ FileChecker( filename=filename, plugins=self.plugins, options=self.options, ) for filename in expand_paths( paths=paths, stdin_display_name=self.options.stdin_display_name, filename_patterns=self.options.filename, exclude=self.exclude, is_running_from_diff=self.options.diff, ) ] self.checkers = [c for c in self._all_checkers if c.should_process] LOG.info("Checking %d files", len(self.checkers))
[docs] def report(self) -> Tuple[int, int]: """Report all of the errors found in the managed file checkers. This iterates over each of the checkers and reports the errors sorted by line number. :returns: A tuple of the total results found and the results reported. """ results_reported = results_found = 0 for checker in self._all_checkers: results = sorted(checker.results, key=lambda tup: (tup[1], tup[2])) filename = checker.display_name with self.style_guide.processing_file(filename): results_reported += self._handle_results(filename, results) results_found += len(results) return (results_found, results_reported)
[docs] def run_parallel(self) -> None: """Run the checkers in parallel.""" # fmt: off final_results: Dict[str, List[Tuple[str, int, int, str, Optional[str]]]] = collections.defaultdict(list) # noqa: E501 final_statistics: Dict[str, Dict[str, int]] = collections.defaultdict(dict) # noqa: E501 # fmt: on pool = _try_initialize_processpool(self.jobs) if pool is None: self.run_serial() return pool_closed = False try: pool_map = pool.imap_unordered( _run_checks, self.checkers, chunksize=calculate_pool_chunksize( len(self.checkers), self.jobs ), ) for ret in pool_map: filename, results, statistics = ret final_results[filename] = results final_statistics[filename] = statistics pool.close() pool.join() pool_closed = True finally: if not pool_closed: pool.terminate() pool.join() for checker in self.checkers: filename = checker.display_name checker.results = final_results[filename] checker.statistics = final_statistics[filename]
[docs] def run_serial(self) -> None: """Run the checkers in serial.""" for checker in self.checkers: checker.run_checks()
[docs] def run(self) -> None: """Run all the checkers. This will intelligently decide whether to run the checks in parallel or whether to run them in serial. If running the checks in parallel causes a problem (e.g., :issue:`117`) this also implements fallback to serial processing. """ try: if self.jobs > 1 and len(self.checkers) > 1: self.run_parallel() else: self.run_serial() except KeyboardInterrupt: LOG.warning("Flake8 was interrupted by the user") raise exceptions.EarlyQuit("Early quit while running checks")
[docs] def start(self, paths: Optional[List[str]] = None) -> None: """Start checking files. :param paths: Path names to check. This is passed directly to :meth:`~Manager.make_checkers`. """ LOG.info("Making checkers") self.make_checkers(paths)
[docs] def stop(self) -> None: """Stop checking files.""" self._process_statistics()
[docs]class FileChecker: """Manage running checks for a file and aggregate the results.""" def __init__( self, *, filename: str, plugins: Checkers, options: argparse.Namespace, ) -> None: """Initialize our file checker.""" self.options = options self.filename = filename self.plugins = plugins self.results: Results = [] self.statistics = { "tokens": 0, "logical lines": 0, "physical lines": 0, } self.processor = self._make_processor() self.display_name = filename self.should_process = False if self.processor is not None: self.display_name = self.processor.filename self.should_process = not self.processor.should_ignore_file() self.statistics["physical lines"] = len(self.processor.lines) def __repr__(self) -> str: """Provide helpful debugging representation.""" return f"FileChecker for {self.filename}" def _make_processor(self) -> Optional[processor.FileProcessor]: try: return processor.FileProcessor(self.filename, self.options) except OSError as e: # If we can not read the file due to an IOError (e.g., the file # does not exist or we do not have the permissions to open it) # then we need to format that exception for the user. # NOTE(sigmavirus24): Historically, pep8 has always reported this # as an E902. We probably *want* a better error code for this # going forward. self.report("E902", 0, 0, f"{type(e).__name__}: {e}") return None
[docs] def report( self, error_code: Optional[str], line_number: int, column: int, text: str, ) -> str: """Report an error by storing it in the results list.""" if error_code is None: error_code, text = text.split(" ", 1) # If we're recovering from a problem in _make_processor, we will not # have this attribute. if hasattr(self, "processor") and self.processor is not None: line = self.processor.noqa_line_for(line_number) else: line = None self.results.append((error_code, line_number, column, text, line)) return error_code
[docs] def run_check(self, plugin: LoadedPlugin, **arguments: Any) -> Any: """Run the check in a single plugin.""" assert self.processor is not None try: params = self.processor.keyword_arguments_for( plugin.parameters, arguments ) except AttributeError as ae: raise exceptions.PluginRequestedUnknownParameters( plugin_name=plugin.display_name, exception=ae ) try: return plugin.obj(**arguments, **params) except Exception as all_exc: LOG.critical( "Plugin %s raised an unexpected exception", plugin.display_name, exc_info=True, ) raise exceptions.PluginExecutionFailed( filename=self.filename, plugin_name=plugin.display_name, exception=all_exc, )
@staticmethod def _extract_syntax_information(exception: Exception) -> Tuple[int, int]: if ( len(exception.args) > 1 and exception.args[1] and len(exception.args[1]) > 2 ): token = exception.args[1] row, column = token[1:3] elif ( isinstance(exception, tokenize.TokenError) and len(exception.args) == 2 and len(exception.args[1]) == 2 ): token = () row, column = exception.args[1] else: token = () row, column = (1, 0) if ( column > 0 and token and isinstance(exception, SyntaxError) and len(token) == 4 # Python 3.9 or earlier ): # NOTE(sigmavirus24): SyntaxErrors report 1-indexed column # numbers. We need to decrement the column number by 1 at # least. column_offset = 1 row_offset = 0 # See also: https://github.com/pycqa/flake8/issues/169, # https://github.com/PyCQA/flake8/issues/1372 # On Python 3.9 and earlier, token will be a 4-item tuple with the # last item being the string. Starting with 3.10, they added to # the tuple so now instead of it ending with the code that failed # to parse, it ends with the end of the section of code that # failed to parse. Luckily the absolute position in the tuple is # stable across versions so we can use that here physical_line = token[3] # NOTE(sigmavirus24): Not all "tokens" have a string as the last # argument. In this event, let's skip trying to find the correct # column and row values. if physical_line is not None: # NOTE(sigmavirus24): SyntaxErrors also don't exactly have a # "physical" line so much as what was accumulated by the point # tokenizing failed. # See also: https://github.com/pycqa/flake8/issues/169 lines = physical_line.rstrip("\n").split("\n") row_offset = len(lines) - 1 logical_line = lines[0] logical_line_length = len(logical_line) if column > logical_line_length: column = logical_line_length row -= row_offset column -= column_offset return row, column
[docs] def run_ast_checks(self) -> None: """Run all checks expecting an abstract syntax tree.""" assert self.processor is not None ast = self.processor.build_ast() for plugin in self.plugins.tree: checker = self.run_check(plugin, tree=ast) # If the plugin uses a class, call the run method of it, otherwise # the call should return something iterable itself try: runner = checker.run() except AttributeError: runner = checker for (line_number, offset, text, _) in runner: self.report( error_code=None, line_number=line_number, column=offset, text=text, )
[docs] def run_logical_checks(self) -> None: """Run all checks expecting a logical line.""" assert self.processor is not None comments, logical_line, mapping = self.processor.build_logical_line() if not mapping: return self.processor.update_state(mapping) LOG.debug('Logical line: "%s"', logical_line.rstrip()) for plugin in self.plugins.logical_line: self.processor.update_checker_state_for(plugin) results = self.run_check(plugin, logical_line=logical_line) or () for offset, text in results: line_number, column_offset = find_offset(offset, mapping) if line_number == column_offset == 0: LOG.warning("position of error out of bounds: %s", plugin) self.report( error_code=None, line_number=line_number, column=column_offset, text=text, ) self.processor.next_logical_line()
[docs] def run_physical_checks(self, physical_line: str) -> None: """Run all checks for a given physical line. A single physical check may return multiple errors. """ assert self.processor is not None for plugin in self.plugins.physical_line: self.processor.update_checker_state_for(plugin) result = self.run_check(plugin, physical_line=physical_line) if result is not None: # This is a single result if first element is an int column_offset = None try: column_offset = result[0] except (IndexError, TypeError): pass if isinstance(column_offset, int): # If we only have a single result, convert to a collection result = (result,) for result_single in result: column_offset, text = result_single self.report( error_code=None, line_number=self.processor.line_number, column=column_offset, text=text, )
[docs] def process_tokens(self) -> None: """Process tokens and trigger checks. Instead of using this directly, you should use :meth:`flake8.checker.FileChecker.run_checks`. """ assert self.processor is not None parens = 0 statistics = self.statistics file_processor = self.processor prev_physical = "" for token in file_processor.generate_tokens(): statistics["tokens"] += 1 self.check_physical_eol(token, prev_physical) token_type, text = token[0:2] if token_type == tokenize.OP: parens = processor.count_parentheses(parens, text) elif parens == 0: if processor.token_is_newline(token): self.handle_newline(token_type) prev_physical = token[4] if file_processor.tokens: # If any tokens are left over, process them self.run_physical_checks(file_processor.lines[-1]) self.run_logical_checks()
[docs] def run_checks(self) -> Tuple[str, Results, Dict[str, int]]: """Run checks against the file.""" assert self.processor is not None try: self.run_ast_checks() self.process_tokens() except (SyntaxError, tokenize.TokenError) as e: code = "E902" if isinstance(e, tokenize.TokenError) else "E999" row, column = self._extract_syntax_information(e) self.report(code, row, column, f"{type(e).__name__}: {e.args[0]}") return self.filename, self.results, self.statistics logical_lines = self.processor.statistics["logical lines"] self.statistics["logical lines"] = logical_lines return self.filename, self.results, self.statistics
[docs] def handle_newline(self, token_type: int) -> None: """Handle the logic when encountering a newline token.""" assert self.processor is not None if token_type == tokenize.NEWLINE: self.run_logical_checks() self.processor.reset_blank_before() elif len(self.processor.tokens) == 1: # The physical line contains only this token. self.processor.visited_new_blank_line() self.processor.delete_first_token() else: self.run_logical_checks()
[docs] def check_physical_eol( self, token: tokenize.TokenInfo, prev_physical: str ) -> None: """Run physical checks if and only if it is at the end of the line.""" assert self.processor is not None # a newline token ends a single physical line. if processor.is_eol_token(token): # if the file does not end with a newline, the NEWLINE # token is inserted by the parser, but it does not contain # the previous physical line in `token[4]` if token[4] == "": self.run_physical_checks(prev_physical) else: self.run_physical_checks(token[4]) elif processor.is_multiline_string(token): # Less obviously, a string that contains newlines is a # multiline string, either triple-quoted or with internal # newlines backslash-escaped. Check every physical line in the # string *except* for the last one: its newline is outside of # the multiline string, so we consider it a regular physical # line, and will check it like any other physical line. # # Subtleties: # - have to wind self.line_number back because initially it # points to the last line of the string, and we want # check_physical() to give accurate feedback line_no = token[2][0] with self.processor.inside_multiline(line_number=line_no): for line in self.processor.split_line(token): self.run_physical_checks(line)
def _pool_init() -> None: """Ensure correct signaling of ^C using multiprocessing.Pool.""" signal.signal(signal.SIGINT, signal.SIG_IGN) def _try_initialize_processpool( job_count: int, ) -> Optional[multiprocessing.pool.Pool]: """Return a new process pool instance if we are able to create one.""" try: return multiprocessing.Pool(job_count, _pool_init) except OSError as err: if err.errno not in SERIAL_RETRY_ERRNOS: raise except ImportError: pass return None def calculate_pool_chunksize(num_checkers: int, num_jobs: int) -> int: """Determine the chunksize for the multiprocessing Pool. - For chunksize, see: https://docs.python.org/3/library/multiprocessing.html#multiprocessing.pool.Pool.imap # noqa - This formula, while not perfect, aims to give each worker two batches of work. - See: https://github.com/pycqa/flake8/issues/829#note_18878876 - See: https://github.com/pycqa/flake8/issues/197 """ return max(num_checkers // (num_jobs * 2), 1) def _run_checks(checker: FileChecker) -> Tuple[str, Results, Dict[str, int]]: return checker.run_checks() def find_offset( offset: int, mapping: processor._LogicalMapping ) -> Tuple[int, int]: """Find the offset tuple for a single offset.""" if isinstance(offset, tuple): return offset for token in mapping: token_offset = token[0] if offset <= token_offset: position = token[1] break else: position = (0, 0) offset = token_offset = 0 return (position[0], position[1] + offset - token_offset)