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 def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool:
213 return all(feature in VERSION_TO_FEATURES[version] for version in target_versions)
216 def find_pyproject_toml(path_search_start: str) -> Optional[str]:
217 """Find the absolute filepath to a pyproject.toml if it exists"""
218 path_project_root = find_project_root(path_search_start)
219 path_pyproject_toml = path_project_root / "pyproject.toml"
220 return str(path_pyproject_toml) if path_pyproject_toml.is_file() else None
223 def parse_pyproject_toml(path_config: str) -> Dict[str, Any]:
224 """Parse a pyproject toml file, pulling out relevant parts for Black
226 If parsing fails, will raise a toml.TomlDecodeError
228 pyproject_toml = toml.load(path_config)
229 config = pyproject_toml.get("tool", {}).get("black", {})
230 return {k.replace("--", "").replace("-", "_"): v for k, v in config.items()}
233 def read_pyproject_toml(
234 ctx: click.Context, param: click.Parameter, value: Union[str, int, bool, None]
236 """Inject Black configuration from "pyproject.toml" into defaults in `ctx`.
238 Returns the path to a successfully found and read configuration file, None
241 assert not isinstance(value, (int, bool)), "Invalid parameter type passed"
243 value = find_pyproject_toml(ctx.params.get("src", ()))
248 config = parse_pyproject_toml(value)
249 except (toml.TomlDecodeError, OSError) as e:
250 raise click.FileError(
251 filename=value, hint=f"Error reading configuration file: {e}"
257 if ctx.default_map is None:
259 ctx.default_map.update(config) # type: ignore # bad types in .pyi
263 def target_version_option_callback(
264 c: click.Context, p: Union[click.Option, click.Parameter], v: Tuple[str, ...]
265 ) -> List[TargetVersion]:
266 """Compute the target versions from a --target-version flag.
268 This is its own function because mypy couldn't infer the type correctly
269 when it was a lambda, causing mypyc trouble.
271 return [TargetVersion[val.upper()] for val in v]
274 @click.command(context_settings=dict(help_option_names=["-h", "--help"]))
275 @click.option("-c", "--code", type=str, help="Format the code passed in as a string.")
280 default=DEFAULT_LINE_LENGTH,
281 help="How many characters per line to allow.",
287 type=click.Choice([v.name.lower() for v in TargetVersion]),
288 callback=target_version_option_callback,
291 "Python versions that should be supported by Black's output. [default: "
292 "per-file auto-detection]"
299 "Allow using Python 3.6-only syntax on all input files. This will put "
300 "trailing commas in function signatures and calls also after *args and "
301 "**kwargs. Deprecated; use --target-version instead. "
302 "[default: per-file auto-detection]"
309 "Format all input files like typing stubs regardless of file extension "
310 "(useful when piping source on standard input)."
315 "--skip-string-normalization",
317 help="Don't normalize string quotes or prefixes.",
323 "Don't write the files back, just return the status. Return code 0 "
324 "means nothing would change. Return code 1 means some files would be "
325 "reformatted. Return code 123 means there was an internal error."
331 help="Don't write the files back, just output a diff for each file on stdout.",
336 help="If --fast given, skip temporary sanity checks. [default: --safe]",
341 default=DEFAULT_INCLUDES,
343 "A regular expression that matches files and directories that should be "
344 "included on recursive searches. An empty value means all files are "
345 "included regardless of the name. Use forward slashes for directories on "
346 "all platforms (Windows, too). Exclusions are calculated first, inclusions "
354 default=DEFAULT_EXCLUDES,
356 "A regular expression that matches files and directories that should be "
357 "excluded on recursive searches. An empty value means no paths are excluded. "
358 "Use forward slashes for directories on all platforms (Windows, too). "
359 "Exclusions are calculated first, inclusions later."
368 "Don't emit non-error messages to stderr. Errors are still emitted; "
369 "silence those with 2>/dev/null."
377 "Also emit messages to stderr about files that were not changed or were "
378 "ignored due to --exclude=."
381 @click.version_option(version=__version__)
386 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
393 exists=False, file_okay=True, dir_okay=False, readable=True, allow_dash=False
396 callback=read_pyproject_toml,
397 help="Read configuration from PATH.",
404 target_version: List[TargetVersion],
410 skip_string_normalization: bool,
415 src: Tuple[str, ...],
416 config: Optional[str],
418 """The uncompromising code formatter."""
419 write_back = WriteBack.from_configuration(check=check, diff=diff)
422 err(f"Cannot use both --target-version and --py36")
425 versions = set(target_version)
428 "--py36 is deprecated and will be removed in a future version. "
429 "Use --target-version py36 instead."
431 versions = PY36_VERSIONS
433 # We'll autodetect later.
436 target_versions=versions,
437 line_length=line_length,
439 string_normalization=not skip_string_normalization,
441 if config and verbose:
442 out(f"Using configuration from {config}.", bold=False, fg="blue")
444 print(format_str(code, mode=mode))
447 include_regex = re_compile_maybe_verbose(include)
449 err(f"Invalid regular expression for include given: {include!r}")
452 exclude_regex = re_compile_maybe_verbose(exclude)
454 err(f"Invalid regular expression for exclude given: {exclude!r}")
456 report = Report(check=check, diff=diff, quiet=quiet, verbose=verbose)
457 root = find_project_root(src)
458 sources: Set[Path] = set()
459 path_empty(src, quiet, verbose, ctx)
464 gen_python_files_in_dir(
465 p, root, include_regex, exclude_regex, report, get_gitignore(root)
468 elif p.is_file() or s == "-":
469 # if a file was explicitly given, we don't care about its extension
472 err(f"invalid path: {s}")
473 if len(sources) == 0:
474 if verbose or not quiet:
475 out("No Python files are present to be formatted. Nothing to do 😴")
478 if len(sources) == 1:
482 write_back=write_back,
488 sources=sources, fast=fast, write_back=write_back, mode=mode, report=report
491 if verbose or not quiet:
492 out("Oh no! 💥 💔 💥" if report.return_code else "All done! ✨ 🍰 ✨")
493 click.secho(str(report), err=True)
494 ctx.exit(report.return_code)
498 src: Tuple[str, ...], quiet: bool, verbose: bool, ctx: click.Context
501 Exit if there is no `src` provided for formatting
504 if verbose or not quiet:
505 out("No Path provided. Nothing to do 😴")
510 src: Path, fast: bool, write_back: WriteBack, mode: FileMode, report: "Report"
512 """Reformat a single file under `src` without spawning child processes.
514 `fast`, `write_back`, and `mode` options are passed to
515 :func:`format_file_in_place` or :func:`format_stdin_to_stdout`.
519 if not src.is_file() and str(src) == "-":
520 if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode):
521 changed = Changed.YES
524 if write_back != WriteBack.DIFF:
525 cache = read_cache(mode)
526 res_src = src.resolve()
527 if res_src in cache and cache[res_src] == get_cache_info(res_src):
528 changed = Changed.CACHED
529 if changed is not Changed.CACHED and format_file_in_place(
530 src, fast=fast, write_back=write_back, mode=mode
532 changed = Changed.YES
533 if (write_back is WriteBack.YES and changed is not Changed.CACHED) or (
534 write_back is WriteBack.CHECK and changed is Changed.NO
536 write_cache(cache, [src], mode)
537 report.done(src, changed)
538 except Exception as exc:
539 report.failed(src, str(exc))
545 write_back: WriteBack,
549 """Reformat multiple files using a ProcessPoolExecutor."""
550 loop = asyncio.get_event_loop()
551 worker_count = os.cpu_count()
552 if sys.platform == "win32":
553 # Work around https://bugs.python.org/issue26903
554 worker_count = min(worker_count, 61)
555 executor = ProcessPoolExecutor(max_workers=worker_count)
557 loop.run_until_complete(
561 write_back=write_back,
573 async def schedule_formatting(
576 write_back: WriteBack,
579 loop: asyncio.AbstractEventLoop,
582 """Run formatting of `sources` in parallel using the provided `executor`.
584 (Use ProcessPoolExecutors for actual parallelism.)
586 `write_back`, `fast`, and `mode` options are passed to
587 :func:`format_file_in_place`.
590 if write_back != WriteBack.DIFF:
591 cache = read_cache(mode)
592 sources, cached = filter_cached(cache, sources)
593 for src in sorted(cached):
594 report.done(src, Changed.CACHED)
599 sources_to_cache = []
601 if write_back == WriteBack.DIFF:
602 # For diff output, we need locks to ensure we don't interleave output
603 # from different processes.
605 lock = manager.Lock()
607 asyncio.ensure_future(
608 loop.run_in_executor(
609 executor, format_file_in_place, src, fast, mode, write_back, lock
612 for src in sorted(sources)
614 pending: Iterable["asyncio.Future[bool]"] = tasks.keys()
616 loop.add_signal_handler(signal.SIGINT, cancel, pending)
617 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
618 except NotImplementedError:
619 # There are no good alternatives for these on Windows.
622 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
624 src = tasks.pop(task)
626 cancelled.append(task)
627 elif task.exception():
628 report.failed(src, str(task.exception()))
630 changed = Changed.YES if task.result() else Changed.NO
631 # If the file was written back or was successfully checked as
632 # well-formatted, store this information in the cache.
633 if write_back is WriteBack.YES or (
634 write_back is WriteBack.CHECK and changed is Changed.NO
636 sources_to_cache.append(src)
637 report.done(src, changed)
639 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
641 write_cache(cache, sources_to_cache, mode)
644 def format_file_in_place(
648 write_back: WriteBack = WriteBack.NO,
649 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
651 """Format file under `src` path. Return True if changed.
653 If `write_back` is DIFF, write a diff to stdout. If it is YES, write reformatted
655 `mode` and `fast` options are passed to :func:`format_file_contents`.
657 if src.suffix == ".pyi":
658 mode = replace(mode, is_pyi=True)
660 then = datetime.utcfromtimestamp(src.stat().st_mtime)
661 with open(src, "rb") as buf:
662 src_contents, encoding, newline = decode_bytes(buf.read())
664 dst_contents = format_file_contents(src_contents, fast=fast, mode=mode)
665 except NothingChanged:
668 if write_back == WriteBack.YES:
669 with open(src, "w", encoding=encoding, newline=newline) as f:
670 f.write(dst_contents)
671 elif write_back == WriteBack.DIFF:
672 now = datetime.utcnow()
673 src_name = f"{src}\t{then} +0000"
674 dst_name = f"{src}\t{now} +0000"
675 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
677 with lock or nullcontext():
678 f = io.TextIOWrapper(
684 f.write(diff_contents)
690 def format_stdin_to_stdout(
691 fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: FileMode
693 """Format file on stdin. Return True if changed.
695 If `write_back` is YES, write reformatted code back to stdout. If it is DIFF,
696 write a diff to stdout. The `mode` argument is passed to
697 :func:`format_file_contents`.
699 then = datetime.utcnow()
700 src, encoding, newline = decode_bytes(sys.stdin.buffer.read())
703 dst = format_file_contents(src, fast=fast, mode=mode)
706 except NothingChanged:
710 f = io.TextIOWrapper(
711 sys.stdout.buffer, encoding=encoding, newline=newline, write_through=True
713 if write_back == WriteBack.YES:
715 elif write_back == WriteBack.DIFF:
716 now = datetime.utcnow()
717 src_name = f"STDIN\t{then} +0000"
718 dst_name = f"STDOUT\t{now} +0000"
719 f.write(diff(src, dst, src_name, dst_name))
723 def format_file_contents(
724 src_contents: str, *, fast: bool, mode: FileMode
726 """Reformat contents a file and return new contents.
728 If `fast` is False, additionally confirm that the reformatted code is
729 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
730 `mode` is passed to :func:`format_str`.
732 if src_contents.strip() == "":
735 dst_contents = format_str(src_contents, mode=mode)
736 if src_contents == dst_contents:
740 assert_equivalent(src_contents, dst_contents)
741 assert_stable(src_contents, dst_contents, mode=mode)
745 def format_str(src_contents: str, *, mode: FileMode) -> FileContent:
746 """Reformat a string and return new contents.
748 `mode` determines formatting options, such as how many characters per line are
752 >>> print(black.format_str("def f(arg:str='')->None:...", mode=FileMode()))
753 def f(arg: str = "") -> None:
756 A more complex example:
758 ... black.format_str(
759 ... "def f(arg:str='')->None: hey",
760 ... mode=black.FileMode(
761 ... target_versions={black.TargetVersion.PY36},
763 ... string_normalization=False,
774 src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
776 future_imports = get_future_imports(src_node)
777 if mode.target_versions:
778 versions = mode.target_versions
780 versions = detect_target_versions(src_node)
781 normalize_fmt_off(src_node)
782 lines = LineGenerator(
783 remove_u_prefix="unicode_literals" in future_imports
784 or supports_feature(versions, Feature.UNICODE_LITERALS),
786 normalize_strings=mode.string_normalization,
788 elt = EmptyLineTracker(is_pyi=mode.is_pyi)
791 split_line_features = {
793 for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF}
794 if supports_feature(versions, feature)
796 for current_line in lines.visit(src_node):
797 dst_contents.append(str(empty_line) * after)
798 before, after = elt.maybe_empty_lines(current_line)
799 dst_contents.append(str(empty_line) * before)
800 for line in split_line(
801 current_line, line_length=mode.line_length, features=split_line_features
803 dst_contents.append(str(line))
804 return "".join(dst_contents)
807 def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]:
808 """Return a tuple of (decoded_contents, encoding, newline).
810 `newline` is either CRLF or LF but `decoded_contents` is decoded with
811 universal newlines (i.e. only contains LF).
813 srcbuf = io.BytesIO(src)
814 encoding, lines = tokenize.detect_encoding(srcbuf.readline)
816 return "", encoding, "\n"
818 newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n"
820 with io.TextIOWrapper(srcbuf, encoding) as tiow:
821 return tiow.read(), encoding, newline
824 def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
825 if not target_versions:
826 # No target_version specified, so try all grammars.
829 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords,
831 pygram.python_grammar_no_print_statement_no_exec_statement,
832 # Python 2.7 with future print_function import
833 pygram.python_grammar_no_print_statement,
835 pygram.python_grammar,
838 if all(version.is_python2() for version in target_versions):
839 # Python 2-only code, so try Python 2 grammars.
841 # Python 2.7 with future print_function import
842 pygram.python_grammar_no_print_statement,
844 pygram.python_grammar,
847 # Python 3-compatible code, so only try Python 3 grammar.
849 # If we have to parse both, try to parse async as a keyword first
850 if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
853 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords
855 if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
857 grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
858 # At least one of the above branches must have been taken, because every Python
859 # version has exactly one of the two 'ASYNC_*' flags
863 def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
864 """Given a string with source, return the lib2to3 Node."""
865 if src_txt[-1:] != "\n":
868 for grammar in get_grammars(set(target_versions)):
869 drv = driver.Driver(grammar, pytree.convert)
871 result = drv.parse_string(src_txt, True)
874 except ParseError as pe:
875 lineno, column = pe.context[1]
876 lines = src_txt.splitlines()
878 faulty_line = lines[lineno - 1]
880 faulty_line = "<line number missing in source>"
881 exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
885 if isinstance(result, Leaf):
886 result = Node(syms.file_input, [result])
890 def lib2to3_unparse(node: Node) -> str:
891 """Given a lib2to3 node, return its string representation."""
899 class Visitor(Generic[T]):
900 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
902 def visit(self, node: LN) -> Iterator[T]:
903 """Main method to visit `node` and its children.
905 It tries to find a `visit_*()` method for the given `node.type`, like
906 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
907 If no dedicated `visit_*()` method is found, chooses `visit_default()`
910 Then yields objects of type `T` from the selected visitor.
913 name = token.tok_name[node.type]
915 name = str(type_repr(node.type))
916 # We explicitly branch on whether a visitor exists (instead of
917 # using self.visit_default as the default arg to getattr) in order
918 # to save needing to create a bound method object and so mypyc can
919 # generate a native call to visit_default.
920 visitf = getattr(self, f"visit_{name}", None)
922 yield from visitf(node)
924 yield from self.visit_default(node)
926 def visit_default(self, node: LN) -> Iterator[T]:
927 """Default `visit_*()` implementation. Recurses to children of `node`."""
928 if isinstance(node, Node):
929 for child in node.children:
930 yield from self.visit(child)
934 class DebugVisitor(Visitor[T]):
937 def visit_default(self, node: LN) -> Iterator[T]:
938 indent = " " * (2 * self.tree_depth)
939 if isinstance(node, Node):
940 _type = type_repr(node.type)
941 out(f"{indent}{_type}", fg="yellow")
943 for child in node.children:
944 yield from self.visit(child)
947 out(f"{indent}/{_type}", fg="yellow", bold=False)
949 _type = token.tok_name.get(node.type, str(node.type))
950 out(f"{indent}{_type}", fg="blue", nl=False)
952 # We don't have to handle prefixes for `Node` objects since
953 # that delegates to the first child anyway.
954 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
955 out(f" {node.value!r}", fg="blue", bold=False)
958 def show(cls, code: Union[str, Leaf, Node]) -> None:
959 """Pretty-print the lib2to3 AST of a given string of `code`.
961 Convenience method for debugging.
963 v: DebugVisitor[None] = DebugVisitor()
964 if isinstance(code, str):
965 code = lib2to3_parse(code)
969 WHITESPACE: Final = {token.DEDENT, token.INDENT, token.NEWLINE}
980 STANDALONE_COMMENT: Final = 153
981 token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT"
982 LOGIC_OPERATORS: Final = {"and", "or"}
983 COMPARATORS: Final = {
991 MATH_OPERATORS: Final = {
1007 STARS: Final = {token.STAR, token.DOUBLESTAR}
1008 VARARGS_SPECIALS: Final = STARS | {token.SLASH}
1009 VARARGS_PARENTS: Final = {
1011 syms.argument, # double star in arglist
1012 syms.trailer, # single argument to call
1014 syms.varargslist, # lambdas
1016 UNPACKING_PARENTS: Final = {
1017 syms.atom, # single element of a list or set literal
1021 syms.testlist_star_expr,
1023 TEST_DESCENDANTS: Final = {
1040 ASSIGNMENTS: Final = {
1056 COMPREHENSION_PRIORITY: Final = 20
1057 COMMA_PRIORITY: Final = 18
1058 TERNARY_PRIORITY: Final = 16
1059 LOGIC_PRIORITY: Final = 14
1060 STRING_PRIORITY: Final = 12
1061 COMPARATOR_PRIORITY: Final = 10
1062 MATH_PRIORITIES: Final = {
1064 token.CIRCUMFLEX: 8,
1067 token.RIGHTSHIFT: 6,
1072 token.DOUBLESLASH: 4,
1076 token.DOUBLESTAR: 2,
1078 DOT_PRIORITY: Final = 1
1082 class BracketTracker:
1083 """Keeps track of brackets on a line."""
1086 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = field(default_factory=dict)
1087 delimiters: Dict[LeafID, Priority] = field(default_factory=dict)
1088 previous: Optional[Leaf] = None
1089 _for_loop_depths: List[int] = field(default_factory=list)
1090 _lambda_argument_depths: List[int] = field(default_factory=list)
1092 def mark(self, leaf: Leaf) -> None:
1093 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
1095 All leaves receive an int `bracket_depth` field that stores how deep
1096 within brackets a given leaf is. 0 means there are no enclosing brackets
1097 that started on this line.
1099 If a leaf is itself a closing bracket, it receives an `opening_bracket`
1100 field that it forms a pair with. This is a one-directional link to
1101 avoid reference cycles.
1103 If a leaf is a delimiter (a token on which Black can split the line if
1104 needed) and it's on depth 0, its `id()` is stored in the tracker's
1107 if leaf.type == token.COMMENT:
1110 self.maybe_decrement_after_for_loop_variable(leaf)
1111 self.maybe_decrement_after_lambda_arguments(leaf)
1112 if leaf.type in CLOSING_BRACKETS:
1114 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
1115 leaf.opening_bracket = opening_bracket
1116 leaf.bracket_depth = self.depth
1118 delim = is_split_before_delimiter(leaf, self.previous)
1119 if delim and self.previous is not None:
1120 self.delimiters[id(self.previous)] = delim
1122 delim = is_split_after_delimiter(leaf, self.previous)
1124 self.delimiters[id(leaf)] = delim
1125 if leaf.type in OPENING_BRACKETS:
1126 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
1128 self.previous = leaf
1129 self.maybe_increment_lambda_arguments(leaf)
1130 self.maybe_increment_for_loop_variable(leaf)
1132 def any_open_brackets(self) -> bool:
1133 """Return True if there is an yet unmatched open bracket on the line."""
1134 return bool(self.bracket_match)
1136 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> Priority:
1137 """Return the highest priority of a delimiter found on the line.
1139 Values are consistent with what `is_split_*_delimiter()` return.
1140 Raises ValueError on no delimiters.
1142 return max(v for k, v in self.delimiters.items() if k not in exclude)
1144 def delimiter_count_with_priority(self, priority: Priority = 0) -> int:
1145 """Return the number of delimiters with the given `priority`.
1147 If no `priority` is passed, defaults to max priority on the line.
1149 if not self.delimiters:
1152 priority = priority or self.max_delimiter_priority()
1153 return sum(1 for p in self.delimiters.values() if p == priority)
1155 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1156 """In a for loop, or comprehension, the variables are often unpacks.
1158 To avoid splitting on the comma in this situation, increase the depth of
1159 tokens between `for` and `in`.
1161 if leaf.type == token.NAME and leaf.value == "for":
1163 self._for_loop_depths.append(self.depth)
1168 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
1169 """See `maybe_increment_for_loop_variable` above for explanation."""
1171 self._for_loop_depths
1172 and self._for_loop_depths[-1] == self.depth
1173 and leaf.type == token.NAME
1174 and leaf.value == "in"
1177 self._for_loop_depths.pop()
1182 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
1183 """In a lambda expression, there might be more than one argument.
1185 To avoid splitting on the comma in this situation, increase the depth of
1186 tokens between `lambda` and `:`.
1188 if leaf.type == token.NAME and leaf.value == "lambda":
1190 self._lambda_argument_depths.append(self.depth)
1195 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
1196 """See `maybe_increment_lambda_arguments` above for explanation."""
1198 self._lambda_argument_depths
1199 and self._lambda_argument_depths[-1] == self.depth
1200 and leaf.type == token.COLON
1203 self._lambda_argument_depths.pop()
1208 def get_open_lsqb(self) -> Optional[Leaf]:
1209 """Return the most recent opening square bracket (if any)."""
1210 return self.bracket_match.get((self.depth - 1, token.RSQB))
1215 """Holds leaves and comments. Can be printed with `str(line)`."""
1218 leaves: List[Leaf] = field(default_factory=list)
1219 # keys ordered like `leaves`
1220 comments: Dict[LeafID, List[Leaf]] = field(default_factory=dict)
1221 bracket_tracker: BracketTracker = field(default_factory=BracketTracker)
1222 inside_brackets: bool = False
1223 should_explode: bool = False
1225 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
1226 """Add a new `leaf` to the end of the line.
1228 Unless `preformatted` is True, the `leaf` will receive a new consistent
1229 whitespace prefix and metadata applied by :class:`BracketTracker`.
1230 Trailing commas are maybe removed, unpacked for loop variables are
1231 demoted from being delimiters.
1233 Inline comments are put aside.
1235 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
1239 if token.COLON == leaf.type and self.is_class_paren_empty:
1240 del self.leaves[-2:]
1241 if self.leaves and not preformatted:
1242 # Note: at this point leaf.prefix should be empty except for
1243 # imports, for which we only preserve newlines.
1244 leaf.prefix += whitespace(
1245 leaf, complex_subscript=self.is_complex_subscript(leaf)
1247 if self.inside_brackets or not preformatted:
1248 self.bracket_tracker.mark(leaf)
1249 self.maybe_remove_trailing_comma(leaf)
1250 if not self.append_comment(leaf):
1251 self.leaves.append(leaf)
1253 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
1254 """Like :func:`append()` but disallow invalid standalone comment structure.
1256 Raises ValueError when any `leaf` is appended after a standalone comment
1257 or when a standalone comment is not the first leaf on the line.
1259 if self.bracket_tracker.depth == 0:
1261 raise ValueError("cannot append to standalone comments")
1263 if self.leaves and leaf.type == STANDALONE_COMMENT:
1265 "cannot append standalone comments to a populated line"
1268 self.append(leaf, preformatted=preformatted)
1271 def is_comment(self) -> bool:
1272 """Is this line a standalone comment?"""
1273 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
1276 def is_decorator(self) -> bool:
1277 """Is this line a decorator?"""
1278 return bool(self) and self.leaves[0].type == token.AT
1281 def is_import(self) -> bool:
1282 """Is this an import line?"""
1283 return bool(self) and is_import(self.leaves[0])
1286 def is_class(self) -> bool:
1287 """Is this line a class definition?"""
1290 and self.leaves[0].type == token.NAME
1291 and self.leaves[0].value == "class"
1295 def is_stub_class(self) -> bool:
1296 """Is this line a class definition with a body consisting only of "..."?"""
1297 return self.is_class and self.leaves[-3:] == [
1298 Leaf(token.DOT, ".") for _ in range(3)
1302 def is_collection_with_optional_trailing_comma(self) -> bool:
1303 """Is this line a collection literal with a trailing comma that's optional?
1305 Note that the trailing comma in a 1-tuple is not optional.
1307 if not self.leaves or len(self.leaves) < 4:
1310 # Look for and address a trailing colon.
1311 if self.leaves[-1].type == token.COLON:
1312 closer = self.leaves[-2]
1315 closer = self.leaves[-1]
1317 if closer.type not in CLOSING_BRACKETS or self.inside_brackets:
1320 if closer.type == token.RPAR:
1321 # Tuples require an extra check, because if there's only
1322 # one element in the tuple removing the comma unmakes the
1325 # We also check for parens before looking for the trailing
1326 # comma because in some cases (eg assigning a dict
1327 # literal) the literal gets wrapped in temporary parens
1328 # during parsing. This case is covered by the
1329 # collections.py test data.
1330 opener = closer.opening_bracket
1331 for _open_index, leaf in enumerate(self.leaves):
1336 # Couldn't find the matching opening paren, play it safe.
1340 comma_depth = self.leaves[close_index - 1].bracket_depth
1341 for leaf in self.leaves[_open_index + 1 : close_index]:
1342 if leaf.bracket_depth == comma_depth and leaf.type == token.COMMA:
1345 # We haven't looked yet for the trailing comma because
1346 # we might also have caught noop parens.
1347 return self.leaves[close_index - 1].type == token.COMMA
1350 return False # it's either a one-tuple or didn't have a trailing comma
1352 if self.leaves[close_index - 1].type in CLOSING_BRACKETS:
1354 closer = self.leaves[close_index]
1355 if closer.type == token.RPAR:
1356 # TODO: this is a gut feeling. Will we ever see this?
1359 if self.leaves[close_index - 1].type != token.COMMA:
1365 def is_def(self) -> bool:
1366 """Is this a function definition? (Also returns True for async defs.)"""
1368 first_leaf = self.leaves[0]
1373 second_leaf: Optional[Leaf] = self.leaves[1]
1376 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
1377 first_leaf.type == token.ASYNC
1378 and second_leaf is not None
1379 and second_leaf.type == token.NAME
1380 and second_leaf.value == "def"
1384 def is_class_paren_empty(self) -> bool:
1385 """Is this a class with no base classes but using parentheses?
1387 Those are unnecessary and should be removed.
1391 and len(self.leaves) == 4
1393 and self.leaves[2].type == token.LPAR
1394 and self.leaves[2].value == "("
1395 and self.leaves[3].type == token.RPAR
1396 and self.leaves[3].value == ")"
1400 def is_triple_quoted_string(self) -> bool:
1401 """Is the line a triple quoted string?"""
1404 and self.leaves[0].type == token.STRING
1405 and self.leaves[0].value.startswith(('"""', "'''"))
1408 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
1409 """If so, needs to be split before emitting."""
1410 for leaf in self.leaves:
1411 if leaf.type == STANDALONE_COMMENT and leaf.bracket_depth <= depth_limit:
1416 def contains_uncollapsable_type_comments(self) -> bool:
1419 last_leaf = self.leaves[-1]
1420 ignored_ids.add(id(last_leaf))
1421 if last_leaf.type == token.COMMA or (
1422 last_leaf.type == token.RPAR and not last_leaf.value
1424 # When trailing commas or optional parens are inserted by Black for
1425 # consistency, comments after the previous last element are not moved
1426 # (they don't have to, rendering will still be correct). So we ignore
1427 # trailing commas and invisible.
1428 last_leaf = self.leaves[-2]
1429 ignored_ids.add(id(last_leaf))
1433 # A type comment is uncollapsable if it is attached to a leaf
1434 # that isn't at the end of the line (since that could cause it
1435 # to get associated to a different argument) or if there are
1436 # comments before it (since that could cause it to get hidden
1438 comment_seen = False
1439 for leaf_id, comments in self.comments.items():
1440 for comment in comments:
1441 if is_type_comment(comment):
1442 if comment_seen or (
1443 not is_type_comment(comment, " ignore")
1444 and leaf_id not in ignored_ids
1452 def contains_unsplittable_type_ignore(self) -> bool:
1456 # If a 'type: ignore' is attached to the end of a line, we
1457 # can't split the line, because we can't know which of the
1458 # subexpressions the ignore was meant to apply to.
1460 # We only want this to apply to actual physical lines from the
1461 # original source, though: we don't want the presence of a
1462 # 'type: ignore' at the end of a multiline expression to
1463 # justify pushing it all onto one line. Thus we
1464 # (unfortunately) need to check the actual source lines and
1465 # only report an unsplittable 'type: ignore' if this line was
1466 # one line in the original code.
1468 # Grab the first and last line numbers, skipping generated leaves
1469 first_line = next((l.lineno for l in self.leaves if l.lineno != 0), 0)
1470 last_line = next((l.lineno for l in reversed(self.leaves) if l.lineno != 0), 0)
1472 if first_line == last_line:
1473 # We look at the last two leaves since a comma or an
1474 # invisible paren could have been added at the end of the
1476 for node in self.leaves[-2:]:
1477 for comment in self.comments.get(id(node), []):
1478 if is_type_comment(comment, " ignore"):
1483 def contains_multiline_strings(self) -> bool:
1484 return any(is_multiline_string(leaf) for leaf in self.leaves)
1486 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1487 """Remove trailing comma if there is one and it's safe."""
1488 if not (self.leaves and self.leaves[-1].type == token.COMMA):
1491 # We remove trailing commas only in the case of importing a
1492 # single name from a module.
1496 and len(self.leaves) > 4
1497 and self.leaves[-1].type == token.COMMA
1498 and closing.type in CLOSING_BRACKETS
1499 and self.leaves[-4].type == token.NAME
1501 # regular `from foo import bar,`
1502 self.leaves[-4].value == "import"
1503 # `from foo import (bar as baz,)
1505 len(self.leaves) > 6
1506 and self.leaves[-6].value == "import"
1507 and self.leaves[-3].value == "as"
1509 # `from foo import bar as baz,`
1511 len(self.leaves) > 5
1512 and self.leaves[-5].value == "import"
1513 and self.leaves[-3].value == "as"
1516 and closing.type == token.RPAR
1520 self.remove_trailing_comma()
1523 def append_comment(self, comment: Leaf) -> bool:
1524 """Add an inline or standalone comment to the line."""
1526 comment.type == STANDALONE_COMMENT
1527 and self.bracket_tracker.any_open_brackets()
1532 if comment.type != token.COMMENT:
1536 comment.type = STANDALONE_COMMENT
1540 last_leaf = self.leaves[-1]
1542 last_leaf.type == token.RPAR
1543 and not last_leaf.value
1544 and last_leaf.parent
1545 and len(list(last_leaf.parent.leaves())) <= 3
1546 and not is_type_comment(comment)
1548 # Comments on an optional parens wrapping a single leaf should belong to
1549 # the wrapped node except if it's a type comment. Pinning the comment like
1550 # this avoids unstable formatting caused by comment migration.
1551 if len(self.leaves) < 2:
1552 comment.type = STANDALONE_COMMENT
1556 last_leaf = self.leaves[-2]
1557 self.comments.setdefault(id(last_leaf), []).append(comment)
1560 def comments_after(self, leaf: Leaf) -> List[Leaf]:
1561 """Generate comments that should appear directly after `leaf`."""
1562 return self.comments.get(id(leaf), [])
1564 def remove_trailing_comma(self) -> None:
1565 """Remove the trailing comma and moves the comments attached to it."""
1566 trailing_comma = self.leaves.pop()
1567 trailing_comma_comments = self.comments.pop(id(trailing_comma), [])
1568 self.comments.setdefault(id(self.leaves[-1]), []).extend(
1569 trailing_comma_comments
1572 def is_complex_subscript(self, leaf: Leaf) -> bool:
1573 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1574 open_lsqb = self.bracket_tracker.get_open_lsqb()
1575 if open_lsqb is None:
1578 subscript_start = open_lsqb.next_sibling
1580 if isinstance(subscript_start, Node):
1581 if subscript_start.type == syms.listmaker:
1584 if subscript_start.type == syms.subscriptlist:
1585 subscript_start = child_towards(subscript_start, leaf)
1586 return subscript_start is not None and any(
1587 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1590 def __str__(self) -> str:
1591 """Render the line."""
1595 indent = " " * self.depth
1596 leaves = iter(self.leaves)
1597 first = next(leaves)
1598 res = f"{first.prefix}{indent}{first.value}"
1601 for comment in itertools.chain.from_iterable(self.comments.values()):
1605 def __bool__(self) -> bool:
1606 """Return True if the line has leaves or comments."""
1607 return bool(self.leaves or self.comments)
1611 class EmptyLineTracker:
1612 """Provides a stateful method that returns the number of potential extra
1613 empty lines needed before and after the currently processed line.
1615 Note: this tracker works on lines that haven't been split yet. It assumes
1616 the prefix of the first leaf consists of optional newlines. Those newlines
1617 are consumed by `maybe_empty_lines()` and included in the computation.
1620 is_pyi: bool = False
1621 previous_line: Optional[Line] = None
1622 previous_after: int = 0
1623 previous_defs: List[int] = field(default_factory=list)
1625 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1626 """Return the number of extra empty lines before and after the `current_line`.
1628 This is for separating `def`, `async def` and `class` with extra empty
1629 lines (two on module-level).
1631 before, after = self._maybe_empty_lines(current_line)
1633 # Black should not insert empty lines at the beginning
1636 if self.previous_line is None
1637 else before - self.previous_after
1639 self.previous_after = after
1640 self.previous_line = current_line
1641 return before, after
1643 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1645 if current_line.depth == 0:
1646 max_allowed = 1 if self.is_pyi else 2
1647 if current_line.leaves:
1648 # Consume the first leaf's extra newlines.
1649 first_leaf = current_line.leaves[0]
1650 before = first_leaf.prefix.count("\n")
1651 before = min(before, max_allowed)
1652 first_leaf.prefix = ""
1655 depth = current_line.depth
1656 while self.previous_defs and self.previous_defs[-1] >= depth:
1657 self.previous_defs.pop()
1659 before = 0 if depth else 1
1661 before = 1 if depth else 2
1662 if current_line.is_decorator or current_line.is_def or current_line.is_class:
1663 return self._maybe_empty_lines_for_class_or_def(current_line, before)
1667 and self.previous_line.is_import
1668 and not current_line.is_import
1669 and depth == self.previous_line.depth
1671 return (before or 1), 0
1675 and self.previous_line.is_class
1676 and current_line.is_triple_quoted_string
1682 def _maybe_empty_lines_for_class_or_def(
1683 self, current_line: Line, before: int
1684 ) -> Tuple[int, int]:
1685 if not current_line.is_decorator:
1686 self.previous_defs.append(current_line.depth)
1687 if self.previous_line is None:
1688 # Don't insert empty lines before the first line in the file.
1691 if self.previous_line.is_decorator:
1694 if self.previous_line.depth < current_line.depth and (
1695 self.previous_line.is_class or self.previous_line.is_def
1700 self.previous_line.is_comment
1701 and self.previous_line.depth == current_line.depth
1707 if self.previous_line.depth > current_line.depth:
1709 elif current_line.is_class or self.previous_line.is_class:
1710 if current_line.is_stub_class and self.previous_line.is_stub_class:
1711 # No blank line between classes with an empty body
1715 elif current_line.is_def and not self.previous_line.is_def:
1716 # Blank line between a block of functions and a block of non-functions
1722 if current_line.depth and newlines:
1728 class LineGenerator(Visitor[Line]):
1729 """Generates reformatted Line objects. Empty lines are not emitted.
1731 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1732 in ways that will no longer stringify to valid Python code on the tree.
1735 is_pyi: bool = False
1736 normalize_strings: bool = True
1737 current_line: Line = field(default_factory=Line)
1738 remove_u_prefix: bool = False
1740 def line(self, indent: int = 0) -> Iterator[Line]:
1743 If the line is empty, only emit if it makes sense.
1744 If the line is too long, split it first and then generate.
1746 If any lines were generated, set up a new current_line.
1748 if not self.current_line:
1749 self.current_line.depth += indent
1750 return # Line is empty, don't emit. Creating a new one unnecessary.
1752 complete_line = self.current_line
1753 self.current_line = Line(depth=complete_line.depth + indent)
1756 def visit_default(self, node: LN) -> Iterator[Line]:
1757 """Default `visit_*()` implementation. Recurses to children of `node`."""
1758 if isinstance(node, Leaf):
1759 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1760 for comment in generate_comments(node):
1761 if any_open_brackets:
1762 # any comment within brackets is subject to splitting
1763 self.current_line.append(comment)
1764 elif comment.type == token.COMMENT:
1765 # regular trailing comment
1766 self.current_line.append(comment)
1767 yield from self.line()
1770 # regular standalone comment
1771 yield from self.line()
1773 self.current_line.append(comment)
1774 yield from self.line()
1776 normalize_prefix(node, inside_brackets=any_open_brackets)
1777 if self.normalize_strings and node.type == token.STRING:
1778 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1779 normalize_string_quotes(node)
1780 if node.type == token.NUMBER:
1781 normalize_numeric_literal(node)
1782 if node.type not in WHITESPACE:
1783 self.current_line.append(node)
1784 yield from super().visit_default(node)
1786 def visit_INDENT(self, node: Leaf) -> Iterator[Line]:
1787 """Increase indentation level, maybe yield a line."""
1788 # In blib2to3 INDENT never holds comments.
1789 yield from self.line(+1)
1790 yield from self.visit_default(node)
1792 def visit_DEDENT(self, node: Leaf) -> Iterator[Line]:
1793 """Decrease indentation level, maybe yield a line."""
1794 # The current line might still wait for trailing comments. At DEDENT time
1795 # there won't be any (they would be prefixes on the preceding NEWLINE).
1796 # Emit the line then.
1797 yield from self.line()
1799 # While DEDENT has no value, its prefix may contain standalone comments
1800 # that belong to the current indentation level. Get 'em.
1801 yield from self.visit_default(node)
1803 # Finally, emit the dedent.
1804 yield from self.line(-1)
1807 self, node: Node, keywords: Set[str], parens: Set[str]
1808 ) -> Iterator[Line]:
1809 """Visit a statement.
1811 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1812 `def`, `with`, `class`, `assert` and assignments.
1814 The relevant Python language `keywords` for a given statement will be
1815 NAME leaves within it. This methods puts those on a separate line.
1817 `parens` holds a set of string leaf values immediately after which
1818 invisible parens should be put.
1820 normalize_invisible_parens(node, parens_after=parens)
1821 for child in node.children:
1822 if child.type == token.NAME and child.value in keywords: # type: ignore
1823 yield from self.line()
1825 yield from self.visit(child)
1827 def visit_suite(self, node: Node) -> Iterator[Line]:
1828 """Visit a suite."""
1829 if self.is_pyi and is_stub_suite(node):
1830 yield from self.visit(node.children[2])
1832 yield from self.visit_default(node)
1834 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1835 """Visit a statement without nested statements."""
1836 is_suite_like = node.parent and node.parent.type in STATEMENT
1838 if self.is_pyi and is_stub_body(node):
1839 yield from self.visit_default(node)
1841 yield from self.line(+1)
1842 yield from self.visit_default(node)
1843 yield from self.line(-1)
1846 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1847 yield from self.line()
1848 yield from self.visit_default(node)
1850 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1851 """Visit `async def`, `async for`, `async with`."""
1852 yield from self.line()
1854 children = iter(node.children)
1855 for child in children:
1856 yield from self.visit(child)
1858 if child.type == token.ASYNC:
1861 internal_stmt = next(children)
1862 for child in internal_stmt.children:
1863 yield from self.visit(child)
1865 def visit_decorators(self, node: Node) -> Iterator[Line]:
1866 """Visit decorators."""
1867 for child in node.children:
1868 yield from self.line()
1869 yield from self.visit(child)
1871 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1872 """Remove a semicolon and put the other statement on a separate line."""
1873 yield from self.line()
1875 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1876 """End of file. Process outstanding comments and end with a newline."""
1877 yield from self.visit_default(leaf)
1878 yield from self.line()
1880 def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]:
1881 if not self.current_line.bracket_tracker.any_open_brackets():
1882 yield from self.line()
1883 yield from self.visit_default(leaf)
1885 def visit_factor(self, node: Node) -> Iterator[Line]:
1886 """Force parentheses between a unary op and a binary power:
1888 -2 ** 8 -> -(2 ** 8)
1890 _operator, operand = node.children
1892 operand.type == syms.power
1893 and len(operand.children) == 3
1894 and operand.children[1].type == token.DOUBLESTAR
1896 lpar = Leaf(token.LPAR, "(")
1897 rpar = Leaf(token.RPAR, ")")
1898 index = operand.remove() or 0
1899 node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
1900 yield from self.visit_default(node)
1902 def __post_init__(self) -> None:
1903 """You are in a twisty little maze of passages."""
1906 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1907 self.visit_if_stmt = partial(
1908 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1910 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1911 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1912 self.visit_try_stmt = partial(
1913 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1915 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1916 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1917 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1918 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1919 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1920 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1921 self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
1922 self.visit_del_stmt = partial(v, keywords=Ø, parens={"del"})
1923 self.visit_async_funcdef = self.visit_async_stmt
1924 self.visit_decorated = self.visit_decorators
1927 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1928 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1929 OPENING_BRACKETS = set(BRACKET.keys())
1930 CLOSING_BRACKETS = set(BRACKET.values())
1931 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1932 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1935 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901
1936 """Return whitespace prefix if needed for the given `leaf`.
1938 `complex_subscript` signals whether the given leaf is part of a subscription
1939 which has non-trivial arguments, like arithmetic expressions or function calls.
1947 if t in ALWAYS_NO_SPACE:
1950 if t == token.COMMENT:
1953 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1954 if t == token.COLON and p.type not in {
1961 prev = leaf.prev_sibling
1963 prevp = preceding_leaf(p)
1964 if not prevp or prevp.type in OPENING_BRACKETS:
1967 if t == token.COLON:
1968 if prevp.type == token.COLON:
1971 elif prevp.type != token.COMMA and not complex_subscript:
1976 if prevp.type == token.EQUAL:
1978 if prevp.parent.type in {
1986 elif prevp.parent.type == syms.typedargslist:
1987 # A bit hacky: if the equal sign has whitespace, it means we
1988 # previously found it's a typed argument. So, we're using
1992 elif prevp.type in VARARGS_SPECIALS:
1993 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1996 elif prevp.type == token.COLON:
1997 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1998 return SPACE if complex_subscript else NO
2002 and prevp.parent.type == syms.factor
2003 and prevp.type in MATH_OPERATORS
2008 prevp.type == token.RIGHTSHIFT
2010 and prevp.parent.type == syms.shift_expr
2011 and prevp.prev_sibling
2012 and prevp.prev_sibling.type == token.NAME
2013 and prevp.prev_sibling.value == "print" # type: ignore
2015 # Python 2 print chevron
2018 elif prev.type in OPENING_BRACKETS:
2021 if p.type in {syms.parameters, syms.arglist}:
2022 # untyped function signatures or calls
2023 if not prev or prev.type != token.COMMA:
2026 elif p.type == syms.varargslist:
2028 if prev and prev.type != token.COMMA:
2031 elif p.type == syms.typedargslist:
2032 # typed function signatures
2036 if t == token.EQUAL:
2037 if prev.type != syms.tname:
2040 elif prev.type == token.EQUAL:
2041 # A bit hacky: if the equal sign has whitespace, it means we
2042 # previously found it's a typed argument. So, we're using that, too.
2045 elif prev.type != token.COMMA:
2048 elif p.type == syms.tname:
2051 prevp = preceding_leaf(p)
2052 if not prevp or prevp.type != token.COMMA:
2055 elif p.type == syms.trailer:
2056 # attributes and calls
2057 if t == token.LPAR or t == token.RPAR:
2062 prevp = preceding_leaf(p)
2063 if not prevp or prevp.type != token.NUMBER:
2066 elif t == token.LSQB:
2069 elif prev.type != token.COMMA:
2072 elif p.type == syms.argument:
2074 if t == token.EQUAL:
2078 prevp = preceding_leaf(p)
2079 if not prevp or prevp.type == token.LPAR:
2082 elif prev.type in {token.EQUAL} | VARARGS_SPECIALS:
2085 elif p.type == syms.decorator:
2089 elif p.type == syms.dotted_name:
2093 prevp = preceding_leaf(p)
2094 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
2097 elif p.type == syms.classdef:
2101 if prev and prev.type == token.LPAR:
2104 elif p.type in {syms.subscript, syms.sliceop}:
2107 assert p.parent is not None, "subscripts are always parented"
2108 if p.parent.type == syms.subscriptlist:
2113 elif not complex_subscript:
2116 elif p.type == syms.atom:
2117 if prev and t == token.DOT:
2118 # dots, but not the first one.
2121 elif p.type == syms.dictsetmaker:
2123 if prev and prev.type == token.DOUBLESTAR:
2126 elif p.type in {syms.factor, syms.star_expr}:
2129 prevp = preceding_leaf(p)
2130 if not prevp or prevp.type in OPENING_BRACKETS:
2133 prevp_parent = prevp.parent
2134 assert prevp_parent is not None
2135 if prevp.type == token.COLON and prevp_parent.type in {
2141 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
2144 elif t in {token.NAME, token.NUMBER, token.STRING}:
2147 elif p.type == syms.import_from:
2149 if prev and prev.type == token.DOT:
2152 elif t == token.NAME:
2156 if prev and prev.type == token.DOT:
2159 elif p.type == syms.sliceop:
2165 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
2166 """Return the first leaf that precedes `node`, if any."""
2168 res = node.prev_sibling
2170 if isinstance(res, Leaf):
2174 return list(res.leaves())[-1]
2183 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
2184 """Return the child of `ancestor` that contains `descendant`."""
2185 node: Optional[LN] = descendant
2186 while node and node.parent != ancestor:
2191 def container_of(leaf: Leaf) -> LN:
2192 """Return `leaf` or one of its ancestors that is the topmost container of it.
2194 By "container" we mean a node where `leaf` is the very first child.
2196 same_prefix = leaf.prefix
2197 container: LN = leaf
2199 parent = container.parent
2203 if parent.children[0].prefix != same_prefix:
2206 if parent.type == syms.file_input:
2209 if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS:
2216 def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2217 """Return the priority of the `leaf` delimiter, given a line break after it.
2219 The delimiter priorities returned here are from those delimiters that would
2220 cause a line break after themselves.
2222 Higher numbers are higher priority.
2224 if leaf.type == token.COMMA:
2225 return COMMA_PRIORITY
2230 def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2231 """Return the priority of the `leaf` delimiter, given a line break before it.
2233 The delimiter priorities returned here are from those delimiters that would
2234 cause a line break before themselves.
2236 Higher numbers are higher priority.
2238 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
2239 # * and ** might also be MATH_OPERATORS but in this case they are not.
2240 # Don't treat them as a delimiter.
2244 leaf.type == token.DOT
2246 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
2247 and (previous is None or previous.type in CLOSING_BRACKETS)
2252 leaf.type in MATH_OPERATORS
2254 and leaf.parent.type not in {syms.factor, syms.star_expr}
2256 return MATH_PRIORITIES[leaf.type]
2258 if leaf.type in COMPARATORS:
2259 return COMPARATOR_PRIORITY
2262 leaf.type == token.STRING
2263 and previous is not None
2264 and previous.type == token.STRING
2266 return STRING_PRIORITY
2268 if leaf.type not in {token.NAME, token.ASYNC}:
2274 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
2275 or leaf.type == token.ASYNC
2278 not isinstance(leaf.prev_sibling, Leaf)
2279 or leaf.prev_sibling.value != "async"
2281 return COMPREHENSION_PRIORITY
2286 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
2288 return COMPREHENSION_PRIORITY
2290 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
2291 return TERNARY_PRIORITY
2293 if leaf.value == "is":
2294 return COMPARATOR_PRIORITY
2299 and leaf.parent.type in {syms.comp_op, syms.comparison}
2301 previous is not None
2302 and previous.type == token.NAME
2303 and previous.value == "not"
2306 return COMPARATOR_PRIORITY
2311 and leaf.parent.type == syms.comp_op
2313 previous is not None
2314 and previous.type == token.NAME
2315 and previous.value == "is"
2318 return COMPARATOR_PRIORITY
2320 if leaf.value in LOGIC_OPERATORS and leaf.parent:
2321 return LOGIC_PRIORITY
2326 FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"}
2327 FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"}
2330 def generate_comments(leaf: LN) -> Iterator[Leaf]:
2331 """Clean the prefix of the `leaf` and generate comments from it, if any.
2333 Comments in lib2to3 are shoved into the whitespace prefix. This happens
2334 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
2335 move because it does away with modifying the grammar to include all the
2336 possible places in which comments can be placed.
2338 The sad consequence for us though is that comments don't "belong" anywhere.
2339 This is why this function generates simple parentless Leaf objects for
2340 comments. We simply don't know what the correct parent should be.
2342 No matter though, we can live without this. We really only need to
2343 differentiate between inline and standalone comments. The latter don't
2344 share the line with any code.
2346 Inline comments are emitted as regular token.COMMENT leaves. Standalone
2347 are emitted with a fake STANDALONE_COMMENT token identifier.
2349 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
2350 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
2355 """Describes a piece of syntax that is a comment.
2357 It's not a :class:`blib2to3.pytree.Leaf` so that:
2359 * it can be cached (`Leaf` objects should not be reused more than once as
2360 they store their lineno, column, prefix, and parent information);
2361 * `newlines` and `consumed` fields are kept separate from the `value`. This
2362 simplifies handling of special marker comments like ``# fmt: off/on``.
2365 type: int # token.COMMENT or STANDALONE_COMMENT
2366 value: str # content of the comment
2367 newlines: int # how many newlines before the comment
2368 consumed: int # how many characters of the original leaf's prefix did we consume
2371 @lru_cache(maxsize=4096)
2372 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
2373 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
2374 result: List[ProtoComment] = []
2375 if not prefix or "#" not in prefix:
2381 for index, line in enumerate(prefix.split("\n")):
2382 consumed += len(line) + 1 # adding the length of the split '\n'
2383 line = line.lstrip()
2386 if not line.startswith("#"):
2387 # Escaped newlines outside of a comment are not really newlines at
2388 # all. We treat a single-line comment following an escaped newline
2389 # as a simple trailing comment.
2390 if line.endswith("\\"):
2394 if index == ignored_lines and not is_endmarker:
2395 comment_type = token.COMMENT # simple trailing comment
2397 comment_type = STANDALONE_COMMENT
2398 comment = make_comment(line)
2401 type=comment_type, value=comment, newlines=nlines, consumed=consumed
2408 def make_comment(content: str) -> str:
2409 """Return a consistently formatted comment from the given `content` string.
2411 All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single
2412 space between the hash sign and the content.
2414 If `content` didn't start with a hash sign, one is provided.
2416 content = content.rstrip()
2420 if content[0] == "#":
2421 content = content[1:]
2422 if content and content[0] not in " !:#'%":
2423 content = " " + content
2424 return "#" + content
2430 inner: bool = False,
2431 features: Collection[Feature] = (),
2432 ) -> Iterator[Line]:
2433 """Split a `line` into potentially many lines.
2435 They should fit in the allotted `line_length` but might not be able to.
2436 `inner` signifies that there were a pair of brackets somewhere around the
2437 current `line`, possibly transitively. This means we can fallback to splitting
2438 by delimiters if the LHS/RHS don't yield any results.
2440 `features` are syntactical features that may be used in the output.
2446 line_str = str(line).strip("\n")
2449 not line.contains_uncollapsable_type_comments()
2450 and not line.should_explode
2451 and not line.is_collection_with_optional_trailing_comma
2453 is_line_short_enough(line, line_length=line_length, line_str=line_str)
2454 or line.contains_unsplittable_type_ignore()
2460 split_funcs: List[SplitFunc]
2462 split_funcs = [left_hand_split]
2465 def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]:
2466 for omit in generate_trailers_to_omit(line, line_length):
2467 lines = list(right_hand_split(line, line_length, features, omit=omit))
2468 if is_line_short_enough(lines[0], line_length=line_length):
2472 # All splits failed, best effort split with no omits.
2473 # This mostly happens to multiline strings that are by definition
2474 # reported as not fitting a single line.
2475 # line_length=1 here was historically a bug that somehow became a feature.
2476 # See #762 and #781 for the full story.
2477 yield from right_hand_split(line, line_length=1, features=features)
2479 if line.inside_brackets:
2480 split_funcs = [delimiter_split, standalone_comment_split, rhs]
2483 for split_func in split_funcs:
2484 # We are accumulating lines in `result` because we might want to abort
2485 # mission and return the original line in the end, or attempt a different
2487 result: List[Line] = []
2489 for l in split_func(line, features):
2490 if str(l).strip("\n") == line_str:
2491 raise CannotSplit("Split function returned an unchanged result")
2495 l, line_length=line_length, inner=True, features=features
2509 def left_hand_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2510 """Split line into many lines, starting with the first matching bracket pair.
2512 Note: this usually looks weird, only use this for function definitions.
2513 Prefer RHS otherwise. This is why this function is not symmetrical with
2514 :func:`right_hand_split` which also handles optional parentheses.
2516 tail_leaves: List[Leaf] = []
2517 body_leaves: List[Leaf] = []
2518 head_leaves: List[Leaf] = []
2519 current_leaves = head_leaves
2520 matching_bracket: Optional[Leaf] = None
2521 for leaf in line.leaves:
2523 current_leaves is body_leaves
2524 and leaf.type in CLOSING_BRACKETS
2525 and leaf.opening_bracket is matching_bracket
2527 current_leaves = tail_leaves if body_leaves else head_leaves
2528 current_leaves.append(leaf)
2529 if current_leaves is head_leaves:
2530 if leaf.type in OPENING_BRACKETS:
2531 matching_bracket = leaf
2532 current_leaves = body_leaves
2533 if not matching_bracket:
2534 raise CannotSplit("No brackets found")
2536 head = bracket_split_build_line(head_leaves, line, matching_bracket)
2537 body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True)
2538 tail = bracket_split_build_line(tail_leaves, line, matching_bracket)
2539 bracket_split_succeeded_or_raise(head, body, tail)
2540 for result in (head, body, tail):
2545 def right_hand_split(
2548 features: Collection[Feature] = (),
2549 omit: Collection[LeafID] = (),
2550 ) -> Iterator[Line]:
2551 """Split line into many lines, starting with the last matching bracket pair.
2553 If the split was by optional parentheses, attempt splitting without them, too.
2554 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2557 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2559 tail_leaves: List[Leaf] = []
2560 body_leaves: List[Leaf] = []
2561 head_leaves: List[Leaf] = []
2562 current_leaves = tail_leaves
2563 opening_bracket: Optional[Leaf] = None
2564 closing_bracket: Optional[Leaf] = None
2565 for leaf in reversed(line.leaves):
2566 if current_leaves is body_leaves:
2567 if leaf is opening_bracket:
2568 current_leaves = head_leaves if body_leaves else tail_leaves
2569 current_leaves.append(leaf)
2570 if current_leaves is tail_leaves:
2571 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2572 opening_bracket = leaf.opening_bracket
2573 closing_bracket = leaf
2574 current_leaves = body_leaves
2575 if not (opening_bracket and closing_bracket and head_leaves):
2576 # If there is no opening or closing_bracket that means the split failed and
2577 # all content is in the tail. Otherwise, if `head_leaves` are empty, it means
2578 # the matching `opening_bracket` wasn't available on `line` anymore.
2579 raise CannotSplit("No brackets found")
2581 tail_leaves.reverse()
2582 body_leaves.reverse()
2583 head_leaves.reverse()
2584 head = bracket_split_build_line(head_leaves, line, opening_bracket)
2585 body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True)
2586 tail = bracket_split_build_line(tail_leaves, line, opening_bracket)
2587 bracket_split_succeeded_or_raise(head, body, tail)
2589 # the body shouldn't be exploded
2590 not body.should_explode
2591 # the opening bracket is an optional paren
2592 and opening_bracket.type == token.LPAR
2593 and not opening_bracket.value
2594 # the closing bracket is an optional paren
2595 and closing_bracket.type == token.RPAR
2596 and not closing_bracket.value
2597 # it's not an import (optional parens are the only thing we can split on
2598 # in this case; attempting a split without them is a waste of time)
2599 and not line.is_import
2600 # there are no standalone comments in the body
2601 and not body.contains_standalone_comments(0)
2602 # and we can actually remove the parens
2603 and can_omit_invisible_parens(body, line_length)
2605 omit = {id(closing_bracket), *omit}
2607 yield from right_hand_split(line, line_length, features=features, omit=omit)
2613 or is_line_short_enough(body, line_length=line_length)
2616 "Splitting failed, body is still too long and can't be split."
2619 elif head.contains_multiline_strings() or tail.contains_multiline_strings():
2621 "The current optional pair of parentheses is bound to fail to "
2622 "satisfy the splitting algorithm because the head or the tail "
2623 "contains multiline strings which by definition never fit one "
2627 ensure_visible(opening_bracket)
2628 ensure_visible(closing_bracket)
2629 for result in (head, body, tail):
2634 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2635 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2637 Do nothing otherwise.
2639 A left- or right-hand split is based on a pair of brackets. Content before
2640 (and including) the opening bracket is left on one line, content inside the
2641 brackets is put on a separate line, and finally content starting with and
2642 following the closing bracket is put on a separate line.
2644 Those are called `head`, `body`, and `tail`, respectively. If the split
2645 produced the same line (all content in `head`) or ended up with an empty `body`
2646 and the `tail` is just the closing bracket, then it's considered failed.
2648 tail_len = len(str(tail).strip())
2651 raise CannotSplit("Splitting brackets produced the same line")
2655 f"Splitting brackets on an empty body to save "
2656 f"{tail_len} characters is not worth it"
2660 def bracket_split_build_line(
2661 leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False
2663 """Return a new line with given `leaves` and respective comments from `original`.
2665 If `is_body` is True, the result line is one-indented inside brackets and as such
2666 has its first leaf's prefix normalized and a trailing comma added when expected.
2668 result = Line(depth=original.depth)
2670 result.inside_brackets = True
2673 # Since body is a new indent level, remove spurious leading whitespace.
2674 normalize_prefix(leaves[0], inside_brackets=True)
2675 # Ensure a trailing comma for imports and standalone function arguments, but
2676 # be careful not to add one after any comments or within type annotations.
2679 and opening_bracket.value == "("
2680 and not any(l.type == token.COMMA for l in leaves)
2683 if original.is_import or no_commas:
2684 for i in range(len(leaves) - 1, -1, -1):
2685 if leaves[i].type == STANDALONE_COMMENT:
2688 if leaves[i].type != token.COMMA:
2689 leaves.insert(i + 1, Leaf(token.COMMA, ","))
2694 result.append(leaf, preformatted=True)
2695 for comment_after in original.comments_after(leaf):
2696 result.append(comment_after, preformatted=True)
2698 result.should_explode = should_explode(result, opening_bracket)
2702 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2703 """Normalize prefix of the first leaf in every line returned by `split_func`.
2705 This is a decorator over relevant split functions.
2709 def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2710 for l in split_func(line, features):
2711 normalize_prefix(l.leaves[0], inside_brackets=True)
2714 return split_wrapper
2717 @dont_increase_indentation
2718 def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2719 """Split according to delimiters of the highest priority.
2721 If the appropriate Features are given, the split will add trailing commas
2722 also in function signatures and calls that contain `*` and `**`.
2725 last_leaf = line.leaves[-1]
2727 raise CannotSplit("Line empty")
2729 bt = line.bracket_tracker
2731 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2733 raise CannotSplit("No delimiters found")
2735 if delimiter_priority == DOT_PRIORITY:
2736 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2737 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2739 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2740 lowest_depth = sys.maxsize
2741 trailing_comma_safe = True
2743 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2744 """Append `leaf` to current line or to new line if appending impossible."""
2745 nonlocal current_line
2747 current_line.append_safe(leaf, preformatted=True)
2751 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2752 current_line.append(leaf)
2754 for leaf in line.leaves:
2755 yield from append_to_line(leaf)
2757 for comment_after in line.comments_after(leaf):
2758 yield from append_to_line(comment_after)
2760 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2761 if leaf.bracket_depth == lowest_depth:
2762 if is_vararg(leaf, within={syms.typedargslist}):
2763 trailing_comma_safe = (
2764 trailing_comma_safe and Feature.TRAILING_COMMA_IN_DEF in features
2766 elif is_vararg(leaf, within={syms.arglist, syms.argument}):
2767 trailing_comma_safe = (
2768 trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features
2771 leaf_priority = bt.delimiters.get(id(leaf))
2772 if leaf_priority == delimiter_priority:
2775 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2779 and delimiter_priority == COMMA_PRIORITY
2780 and current_line.leaves[-1].type != token.COMMA
2781 and current_line.leaves[-1].type != STANDALONE_COMMENT
2783 current_line.append(Leaf(token.COMMA, ","))
2787 @dont_increase_indentation
2788 def standalone_comment_split(
2789 line: Line, features: Collection[Feature] = ()
2790 ) -> Iterator[Line]:
2791 """Split standalone comments from the rest of the line."""
2792 if not line.contains_standalone_comments(0):
2793 raise CannotSplit("Line does not have any standalone comments")
2795 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2797 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2798 """Append `leaf` to current line or to new line if appending impossible."""
2799 nonlocal current_line
2801 current_line.append_safe(leaf, preformatted=True)
2805 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2806 current_line.append(leaf)
2808 for leaf in line.leaves:
2809 yield from append_to_line(leaf)
2811 for comment_after in line.comments_after(leaf):
2812 yield from append_to_line(comment_after)
2818 def is_import(leaf: Leaf) -> bool:
2819 """Return True if the given leaf starts an import statement."""
2826 (v == "import" and p and p.type == syms.import_name)
2827 or (v == "from" and p and p.type == syms.import_from)
2832 def is_type_comment(leaf: Leaf, suffix: str = "") -> bool:
2833 """Return True if the given leaf is a special comment.
2834 Only returns true for type comments for now."""
2837 return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix)
2840 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2841 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2844 Note: don't use backslashes for formatting or you'll lose your voting rights.
2846 if not inside_brackets:
2847 spl = leaf.prefix.split("#")
2848 if "\\" not in spl[0]:
2849 nl_count = spl[-1].count("\n")
2852 leaf.prefix = "\n" * nl_count
2858 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2859 """Make all string prefixes lowercase.
2861 If remove_u_prefix is given, also removes any u prefix from the string.
2863 Note: Mutates its argument.
2865 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2866 assert match is not None, f"failed to match string {leaf.value!r}"
2867 orig_prefix = match.group(1)
2868 new_prefix = orig_prefix.replace("F", "f").replace("B", "b").replace("U", "u")
2870 new_prefix = new_prefix.replace("u", "")
2871 leaf.value = f"{new_prefix}{match.group(2)}"
2874 def normalize_string_quotes(leaf: Leaf) -> None:
2875 """Prefer double quotes but only if it doesn't cause more escaping.
2877 Adds or removes backslashes as appropriate. Doesn't parse and fix
2878 strings nested in f-strings (yet).
2880 Note: Mutates its argument.
2882 value = leaf.value.lstrip("furbFURB")
2883 if value[:3] == '"""':
2886 elif value[:3] == "'''":
2889 elif value[0] == '"':
2895 first_quote_pos = leaf.value.find(orig_quote)
2896 if first_quote_pos == -1:
2897 return # There's an internal error
2899 prefix = leaf.value[:first_quote_pos]
2900 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2901 escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}")
2902 escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}")
2903 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2904 if "r" in prefix.casefold():
2905 if unescaped_new_quote.search(body):
2906 # There's at least one unescaped new_quote in this raw string
2907 # so converting is impossible
2910 # Do not introduce or remove backslashes in raw strings
2913 # remove unnecessary escapes
2914 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2915 if body != new_body:
2916 # Consider the string without unnecessary escapes as the original
2918 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2919 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2920 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2921 if "f" in prefix.casefold():
2922 matches = re.findall(
2924 (?:[^{]|^)\{ # start of the string or a non-{ followed by a single {
2925 ([^{].*?) # contents of the brackets except if begins with {{
2926 \}(?:[^}]|$) # A } followed by end of the string or a non-}
2933 # Do not introduce backslashes in interpolated expressions
2936 if new_quote == '"""' and new_body[-1:] == '"':
2938 new_body = new_body[:-1] + '\\"'
2939 orig_escape_count = body.count("\\")
2940 new_escape_count = new_body.count("\\")
2941 if new_escape_count > orig_escape_count:
2942 return # Do not introduce more escaping
2944 if new_escape_count == orig_escape_count and orig_quote == '"':
2945 return # Prefer double quotes
2947 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2950 def normalize_numeric_literal(leaf: Leaf) -> None:
2951 """Normalizes numeric (float, int, and complex) literals.
2953 All letters used in the representation are normalized to lowercase (except
2954 in Python 2 long literals).
2956 text = leaf.value.lower()
2957 if text.startswith(("0o", "0b")):
2958 # Leave octal and binary literals alone.
2960 elif text.startswith("0x"):
2961 # Change hex literals to upper case.
2962 before, after = text[:2], text[2:]
2963 text = f"{before}{after.upper()}"
2965 before, after = text.split("e")
2967 if after.startswith("-"):
2970 elif after.startswith("+"):
2972 before = format_float_or_int_string(before)
2973 text = f"{before}e{sign}{after}"
2974 elif text.endswith(("j", "l")):
2977 # Capitalize in "2L" because "l" looks too similar to "1".
2980 text = f"{format_float_or_int_string(number)}{suffix}"
2982 text = format_float_or_int_string(text)
2986 def format_float_or_int_string(text: str) -> str:
2987 """Formats a float string like "1.0"."""
2991 before, after = text.split(".")
2992 return f"{before or 0}.{after or 0}"
2995 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2996 """Make existing optional parentheses invisible or create new ones.
2998 `parens_after` is a set of string leaf values immediately after which parens
3001 Standardizes on visible parentheses for single-element tuples, and keeps
3002 existing visible parentheses for other tuples and generator expressions.
3004 for pc in list_comments(node.prefix, is_endmarker=False):
3005 if pc.value in FMT_OFF:
3006 # This `node` has a prefix with `# fmt: off`, don't mess with parens.
3009 for index, child in enumerate(list(node.children)):
3010 # Add parentheses around long tuple unpacking in assignments.
3013 and isinstance(child, Node)
3014 and child.type == syms.testlist_star_expr
3019 if is_walrus_assignment(child):
3022 if child.type == syms.atom:
3023 if maybe_make_parens_invisible_in_atom(child, parent=node):
3024 wrap_in_parentheses(node, child, visible=False)
3025 elif is_one_tuple(child):
3026 wrap_in_parentheses(node, child, visible=True)
3027 elif node.type == syms.import_from:
3028 # "import from" nodes store parentheses directly as part of
3030 if child.type == token.LPAR:
3031 # make parentheses invisible
3032 child.value = "" # type: ignore
3033 node.children[-1].value = "" # type: ignore
3034 elif child.type != token.STAR:
3035 # insert invisible parentheses
3036 node.insert_child(index, Leaf(token.LPAR, ""))
3037 node.append_child(Leaf(token.RPAR, ""))
3040 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
3041 wrap_in_parentheses(node, child, visible=False)
3043 check_lpar = isinstance(child, Leaf) and child.value in parens_after
3046 def normalize_fmt_off(node: Node) -> None:
3047 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
3050 try_again = convert_one_fmt_off_pair(node)
3053 def convert_one_fmt_off_pair(node: Node) -> bool:
3054 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
3056 Returns True if a pair was converted.
3058 for leaf in node.leaves():
3059 previous_consumed = 0
3060 for comment in list_comments(leaf.prefix, is_endmarker=False):
3061 if comment.value in FMT_OFF:
3062 # We only want standalone comments. If there's no previous leaf or
3063 # the previous leaf is indentation, it's a standalone comment in
3065 if comment.type != STANDALONE_COMMENT:
3066 prev = preceding_leaf(leaf)
3067 if prev and prev.type not in WHITESPACE:
3070 ignored_nodes = list(generate_ignored_nodes(leaf))
3071 if not ignored_nodes:
3074 first = ignored_nodes[0] # Can be a container node with the `leaf`.
3075 parent = first.parent
3076 prefix = first.prefix
3077 first.prefix = prefix[comment.consumed :]
3079 comment.value + "\n" + "".join(str(n) for n in ignored_nodes)
3081 if hidden_value.endswith("\n"):
3082 # That happens when one of the `ignored_nodes` ended with a NEWLINE
3083 # leaf (possibly followed by a DEDENT).
3084 hidden_value = hidden_value[:-1]
3085 first_idx: Optional[int] = None
3086 for ignored in ignored_nodes:
3087 index = ignored.remove()
3088 if first_idx is None:
3090 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
3091 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
3092 parent.insert_child(
3097 prefix=prefix[:previous_consumed] + "\n" * comment.newlines,
3102 previous_consumed = comment.consumed
3107 def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]:
3108 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
3110 Stops at the end of the block.
3112 container: Optional[LN] = container_of(leaf)
3113 while container is not None and container.type != token.ENDMARKER:
3115 for comment in list_comments(container.prefix, is_endmarker=False):
3116 if comment.value in FMT_ON:
3118 elif comment.value in FMT_OFF:
3125 container = container.next_sibling
3128 def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool:
3129 """If it's safe, make the parens in the atom `node` invisible, recursively.
3130 Additionally, remove repeated, adjacent invisible parens from the atom `node`
3131 as they are redundant.
3133 Returns whether the node should itself be wrapped in invisible parentheses.
3137 node.type != syms.atom
3138 or is_empty_tuple(node)
3139 or is_one_tuple(node)
3140 or (is_yield(node) and parent.type != syms.expr_stmt)
3141 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
3145 first = node.children[0]
3146 last = node.children[-1]
3147 if first.type == token.LPAR and last.type == token.RPAR:
3148 middle = node.children[1]
3149 # make parentheses invisible
3150 first.value = "" # type: ignore
3151 last.value = "" # type: ignore
3152 maybe_make_parens_invisible_in_atom(middle, parent=parent)
3154 if is_atom_with_invisible_parens(middle):
3155 # Strip the invisible parens from `middle` by replacing
3156 # it with the child in-between the invisible parens
3157 middle.replace(middle.children[1])
3164 def is_atom_with_invisible_parens(node: LN) -> bool:
3165 """Given a `LN`, determines whether it's an atom `node` with invisible
3166 parens. Useful in dedupe-ing and normalizing parens.
3168 if isinstance(node, Leaf) or node.type != syms.atom:
3171 first, last = node.children[0], node.children[-1]
3173 isinstance(first, Leaf)
3174 and first.type == token.LPAR
3175 and first.value == ""
3176 and isinstance(last, Leaf)
3177 and last.type == token.RPAR
3178 and last.value == ""
3182 def is_empty_tuple(node: LN) -> bool:
3183 """Return True if `node` holds an empty tuple."""
3185 node.type == syms.atom
3186 and len(node.children) == 2
3187 and node.children[0].type == token.LPAR
3188 and node.children[1].type == token.RPAR
3192 def unwrap_singleton_parenthesis(node: LN) -> Optional[LN]:
3193 """Returns `wrapped` if `node` is of the shape ( wrapped ).
3195 Parenthesis can be optional. Returns None otherwise"""
3196 if len(node.children) != 3:
3199 lpar, wrapped, rpar = node.children
3200 if not (lpar.type == token.LPAR and rpar.type == token.RPAR):
3206 def wrap_in_parentheses(parent: Node, child: LN, *, visible: bool = True) -> None:
3207 """Wrap `child` in parentheses.
3209 This replaces `child` with an atom holding the parentheses and the old
3210 child. That requires moving the prefix.
3212 If `visible` is False, the leaves will be valueless (and thus invisible).
3214 lpar = Leaf(token.LPAR, "(" if visible else "")
3215 rpar = Leaf(token.RPAR, ")" if visible else "")
3216 prefix = child.prefix
3218 index = child.remove() or 0
3219 new_child = Node(syms.atom, [lpar, child, rpar])
3220 new_child.prefix = prefix
3221 parent.insert_child(index, new_child)
3224 def is_one_tuple(node: LN) -> bool:
3225 """Return True if `node` holds a tuple with one element, with or without parens."""
3226 if node.type == syms.atom:
3227 gexp = unwrap_singleton_parenthesis(node)
3228 if gexp is None or gexp.type != syms.testlist_gexp:
3231 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
3234 node.type in IMPLICIT_TUPLE
3235 and len(node.children) == 2
3236 and node.children[1].type == token.COMMA
3240 def is_walrus_assignment(node: LN) -> bool:
3241 """Return True iff `node` is of the shape ( test := test )"""
3242 inner = unwrap_singleton_parenthesis(node)
3243 return inner is not None and inner.type == syms.namedexpr_test
3246 def is_yield(node: LN) -> bool:
3247 """Return True if `node` holds a `yield` or `yield from` expression."""
3248 if node.type == syms.yield_expr:
3251 if node.type == token.NAME and node.value == "yield": # type: ignore
3254 if node.type != syms.atom:
3257 if len(node.children) != 3:
3260 lpar, expr, rpar = node.children
3261 if lpar.type == token.LPAR and rpar.type == token.RPAR:
3262 return is_yield(expr)
3267 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
3268 """Return True if `leaf` is a star or double star in a vararg or kwarg.
3270 If `within` includes VARARGS_PARENTS, this applies to function signatures.
3271 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
3272 extended iterable unpacking (PEP 3132) and additional unpacking
3273 generalizations (PEP 448).
3275 if leaf.type not in VARARGS_SPECIALS or not leaf.parent:
3279 if p.type == syms.star_expr:
3280 # Star expressions are also used as assignment targets in extended
3281 # iterable unpacking (PEP 3132). See what its parent is instead.
3287 return p.type in within
3290 def is_multiline_string(leaf: Leaf) -> bool:
3291 """Return True if `leaf` is a multiline string that actually spans many lines."""
3292 value = leaf.value.lstrip("furbFURB")
3293 return value[:3] in {'"""', "'''"} and "\n" in value
3296 def is_stub_suite(node: Node) -> bool:
3297 """Return True if `node` is a suite with a stub body."""
3299 len(node.children) != 4
3300 or node.children[0].type != token.NEWLINE
3301 or node.children[1].type != token.INDENT
3302 or node.children[3].type != token.DEDENT
3306 return is_stub_body(node.children[2])
3309 def is_stub_body(node: LN) -> bool:
3310 """Return True if `node` is a simple statement containing an ellipsis."""
3311 if not isinstance(node, Node) or node.type != syms.simple_stmt:
3314 if len(node.children) != 2:
3317 child = node.children[0]
3319 child.type == syms.atom
3320 and len(child.children) == 3
3321 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
3325 def max_delimiter_priority_in_atom(node: LN) -> Priority:
3326 """Return maximum delimiter priority inside `node`.
3328 This is specific to atoms with contents contained in a pair of parentheses.
3329 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
3331 if node.type != syms.atom:
3334 first = node.children[0]
3335 last = node.children[-1]
3336 if not (first.type == token.LPAR and last.type == token.RPAR):
3339 bt = BracketTracker()
3340 for c in node.children[1:-1]:
3341 if isinstance(c, Leaf):
3344 for leaf in c.leaves():
3347 return bt.max_delimiter_priority()
3353 def ensure_visible(leaf: Leaf) -> None:
3354 """Make sure parentheses are visible.
3356 They could be invisible as part of some statements (see
3357 :func:`normalize_invisible_parens` and :func:`visit_import_from`).
3359 if leaf.type == token.LPAR:
3361 elif leaf.type == token.RPAR:
3365 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
3366 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
3369 opening_bracket.parent
3370 and opening_bracket.parent.type in {syms.atom, syms.import_from}
3371 and opening_bracket.value in "[{("
3376 last_leaf = line.leaves[-1]
3377 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
3378 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
3379 except (IndexError, ValueError):
3382 return max_priority == COMMA_PRIORITY
3385 def get_features_used(node: Node) -> Set[Feature]:
3386 """Return a set of (relatively) new Python features used in this file.
3388 Currently looking for:
3390 - underscores in numeric literals;
3391 - trailing commas after * or ** in function signatures and calls;
3392 - positional only arguments in function signatures and lambdas;
3394 features: Set[Feature] = set()
3395 for n in node.pre_order():
3396 if n.type == token.STRING:
3397 value_head = n.value[:2] # type: ignore
3398 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
3399 features.add(Feature.F_STRINGS)
3401 elif n.type == token.NUMBER:
3402 if "_" in n.value: # type: ignore
3403 features.add(Feature.NUMERIC_UNDERSCORES)
3405 elif n.type == token.SLASH:
3406 if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}:
3407 features.add(Feature.POS_ONLY_ARGUMENTS)
3409 elif n.type == token.COLONEQUAL:
3410 features.add(Feature.ASSIGNMENT_EXPRESSIONS)
3413 n.type in {syms.typedargslist, syms.arglist}
3415 and n.children[-1].type == token.COMMA
3417 if n.type == syms.typedargslist:
3418 feature = Feature.TRAILING_COMMA_IN_DEF
3420 feature = Feature.TRAILING_COMMA_IN_CALL
3422 for ch in n.children:
3423 if ch.type in STARS:
3424 features.add(feature)
3426 if ch.type == syms.argument:
3427 for argch in ch.children:
3428 if argch.type in STARS:
3429 features.add(feature)
3434 def detect_target_versions(node: Node) -> Set[TargetVersion]:
3435 """Detect the version to target based on the nodes used."""
3436 features = get_features_used(node)
3438 version for version in TargetVersion if features <= VERSION_TO_FEATURES[version]
3442 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
3443 """Generate sets of closing bracket IDs that should be omitted in a RHS.
3445 Brackets can be omitted if the entire trailer up to and including
3446 a preceding closing bracket fits in one line.
3448 Yielded sets are cumulative (contain results of previous yields, too). First
3452 omit: Set[LeafID] = set()
3455 length = 4 * line.depth
3456 opening_bracket: Optional[Leaf] = None
3457 closing_bracket: Optional[Leaf] = None
3458 inner_brackets: Set[LeafID] = set()
3459 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
3460 length += leaf_length
3461 if length > line_length:
3464 has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
3465 if leaf.type == STANDALONE_COMMENT or has_inline_comment:
3469 if leaf is opening_bracket:
3470 opening_bracket = None
3471 elif leaf.type in CLOSING_BRACKETS:
3472 inner_brackets.add(id(leaf))
3473 elif leaf.type in CLOSING_BRACKETS:
3474 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
3475 # Empty brackets would fail a split so treat them as "inner"
3476 # brackets (e.g. only add them to the `omit` set if another
3477 # pair of brackets was good enough.
3478 inner_brackets.add(id(leaf))
3482 omit.add(id(closing_bracket))
3483 omit.update(inner_brackets)
3484 inner_brackets.clear()
3488 opening_bracket = leaf.opening_bracket
3489 closing_bracket = leaf
3492 def get_future_imports(node: Node) -> Set[str]:
3493 """Return a set of __future__ imports in the file."""
3494 imports: Set[str] = set()
3496 def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]:
3497 for child in children:
3498 if isinstance(child, Leaf):
3499 if child.type == token.NAME:
3502 elif child.type == syms.import_as_name:
3503 orig_name = child.children[0]
3504 assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
3505 assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
3506 yield orig_name.value
3508 elif child.type == syms.import_as_names:
3509 yield from get_imports_from_children(child.children)
3512 raise AssertionError("Invalid syntax parsing imports")
3514 for child in node.children:
3515 if child.type != syms.simple_stmt:
3518 first_child = child.children[0]
3519 if isinstance(first_child, Leaf):
3520 # Continue looking if we see a docstring; otherwise stop.
3522 len(child.children) == 2
3523 and first_child.type == token.STRING
3524 and child.children[1].type == token.NEWLINE
3530 elif first_child.type == syms.import_from:
3531 module_name = first_child.children[1]
3532 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
3535 imports |= set(get_imports_from_children(first_child.children[3:]))
3543 def get_gitignore(root: Path) -> PathSpec:
3544 """ Return a PathSpec matching gitignore content if present."""
3545 gitignore = root / ".gitignore"
3546 lines: List[str] = []
3547 if gitignore.is_file():
3548 with gitignore.open() as gf:
3549 lines = gf.readlines()
3550 return PathSpec.from_lines("gitwildmatch", lines)
3553 def gen_python_files_in_dir(
3556 include: Pattern[str],
3557 exclude: Pattern[str],
3559 gitignore: PathSpec,
3560 ) -> Iterator[Path]:
3561 """Generate all files under `path` whose paths are not excluded by the
3562 `exclude` regex, but are included by the `include` regex.
3564 Symbolic links pointing outside of the `root` directory are ignored.
3566 `report` is where output about exclusions goes.
3568 assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
3569 for child in path.iterdir():
3570 # First ignore files matching .gitignore
3571 if gitignore.match_file(child.as_posix()):
3572 report.path_ignored(child, f"matches the .gitignore file content")
3575 # Then ignore with `exclude` option.
3577 normalized_path = "/" + child.resolve().relative_to(root).as_posix()
3578 except OSError as e:
3579 report.path_ignored(child, f"cannot be read because {e}")
3583 if child.is_symlink():
3584 report.path_ignored(
3585 child, f"is a symbolic link that points outside {root}"
3592 normalized_path += "/"
3594 exclude_match = exclude.search(normalized_path)
3595 if exclude_match and exclude_match.group(0):
3596 report.path_ignored(child, f"matches the --exclude regular expression")
3600 yield from gen_python_files_in_dir(
3601 child, root, include, exclude, report, gitignore
3604 elif child.is_file():
3605 include_match = include.search(normalized_path)
3611 def find_project_root(srcs: Iterable[str]) -> Path:
3612 """Return a directory containing .git, .hg, or pyproject.toml.
3614 That directory can be one of the directories passed in `srcs` or their
3617 If no directory in the tree contains a marker that would specify it's the
3618 project root, the root of the file system is returned.
3621 return Path("/").resolve()
3623 common_base = min(Path(src).resolve() for src in srcs)
3624 if common_base.is_dir():
3625 # Append a fake file so `parents` below returns `common_base_dir`, too.
3626 common_base /= "fake-file"
3627 for directory in common_base.parents:
3628 if (directory / ".git").exists():
3631 if (directory / ".hg").is_dir():
3634 if (directory / "pyproject.toml").is_file():
3642 """Provides a reformatting counter. Can be rendered with `str(report)`."""
3647 verbose: bool = False
3648 change_count: int = 0
3650 failure_count: int = 0
3652 def done(self, src: Path, changed: Changed) -> None:
3653 """Increment the counter for successful reformatting. Write out a message."""
3654 if changed is Changed.YES:
3655 reformatted = "would reformat" if self.check or self.diff else "reformatted"
3656 if self.verbose or not self.quiet:
3657 out(f"{reformatted} {src}")
3658 self.change_count += 1
3661 if changed is Changed.NO:
3662 msg = f"{src} already well formatted, good job."
3664 msg = f"{src} wasn't modified on disk since last run."
3665 out(msg, bold=False)
3666 self.same_count += 1
3668 def failed(self, src: Path, message: str) -> None:
3669 """Increment the counter for failed reformatting. Write out a message."""
3670 err(f"error: cannot format {src}: {message}")
3671 self.failure_count += 1
3673 def path_ignored(self, path: Path, message: str) -> None:
3675 out(f"{path} ignored: {message}", bold=False)
3678 def return_code(self) -> int:
3679 """Return the exit code that the app should use.
3681 This considers the current state of changed files and failures:
3682 - if there were any failures, return 123;
3683 - if any files were changed and --check is being used, return 1;
3684 - otherwise return 0.
3686 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
3687 # 126 we have special return codes reserved by the shell.
3688 if self.failure_count:
3691 elif self.change_count and self.check:
3696 def __str__(self) -> str:
3697 """Render a color report of the current state.
3699 Use `click.unstyle` to remove colors.
3701 if self.check or self.diff:
3702 reformatted = "would be reformatted"
3703 unchanged = "would be left unchanged"
3704 failed = "would fail to reformat"
3706 reformatted = "reformatted"
3707 unchanged = "left unchanged"
3708 failed = "failed to reformat"
3710 if self.change_count:
3711 s = "s" if self.change_count > 1 else ""
3713 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
3716 s = "s" if self.same_count > 1 else ""
3717 report.append(f"{self.same_count} file{s} {unchanged}")
3718 if self.failure_count:
3719 s = "s" if self.failure_count > 1 else ""
3721 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
3723 return ", ".join(report) + "."
3726 def parse_ast(src: str) -> Union[ast.AST, ast3.AST, ast27.AST]:
3727 filename = "<unknown>"
3728 if sys.version_info >= (3, 8):
3729 # TODO: support Python 4+ ;)
3730 for minor_version in range(sys.version_info[1], 4, -1):
3732 return ast.parse(src, filename, feature_version=(3, minor_version))
3736 for feature_version in (7, 6):
3738 return ast3.parse(src, filename, feature_version=feature_version)
3742 return ast27.parse(src)
3745 def _fixup_ast_constants(
3746 node: Union[ast.AST, ast3.AST, ast27.AST]
3747 ) -> Union[ast.AST, ast3.AST, ast27.AST]:
3748 """Map ast nodes deprecated in 3.8 to Constant."""
3749 if isinstance(node, (ast.Str, ast3.Str, ast27.Str, ast.Bytes, ast3.Bytes)):
3750 return ast.Constant(value=node.s)
3752 if isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
3753 return ast.Constant(value=node.n)
3755 if isinstance(node, (ast.NameConstant, ast3.NameConstant)):
3756 return ast.Constant(value=node.value)
3761 def assert_equivalent(src: str, dst: str) -> None:
3762 """Raise AssertionError if `src` and `dst` aren't equivalent."""
3764 def _v(node: Union[ast.AST, ast3.AST, ast27.AST], depth: int = 0) -> Iterator[str]:
3765 """Simple visitor generating strings to compare ASTs by content."""
3767 node = _fixup_ast_constants(node)
3769 yield f"{' ' * depth}{node.__class__.__name__}("
3771 for field in sorted(node._fields): # noqa: F402
3772 # TypeIgnore has only one field 'lineno' which breaks this comparison
3773 type_ignore_classes = (ast3.TypeIgnore, ast27.TypeIgnore)
3774 if sys.version_info >= (3, 8):
3775 type_ignore_classes += (ast.TypeIgnore,)
3776 if isinstance(node, type_ignore_classes):
3780 value = getattr(node, field)
3781 except AttributeError:
3784 yield f"{' ' * (depth+1)}{field}="
3786 if isinstance(value, list):
3788 # Ignore nested tuples within del statements, because we may insert
3789 # parentheses and they change the AST.
3792 and isinstance(node, (ast.Delete, ast3.Delete, ast27.Delete))
3793 and isinstance(item, (ast.Tuple, ast3.Tuple, ast27.Tuple))
3795 for item in item.elts:
3796 yield from _v(item, depth + 2)
3798 elif isinstance(item, (ast.AST, ast3.AST, ast27.AST)):
3799 yield from _v(item, depth + 2)
3801 elif isinstance(value, (ast.AST, ast3.AST, ast27.AST)):
3802 yield from _v(value, depth + 2)
3805 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
3807 yield f"{' ' * depth}) # /{node.__class__.__name__}"
3810 src_ast = parse_ast(src)
3811 except Exception as exc:
3812 raise AssertionError(
3813 f"cannot use --safe with this file; failed to parse source file. "
3814 f"AST error message: {exc}"
3818 dst_ast = parse_ast(dst)
3819 except Exception as exc:
3820 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
3821 raise AssertionError(
3822 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
3823 f"Please report a bug on https://github.com/psf/black/issues. "
3824 f"This invalid output might be helpful: {log}"
3827 src_ast_str = "\n".join(_v(src_ast))
3828 dst_ast_str = "\n".join(_v(dst_ast))
3829 if src_ast_str != dst_ast_str:
3830 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
3831 raise AssertionError(
3832 f"INTERNAL ERROR: Black produced code that is not equivalent to "
3834 f"Please report a bug on https://github.com/psf/black/issues. "
3835 f"This diff might be helpful: {log}"
3839 def assert_stable(src: str, dst: str, mode: FileMode) -> None:
3840 """Raise AssertionError if `dst` reformats differently the second time."""
3841 newdst = format_str(dst, mode=mode)
3844 diff(src, dst, "source", "first pass"),
3845 diff(dst, newdst, "first pass", "second pass"),
3847 raise AssertionError(
3848 f"INTERNAL ERROR: Black produced different code on the second pass "
3849 f"of the formatter. "
3850 f"Please report a bug on https://github.com/psf/black/issues. "
3851 f"This diff might be helpful: {log}"
3855 @mypyc_attr(patchable=True)
3856 def dump_to_file(*output: str) -> str:
3857 """Dump `output` to a temporary file. Return path to the file."""
3858 with tempfile.NamedTemporaryFile(
3859 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
3861 for lines in output:
3863 if lines and lines[-1] != "\n":
3869 def nullcontext() -> Iterator[None]:
3870 """Return an empty context manager.
3872 To be used like `nullcontext` in Python 3.7.
3877 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
3878 """Return a unified diff string between strings `a` and `b`."""
3881 a_lines = [line + "\n" for line in a.split("\n")]
3882 b_lines = [line + "\n" for line in b.split("\n")]
3884 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
3888 def cancel(tasks: Iterable["asyncio.Task[Any]"]) -> None:
3889 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
3895 def shutdown(loop: asyncio.AbstractEventLoop) -> None:
3896 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
3898 if sys.version_info[:2] >= (3, 7):
3899 all_tasks = asyncio.all_tasks
3901 all_tasks = asyncio.Task.all_tasks
3902 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
3903 to_cancel = [task for task in all_tasks(loop) if not task.done()]
3907 for task in to_cancel:
3909 loop.run_until_complete(
3910 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
3913 # `concurrent.futures.Future` objects cannot be cancelled once they
3914 # are already running. There might be some when the `shutdown()` happened.
3915 # Silence their logger's spew about the event loop being closed.
3916 cf_logger = logging.getLogger("concurrent.futures")
3917 cf_logger.setLevel(logging.CRITICAL)
3921 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
3922 """Replace `regex` with `replacement` twice on `original`.
3924 This is used by string normalization to perform replaces on
3925 overlapping matches.
3927 return regex.sub(replacement, regex.sub(replacement, original))
3930 def re_compile_maybe_verbose(regex: str) -> Pattern[str]:
3931 """Compile a regular expression string in `regex`.
3933 If it contains newlines, use verbose mode.
3936 regex = "(?x)" + regex
3937 compiled: Pattern[str] = re.compile(regex)
3941 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
3942 """Like `reversed(enumerate(sequence))` if that were possible."""
3943 index = len(sequence) - 1
3944 for element in reversed(sequence):
3945 yield (index, element)
3949 def enumerate_with_length(
3950 line: Line, reversed: bool = False
3951 ) -> Iterator[Tuple[Index, Leaf, int]]:
3952 """Return an enumeration of leaves with their length.
3954 Stops prematurely on multiline strings and standalone comments.
3957 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
3958 enumerate_reversed if reversed else enumerate,
3960 for index, leaf in op(line.leaves):
3961 length = len(leaf.prefix) + len(leaf.value)
3962 if "\n" in leaf.value:
3963 return # Multiline strings, we can't continue.
3965 for comment in line.comments_after(leaf):
3966 length += len(comment.value)
3968 yield index, leaf, length
3971 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
3972 """Return True if `line` is no longer than `line_length`.
3974 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
3977 line_str = str(line).strip("\n")
3979 len(line_str) <= line_length
3980 and "\n" not in line_str # multiline strings
3981 and not line.contains_standalone_comments()
3985 def can_be_split(line: Line) -> bool:
3986 """Return False if the line cannot be split *for sure*.
3988 This is not an exhaustive search but a cheap heuristic that we can use to
3989 avoid some unfortunate formattings (mostly around wrapping unsplittable code
3990 in unnecessary parentheses).
3992 leaves = line.leaves
3996 if leaves[0].type == token.STRING and leaves[1].type == token.DOT:
4000 for leaf in leaves[-2::-1]:
4001 if leaf.type in OPENING_BRACKETS:
4002 if next.type not in CLOSING_BRACKETS:
4006 elif leaf.type == token.DOT:
4008 elif leaf.type == token.NAME:
4009 if not (next.type == token.DOT or next.type in OPENING_BRACKETS):
4012 elif leaf.type not in CLOSING_BRACKETS:
4015 if dot_count > 1 and call_count > 1:
4021 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
4022 """Does `line` have a shape safe to reformat without optional parens around it?
4024 Returns True for only a subset of potentially nice looking formattings but
4025 the point is to not return false positives that end up producing lines that
4028 bt = line.bracket_tracker
4029 if not bt.delimiters:
4030 # Without delimiters the optional parentheses are useless.
4033 max_priority = bt.max_delimiter_priority()
4034 if bt.delimiter_count_with_priority(max_priority) > 1:
4035 # With more than one delimiter of a kind the optional parentheses read better.
4038 if max_priority == DOT_PRIORITY:
4039 # A single stranded method call doesn't require optional parentheses.
4042 assert len(line.leaves) >= 2, "Stranded delimiter"
4044 first = line.leaves[0]
4045 second = line.leaves[1]
4046 penultimate = line.leaves[-2]
4047 last = line.leaves[-1]
4049 # With a single delimiter, omit if the expression starts or ends with
4051 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
4053 length = 4 * line.depth
4054 for _index, leaf, leaf_length in enumerate_with_length(line):
4055 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
4058 length += leaf_length
4059 if length > line_length:
4062 if leaf.type in OPENING_BRACKETS:
4063 # There are brackets we can further split on.
4067 # checked the entire string and line length wasn't exceeded
4068 if len(line.leaves) == _index + 1:
4071 # Note: we are not returning False here because a line might have *both*
4072 # a leading opening bracket and a trailing closing bracket. If the
4073 # opening bracket doesn't match our rule, maybe the closing will.
4076 last.type == token.RPAR
4077 or last.type == token.RBRACE
4079 # don't use indexing for omitting optional parentheses;
4081 last.type == token.RSQB
4083 and last.parent.type != syms.trailer
4086 if penultimate.type in OPENING_BRACKETS:
4087 # Empty brackets don't help.
4090 if is_multiline_string(first):
4091 # Additional wrapping of a multiline string in this situation is
4095 length = 4 * line.depth
4096 seen_other_brackets = False
4097 for _index, leaf, leaf_length in enumerate_with_length(line):
4098 length += leaf_length
4099 if leaf is last.opening_bracket:
4100 if seen_other_brackets or length <= line_length:
4103 elif leaf.type in OPENING_BRACKETS:
4104 # There are brackets we can further split on.
4105 seen_other_brackets = True
4110 def get_cache_file(mode: FileMode) -> Path:
4111 return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle"
4114 def read_cache(mode: FileMode) -> Cache:
4115 """Read the cache if it exists and is well formed.
4117 If it is not well formed, the call to write_cache later should resolve the issue.
4119 cache_file = get_cache_file(mode)
4120 if not cache_file.exists():
4123 with cache_file.open("rb") as fobj:
4125 cache: Cache = pickle.load(fobj)
4126 except (pickle.UnpicklingError, ValueError):
4132 def get_cache_info(path: Path) -> CacheInfo:
4133 """Return the information used to check if a file is already formatted or not."""
4135 return stat.st_mtime, stat.st_size
4138 def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]:
4139 """Split an iterable of paths in `sources` into two sets.
4141 The first contains paths of files that modified on disk or are not in the
4142 cache. The other contains paths to non-modified files.
4144 todo, done = set(), set()
4147 if cache.get(src) != get_cache_info(src):
4154 def write_cache(cache: Cache, sources: Iterable[Path], mode: FileMode) -> None:
4155 """Update the cache file."""
4156 cache_file = get_cache_file(mode)
4158 CACHE_DIR.mkdir(parents=True, exist_ok=True)
4159 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
4160 with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
4161 pickle.dump(new_cache, f, protocol=4)
4162 os.replace(f.name, cache_file)
4167 def patch_click() -> None:
4168 """Make Click not crash.
4170 On certain misconfigured environments, Python 3 selects the ASCII encoding as the
4171 default which restricts paths that it can access during the lifetime of the
4172 application. Click refuses to work in this scenario by raising a RuntimeError.
4174 In case of Black the likelihood that non-ASCII characters are going to be used in
4175 file paths is minimal since it's Python source code. Moreover, this crash was
4176 spurious on Python 3.7 thanks to PEP 538 and PEP 540.
4179 from click import core
4180 from click import _unicodefun # type: ignore
4181 except ModuleNotFoundError:
4184 for module in (core, _unicodefun):
4185 if hasattr(module, "_verify_python3_env"):
4186 module._verify_python3_env = lambda: None
4189 def patched_main() -> None:
4195 if __name__ == "__main__":