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.
3 from concurrent.futures import Executor, ProcessPoolExecutor
4 from contextlib import contextmanager
5 from datetime import datetime
7 from functools import lru_cache, partial, wraps
11 from multiprocessing import Manager, freeze_support
13 from pathlib import Path
40 from typing_extensions import Final
41 from mypy_extensions import mypyc_attr
43 from appdirs import user_cache_dir
44 from dataclasses import dataclass, field, replace
47 from typed_ast import ast3, ast27
48 from pathspec import PathSpec
51 from blib2to3.pytree import Node, Leaf, type_repr
52 from blib2to3 import pygram, pytree
53 from blib2to3.pgen2 import driver, token
54 from blib2to3.pgen2.grammar import Grammar
55 from blib2to3.pgen2.parse import ParseError
57 from _black_version import version as __version__
59 DEFAULT_LINE_LENGTH = 88
60 DEFAULT_EXCLUDES = r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|\.svn|_build|buck-out|build|dist)/" # noqa: B950
61 DEFAULT_INCLUDES = r"\.pyi?$"
62 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
74 LN = Union[Leaf, Node]
75 SplitFunc = Callable[["Line", Collection["Feature"]], Iterator["Line"]]
78 CacheInfo = Tuple[Timestamp, FileSize]
79 Cache = Dict[Path, CacheInfo]
80 out = partial(click.secho, bold=True, err=True)
81 err = partial(click.secho, fg="red", err=True)
83 pygram.initialize(CACHE_DIR)
84 syms = pygram.python_symbols
87 class NothingChanged(UserWarning):
88 """Raised when reformatted code is the same as source."""
91 class CannotSplit(Exception):
92 """A readable split that fits the allotted line length is impossible."""
95 class InvalidInput(ValueError):
96 """Raised when input source code fails all parse attempts."""
99 class WriteBack(Enum):
106 def from_configuration(cls, *, check: bool, diff: bool) -> "WriteBack":
107 if check and not diff:
110 return cls.DIFF if diff else cls.YES
119 class TargetVersion(Enum):
128 def is_python2(self) -> bool:
129 return self is TargetVersion.PY27
132 PY36_VERSIONS = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38}
136 # All string literals are unicode
139 NUMERIC_UNDERSCORES = 3
140 TRAILING_COMMA_IN_CALL = 4
141 TRAILING_COMMA_IN_DEF = 5
142 # The following two feature-flags are mutually exclusive, and exactly one should be
143 # set for every version of python.
144 ASYNC_IDENTIFIERS = 6
146 ASSIGNMENT_EXPRESSIONS = 8
147 POS_ONLY_ARGUMENTS = 9
150 VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = {
151 TargetVersion.PY27: {Feature.ASYNC_IDENTIFIERS},
152 TargetVersion.PY33: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
153 TargetVersion.PY34: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
154 TargetVersion.PY35: {
155 Feature.UNICODE_LITERALS,
156 Feature.TRAILING_COMMA_IN_CALL,
157 Feature.ASYNC_IDENTIFIERS,
159 TargetVersion.PY36: {
160 Feature.UNICODE_LITERALS,
162 Feature.NUMERIC_UNDERSCORES,
163 Feature.TRAILING_COMMA_IN_CALL,
164 Feature.TRAILING_COMMA_IN_DEF,
165 Feature.ASYNC_IDENTIFIERS,
167 TargetVersion.PY37: {
168 Feature.UNICODE_LITERALS,
170 Feature.NUMERIC_UNDERSCORES,
171 Feature.TRAILING_COMMA_IN_CALL,
172 Feature.TRAILING_COMMA_IN_DEF,
173 Feature.ASYNC_KEYWORDS,
175 TargetVersion.PY38: {
176 Feature.UNICODE_LITERALS,
178 Feature.NUMERIC_UNDERSCORES,
179 Feature.TRAILING_COMMA_IN_CALL,
180 Feature.TRAILING_COMMA_IN_DEF,
181 Feature.ASYNC_KEYWORDS,
182 Feature.ASSIGNMENT_EXPRESSIONS,
183 Feature.POS_ONLY_ARGUMENTS,
190 target_versions: Set[TargetVersion] = field(default_factory=set)
191 line_length: int = DEFAULT_LINE_LENGTH
192 string_normalization: bool = True
195 def get_cache_key(self) -> str:
196 if self.target_versions:
197 version_str = ",".join(
199 for version in sorted(self.target_versions, key=lambda v: v.value)
205 str(self.line_length),
206 str(int(self.string_normalization)),
207 str(int(self.is_pyi)),
209 return ".".join(parts)
212 # Legacy name, left for integrations.
216 def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool:
217 return all(feature in VERSION_TO_FEATURES[version] for version in target_versions)
220 def find_pyproject_toml(path_search_start: str) -> Optional[str]:
221 """Find the absolute filepath to a pyproject.toml if it exists"""
222 path_project_root = find_project_root(path_search_start)
223 path_pyproject_toml = path_project_root / "pyproject.toml"
224 return str(path_pyproject_toml) if path_pyproject_toml.is_file() else None
227 def parse_pyproject_toml(path_config: str) -> Dict[str, Any]:
228 """Parse a pyproject toml file, pulling out relevant parts for Black
230 If parsing fails, will raise a toml.TomlDecodeError
232 pyproject_toml = toml.load(path_config)
233 config = pyproject_toml.get("tool", {}).get("black", {})
234 return {k.replace("--", "").replace("-", "_"): v for k, v in config.items()}
237 def read_pyproject_toml(
238 ctx: click.Context, param: click.Parameter, value: Optional[str]
240 """Inject Black configuration from "pyproject.toml" into defaults in `ctx`.
242 Returns the path to a successfully found and read configuration file, None
246 value = find_pyproject_toml(ctx.params.get("src", ()))
251 config = parse_pyproject_toml(value)
252 except (toml.TomlDecodeError, OSError) as e:
253 raise click.FileError(
254 filename=value, hint=f"Error reading configuration file: {e}"
260 default_map: Dict[str, Any] = {}
262 default_map.update(ctx.default_map)
263 default_map.update(config)
265 ctx.default_map = default_map
269 def target_version_option_callback(
270 c: click.Context, p: Union[click.Option, click.Parameter], v: Tuple[str, ...]
271 ) -> List[TargetVersion]:
272 """Compute the target versions from a --target-version flag.
274 This is its own function because mypy couldn't infer the type correctly
275 when it was a lambda, causing mypyc trouble.
277 return [TargetVersion[val.upper()] for val in v]
280 @click.command(context_settings=dict(help_option_names=["-h", "--help"]))
281 @click.option("-c", "--code", type=str, help="Format the code passed in as a string.")
286 default=DEFAULT_LINE_LENGTH,
287 help="How many characters per line to allow.",
293 type=click.Choice([v.name.lower() for v in TargetVersion]),
294 callback=target_version_option_callback,
297 "Python versions that should be supported by Black's output. [default: "
298 "per-file auto-detection]"
305 "Allow using Python 3.6-only syntax on all input files. This will put "
306 "trailing commas in function signatures and calls also after *args and "
307 "**kwargs. Deprecated; use --target-version instead. "
308 "[default: per-file auto-detection]"
315 "Format all input files like typing stubs regardless of file extension "
316 "(useful when piping source on standard input)."
321 "--skip-string-normalization",
323 help="Don't normalize string quotes or prefixes.",
329 "Don't write the files back, just return the status. Return code 0 "
330 "means nothing would change. Return code 1 means some files would be "
331 "reformatted. Return code 123 means there was an internal error."
337 help="Don't write the files back, just output a diff for each file on stdout.",
342 help="If --fast given, skip temporary sanity checks. [default: --safe]",
347 default=DEFAULT_INCLUDES,
349 "A regular expression that matches files and directories that should be "
350 "included on recursive searches. An empty value means all files are "
351 "included regardless of the name. Use forward slashes for directories on "
352 "all platforms (Windows, too). Exclusions are calculated first, inclusions "
360 default=DEFAULT_EXCLUDES,
362 "A regular expression that matches files and directories that should be "
363 "excluded on recursive searches. An empty value means no paths are excluded. "
364 "Use forward slashes for directories on all platforms (Windows, too). "
365 "Exclusions are calculated first, inclusions later."
374 "Don't emit non-error messages to stderr. Errors are still emitted; "
375 "silence those with 2>/dev/null."
383 "Also emit messages to stderr about files that were not changed or were "
384 "ignored due to --exclude=."
387 @click.version_option(version=__version__)
392 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
407 callback=read_pyproject_toml,
408 help="Read configuration from PATH.",
415 target_version: List[TargetVersion],
421 skip_string_normalization: bool,
426 src: Tuple[str, ...],
427 config: Optional[str],
429 """The uncompromising code formatter."""
430 write_back = WriteBack.from_configuration(check=check, diff=diff)
433 err("Cannot use both --target-version and --py36")
436 versions = set(target_version)
439 "--py36 is deprecated and will be removed in a future version. "
440 "Use --target-version py36 instead."
442 versions = PY36_VERSIONS
444 # We'll autodetect later.
447 target_versions=versions,
448 line_length=line_length,
450 string_normalization=not skip_string_normalization,
452 if config and verbose:
453 out(f"Using configuration from {config}.", bold=False, fg="blue")
455 print(format_str(code, mode=mode))
458 include_regex = re_compile_maybe_verbose(include)
460 err(f"Invalid regular expression for include given: {include!r}")
463 exclude_regex = re_compile_maybe_verbose(exclude)
465 err(f"Invalid regular expression for exclude given: {exclude!r}")
467 report = Report(check=check, diff=diff, quiet=quiet, verbose=verbose)
468 root = find_project_root(src)
469 sources: Set[Path] = set()
470 path_empty(src, quiet, verbose, ctx)
475 gen_python_files_in_dir(
476 p, root, include_regex, exclude_regex, report, get_gitignore(root)
479 elif p.is_file() or s == "-":
480 # if a file was explicitly given, we don't care about its extension
483 err(f"invalid path: {s}")
484 if len(sources) == 0:
485 if verbose or not quiet:
486 out("No Python files are present to be formatted. Nothing to do 😴")
489 if len(sources) == 1:
493 write_back=write_back,
499 sources=sources, fast=fast, write_back=write_back, mode=mode, report=report
502 if verbose or not quiet:
503 out("Oh no! 💥 💔 💥" if report.return_code else "All done! ✨ 🍰 ✨")
504 click.secho(str(report), err=True)
505 ctx.exit(report.return_code)
509 src: Tuple[str, ...], quiet: bool, verbose: bool, ctx: click.Context
512 Exit if there is no `src` provided for formatting
515 if verbose or not quiet:
516 out("No Path provided. Nothing to do 😴")
521 src: Path, fast: bool, write_back: WriteBack, mode: Mode, report: "Report"
523 """Reformat a single file under `src` without spawning child processes.
525 `fast`, `write_back`, and `mode` options are passed to
526 :func:`format_file_in_place` or :func:`format_stdin_to_stdout`.
530 if not src.is_file() and str(src) == "-":
531 if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode):
532 changed = Changed.YES
535 if write_back != WriteBack.DIFF:
536 cache = read_cache(mode)
537 res_src = src.resolve()
538 if res_src in cache and cache[res_src] == get_cache_info(res_src):
539 changed = Changed.CACHED
540 if changed is not Changed.CACHED and format_file_in_place(
541 src, fast=fast, write_back=write_back, mode=mode
543 changed = Changed.YES
544 if (write_back is WriteBack.YES and changed is not Changed.CACHED) or (
545 write_back is WriteBack.CHECK and changed is Changed.NO
547 write_cache(cache, [src], mode)
548 report.done(src, changed)
549 except Exception as exc:
550 report.failed(src, str(exc))
554 sources: Set[Path], fast: bool, write_back: WriteBack, mode: Mode, report: "Report"
556 """Reformat multiple files using a ProcessPoolExecutor."""
557 loop = asyncio.get_event_loop()
558 worker_count = os.cpu_count()
559 if sys.platform == "win32":
560 # Work around https://bugs.python.org/issue26903
561 worker_count = min(worker_count, 61)
562 executor = ProcessPoolExecutor(max_workers=worker_count)
564 loop.run_until_complete(
568 write_back=write_back,
580 async def schedule_formatting(
583 write_back: WriteBack,
586 loop: asyncio.AbstractEventLoop,
589 """Run formatting of `sources` in parallel using the provided `executor`.
591 (Use ProcessPoolExecutors for actual parallelism.)
593 `write_back`, `fast`, and `mode` options are passed to
594 :func:`format_file_in_place`.
597 if write_back != WriteBack.DIFF:
598 cache = read_cache(mode)
599 sources, cached = filter_cached(cache, sources)
600 for src in sorted(cached):
601 report.done(src, Changed.CACHED)
606 sources_to_cache = []
608 if write_back == WriteBack.DIFF:
609 # For diff output, we need locks to ensure we don't interleave output
610 # from different processes.
612 lock = manager.Lock()
614 asyncio.ensure_future(
615 loop.run_in_executor(
616 executor, format_file_in_place, src, fast, mode, write_back, lock
619 for src in sorted(sources)
621 pending: Iterable["asyncio.Future[bool]"] = tasks.keys()
623 loop.add_signal_handler(signal.SIGINT, cancel, pending)
624 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
625 except NotImplementedError:
626 # There are no good alternatives for these on Windows.
629 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
631 src = tasks.pop(task)
633 cancelled.append(task)
634 elif task.exception():
635 report.failed(src, str(task.exception()))
637 changed = Changed.YES if task.result() else Changed.NO
638 # If the file was written back or was successfully checked as
639 # well-formatted, store this information in the cache.
640 if write_back is WriteBack.YES or (
641 write_back is WriteBack.CHECK and changed is Changed.NO
643 sources_to_cache.append(src)
644 report.done(src, changed)
646 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
648 write_cache(cache, sources_to_cache, mode)
651 def format_file_in_place(
655 write_back: WriteBack = WriteBack.NO,
656 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
658 """Format file under `src` path. Return True if changed.
660 If `write_back` is DIFF, write a diff to stdout. If it is YES, write reformatted
662 `mode` and `fast` options are passed to :func:`format_file_contents`.
664 if src.suffix == ".pyi":
665 mode = replace(mode, is_pyi=True)
667 then = datetime.utcfromtimestamp(src.stat().st_mtime)
668 with open(src, "rb") as buf:
669 src_contents, encoding, newline = decode_bytes(buf.read())
671 dst_contents = format_file_contents(src_contents, fast=fast, mode=mode)
672 except NothingChanged:
675 if write_back == WriteBack.YES:
676 with open(src, "w", encoding=encoding, newline=newline) as f:
677 f.write(dst_contents)
678 elif write_back == WriteBack.DIFF:
679 now = datetime.utcnow()
680 src_name = f"{src}\t{then} +0000"
681 dst_name = f"{src}\t{now} +0000"
682 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
684 with lock or nullcontext():
685 f = io.TextIOWrapper(
691 f.write(diff_contents)
697 def format_stdin_to_stdout(
698 fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: Mode
700 """Format file on stdin. Return True if changed.
702 If `write_back` is YES, write reformatted code back to stdout. If it is DIFF,
703 write a diff to stdout. The `mode` argument is passed to
704 :func:`format_file_contents`.
706 then = datetime.utcnow()
707 src, encoding, newline = decode_bytes(sys.stdin.buffer.read())
710 dst = format_file_contents(src, fast=fast, mode=mode)
713 except NothingChanged:
717 f = io.TextIOWrapper(
718 sys.stdout.buffer, encoding=encoding, newline=newline, write_through=True
720 if write_back == WriteBack.YES:
722 elif write_back == WriteBack.DIFF:
723 now = datetime.utcnow()
724 src_name = f"STDIN\t{then} +0000"
725 dst_name = f"STDOUT\t{now} +0000"
726 f.write(diff(src, dst, src_name, dst_name))
730 def format_file_contents(src_contents: str, *, fast: bool, mode: Mode) -> FileContent:
731 """Reformat contents a file and return new contents.
733 If `fast` is False, additionally confirm that the reformatted code is
734 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
735 `mode` is passed to :func:`format_str`.
737 if src_contents.strip() == "":
740 dst_contents = format_str(src_contents, mode=mode)
741 if src_contents == dst_contents:
745 assert_equivalent(src_contents, dst_contents)
746 assert_stable(src_contents, dst_contents, mode=mode)
750 def format_str(src_contents: str, *, mode: Mode) -> FileContent:
751 """Reformat a string and return new contents.
753 `mode` determines formatting options, such as how many characters per line are
757 >>> print(black.format_str("def f(arg:str='')->None:...", mode=Mode()))
758 def f(arg: str = "") -> None:
761 A more complex example:
763 ... black.format_str(
764 ... "def f(arg:str='')->None: hey",
766 ... target_versions={black.TargetVersion.PY36},
768 ... string_normalization=False,
779 src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
781 future_imports = get_future_imports(src_node)
782 if mode.target_versions:
783 versions = mode.target_versions
785 versions = detect_target_versions(src_node)
786 normalize_fmt_off(src_node)
787 lines = LineGenerator(
788 remove_u_prefix="unicode_literals" in future_imports
789 or supports_feature(versions, Feature.UNICODE_LITERALS),
791 normalize_strings=mode.string_normalization,
793 elt = EmptyLineTracker(is_pyi=mode.is_pyi)
796 split_line_features = {
798 for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF}
799 if supports_feature(versions, feature)
801 for current_line in lines.visit(src_node):
802 dst_contents.append(str(empty_line) * after)
803 before, after = elt.maybe_empty_lines(current_line)
804 dst_contents.append(str(empty_line) * before)
805 for line in split_line(
806 current_line, line_length=mode.line_length, features=split_line_features
808 dst_contents.append(str(line))
809 return "".join(dst_contents)
812 def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]:
813 """Return a tuple of (decoded_contents, encoding, newline).
815 `newline` is either CRLF or LF but `decoded_contents` is decoded with
816 universal newlines (i.e. only contains LF).
818 srcbuf = io.BytesIO(src)
819 encoding, lines = tokenize.detect_encoding(srcbuf.readline)
821 return "", encoding, "\n"
823 newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n"
825 with io.TextIOWrapper(srcbuf, encoding) as tiow:
826 return tiow.read(), encoding, newline
829 def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
830 if not target_versions:
831 # No target_version specified, so try all grammars.
834 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords,
836 pygram.python_grammar_no_print_statement_no_exec_statement,
837 # Python 2.7 with future print_function import
838 pygram.python_grammar_no_print_statement,
840 pygram.python_grammar,
843 if all(version.is_python2() for version in target_versions):
844 # Python 2-only code, so try Python 2 grammars.
846 # Python 2.7 with future print_function import
847 pygram.python_grammar_no_print_statement,
849 pygram.python_grammar,
852 # Python 3-compatible code, so only try Python 3 grammar.
854 # If we have to parse both, try to parse async as a keyword first
855 if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
858 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords
860 if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
862 grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
863 # At least one of the above branches must have been taken, because every Python
864 # version has exactly one of the two 'ASYNC_*' flags
868 def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
869 """Given a string with source, return the lib2to3 Node."""
870 if src_txt[-1:] != "\n":
873 for grammar in get_grammars(set(target_versions)):
874 drv = driver.Driver(grammar, pytree.convert)
876 result = drv.parse_string(src_txt, True)
879 except ParseError as pe:
880 lineno, column = pe.context[1]
881 lines = src_txt.splitlines()
883 faulty_line = lines[lineno - 1]
885 faulty_line = "<line number missing in source>"
886 exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
890 if isinstance(result, Leaf):
891 result = Node(syms.file_input, [result])
895 def lib2to3_unparse(node: Node) -> str:
896 """Given a lib2to3 node, return its string representation."""
904 class Visitor(Generic[T]):
905 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
907 def visit(self, node: LN) -> Iterator[T]:
908 """Main method to visit `node` and its children.
910 It tries to find a `visit_*()` method for the given `node.type`, like
911 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
912 If no dedicated `visit_*()` method is found, chooses `visit_default()`
915 Then yields objects of type `T` from the selected visitor.
918 name = token.tok_name[node.type]
920 name = str(type_repr(node.type))
921 # We explicitly branch on whether a visitor exists (instead of
922 # using self.visit_default as the default arg to getattr) in order
923 # to save needing to create a bound method object and so mypyc can
924 # generate a native call to visit_default.
925 visitf = getattr(self, f"visit_{name}", None)
927 yield from visitf(node)
929 yield from self.visit_default(node)
931 def visit_default(self, node: LN) -> Iterator[T]:
932 """Default `visit_*()` implementation. Recurses to children of `node`."""
933 if isinstance(node, Node):
934 for child in node.children:
935 yield from self.visit(child)
939 class DebugVisitor(Visitor[T]):
942 def visit_default(self, node: LN) -> Iterator[T]:
943 indent = " " * (2 * self.tree_depth)
944 if isinstance(node, Node):
945 _type = type_repr(node.type)
946 out(f"{indent}{_type}", fg="yellow")
948 for child in node.children:
949 yield from self.visit(child)
952 out(f"{indent}/{_type}", fg="yellow", bold=False)
954 _type = token.tok_name.get(node.type, str(node.type))
955 out(f"{indent}{_type}", fg="blue", nl=False)
957 # We don't have to handle prefixes for `Node` objects since
958 # that delegates to the first child anyway.
959 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
960 out(f" {node.value!r}", fg="blue", bold=False)
963 def show(cls, code: Union[str, Leaf, Node]) -> None:
964 """Pretty-print the lib2to3 AST of a given string of `code`.
966 Convenience method for debugging.
968 v: DebugVisitor[None] = DebugVisitor()
969 if isinstance(code, str):
970 code = lib2to3_parse(code)
974 WHITESPACE: Final = {token.DEDENT, token.INDENT, token.NEWLINE}
985 STANDALONE_COMMENT: Final = 153
986 token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT"
987 LOGIC_OPERATORS: Final = {"and", "or"}
988 COMPARATORS: Final = {
996 MATH_OPERATORS: Final = {
1012 STARS: Final = {token.STAR, token.DOUBLESTAR}
1013 VARARGS_SPECIALS: Final = STARS | {token.SLASH}
1014 VARARGS_PARENTS: Final = {
1016 syms.argument, # double star in arglist
1017 syms.trailer, # single argument to call
1019 syms.varargslist, # lambdas
1021 UNPACKING_PARENTS: Final = {
1022 syms.atom, # single element of a list or set literal
1026 syms.testlist_star_expr,
1028 TEST_DESCENDANTS: Final = {
1045 ASSIGNMENTS: Final = {
1061 COMPREHENSION_PRIORITY: Final = 20
1062 COMMA_PRIORITY: Final = 18
1063 TERNARY_PRIORITY: Final = 16
1064 LOGIC_PRIORITY: Final = 14
1065 STRING_PRIORITY: Final = 12
1066 COMPARATOR_PRIORITY: Final = 10
1067 MATH_PRIORITIES: Final = {
1069 token.CIRCUMFLEX: 8,
1072 token.RIGHTSHIFT: 6,
1077 token.DOUBLESLASH: 4,
1081 token.DOUBLESTAR: 2,
1083 DOT_PRIORITY: Final = 1
1087 class BracketTracker:
1088 """Keeps track of brackets on a line."""
1091 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = field(default_factory=dict)
1092 delimiters: Dict[LeafID, Priority] = field(default_factory=dict)
1093 previous: Optional[Leaf] = None
1094 _for_loop_depths: List[int] = field(default_factory=list)
1095 _lambda_argument_depths: List[int] = field(default_factory=list)
1097 def mark(self, leaf: Leaf) -> None:
1098 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
1100 All leaves receive an int `bracket_depth` field that stores how deep
1101 within brackets a given leaf is. 0 means there are no enclosing brackets
1102 that started on this line.
1104 If a leaf is itself a closing bracket, it receives an `opening_bracket`
1105 field that it forms a pair with. This is a one-directional link to
1106 avoid reference cycles.
1108 If a leaf is a delimiter (a token on which Black can split the line if
1109 needed) and it's on depth 0, its `id()` is stored in the tracker's
1112 if leaf.type == token.COMMENT:
1115 self.maybe_decrement_after_for_loop_variable(leaf)
1116 self.maybe_decrement_after_lambda_arguments(leaf)
1117 if leaf.type in CLOSING_BRACKETS:
1119 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
1120 leaf.opening_bracket = opening_bracket
1121 leaf.bracket_depth = self.depth
1123 delim = is_split_before_delimiter(leaf, self.previous)
1124 if delim and self.previous is not None:
1125 self.delimiters[id(self.previous)] = delim
1127 delim = is_split_after_delimiter(leaf, self.previous)
1129 self.delimiters[id(leaf)] = delim
1130 if leaf.type in OPENING_BRACKETS:
1131 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
1133 self.previous = leaf
1134 self.maybe_increment_lambda_arguments(leaf)
1135 self.maybe_increment_for_loop_variable(leaf)
1137 def any_open_brackets(self) -> bool:
1138 """Return True if there is an yet unmatched open bracket on the line."""
1139 return bool(self.bracket_match)
1141 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> Priority:
1142 """Return the highest priority of a delimiter found on the line.
1144 Values are consistent with what `is_split_*_delimiter()` return.
1145 Raises ValueError on no delimiters.
1147 return max(v for k, v in self.delimiters.items() if k not in exclude)
1149 def delimiter_count_with_priority(self, priority: Priority = 0) -> int:
1150 """Return the number of delimiters with the given `priority`.
1152 If no `priority` is passed, defaults to max priority on the line.
1154 if not self.delimiters:
1157 priority = priority or self.max_delimiter_priority()
1158 return sum(1 for p in self.delimiters.values() if p == priority)
1160 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1161 """In a for loop, or comprehension, the variables are often unpacks.
1163 To avoid splitting on the comma in this situation, increase the depth of
1164 tokens between `for` and `in`.
1166 if leaf.type == token.NAME and leaf.value == "for":
1168 self._for_loop_depths.append(self.depth)
1173 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
1174 """See `maybe_increment_for_loop_variable` above for explanation."""
1176 self._for_loop_depths
1177 and self._for_loop_depths[-1] == self.depth
1178 and leaf.type == token.NAME
1179 and leaf.value == "in"
1182 self._for_loop_depths.pop()
1187 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
1188 """In a lambda expression, there might be more than one argument.
1190 To avoid splitting on the comma in this situation, increase the depth of
1191 tokens between `lambda` and `:`.
1193 if leaf.type == token.NAME and leaf.value == "lambda":
1195 self._lambda_argument_depths.append(self.depth)
1200 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
1201 """See `maybe_increment_lambda_arguments` above for explanation."""
1203 self._lambda_argument_depths
1204 and self._lambda_argument_depths[-1] == self.depth
1205 and leaf.type == token.COLON
1208 self._lambda_argument_depths.pop()
1213 def get_open_lsqb(self) -> Optional[Leaf]:
1214 """Return the most recent opening square bracket (if any)."""
1215 return self.bracket_match.get((self.depth - 1, token.RSQB))
1220 """Holds leaves and comments. Can be printed with `str(line)`."""
1223 leaves: List[Leaf] = field(default_factory=list)
1224 # keys ordered like `leaves`
1225 comments: Dict[LeafID, List[Leaf]] = field(default_factory=dict)
1226 bracket_tracker: BracketTracker = field(default_factory=BracketTracker)
1227 inside_brackets: bool = False
1228 should_explode: bool = False
1230 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
1231 """Add a new `leaf` to the end of the line.
1233 Unless `preformatted` is True, the `leaf` will receive a new consistent
1234 whitespace prefix and metadata applied by :class:`BracketTracker`.
1235 Trailing commas are maybe removed, unpacked for loop variables are
1236 demoted from being delimiters.
1238 Inline comments are put aside.
1240 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
1244 if token.COLON == leaf.type and self.is_class_paren_empty:
1245 del self.leaves[-2:]
1246 if self.leaves and not preformatted:
1247 # Note: at this point leaf.prefix should be empty except for
1248 # imports, for which we only preserve newlines.
1249 leaf.prefix += whitespace(
1250 leaf, complex_subscript=self.is_complex_subscript(leaf)
1252 if self.inside_brackets or not preformatted:
1253 self.bracket_tracker.mark(leaf)
1254 self.maybe_remove_trailing_comma(leaf)
1255 if not self.append_comment(leaf):
1256 self.leaves.append(leaf)
1258 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
1259 """Like :func:`append()` but disallow invalid standalone comment structure.
1261 Raises ValueError when any `leaf` is appended after a standalone comment
1262 or when a standalone comment is not the first leaf on the line.
1264 if self.bracket_tracker.depth == 0:
1266 raise ValueError("cannot append to standalone comments")
1268 if self.leaves and leaf.type == STANDALONE_COMMENT:
1270 "cannot append standalone comments to a populated line"
1273 self.append(leaf, preformatted=preformatted)
1276 def is_comment(self) -> bool:
1277 """Is this line a standalone comment?"""
1278 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
1281 def is_decorator(self) -> bool:
1282 """Is this line a decorator?"""
1283 return bool(self) and self.leaves[0].type == token.AT
1286 def is_import(self) -> bool:
1287 """Is this an import line?"""
1288 return bool(self) and is_import(self.leaves[0])
1291 def is_class(self) -> bool:
1292 """Is this line a class definition?"""
1295 and self.leaves[0].type == token.NAME
1296 and self.leaves[0].value == "class"
1300 def is_stub_class(self) -> bool:
1301 """Is this line a class definition with a body consisting only of "..."?"""
1302 return self.is_class and self.leaves[-3:] == [
1303 Leaf(token.DOT, ".") for _ in range(3)
1307 def is_collection_with_optional_trailing_comma(self) -> bool:
1308 """Is this line a collection literal with a trailing comma that's optional?
1310 Note that the trailing comma in a 1-tuple is not optional.
1312 if not self.leaves or len(self.leaves) < 4:
1315 # Look for and address a trailing colon.
1316 if self.leaves[-1].type == token.COLON:
1317 closer = self.leaves[-2]
1320 closer = self.leaves[-1]
1322 if closer.type not in CLOSING_BRACKETS or self.inside_brackets:
1325 if closer.type == token.RPAR:
1326 # Tuples require an extra check, because if there's only
1327 # one element in the tuple removing the comma unmakes the
1330 # We also check for parens before looking for the trailing
1331 # comma because in some cases (eg assigning a dict
1332 # literal) the literal gets wrapped in temporary parens
1333 # during parsing. This case is covered by the
1334 # collections.py test data.
1335 opener = closer.opening_bracket
1336 for _open_index, leaf in enumerate(self.leaves):
1341 # Couldn't find the matching opening paren, play it safe.
1345 comma_depth = self.leaves[close_index - 1].bracket_depth
1346 for leaf in self.leaves[_open_index + 1 : close_index]:
1347 if leaf.bracket_depth == comma_depth and leaf.type == token.COMMA:
1350 # We haven't looked yet for the trailing comma because
1351 # we might also have caught noop parens.
1352 return self.leaves[close_index - 1].type == token.COMMA
1355 return False # it's either a one-tuple or didn't have a trailing comma
1357 if self.leaves[close_index - 1].type in CLOSING_BRACKETS:
1359 closer = self.leaves[close_index]
1360 if closer.type == token.RPAR:
1361 # TODO: this is a gut feeling. Will we ever see this?
1364 if self.leaves[close_index - 1].type != token.COMMA:
1370 def is_def(self) -> bool:
1371 """Is this a function definition? (Also returns True for async defs.)"""
1373 first_leaf = self.leaves[0]
1378 second_leaf: Optional[Leaf] = self.leaves[1]
1381 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
1382 first_leaf.type == token.ASYNC
1383 and second_leaf is not None
1384 and second_leaf.type == token.NAME
1385 and second_leaf.value == "def"
1389 def is_class_paren_empty(self) -> bool:
1390 """Is this a class with no base classes but using parentheses?
1392 Those are unnecessary and should be removed.
1396 and len(self.leaves) == 4
1398 and self.leaves[2].type == token.LPAR
1399 and self.leaves[2].value == "("
1400 and self.leaves[3].type == token.RPAR
1401 and self.leaves[3].value == ")"
1405 def is_triple_quoted_string(self) -> bool:
1406 """Is the line a triple quoted string?"""
1409 and self.leaves[0].type == token.STRING
1410 and self.leaves[0].value.startswith(('"""', "'''"))
1413 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
1414 """If so, needs to be split before emitting."""
1415 for leaf in self.leaves:
1416 if leaf.type == STANDALONE_COMMENT and leaf.bracket_depth <= depth_limit:
1421 def contains_uncollapsable_type_comments(self) -> bool:
1424 last_leaf = self.leaves[-1]
1425 ignored_ids.add(id(last_leaf))
1426 if last_leaf.type == token.COMMA or (
1427 last_leaf.type == token.RPAR and not last_leaf.value
1429 # When trailing commas or optional parens are inserted by Black for
1430 # consistency, comments after the previous last element are not moved
1431 # (they don't have to, rendering will still be correct). So we ignore
1432 # trailing commas and invisible.
1433 last_leaf = self.leaves[-2]
1434 ignored_ids.add(id(last_leaf))
1438 # A type comment is uncollapsable if it is attached to a leaf
1439 # that isn't at the end of the line (since that could cause it
1440 # to get associated to a different argument) or if there are
1441 # comments before it (since that could cause it to get hidden
1443 comment_seen = False
1444 for leaf_id, comments in self.comments.items():
1445 for comment in comments:
1446 if is_type_comment(comment):
1447 if comment_seen or (
1448 not is_type_comment(comment, " ignore")
1449 and leaf_id not in ignored_ids
1457 def contains_unsplittable_type_ignore(self) -> bool:
1461 # If a 'type: ignore' is attached to the end of a line, we
1462 # can't split the line, because we can't know which of the
1463 # subexpressions the ignore was meant to apply to.
1465 # We only want this to apply to actual physical lines from the
1466 # original source, though: we don't want the presence of a
1467 # 'type: ignore' at the end of a multiline expression to
1468 # justify pushing it all onto one line. Thus we
1469 # (unfortunately) need to check the actual source lines and
1470 # only report an unsplittable 'type: ignore' if this line was
1471 # one line in the original code.
1473 # Grab the first and last line numbers, skipping generated leaves
1474 first_line = next((l.lineno for l in self.leaves if l.lineno != 0), 0)
1475 last_line = next((l.lineno for l in reversed(self.leaves) if l.lineno != 0), 0)
1477 if first_line == last_line:
1478 # We look at the last two leaves since a comma or an
1479 # invisible paren could have been added at the end of the
1481 for node in self.leaves[-2:]:
1482 for comment in self.comments.get(id(node), []):
1483 if is_type_comment(comment, " ignore"):
1488 def contains_multiline_strings(self) -> bool:
1489 return any(is_multiline_string(leaf) for leaf in self.leaves)
1491 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1492 """Remove trailing comma if there is one and it's safe."""
1493 if not (self.leaves and self.leaves[-1].type == token.COMMA):
1496 # We remove trailing commas only in the case of importing a
1497 # single name from a module.
1501 and len(self.leaves) > 4
1502 and self.leaves[-1].type == token.COMMA
1503 and closing.type in CLOSING_BRACKETS
1504 and self.leaves[-4].type == token.NAME
1506 # regular `from foo import bar,`
1507 self.leaves[-4].value == "import"
1508 # `from foo import (bar as baz,)
1510 len(self.leaves) > 6
1511 and self.leaves[-6].value == "import"
1512 and self.leaves[-3].value == "as"
1514 # `from foo import bar as baz,`
1516 len(self.leaves) > 5
1517 and self.leaves[-5].value == "import"
1518 and self.leaves[-3].value == "as"
1521 and closing.type == token.RPAR
1525 self.remove_trailing_comma()
1528 def append_comment(self, comment: Leaf) -> bool:
1529 """Add an inline or standalone comment to the line."""
1531 comment.type == STANDALONE_COMMENT
1532 and self.bracket_tracker.any_open_brackets()
1537 if comment.type != token.COMMENT:
1541 comment.type = STANDALONE_COMMENT
1545 last_leaf = self.leaves[-1]
1547 last_leaf.type == token.RPAR
1548 and not last_leaf.value
1549 and last_leaf.parent
1550 and len(list(last_leaf.parent.leaves())) <= 3
1551 and not is_type_comment(comment)
1553 # Comments on an optional parens wrapping a single leaf should belong to
1554 # the wrapped node except if it's a type comment. Pinning the comment like
1555 # this avoids unstable formatting caused by comment migration.
1556 if len(self.leaves) < 2:
1557 comment.type = STANDALONE_COMMENT
1561 last_leaf = self.leaves[-2]
1562 self.comments.setdefault(id(last_leaf), []).append(comment)
1565 def comments_after(self, leaf: Leaf) -> List[Leaf]:
1566 """Generate comments that should appear directly after `leaf`."""
1567 return self.comments.get(id(leaf), [])
1569 def remove_trailing_comma(self) -> None:
1570 """Remove the trailing comma and moves the comments attached to it."""
1571 trailing_comma = self.leaves.pop()
1572 trailing_comma_comments = self.comments.pop(id(trailing_comma), [])
1573 self.comments.setdefault(id(self.leaves[-1]), []).extend(
1574 trailing_comma_comments
1577 def is_complex_subscript(self, leaf: Leaf) -> bool:
1578 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1579 open_lsqb = self.bracket_tracker.get_open_lsqb()
1580 if open_lsqb is None:
1583 subscript_start = open_lsqb.next_sibling
1585 if isinstance(subscript_start, Node):
1586 if subscript_start.type == syms.listmaker:
1589 if subscript_start.type == syms.subscriptlist:
1590 subscript_start = child_towards(subscript_start, leaf)
1591 return subscript_start is not None and any(
1592 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1595 def __str__(self) -> str:
1596 """Render the line."""
1600 indent = " " * self.depth
1601 leaves = iter(self.leaves)
1602 first = next(leaves)
1603 res = f"{first.prefix}{indent}{first.value}"
1606 for comment in itertools.chain.from_iterable(self.comments.values()):
1610 def __bool__(self) -> bool:
1611 """Return True if the line has leaves or comments."""
1612 return bool(self.leaves or self.comments)
1616 class EmptyLineTracker:
1617 """Provides a stateful method that returns the number of potential extra
1618 empty lines needed before and after the currently processed line.
1620 Note: this tracker works on lines that haven't been split yet. It assumes
1621 the prefix of the first leaf consists of optional newlines. Those newlines
1622 are consumed by `maybe_empty_lines()` and included in the computation.
1625 is_pyi: bool = False
1626 previous_line: Optional[Line] = None
1627 previous_after: int = 0
1628 previous_defs: List[int] = field(default_factory=list)
1630 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1631 """Return the number of extra empty lines before and after the `current_line`.
1633 This is for separating `def`, `async def` and `class` with extra empty
1634 lines (two on module-level).
1636 before, after = self._maybe_empty_lines(current_line)
1638 # Black should not insert empty lines at the beginning
1641 if self.previous_line is None
1642 else before - self.previous_after
1644 self.previous_after = after
1645 self.previous_line = current_line
1646 return before, after
1648 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1650 if current_line.depth == 0:
1651 max_allowed = 1 if self.is_pyi else 2
1652 if current_line.leaves:
1653 # Consume the first leaf's extra newlines.
1654 first_leaf = current_line.leaves[0]
1655 before = first_leaf.prefix.count("\n")
1656 before = min(before, max_allowed)
1657 first_leaf.prefix = ""
1660 depth = current_line.depth
1661 while self.previous_defs and self.previous_defs[-1] >= depth:
1662 self.previous_defs.pop()
1664 before = 0 if depth else 1
1666 before = 1 if depth else 2
1667 if current_line.is_decorator or current_line.is_def or current_line.is_class:
1668 return self._maybe_empty_lines_for_class_or_def(current_line, before)
1672 and self.previous_line.is_import
1673 and not current_line.is_import
1674 and depth == self.previous_line.depth
1676 return (before or 1), 0
1680 and self.previous_line.is_class
1681 and current_line.is_triple_quoted_string
1687 def _maybe_empty_lines_for_class_or_def(
1688 self, current_line: Line, before: int
1689 ) -> Tuple[int, int]:
1690 if not current_line.is_decorator:
1691 self.previous_defs.append(current_line.depth)
1692 if self.previous_line is None:
1693 # Don't insert empty lines before the first line in the file.
1696 if self.previous_line.is_decorator:
1699 if self.previous_line.depth < current_line.depth and (
1700 self.previous_line.is_class or self.previous_line.is_def
1705 self.previous_line.is_comment
1706 and self.previous_line.depth == current_line.depth
1712 if self.previous_line.depth > current_line.depth:
1714 elif current_line.is_class or self.previous_line.is_class:
1715 if current_line.is_stub_class and self.previous_line.is_stub_class:
1716 # No blank line between classes with an empty body
1720 elif current_line.is_def and not self.previous_line.is_def:
1721 # Blank line between a block of functions and a block of non-functions
1727 if current_line.depth and newlines:
1733 class LineGenerator(Visitor[Line]):
1734 """Generates reformatted Line objects. Empty lines are not emitted.
1736 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1737 in ways that will no longer stringify to valid Python code on the tree.
1740 is_pyi: bool = False
1741 normalize_strings: bool = True
1742 current_line: Line = field(default_factory=Line)
1743 remove_u_prefix: bool = False
1745 def line(self, indent: int = 0) -> Iterator[Line]:
1748 If the line is empty, only emit if it makes sense.
1749 If the line is too long, split it first and then generate.
1751 If any lines were generated, set up a new current_line.
1753 if not self.current_line:
1754 self.current_line.depth += indent
1755 return # Line is empty, don't emit. Creating a new one unnecessary.
1757 complete_line = self.current_line
1758 self.current_line = Line(depth=complete_line.depth + indent)
1761 def visit_default(self, node: LN) -> Iterator[Line]:
1762 """Default `visit_*()` implementation. Recurses to children of `node`."""
1763 if isinstance(node, Leaf):
1764 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1765 for comment in generate_comments(node):
1766 if any_open_brackets:
1767 # any comment within brackets is subject to splitting
1768 self.current_line.append(comment)
1769 elif comment.type == token.COMMENT:
1770 # regular trailing comment
1771 self.current_line.append(comment)
1772 yield from self.line()
1775 # regular standalone comment
1776 yield from self.line()
1778 self.current_line.append(comment)
1779 yield from self.line()
1781 normalize_prefix(node, inside_brackets=any_open_brackets)
1782 if self.normalize_strings and node.type == token.STRING:
1783 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1784 normalize_string_quotes(node)
1785 if node.type == token.NUMBER:
1786 normalize_numeric_literal(node)
1787 if node.type not in WHITESPACE:
1788 self.current_line.append(node)
1789 yield from super().visit_default(node)
1791 def visit_INDENT(self, node: Leaf) -> Iterator[Line]:
1792 """Increase indentation level, maybe yield a line."""
1793 # In blib2to3 INDENT never holds comments.
1794 yield from self.line(+1)
1795 yield from self.visit_default(node)
1797 def visit_DEDENT(self, node: Leaf) -> Iterator[Line]:
1798 """Decrease indentation level, maybe yield a line."""
1799 # The current line might still wait for trailing comments. At DEDENT time
1800 # there won't be any (they would be prefixes on the preceding NEWLINE).
1801 # Emit the line then.
1802 yield from self.line()
1804 # While DEDENT has no value, its prefix may contain standalone comments
1805 # that belong to the current indentation level. Get 'em.
1806 yield from self.visit_default(node)
1808 # Finally, emit the dedent.
1809 yield from self.line(-1)
1812 self, node: Node, keywords: Set[str], parens: Set[str]
1813 ) -> Iterator[Line]:
1814 """Visit a statement.
1816 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1817 `def`, `with`, `class`, `assert` and assignments.
1819 The relevant Python language `keywords` for a given statement will be
1820 NAME leaves within it. This methods puts those on a separate line.
1822 `parens` holds a set of string leaf values immediately after which
1823 invisible parens should be put.
1825 normalize_invisible_parens(node, parens_after=parens)
1826 for child in node.children:
1827 if child.type == token.NAME and child.value in keywords: # type: ignore
1828 yield from self.line()
1830 yield from self.visit(child)
1832 def visit_suite(self, node: Node) -> Iterator[Line]:
1833 """Visit a suite."""
1834 if self.is_pyi and is_stub_suite(node):
1835 yield from self.visit(node.children[2])
1837 yield from self.visit_default(node)
1839 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1840 """Visit a statement without nested statements."""
1841 is_suite_like = node.parent and node.parent.type in STATEMENT
1843 if self.is_pyi and is_stub_body(node):
1844 yield from self.visit_default(node)
1846 yield from self.line(+1)
1847 yield from self.visit_default(node)
1848 yield from self.line(-1)
1851 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1852 yield from self.line()
1853 yield from self.visit_default(node)
1855 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1856 """Visit `async def`, `async for`, `async with`."""
1857 yield from self.line()
1859 children = iter(node.children)
1860 for child in children:
1861 yield from self.visit(child)
1863 if child.type == token.ASYNC:
1866 internal_stmt = next(children)
1867 for child in internal_stmt.children:
1868 yield from self.visit(child)
1870 def visit_decorators(self, node: Node) -> Iterator[Line]:
1871 """Visit decorators."""
1872 for child in node.children:
1873 yield from self.line()
1874 yield from self.visit(child)
1876 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1877 """Remove a semicolon and put the other statement on a separate line."""
1878 yield from self.line()
1880 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1881 """End of file. Process outstanding comments and end with a newline."""
1882 yield from self.visit_default(leaf)
1883 yield from self.line()
1885 def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]:
1886 if not self.current_line.bracket_tracker.any_open_brackets():
1887 yield from self.line()
1888 yield from self.visit_default(leaf)
1890 def visit_factor(self, node: Node) -> Iterator[Line]:
1891 """Force parentheses between a unary op and a binary power:
1893 -2 ** 8 -> -(2 ** 8)
1895 _operator, operand = node.children
1897 operand.type == syms.power
1898 and len(operand.children) == 3
1899 and operand.children[1].type == token.DOUBLESTAR
1901 lpar = Leaf(token.LPAR, "(")
1902 rpar = Leaf(token.RPAR, ")")
1903 index = operand.remove() or 0
1904 node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
1905 yield from self.visit_default(node)
1907 def __post_init__(self) -> None:
1908 """You are in a twisty little maze of passages."""
1911 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1912 self.visit_if_stmt = partial(
1913 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1915 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1916 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1917 self.visit_try_stmt = partial(
1918 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1920 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1921 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1922 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1923 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1924 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1925 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1926 self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
1927 self.visit_del_stmt = partial(v, keywords=Ø, parens={"del"})
1928 self.visit_async_funcdef = self.visit_async_stmt
1929 self.visit_decorated = self.visit_decorators
1932 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1933 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1934 OPENING_BRACKETS = set(BRACKET.keys())
1935 CLOSING_BRACKETS = set(BRACKET.values())
1936 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1937 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1940 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901
1941 """Return whitespace prefix if needed for the given `leaf`.
1943 `complex_subscript` signals whether the given leaf is part of a subscription
1944 which has non-trivial arguments, like arithmetic expressions or function calls.
1952 if t in ALWAYS_NO_SPACE:
1955 if t == token.COMMENT:
1958 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1959 if t == token.COLON and p.type not in {
1966 prev = leaf.prev_sibling
1968 prevp = preceding_leaf(p)
1969 if not prevp or prevp.type in OPENING_BRACKETS:
1972 if t == token.COLON:
1973 if prevp.type == token.COLON:
1976 elif prevp.type != token.COMMA and not complex_subscript:
1981 if prevp.type == token.EQUAL:
1983 if prevp.parent.type in {
1991 elif prevp.parent.type == syms.typedargslist:
1992 # A bit hacky: if the equal sign has whitespace, it means we
1993 # previously found it's a typed argument. So, we're using
1997 elif prevp.type in VARARGS_SPECIALS:
1998 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
2001 elif prevp.type == token.COLON:
2002 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
2003 return SPACE if complex_subscript else NO
2007 and prevp.parent.type == syms.factor
2008 and prevp.type in MATH_OPERATORS
2013 prevp.type == token.RIGHTSHIFT
2015 and prevp.parent.type == syms.shift_expr
2016 and prevp.prev_sibling
2017 and prevp.prev_sibling.type == token.NAME
2018 and prevp.prev_sibling.value == "print" # type: ignore
2020 # Python 2 print chevron
2023 elif prev.type in OPENING_BRACKETS:
2026 if p.type in {syms.parameters, syms.arglist}:
2027 # untyped function signatures or calls
2028 if not prev or prev.type != token.COMMA:
2031 elif p.type == syms.varargslist:
2033 if prev and prev.type != token.COMMA:
2036 elif p.type == syms.typedargslist:
2037 # typed function signatures
2041 if t == token.EQUAL:
2042 if prev.type != syms.tname:
2045 elif prev.type == token.EQUAL:
2046 # A bit hacky: if the equal sign has whitespace, it means we
2047 # previously found it's a typed argument. So, we're using that, too.
2050 elif prev.type != token.COMMA:
2053 elif p.type == syms.tname:
2056 prevp = preceding_leaf(p)
2057 if not prevp or prevp.type != token.COMMA:
2060 elif p.type == syms.trailer:
2061 # attributes and calls
2062 if t == token.LPAR or t == token.RPAR:
2067 prevp = preceding_leaf(p)
2068 if not prevp or prevp.type != token.NUMBER:
2071 elif t == token.LSQB:
2074 elif prev.type != token.COMMA:
2077 elif p.type == syms.argument:
2079 if t == token.EQUAL:
2083 prevp = preceding_leaf(p)
2084 if not prevp or prevp.type == token.LPAR:
2087 elif prev.type in {token.EQUAL} | VARARGS_SPECIALS:
2090 elif p.type == syms.decorator:
2094 elif p.type == syms.dotted_name:
2098 prevp = preceding_leaf(p)
2099 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
2102 elif p.type == syms.classdef:
2106 if prev and prev.type == token.LPAR:
2109 elif p.type in {syms.subscript, syms.sliceop}:
2112 assert p.parent is not None, "subscripts are always parented"
2113 if p.parent.type == syms.subscriptlist:
2118 elif not complex_subscript:
2121 elif p.type == syms.atom:
2122 if prev and t == token.DOT:
2123 # dots, but not the first one.
2126 elif p.type == syms.dictsetmaker:
2128 if prev and prev.type == token.DOUBLESTAR:
2131 elif p.type in {syms.factor, syms.star_expr}:
2134 prevp = preceding_leaf(p)
2135 if not prevp or prevp.type in OPENING_BRACKETS:
2138 prevp_parent = prevp.parent
2139 assert prevp_parent is not None
2140 if prevp.type == token.COLON and prevp_parent.type in {
2146 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
2149 elif t in {token.NAME, token.NUMBER, token.STRING}:
2152 elif p.type == syms.import_from:
2154 if prev and prev.type == token.DOT:
2157 elif t == token.NAME:
2161 if prev and prev.type == token.DOT:
2164 elif p.type == syms.sliceop:
2170 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
2171 """Return the first leaf that precedes `node`, if any."""
2173 res = node.prev_sibling
2175 if isinstance(res, Leaf):
2179 return list(res.leaves())[-1]
2188 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
2189 """Return the child of `ancestor` that contains `descendant`."""
2190 node: Optional[LN] = descendant
2191 while node and node.parent != ancestor:
2196 def container_of(leaf: Leaf) -> LN:
2197 """Return `leaf` or one of its ancestors that is the topmost container of it.
2199 By "container" we mean a node where `leaf` is the very first child.
2201 same_prefix = leaf.prefix
2202 container: LN = leaf
2204 parent = container.parent
2208 if parent.children[0].prefix != same_prefix:
2211 if parent.type == syms.file_input:
2214 if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS:
2221 def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2222 """Return the priority of the `leaf` delimiter, given a line break after it.
2224 The delimiter priorities returned here are from those delimiters that would
2225 cause a line break after themselves.
2227 Higher numbers are higher priority.
2229 if leaf.type == token.COMMA:
2230 return COMMA_PRIORITY
2235 def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2236 """Return the priority of the `leaf` delimiter, given a line break before it.
2238 The delimiter priorities returned here are from those delimiters that would
2239 cause a line break before themselves.
2241 Higher numbers are higher priority.
2243 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
2244 # * and ** might also be MATH_OPERATORS but in this case they are not.
2245 # Don't treat them as a delimiter.
2249 leaf.type == token.DOT
2251 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
2252 and (previous is None or previous.type in CLOSING_BRACKETS)
2257 leaf.type in MATH_OPERATORS
2259 and leaf.parent.type not in {syms.factor, syms.star_expr}
2261 return MATH_PRIORITIES[leaf.type]
2263 if leaf.type in COMPARATORS:
2264 return COMPARATOR_PRIORITY
2267 leaf.type == token.STRING
2268 and previous is not None
2269 and previous.type == token.STRING
2271 return STRING_PRIORITY
2273 if leaf.type not in {token.NAME, token.ASYNC}:
2279 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
2280 or leaf.type == token.ASYNC
2283 not isinstance(leaf.prev_sibling, Leaf)
2284 or leaf.prev_sibling.value != "async"
2286 return COMPREHENSION_PRIORITY
2291 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
2293 return COMPREHENSION_PRIORITY
2295 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
2296 return TERNARY_PRIORITY
2298 if leaf.value == "is":
2299 return COMPARATOR_PRIORITY
2304 and leaf.parent.type in {syms.comp_op, syms.comparison}
2306 previous is not None
2307 and previous.type == token.NAME
2308 and previous.value == "not"
2311 return COMPARATOR_PRIORITY
2316 and leaf.parent.type == syms.comp_op
2318 previous is not None
2319 and previous.type == token.NAME
2320 and previous.value == "is"
2323 return COMPARATOR_PRIORITY
2325 if leaf.value in LOGIC_OPERATORS and leaf.parent:
2326 return LOGIC_PRIORITY
2331 FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"}
2332 FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"}
2335 def generate_comments(leaf: LN) -> Iterator[Leaf]:
2336 """Clean the prefix of the `leaf` and generate comments from it, if any.
2338 Comments in lib2to3 are shoved into the whitespace prefix. This happens
2339 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
2340 move because it does away with modifying the grammar to include all the
2341 possible places in which comments can be placed.
2343 The sad consequence for us though is that comments don't "belong" anywhere.
2344 This is why this function generates simple parentless Leaf objects for
2345 comments. We simply don't know what the correct parent should be.
2347 No matter though, we can live without this. We really only need to
2348 differentiate between inline and standalone comments. The latter don't
2349 share the line with any code.
2351 Inline comments are emitted as regular token.COMMENT leaves. Standalone
2352 are emitted with a fake STANDALONE_COMMENT token identifier.
2354 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
2355 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
2360 """Describes a piece of syntax that is a comment.
2362 It's not a :class:`blib2to3.pytree.Leaf` so that:
2364 * it can be cached (`Leaf` objects should not be reused more than once as
2365 they store their lineno, column, prefix, and parent information);
2366 * `newlines` and `consumed` fields are kept separate from the `value`. This
2367 simplifies handling of special marker comments like ``# fmt: off/on``.
2370 type: int # token.COMMENT or STANDALONE_COMMENT
2371 value: str # content of the comment
2372 newlines: int # how many newlines before the comment
2373 consumed: int # how many characters of the original leaf's prefix did we consume
2376 @lru_cache(maxsize=4096)
2377 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
2378 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
2379 result: List[ProtoComment] = []
2380 if not prefix or "#" not in prefix:
2386 for index, line in enumerate(prefix.split("\n")):
2387 consumed += len(line) + 1 # adding the length of the split '\n'
2388 line = line.lstrip()
2391 if not line.startswith("#"):
2392 # Escaped newlines outside of a comment are not really newlines at
2393 # all. We treat a single-line comment following an escaped newline
2394 # as a simple trailing comment.
2395 if line.endswith("\\"):
2399 if index == ignored_lines and not is_endmarker:
2400 comment_type = token.COMMENT # simple trailing comment
2402 comment_type = STANDALONE_COMMENT
2403 comment = make_comment(line)
2406 type=comment_type, value=comment, newlines=nlines, consumed=consumed
2413 def make_comment(content: str) -> str:
2414 """Return a consistently formatted comment from the given `content` string.
2416 All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single
2417 space between the hash sign and the content.
2419 If `content` didn't start with a hash sign, one is provided.
2421 content = content.rstrip()
2425 if content[0] == "#":
2426 content = content[1:]
2427 if content and content[0] not in " !:#'%":
2428 content = " " + content
2429 return "#" + content
2435 inner: bool = False,
2436 features: Collection[Feature] = (),
2437 ) -> Iterator[Line]:
2438 """Split a `line` into potentially many lines.
2440 They should fit in the allotted `line_length` but might not be able to.
2441 `inner` signifies that there were a pair of brackets somewhere around the
2442 current `line`, possibly transitively. This means we can fallback to splitting
2443 by delimiters if the LHS/RHS don't yield any results.
2445 `features` are syntactical features that may be used in the output.
2451 line_str = str(line).strip("\n")
2454 not line.contains_uncollapsable_type_comments()
2455 and not line.should_explode
2456 and not line.is_collection_with_optional_trailing_comma
2458 is_line_short_enough(line, line_length=line_length, line_str=line_str)
2459 or line.contains_unsplittable_type_ignore()
2465 split_funcs: List[SplitFunc]
2467 split_funcs = [left_hand_split]
2470 def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]:
2471 for omit in generate_trailers_to_omit(line, line_length):
2472 lines = list(right_hand_split(line, line_length, features, omit=omit))
2473 if is_line_short_enough(lines[0], line_length=line_length):
2477 # All splits failed, best effort split with no omits.
2478 # This mostly happens to multiline strings that are by definition
2479 # reported as not fitting a single line.
2480 # line_length=1 here was historically a bug that somehow became a feature.
2481 # See #762 and #781 for the full story.
2482 yield from right_hand_split(line, line_length=1, features=features)
2484 if line.inside_brackets:
2485 split_funcs = [delimiter_split, standalone_comment_split, rhs]
2488 for split_func in split_funcs:
2489 # We are accumulating lines in `result` because we might want to abort
2490 # mission and return the original line in the end, or attempt a different
2492 result: List[Line] = []
2494 for l in split_func(line, features):
2495 if str(l).strip("\n") == line_str:
2496 raise CannotSplit("Split function returned an unchanged result")
2500 l, line_length=line_length, inner=True, features=features
2514 def left_hand_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2515 """Split line into many lines, starting with the first matching bracket pair.
2517 Note: this usually looks weird, only use this for function definitions.
2518 Prefer RHS otherwise. This is why this function is not symmetrical with
2519 :func:`right_hand_split` which also handles optional parentheses.
2521 tail_leaves: List[Leaf] = []
2522 body_leaves: List[Leaf] = []
2523 head_leaves: List[Leaf] = []
2524 current_leaves = head_leaves
2525 matching_bracket: Optional[Leaf] = None
2526 for leaf in line.leaves:
2528 current_leaves is body_leaves
2529 and leaf.type in CLOSING_BRACKETS
2530 and leaf.opening_bracket is matching_bracket
2532 current_leaves = tail_leaves if body_leaves else head_leaves
2533 current_leaves.append(leaf)
2534 if current_leaves is head_leaves:
2535 if leaf.type in OPENING_BRACKETS:
2536 matching_bracket = leaf
2537 current_leaves = body_leaves
2538 if not matching_bracket:
2539 raise CannotSplit("No brackets found")
2541 head = bracket_split_build_line(head_leaves, line, matching_bracket)
2542 body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True)
2543 tail = bracket_split_build_line(tail_leaves, line, matching_bracket)
2544 bracket_split_succeeded_or_raise(head, body, tail)
2545 for result in (head, body, tail):
2550 def right_hand_split(
2553 features: Collection[Feature] = (),
2554 omit: Collection[LeafID] = (),
2555 ) -> Iterator[Line]:
2556 """Split line into many lines, starting with the last matching bracket pair.
2558 If the split was by optional parentheses, attempt splitting without them, too.
2559 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2562 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2564 tail_leaves: List[Leaf] = []
2565 body_leaves: List[Leaf] = []
2566 head_leaves: List[Leaf] = []
2567 current_leaves = tail_leaves
2568 opening_bracket: Optional[Leaf] = None
2569 closing_bracket: Optional[Leaf] = None
2570 for leaf in reversed(line.leaves):
2571 if current_leaves is body_leaves:
2572 if leaf is opening_bracket:
2573 current_leaves = head_leaves if body_leaves else tail_leaves
2574 current_leaves.append(leaf)
2575 if current_leaves is tail_leaves:
2576 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2577 opening_bracket = leaf.opening_bracket
2578 closing_bracket = leaf
2579 current_leaves = body_leaves
2580 if not (opening_bracket and closing_bracket and head_leaves):
2581 # If there is no opening or closing_bracket that means the split failed and
2582 # all content is in the tail. Otherwise, if `head_leaves` are empty, it means
2583 # the matching `opening_bracket` wasn't available on `line` anymore.
2584 raise CannotSplit("No brackets found")
2586 tail_leaves.reverse()
2587 body_leaves.reverse()
2588 head_leaves.reverse()
2589 head = bracket_split_build_line(head_leaves, line, opening_bracket)
2590 body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True)
2591 tail = bracket_split_build_line(tail_leaves, line, opening_bracket)
2592 bracket_split_succeeded_or_raise(head, body, tail)
2594 # the body shouldn't be exploded
2595 not body.should_explode
2596 # the opening bracket is an optional paren
2597 and opening_bracket.type == token.LPAR
2598 and not opening_bracket.value
2599 # the closing bracket is an optional paren
2600 and closing_bracket.type == token.RPAR
2601 and not closing_bracket.value
2602 # it's not an import (optional parens are the only thing we can split on
2603 # in this case; attempting a split without them is a waste of time)
2604 and not line.is_import
2605 # there are no standalone comments in the body
2606 and not body.contains_standalone_comments(0)
2607 # and we can actually remove the parens
2608 and can_omit_invisible_parens(body, line_length)
2610 omit = {id(closing_bracket), *omit}
2612 yield from right_hand_split(line, line_length, features=features, omit=omit)
2618 or is_line_short_enough(body, line_length=line_length)
2621 "Splitting failed, body is still too long and can't be split."
2624 elif head.contains_multiline_strings() or tail.contains_multiline_strings():
2626 "The current optional pair of parentheses is bound to fail to "
2627 "satisfy the splitting algorithm because the head or the tail "
2628 "contains multiline strings which by definition never fit one "
2632 ensure_visible(opening_bracket)
2633 ensure_visible(closing_bracket)
2634 for result in (head, body, tail):
2639 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2640 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2642 Do nothing otherwise.
2644 A left- or right-hand split is based on a pair of brackets. Content before
2645 (and including) the opening bracket is left on one line, content inside the
2646 brackets is put on a separate line, and finally content starting with and
2647 following the closing bracket is put on a separate line.
2649 Those are called `head`, `body`, and `tail`, respectively. If the split
2650 produced the same line (all content in `head`) or ended up with an empty `body`
2651 and the `tail` is just the closing bracket, then it's considered failed.
2653 tail_len = len(str(tail).strip())
2656 raise CannotSplit("Splitting brackets produced the same line")
2660 f"Splitting brackets on an empty body to save "
2661 f"{tail_len} characters is not worth it"
2665 def bracket_split_build_line(
2666 leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False
2668 """Return a new line with given `leaves` and respective comments from `original`.
2670 If `is_body` is True, the result line is one-indented inside brackets and as such
2671 has its first leaf's prefix normalized and a trailing comma added when expected.
2673 result = Line(depth=original.depth)
2675 result.inside_brackets = True
2678 # Since body is a new indent level, remove spurious leading whitespace.
2679 normalize_prefix(leaves[0], inside_brackets=True)
2680 # Ensure a trailing comma for imports and standalone function arguments, but
2681 # be careful not to add one after any comments or within type annotations.
2684 and opening_bracket.value == "("
2685 and not any(l.type == token.COMMA for l in leaves)
2688 if original.is_import or no_commas:
2689 for i in range(len(leaves) - 1, -1, -1):
2690 if leaves[i].type == STANDALONE_COMMENT:
2693 if leaves[i].type != token.COMMA:
2694 leaves.insert(i + 1, Leaf(token.COMMA, ","))
2699 result.append(leaf, preformatted=True)
2700 for comment_after in original.comments_after(leaf):
2701 result.append(comment_after, preformatted=True)
2703 result.should_explode = should_explode(result, opening_bracket)
2707 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2708 """Normalize prefix of the first leaf in every line returned by `split_func`.
2710 This is a decorator over relevant split functions.
2714 def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2715 for l in split_func(line, features):
2716 normalize_prefix(l.leaves[0], inside_brackets=True)
2719 return split_wrapper
2722 @dont_increase_indentation
2723 def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2724 """Split according to delimiters of the highest priority.
2726 If the appropriate Features are given, the split will add trailing commas
2727 also in function signatures and calls that contain `*` and `**`.
2730 last_leaf = line.leaves[-1]
2732 raise CannotSplit("Line empty")
2734 bt = line.bracket_tracker
2736 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2738 raise CannotSplit("No delimiters found")
2740 if delimiter_priority == DOT_PRIORITY:
2741 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2742 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2744 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2745 lowest_depth = sys.maxsize
2746 trailing_comma_safe = True
2748 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2749 """Append `leaf` to current line or to new line if appending impossible."""
2750 nonlocal current_line
2752 current_line.append_safe(leaf, preformatted=True)
2756 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2757 current_line.append(leaf)
2759 for leaf in line.leaves:
2760 yield from append_to_line(leaf)
2762 for comment_after in line.comments_after(leaf):
2763 yield from append_to_line(comment_after)
2765 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2766 if leaf.bracket_depth == lowest_depth:
2767 if is_vararg(leaf, within={syms.typedargslist}):
2768 trailing_comma_safe = (
2769 trailing_comma_safe and Feature.TRAILING_COMMA_IN_DEF in features
2771 elif is_vararg(leaf, within={syms.arglist, syms.argument}):
2772 trailing_comma_safe = (
2773 trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features
2776 leaf_priority = bt.delimiters.get(id(leaf))
2777 if leaf_priority == delimiter_priority:
2780 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2784 and delimiter_priority == COMMA_PRIORITY
2785 and current_line.leaves[-1].type != token.COMMA
2786 and current_line.leaves[-1].type != STANDALONE_COMMENT
2788 current_line.append(Leaf(token.COMMA, ","))
2792 @dont_increase_indentation
2793 def standalone_comment_split(
2794 line: Line, features: Collection[Feature] = ()
2795 ) -> Iterator[Line]:
2796 """Split standalone comments from the rest of the line."""
2797 if not line.contains_standalone_comments(0):
2798 raise CannotSplit("Line does not have any standalone comments")
2800 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2802 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2803 """Append `leaf` to current line or to new line if appending impossible."""
2804 nonlocal current_line
2806 current_line.append_safe(leaf, preformatted=True)
2810 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2811 current_line.append(leaf)
2813 for leaf in line.leaves:
2814 yield from append_to_line(leaf)
2816 for comment_after in line.comments_after(leaf):
2817 yield from append_to_line(comment_after)
2823 def is_import(leaf: Leaf) -> bool:
2824 """Return True if the given leaf starts an import statement."""
2831 (v == "import" and p and p.type == syms.import_name)
2832 or (v == "from" and p and p.type == syms.import_from)
2837 def is_type_comment(leaf: Leaf, suffix: str = "") -> bool:
2838 """Return True if the given leaf is a special comment.
2839 Only returns true for type comments for now."""
2842 return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix)
2845 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2846 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2849 Note: don't use backslashes for formatting or you'll lose your voting rights.
2851 if not inside_brackets:
2852 spl = leaf.prefix.split("#")
2853 if "\\" not in spl[0]:
2854 nl_count = spl[-1].count("\n")
2857 leaf.prefix = "\n" * nl_count
2863 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2864 """Make all string prefixes lowercase.
2866 If remove_u_prefix is given, also removes any u prefix from the string.
2868 Note: Mutates its argument.
2870 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2871 assert match is not None, f"failed to match string {leaf.value!r}"
2872 orig_prefix = match.group(1)
2873 new_prefix = orig_prefix.replace("F", "f").replace("B", "b").replace("U", "u")
2875 new_prefix = new_prefix.replace("u", "")
2876 leaf.value = f"{new_prefix}{match.group(2)}"
2879 def normalize_string_quotes(leaf: Leaf) -> None:
2880 """Prefer double quotes but only if it doesn't cause more escaping.
2882 Adds or removes backslashes as appropriate. Doesn't parse and fix
2883 strings nested in f-strings (yet).
2885 Note: Mutates its argument.
2887 value = leaf.value.lstrip("furbFURB")
2888 if value[:3] == '"""':
2891 elif value[:3] == "'''":
2894 elif value[0] == '"':
2900 first_quote_pos = leaf.value.find(orig_quote)
2901 if first_quote_pos == -1:
2902 return # There's an internal error
2904 prefix = leaf.value[:first_quote_pos]
2905 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2906 escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}")
2907 escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}")
2908 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2909 if "r" in prefix.casefold():
2910 if unescaped_new_quote.search(body):
2911 # There's at least one unescaped new_quote in this raw string
2912 # so converting is impossible
2915 # Do not introduce or remove backslashes in raw strings
2918 # remove unnecessary escapes
2919 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2920 if body != new_body:
2921 # Consider the string without unnecessary escapes as the original
2923 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2924 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2925 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2926 if "f" in prefix.casefold():
2927 matches = re.findall(
2929 (?:[^{]|^)\{ # start of the string or a non-{ followed by a single {
2930 ([^{].*?) # contents of the brackets except if begins with {{
2931 \}(?:[^}]|$) # A } followed by end of the string or a non-}
2938 # Do not introduce backslashes in interpolated expressions
2941 if new_quote == '"""' and new_body[-1:] == '"':
2943 new_body = new_body[:-1] + '\\"'
2944 orig_escape_count = body.count("\\")
2945 new_escape_count = new_body.count("\\")
2946 if new_escape_count > orig_escape_count:
2947 return # Do not introduce more escaping
2949 if new_escape_count == orig_escape_count and orig_quote == '"':
2950 return # Prefer double quotes
2952 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2955 def normalize_numeric_literal(leaf: Leaf) -> None:
2956 """Normalizes numeric (float, int, and complex) literals.
2958 All letters used in the representation are normalized to lowercase (except
2959 in Python 2 long literals).
2961 text = leaf.value.lower()
2962 if text.startswith(("0o", "0b")):
2963 # Leave octal and binary literals alone.
2965 elif text.startswith("0x"):
2966 # Change hex literals to upper case.
2967 before, after = text[:2], text[2:]
2968 text = f"{before}{after.upper()}"
2970 before, after = text.split("e")
2972 if after.startswith("-"):
2975 elif after.startswith("+"):
2977 before = format_float_or_int_string(before)
2978 text = f"{before}e{sign}{after}"
2979 elif text.endswith(("j", "l")):
2982 # Capitalize in "2L" because "l" looks too similar to "1".
2985 text = f"{format_float_or_int_string(number)}{suffix}"
2987 text = format_float_or_int_string(text)
2991 def format_float_or_int_string(text: str) -> str:
2992 """Formats a float string like "1.0"."""
2996 before, after = text.split(".")
2997 return f"{before or 0}.{after or 0}"
3000 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
3001 """Make existing optional parentheses invisible or create new ones.
3003 `parens_after` is a set of string leaf values immediately after which parens
3006 Standardizes on visible parentheses for single-element tuples, and keeps
3007 existing visible parentheses for other tuples and generator expressions.
3009 for pc in list_comments(node.prefix, is_endmarker=False):
3010 if pc.value in FMT_OFF:
3011 # This `node` has a prefix with `# fmt: off`, don't mess with parens.
3014 for index, child in enumerate(list(node.children)):
3015 # Add parentheses around long tuple unpacking in assignments.
3018 and isinstance(child, Node)
3019 and child.type == syms.testlist_star_expr
3024 if is_walrus_assignment(child):
3027 if child.type == syms.atom:
3028 if maybe_make_parens_invisible_in_atom(child, parent=node):
3029 wrap_in_parentheses(node, child, visible=False)
3030 elif is_one_tuple(child):
3031 wrap_in_parentheses(node, child, visible=True)
3032 elif node.type == syms.import_from:
3033 # "import from" nodes store parentheses directly as part of
3035 if child.type == token.LPAR:
3036 # make parentheses invisible
3037 child.value = "" # type: ignore
3038 node.children[-1].value = "" # type: ignore
3039 elif child.type != token.STAR:
3040 # insert invisible parentheses
3041 node.insert_child(index, Leaf(token.LPAR, ""))
3042 node.append_child(Leaf(token.RPAR, ""))
3045 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
3046 wrap_in_parentheses(node, child, visible=False)
3048 check_lpar = isinstance(child, Leaf) and child.value in parens_after
3051 def normalize_fmt_off(node: Node) -> None:
3052 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
3055 try_again = convert_one_fmt_off_pair(node)
3058 def convert_one_fmt_off_pair(node: Node) -> bool:
3059 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
3061 Returns True if a pair was converted.
3063 for leaf in node.leaves():
3064 previous_consumed = 0
3065 for comment in list_comments(leaf.prefix, is_endmarker=False):
3066 if comment.value in FMT_OFF:
3067 # We only want standalone comments. If there's no previous leaf or
3068 # the previous leaf is indentation, it's a standalone comment in
3070 if comment.type != STANDALONE_COMMENT:
3071 prev = preceding_leaf(leaf)
3072 if prev and prev.type not in WHITESPACE:
3075 ignored_nodes = list(generate_ignored_nodes(leaf))
3076 if not ignored_nodes:
3079 first = ignored_nodes[0] # Can be a container node with the `leaf`.
3080 parent = first.parent
3081 prefix = first.prefix
3082 first.prefix = prefix[comment.consumed :]
3084 comment.value + "\n" + "".join(str(n) for n in ignored_nodes)
3086 if hidden_value.endswith("\n"):
3087 # That happens when one of the `ignored_nodes` ended with a NEWLINE
3088 # leaf (possibly followed by a DEDENT).
3089 hidden_value = hidden_value[:-1]
3090 first_idx: Optional[int] = None
3091 for ignored in ignored_nodes:
3092 index = ignored.remove()
3093 if first_idx is None:
3095 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
3096 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
3097 parent.insert_child(
3102 prefix=prefix[:previous_consumed] + "\n" * comment.newlines,
3107 previous_consumed = comment.consumed
3112 def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]:
3113 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
3115 Stops at the end of the block.
3117 container: Optional[LN] = container_of(leaf)
3118 while container is not None and container.type != token.ENDMARKER:
3120 for comment in list_comments(container.prefix, is_endmarker=False):
3121 if comment.value in FMT_ON:
3123 elif comment.value in FMT_OFF:
3130 container = container.next_sibling
3133 def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool:
3134 """If it's safe, make the parens in the atom `node` invisible, recursively.
3135 Additionally, remove repeated, adjacent invisible parens from the atom `node`
3136 as they are redundant.
3138 Returns whether the node should itself be wrapped in invisible parentheses.
3142 node.type != syms.atom
3143 or is_empty_tuple(node)
3144 or is_one_tuple(node)
3145 or (is_yield(node) and parent.type != syms.expr_stmt)
3146 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
3150 first = node.children[0]
3151 last = node.children[-1]
3152 if first.type == token.LPAR and last.type == token.RPAR:
3153 middle = node.children[1]
3154 # make parentheses invisible
3155 first.value = "" # type: ignore
3156 last.value = "" # type: ignore
3157 maybe_make_parens_invisible_in_atom(middle, parent=parent)
3159 if is_atom_with_invisible_parens(middle):
3160 # Strip the invisible parens from `middle` by replacing
3161 # it with the child in-between the invisible parens
3162 middle.replace(middle.children[1])
3169 def is_atom_with_invisible_parens(node: LN) -> bool:
3170 """Given a `LN`, determines whether it's an atom `node` with invisible
3171 parens. Useful in dedupe-ing and normalizing parens.
3173 if isinstance(node, Leaf) or node.type != syms.atom:
3176 first, last = node.children[0], node.children[-1]
3178 isinstance(first, Leaf)
3179 and first.type == token.LPAR
3180 and first.value == ""
3181 and isinstance(last, Leaf)
3182 and last.type == token.RPAR
3183 and last.value == ""
3187 def is_empty_tuple(node: LN) -> bool:
3188 """Return True if `node` holds an empty tuple."""
3190 node.type == syms.atom
3191 and len(node.children) == 2
3192 and node.children[0].type == token.LPAR
3193 and node.children[1].type == token.RPAR
3197 def unwrap_singleton_parenthesis(node: LN) -> Optional[LN]:
3198 """Returns `wrapped` if `node` is of the shape ( wrapped ).
3200 Parenthesis can be optional. Returns None otherwise"""
3201 if len(node.children) != 3:
3204 lpar, wrapped, rpar = node.children
3205 if not (lpar.type == token.LPAR and rpar.type == token.RPAR):
3211 def wrap_in_parentheses(parent: Node, child: LN, *, visible: bool = True) -> None:
3212 """Wrap `child` in parentheses.
3214 This replaces `child` with an atom holding the parentheses and the old
3215 child. That requires moving the prefix.
3217 If `visible` is False, the leaves will be valueless (and thus invisible).
3219 lpar = Leaf(token.LPAR, "(" if visible else "")
3220 rpar = Leaf(token.RPAR, ")" if visible else "")
3221 prefix = child.prefix
3223 index = child.remove() or 0
3224 new_child = Node(syms.atom, [lpar, child, rpar])
3225 new_child.prefix = prefix
3226 parent.insert_child(index, new_child)
3229 def is_one_tuple(node: LN) -> bool:
3230 """Return True if `node` holds a tuple with one element, with or without parens."""
3231 if node.type == syms.atom:
3232 gexp = unwrap_singleton_parenthesis(node)
3233 if gexp is None or gexp.type != syms.testlist_gexp:
3236 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
3239 node.type in IMPLICIT_TUPLE
3240 and len(node.children) == 2
3241 and node.children[1].type == token.COMMA
3245 def is_walrus_assignment(node: LN) -> bool:
3246 """Return True iff `node` is of the shape ( test := test )"""
3247 inner = unwrap_singleton_parenthesis(node)
3248 return inner is not None and inner.type == syms.namedexpr_test
3251 def is_yield(node: LN) -> bool:
3252 """Return True if `node` holds a `yield` or `yield from` expression."""
3253 if node.type == syms.yield_expr:
3256 if node.type == token.NAME and node.value == "yield": # type: ignore
3259 if node.type != syms.atom:
3262 if len(node.children) != 3:
3265 lpar, expr, rpar = node.children
3266 if lpar.type == token.LPAR and rpar.type == token.RPAR:
3267 return is_yield(expr)
3272 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
3273 """Return True if `leaf` is a star or double star in a vararg or kwarg.
3275 If `within` includes VARARGS_PARENTS, this applies to function signatures.
3276 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
3277 extended iterable unpacking (PEP 3132) and additional unpacking
3278 generalizations (PEP 448).
3280 if leaf.type not in VARARGS_SPECIALS or not leaf.parent:
3284 if p.type == syms.star_expr:
3285 # Star expressions are also used as assignment targets in extended
3286 # iterable unpacking (PEP 3132). See what its parent is instead.
3292 return p.type in within
3295 def is_multiline_string(leaf: Leaf) -> bool:
3296 """Return True if `leaf` is a multiline string that actually spans many lines."""
3297 value = leaf.value.lstrip("furbFURB")
3298 return value[:3] in {'"""', "'''"} and "\n" in value
3301 def is_stub_suite(node: Node) -> bool:
3302 """Return True if `node` is a suite with a stub body."""
3304 len(node.children) != 4
3305 or node.children[0].type != token.NEWLINE
3306 or node.children[1].type != token.INDENT
3307 or node.children[3].type != token.DEDENT
3311 return is_stub_body(node.children[2])
3314 def is_stub_body(node: LN) -> bool:
3315 """Return True if `node` is a simple statement containing an ellipsis."""
3316 if not isinstance(node, Node) or node.type != syms.simple_stmt:
3319 if len(node.children) != 2:
3322 child = node.children[0]
3324 child.type == syms.atom
3325 and len(child.children) == 3
3326 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
3330 def max_delimiter_priority_in_atom(node: LN) -> Priority:
3331 """Return maximum delimiter priority inside `node`.
3333 This is specific to atoms with contents contained in a pair of parentheses.
3334 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
3336 if node.type != syms.atom:
3339 first = node.children[0]
3340 last = node.children[-1]
3341 if not (first.type == token.LPAR and last.type == token.RPAR):
3344 bt = BracketTracker()
3345 for c in node.children[1:-1]:
3346 if isinstance(c, Leaf):
3349 for leaf in c.leaves():
3352 return bt.max_delimiter_priority()
3358 def ensure_visible(leaf: Leaf) -> None:
3359 """Make sure parentheses are visible.
3361 They could be invisible as part of some statements (see
3362 :func:`normalize_invisible_parens` and :func:`visit_import_from`).
3364 if leaf.type == token.LPAR:
3366 elif leaf.type == token.RPAR:
3370 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
3371 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
3374 opening_bracket.parent
3375 and opening_bracket.parent.type in {syms.atom, syms.import_from}
3376 and opening_bracket.value in "[{("
3381 last_leaf = line.leaves[-1]
3382 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
3383 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
3384 except (IndexError, ValueError):
3387 return max_priority == COMMA_PRIORITY
3390 def get_features_used(node: Node) -> Set[Feature]:
3391 """Return a set of (relatively) new Python features used in this file.
3393 Currently looking for:
3395 - underscores in numeric literals;
3396 - trailing commas after * or ** in function signatures and calls;
3397 - positional only arguments in function signatures and lambdas;
3399 features: Set[Feature] = set()
3400 for n in node.pre_order():
3401 if n.type == token.STRING:
3402 value_head = n.value[:2] # type: ignore
3403 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
3404 features.add(Feature.F_STRINGS)
3406 elif n.type == token.NUMBER:
3407 if "_" in n.value: # type: ignore
3408 features.add(Feature.NUMERIC_UNDERSCORES)
3410 elif n.type == token.SLASH:
3411 if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}:
3412 features.add(Feature.POS_ONLY_ARGUMENTS)
3414 elif n.type == token.COLONEQUAL:
3415 features.add(Feature.ASSIGNMENT_EXPRESSIONS)
3418 n.type in {syms.typedargslist, syms.arglist}
3420 and n.children[-1].type == token.COMMA
3422 if n.type == syms.typedargslist:
3423 feature = Feature.TRAILING_COMMA_IN_DEF
3425 feature = Feature.TRAILING_COMMA_IN_CALL
3427 for ch in n.children:
3428 if ch.type in STARS:
3429 features.add(feature)
3431 if ch.type == syms.argument:
3432 for argch in ch.children:
3433 if argch.type in STARS:
3434 features.add(feature)
3439 def detect_target_versions(node: Node) -> Set[TargetVersion]:
3440 """Detect the version to target based on the nodes used."""
3441 features = get_features_used(node)
3443 version for version in TargetVersion if features <= VERSION_TO_FEATURES[version]
3447 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
3448 """Generate sets of closing bracket IDs that should be omitted in a RHS.
3450 Brackets can be omitted if the entire trailer up to and including
3451 a preceding closing bracket fits in one line.
3453 Yielded sets are cumulative (contain results of previous yields, too). First
3457 omit: Set[LeafID] = set()
3460 length = 4 * line.depth
3461 opening_bracket: Optional[Leaf] = None
3462 closing_bracket: Optional[Leaf] = None
3463 inner_brackets: Set[LeafID] = set()
3464 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
3465 length += leaf_length
3466 if length > line_length:
3469 has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
3470 if leaf.type == STANDALONE_COMMENT or has_inline_comment:
3474 if leaf is opening_bracket:
3475 opening_bracket = None
3476 elif leaf.type in CLOSING_BRACKETS:
3477 inner_brackets.add(id(leaf))
3478 elif leaf.type in CLOSING_BRACKETS:
3479 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
3480 # Empty brackets would fail a split so treat them as "inner"
3481 # brackets (e.g. only add them to the `omit` set if another
3482 # pair of brackets was good enough.
3483 inner_brackets.add(id(leaf))
3487 omit.add(id(closing_bracket))
3488 omit.update(inner_brackets)
3489 inner_brackets.clear()
3493 opening_bracket = leaf.opening_bracket
3494 closing_bracket = leaf
3497 def get_future_imports(node: Node) -> Set[str]:
3498 """Return a set of __future__ imports in the file."""
3499 imports: Set[str] = set()
3501 def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]:
3502 for child in children:
3503 if isinstance(child, Leaf):
3504 if child.type == token.NAME:
3507 elif child.type == syms.import_as_name:
3508 orig_name = child.children[0]
3509 assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
3510 assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
3511 yield orig_name.value
3513 elif child.type == syms.import_as_names:
3514 yield from get_imports_from_children(child.children)
3517 raise AssertionError("Invalid syntax parsing imports")
3519 for child in node.children:
3520 if child.type != syms.simple_stmt:
3523 first_child = child.children[0]
3524 if isinstance(first_child, Leaf):
3525 # Continue looking if we see a docstring; otherwise stop.
3527 len(child.children) == 2
3528 and first_child.type == token.STRING
3529 and child.children[1].type == token.NEWLINE
3535 elif first_child.type == syms.import_from:
3536 module_name = first_child.children[1]
3537 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
3540 imports |= set(get_imports_from_children(first_child.children[3:]))
3548 def get_gitignore(root: Path) -> PathSpec:
3549 """ Return a PathSpec matching gitignore content if present."""
3550 gitignore = root / ".gitignore"
3551 lines: List[str] = []
3552 if gitignore.is_file():
3553 with gitignore.open() as gf:
3554 lines = gf.readlines()
3555 return PathSpec.from_lines("gitwildmatch", lines)
3558 def gen_python_files_in_dir(
3561 include: Pattern[str],
3562 exclude: Pattern[str],
3564 gitignore: PathSpec,
3565 ) -> Iterator[Path]:
3566 """Generate all files under `path` whose paths are not excluded by the
3567 `exclude` regex, but are included by the `include` regex.
3569 Symbolic links pointing outside of the `root` directory are ignored.
3571 `report` is where output about exclusions goes.
3573 assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
3574 for child in path.iterdir():
3575 # First ignore files matching .gitignore
3576 if gitignore.match_file(child.as_posix()):
3577 report.path_ignored(child, "matches the .gitignore file content")
3580 # Then ignore with `exclude` option.
3582 normalized_path = "/" + child.resolve().relative_to(root).as_posix()
3583 except OSError as e:
3584 report.path_ignored(child, f"cannot be read because {e}")
3588 if child.is_symlink():
3589 report.path_ignored(
3590 child, f"is a symbolic link that points outside {root}"
3597 normalized_path += "/"
3599 exclude_match = exclude.search(normalized_path)
3600 if exclude_match and exclude_match.group(0):
3601 report.path_ignored(child, "matches the --exclude regular expression")
3605 yield from gen_python_files_in_dir(
3606 child, root, include, exclude, report, gitignore
3609 elif child.is_file():
3610 include_match = include.search(normalized_path)
3616 def find_project_root(srcs: Iterable[str]) -> Path:
3617 """Return a directory containing .git, .hg, or pyproject.toml.
3619 That directory can be one of the directories passed in `srcs` or their
3622 If no directory in the tree contains a marker that would specify it's the
3623 project root, the root of the file system is returned.
3626 return Path("/").resolve()
3628 common_base = min(Path(src).resolve() for src in srcs)
3629 if common_base.is_dir():
3630 # Append a fake file so `parents` below returns `common_base_dir`, too.
3631 common_base /= "fake-file"
3632 for directory in common_base.parents:
3633 if (directory / ".git").exists():
3636 if (directory / ".hg").is_dir():
3639 if (directory / "pyproject.toml").is_file():
3647 """Provides a reformatting counter. Can be rendered with `str(report)`."""
3652 verbose: bool = False
3653 change_count: int = 0
3655 failure_count: int = 0
3657 def done(self, src: Path, changed: Changed) -> None:
3658 """Increment the counter for successful reformatting. Write out a message."""
3659 if changed is Changed.YES:
3660 reformatted = "would reformat" if self.check or self.diff else "reformatted"
3661 if self.verbose or not self.quiet:
3662 out(f"{reformatted} {src}")
3663 self.change_count += 1
3666 if changed is Changed.NO:
3667 msg = f"{src} already well formatted, good job."
3669 msg = f"{src} wasn't modified on disk since last run."
3670 out(msg, bold=False)
3671 self.same_count += 1
3673 def failed(self, src: Path, message: str) -> None:
3674 """Increment the counter for failed reformatting. Write out a message."""
3675 err(f"error: cannot format {src}: {message}")
3676 self.failure_count += 1
3678 def path_ignored(self, path: Path, message: str) -> None:
3680 out(f"{path} ignored: {message}", bold=False)
3683 def return_code(self) -> int:
3684 """Return the exit code that the app should use.
3686 This considers the current state of changed files and failures:
3687 - if there were any failures, return 123;
3688 - if any files were changed and --check is being used, return 1;
3689 - otherwise return 0.
3691 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
3692 # 126 we have special return codes reserved by the shell.
3693 if self.failure_count:
3696 elif self.change_count and self.check:
3701 def __str__(self) -> str:
3702 """Render a color report of the current state.
3704 Use `click.unstyle` to remove colors.
3706 if self.check or self.diff:
3707 reformatted = "would be reformatted"
3708 unchanged = "would be left unchanged"
3709 failed = "would fail to reformat"
3711 reformatted = "reformatted"
3712 unchanged = "left unchanged"
3713 failed = "failed to reformat"
3715 if self.change_count:
3716 s = "s" if self.change_count > 1 else ""
3718 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
3721 s = "s" if self.same_count > 1 else ""
3722 report.append(f"{self.same_count} file{s} {unchanged}")
3723 if self.failure_count:
3724 s = "s" if self.failure_count > 1 else ""
3726 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
3728 return ", ".join(report) + "."
3731 def parse_ast(src: str) -> Union[ast.AST, ast3.AST, ast27.AST]:
3732 filename = "<unknown>"
3733 if sys.version_info >= (3, 8):
3734 # TODO: support Python 4+ ;)
3735 for minor_version in range(sys.version_info[1], 4, -1):
3737 return ast.parse(src, filename, feature_version=(3, minor_version))
3741 for feature_version in (7, 6):
3743 return ast3.parse(src, filename, feature_version=feature_version)
3747 return ast27.parse(src)
3750 def _fixup_ast_constants(
3751 node: Union[ast.AST, ast3.AST, ast27.AST]
3752 ) -> Union[ast.AST, ast3.AST, ast27.AST]:
3753 """Map ast nodes deprecated in 3.8 to Constant."""
3754 if isinstance(node, (ast.Str, ast3.Str, ast27.Str, ast.Bytes, ast3.Bytes)):
3755 return ast.Constant(value=node.s)
3757 if isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
3758 return ast.Constant(value=node.n)
3760 if isinstance(node, (ast.NameConstant, ast3.NameConstant)):
3761 return ast.Constant(value=node.value)
3766 def assert_equivalent(src: str, dst: str) -> None:
3767 """Raise AssertionError if `src` and `dst` aren't equivalent."""
3769 def _v(node: Union[ast.AST, ast3.AST, ast27.AST], depth: int = 0) -> Iterator[str]:
3770 """Simple visitor generating strings to compare ASTs by content."""
3772 node = _fixup_ast_constants(node)
3774 yield f"{' ' * depth}{node.__class__.__name__}("
3776 for field in sorted(node._fields): # noqa: F402
3777 # TypeIgnore has only one field 'lineno' which breaks this comparison
3778 type_ignore_classes = (ast3.TypeIgnore, ast27.TypeIgnore)
3779 if sys.version_info >= (3, 8):
3780 type_ignore_classes += (ast.TypeIgnore,)
3781 if isinstance(node, type_ignore_classes):
3785 value = getattr(node, field)
3786 except AttributeError:
3789 yield f"{' ' * (depth+1)}{field}="
3791 if isinstance(value, list):
3793 # Ignore nested tuples within del statements, because we may insert
3794 # parentheses and they change the AST.
3797 and isinstance(node, (ast.Delete, ast3.Delete, ast27.Delete))
3798 and isinstance(item, (ast.Tuple, ast3.Tuple, ast27.Tuple))
3800 for item in item.elts:
3801 yield from _v(item, depth + 2)
3803 elif isinstance(item, (ast.AST, ast3.AST, ast27.AST)):
3804 yield from _v(item, depth + 2)
3806 elif isinstance(value, (ast.AST, ast3.AST, ast27.AST)):
3807 yield from _v(value, depth + 2)
3810 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
3812 yield f"{' ' * depth}) # /{node.__class__.__name__}"
3815 src_ast = parse_ast(src)
3816 except Exception as exc:
3817 raise AssertionError(
3818 f"cannot use --safe with this file; failed to parse source file. "
3819 f"AST error message: {exc}"
3823 dst_ast = parse_ast(dst)
3824 except Exception as exc:
3825 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
3826 raise AssertionError(
3827 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
3828 f"Please report a bug on https://github.com/psf/black/issues. "
3829 f"This invalid output might be helpful: {log}"
3832 src_ast_str = "\n".join(_v(src_ast))
3833 dst_ast_str = "\n".join(_v(dst_ast))
3834 if src_ast_str != dst_ast_str:
3835 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
3836 raise AssertionError(
3837 f"INTERNAL ERROR: Black produced code that is not equivalent to "
3839 f"Please report a bug on https://github.com/psf/black/issues. "
3840 f"This diff might be helpful: {log}"
3844 def assert_stable(src: str, dst: str, mode: Mode) -> None:
3845 """Raise AssertionError if `dst` reformats differently the second time."""
3846 newdst = format_str(dst, mode=mode)
3849 diff(src, dst, "source", "first pass"),
3850 diff(dst, newdst, "first pass", "second pass"),
3852 raise AssertionError(
3853 f"INTERNAL ERROR: Black produced different code on the second pass "
3854 f"of the formatter. "
3855 f"Please report a bug on https://github.com/psf/black/issues. "
3856 f"This diff might be helpful: {log}"
3860 @mypyc_attr(patchable=True)
3861 def dump_to_file(*output: str) -> str:
3862 """Dump `output` to a temporary file. Return path to the file."""
3863 with tempfile.NamedTemporaryFile(
3864 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
3866 for lines in output:
3868 if lines and lines[-1] != "\n":
3874 def nullcontext() -> Iterator[None]:
3875 """Return an empty context manager.
3877 To be used like `nullcontext` in Python 3.7.
3882 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
3883 """Return a unified diff string between strings `a` and `b`."""
3886 a_lines = [line + "\n" for line in a.splitlines()]
3887 b_lines = [line + "\n" for line in b.splitlines()]
3889 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
3893 def cancel(tasks: Iterable["asyncio.Task[Any]"]) -> None:
3894 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
3900 def shutdown(loop: asyncio.AbstractEventLoop) -> None:
3901 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
3903 if sys.version_info[:2] >= (3, 7):
3904 all_tasks = asyncio.all_tasks
3906 all_tasks = asyncio.Task.all_tasks
3907 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
3908 to_cancel = [task for task in all_tasks(loop) if not task.done()]
3912 for task in to_cancel:
3914 loop.run_until_complete(
3915 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
3918 # `concurrent.futures.Future` objects cannot be cancelled once they
3919 # are already running. There might be some when the `shutdown()` happened.
3920 # Silence their logger's spew about the event loop being closed.
3921 cf_logger = logging.getLogger("concurrent.futures")
3922 cf_logger.setLevel(logging.CRITICAL)
3926 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
3927 """Replace `regex` with `replacement` twice on `original`.
3929 This is used by string normalization to perform replaces on
3930 overlapping matches.
3932 return regex.sub(replacement, regex.sub(replacement, original))
3935 def re_compile_maybe_verbose(regex: str) -> Pattern[str]:
3936 """Compile a regular expression string in `regex`.
3938 If it contains newlines, use verbose mode.
3941 regex = "(?x)" + regex
3942 compiled: Pattern[str] = re.compile(regex)
3946 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
3947 """Like `reversed(enumerate(sequence))` if that were possible."""
3948 index = len(sequence) - 1
3949 for element in reversed(sequence):
3950 yield (index, element)
3954 def enumerate_with_length(
3955 line: Line, reversed: bool = False
3956 ) -> Iterator[Tuple[Index, Leaf, int]]:
3957 """Return an enumeration of leaves with their length.
3959 Stops prematurely on multiline strings and standalone comments.
3962 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
3963 enumerate_reversed if reversed else enumerate,
3965 for index, leaf in op(line.leaves):
3966 length = len(leaf.prefix) + len(leaf.value)
3967 if "\n" in leaf.value:
3968 return # Multiline strings, we can't continue.
3970 for comment in line.comments_after(leaf):
3971 length += len(comment.value)
3973 yield index, leaf, length
3976 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
3977 """Return True if `line` is no longer than `line_length`.
3979 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
3982 line_str = str(line).strip("\n")
3984 len(line_str) <= line_length
3985 and "\n" not in line_str # multiline strings
3986 and not line.contains_standalone_comments()
3990 def can_be_split(line: Line) -> bool:
3991 """Return False if the line cannot be split *for sure*.
3993 This is not an exhaustive search but a cheap heuristic that we can use to
3994 avoid some unfortunate formattings (mostly around wrapping unsplittable code
3995 in unnecessary parentheses).
3997 leaves = line.leaves
4001 if leaves[0].type == token.STRING and leaves[1].type == token.DOT:
4005 for leaf in leaves[-2::-1]:
4006 if leaf.type in OPENING_BRACKETS:
4007 if next.type not in CLOSING_BRACKETS:
4011 elif leaf.type == token.DOT:
4013 elif leaf.type == token.NAME:
4014 if not (next.type == token.DOT or next.type in OPENING_BRACKETS):
4017 elif leaf.type not in CLOSING_BRACKETS:
4020 if dot_count > 1 and call_count > 1:
4026 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
4027 """Does `line` have a shape safe to reformat without optional parens around it?
4029 Returns True for only a subset of potentially nice looking formattings but
4030 the point is to not return false positives that end up producing lines that
4033 bt = line.bracket_tracker
4034 if not bt.delimiters:
4035 # Without delimiters the optional parentheses are useless.
4038 max_priority = bt.max_delimiter_priority()
4039 if bt.delimiter_count_with_priority(max_priority) > 1:
4040 # With more than one delimiter of a kind the optional parentheses read better.
4043 if max_priority == DOT_PRIORITY:
4044 # A single stranded method call doesn't require optional parentheses.
4047 assert len(line.leaves) >= 2, "Stranded delimiter"
4049 first = line.leaves[0]
4050 second = line.leaves[1]
4051 penultimate = line.leaves[-2]
4052 last = line.leaves[-1]
4054 # With a single delimiter, omit if the expression starts or ends with
4056 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
4058 length = 4 * line.depth
4059 for _index, leaf, leaf_length in enumerate_with_length(line):
4060 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
4063 length += leaf_length
4064 if length > line_length:
4067 if leaf.type in OPENING_BRACKETS:
4068 # There are brackets we can further split on.
4072 # checked the entire string and line length wasn't exceeded
4073 if len(line.leaves) == _index + 1:
4076 # Note: we are not returning False here because a line might have *both*
4077 # a leading opening bracket and a trailing closing bracket. If the
4078 # opening bracket doesn't match our rule, maybe the closing will.
4081 last.type == token.RPAR
4082 or last.type == token.RBRACE
4084 # don't use indexing for omitting optional parentheses;
4086 last.type == token.RSQB
4088 and last.parent.type != syms.trailer
4091 if penultimate.type in OPENING_BRACKETS:
4092 # Empty brackets don't help.
4095 if is_multiline_string(first):
4096 # Additional wrapping of a multiline string in this situation is
4100 length = 4 * line.depth
4101 seen_other_brackets = False
4102 for _index, leaf, leaf_length in enumerate_with_length(line):
4103 length += leaf_length
4104 if leaf is last.opening_bracket:
4105 if seen_other_brackets or length <= line_length:
4108 elif leaf.type in OPENING_BRACKETS:
4109 # There are brackets we can further split on.
4110 seen_other_brackets = True
4115 def get_cache_file(mode: Mode) -> Path:
4116 return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle"
4119 def read_cache(mode: Mode) -> Cache:
4120 """Read the cache if it exists and is well formed.
4122 If it is not well formed, the call to write_cache later should resolve the issue.
4124 cache_file = get_cache_file(mode)
4125 if not cache_file.exists():
4128 with cache_file.open("rb") as fobj:
4130 cache: Cache = pickle.load(fobj)
4131 except (pickle.UnpicklingError, ValueError):
4137 def get_cache_info(path: Path) -> CacheInfo:
4138 """Return the information used to check if a file is already formatted or not."""
4140 return stat.st_mtime, stat.st_size
4143 def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]:
4144 """Split an iterable of paths in `sources` into two sets.
4146 The first contains paths of files that modified on disk or are not in the
4147 cache. The other contains paths to non-modified files.
4149 todo, done = set(), set()
4152 if cache.get(src) != get_cache_info(src):
4159 def write_cache(cache: Cache, sources: Iterable[Path], mode: Mode) -> None:
4160 """Update the cache file."""
4161 cache_file = get_cache_file(mode)
4163 CACHE_DIR.mkdir(parents=True, exist_ok=True)
4164 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
4165 with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
4166 pickle.dump(new_cache, f, protocol=4)
4167 os.replace(f.name, cache_file)
4172 def patch_click() -> None:
4173 """Make Click not crash.
4175 On certain misconfigured environments, Python 3 selects the ASCII encoding as the
4176 default which restricts paths that it can access during the lifetime of the
4177 application. Click refuses to work in this scenario by raising a RuntimeError.
4179 In case of Black the likelihood that non-ASCII characters are going to be used in
4180 file paths is minimal since it's Python source code. Moreover, this crash was
4181 spurious on Python 3.7 thanks to PEP 538 and PEP 540.
4184 from click import core
4185 from click import _unicodefun # type: ignore
4186 except ModuleNotFoundError:
4189 for module in (core, _unicodefun):
4190 if hasattr(module, "_verify_python3_env"):
4191 module._verify_python3_env = lambda: None
4194 def patched_main() -> None:
4200 if __name__ == "__main__":