All patches and comments are welcome. Please squash your changes to logical
commits before using git-format-patch and git-send-email to
patches@git.madduck.net.
If you'd read over the Git project's submission guidelines and adhered to them,
I'd be especially grateful.
2 from asyncio.base_events import BaseEventLoop
3 from concurrent.futures import Executor, ProcessPoolExecutor
4 from datetime import datetime
6 from functools import lru_cache, partial, wraps
10 from multiprocessing import Manager, freeze_support
12 from pathlib import Path
39 from appdirs import user_cache_dir
40 from attr import dataclass, evolve, Factory
45 from blib2to3.pytree import Node, Leaf, type_repr
46 from blib2to3 import pygram, pytree
47 from blib2to3.pgen2 import driver, token
48 from blib2to3.pgen2.grammar import Grammar
49 from blib2to3.pgen2.parse import ParseError
52 __version__ = "18.9b0"
53 DEFAULT_LINE_LENGTH = 88
55 r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|_build|buck-out|build|dist)/"
57 DEFAULT_INCLUDES = r"\.pyi?$"
58 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
70 LN = Union[Leaf, Node]
71 SplitFunc = Callable[["Line", bool], Iterator["Line"]]
74 CacheInfo = Tuple[Timestamp, FileSize]
75 Cache = Dict[Path, CacheInfo]
76 out = partial(click.secho, bold=True, err=True)
77 err = partial(click.secho, fg="red", err=True)
79 pygram.initialize(CACHE_DIR)
80 syms = pygram.python_symbols
83 class NothingChanged(UserWarning):
84 """Raised when reformatted code is the same as source."""
87 class CannotSplit(Exception):
88 """A readable split that fits the allotted line length is impossible."""
91 class InvalidInput(ValueError):
92 """Raised when input source code fails all parse attempts."""
95 class WriteBack(Enum):
102 def from_configuration(cls, *, check: bool, diff: bool) -> "WriteBack":
103 if check and not diff:
106 return cls.DIFF if diff else cls.YES
115 class TargetVersion(Enum):
125 def is_python2(self) -> bool:
126 return self is TargetVersion.CPY27
129 PY36_VERSIONS = {TargetVersion.CPY36, TargetVersion.CPY37, TargetVersion.CPY38}
133 # All string literals are unicode
136 NUMERIC_UNDERSCORES = 3
140 VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = {
141 TargetVersion.CPY27: set(),
142 TargetVersion.PYPY35: {Feature.UNICODE_LITERALS, Feature.F_STRINGS},
143 TargetVersion.CPY33: {Feature.UNICODE_LITERALS},
144 TargetVersion.CPY34: {Feature.UNICODE_LITERALS},
145 TargetVersion.CPY35: {Feature.UNICODE_LITERALS, Feature.TRAILING_COMMA},
146 TargetVersion.CPY36: {
147 Feature.UNICODE_LITERALS,
149 Feature.NUMERIC_UNDERSCORES,
150 Feature.TRAILING_COMMA,
152 TargetVersion.CPY37: {
153 Feature.UNICODE_LITERALS,
155 Feature.NUMERIC_UNDERSCORES,
156 Feature.TRAILING_COMMA,
158 TargetVersion.CPY38: {
159 Feature.UNICODE_LITERALS,
161 Feature.NUMERIC_UNDERSCORES,
162 Feature.TRAILING_COMMA,
169 target_versions: Set[TargetVersion] = Factory(set)
170 line_length: int = DEFAULT_LINE_LENGTH
171 string_normalization: bool = True
174 def get_cache_key(self) -> str:
175 if self.target_versions:
176 version_str = ",".join(
178 for version in sorted(self.target_versions, key=lambda v: v.value)
184 str(self.line_length),
185 str(int(self.string_normalization)),
186 str(int(self.is_pyi)),
188 return ".".join(parts)
191 def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool:
192 return all(feature in VERSION_TO_FEATURES[version] for version in target_versions)
195 def read_pyproject_toml(
196 ctx: click.Context, param: click.Parameter, value: Union[str, int, bool, None]
198 """Inject Black configuration from "pyproject.toml" into defaults in `ctx`.
200 Returns the path to a successfully found and read configuration file, None
203 assert not isinstance(value, (int, bool)), "Invalid parameter type passed"
205 root = find_project_root(ctx.params.get("src", ()))
206 path = root / "pyproject.toml"
213 pyproject_toml = toml.load(value)
214 config = pyproject_toml.get("tool", {}).get("black", {})
215 except (toml.TomlDecodeError, OSError) as e:
216 raise click.FileError(
217 filename=value, hint=f"Error reading configuration file: {e}"
223 if ctx.default_map is None:
225 ctx.default_map.update( # type: ignore # bad types in .pyi
226 {k.replace("--", "").replace("-", "_"): v for k, v in config.items()}
231 @click.command(context_settings=dict(help_option_names=["-h", "--help"]))
236 default=DEFAULT_LINE_LENGTH,
237 help="How many characters per line to allow.",
243 type=click.Choice([v.name.lower() for v in TargetVersion]),
244 callback=lambda c, p, v: [TargetVersion[val.upper()] for val in v],
247 "Python versions that should be supported by Black's output. [default: "
248 "per-file auto-detection]"
255 "Allow using Python 3.6-only syntax on all input files. This will put "
256 "trailing commas in function signatures and calls also after *args and "
257 "**kwargs. [default: per-file auto-detection]"
264 "Format all input files like typing stubs regardless of file extension "
265 "(useful when piping source on standard input)."
270 "--skip-string-normalization",
272 help="Don't normalize string quotes or prefixes.",
278 "Don't write the files back, just return the status. Return code 0 "
279 "means nothing would change. Return code 1 means some files would be "
280 "reformatted. Return code 123 means there was an internal error."
286 help="Don't write the files back, just output a diff for each file on stdout.",
291 help="If --fast given, skip temporary sanity checks. [default: --safe]",
296 default=DEFAULT_INCLUDES,
298 "A regular expression that matches files and directories that should be "
299 "included on recursive searches. An empty value means all files are "
300 "included regardless of the name. Use forward slashes for directories on "
301 "all platforms (Windows, too). Exclusions are calculated first, inclusions "
309 default=DEFAULT_EXCLUDES,
311 "A regular expression that matches files and directories that should be "
312 "excluded on recursive searches. An empty value means no paths are excluded. "
313 "Use forward slashes for directories on all platforms (Windows, too). "
314 "Exclusions are calculated first, inclusions later."
323 "Don't emit non-error messages to stderr. Errors are still emitted, "
324 "silence those with 2>/dev/null."
332 "Also emit messages to stderr about files that were not changed or were "
333 "ignored due to --exclude=."
336 @click.version_option(version=__version__)
341 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
348 exists=False, file_okay=True, dir_okay=False, readable=True, allow_dash=False
351 callback=read_pyproject_toml,
352 help="Read configuration from PATH.",
358 target_version: List[TargetVersion],
364 skip_string_normalization: bool,
370 config: Optional[str],
372 """The uncompromising code formatter."""
373 write_back = WriteBack.from_configuration(check=check, diff=diff)
376 err(f"Cannot use both --target-version and --py36")
379 versions = set(target_version)
381 versions = PY36_VERSIONS
383 # We'll autodetect later.
386 target_versions=versions,
387 line_length=line_length,
389 string_normalization=not skip_string_normalization,
391 if config and verbose:
392 out(f"Using configuration from {config}.", bold=False, fg="blue")
394 include_regex = re_compile_maybe_verbose(include)
396 err(f"Invalid regular expression for include given: {include!r}")
399 exclude_regex = re_compile_maybe_verbose(exclude)
401 err(f"Invalid regular expression for exclude given: {exclude!r}")
403 report = Report(check=check, quiet=quiet, verbose=verbose)
404 root = find_project_root(src)
405 sources: Set[Path] = set()
410 gen_python_files_in_dir(p, root, include_regex, exclude_regex, report)
412 elif p.is_file() or s == "-":
413 # if a file was explicitly given, we don't care about its extension
416 err(f"invalid path: {s}")
417 if len(sources) == 0:
418 if verbose or not quiet:
419 out("No paths given. Nothing to do 😴")
422 if len(sources) == 1:
426 write_back=write_back,
431 loop = asyncio.get_event_loop()
432 executor = ProcessPoolExecutor(max_workers=os.cpu_count())
434 loop.run_until_complete(
438 write_back=write_back,
447 if verbose or not quiet:
448 bang = "💥 💔 💥" if report.return_code else "✨ 🍰 ✨"
449 out(f"All done! {bang}")
450 click.secho(str(report), err=True)
451 ctx.exit(report.return_code)
455 src: Path, fast: bool, write_back: WriteBack, mode: FileMode, report: "Report"
457 """Reformat a single file under `src` without spawning child processes.
459 If `quiet` is True, non-error messages are not output. `line_length`,
460 `write_back`, `fast` and `pyi` options are passed to
461 :func:`format_file_in_place` or :func:`format_stdin_to_stdout`.
465 if not src.is_file() and str(src) == "-":
466 if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode):
467 changed = Changed.YES
470 if write_back != WriteBack.DIFF:
471 cache = read_cache(mode)
472 res_src = src.resolve()
473 if res_src in cache and cache[res_src] == get_cache_info(res_src):
474 changed = Changed.CACHED
475 if changed is not Changed.CACHED and format_file_in_place(
476 src, fast=fast, write_back=write_back, mode=mode
478 changed = Changed.YES
479 if (write_back is WriteBack.YES and changed is not Changed.CACHED) or (
480 write_back is WriteBack.CHECK and changed is Changed.NO
482 write_cache(cache, [src], mode)
483 report.done(src, changed)
484 except Exception as exc:
485 report.failed(src, str(exc))
488 async def schedule_formatting(
491 write_back: WriteBack,
497 """Run formatting of `sources` in parallel using the provided `executor`.
499 (Use ProcessPoolExecutors for actual parallelism.)
501 `line_length`, `write_back`, `fast`, and `pyi` options are passed to
502 :func:`format_file_in_place`.
505 if write_back != WriteBack.DIFF:
506 cache = read_cache(mode)
507 sources, cached = filter_cached(cache, sources)
508 for src in sorted(cached):
509 report.done(src, Changed.CACHED)
514 sources_to_cache = []
516 if write_back == WriteBack.DIFF:
517 # For diff output, we need locks to ensure we don't interleave output
518 # from different processes.
520 lock = manager.Lock()
522 loop.run_in_executor(
523 executor, format_file_in_place, src, fast, mode, write_back, lock
525 for src in sorted(sources)
527 pending: Iterable[asyncio.Task] = tasks.keys()
529 loop.add_signal_handler(signal.SIGINT, cancel, pending)
530 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
531 except NotImplementedError:
532 # There are no good alternatives for these on Windows.
535 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
537 src = tasks.pop(task)
539 cancelled.append(task)
540 elif task.exception():
541 report.failed(src, str(task.exception()))
543 changed = Changed.YES if task.result() else Changed.NO
544 # If the file was written back or was successfully checked as
545 # well-formatted, store this information in the cache.
546 if write_back is WriteBack.YES or (
547 write_back is WriteBack.CHECK and changed is Changed.NO
549 sources_to_cache.append(src)
550 report.done(src, changed)
552 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
554 write_cache(cache, sources_to_cache, mode)
557 def format_file_in_place(
561 write_back: WriteBack = WriteBack.NO,
562 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
564 """Format file under `src` path. Return True if changed.
566 If `write_back` is DIFF, write a diff to stdout. If it is YES, write reformatted
568 `line_length` and `fast` options are passed to :func:`format_file_contents`.
570 if src.suffix == ".pyi":
571 mode = evolve(mode, is_pyi=True)
573 then = datetime.utcfromtimestamp(src.stat().st_mtime)
574 with open(src, "rb") as buf:
575 src_contents, encoding, newline = decode_bytes(buf.read())
577 dst_contents = format_file_contents(src_contents, fast=fast, mode=mode)
578 except NothingChanged:
581 if write_back == write_back.YES:
582 with open(src, "w", encoding=encoding, newline=newline) as f:
583 f.write(dst_contents)
584 elif write_back == write_back.DIFF:
585 now = datetime.utcnow()
586 src_name = f"{src}\t{then} +0000"
587 dst_name = f"{src}\t{now} +0000"
588 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
592 f = io.TextIOWrapper(
598 f.write(diff_contents)
606 def format_stdin_to_stdout(
607 fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: FileMode
609 """Format file on stdin. Return True if changed.
611 If `write_back` is YES, write reformatted code back to stdout. If it is DIFF,
612 write a diff to stdout. The `mode` argument is passed to
613 :func:`format_file_contents`.
615 then = datetime.utcnow()
616 src, encoding, newline = decode_bytes(sys.stdin.buffer.read())
619 dst = format_file_contents(src, fast=fast, mode=mode)
622 except NothingChanged:
626 f = io.TextIOWrapper(
627 sys.stdout.buffer, encoding=encoding, newline=newline, write_through=True
629 if write_back == WriteBack.YES:
631 elif write_back == WriteBack.DIFF:
632 now = datetime.utcnow()
633 src_name = f"STDIN\t{then} +0000"
634 dst_name = f"STDOUT\t{now} +0000"
635 f.write(diff(src, dst, src_name, dst_name))
639 def format_file_contents(
640 src_contents: str, *, fast: bool, mode: FileMode
642 """Reformat contents a file and return new contents.
644 If `fast` is False, additionally confirm that the reformatted code is
645 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
646 `line_length` is passed to :func:`format_str`.
648 if src_contents.strip() == "":
651 dst_contents = format_str(src_contents, mode=mode)
652 if src_contents == dst_contents:
656 assert_equivalent(src_contents, dst_contents)
657 assert_stable(src_contents, dst_contents, mode=mode)
661 def format_str(src_contents: str, *, mode: FileMode) -> FileContent:
662 """Reformat a string and return new contents.
664 `line_length` determines how many characters per line are allowed.
666 src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
668 future_imports = get_future_imports(src_node)
669 if mode.target_versions:
670 versions = mode.target_versions
672 versions = detect_target_versions(src_node)
673 normalize_fmt_off(src_node)
674 lines = LineGenerator(
675 remove_u_prefix="unicode_literals" in future_imports
676 or supports_feature(versions, Feature.UNICODE_LITERALS),
678 normalize_strings=mode.string_normalization,
680 elt = EmptyLineTracker(is_pyi=mode.is_pyi)
683 for current_line in lines.visit(src_node):
684 for _ in range(after):
685 dst_contents += str(empty_line)
686 before, after = elt.maybe_empty_lines(current_line)
687 for _ in range(before):
688 dst_contents += str(empty_line)
689 for line in split_line(
691 line_length=mode.line_length,
692 supports_trailing_commas=supports_feature(versions, Feature.TRAILING_COMMA),
694 dst_contents += str(line)
698 def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]:
699 """Return a tuple of (decoded_contents, encoding, newline).
701 `newline` is either CRLF or LF but `decoded_contents` is decoded with
702 universal newlines (i.e. only contains LF).
704 srcbuf = io.BytesIO(src)
705 encoding, lines = tokenize.detect_encoding(srcbuf.readline)
707 return "", encoding, "\n"
709 newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n"
711 with io.TextIOWrapper(srcbuf, encoding) as tiow:
712 return tiow.read(), encoding, newline
716 pygram.python_grammar_no_print_statement_no_exec_statement,
717 pygram.python_grammar_no_print_statement,
718 pygram.python_grammar,
722 def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
723 if not target_versions:
725 elif all(not version.is_python2() for version in target_versions):
726 # Python 2-compatible code, so don't try Python 3 grammar.
728 pygram.python_grammar_no_print_statement_no_exec_statement,
729 pygram.python_grammar_no_print_statement,
732 return [pygram.python_grammar]
735 def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
736 """Given a string with source, return the lib2to3 Node."""
737 if src_txt[-1:] != "\n":
740 for grammar in get_grammars(set(target_versions)):
741 drv = driver.Driver(grammar, pytree.convert)
743 result = drv.parse_string(src_txt, True)
746 except ParseError as pe:
747 lineno, column = pe.context[1]
748 lines = src_txt.splitlines()
750 faulty_line = lines[lineno - 1]
752 faulty_line = "<line number missing in source>"
753 exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
757 if isinstance(result, Leaf):
758 result = Node(syms.file_input, [result])
762 def lib2to3_unparse(node: Node) -> str:
763 """Given a lib2to3 node, return its string representation."""
771 class Visitor(Generic[T]):
772 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
774 def visit(self, node: LN) -> Iterator[T]:
775 """Main method to visit `node` and its children.
777 It tries to find a `visit_*()` method for the given `node.type`, like
778 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
779 If no dedicated `visit_*()` method is found, chooses `visit_default()`
782 Then yields objects of type `T` from the selected visitor.
785 name = token.tok_name[node.type]
787 name = type_repr(node.type)
788 yield from getattr(self, f"visit_{name}", self.visit_default)(node)
790 def visit_default(self, node: LN) -> Iterator[T]:
791 """Default `visit_*()` implementation. Recurses to children of `node`."""
792 if isinstance(node, Node):
793 for child in node.children:
794 yield from self.visit(child)
798 class DebugVisitor(Visitor[T]):
801 def visit_default(self, node: LN) -> Iterator[T]:
802 indent = " " * (2 * self.tree_depth)
803 if isinstance(node, Node):
804 _type = type_repr(node.type)
805 out(f"{indent}{_type}", fg="yellow")
807 for child in node.children:
808 yield from self.visit(child)
811 out(f"{indent}/{_type}", fg="yellow", bold=False)
813 _type = token.tok_name.get(node.type, str(node.type))
814 out(f"{indent}{_type}", fg="blue", nl=False)
816 # We don't have to handle prefixes for `Node` objects since
817 # that delegates to the first child anyway.
818 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
819 out(f" {node.value!r}", fg="blue", bold=False)
822 def show(cls, code: Union[str, Leaf, Node]) -> None:
823 """Pretty-print the lib2to3 AST of a given string of `code`.
825 Convenience method for debugging.
827 v: DebugVisitor[None] = DebugVisitor()
828 if isinstance(code, str):
829 code = lib2to3_parse(code)
833 WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE}
844 STANDALONE_COMMENT = 153
845 token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT"
846 LOGIC_OPERATORS = {"and", "or"}
871 STARS = {token.STAR, token.DOUBLESTAR}
874 syms.argument, # double star in arglist
875 syms.trailer, # single argument to call
877 syms.varargslist, # lambdas
879 UNPACKING_PARENTS = {
880 syms.atom, # single element of a list or set literal
884 syms.testlist_star_expr,
919 COMPREHENSION_PRIORITY = 20
921 TERNARY_PRIORITY = 16
924 COMPARATOR_PRIORITY = 10
935 token.DOUBLESLASH: 4,
945 class BracketTracker:
946 """Keeps track of brackets on a line."""
949 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
950 delimiters: Dict[LeafID, Priority] = Factory(dict)
951 previous: Optional[Leaf] = None
952 _for_loop_depths: List[int] = Factory(list)
953 _lambda_argument_depths: List[int] = Factory(list)
955 def mark(self, leaf: Leaf) -> None:
956 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
958 All leaves receive an int `bracket_depth` field that stores how deep
959 within brackets a given leaf is. 0 means there are no enclosing brackets
960 that started on this line.
962 If a leaf is itself a closing bracket, it receives an `opening_bracket`
963 field that it forms a pair with. This is a one-directional link to
964 avoid reference cycles.
966 If a leaf is a delimiter (a token on which Black can split the line if
967 needed) and it's on depth 0, its `id()` is stored in the tracker's
970 if leaf.type == token.COMMENT:
973 self.maybe_decrement_after_for_loop_variable(leaf)
974 self.maybe_decrement_after_lambda_arguments(leaf)
975 if leaf.type in CLOSING_BRACKETS:
977 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
978 leaf.opening_bracket = opening_bracket
979 leaf.bracket_depth = self.depth
981 delim = is_split_before_delimiter(leaf, self.previous)
982 if delim and self.previous is not None:
983 self.delimiters[id(self.previous)] = delim
985 delim = is_split_after_delimiter(leaf, self.previous)
987 self.delimiters[id(leaf)] = delim
988 if leaf.type in OPENING_BRACKETS:
989 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
992 self.maybe_increment_lambda_arguments(leaf)
993 self.maybe_increment_for_loop_variable(leaf)
995 def any_open_brackets(self) -> bool:
996 """Return True if there is an yet unmatched open bracket on the line."""
997 return bool(self.bracket_match)
999 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int:
1000 """Return the highest priority of a delimiter found on the line.
1002 Values are consistent with what `is_split_*_delimiter()` return.
1003 Raises ValueError on no delimiters.
1005 return max(v for k, v in self.delimiters.items() if k not in exclude)
1007 def delimiter_count_with_priority(self, priority: int = 0) -> int:
1008 """Return the number of delimiters with the given `priority`.
1010 If no `priority` is passed, defaults to max priority on the line.
1012 if not self.delimiters:
1015 priority = priority or self.max_delimiter_priority()
1016 return sum(1 for p in self.delimiters.values() if p == priority)
1018 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1019 """In a for loop, or comprehension, the variables are often unpacks.
1021 To avoid splitting on the comma in this situation, increase the depth of
1022 tokens between `for` and `in`.
1024 if leaf.type == token.NAME and leaf.value == "for":
1026 self._for_loop_depths.append(self.depth)
1031 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
1032 """See `maybe_increment_for_loop_variable` above for explanation."""
1034 self._for_loop_depths
1035 and self._for_loop_depths[-1] == self.depth
1036 and leaf.type == token.NAME
1037 and leaf.value == "in"
1040 self._for_loop_depths.pop()
1045 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
1046 """In a lambda expression, there might be more than one argument.
1048 To avoid splitting on the comma in this situation, increase the depth of
1049 tokens between `lambda` and `:`.
1051 if leaf.type == token.NAME and leaf.value == "lambda":
1053 self._lambda_argument_depths.append(self.depth)
1058 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
1059 """See `maybe_increment_lambda_arguments` above for explanation."""
1061 self._lambda_argument_depths
1062 and self._lambda_argument_depths[-1] == self.depth
1063 and leaf.type == token.COLON
1066 self._lambda_argument_depths.pop()
1071 def get_open_lsqb(self) -> Optional[Leaf]:
1072 """Return the most recent opening square bracket (if any)."""
1073 return self.bracket_match.get((self.depth - 1, token.RSQB))
1078 """Holds leaves and comments. Can be printed with `str(line)`."""
1081 leaves: List[Leaf] = Factory(list)
1082 # The LeafID keys of comments must remain ordered by the corresponding leaf's index
1084 comments: Dict[LeafID, List[Leaf]] = Factory(dict)
1085 bracket_tracker: BracketTracker = Factory(BracketTracker)
1086 inside_brackets: bool = False
1087 should_explode: bool = False
1089 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
1090 """Add a new `leaf` to the end of the line.
1092 Unless `preformatted` is True, the `leaf` will receive a new consistent
1093 whitespace prefix and metadata applied by :class:`BracketTracker`.
1094 Trailing commas are maybe removed, unpacked for loop variables are
1095 demoted from being delimiters.
1097 Inline comments are put aside.
1099 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
1103 if token.COLON == leaf.type and self.is_class_paren_empty:
1104 del self.leaves[-2:]
1105 if self.leaves and not preformatted:
1106 # Note: at this point leaf.prefix should be empty except for
1107 # imports, for which we only preserve newlines.
1108 leaf.prefix += whitespace(
1109 leaf, complex_subscript=self.is_complex_subscript(leaf)
1111 if self.inside_brackets or not preformatted:
1112 self.bracket_tracker.mark(leaf)
1113 self.maybe_remove_trailing_comma(leaf)
1114 if not self.append_comment(leaf):
1115 self.leaves.append(leaf)
1117 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
1118 """Like :func:`append()` but disallow invalid standalone comment structure.
1120 Raises ValueError when any `leaf` is appended after a standalone comment
1121 or when a standalone comment is not the first leaf on the line.
1123 if self.bracket_tracker.depth == 0:
1125 raise ValueError("cannot append to standalone comments")
1127 if self.leaves and leaf.type == STANDALONE_COMMENT:
1129 "cannot append standalone comments to a populated line"
1132 self.append(leaf, preformatted=preformatted)
1135 def is_comment(self) -> bool:
1136 """Is this line a standalone comment?"""
1137 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
1140 def is_decorator(self) -> bool:
1141 """Is this line a decorator?"""
1142 return bool(self) and self.leaves[0].type == token.AT
1145 def is_import(self) -> bool:
1146 """Is this an import line?"""
1147 return bool(self) and is_import(self.leaves[0])
1150 def is_class(self) -> bool:
1151 """Is this line a class definition?"""
1154 and self.leaves[0].type == token.NAME
1155 and self.leaves[0].value == "class"
1159 def is_stub_class(self) -> bool:
1160 """Is this line a class definition with a body consisting only of "..."?"""
1161 return self.is_class and self.leaves[-3:] == [
1162 Leaf(token.DOT, ".") for _ in range(3)
1166 def is_def(self) -> bool:
1167 """Is this a function definition? (Also returns True for async defs.)"""
1169 first_leaf = self.leaves[0]
1174 second_leaf: Optional[Leaf] = self.leaves[1]
1177 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
1178 first_leaf.type == token.ASYNC
1179 and second_leaf is not None
1180 and second_leaf.type == token.NAME
1181 and second_leaf.value == "def"
1185 def is_class_paren_empty(self) -> bool:
1186 """Is this a class with no base classes but using parentheses?
1188 Those are unnecessary and should be removed.
1192 and len(self.leaves) == 4
1194 and self.leaves[2].type == token.LPAR
1195 and self.leaves[2].value == "("
1196 and self.leaves[3].type == token.RPAR
1197 and self.leaves[3].value == ")"
1201 def is_triple_quoted_string(self) -> bool:
1202 """Is the line a triple quoted string?"""
1205 and self.leaves[0].type == token.STRING
1206 and self.leaves[0].value.startswith(('"""', "'''"))
1209 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
1210 """If so, needs to be split before emitting."""
1211 for leaf in self.leaves:
1212 if leaf.type == STANDALONE_COMMENT:
1213 if leaf.bracket_depth <= depth_limit:
1218 def contains_multiline_strings(self) -> bool:
1219 for leaf in self.leaves:
1220 if is_multiline_string(leaf):
1225 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1226 """Remove trailing comma if there is one and it's safe."""
1229 and self.leaves[-1].type == token.COMMA
1230 and closing.type in CLOSING_BRACKETS
1234 if closing.type == token.RBRACE:
1235 self.remove_trailing_comma()
1238 if closing.type == token.RSQB:
1239 comma = self.leaves[-1]
1240 if comma.parent and comma.parent.type == syms.listmaker:
1241 self.remove_trailing_comma()
1244 # For parens let's check if it's safe to remove the comma.
1245 # Imports are always safe.
1247 self.remove_trailing_comma()
1250 # Otherwise, if the trailing one is the only one, we might mistakenly
1251 # change a tuple into a different type by removing the comma.
1252 depth = closing.bracket_depth + 1
1254 opening = closing.opening_bracket
1255 for _opening_index, leaf in enumerate(self.leaves):
1262 for leaf in self.leaves[_opening_index + 1 :]:
1266 bracket_depth = leaf.bracket_depth
1267 if bracket_depth == depth and leaf.type == token.COMMA:
1269 if leaf.parent and leaf.parent.type == syms.arglist:
1274 self.remove_trailing_comma()
1279 def append_comment(self, comment: Leaf) -> bool:
1280 """Add an inline or standalone comment to the line."""
1282 comment.type == STANDALONE_COMMENT
1283 and self.bracket_tracker.any_open_brackets()
1288 if comment.type != token.COMMENT:
1292 comment.type = STANDALONE_COMMENT
1297 leaf_id = id(self.leaves[-1])
1298 if leaf_id not in self.comments:
1299 self.comments[leaf_id] = [comment]
1301 self.comments[leaf_id].append(comment)
1304 def comments_after(self, leaf: Leaf) -> List[Leaf]:
1305 """Generate comments that should appear directly after `leaf`."""
1306 return self.comments.get(id(leaf), [])
1308 def remove_trailing_comma(self) -> None:
1309 """Remove the trailing comma and moves the comments attached to it."""
1310 # Remember, the LeafID keys of self.comments are ordered by the
1311 # corresponding leaf's index in self.leaves
1312 # If id(self.leaves[-2]) is in self.comments, the order doesn't change.
1313 # Otherwise, we insert it into self.comments, and it becomes the last entry.
1314 # However, since we delete id(self.leaves[-1]) from self.comments, the invariant
1316 self.comments.setdefault(id(self.leaves[-2]), []).extend(
1317 self.comments.get(id(self.leaves[-1]), [])
1319 self.comments.pop(id(self.leaves[-1]), None)
1322 def is_complex_subscript(self, leaf: Leaf) -> bool:
1323 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1324 open_lsqb = self.bracket_tracker.get_open_lsqb()
1325 if open_lsqb is None:
1328 subscript_start = open_lsqb.next_sibling
1330 if isinstance(subscript_start, Node):
1331 if subscript_start.type == syms.listmaker:
1334 if subscript_start.type == syms.subscriptlist:
1335 subscript_start = child_towards(subscript_start, leaf)
1336 return subscript_start is not None and any(
1337 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1340 def __str__(self) -> str:
1341 """Render the line."""
1345 indent = " " * self.depth
1346 leaves = iter(self.leaves)
1347 first = next(leaves)
1348 res = f"{first.prefix}{indent}{first.value}"
1351 for comment in itertools.chain.from_iterable(self.comments.values()):
1355 def __bool__(self) -> bool:
1356 """Return True if the line has leaves or comments."""
1357 return bool(self.leaves or self.comments)
1361 class EmptyLineTracker:
1362 """Provides a stateful method that returns the number of potential extra
1363 empty lines needed before and after the currently processed line.
1365 Note: this tracker works on lines that haven't been split yet. It assumes
1366 the prefix of the first leaf consists of optional newlines. Those newlines
1367 are consumed by `maybe_empty_lines()` and included in the computation.
1370 is_pyi: bool = False
1371 previous_line: Optional[Line] = None
1372 previous_after: int = 0
1373 previous_defs: List[int] = Factory(list)
1375 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1376 """Return the number of extra empty lines before and after the `current_line`.
1378 This is for separating `def`, `async def` and `class` with extra empty
1379 lines (two on module-level).
1381 before, after = self._maybe_empty_lines(current_line)
1382 before -= self.previous_after
1383 self.previous_after = after
1384 self.previous_line = current_line
1385 return before, after
1387 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1389 if current_line.depth == 0:
1390 max_allowed = 1 if self.is_pyi else 2
1391 if current_line.leaves:
1392 # Consume the first leaf's extra newlines.
1393 first_leaf = current_line.leaves[0]
1394 before = first_leaf.prefix.count("\n")
1395 before = min(before, max_allowed)
1396 first_leaf.prefix = ""
1399 depth = current_line.depth
1400 while self.previous_defs and self.previous_defs[-1] >= depth:
1401 self.previous_defs.pop()
1403 before = 0 if depth else 1
1405 before = 1 if depth else 2
1406 if current_line.is_decorator or current_line.is_def or current_line.is_class:
1407 return self._maybe_empty_lines_for_class_or_def(current_line, before)
1411 and self.previous_line.is_import
1412 and not current_line.is_import
1413 and depth == self.previous_line.depth
1415 return (before or 1), 0
1419 and self.previous_line.is_class
1420 and current_line.is_triple_quoted_string
1426 def _maybe_empty_lines_for_class_or_def(
1427 self, current_line: Line, before: int
1428 ) -> Tuple[int, int]:
1429 if not current_line.is_decorator:
1430 self.previous_defs.append(current_line.depth)
1431 if self.previous_line is None:
1432 # Don't insert empty lines before the first line in the file.
1435 if self.previous_line.is_decorator:
1438 if self.previous_line.depth < current_line.depth and (
1439 self.previous_line.is_class or self.previous_line.is_def
1444 self.previous_line.is_comment
1445 and self.previous_line.depth == current_line.depth
1451 if self.previous_line.depth > current_line.depth:
1453 elif current_line.is_class or self.previous_line.is_class:
1454 if current_line.is_stub_class and self.previous_line.is_stub_class:
1455 # No blank line between classes with an empty body
1459 elif current_line.is_def and not self.previous_line.is_def:
1460 # Blank line between a block of functions and a block of non-functions
1466 if current_line.depth and newlines:
1472 class LineGenerator(Visitor[Line]):
1473 """Generates reformatted Line objects. Empty lines are not emitted.
1475 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1476 in ways that will no longer stringify to valid Python code on the tree.
1479 is_pyi: bool = False
1480 normalize_strings: bool = True
1481 current_line: Line = Factory(Line)
1482 remove_u_prefix: bool = False
1484 def line(self, indent: int = 0) -> Iterator[Line]:
1487 If the line is empty, only emit if it makes sense.
1488 If the line is too long, split it first and then generate.
1490 If any lines were generated, set up a new current_line.
1492 if not self.current_line:
1493 self.current_line.depth += indent
1494 return # Line is empty, don't emit. Creating a new one unnecessary.
1496 complete_line = self.current_line
1497 self.current_line = Line(depth=complete_line.depth + indent)
1500 def visit_default(self, node: LN) -> Iterator[Line]:
1501 """Default `visit_*()` implementation. Recurses to children of `node`."""
1502 if isinstance(node, Leaf):
1503 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1504 for comment in generate_comments(node):
1505 if any_open_brackets:
1506 # any comment within brackets is subject to splitting
1507 self.current_line.append(comment)
1508 elif comment.type == token.COMMENT:
1509 # regular trailing comment
1510 self.current_line.append(comment)
1511 yield from self.line()
1514 # regular standalone comment
1515 yield from self.line()
1517 self.current_line.append(comment)
1518 yield from self.line()
1520 normalize_prefix(node, inside_brackets=any_open_brackets)
1521 if self.normalize_strings and node.type == token.STRING:
1522 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1523 normalize_string_quotes(node)
1524 if node.type == token.NUMBER:
1525 normalize_numeric_literal(node)
1526 if node.type not in WHITESPACE:
1527 self.current_line.append(node)
1528 yield from super().visit_default(node)
1530 def visit_INDENT(self, node: Node) -> Iterator[Line]:
1531 """Increase indentation level, maybe yield a line."""
1532 # In blib2to3 INDENT never holds comments.
1533 yield from self.line(+1)
1534 yield from self.visit_default(node)
1536 def visit_DEDENT(self, node: Node) -> Iterator[Line]:
1537 """Decrease indentation level, maybe yield a line."""
1538 # The current line might still wait for trailing comments. At DEDENT time
1539 # there won't be any (they would be prefixes on the preceding NEWLINE).
1540 # Emit the line then.
1541 yield from self.line()
1543 # While DEDENT has no value, its prefix may contain standalone comments
1544 # that belong to the current indentation level. Get 'em.
1545 yield from self.visit_default(node)
1547 # Finally, emit the dedent.
1548 yield from self.line(-1)
1551 self, node: Node, keywords: Set[str], parens: Set[str]
1552 ) -> Iterator[Line]:
1553 """Visit a statement.
1555 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1556 `def`, `with`, `class`, `assert` and assignments.
1558 The relevant Python language `keywords` for a given statement will be
1559 NAME leaves within it. This methods puts those on a separate line.
1561 `parens` holds a set of string leaf values immediately after which
1562 invisible parens should be put.
1564 normalize_invisible_parens(node, parens_after=parens)
1565 for child in node.children:
1566 if child.type == token.NAME and child.value in keywords: # type: ignore
1567 yield from self.line()
1569 yield from self.visit(child)
1571 def visit_suite(self, node: Node) -> Iterator[Line]:
1572 """Visit a suite."""
1573 if self.is_pyi and is_stub_suite(node):
1574 yield from self.visit(node.children[2])
1576 yield from self.visit_default(node)
1578 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1579 """Visit a statement without nested statements."""
1580 is_suite_like = node.parent and node.parent.type in STATEMENT
1582 if self.is_pyi and is_stub_body(node):
1583 yield from self.visit_default(node)
1585 yield from self.line(+1)
1586 yield from self.visit_default(node)
1587 yield from self.line(-1)
1590 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1591 yield from self.line()
1592 yield from self.visit_default(node)
1594 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1595 """Visit `async def`, `async for`, `async with`."""
1596 yield from self.line()
1598 children = iter(node.children)
1599 for child in children:
1600 yield from self.visit(child)
1602 if child.type == token.ASYNC:
1605 internal_stmt = next(children)
1606 for child in internal_stmt.children:
1607 yield from self.visit(child)
1609 def visit_decorators(self, node: Node) -> Iterator[Line]:
1610 """Visit decorators."""
1611 for child in node.children:
1612 yield from self.line()
1613 yield from self.visit(child)
1615 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1616 """Remove a semicolon and put the other statement on a separate line."""
1617 yield from self.line()
1619 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1620 """End of file. Process outstanding comments and end with a newline."""
1621 yield from self.visit_default(leaf)
1622 yield from self.line()
1624 def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]:
1625 if not self.current_line.bracket_tracker.any_open_brackets():
1626 yield from self.line()
1627 yield from self.visit_default(leaf)
1629 def __attrs_post_init__(self) -> None:
1630 """You are in a twisty little maze of passages."""
1633 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1634 self.visit_if_stmt = partial(
1635 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1637 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1638 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1639 self.visit_try_stmt = partial(
1640 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1642 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1643 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1644 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1645 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1646 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1647 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1648 self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
1649 self.visit_del_stmt = partial(v, keywords=Ø, parens={"del"})
1650 self.visit_async_funcdef = self.visit_async_stmt
1651 self.visit_decorated = self.visit_decorators
1654 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1655 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1656 OPENING_BRACKETS = set(BRACKET.keys())
1657 CLOSING_BRACKETS = set(BRACKET.values())
1658 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1659 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1662 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901
1663 """Return whitespace prefix if needed for the given `leaf`.
1665 `complex_subscript` signals whether the given leaf is part of a subscription
1666 which has non-trivial arguments, like arithmetic expressions or function calls.
1674 if t in ALWAYS_NO_SPACE:
1677 if t == token.COMMENT:
1680 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1681 if t == token.COLON and p.type not in {
1688 prev = leaf.prev_sibling
1690 prevp = preceding_leaf(p)
1691 if not prevp or prevp.type in OPENING_BRACKETS:
1694 if t == token.COLON:
1695 if prevp.type == token.COLON:
1698 elif prevp.type != token.COMMA and not complex_subscript:
1703 if prevp.type == token.EQUAL:
1705 if prevp.parent.type in {
1713 elif prevp.parent.type == syms.typedargslist:
1714 # A bit hacky: if the equal sign has whitespace, it means we
1715 # previously found it's a typed argument. So, we're using
1719 elif prevp.type in STARS:
1720 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1723 elif prevp.type == token.COLON:
1724 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1725 return SPACE if complex_subscript else NO
1729 and prevp.parent.type == syms.factor
1730 and prevp.type in MATH_OPERATORS
1735 prevp.type == token.RIGHTSHIFT
1737 and prevp.parent.type == syms.shift_expr
1738 and prevp.prev_sibling
1739 and prevp.prev_sibling.type == token.NAME
1740 and prevp.prev_sibling.value == "print" # type: ignore
1742 # Python 2 print chevron
1745 elif prev.type in OPENING_BRACKETS:
1748 if p.type in {syms.parameters, syms.arglist}:
1749 # untyped function signatures or calls
1750 if not prev or prev.type != token.COMMA:
1753 elif p.type == syms.varargslist:
1755 if prev and prev.type != token.COMMA:
1758 elif p.type == syms.typedargslist:
1759 # typed function signatures
1763 if t == token.EQUAL:
1764 if prev.type != syms.tname:
1767 elif prev.type == token.EQUAL:
1768 # A bit hacky: if the equal sign has whitespace, it means we
1769 # previously found it's a typed argument. So, we're using that, too.
1772 elif prev.type != token.COMMA:
1775 elif p.type == syms.tname:
1778 prevp = preceding_leaf(p)
1779 if not prevp or prevp.type != token.COMMA:
1782 elif p.type == syms.trailer:
1783 # attributes and calls
1784 if t == token.LPAR or t == token.RPAR:
1789 prevp = preceding_leaf(p)
1790 if not prevp or prevp.type != token.NUMBER:
1793 elif t == token.LSQB:
1796 elif prev.type != token.COMMA:
1799 elif p.type == syms.argument:
1801 if t == token.EQUAL:
1805 prevp = preceding_leaf(p)
1806 if not prevp or prevp.type == token.LPAR:
1809 elif prev.type in {token.EQUAL} | STARS:
1812 elif p.type == syms.decorator:
1816 elif p.type == syms.dotted_name:
1820 prevp = preceding_leaf(p)
1821 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
1824 elif p.type == syms.classdef:
1828 if prev and prev.type == token.LPAR:
1831 elif p.type in {syms.subscript, syms.sliceop}:
1834 assert p.parent is not None, "subscripts are always parented"
1835 if p.parent.type == syms.subscriptlist:
1840 elif not complex_subscript:
1843 elif p.type == syms.atom:
1844 if prev and t == token.DOT:
1845 # dots, but not the first one.
1848 elif p.type == syms.dictsetmaker:
1850 if prev and prev.type == token.DOUBLESTAR:
1853 elif p.type in {syms.factor, syms.star_expr}:
1856 prevp = preceding_leaf(p)
1857 if not prevp or prevp.type in OPENING_BRACKETS:
1860 prevp_parent = prevp.parent
1861 assert prevp_parent is not None
1862 if prevp.type == token.COLON and prevp_parent.type in {
1868 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
1871 elif t in {token.NAME, token.NUMBER, token.STRING}:
1874 elif p.type == syms.import_from:
1876 if prev and prev.type == token.DOT:
1879 elif t == token.NAME:
1883 if prev and prev.type == token.DOT:
1886 elif p.type == syms.sliceop:
1892 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
1893 """Return the first leaf that precedes `node`, if any."""
1895 res = node.prev_sibling
1897 if isinstance(res, Leaf):
1901 return list(res.leaves())[-1]
1910 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
1911 """Return the child of `ancestor` that contains `descendant`."""
1912 node: Optional[LN] = descendant
1913 while node and node.parent != ancestor:
1918 def container_of(leaf: Leaf) -> LN:
1919 """Return `leaf` or one of its ancestors that is the topmost container of it.
1921 By "container" we mean a node where `leaf` is the very first child.
1923 same_prefix = leaf.prefix
1924 container: LN = leaf
1926 parent = container.parent
1930 if parent.children[0].prefix != same_prefix:
1933 if parent.type == syms.file_input:
1936 if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS:
1943 def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int:
1944 """Return the priority of the `leaf` delimiter, given a line break after it.
1946 The delimiter priorities returned here are from those delimiters that would
1947 cause a line break after themselves.
1949 Higher numbers are higher priority.
1951 if leaf.type == token.COMMA:
1952 return COMMA_PRIORITY
1957 def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int:
1958 """Return the priority of the `leaf` delimiter, given a line break before it.
1960 The delimiter priorities returned here are from those delimiters that would
1961 cause a line break before themselves.
1963 Higher numbers are higher priority.
1965 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1966 # * and ** might also be MATH_OPERATORS but in this case they are not.
1967 # Don't treat them as a delimiter.
1971 leaf.type == token.DOT
1973 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
1974 and (previous is None or previous.type in CLOSING_BRACKETS)
1979 leaf.type in MATH_OPERATORS
1981 and leaf.parent.type not in {syms.factor, syms.star_expr}
1983 return MATH_PRIORITIES[leaf.type]
1985 if leaf.type in COMPARATORS:
1986 return COMPARATOR_PRIORITY
1989 leaf.type == token.STRING
1990 and previous is not None
1991 and previous.type == token.STRING
1993 return STRING_PRIORITY
1995 if leaf.type not in {token.NAME, token.ASYNC}:
2001 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
2002 or leaf.type == token.ASYNC
2005 not isinstance(leaf.prev_sibling, Leaf)
2006 or leaf.prev_sibling.value != "async"
2008 return COMPREHENSION_PRIORITY
2013 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
2015 return COMPREHENSION_PRIORITY
2017 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
2018 return TERNARY_PRIORITY
2020 if leaf.value == "is":
2021 return COMPARATOR_PRIORITY
2026 and leaf.parent.type in {syms.comp_op, syms.comparison}
2028 previous is not None
2029 and previous.type == token.NAME
2030 and previous.value == "not"
2033 return COMPARATOR_PRIORITY
2038 and leaf.parent.type == syms.comp_op
2040 previous is not None
2041 and previous.type == token.NAME
2042 and previous.value == "is"
2045 return COMPARATOR_PRIORITY
2047 if leaf.value in LOGIC_OPERATORS and leaf.parent:
2048 return LOGIC_PRIORITY
2053 FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"}
2054 FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"}
2057 def generate_comments(leaf: LN) -> Iterator[Leaf]:
2058 """Clean the prefix of the `leaf` and generate comments from it, if any.
2060 Comments in lib2to3 are shoved into the whitespace prefix. This happens
2061 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
2062 move because it does away with modifying the grammar to include all the
2063 possible places in which comments can be placed.
2065 The sad consequence for us though is that comments don't "belong" anywhere.
2066 This is why this function generates simple parentless Leaf objects for
2067 comments. We simply don't know what the correct parent should be.
2069 No matter though, we can live without this. We really only need to
2070 differentiate between inline and standalone comments. The latter don't
2071 share the line with any code.
2073 Inline comments are emitted as regular token.COMMENT leaves. Standalone
2074 are emitted with a fake STANDALONE_COMMENT token identifier.
2076 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
2077 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
2082 """Describes a piece of syntax that is a comment.
2084 It's not a :class:`blib2to3.pytree.Leaf` so that:
2086 * it can be cached (`Leaf` objects should not be reused more than once as
2087 they store their lineno, column, prefix, and parent information);
2088 * `newlines` and `consumed` fields are kept separate from the `value`. This
2089 simplifies handling of special marker comments like ``# fmt: off/on``.
2092 type: int # token.COMMENT or STANDALONE_COMMENT
2093 value: str # content of the comment
2094 newlines: int # how many newlines before the comment
2095 consumed: int # how many characters of the original leaf's prefix did we consume
2098 @lru_cache(maxsize=4096)
2099 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
2100 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
2101 result: List[ProtoComment] = []
2102 if not prefix or "#" not in prefix:
2107 for index, line in enumerate(prefix.split("\n")):
2108 consumed += len(line) + 1 # adding the length of the split '\n'
2109 line = line.lstrip()
2112 if not line.startswith("#"):
2115 if index == 0 and not is_endmarker:
2116 comment_type = token.COMMENT # simple trailing comment
2118 comment_type = STANDALONE_COMMENT
2119 comment = make_comment(line)
2122 type=comment_type, value=comment, newlines=nlines, consumed=consumed
2129 def make_comment(content: str) -> str:
2130 """Return a consistently formatted comment from the given `content` string.
2132 All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single
2133 space between the hash sign and the content.
2135 If `content` didn't start with a hash sign, one is provided.
2137 content = content.rstrip()
2141 if content[0] == "#":
2142 content = content[1:]
2143 if content and content[0] not in " !:#'%":
2144 content = " " + content
2145 return "#" + content
2151 inner: bool = False,
2152 supports_trailing_commas: bool = False,
2153 ) -> Iterator[Line]:
2154 """Split a `line` into potentially many lines.
2156 They should fit in the allotted `line_length` but might not be able to.
2157 `inner` signifies that there were a pair of brackets somewhere around the
2158 current `line`, possibly transitively. This means we can fallback to splitting
2159 by delimiters if the LHS/RHS don't yield any results.
2161 If `supports_trailing_commas` is True, splitting may use the TRAILING_COMMA feature.
2167 line_str = str(line).strip("\n")
2169 # we don't want to split special comments like type annotations
2170 # https://github.com/python/typing/issues/186
2171 has_special_comment = False
2172 for leaf in line.leaves:
2173 for comment in line.comments_after(leaf):
2174 if leaf.type == token.COMMA and is_special_comment(comment):
2175 has_special_comment = True
2178 not has_special_comment
2179 and not line.should_explode
2180 and is_line_short_enough(line, line_length=line_length, line_str=line_str)
2185 split_funcs: List[SplitFunc]
2187 split_funcs = [left_hand_split]
2190 def rhs(line: Line, supports_trailing_commas: bool = False) -> Iterator[Line]:
2191 for omit in generate_trailers_to_omit(line, line_length):
2194 line, line_length, supports_trailing_commas, omit=omit
2197 if is_line_short_enough(lines[0], line_length=line_length):
2201 # All splits failed, best effort split with no omits.
2202 # This mostly happens to multiline strings that are by definition
2203 # reported as not fitting a single line.
2204 yield from right_hand_split(line, supports_trailing_commas)
2206 if line.inside_brackets:
2207 split_funcs = [delimiter_split, standalone_comment_split, rhs]
2210 for split_func in split_funcs:
2211 # We are accumulating lines in `result` because we might want to abort
2212 # mission and return the original line in the end, or attempt a different
2214 result: List[Line] = []
2216 for l in split_func(line, supports_trailing_commas):
2217 if str(l).strip("\n") == line_str:
2218 raise CannotSplit("Split function returned an unchanged result")
2223 line_length=line_length,
2225 supports_trailing_commas=supports_trailing_commas,
2239 def left_hand_split(
2240 line: Line, supports_trailing_commas: bool = False
2241 ) -> Iterator[Line]:
2242 """Split line into many lines, starting with the first matching bracket pair.
2244 Note: this usually looks weird, only use this for function definitions.
2245 Prefer RHS otherwise. This is why this function is not symmetrical with
2246 :func:`right_hand_split` which also handles optional parentheses.
2248 tail_leaves: List[Leaf] = []
2249 body_leaves: List[Leaf] = []
2250 head_leaves: List[Leaf] = []
2251 current_leaves = head_leaves
2252 matching_bracket = None
2253 for leaf in line.leaves:
2255 current_leaves is body_leaves
2256 and leaf.type in CLOSING_BRACKETS
2257 and leaf.opening_bracket is matching_bracket
2259 current_leaves = tail_leaves if body_leaves else head_leaves
2260 current_leaves.append(leaf)
2261 if current_leaves is head_leaves:
2262 if leaf.type in OPENING_BRACKETS:
2263 matching_bracket = leaf
2264 current_leaves = body_leaves
2265 if not matching_bracket:
2266 raise CannotSplit("No brackets found")
2268 head = bracket_split_build_line(head_leaves, line, matching_bracket)
2269 body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True)
2270 tail = bracket_split_build_line(tail_leaves, line, matching_bracket)
2271 bracket_split_succeeded_or_raise(head, body, tail)
2272 for result in (head, body, tail):
2277 def right_hand_split(
2280 supports_trailing_commas: bool = False,
2281 omit: Collection[LeafID] = (),
2282 ) -> Iterator[Line]:
2283 """Split line into many lines, starting with the last matching bracket pair.
2285 If the split was by optional parentheses, attempt splitting without them, too.
2286 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2289 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2291 tail_leaves: List[Leaf] = []
2292 body_leaves: List[Leaf] = []
2293 head_leaves: List[Leaf] = []
2294 current_leaves = tail_leaves
2295 opening_bracket = None
2296 closing_bracket = None
2297 for leaf in reversed(line.leaves):
2298 if current_leaves is body_leaves:
2299 if leaf is opening_bracket:
2300 current_leaves = head_leaves if body_leaves else tail_leaves
2301 current_leaves.append(leaf)
2302 if current_leaves is tail_leaves:
2303 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2304 opening_bracket = leaf.opening_bracket
2305 closing_bracket = leaf
2306 current_leaves = body_leaves
2307 if not (opening_bracket and closing_bracket and head_leaves):
2308 # If there is no opening or closing_bracket that means the split failed and
2309 # all content is in the tail. Otherwise, if `head_leaves` are empty, it means
2310 # the matching `opening_bracket` wasn't available on `line` anymore.
2311 raise CannotSplit("No brackets found")
2313 tail_leaves.reverse()
2314 body_leaves.reverse()
2315 head_leaves.reverse()
2316 head = bracket_split_build_line(head_leaves, line, opening_bracket)
2317 body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True)
2318 tail = bracket_split_build_line(tail_leaves, line, opening_bracket)
2319 bracket_split_succeeded_or_raise(head, body, tail)
2321 # the body shouldn't be exploded
2322 not body.should_explode
2323 # the opening bracket is an optional paren
2324 and opening_bracket.type == token.LPAR
2325 and not opening_bracket.value
2326 # the closing bracket is an optional paren
2327 and closing_bracket.type == token.RPAR
2328 and not closing_bracket.value
2329 # it's not an import (optional parens are the only thing we can split on
2330 # in this case; attempting a split without them is a waste of time)
2331 and not line.is_import
2332 # there are no standalone comments in the body
2333 and not body.contains_standalone_comments(0)
2334 # and we can actually remove the parens
2335 and can_omit_invisible_parens(body, line_length)
2337 omit = {id(closing_bracket), *omit}
2339 yield from right_hand_split(
2342 supports_trailing_commas=supports_trailing_commas,
2350 or is_line_short_enough(body, line_length=line_length)
2353 "Splitting failed, body is still too long and can't be split."
2356 elif head.contains_multiline_strings() or tail.contains_multiline_strings():
2358 "The current optional pair of parentheses is bound to fail to "
2359 "satisfy the splitting algorithm because the head or the tail "
2360 "contains multiline strings which by definition never fit one "
2364 ensure_visible(opening_bracket)
2365 ensure_visible(closing_bracket)
2366 for result in (head, body, tail):
2371 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2372 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2374 Do nothing otherwise.
2376 A left- or right-hand split is based on a pair of brackets. Content before
2377 (and including) the opening bracket is left on one line, content inside the
2378 brackets is put on a separate line, and finally content starting with and
2379 following the closing bracket is put on a separate line.
2381 Those are called `head`, `body`, and `tail`, respectively. If the split
2382 produced the same line (all content in `head`) or ended up with an empty `body`
2383 and the `tail` is just the closing bracket, then it's considered failed.
2385 tail_len = len(str(tail).strip())
2388 raise CannotSplit("Splitting brackets produced the same line")
2392 f"Splitting brackets on an empty body to save "
2393 f"{tail_len} characters is not worth it"
2397 def bracket_split_build_line(
2398 leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False
2400 """Return a new line with given `leaves` and respective comments from `original`.
2402 If `is_body` is True, the result line is one-indented inside brackets and as such
2403 has its first leaf's prefix normalized and a trailing comma added when expected.
2405 result = Line(depth=original.depth)
2407 result.inside_brackets = True
2410 # Since body is a new indent level, remove spurious leading whitespace.
2411 normalize_prefix(leaves[0], inside_brackets=True)
2412 # Ensure a trailing comma when expected.
2413 if original.is_import:
2414 if leaves[-1].type != token.COMMA:
2415 leaves.append(Leaf(token.COMMA, ","))
2418 result.append(leaf, preformatted=True)
2419 for comment_after in original.comments_after(leaf):
2420 result.append(comment_after, preformatted=True)
2422 result.should_explode = should_explode(result, opening_bracket)
2426 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2427 """Normalize prefix of the first leaf in every line returned by `split_func`.
2429 This is a decorator over relevant split functions.
2434 line: Line, supports_trailing_commas: bool = False
2435 ) -> Iterator[Line]:
2436 for l in split_func(line, supports_trailing_commas):
2437 normalize_prefix(l.leaves[0], inside_brackets=True)
2440 return split_wrapper
2443 @dont_increase_indentation
2444 def delimiter_split(
2445 line: Line, supports_trailing_commas: bool = False
2446 ) -> Iterator[Line]:
2447 """Split according to delimiters of the highest priority.
2449 If `py36` is True, the split will add trailing commas also in function
2450 signatures that contain `*` and `**`.
2453 last_leaf = line.leaves[-1]
2455 raise CannotSplit("Line empty")
2457 bt = line.bracket_tracker
2459 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2461 raise CannotSplit("No delimiters found")
2463 if delimiter_priority == DOT_PRIORITY:
2464 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2465 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2467 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2468 lowest_depth = sys.maxsize
2469 trailing_comma_safe = True
2471 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2472 """Append `leaf` to current line or to new line if appending impossible."""
2473 nonlocal current_line
2475 current_line.append_safe(leaf, preformatted=True)
2479 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2480 current_line.append(leaf)
2482 for leaf in line.leaves:
2483 yield from append_to_line(leaf)
2485 for comment_after in line.comments_after(leaf):
2486 yield from append_to_line(comment_after)
2488 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2489 if leaf.bracket_depth == lowest_depth and is_vararg(
2490 leaf, within=VARARGS_PARENTS
2492 trailing_comma_safe = trailing_comma_safe and supports_trailing_commas
2493 leaf_priority = bt.delimiters.get(id(leaf))
2494 if leaf_priority == delimiter_priority:
2497 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2501 and delimiter_priority == COMMA_PRIORITY
2502 and current_line.leaves[-1].type != token.COMMA
2503 and current_line.leaves[-1].type != STANDALONE_COMMENT
2505 current_line.append(Leaf(token.COMMA, ","))
2509 @dont_increase_indentation
2510 def standalone_comment_split(
2511 line: Line, supports_trailing_commas: bool = False
2512 ) -> Iterator[Line]:
2513 """Split standalone comments from the rest of the line."""
2514 if not line.contains_standalone_comments(0):
2515 raise CannotSplit("Line does not have any standalone comments")
2517 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2519 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2520 """Append `leaf` to current line or to new line if appending impossible."""
2521 nonlocal current_line
2523 current_line.append_safe(leaf, preformatted=True)
2527 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2528 current_line.append(leaf)
2530 for leaf in line.leaves:
2531 yield from append_to_line(leaf)
2533 for comment_after in line.comments_after(leaf):
2534 yield from append_to_line(comment_after)
2540 def is_import(leaf: Leaf) -> bool:
2541 """Return True if the given leaf starts an import statement."""
2548 (v == "import" and p and p.type == syms.import_name)
2549 or (v == "from" and p and p.type == syms.import_from)
2554 def is_special_comment(leaf: Leaf) -> bool:
2555 """Return True if the given leaf is a special comment.
2556 Only returns true for type comments for now."""
2560 (t == token.COMMENT or t == STANDALONE_COMMENT) and (v.startswith("# type:"))
2564 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2565 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2568 Note: don't use backslashes for formatting or you'll lose your voting rights.
2570 if not inside_brackets:
2571 spl = leaf.prefix.split("#")
2572 if "\\" not in spl[0]:
2573 nl_count = spl[-1].count("\n")
2576 leaf.prefix = "\n" * nl_count
2582 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2583 """Make all string prefixes lowercase.
2585 If remove_u_prefix is given, also removes any u prefix from the string.
2587 Note: Mutates its argument.
2589 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2590 assert match is not None, f"failed to match string {leaf.value!r}"
2591 orig_prefix = match.group(1)
2592 new_prefix = orig_prefix.lower()
2594 new_prefix = new_prefix.replace("u", "")
2595 leaf.value = f"{new_prefix}{match.group(2)}"
2598 def normalize_string_quotes(leaf: Leaf) -> None:
2599 """Prefer double quotes but only if it doesn't cause more escaping.
2601 Adds or removes backslashes as appropriate. Doesn't parse and fix
2602 strings nested in f-strings (yet).
2604 Note: Mutates its argument.
2606 value = leaf.value.lstrip("furbFURB")
2607 if value[:3] == '"""':
2610 elif value[:3] == "'''":
2613 elif value[0] == '"':
2619 first_quote_pos = leaf.value.find(orig_quote)
2620 if first_quote_pos == -1:
2621 return # There's an internal error
2623 prefix = leaf.value[:first_quote_pos]
2624 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2625 escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}")
2626 escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}")
2627 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2628 if "r" in prefix.casefold():
2629 if unescaped_new_quote.search(body):
2630 # There's at least one unescaped new_quote in this raw string
2631 # so converting is impossible
2634 # Do not introduce or remove backslashes in raw strings
2637 # remove unnecessary escapes
2638 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2639 if body != new_body:
2640 # Consider the string without unnecessary escapes as the original
2642 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2643 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2644 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2645 if "f" in prefix.casefold():
2646 matches = re.findall(r"[^{]\{(.*?)\}[^}]", new_body)
2649 # Do not introduce backslashes in interpolated expressions
2651 if new_quote == '"""' and new_body[-1:] == '"':
2653 new_body = new_body[:-1] + '\\"'
2654 orig_escape_count = body.count("\\")
2655 new_escape_count = new_body.count("\\")
2656 if new_escape_count > orig_escape_count:
2657 return # Do not introduce more escaping
2659 if new_escape_count == orig_escape_count and orig_quote == '"':
2660 return # Prefer double quotes
2662 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2665 def normalize_numeric_literal(leaf: Leaf) -> None:
2666 """Normalizes numeric (float, int, and complex) literals.
2668 All letters used in the representation are normalized to lowercase (except
2669 in Python 2 long literals).
2671 text = leaf.value.lower()
2672 if text.startswith(("0o", "0b")):
2673 # Leave octal and binary literals alone.
2675 elif text.startswith("0x"):
2676 # Change hex literals to upper case.
2677 before, after = text[:2], text[2:]
2678 text = f"{before}{after.upper()}"
2680 before, after = text.split("e")
2682 if after.startswith("-"):
2685 elif after.startswith("+"):
2687 before = format_float_or_int_string(before)
2688 text = f"{before}e{sign}{after}"
2689 elif text.endswith(("j", "l")):
2692 # Capitalize in "2L" because "l" looks too similar to "1".
2695 text = f"{format_float_or_int_string(number)}{suffix}"
2697 text = format_float_or_int_string(text)
2701 def format_float_or_int_string(text: str) -> str:
2702 """Formats a float string like "1.0"."""
2706 before, after = text.split(".")
2707 return f"{before or 0}.{after or 0}"
2710 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2711 """Make existing optional parentheses invisible or create new ones.
2713 `parens_after` is a set of string leaf values immeditely after which parens
2716 Standardizes on visible parentheses for single-element tuples, and keeps
2717 existing visible parentheses for other tuples and generator expressions.
2719 for pc in list_comments(node.prefix, is_endmarker=False):
2720 if pc.value in FMT_OFF:
2721 # This `node` has a prefix with `# fmt: off`, don't mess with parens.
2725 for index, child in enumerate(list(node.children)):
2727 if child.type == syms.atom:
2728 if maybe_make_parens_invisible_in_atom(child):
2729 lpar = Leaf(token.LPAR, "")
2730 rpar = Leaf(token.RPAR, "")
2731 index = child.remove() or 0
2732 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2733 elif is_one_tuple(child):
2734 # wrap child in visible parentheses
2735 lpar = Leaf(token.LPAR, "(")
2736 rpar = Leaf(token.RPAR, ")")
2738 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2739 elif node.type == syms.import_from:
2740 # "import from" nodes store parentheses directly as part of
2742 if child.type == token.LPAR:
2743 # make parentheses invisible
2744 child.value = "" # type: ignore
2745 node.children[-1].value = "" # type: ignore
2746 elif child.type != token.STAR:
2747 # insert invisible parentheses
2748 node.insert_child(index, Leaf(token.LPAR, ""))
2749 node.append_child(Leaf(token.RPAR, ""))
2752 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
2753 # wrap child in invisible parentheses
2754 lpar = Leaf(token.LPAR, "")
2755 rpar = Leaf(token.RPAR, "")
2756 index = child.remove() or 0
2757 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2759 check_lpar = isinstance(child, Leaf) and child.value in parens_after
2762 def normalize_fmt_off(node: Node) -> None:
2763 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
2766 try_again = convert_one_fmt_off_pair(node)
2769 def convert_one_fmt_off_pair(node: Node) -> bool:
2770 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
2772 Returns True if a pair was converted.
2774 for leaf in node.leaves():
2775 previous_consumed = 0
2776 for comment in list_comments(leaf.prefix, is_endmarker=False):
2777 if comment.value in FMT_OFF:
2778 # We only want standalone comments. If there's no previous leaf or
2779 # the previous leaf is indentation, it's a standalone comment in
2781 if comment.type != STANDALONE_COMMENT:
2782 prev = preceding_leaf(leaf)
2783 if prev and prev.type not in WHITESPACE:
2786 ignored_nodes = list(generate_ignored_nodes(leaf))
2787 if not ignored_nodes:
2790 first = ignored_nodes[0] # Can be a container node with the `leaf`.
2791 parent = first.parent
2792 prefix = first.prefix
2793 first.prefix = prefix[comment.consumed :]
2795 comment.value + "\n" + "".join(str(n) for n in ignored_nodes)
2797 if hidden_value.endswith("\n"):
2798 # That happens when one of the `ignored_nodes` ended with a NEWLINE
2799 # leaf (possibly followed by a DEDENT).
2800 hidden_value = hidden_value[:-1]
2802 for ignored in ignored_nodes:
2803 index = ignored.remove()
2804 if first_idx is None:
2806 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
2807 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
2808 parent.insert_child(
2813 prefix=prefix[:previous_consumed] + "\n" * comment.newlines,
2818 previous_consumed = comment.consumed
2823 def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]:
2824 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
2826 Stops at the end of the block.
2828 container: Optional[LN] = container_of(leaf)
2829 while container is not None and container.type != token.ENDMARKER:
2830 for comment in list_comments(container.prefix, is_endmarker=False):
2831 if comment.value in FMT_ON:
2836 container = container.next_sibling
2839 def maybe_make_parens_invisible_in_atom(node: LN) -> bool:
2840 """If it's safe, make the parens in the atom `node` invisible, recursively.
2842 Returns whether the node should itself be wrapped in invisible parentheses.
2846 node.type != syms.atom
2847 or is_empty_tuple(node)
2848 or is_one_tuple(node)
2850 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
2854 first = node.children[0]
2855 last = node.children[-1]
2856 if first.type == token.LPAR and last.type == token.RPAR:
2857 # make parentheses invisible
2858 first.value = "" # type: ignore
2859 last.value = "" # type: ignore
2860 if len(node.children) > 1:
2861 maybe_make_parens_invisible_in_atom(node.children[1])
2867 def is_empty_tuple(node: LN) -> bool:
2868 """Return True if `node` holds an empty tuple."""
2870 node.type == syms.atom
2871 and len(node.children) == 2
2872 and node.children[0].type == token.LPAR
2873 and node.children[1].type == token.RPAR
2877 def is_one_tuple(node: LN) -> bool:
2878 """Return True if `node` holds a tuple with one element, with or without parens."""
2879 if node.type == syms.atom:
2880 if len(node.children) != 3:
2883 lpar, gexp, rpar = node.children
2885 lpar.type == token.LPAR
2886 and gexp.type == syms.testlist_gexp
2887 and rpar.type == token.RPAR
2891 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
2894 node.type in IMPLICIT_TUPLE
2895 and len(node.children) == 2
2896 and node.children[1].type == token.COMMA
2900 def is_yield(node: LN) -> bool:
2901 """Return True if `node` holds a `yield` or `yield from` expression."""
2902 if node.type == syms.yield_expr:
2905 if node.type == token.NAME and node.value == "yield": # type: ignore
2908 if node.type != syms.atom:
2911 if len(node.children) != 3:
2914 lpar, expr, rpar = node.children
2915 if lpar.type == token.LPAR and rpar.type == token.RPAR:
2916 return is_yield(expr)
2921 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
2922 """Return True if `leaf` is a star or double star in a vararg or kwarg.
2924 If `within` includes VARARGS_PARENTS, this applies to function signatures.
2925 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
2926 extended iterable unpacking (PEP 3132) and additional unpacking
2927 generalizations (PEP 448).
2929 if leaf.type not in STARS or not leaf.parent:
2933 if p.type == syms.star_expr:
2934 # Star expressions are also used as assignment targets in extended
2935 # iterable unpacking (PEP 3132). See what its parent is instead.
2941 return p.type in within
2944 def is_multiline_string(leaf: Leaf) -> bool:
2945 """Return True if `leaf` is a multiline string that actually spans many lines."""
2946 value = leaf.value.lstrip("furbFURB")
2947 return value[:3] in {'"""', "'''"} and "\n" in value
2950 def is_stub_suite(node: Node) -> bool:
2951 """Return True if `node` is a suite with a stub body."""
2953 len(node.children) != 4
2954 or node.children[0].type != token.NEWLINE
2955 or node.children[1].type != token.INDENT
2956 or node.children[3].type != token.DEDENT
2960 return is_stub_body(node.children[2])
2963 def is_stub_body(node: LN) -> bool:
2964 """Return True if `node` is a simple statement containing an ellipsis."""
2965 if not isinstance(node, Node) or node.type != syms.simple_stmt:
2968 if len(node.children) != 2:
2971 child = node.children[0]
2973 child.type == syms.atom
2974 and len(child.children) == 3
2975 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
2979 def max_delimiter_priority_in_atom(node: LN) -> int:
2980 """Return maximum delimiter priority inside `node`.
2982 This is specific to atoms with contents contained in a pair of parentheses.
2983 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
2985 if node.type != syms.atom:
2988 first = node.children[0]
2989 last = node.children[-1]
2990 if not (first.type == token.LPAR and last.type == token.RPAR):
2993 bt = BracketTracker()
2994 for c in node.children[1:-1]:
2995 if isinstance(c, Leaf):
2998 for leaf in c.leaves():
3001 return bt.max_delimiter_priority()
3007 def ensure_visible(leaf: Leaf) -> None:
3008 """Make sure parentheses are visible.
3010 They could be invisible as part of some statements (see
3011 :func:`normalize_invible_parens` and :func:`visit_import_from`).
3013 if leaf.type == token.LPAR:
3015 elif leaf.type == token.RPAR:
3019 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
3020 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
3023 opening_bracket.parent
3024 and opening_bracket.parent.type in {syms.atom, syms.import_from}
3025 and opening_bracket.value in "[{("
3030 last_leaf = line.leaves[-1]
3031 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
3032 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
3033 except (IndexError, ValueError):
3036 return max_priority == COMMA_PRIORITY
3039 def get_features_used(node: Node) -> Set[Feature]:
3040 """Return a set of (relatively) new Python features used in this file.
3042 Currently looking for:
3044 - underscores in numeric literals; and
3045 - trailing commas after * or ** in function signatures and calls.
3047 features: Set[Feature] = set()
3048 for n in node.pre_order():
3049 if n.type == token.STRING:
3050 value_head = n.value[:2] # type: ignore
3051 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
3052 features.add(Feature.F_STRINGS)
3054 elif n.type == token.NUMBER:
3055 if "_" in n.value: # type: ignore
3056 features.add(Feature.NUMERIC_UNDERSCORES)
3059 n.type in {syms.typedargslist, syms.arglist}
3061 and n.children[-1].type == token.COMMA
3063 for ch in n.children:
3064 if ch.type in STARS:
3065 features.add(Feature.TRAILING_COMMA)
3067 if ch.type == syms.argument:
3068 for argch in ch.children:
3069 if argch.type in STARS:
3070 features.add(Feature.TRAILING_COMMA)
3075 def detect_target_versions(node: Node) -> Set[TargetVersion]:
3076 """Detect the version to target based on the nodes used."""
3077 features = get_features_used(node)
3079 version for version in TargetVersion if features <= VERSION_TO_FEATURES[version]
3083 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
3084 """Generate sets of closing bracket IDs that should be omitted in a RHS.
3086 Brackets can be omitted if the entire trailer up to and including
3087 a preceding closing bracket fits in one line.
3089 Yielded sets are cumulative (contain results of previous yields, too). First
3093 omit: Set[LeafID] = set()
3096 length = 4 * line.depth
3097 opening_bracket = None
3098 closing_bracket = None
3099 inner_brackets: Set[LeafID] = set()
3100 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
3101 length += leaf_length
3102 if length > line_length:
3105 has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
3106 if leaf.type == STANDALONE_COMMENT or has_inline_comment:
3110 if leaf is opening_bracket:
3111 opening_bracket = None
3112 elif leaf.type in CLOSING_BRACKETS:
3113 inner_brackets.add(id(leaf))
3114 elif leaf.type in CLOSING_BRACKETS:
3115 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
3116 # Empty brackets would fail a split so treat them as "inner"
3117 # brackets (e.g. only add them to the `omit` set if another
3118 # pair of brackets was good enough.
3119 inner_brackets.add(id(leaf))
3123 omit.add(id(closing_bracket))
3124 omit.update(inner_brackets)
3125 inner_brackets.clear()
3129 opening_bracket = leaf.opening_bracket
3130 closing_bracket = leaf
3133 def get_future_imports(node: Node) -> Set[str]:
3134 """Return a set of __future__ imports in the file."""
3135 imports: Set[str] = set()
3137 def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]:
3138 for child in children:
3139 if isinstance(child, Leaf):
3140 if child.type == token.NAME:
3142 elif child.type == syms.import_as_name:
3143 orig_name = child.children[0]
3144 assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
3145 assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
3146 yield orig_name.value
3147 elif child.type == syms.import_as_names:
3148 yield from get_imports_from_children(child.children)
3150 assert False, "Invalid syntax parsing imports"
3152 for child in node.children:
3153 if child.type != syms.simple_stmt:
3155 first_child = child.children[0]
3156 if isinstance(first_child, Leaf):
3157 # Continue looking if we see a docstring; otherwise stop.
3159 len(child.children) == 2
3160 and first_child.type == token.STRING
3161 and child.children[1].type == token.NEWLINE
3166 elif first_child.type == syms.import_from:
3167 module_name = first_child.children[1]
3168 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
3170 imports |= set(get_imports_from_children(first_child.children[3:]))
3176 def gen_python_files_in_dir(
3179 include: Pattern[str],
3180 exclude: Pattern[str],
3182 ) -> Iterator[Path]:
3183 """Generate all files under `path` whose paths are not excluded by the
3184 `exclude` regex, but are included by the `include` regex.
3186 Symbolic links pointing outside of the `root` directory are ignored.
3188 `report` is where output about exclusions goes.
3190 assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
3191 for child in path.iterdir():
3193 normalized_path = "/" + child.resolve().relative_to(root).as_posix()
3195 if child.is_symlink():
3196 report.path_ignored(
3197 child, f"is a symbolic link that points outside {root}"
3204 normalized_path += "/"
3205 exclude_match = exclude.search(normalized_path)
3206 if exclude_match and exclude_match.group(0):
3207 report.path_ignored(child, f"matches the --exclude regular expression")
3211 yield from gen_python_files_in_dir(child, root, include, exclude, report)
3213 elif child.is_file():
3214 include_match = include.search(normalized_path)
3220 def find_project_root(srcs: Iterable[str]) -> Path:
3221 """Return a directory containing .git, .hg, or pyproject.toml.
3223 That directory can be one of the directories passed in `srcs` or their
3226 If no directory in the tree contains a marker that would specify it's the
3227 project root, the root of the file system is returned.
3230 return Path("/").resolve()
3232 common_base = min(Path(src).resolve() for src in srcs)
3233 if common_base.is_dir():
3234 # Append a fake file so `parents` below returns `common_base_dir`, too.
3235 common_base /= "fake-file"
3236 for directory in common_base.parents:
3237 if (directory / ".git").is_dir():
3240 if (directory / ".hg").is_dir():
3243 if (directory / "pyproject.toml").is_file():
3251 """Provides a reformatting counter. Can be rendered with `str(report)`."""
3255 verbose: bool = False
3256 change_count: int = 0
3258 failure_count: int = 0
3260 def done(self, src: Path, changed: Changed) -> None:
3261 """Increment the counter for successful reformatting. Write out a message."""
3262 if changed is Changed.YES:
3263 reformatted = "would reformat" if self.check else "reformatted"
3264 if self.verbose or not self.quiet:
3265 out(f"{reformatted} {src}")
3266 self.change_count += 1
3269 if changed is Changed.NO:
3270 msg = f"{src} already well formatted, good job."
3272 msg = f"{src} wasn't modified on disk since last run."
3273 out(msg, bold=False)
3274 self.same_count += 1
3276 def failed(self, src: Path, message: str) -> None:
3277 """Increment the counter for failed reformatting. Write out a message."""
3278 err(f"error: cannot format {src}: {message}")
3279 self.failure_count += 1
3281 def path_ignored(self, path: Path, message: str) -> None:
3283 out(f"{path} ignored: {message}", bold=False)
3286 def return_code(self) -> int:
3287 """Return the exit code that the app should use.
3289 This considers the current state of changed files and failures:
3290 - if there were any failures, return 123;
3291 - if any files were changed and --check is being used, return 1;
3292 - otherwise return 0.
3294 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
3295 # 126 we have special return codes reserved by the shell.
3296 if self.failure_count:
3299 elif self.change_count and self.check:
3304 def __str__(self) -> str:
3305 """Render a color report of the current state.
3307 Use `click.unstyle` to remove colors.
3310 reformatted = "would be reformatted"
3311 unchanged = "would be left unchanged"
3312 failed = "would fail to reformat"
3314 reformatted = "reformatted"
3315 unchanged = "left unchanged"
3316 failed = "failed to reformat"
3318 if self.change_count:
3319 s = "s" if self.change_count > 1 else ""
3321 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
3324 s = "s" if self.same_count > 1 else ""
3325 report.append(f"{self.same_count} file{s} {unchanged}")
3326 if self.failure_count:
3327 s = "s" if self.failure_count > 1 else ""
3329 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
3331 return ", ".join(report) + "."
3334 def assert_equivalent(src: str, dst: str) -> None:
3335 """Raise AssertionError if `src` and `dst` aren't equivalent."""
3340 def _v(node: ast.AST, depth: int = 0) -> Iterator[str]:
3341 """Simple visitor generating strings to compare ASTs by content."""
3342 yield f"{' ' * depth}{node.__class__.__name__}("
3344 for field in sorted(node._fields):
3346 value = getattr(node, field)
3347 except AttributeError:
3350 yield f"{' ' * (depth+1)}{field}="
3352 if isinstance(value, list):
3354 # Ignore nested tuples within del statements, because we may insert
3355 # parentheses and they change the AST.
3358 and isinstance(node, ast.Delete)
3359 and isinstance(item, ast.Tuple)
3361 for item in item.elts:
3362 yield from _v(item, depth + 2)
3363 elif isinstance(item, ast.AST):
3364 yield from _v(item, depth + 2)
3366 elif isinstance(value, ast.AST):
3367 yield from _v(value, depth + 2)
3370 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
3372 yield f"{' ' * depth}) # /{node.__class__.__name__}"
3375 src_ast = ast.parse(src)
3376 except Exception as exc:
3377 major, minor = sys.version_info[:2]
3378 raise AssertionError(
3379 f"cannot use --safe with this file; failed to parse source file "
3380 f"with Python {major}.{minor}'s builtin AST. Re-run with --fast "
3381 f"or stop using deprecated Python 2 syntax. AST error message: {exc}"
3385 dst_ast = ast.parse(dst)
3386 except Exception as exc:
3387 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
3388 raise AssertionError(
3389 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
3390 f"Please report a bug on https://github.com/ambv/black/issues. "
3391 f"This invalid output might be helpful: {log}"
3394 src_ast_str = "\n".join(_v(src_ast))
3395 dst_ast_str = "\n".join(_v(dst_ast))
3396 if src_ast_str != dst_ast_str:
3397 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
3398 raise AssertionError(
3399 f"INTERNAL ERROR: Black produced code that is not equivalent to "
3401 f"Please report a bug on https://github.com/ambv/black/issues. "
3402 f"This diff might be helpful: {log}"
3406 def assert_stable(src: str, dst: str, mode: FileMode) -> None:
3407 """Raise AssertionError if `dst` reformats differently the second time."""
3408 newdst = format_str(dst, mode=mode)
3411 diff(src, dst, "source", "first pass"),
3412 diff(dst, newdst, "first pass", "second pass"),
3414 raise AssertionError(
3415 f"INTERNAL ERROR: Black produced different code on the second pass "
3416 f"of the formatter. "
3417 f"Please report a bug on https://github.com/ambv/black/issues. "
3418 f"This diff might be helpful: {log}"
3422 def dump_to_file(*output: str) -> str:
3423 """Dump `output` to a temporary file. Return path to the file."""
3426 with tempfile.NamedTemporaryFile(
3427 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
3429 for lines in output:
3431 if lines and lines[-1] != "\n":
3436 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
3437 """Return a unified diff string between strings `a` and `b`."""
3440 a_lines = [line + "\n" for line in a.split("\n")]
3441 b_lines = [line + "\n" for line in b.split("\n")]
3443 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
3447 def cancel(tasks: Iterable[asyncio.Task]) -> None:
3448 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
3454 def shutdown(loop: BaseEventLoop) -> None:
3455 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
3457 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
3458 to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()]
3462 for task in to_cancel:
3464 loop.run_until_complete(
3465 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
3468 # `concurrent.futures.Future` objects cannot be cancelled once they
3469 # are already running. There might be some when the `shutdown()` happened.
3470 # Silence their logger's spew about the event loop being closed.
3471 cf_logger = logging.getLogger("concurrent.futures")
3472 cf_logger.setLevel(logging.CRITICAL)
3476 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
3477 """Replace `regex` with `replacement` twice on `original`.
3479 This is used by string normalization to perform replaces on
3480 overlapping matches.
3482 return regex.sub(replacement, regex.sub(replacement, original))
3485 def re_compile_maybe_verbose(regex: str) -> Pattern[str]:
3486 """Compile a regular expression string in `regex`.
3488 If it contains newlines, use verbose mode.
3491 regex = "(?x)" + regex
3492 return re.compile(regex)
3495 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
3496 """Like `reversed(enumerate(sequence))` if that were possible."""
3497 index = len(sequence) - 1
3498 for element in reversed(sequence):
3499 yield (index, element)
3503 def enumerate_with_length(
3504 line: Line, reversed: bool = False
3505 ) -> Iterator[Tuple[Index, Leaf, int]]:
3506 """Return an enumeration of leaves with their length.
3508 Stops prematurely on multiline strings and standalone comments.
3511 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
3512 enumerate_reversed if reversed else enumerate,
3514 for index, leaf in op(line.leaves):
3515 length = len(leaf.prefix) + len(leaf.value)
3516 if "\n" in leaf.value:
3517 return # Multiline strings, we can't continue.
3519 comment: Optional[Leaf]
3520 for comment in line.comments_after(leaf):
3521 length += len(comment.value)
3523 yield index, leaf, length
3526 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
3527 """Return True if `line` is no longer than `line_length`.
3529 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
3532 line_str = str(line).strip("\n")
3534 len(line_str) <= line_length
3535 and "\n" not in line_str # multiline strings
3536 and not line.contains_standalone_comments()
3540 def can_be_split(line: Line) -> bool:
3541 """Return False if the line cannot be split *for sure*.
3543 This is not an exhaustive search but a cheap heuristic that we can use to
3544 avoid some unfortunate formattings (mostly around wrapping unsplittable code
3545 in unnecessary parentheses).
3547 leaves = line.leaves
3551 if leaves[0].type == token.STRING and leaves[1].type == token.DOT:
3555 for leaf in leaves[-2::-1]:
3556 if leaf.type in OPENING_BRACKETS:
3557 if next.type not in CLOSING_BRACKETS:
3561 elif leaf.type == token.DOT:
3563 elif leaf.type == token.NAME:
3564 if not (next.type == token.DOT or next.type in OPENING_BRACKETS):
3567 elif leaf.type not in CLOSING_BRACKETS:
3570 if dot_count > 1 and call_count > 1:
3576 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
3577 """Does `line` have a shape safe to reformat without optional parens around it?
3579 Returns True for only a subset of potentially nice looking formattings but
3580 the point is to not return false positives that end up producing lines that
3583 bt = line.bracket_tracker
3584 if not bt.delimiters:
3585 # Without delimiters the optional parentheses are useless.
3588 max_priority = bt.max_delimiter_priority()
3589 if bt.delimiter_count_with_priority(max_priority) > 1:
3590 # With more than one delimiter of a kind the optional parentheses read better.
3593 if max_priority == DOT_PRIORITY:
3594 # A single stranded method call doesn't require optional parentheses.
3597 assert len(line.leaves) >= 2, "Stranded delimiter"
3599 first = line.leaves[0]
3600 second = line.leaves[1]
3601 penultimate = line.leaves[-2]
3602 last = line.leaves[-1]
3604 # With a single delimiter, omit if the expression starts or ends with
3606 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
3608 length = 4 * line.depth
3609 for _index, leaf, leaf_length in enumerate_with_length(line):
3610 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
3613 length += leaf_length
3614 if length > line_length:
3617 if leaf.type in OPENING_BRACKETS:
3618 # There are brackets we can further split on.
3622 # checked the entire string and line length wasn't exceeded
3623 if len(line.leaves) == _index + 1:
3626 # Note: we are not returning False here because a line might have *both*
3627 # a leading opening bracket and a trailing closing bracket. If the
3628 # opening bracket doesn't match our rule, maybe the closing will.
3631 last.type == token.RPAR
3632 or last.type == token.RBRACE
3634 # don't use indexing for omitting optional parentheses;
3636 last.type == token.RSQB
3638 and last.parent.type != syms.trailer
3641 if penultimate.type in OPENING_BRACKETS:
3642 # Empty brackets don't help.
3645 if is_multiline_string(first):
3646 # Additional wrapping of a multiline string in this situation is
3650 length = 4 * line.depth
3651 seen_other_brackets = False
3652 for _index, leaf, leaf_length in enumerate_with_length(line):
3653 length += leaf_length
3654 if leaf is last.opening_bracket:
3655 if seen_other_brackets or length <= line_length:
3658 elif leaf.type in OPENING_BRACKETS:
3659 # There are brackets we can further split on.
3660 seen_other_brackets = True
3665 def get_cache_file(mode: FileMode) -> Path:
3666 return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle"
3669 def read_cache(mode: FileMode) -> Cache:
3670 """Read the cache if it exists and is well formed.
3672 If it is not well formed, the call to write_cache later should resolve the issue.
3674 cache_file = get_cache_file(mode)
3675 if not cache_file.exists():
3678 with cache_file.open("rb") as fobj:
3680 cache: Cache = pickle.load(fobj)
3681 except pickle.UnpicklingError:
3687 def get_cache_info(path: Path) -> CacheInfo:
3688 """Return the information used to check if a file is already formatted or not."""
3690 return stat.st_mtime, stat.st_size
3693 def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]:
3694 """Split an iterable of paths in `sources` into two sets.
3696 The first contains paths of files that modified on disk or are not in the
3697 cache. The other contains paths to non-modified files.
3699 todo, done = set(), set()
3702 if cache.get(src) != get_cache_info(src):
3709 def write_cache(cache: Cache, sources: Iterable[Path], mode: FileMode) -> None:
3710 """Update the cache file."""
3711 cache_file = get_cache_file(mode)
3713 CACHE_DIR.mkdir(parents=True, exist_ok=True)
3714 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
3715 with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
3716 pickle.dump(new_cache, f, protocol=pickle.HIGHEST_PROTOCOL)
3717 os.replace(f.name, cache_file)
3722 def patch_click() -> None:
3723 """Make Click not crash.
3725 On certain misconfigured environments, Python 3 selects the ASCII encoding as the
3726 default which restricts paths that it can access during the lifetime of the
3727 application. Click refuses to work in this scenario by raising a RuntimeError.
3729 In case of Black the likelihood that non-ASCII characters are going to be used in
3730 file paths is minimal since it's Python source code. Moreover, this crash was
3731 spurious on Python 3.7 thanks to PEP 538 and PEP 540.
3734 from click import core
3735 from click import _unicodefun # type: ignore
3736 except ModuleNotFoundError:
3739 for module in (core, _unicodefun):
3740 if hasattr(module, "_verify_python3_env"):
3741 module._verify_python3_env = lambda: None
3744 def patched_main() -> None:
3750 if __name__ == "__main__":