All patches and comments are welcome. Please squash your changes to logical
commits before using git-format-patch and git-send-email to
patches@git.madduck.net.
If you'd read over the Git project's submission guidelines and adhered to them,
I'd be especially grateful.
3 from concurrent.futures import Executor, ProcessPoolExecutor
4 from contextlib import contextmanager
5 from datetime import datetime
7 from functools import lru_cache, partial, wraps
11 from multiprocessing import Manager, freeze_support
13 from pathlib import Path
40 from typing_extensions import Final
41 from mypy_extensions import mypyc_attr
43 from appdirs import user_cache_dir
44 from dataclasses import dataclass, field, replace
47 from typed_ast import ast3, ast27
48 from pathspec import PathSpec
51 from blib2to3.pytree import Node, Leaf, type_repr
52 from blib2to3 import pygram, pytree
53 from blib2to3.pgen2 import driver, token
54 from blib2to3.pgen2.grammar import Grammar
55 from blib2to3.pgen2.parse import ParseError
57 from _black_version import version as __version__
59 DEFAULT_LINE_LENGTH = 88
60 DEFAULT_EXCLUDES = r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|\.svn|_build|buck-out|build|dist)/" # noqa: B950
61 DEFAULT_INCLUDES = r"\.pyi?$"
62 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
74 LN = Union[Leaf, Node]
75 SplitFunc = Callable[["Line", Collection["Feature"]], Iterator["Line"]]
78 CacheInfo = Tuple[Timestamp, FileSize]
79 Cache = Dict[Path, CacheInfo]
80 out = partial(click.secho, bold=True, err=True)
81 err = partial(click.secho, fg="red", err=True)
83 pygram.initialize(CACHE_DIR)
84 syms = pygram.python_symbols
87 class NothingChanged(UserWarning):
88 """Raised when reformatted code is the same as source."""
91 class CannotSplit(Exception):
92 """A readable split that fits the allotted line length is impossible."""
95 class InvalidInput(ValueError):
96 """Raised when input source code fails all parse attempts."""
99 class WriteBack(Enum):
106 def from_configuration(cls, *, check: bool, diff: bool) -> "WriteBack":
107 if check and not diff:
110 return cls.DIFF if diff else cls.YES
119 class TargetVersion(Enum):
128 def is_python2(self) -> bool:
129 return self is TargetVersion.PY27
132 PY36_VERSIONS = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38}
136 # All string literals are unicode
139 NUMERIC_UNDERSCORES = 3
140 TRAILING_COMMA_IN_CALL = 4
141 TRAILING_COMMA_IN_DEF = 5
142 # The following two feature-flags are mutually exclusive, and exactly one should be
143 # set for every version of python.
144 ASYNC_IDENTIFIERS = 6
146 ASSIGNMENT_EXPRESSIONS = 8
147 POS_ONLY_ARGUMENTS = 9
150 VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = {
151 TargetVersion.PY27: {Feature.ASYNC_IDENTIFIERS},
152 TargetVersion.PY33: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
153 TargetVersion.PY34: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
154 TargetVersion.PY35: {
155 Feature.UNICODE_LITERALS,
156 Feature.TRAILING_COMMA_IN_CALL,
157 Feature.ASYNC_IDENTIFIERS,
159 TargetVersion.PY36: {
160 Feature.UNICODE_LITERALS,
162 Feature.NUMERIC_UNDERSCORES,
163 Feature.TRAILING_COMMA_IN_CALL,
164 Feature.TRAILING_COMMA_IN_DEF,
165 Feature.ASYNC_IDENTIFIERS,
167 TargetVersion.PY37: {
168 Feature.UNICODE_LITERALS,
170 Feature.NUMERIC_UNDERSCORES,
171 Feature.TRAILING_COMMA_IN_CALL,
172 Feature.TRAILING_COMMA_IN_DEF,
173 Feature.ASYNC_KEYWORDS,
175 TargetVersion.PY38: {
176 Feature.UNICODE_LITERALS,
178 Feature.NUMERIC_UNDERSCORES,
179 Feature.TRAILING_COMMA_IN_CALL,
180 Feature.TRAILING_COMMA_IN_DEF,
181 Feature.ASYNC_KEYWORDS,
182 Feature.ASSIGNMENT_EXPRESSIONS,
183 Feature.POS_ONLY_ARGUMENTS,
190 target_versions: Set[TargetVersion] = field(default_factory=set)
191 line_length: int = DEFAULT_LINE_LENGTH
192 string_normalization: bool = True
195 def get_cache_key(self) -> str:
196 if self.target_versions:
197 version_str = ",".join(
199 for version in sorted(self.target_versions, key=lambda v: v.value)
205 str(self.line_length),
206 str(int(self.string_normalization)),
207 str(int(self.is_pyi)),
209 return ".".join(parts)
212 # Legacy name, left for integrations.
216 def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool:
217 return all(feature in VERSION_TO_FEATURES[version] for version in target_versions)
220 def find_pyproject_toml(path_search_start: str) -> Optional[str]:
221 """Find the absolute filepath to a pyproject.toml if it exists"""
222 path_project_root = find_project_root(path_search_start)
223 path_pyproject_toml = path_project_root / "pyproject.toml"
224 return str(path_pyproject_toml) if path_pyproject_toml.is_file() else None
227 def parse_pyproject_toml(path_config: str) -> Dict[str, Any]:
228 """Parse a pyproject toml file, pulling out relevant parts for Black
230 If parsing fails, will raise a toml.TomlDecodeError
232 pyproject_toml = toml.load(path_config)
233 config = pyproject_toml.get("tool", {}).get("black", {})
234 return {k.replace("--", "").replace("-", "_"): v for k, v in config.items()}
237 def read_pyproject_toml(
238 ctx: click.Context, param: click.Parameter, value: Union[str, int, bool, None]
240 """Inject Black configuration from "pyproject.toml" into defaults in `ctx`.
242 Returns the path to a successfully found and read configuration file, None
245 assert not isinstance(value, (int, bool)), "Invalid parameter type passed"
247 value = find_pyproject_toml(ctx.params.get("src", ()))
252 config = parse_pyproject_toml(value)
253 except (toml.TomlDecodeError, OSError) as e:
254 raise click.FileError(
255 filename=value, hint=f"Error reading configuration file: {e}"
261 if ctx.default_map is None:
263 ctx.default_map.update(config) # type: ignore # bad types in .pyi
267 def target_version_option_callback(
268 c: click.Context, p: Union[click.Option, click.Parameter], v: Tuple[str, ...]
269 ) -> List[TargetVersion]:
270 """Compute the target versions from a --target-version flag.
272 This is its own function because mypy couldn't infer the type correctly
273 when it was a lambda, causing mypyc trouble.
275 return [TargetVersion[val.upper()] for val in v]
278 @click.command(context_settings=dict(help_option_names=["-h", "--help"]))
279 @click.option("-c", "--code", type=str, help="Format the code passed in as a string.")
284 default=DEFAULT_LINE_LENGTH,
285 help="How many characters per line to allow.",
291 type=click.Choice([v.name.lower() for v in TargetVersion]),
292 callback=target_version_option_callback,
295 "Python versions that should be supported by Black's output. [default: "
296 "per-file auto-detection]"
303 "Allow using Python 3.6-only syntax on all input files. This will put "
304 "trailing commas in function signatures and calls also after *args and "
305 "**kwargs. Deprecated; use --target-version instead. "
306 "[default: per-file auto-detection]"
313 "Format all input files like typing stubs regardless of file extension "
314 "(useful when piping source on standard input)."
319 "--skip-string-normalization",
321 help="Don't normalize string quotes or prefixes.",
327 "Don't write the files back, just return the status. Return code 0 "
328 "means nothing would change. Return code 1 means some files would be "
329 "reformatted. Return code 123 means there was an internal error."
335 help="Don't write the files back, just output a diff for each file on stdout.",
340 help="If --fast given, skip temporary sanity checks. [default: --safe]",
345 default=DEFAULT_INCLUDES,
347 "A regular expression that matches files and directories that should be "
348 "included on recursive searches. An empty value means all files are "
349 "included regardless of the name. Use forward slashes for directories on "
350 "all platforms (Windows, too). Exclusions are calculated first, inclusions "
358 default=DEFAULT_EXCLUDES,
360 "A regular expression that matches files and directories that should be "
361 "excluded on recursive searches. An empty value means no paths are excluded. "
362 "Use forward slashes for directories on all platforms (Windows, too). "
363 "Exclusions are calculated first, inclusions later."
372 "Don't emit non-error messages to stderr. Errors are still emitted; "
373 "silence those with 2>/dev/null."
381 "Also emit messages to stderr about files that were not changed or were "
382 "ignored due to --exclude=."
385 @click.version_option(version=__version__)
390 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
397 exists=True, file_okay=True, dir_okay=False, readable=True, allow_dash=False
400 callback=read_pyproject_toml,
401 help="Read configuration from PATH.",
408 target_version: List[TargetVersion],
414 skip_string_normalization: bool,
419 src: Tuple[str, ...],
420 config: Optional[str],
422 """The uncompromising code formatter."""
423 write_back = WriteBack.from_configuration(check=check, diff=diff)
426 err("Cannot use both --target-version and --py36")
429 versions = set(target_version)
432 "--py36 is deprecated and will be removed in a future version. "
433 "Use --target-version py36 instead."
435 versions = PY36_VERSIONS
437 # We'll autodetect later.
440 target_versions=versions,
441 line_length=line_length,
443 string_normalization=not skip_string_normalization,
445 if config and verbose:
446 out(f"Using configuration from {config}.", bold=False, fg="blue")
448 print(format_str(code, mode=mode))
451 include_regex = re_compile_maybe_verbose(include)
453 err(f"Invalid regular expression for include given: {include!r}")
456 exclude_regex = re_compile_maybe_verbose(exclude)
458 err(f"Invalid regular expression for exclude given: {exclude!r}")
460 report = Report(check=check, diff=diff, quiet=quiet, verbose=verbose)
461 root = find_project_root(src)
462 sources: Set[Path] = set()
463 path_empty(src, quiet, verbose, ctx)
468 gen_python_files_in_dir(
469 p, root, include_regex, exclude_regex, report, get_gitignore(root)
472 elif p.is_file() or s == "-":
473 # if a file was explicitly given, we don't care about its extension
476 err(f"invalid path: {s}")
477 if len(sources) == 0:
478 if verbose or not quiet:
479 out("No Python files are present to be formatted. Nothing to do 😴")
482 if len(sources) == 1:
486 write_back=write_back,
492 sources=sources, fast=fast, write_back=write_back, mode=mode, report=report
495 if verbose or not quiet:
496 out("Oh no! 💥 💔 💥" if report.return_code else "All done! ✨ 🍰 ✨")
497 click.secho(str(report), err=True)
498 ctx.exit(report.return_code)
502 src: Tuple[str, ...], quiet: bool, verbose: bool, ctx: click.Context
505 Exit if there is no `src` provided for formatting
508 if verbose or not quiet:
509 out("No Path provided. Nothing to do 😴")
514 src: Path, fast: bool, write_back: WriteBack, mode: Mode, report: "Report"
516 """Reformat a single file under `src` without spawning child processes.
518 `fast`, `write_back`, and `mode` options are passed to
519 :func:`format_file_in_place` or :func:`format_stdin_to_stdout`.
523 if not src.is_file() and str(src) == "-":
524 if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode):
525 changed = Changed.YES
528 if write_back != WriteBack.DIFF:
529 cache = read_cache(mode)
530 res_src = src.resolve()
531 if res_src in cache and cache[res_src] == get_cache_info(res_src):
532 changed = Changed.CACHED
533 if changed is not Changed.CACHED and format_file_in_place(
534 src, fast=fast, write_back=write_back, mode=mode
536 changed = Changed.YES
537 if (write_back is WriteBack.YES and changed is not Changed.CACHED) or (
538 write_back is WriteBack.CHECK and changed is Changed.NO
540 write_cache(cache, [src], mode)
541 report.done(src, changed)
542 except Exception as exc:
543 report.failed(src, str(exc))
547 sources: Set[Path], fast: bool, write_back: WriteBack, mode: Mode, report: "Report"
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: Mode
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(src_contents: str, *, fast: bool, mode: Mode) -> FileContent:
724 """Reformat contents a file and return new contents.
726 If `fast` is False, additionally confirm that the reformatted code is
727 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
728 `mode` is passed to :func:`format_str`.
730 if src_contents.strip() == "":
733 dst_contents = format_str(src_contents, mode=mode)
734 if src_contents == dst_contents:
738 assert_equivalent(src_contents, dst_contents)
739 assert_stable(src_contents, dst_contents, mode=mode)
743 def format_str(src_contents: str, *, mode: Mode) -> FileContent:
744 """Reformat a string and return new contents.
746 `mode` determines formatting options, such as how many characters per line are
750 >>> print(black.format_str("def f(arg:str='')->None:...", mode=Mode()))
751 def f(arg: str = "") -> None:
754 A more complex example:
756 ... black.format_str(
757 ... "def f(arg:str='')->None: hey",
759 ... target_versions={black.TargetVersion.PY36},
761 ... string_normalization=False,
772 src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
774 future_imports = get_future_imports(src_node)
775 if mode.target_versions:
776 versions = mode.target_versions
778 versions = detect_target_versions(src_node)
779 normalize_fmt_off(src_node)
780 lines = LineGenerator(
781 remove_u_prefix="unicode_literals" in future_imports
782 or supports_feature(versions, Feature.UNICODE_LITERALS),
784 normalize_strings=mode.string_normalization,
786 elt = EmptyLineTracker(is_pyi=mode.is_pyi)
789 split_line_features = {
791 for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF}
792 if supports_feature(versions, feature)
794 for current_line in lines.visit(src_node):
795 dst_contents.append(str(empty_line) * after)
796 before, after = elt.maybe_empty_lines(current_line)
797 dst_contents.append(str(empty_line) * before)
798 for line in split_line(
799 current_line, line_length=mode.line_length, features=split_line_features
801 dst_contents.append(str(line))
802 return "".join(dst_contents)
805 def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]:
806 """Return a tuple of (decoded_contents, encoding, newline).
808 `newline` is either CRLF or LF but `decoded_contents` is decoded with
809 universal newlines (i.e. only contains LF).
811 srcbuf = io.BytesIO(src)
812 encoding, lines = tokenize.detect_encoding(srcbuf.readline)
814 return "", encoding, "\n"
816 newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n"
818 with io.TextIOWrapper(srcbuf, encoding) as tiow:
819 return tiow.read(), encoding, newline
822 def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
823 if not target_versions:
824 # No target_version specified, so try all grammars.
827 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords,
829 pygram.python_grammar_no_print_statement_no_exec_statement,
830 # Python 2.7 with future print_function import
831 pygram.python_grammar_no_print_statement,
833 pygram.python_grammar,
836 if all(version.is_python2() for version in target_versions):
837 # Python 2-only code, so try Python 2 grammars.
839 # Python 2.7 with future print_function import
840 pygram.python_grammar_no_print_statement,
842 pygram.python_grammar,
845 # Python 3-compatible code, so only try Python 3 grammar.
847 # If we have to parse both, try to parse async as a keyword first
848 if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
851 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords
853 if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
855 grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
856 # At least one of the above branches must have been taken, because every Python
857 # version has exactly one of the two 'ASYNC_*' flags
861 def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
862 """Given a string with source, return the lib2to3 Node."""
863 if src_txt[-1:] != "\n":
866 for grammar in get_grammars(set(target_versions)):
867 drv = driver.Driver(grammar, pytree.convert)
869 result = drv.parse_string(src_txt, True)
872 except ParseError as pe:
873 lineno, column = pe.context[1]
874 lines = src_txt.splitlines()
876 faulty_line = lines[lineno - 1]
878 faulty_line = "<line number missing in source>"
879 exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
883 if isinstance(result, Leaf):
884 result = Node(syms.file_input, [result])
888 def lib2to3_unparse(node: Node) -> str:
889 """Given a lib2to3 node, return its string representation."""
897 class Visitor(Generic[T]):
898 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
900 def visit(self, node: LN) -> Iterator[T]:
901 """Main method to visit `node` and its children.
903 It tries to find a `visit_*()` method for the given `node.type`, like
904 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
905 If no dedicated `visit_*()` method is found, chooses `visit_default()`
908 Then yields objects of type `T` from the selected visitor.
911 name = token.tok_name[node.type]
913 name = str(type_repr(node.type))
914 # We explicitly branch on whether a visitor exists (instead of
915 # using self.visit_default as the default arg to getattr) in order
916 # to save needing to create a bound method object and so mypyc can
917 # generate a native call to visit_default.
918 visitf = getattr(self, f"visit_{name}", None)
920 yield from visitf(node)
922 yield from self.visit_default(node)
924 def visit_default(self, node: LN) -> Iterator[T]:
925 """Default `visit_*()` implementation. Recurses to children of `node`."""
926 if isinstance(node, Node):
927 for child in node.children:
928 yield from self.visit(child)
932 class DebugVisitor(Visitor[T]):
935 def visit_default(self, node: LN) -> Iterator[T]:
936 indent = " " * (2 * self.tree_depth)
937 if isinstance(node, Node):
938 _type = type_repr(node.type)
939 out(f"{indent}{_type}", fg="yellow")
941 for child in node.children:
942 yield from self.visit(child)
945 out(f"{indent}/{_type}", fg="yellow", bold=False)
947 _type = token.tok_name.get(node.type, str(node.type))
948 out(f"{indent}{_type}", fg="blue", nl=False)
950 # We don't have to handle prefixes for `Node` objects since
951 # that delegates to the first child anyway.
952 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
953 out(f" {node.value!r}", fg="blue", bold=False)
956 def show(cls, code: Union[str, Leaf, Node]) -> None:
957 """Pretty-print the lib2to3 AST of a given string of `code`.
959 Convenience method for debugging.
961 v: DebugVisitor[None] = DebugVisitor()
962 if isinstance(code, str):
963 code = lib2to3_parse(code)
967 WHITESPACE: Final = {token.DEDENT, token.INDENT, token.NEWLINE}
978 STANDALONE_COMMENT: Final = 153
979 token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT"
980 LOGIC_OPERATORS: Final = {"and", "or"}
981 COMPARATORS: Final = {
989 MATH_OPERATORS: Final = {
1005 STARS: Final = {token.STAR, token.DOUBLESTAR}
1006 VARARGS_SPECIALS: Final = STARS | {token.SLASH}
1007 VARARGS_PARENTS: Final = {
1009 syms.argument, # double star in arglist
1010 syms.trailer, # single argument to call
1012 syms.varargslist, # lambdas
1014 UNPACKING_PARENTS: Final = {
1015 syms.atom, # single element of a list or set literal
1019 syms.testlist_star_expr,
1021 TEST_DESCENDANTS: Final = {
1038 ASSIGNMENTS: Final = {
1054 COMPREHENSION_PRIORITY: Final = 20
1055 COMMA_PRIORITY: Final = 18
1056 TERNARY_PRIORITY: Final = 16
1057 LOGIC_PRIORITY: Final = 14
1058 STRING_PRIORITY: Final = 12
1059 COMPARATOR_PRIORITY: Final = 10
1060 MATH_PRIORITIES: Final = {
1062 token.CIRCUMFLEX: 8,
1065 token.RIGHTSHIFT: 6,
1070 token.DOUBLESLASH: 4,
1074 token.DOUBLESTAR: 2,
1076 DOT_PRIORITY: Final = 1
1080 class BracketTracker:
1081 """Keeps track of brackets on a line."""
1084 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = field(default_factory=dict)
1085 delimiters: Dict[LeafID, Priority] = field(default_factory=dict)
1086 previous: Optional[Leaf] = None
1087 _for_loop_depths: List[int] = field(default_factory=list)
1088 _lambda_argument_depths: List[int] = field(default_factory=list)
1090 def mark(self, leaf: Leaf) -> None:
1091 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
1093 All leaves receive an int `bracket_depth` field that stores how deep
1094 within brackets a given leaf is. 0 means there are no enclosing brackets
1095 that started on this line.
1097 If a leaf is itself a closing bracket, it receives an `opening_bracket`
1098 field that it forms a pair with. This is a one-directional link to
1099 avoid reference cycles.
1101 If a leaf is a delimiter (a token on which Black can split the line if
1102 needed) and it's on depth 0, its `id()` is stored in the tracker's
1105 if leaf.type == token.COMMENT:
1108 self.maybe_decrement_after_for_loop_variable(leaf)
1109 self.maybe_decrement_after_lambda_arguments(leaf)
1110 if leaf.type in CLOSING_BRACKETS:
1112 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
1113 leaf.opening_bracket = opening_bracket
1114 leaf.bracket_depth = self.depth
1116 delim = is_split_before_delimiter(leaf, self.previous)
1117 if delim and self.previous is not None:
1118 self.delimiters[id(self.previous)] = delim
1120 delim = is_split_after_delimiter(leaf, self.previous)
1122 self.delimiters[id(leaf)] = delim
1123 if leaf.type in OPENING_BRACKETS:
1124 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
1126 self.previous = leaf
1127 self.maybe_increment_lambda_arguments(leaf)
1128 self.maybe_increment_for_loop_variable(leaf)
1130 def any_open_brackets(self) -> bool:
1131 """Return True if there is an yet unmatched open bracket on the line."""
1132 return bool(self.bracket_match)
1134 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> Priority:
1135 """Return the highest priority of a delimiter found on the line.
1137 Values are consistent with what `is_split_*_delimiter()` return.
1138 Raises ValueError on no delimiters.
1140 return max(v for k, v in self.delimiters.items() if k not in exclude)
1142 def delimiter_count_with_priority(self, priority: Priority = 0) -> int:
1143 """Return the number of delimiters with the given `priority`.
1145 If no `priority` is passed, defaults to max priority on the line.
1147 if not self.delimiters:
1150 priority = priority or self.max_delimiter_priority()
1151 return sum(1 for p in self.delimiters.values() if p == priority)
1153 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1154 """In a for loop, or comprehension, the variables are often unpacks.
1156 To avoid splitting on the comma in this situation, increase the depth of
1157 tokens between `for` and `in`.
1159 if leaf.type == token.NAME and leaf.value == "for":
1161 self._for_loop_depths.append(self.depth)
1166 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
1167 """See `maybe_increment_for_loop_variable` above for explanation."""
1169 self._for_loop_depths
1170 and self._for_loop_depths[-1] == self.depth
1171 and leaf.type == token.NAME
1172 and leaf.value == "in"
1175 self._for_loop_depths.pop()
1180 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
1181 """In a lambda expression, there might be more than one argument.
1183 To avoid splitting on the comma in this situation, increase the depth of
1184 tokens between `lambda` and `:`.
1186 if leaf.type == token.NAME and leaf.value == "lambda":
1188 self._lambda_argument_depths.append(self.depth)
1193 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
1194 """See `maybe_increment_lambda_arguments` above for explanation."""
1196 self._lambda_argument_depths
1197 and self._lambda_argument_depths[-1] == self.depth
1198 and leaf.type == token.COLON
1201 self._lambda_argument_depths.pop()
1206 def get_open_lsqb(self) -> Optional[Leaf]:
1207 """Return the most recent opening square bracket (if any)."""
1208 return self.bracket_match.get((self.depth - 1, token.RSQB))
1213 """Holds leaves and comments. Can be printed with `str(line)`."""
1216 leaves: List[Leaf] = field(default_factory=list)
1217 # keys ordered like `leaves`
1218 comments: Dict[LeafID, List[Leaf]] = field(default_factory=dict)
1219 bracket_tracker: BracketTracker = field(default_factory=BracketTracker)
1220 inside_brackets: bool = False
1221 should_explode: bool = False
1223 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
1224 """Add a new `leaf` to the end of the line.
1226 Unless `preformatted` is True, the `leaf` will receive a new consistent
1227 whitespace prefix and metadata applied by :class:`BracketTracker`.
1228 Trailing commas are maybe removed, unpacked for loop variables are
1229 demoted from being delimiters.
1231 Inline comments are put aside.
1233 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
1237 if token.COLON == leaf.type and self.is_class_paren_empty:
1238 del self.leaves[-2:]
1239 if self.leaves and not preformatted:
1240 # Note: at this point leaf.prefix should be empty except for
1241 # imports, for which we only preserve newlines.
1242 leaf.prefix += whitespace(
1243 leaf, complex_subscript=self.is_complex_subscript(leaf)
1245 if self.inside_brackets or not preformatted:
1246 self.bracket_tracker.mark(leaf)
1247 self.maybe_remove_trailing_comma(leaf)
1248 if not self.append_comment(leaf):
1249 self.leaves.append(leaf)
1251 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
1252 """Like :func:`append()` but disallow invalid standalone comment structure.
1254 Raises ValueError when any `leaf` is appended after a standalone comment
1255 or when a standalone comment is not the first leaf on the line.
1257 if self.bracket_tracker.depth == 0:
1259 raise ValueError("cannot append to standalone comments")
1261 if self.leaves and leaf.type == STANDALONE_COMMENT:
1263 "cannot append standalone comments to a populated line"
1266 self.append(leaf, preformatted=preformatted)
1269 def is_comment(self) -> bool:
1270 """Is this line a standalone comment?"""
1271 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
1274 def is_decorator(self) -> bool:
1275 """Is this line a decorator?"""
1276 return bool(self) and self.leaves[0].type == token.AT
1279 def is_import(self) -> bool:
1280 """Is this an import line?"""
1281 return bool(self) and is_import(self.leaves[0])
1284 def is_class(self) -> bool:
1285 """Is this line a class definition?"""
1288 and self.leaves[0].type == token.NAME
1289 and self.leaves[0].value == "class"
1293 def is_stub_class(self) -> bool:
1294 """Is this line a class definition with a body consisting only of "..."?"""
1295 return self.is_class and self.leaves[-3:] == [
1296 Leaf(token.DOT, ".") for _ in range(3)
1300 def is_collection_with_optional_trailing_comma(self) -> bool:
1301 """Is this line a collection literal with a trailing comma that's optional?
1303 Note that the trailing comma in a 1-tuple is not optional.
1305 if not self.leaves or len(self.leaves) < 4:
1308 # Look for and address a trailing colon.
1309 if self.leaves[-1].type == token.COLON:
1310 closer = self.leaves[-2]
1313 closer = self.leaves[-1]
1315 if closer.type not in CLOSING_BRACKETS or self.inside_brackets:
1318 if closer.type == token.RPAR:
1319 # Tuples require an extra check, because if there's only
1320 # one element in the tuple removing the comma unmakes the
1323 # We also check for parens before looking for the trailing
1324 # comma because in some cases (eg assigning a dict
1325 # literal) the literal gets wrapped in temporary parens
1326 # during parsing. This case is covered by the
1327 # collections.py test data.
1328 opener = closer.opening_bracket
1329 for _open_index, leaf in enumerate(self.leaves):
1334 # Couldn't find the matching opening paren, play it safe.
1338 comma_depth = self.leaves[close_index - 1].bracket_depth
1339 for leaf in self.leaves[_open_index + 1 : close_index]:
1340 if leaf.bracket_depth == comma_depth and leaf.type == token.COMMA:
1343 # We haven't looked yet for the trailing comma because
1344 # we might also have caught noop parens.
1345 return self.leaves[close_index - 1].type == token.COMMA
1348 return False # it's either a one-tuple or didn't have a trailing comma
1350 if self.leaves[close_index - 1].type in CLOSING_BRACKETS:
1352 closer = self.leaves[close_index]
1353 if closer.type == token.RPAR:
1354 # TODO: this is a gut feeling. Will we ever see this?
1357 if self.leaves[close_index - 1].type != token.COMMA:
1363 def is_def(self) -> bool:
1364 """Is this a function definition? (Also returns True for async defs.)"""
1366 first_leaf = self.leaves[0]
1371 second_leaf: Optional[Leaf] = self.leaves[1]
1374 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
1375 first_leaf.type == token.ASYNC
1376 and second_leaf is not None
1377 and second_leaf.type == token.NAME
1378 and second_leaf.value == "def"
1382 def is_class_paren_empty(self) -> bool:
1383 """Is this a class with no base classes but using parentheses?
1385 Those are unnecessary and should be removed.
1389 and len(self.leaves) == 4
1391 and self.leaves[2].type == token.LPAR
1392 and self.leaves[2].value == "("
1393 and self.leaves[3].type == token.RPAR
1394 and self.leaves[3].value == ")"
1398 def is_triple_quoted_string(self) -> bool:
1399 """Is the line a triple quoted string?"""
1402 and self.leaves[0].type == token.STRING
1403 and self.leaves[0].value.startswith(('"""', "'''"))
1406 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
1407 """If so, needs to be split before emitting."""
1408 for leaf in self.leaves:
1409 if leaf.type == STANDALONE_COMMENT and leaf.bracket_depth <= depth_limit:
1414 def contains_uncollapsable_type_comments(self) -> bool:
1417 last_leaf = self.leaves[-1]
1418 ignored_ids.add(id(last_leaf))
1419 if last_leaf.type == token.COMMA or (
1420 last_leaf.type == token.RPAR and not last_leaf.value
1422 # When trailing commas or optional parens are inserted by Black for
1423 # consistency, comments after the previous last element are not moved
1424 # (they don't have to, rendering will still be correct). So we ignore
1425 # trailing commas and invisible.
1426 last_leaf = self.leaves[-2]
1427 ignored_ids.add(id(last_leaf))
1431 # A type comment is uncollapsable if it is attached to a leaf
1432 # that isn't at the end of the line (since that could cause it
1433 # to get associated to a different argument) or if there are
1434 # comments before it (since that could cause it to get hidden
1436 comment_seen = False
1437 for leaf_id, comments in self.comments.items():
1438 for comment in comments:
1439 if is_type_comment(comment):
1440 if comment_seen or (
1441 not is_type_comment(comment, " ignore")
1442 and leaf_id not in ignored_ids
1450 def contains_unsplittable_type_ignore(self) -> bool:
1454 # If a 'type: ignore' is attached to the end of a line, we
1455 # can't split the line, because we can't know which of the
1456 # subexpressions the ignore was meant to apply to.
1458 # We only want this to apply to actual physical lines from the
1459 # original source, though: we don't want the presence of a
1460 # 'type: ignore' at the end of a multiline expression to
1461 # justify pushing it all onto one line. Thus we
1462 # (unfortunately) need to check the actual source lines and
1463 # only report an unsplittable 'type: ignore' if this line was
1464 # one line in the original code.
1466 # Grab the first and last line numbers, skipping generated leaves
1467 first_line = next((l.lineno for l in self.leaves if l.lineno != 0), 0)
1468 last_line = next((l.lineno for l in reversed(self.leaves) if l.lineno != 0), 0)
1470 if first_line == last_line:
1471 # We look at the last two leaves since a comma or an
1472 # invisible paren could have been added at the end of the
1474 for node in self.leaves[-2:]:
1475 for comment in self.comments.get(id(node), []):
1476 if is_type_comment(comment, " ignore"):
1481 def contains_multiline_strings(self) -> bool:
1482 return any(is_multiline_string(leaf) for leaf in self.leaves)
1484 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1485 """Remove trailing comma if there is one and it's safe."""
1486 if not (self.leaves and self.leaves[-1].type == token.COMMA):
1489 # We remove trailing commas only in the case of importing a
1490 # single name from a module.
1494 and len(self.leaves) > 4
1495 and self.leaves[-1].type == token.COMMA
1496 and closing.type in CLOSING_BRACKETS
1497 and self.leaves[-4].type == token.NAME
1499 # regular `from foo import bar,`
1500 self.leaves[-4].value == "import"
1501 # `from foo import (bar as baz,)
1503 len(self.leaves) > 6
1504 and self.leaves[-6].value == "import"
1505 and self.leaves[-3].value == "as"
1507 # `from foo import bar as baz,`
1509 len(self.leaves) > 5
1510 and self.leaves[-5].value == "import"
1511 and self.leaves[-3].value == "as"
1514 and closing.type == token.RPAR
1518 self.remove_trailing_comma()
1521 def append_comment(self, comment: Leaf) -> bool:
1522 """Add an inline or standalone comment to the line."""
1524 comment.type == STANDALONE_COMMENT
1525 and self.bracket_tracker.any_open_brackets()
1530 if comment.type != token.COMMENT:
1534 comment.type = STANDALONE_COMMENT
1538 last_leaf = self.leaves[-1]
1540 last_leaf.type == token.RPAR
1541 and not last_leaf.value
1542 and last_leaf.parent
1543 and len(list(last_leaf.parent.leaves())) <= 3
1544 and not is_type_comment(comment)
1546 # Comments on an optional parens wrapping a single leaf should belong to
1547 # the wrapped node except if it's a type comment. Pinning the comment like
1548 # this avoids unstable formatting caused by comment migration.
1549 if len(self.leaves) < 2:
1550 comment.type = STANDALONE_COMMENT
1554 last_leaf = self.leaves[-2]
1555 self.comments.setdefault(id(last_leaf), []).append(comment)
1558 def comments_after(self, leaf: Leaf) -> List[Leaf]:
1559 """Generate comments that should appear directly after `leaf`."""
1560 return self.comments.get(id(leaf), [])
1562 def remove_trailing_comma(self) -> None:
1563 """Remove the trailing comma and moves the comments attached to it."""
1564 trailing_comma = self.leaves.pop()
1565 trailing_comma_comments = self.comments.pop(id(trailing_comma), [])
1566 self.comments.setdefault(id(self.leaves[-1]), []).extend(
1567 trailing_comma_comments
1570 def is_complex_subscript(self, leaf: Leaf) -> bool:
1571 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1572 open_lsqb = self.bracket_tracker.get_open_lsqb()
1573 if open_lsqb is None:
1576 subscript_start = open_lsqb.next_sibling
1578 if isinstance(subscript_start, Node):
1579 if subscript_start.type == syms.listmaker:
1582 if subscript_start.type == syms.subscriptlist:
1583 subscript_start = child_towards(subscript_start, leaf)
1584 return subscript_start is not None and any(
1585 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1588 def __str__(self) -> str:
1589 """Render the line."""
1593 indent = " " * self.depth
1594 leaves = iter(self.leaves)
1595 first = next(leaves)
1596 res = f"{first.prefix}{indent}{first.value}"
1599 for comment in itertools.chain.from_iterable(self.comments.values()):
1603 def __bool__(self) -> bool:
1604 """Return True if the line has leaves or comments."""
1605 return bool(self.leaves or self.comments)
1609 class EmptyLineTracker:
1610 """Provides a stateful method that returns the number of potential extra
1611 empty lines needed before and after the currently processed line.
1613 Note: this tracker works on lines that haven't been split yet. It assumes
1614 the prefix of the first leaf consists of optional newlines. Those newlines
1615 are consumed by `maybe_empty_lines()` and included in the computation.
1618 is_pyi: bool = False
1619 previous_line: Optional[Line] = None
1620 previous_after: int = 0
1621 previous_defs: List[int] = field(default_factory=list)
1623 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1624 """Return the number of extra empty lines before and after the `current_line`.
1626 This is for separating `def`, `async def` and `class` with extra empty
1627 lines (two on module-level).
1629 before, after = self._maybe_empty_lines(current_line)
1631 # Black should not insert empty lines at the beginning
1634 if self.previous_line is None
1635 else before - self.previous_after
1637 self.previous_after = after
1638 self.previous_line = current_line
1639 return before, after
1641 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1643 if current_line.depth == 0:
1644 max_allowed = 1 if self.is_pyi else 2
1645 if current_line.leaves:
1646 # Consume the first leaf's extra newlines.
1647 first_leaf = current_line.leaves[0]
1648 before = first_leaf.prefix.count("\n")
1649 before = min(before, max_allowed)
1650 first_leaf.prefix = ""
1653 depth = current_line.depth
1654 while self.previous_defs and self.previous_defs[-1] >= depth:
1655 self.previous_defs.pop()
1657 before = 0 if depth else 1
1659 before = 1 if depth else 2
1660 if current_line.is_decorator or current_line.is_def or current_line.is_class:
1661 return self._maybe_empty_lines_for_class_or_def(current_line, before)
1665 and self.previous_line.is_import
1666 and not current_line.is_import
1667 and depth == self.previous_line.depth
1669 return (before or 1), 0
1673 and self.previous_line.is_class
1674 and current_line.is_triple_quoted_string
1680 def _maybe_empty_lines_for_class_or_def(
1681 self, current_line: Line, before: int
1682 ) -> Tuple[int, int]:
1683 if not current_line.is_decorator:
1684 self.previous_defs.append(current_line.depth)
1685 if self.previous_line is None:
1686 # Don't insert empty lines before the first line in the file.
1689 if self.previous_line.is_decorator:
1692 if self.previous_line.depth < current_line.depth and (
1693 self.previous_line.is_class or self.previous_line.is_def
1698 self.previous_line.is_comment
1699 and self.previous_line.depth == current_line.depth
1705 if self.previous_line.depth > current_line.depth:
1707 elif current_line.is_class or self.previous_line.is_class:
1708 if current_line.is_stub_class and self.previous_line.is_stub_class:
1709 # No blank line between classes with an empty body
1713 elif current_line.is_def and not self.previous_line.is_def:
1714 # Blank line between a block of functions and a block of non-functions
1720 if current_line.depth and newlines:
1726 class LineGenerator(Visitor[Line]):
1727 """Generates reformatted Line objects. Empty lines are not emitted.
1729 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1730 in ways that will no longer stringify to valid Python code on the tree.
1733 is_pyi: bool = False
1734 normalize_strings: bool = True
1735 current_line: Line = field(default_factory=Line)
1736 remove_u_prefix: bool = False
1738 def line(self, indent: int = 0) -> Iterator[Line]:
1741 If the line is empty, only emit if it makes sense.
1742 If the line is too long, split it first and then generate.
1744 If any lines were generated, set up a new current_line.
1746 if not self.current_line:
1747 self.current_line.depth += indent
1748 return # Line is empty, don't emit. Creating a new one unnecessary.
1750 complete_line = self.current_line
1751 self.current_line = Line(depth=complete_line.depth + indent)
1754 def visit_default(self, node: LN) -> Iterator[Line]:
1755 """Default `visit_*()` implementation. Recurses to children of `node`."""
1756 if isinstance(node, Leaf):
1757 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1758 for comment in generate_comments(node):
1759 if any_open_brackets:
1760 # any comment within brackets is subject to splitting
1761 self.current_line.append(comment)
1762 elif comment.type == token.COMMENT:
1763 # regular trailing comment
1764 self.current_line.append(comment)
1765 yield from self.line()
1768 # regular standalone comment
1769 yield from self.line()
1771 self.current_line.append(comment)
1772 yield from self.line()
1774 normalize_prefix(node, inside_brackets=any_open_brackets)
1775 if self.normalize_strings and node.type == token.STRING:
1776 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1777 normalize_string_quotes(node)
1778 if node.type == token.NUMBER:
1779 normalize_numeric_literal(node)
1780 if node.type not in WHITESPACE:
1781 self.current_line.append(node)
1782 yield from super().visit_default(node)
1784 def visit_INDENT(self, node: Leaf) -> Iterator[Line]:
1785 """Increase indentation level, maybe yield a line."""
1786 # In blib2to3 INDENT never holds comments.
1787 yield from self.line(+1)
1788 yield from self.visit_default(node)
1790 def visit_DEDENT(self, node: Leaf) -> Iterator[Line]:
1791 """Decrease indentation level, maybe yield a line."""
1792 # The current line might still wait for trailing comments. At DEDENT time
1793 # there won't be any (they would be prefixes on the preceding NEWLINE).
1794 # Emit the line then.
1795 yield from self.line()
1797 # While DEDENT has no value, its prefix may contain standalone comments
1798 # that belong to the current indentation level. Get 'em.
1799 yield from self.visit_default(node)
1801 # Finally, emit the dedent.
1802 yield from self.line(-1)
1805 self, node: Node, keywords: Set[str], parens: Set[str]
1806 ) -> Iterator[Line]:
1807 """Visit a statement.
1809 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1810 `def`, `with`, `class`, `assert` and assignments.
1812 The relevant Python language `keywords` for a given statement will be
1813 NAME leaves within it. This methods puts those on a separate line.
1815 `parens` holds a set of string leaf values immediately after which
1816 invisible parens should be put.
1818 normalize_invisible_parens(node, parens_after=parens)
1819 for child in node.children:
1820 if child.type == token.NAME and child.value in keywords: # type: ignore
1821 yield from self.line()
1823 yield from self.visit(child)
1825 def visit_suite(self, node: Node) -> Iterator[Line]:
1826 """Visit a suite."""
1827 if self.is_pyi and is_stub_suite(node):
1828 yield from self.visit(node.children[2])
1830 yield from self.visit_default(node)
1832 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1833 """Visit a statement without nested statements."""
1834 is_suite_like = node.parent and node.parent.type in STATEMENT
1836 if self.is_pyi and is_stub_body(node):
1837 yield from self.visit_default(node)
1839 yield from self.line(+1)
1840 yield from self.visit_default(node)
1841 yield from self.line(-1)
1844 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1845 yield from self.line()
1846 yield from self.visit_default(node)
1848 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1849 """Visit `async def`, `async for`, `async with`."""
1850 yield from self.line()
1852 children = iter(node.children)
1853 for child in children:
1854 yield from self.visit(child)
1856 if child.type == token.ASYNC:
1859 internal_stmt = next(children)
1860 for child in internal_stmt.children:
1861 yield from self.visit(child)
1863 def visit_decorators(self, node: Node) -> Iterator[Line]:
1864 """Visit decorators."""
1865 for child in node.children:
1866 yield from self.line()
1867 yield from self.visit(child)
1869 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1870 """Remove a semicolon and put the other statement on a separate line."""
1871 yield from self.line()
1873 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1874 """End of file. Process outstanding comments and end with a newline."""
1875 yield from self.visit_default(leaf)
1876 yield from self.line()
1878 def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]:
1879 if not self.current_line.bracket_tracker.any_open_brackets():
1880 yield from self.line()
1881 yield from self.visit_default(leaf)
1883 def visit_factor(self, node: Node) -> Iterator[Line]:
1884 """Force parentheses between a unary op and a binary power:
1886 -2 ** 8 -> -(2 ** 8)
1888 _operator, operand = node.children
1890 operand.type == syms.power
1891 and len(operand.children) == 3
1892 and operand.children[1].type == token.DOUBLESTAR
1894 lpar = Leaf(token.LPAR, "(")
1895 rpar = Leaf(token.RPAR, ")")
1896 index = operand.remove() or 0
1897 node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
1898 yield from self.visit_default(node)
1900 def __post_init__(self) -> None:
1901 """You are in a twisty little maze of passages."""
1904 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1905 self.visit_if_stmt = partial(
1906 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1908 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1909 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1910 self.visit_try_stmt = partial(
1911 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1913 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1914 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1915 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1916 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1917 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1918 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1919 self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
1920 self.visit_del_stmt = partial(v, keywords=Ø, parens={"del"})
1921 self.visit_async_funcdef = self.visit_async_stmt
1922 self.visit_decorated = self.visit_decorators
1925 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1926 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1927 OPENING_BRACKETS = set(BRACKET.keys())
1928 CLOSING_BRACKETS = set(BRACKET.values())
1929 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1930 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1933 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901
1934 """Return whitespace prefix if needed for the given `leaf`.
1936 `complex_subscript` signals whether the given leaf is part of a subscription
1937 which has non-trivial arguments, like arithmetic expressions or function calls.
1945 if t in ALWAYS_NO_SPACE:
1948 if t == token.COMMENT:
1951 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1952 if t == token.COLON and p.type not in {
1959 prev = leaf.prev_sibling
1961 prevp = preceding_leaf(p)
1962 if not prevp or prevp.type in OPENING_BRACKETS:
1965 if t == token.COLON:
1966 if prevp.type == token.COLON:
1969 elif prevp.type != token.COMMA and not complex_subscript:
1974 if prevp.type == token.EQUAL:
1976 if prevp.parent.type in {
1984 elif prevp.parent.type == syms.typedargslist:
1985 # A bit hacky: if the equal sign has whitespace, it means we
1986 # previously found it's a typed argument. So, we're using
1990 elif prevp.type in VARARGS_SPECIALS:
1991 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1994 elif prevp.type == token.COLON:
1995 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1996 return SPACE if complex_subscript else NO
2000 and prevp.parent.type == syms.factor
2001 and prevp.type in MATH_OPERATORS
2006 prevp.type == token.RIGHTSHIFT
2008 and prevp.parent.type == syms.shift_expr
2009 and prevp.prev_sibling
2010 and prevp.prev_sibling.type == token.NAME
2011 and prevp.prev_sibling.value == "print" # type: ignore
2013 # Python 2 print chevron
2016 elif prev.type in OPENING_BRACKETS:
2019 if p.type in {syms.parameters, syms.arglist}:
2020 # untyped function signatures or calls
2021 if not prev or prev.type != token.COMMA:
2024 elif p.type == syms.varargslist:
2026 if prev and prev.type != token.COMMA:
2029 elif p.type == syms.typedargslist:
2030 # typed function signatures
2034 if t == token.EQUAL:
2035 if prev.type != syms.tname:
2038 elif prev.type == token.EQUAL:
2039 # A bit hacky: if the equal sign has whitespace, it means we
2040 # previously found it's a typed argument. So, we're using that, too.
2043 elif prev.type != token.COMMA:
2046 elif p.type == syms.tname:
2049 prevp = preceding_leaf(p)
2050 if not prevp or prevp.type != token.COMMA:
2053 elif p.type == syms.trailer:
2054 # attributes and calls
2055 if t == token.LPAR or t == token.RPAR:
2060 prevp = preceding_leaf(p)
2061 if not prevp or prevp.type != token.NUMBER:
2064 elif t == token.LSQB:
2067 elif prev.type != token.COMMA:
2070 elif p.type == syms.argument:
2072 if t == token.EQUAL:
2076 prevp = preceding_leaf(p)
2077 if not prevp or prevp.type == token.LPAR:
2080 elif prev.type in {token.EQUAL} | VARARGS_SPECIALS:
2083 elif p.type == syms.decorator:
2087 elif p.type == syms.dotted_name:
2091 prevp = preceding_leaf(p)
2092 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
2095 elif p.type == syms.classdef:
2099 if prev and prev.type == token.LPAR:
2102 elif p.type in {syms.subscript, syms.sliceop}:
2105 assert p.parent is not None, "subscripts are always parented"
2106 if p.parent.type == syms.subscriptlist:
2111 elif not complex_subscript:
2114 elif p.type == syms.atom:
2115 if prev and t == token.DOT:
2116 # dots, but not the first one.
2119 elif p.type == syms.dictsetmaker:
2121 if prev and prev.type == token.DOUBLESTAR:
2124 elif p.type in {syms.factor, syms.star_expr}:
2127 prevp = preceding_leaf(p)
2128 if not prevp or prevp.type in OPENING_BRACKETS:
2131 prevp_parent = prevp.parent
2132 assert prevp_parent is not None
2133 if prevp.type == token.COLON and prevp_parent.type in {
2139 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
2142 elif t in {token.NAME, token.NUMBER, token.STRING}:
2145 elif p.type == syms.import_from:
2147 if prev and prev.type == token.DOT:
2150 elif t == token.NAME:
2154 if prev and prev.type == token.DOT:
2157 elif p.type == syms.sliceop:
2163 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
2164 """Return the first leaf that precedes `node`, if any."""
2166 res = node.prev_sibling
2168 if isinstance(res, Leaf):
2172 return list(res.leaves())[-1]
2181 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
2182 """Return the child of `ancestor` that contains `descendant`."""
2183 node: Optional[LN] = descendant
2184 while node and node.parent != ancestor:
2189 def container_of(leaf: Leaf) -> LN:
2190 """Return `leaf` or one of its ancestors that is the topmost container of it.
2192 By "container" we mean a node where `leaf` is the very first child.
2194 same_prefix = leaf.prefix
2195 container: LN = leaf
2197 parent = container.parent
2201 if parent.children[0].prefix != same_prefix:
2204 if parent.type == syms.file_input:
2207 if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS:
2214 def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2215 """Return the priority of the `leaf` delimiter, given a line break after it.
2217 The delimiter priorities returned here are from those delimiters that would
2218 cause a line break after themselves.
2220 Higher numbers are higher priority.
2222 if leaf.type == token.COMMA:
2223 return COMMA_PRIORITY
2228 def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2229 """Return the priority of the `leaf` delimiter, given a line break before it.
2231 The delimiter priorities returned here are from those delimiters that would
2232 cause a line break before themselves.
2234 Higher numbers are higher priority.
2236 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
2237 # * and ** might also be MATH_OPERATORS but in this case they are not.
2238 # Don't treat them as a delimiter.
2242 leaf.type == token.DOT
2244 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
2245 and (previous is None or previous.type in CLOSING_BRACKETS)
2250 leaf.type in MATH_OPERATORS
2252 and leaf.parent.type not in {syms.factor, syms.star_expr}
2254 return MATH_PRIORITIES[leaf.type]
2256 if leaf.type in COMPARATORS:
2257 return COMPARATOR_PRIORITY
2260 leaf.type == token.STRING
2261 and previous is not None
2262 and previous.type == token.STRING
2264 return STRING_PRIORITY
2266 if leaf.type not in {token.NAME, token.ASYNC}:
2272 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
2273 or leaf.type == token.ASYNC
2276 not isinstance(leaf.prev_sibling, Leaf)
2277 or leaf.prev_sibling.value != "async"
2279 return COMPREHENSION_PRIORITY
2284 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
2286 return COMPREHENSION_PRIORITY
2288 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
2289 return TERNARY_PRIORITY
2291 if leaf.value == "is":
2292 return COMPARATOR_PRIORITY
2297 and leaf.parent.type in {syms.comp_op, syms.comparison}
2299 previous is not None
2300 and previous.type == token.NAME
2301 and previous.value == "not"
2304 return COMPARATOR_PRIORITY
2309 and leaf.parent.type == syms.comp_op
2311 previous is not None
2312 and previous.type == token.NAME
2313 and previous.value == "is"
2316 return COMPARATOR_PRIORITY
2318 if leaf.value in LOGIC_OPERATORS and leaf.parent:
2319 return LOGIC_PRIORITY
2324 FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"}
2325 FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"}
2328 def generate_comments(leaf: LN) -> Iterator[Leaf]:
2329 """Clean the prefix of the `leaf` and generate comments from it, if any.
2331 Comments in lib2to3 are shoved into the whitespace prefix. This happens
2332 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
2333 move because it does away with modifying the grammar to include all the
2334 possible places in which comments can be placed.
2336 The sad consequence for us though is that comments don't "belong" anywhere.
2337 This is why this function generates simple parentless Leaf objects for
2338 comments. We simply don't know what the correct parent should be.
2340 No matter though, we can live without this. We really only need to
2341 differentiate between inline and standalone comments. The latter don't
2342 share the line with any code.
2344 Inline comments are emitted as regular token.COMMENT leaves. Standalone
2345 are emitted with a fake STANDALONE_COMMENT token identifier.
2347 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
2348 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
2353 """Describes a piece of syntax that is a comment.
2355 It's not a :class:`blib2to3.pytree.Leaf` so that:
2357 * it can be cached (`Leaf` objects should not be reused more than once as
2358 they store their lineno, column, prefix, and parent information);
2359 * `newlines` and `consumed` fields are kept separate from the `value`. This
2360 simplifies handling of special marker comments like ``# fmt: off/on``.
2363 type: int # token.COMMENT or STANDALONE_COMMENT
2364 value: str # content of the comment
2365 newlines: int # how many newlines before the comment
2366 consumed: int # how many characters of the original leaf's prefix did we consume
2369 @lru_cache(maxsize=4096)
2370 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
2371 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
2372 result: List[ProtoComment] = []
2373 if not prefix or "#" not in prefix:
2379 for index, line in enumerate(prefix.split("\n")):
2380 consumed += len(line) + 1 # adding the length of the split '\n'
2381 line = line.lstrip()
2384 if not line.startswith("#"):
2385 # Escaped newlines outside of a comment are not really newlines at
2386 # all. We treat a single-line comment following an escaped newline
2387 # as a simple trailing comment.
2388 if line.endswith("\\"):
2392 if index == ignored_lines and not is_endmarker:
2393 comment_type = token.COMMENT # simple trailing comment
2395 comment_type = STANDALONE_COMMENT
2396 comment = make_comment(line)
2399 type=comment_type, value=comment, newlines=nlines, consumed=consumed
2406 def make_comment(content: str) -> str:
2407 """Return a consistently formatted comment from the given `content` string.
2409 All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single
2410 space between the hash sign and the content.
2412 If `content` didn't start with a hash sign, one is provided.
2414 content = content.rstrip()
2418 if content[0] == "#":
2419 content = content[1:]
2420 if content and content[0] not in " !:#'%":
2421 content = " " + content
2422 return "#" + content
2428 inner: bool = False,
2429 features: Collection[Feature] = (),
2430 ) -> Iterator[Line]:
2431 """Split a `line` into potentially many lines.
2433 They should fit in the allotted `line_length` but might not be able to.
2434 `inner` signifies that there were a pair of brackets somewhere around the
2435 current `line`, possibly transitively. This means we can fallback to splitting
2436 by delimiters if the LHS/RHS don't yield any results.
2438 `features` are syntactical features that may be used in the output.
2444 line_str = str(line).strip("\n")
2447 not line.contains_uncollapsable_type_comments()
2448 and not line.should_explode
2449 and not line.is_collection_with_optional_trailing_comma
2451 is_line_short_enough(line, line_length=line_length, line_str=line_str)
2452 or line.contains_unsplittable_type_ignore()
2458 split_funcs: List[SplitFunc]
2460 split_funcs = [left_hand_split]
2463 def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]:
2464 for omit in generate_trailers_to_omit(line, line_length):
2465 lines = list(right_hand_split(line, line_length, features, omit=omit))
2466 if is_line_short_enough(lines[0], line_length=line_length):
2470 # All splits failed, best effort split with no omits.
2471 # This mostly happens to multiline strings that are by definition
2472 # reported as not fitting a single line.
2473 # line_length=1 here was historically a bug that somehow became a feature.
2474 # See #762 and #781 for the full story.
2475 yield from right_hand_split(line, line_length=1, features=features)
2477 if line.inside_brackets:
2478 split_funcs = [delimiter_split, standalone_comment_split, rhs]
2481 for split_func in split_funcs:
2482 # We are accumulating lines in `result` because we might want to abort
2483 # mission and return the original line in the end, or attempt a different
2485 result: List[Line] = []
2487 for l in split_func(line, features):
2488 if str(l).strip("\n") == line_str:
2489 raise CannotSplit("Split function returned an unchanged result")
2493 l, line_length=line_length, inner=True, features=features
2507 def left_hand_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2508 """Split line into many lines, starting with the first matching bracket pair.
2510 Note: this usually looks weird, only use this for function definitions.
2511 Prefer RHS otherwise. This is why this function is not symmetrical with
2512 :func:`right_hand_split` which also handles optional parentheses.
2514 tail_leaves: List[Leaf] = []
2515 body_leaves: List[Leaf] = []
2516 head_leaves: List[Leaf] = []
2517 current_leaves = head_leaves
2518 matching_bracket: Optional[Leaf] = None
2519 for leaf in line.leaves:
2521 current_leaves is body_leaves
2522 and leaf.type in CLOSING_BRACKETS
2523 and leaf.opening_bracket is matching_bracket
2525 current_leaves = tail_leaves if body_leaves else head_leaves
2526 current_leaves.append(leaf)
2527 if current_leaves is head_leaves:
2528 if leaf.type in OPENING_BRACKETS:
2529 matching_bracket = leaf
2530 current_leaves = body_leaves
2531 if not matching_bracket:
2532 raise CannotSplit("No brackets found")
2534 head = bracket_split_build_line(head_leaves, line, matching_bracket)
2535 body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True)
2536 tail = bracket_split_build_line(tail_leaves, line, matching_bracket)
2537 bracket_split_succeeded_or_raise(head, body, tail)
2538 for result in (head, body, tail):
2543 def right_hand_split(
2546 features: Collection[Feature] = (),
2547 omit: Collection[LeafID] = (),
2548 ) -> Iterator[Line]:
2549 """Split line into many lines, starting with the last matching bracket pair.
2551 If the split was by optional parentheses, attempt splitting without them, too.
2552 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2555 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2557 tail_leaves: List[Leaf] = []
2558 body_leaves: List[Leaf] = []
2559 head_leaves: List[Leaf] = []
2560 current_leaves = tail_leaves
2561 opening_bracket: Optional[Leaf] = None
2562 closing_bracket: Optional[Leaf] = None
2563 for leaf in reversed(line.leaves):
2564 if current_leaves is body_leaves:
2565 if leaf is opening_bracket:
2566 current_leaves = head_leaves if body_leaves else tail_leaves
2567 current_leaves.append(leaf)
2568 if current_leaves is tail_leaves:
2569 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2570 opening_bracket = leaf.opening_bracket
2571 closing_bracket = leaf
2572 current_leaves = body_leaves
2573 if not (opening_bracket and closing_bracket and head_leaves):
2574 # If there is no opening or closing_bracket that means the split failed and
2575 # all content is in the tail. Otherwise, if `head_leaves` are empty, it means
2576 # the matching `opening_bracket` wasn't available on `line` anymore.
2577 raise CannotSplit("No brackets found")
2579 tail_leaves.reverse()
2580 body_leaves.reverse()
2581 head_leaves.reverse()
2582 head = bracket_split_build_line(head_leaves, line, opening_bracket)
2583 body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True)
2584 tail = bracket_split_build_line(tail_leaves, line, opening_bracket)
2585 bracket_split_succeeded_or_raise(head, body, tail)
2587 # the body shouldn't be exploded
2588 not body.should_explode
2589 # the opening bracket is an optional paren
2590 and opening_bracket.type == token.LPAR
2591 and not opening_bracket.value
2592 # the closing bracket is an optional paren
2593 and closing_bracket.type == token.RPAR
2594 and not closing_bracket.value
2595 # it's not an import (optional parens are the only thing we can split on
2596 # in this case; attempting a split without them is a waste of time)
2597 and not line.is_import
2598 # there are no standalone comments in the body
2599 and not body.contains_standalone_comments(0)
2600 # and we can actually remove the parens
2601 and can_omit_invisible_parens(body, line_length)
2603 omit = {id(closing_bracket), *omit}
2605 yield from right_hand_split(line, line_length, features=features, omit=omit)
2611 or is_line_short_enough(body, line_length=line_length)
2614 "Splitting failed, body is still too long and can't be split."
2617 elif head.contains_multiline_strings() or tail.contains_multiline_strings():
2619 "The current optional pair of parentheses is bound to fail to "
2620 "satisfy the splitting algorithm because the head or the tail "
2621 "contains multiline strings which by definition never fit one "
2625 ensure_visible(opening_bracket)
2626 ensure_visible(closing_bracket)
2627 for result in (head, body, tail):
2632 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2633 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2635 Do nothing otherwise.
2637 A left- or right-hand split is based on a pair of brackets. Content before
2638 (and including) the opening bracket is left on one line, content inside the
2639 brackets is put on a separate line, and finally content starting with and
2640 following the closing bracket is put on a separate line.
2642 Those are called `head`, `body`, and `tail`, respectively. If the split
2643 produced the same line (all content in `head`) or ended up with an empty `body`
2644 and the `tail` is just the closing bracket, then it's considered failed.
2646 tail_len = len(str(tail).strip())
2649 raise CannotSplit("Splitting brackets produced the same line")
2653 f"Splitting brackets on an empty body to save "
2654 f"{tail_len} characters is not worth it"
2658 def bracket_split_build_line(
2659 leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False
2661 """Return a new line with given `leaves` and respective comments from `original`.
2663 If `is_body` is True, the result line is one-indented inside brackets and as such
2664 has its first leaf's prefix normalized and a trailing comma added when expected.
2666 result = Line(depth=original.depth)
2668 result.inside_brackets = True
2671 # Since body is a new indent level, remove spurious leading whitespace.
2672 normalize_prefix(leaves[0], inside_brackets=True)
2673 # Ensure a trailing comma for imports and standalone function arguments, but
2674 # be careful not to add one after any comments or within type annotations.
2677 and opening_bracket.value == "("
2678 and not any(l.type == token.COMMA for l in leaves)
2681 if original.is_import or no_commas:
2682 for i in range(len(leaves) - 1, -1, -1):
2683 if leaves[i].type == STANDALONE_COMMENT:
2686 if leaves[i].type != token.COMMA:
2687 leaves.insert(i + 1, Leaf(token.COMMA, ","))
2692 result.append(leaf, preformatted=True)
2693 for comment_after in original.comments_after(leaf):
2694 result.append(comment_after, preformatted=True)
2696 result.should_explode = should_explode(result, opening_bracket)
2700 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2701 """Normalize prefix of the first leaf in every line returned by `split_func`.
2703 This is a decorator over relevant split functions.
2707 def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2708 for l in split_func(line, features):
2709 normalize_prefix(l.leaves[0], inside_brackets=True)
2712 return split_wrapper
2715 @dont_increase_indentation
2716 def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2717 """Split according to delimiters of the highest priority.
2719 If the appropriate Features are given, the split will add trailing commas
2720 also in function signatures and calls that contain `*` and `**`.
2723 last_leaf = line.leaves[-1]
2725 raise CannotSplit("Line empty")
2727 bt = line.bracket_tracker
2729 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2731 raise CannotSplit("No delimiters found")
2733 if delimiter_priority == DOT_PRIORITY:
2734 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2735 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2737 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2738 lowest_depth = sys.maxsize
2739 trailing_comma_safe = True
2741 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2742 """Append `leaf` to current line or to new line if appending impossible."""
2743 nonlocal current_line
2745 current_line.append_safe(leaf, preformatted=True)
2749 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2750 current_line.append(leaf)
2752 for leaf in line.leaves:
2753 yield from append_to_line(leaf)
2755 for comment_after in line.comments_after(leaf):
2756 yield from append_to_line(comment_after)
2758 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2759 if leaf.bracket_depth == lowest_depth:
2760 if is_vararg(leaf, within={syms.typedargslist}):
2761 trailing_comma_safe = (
2762 trailing_comma_safe and Feature.TRAILING_COMMA_IN_DEF in features
2764 elif is_vararg(leaf, within={syms.arglist, syms.argument}):
2765 trailing_comma_safe = (
2766 trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features
2769 leaf_priority = bt.delimiters.get(id(leaf))
2770 if leaf_priority == delimiter_priority:
2773 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2777 and delimiter_priority == COMMA_PRIORITY
2778 and current_line.leaves[-1].type != token.COMMA
2779 and current_line.leaves[-1].type != STANDALONE_COMMENT
2781 current_line.append(Leaf(token.COMMA, ","))
2785 @dont_increase_indentation
2786 def standalone_comment_split(
2787 line: Line, features: Collection[Feature] = ()
2788 ) -> Iterator[Line]:
2789 """Split standalone comments from the rest of the line."""
2790 if not line.contains_standalone_comments(0):
2791 raise CannotSplit("Line does not have any standalone comments")
2793 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2795 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2796 """Append `leaf` to current line or to new line if appending impossible."""
2797 nonlocal current_line
2799 current_line.append_safe(leaf, preformatted=True)
2803 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2804 current_line.append(leaf)
2806 for leaf in line.leaves:
2807 yield from append_to_line(leaf)
2809 for comment_after in line.comments_after(leaf):
2810 yield from append_to_line(comment_after)
2816 def is_import(leaf: Leaf) -> bool:
2817 """Return True if the given leaf starts an import statement."""
2824 (v == "import" and p and p.type == syms.import_name)
2825 or (v == "from" and p and p.type == syms.import_from)
2830 def is_type_comment(leaf: Leaf, suffix: str = "") -> bool:
2831 """Return True if the given leaf is a special comment.
2832 Only returns true for type comments for now."""
2835 return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix)
2838 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2839 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2842 Note: don't use backslashes for formatting or you'll lose your voting rights.
2844 if not inside_brackets:
2845 spl = leaf.prefix.split("#")
2846 if "\\" not in spl[0]:
2847 nl_count = spl[-1].count("\n")
2850 leaf.prefix = "\n" * nl_count
2856 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2857 """Make all string prefixes lowercase.
2859 If remove_u_prefix is given, also removes any u prefix from the string.
2861 Note: Mutates its argument.
2863 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2864 assert match is not None, f"failed to match string {leaf.value!r}"
2865 orig_prefix = match.group(1)
2866 new_prefix = orig_prefix.replace("F", "f").replace("B", "b").replace("U", "u")
2868 new_prefix = new_prefix.replace("u", "")
2869 leaf.value = f"{new_prefix}{match.group(2)}"
2872 def normalize_string_quotes(leaf: Leaf) -> None:
2873 """Prefer double quotes but only if it doesn't cause more escaping.
2875 Adds or removes backslashes as appropriate. Doesn't parse and fix
2876 strings nested in f-strings (yet).
2878 Note: Mutates its argument.
2880 value = leaf.value.lstrip("furbFURB")
2881 if value[:3] == '"""':
2884 elif value[:3] == "'''":
2887 elif value[0] == '"':
2893 first_quote_pos = leaf.value.find(orig_quote)
2894 if first_quote_pos == -1:
2895 return # There's an internal error
2897 prefix = leaf.value[:first_quote_pos]
2898 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2899 escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}")
2900 escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}")
2901 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2902 if "r" in prefix.casefold():
2903 if unescaped_new_quote.search(body):
2904 # There's at least one unescaped new_quote in this raw string
2905 # so converting is impossible
2908 # Do not introduce or remove backslashes in raw strings
2911 # remove unnecessary escapes
2912 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2913 if body != new_body:
2914 # Consider the string without unnecessary escapes as the original
2916 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2917 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2918 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2919 if "f" in prefix.casefold():
2920 matches = re.findall(
2922 (?:[^{]|^)\{ # start of the string or a non-{ followed by a single {
2923 ([^{].*?) # contents of the brackets except if begins with {{
2924 \}(?:[^}]|$) # A } followed by end of the string or a non-}
2931 # Do not introduce backslashes in interpolated expressions
2934 if new_quote == '"""' and new_body[-1:] == '"':
2936 new_body = new_body[:-1] + '\\"'
2937 orig_escape_count = body.count("\\")
2938 new_escape_count = new_body.count("\\")
2939 if new_escape_count > orig_escape_count:
2940 return # Do not introduce more escaping
2942 if new_escape_count == orig_escape_count and orig_quote == '"':
2943 return # Prefer double quotes
2945 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2948 def normalize_numeric_literal(leaf: Leaf) -> None:
2949 """Normalizes numeric (float, int, and complex) literals.
2951 All letters used in the representation are normalized to lowercase (except
2952 in Python 2 long literals).
2954 text = leaf.value.lower()
2955 if text.startswith(("0o", "0b")):
2956 # Leave octal and binary literals alone.
2958 elif text.startswith("0x"):
2959 # Change hex literals to upper case.
2960 before, after = text[:2], text[2:]
2961 text = f"{before}{after.upper()}"
2963 before, after = text.split("e")
2965 if after.startswith("-"):
2968 elif after.startswith("+"):
2970 before = format_float_or_int_string(before)
2971 text = f"{before}e{sign}{after}"
2972 elif text.endswith(("j", "l")):
2975 # Capitalize in "2L" because "l" looks too similar to "1".
2978 text = f"{format_float_or_int_string(number)}{suffix}"
2980 text = format_float_or_int_string(text)
2984 def format_float_or_int_string(text: str) -> str:
2985 """Formats a float string like "1.0"."""
2989 before, after = text.split(".")
2990 return f"{before or 0}.{after or 0}"
2993 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2994 """Make existing optional parentheses invisible or create new ones.
2996 `parens_after` is a set of string leaf values immediately after which parens
2999 Standardizes on visible parentheses for single-element tuples, and keeps
3000 existing visible parentheses for other tuples and generator expressions.
3002 for pc in list_comments(node.prefix, is_endmarker=False):
3003 if pc.value in FMT_OFF:
3004 # This `node` has a prefix with `# fmt: off`, don't mess with parens.
3007 for index, child in enumerate(list(node.children)):
3008 # Add parentheses around long tuple unpacking in assignments.
3011 and isinstance(child, Node)
3012 and child.type == syms.testlist_star_expr
3017 if is_walrus_assignment(child):
3020 if child.type == syms.atom:
3021 if maybe_make_parens_invisible_in_atom(child, parent=node):
3022 wrap_in_parentheses(node, child, visible=False)
3023 elif is_one_tuple(child):
3024 wrap_in_parentheses(node, child, visible=True)
3025 elif node.type == syms.import_from:
3026 # "import from" nodes store parentheses directly as part of
3028 if child.type == token.LPAR:
3029 # make parentheses invisible
3030 child.value = "" # type: ignore
3031 node.children[-1].value = "" # type: ignore
3032 elif child.type != token.STAR:
3033 # insert invisible parentheses
3034 node.insert_child(index, Leaf(token.LPAR, ""))
3035 node.append_child(Leaf(token.RPAR, ""))
3038 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
3039 wrap_in_parentheses(node, child, visible=False)
3041 check_lpar = isinstance(child, Leaf) and child.value in parens_after
3044 def normalize_fmt_off(node: Node) -> None:
3045 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
3048 try_again = convert_one_fmt_off_pair(node)
3051 def convert_one_fmt_off_pair(node: Node) -> bool:
3052 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
3054 Returns True if a pair was converted.
3056 for leaf in node.leaves():
3057 previous_consumed = 0
3058 for comment in list_comments(leaf.prefix, is_endmarker=False):
3059 if comment.value in FMT_OFF:
3060 # We only want standalone comments. If there's no previous leaf or
3061 # the previous leaf is indentation, it's a standalone comment in
3063 if comment.type != STANDALONE_COMMENT:
3064 prev = preceding_leaf(leaf)
3065 if prev and prev.type not in WHITESPACE:
3068 ignored_nodes = list(generate_ignored_nodes(leaf))
3069 if not ignored_nodes:
3072 first = ignored_nodes[0] # Can be a container node with the `leaf`.
3073 parent = first.parent
3074 prefix = first.prefix
3075 first.prefix = prefix[comment.consumed :]
3077 comment.value + "\n" + "".join(str(n) for n in ignored_nodes)
3079 if hidden_value.endswith("\n"):
3080 # That happens when one of the `ignored_nodes` ended with a NEWLINE
3081 # leaf (possibly followed by a DEDENT).
3082 hidden_value = hidden_value[:-1]
3083 first_idx: Optional[int] = None
3084 for ignored in ignored_nodes:
3085 index = ignored.remove()
3086 if first_idx is None:
3088 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
3089 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
3090 parent.insert_child(
3095 prefix=prefix[:previous_consumed] + "\n" * comment.newlines,
3100 previous_consumed = comment.consumed
3105 def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]:
3106 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
3108 Stops at the end of the block.
3110 container: Optional[LN] = container_of(leaf)
3111 while container is not None and container.type != token.ENDMARKER:
3113 for comment in list_comments(container.prefix, is_endmarker=False):
3114 if comment.value in FMT_ON:
3116 elif comment.value in FMT_OFF:
3123 container = container.next_sibling
3126 def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool:
3127 """If it's safe, make the parens in the atom `node` invisible, recursively.
3128 Additionally, remove repeated, adjacent invisible parens from the atom `node`
3129 as they are redundant.
3131 Returns whether the node should itself be wrapped in invisible parentheses.
3135 node.type != syms.atom
3136 or is_empty_tuple(node)
3137 or is_one_tuple(node)
3138 or (is_yield(node) and parent.type != syms.expr_stmt)
3139 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
3143 first = node.children[0]
3144 last = node.children[-1]
3145 if first.type == token.LPAR and last.type == token.RPAR:
3146 middle = node.children[1]
3147 # make parentheses invisible
3148 first.value = "" # type: ignore
3149 last.value = "" # type: ignore
3150 maybe_make_parens_invisible_in_atom(middle, parent=parent)
3152 if is_atom_with_invisible_parens(middle):
3153 # Strip the invisible parens from `middle` by replacing
3154 # it with the child in-between the invisible parens
3155 middle.replace(middle.children[1])
3162 def is_atom_with_invisible_parens(node: LN) -> bool:
3163 """Given a `LN`, determines whether it's an atom `node` with invisible
3164 parens. Useful in dedupe-ing and normalizing parens.
3166 if isinstance(node, Leaf) or node.type != syms.atom:
3169 first, last = node.children[0], node.children[-1]
3171 isinstance(first, Leaf)
3172 and first.type == token.LPAR
3173 and first.value == ""
3174 and isinstance(last, Leaf)
3175 and last.type == token.RPAR
3176 and last.value == ""
3180 def is_empty_tuple(node: LN) -> bool:
3181 """Return True if `node` holds an empty tuple."""
3183 node.type == syms.atom
3184 and len(node.children) == 2
3185 and node.children[0].type == token.LPAR
3186 and node.children[1].type == token.RPAR
3190 def unwrap_singleton_parenthesis(node: LN) -> Optional[LN]:
3191 """Returns `wrapped` if `node` is of the shape ( wrapped ).
3193 Parenthesis can be optional. Returns None otherwise"""
3194 if len(node.children) != 3:
3197 lpar, wrapped, rpar = node.children
3198 if not (lpar.type == token.LPAR and rpar.type == token.RPAR):
3204 def wrap_in_parentheses(parent: Node, child: LN, *, visible: bool = True) -> None:
3205 """Wrap `child` in parentheses.
3207 This replaces `child` with an atom holding the parentheses and the old
3208 child. That requires moving the prefix.
3210 If `visible` is False, the leaves will be valueless (and thus invisible).
3212 lpar = Leaf(token.LPAR, "(" if visible else "")
3213 rpar = Leaf(token.RPAR, ")" if visible else "")
3214 prefix = child.prefix
3216 index = child.remove() or 0
3217 new_child = Node(syms.atom, [lpar, child, rpar])
3218 new_child.prefix = prefix
3219 parent.insert_child(index, new_child)
3222 def is_one_tuple(node: LN) -> bool:
3223 """Return True if `node` holds a tuple with one element, with or without parens."""
3224 if node.type == syms.atom:
3225 gexp = unwrap_singleton_parenthesis(node)
3226 if gexp is None or gexp.type != syms.testlist_gexp:
3229 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
3232 node.type in IMPLICIT_TUPLE
3233 and len(node.children) == 2
3234 and node.children[1].type == token.COMMA
3238 def is_walrus_assignment(node: LN) -> bool:
3239 """Return True iff `node` is of the shape ( test := test )"""
3240 inner = unwrap_singleton_parenthesis(node)
3241 return inner is not None and inner.type == syms.namedexpr_test
3244 def is_yield(node: LN) -> bool:
3245 """Return True if `node` holds a `yield` or `yield from` expression."""
3246 if node.type == syms.yield_expr:
3249 if node.type == token.NAME and node.value == "yield": # type: ignore
3252 if node.type != syms.atom:
3255 if len(node.children) != 3:
3258 lpar, expr, rpar = node.children
3259 if lpar.type == token.LPAR and rpar.type == token.RPAR:
3260 return is_yield(expr)
3265 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
3266 """Return True if `leaf` is a star or double star in a vararg or kwarg.
3268 If `within` includes VARARGS_PARENTS, this applies to function signatures.
3269 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
3270 extended iterable unpacking (PEP 3132) and additional unpacking
3271 generalizations (PEP 448).
3273 if leaf.type not in VARARGS_SPECIALS or not leaf.parent:
3277 if p.type == syms.star_expr:
3278 # Star expressions are also used as assignment targets in extended
3279 # iterable unpacking (PEP 3132). See what its parent is instead.
3285 return p.type in within
3288 def is_multiline_string(leaf: Leaf) -> bool:
3289 """Return True if `leaf` is a multiline string that actually spans many lines."""
3290 value = leaf.value.lstrip("furbFURB")
3291 return value[:3] in {'"""', "'''"} and "\n" in value
3294 def is_stub_suite(node: Node) -> bool:
3295 """Return True if `node` is a suite with a stub body."""
3297 len(node.children) != 4
3298 or node.children[0].type != token.NEWLINE
3299 or node.children[1].type != token.INDENT
3300 or node.children[3].type != token.DEDENT
3304 return is_stub_body(node.children[2])
3307 def is_stub_body(node: LN) -> bool:
3308 """Return True if `node` is a simple statement containing an ellipsis."""
3309 if not isinstance(node, Node) or node.type != syms.simple_stmt:
3312 if len(node.children) != 2:
3315 child = node.children[0]
3317 child.type == syms.atom
3318 and len(child.children) == 3
3319 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
3323 def max_delimiter_priority_in_atom(node: LN) -> Priority:
3324 """Return maximum delimiter priority inside `node`.
3326 This is specific to atoms with contents contained in a pair of parentheses.
3327 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
3329 if node.type != syms.atom:
3332 first = node.children[0]
3333 last = node.children[-1]
3334 if not (first.type == token.LPAR and last.type == token.RPAR):
3337 bt = BracketTracker()
3338 for c in node.children[1:-1]:
3339 if isinstance(c, Leaf):
3342 for leaf in c.leaves():
3345 return bt.max_delimiter_priority()
3351 def ensure_visible(leaf: Leaf) -> None:
3352 """Make sure parentheses are visible.
3354 They could be invisible as part of some statements (see
3355 :func:`normalize_invisible_parens` and :func:`visit_import_from`).
3357 if leaf.type == token.LPAR:
3359 elif leaf.type == token.RPAR:
3363 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
3364 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
3367 opening_bracket.parent
3368 and opening_bracket.parent.type in {syms.atom, syms.import_from}
3369 and opening_bracket.value in "[{("
3374 last_leaf = line.leaves[-1]
3375 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
3376 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
3377 except (IndexError, ValueError):
3380 return max_priority == COMMA_PRIORITY
3383 def get_features_used(node: Node) -> Set[Feature]:
3384 """Return a set of (relatively) new Python features used in this file.
3386 Currently looking for:
3388 - underscores in numeric literals;
3389 - trailing commas after * or ** in function signatures and calls;
3390 - positional only arguments in function signatures and lambdas;
3392 features: Set[Feature] = set()
3393 for n in node.pre_order():
3394 if n.type == token.STRING:
3395 value_head = n.value[:2] # type: ignore
3396 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
3397 features.add(Feature.F_STRINGS)
3399 elif n.type == token.NUMBER:
3400 if "_" in n.value: # type: ignore
3401 features.add(Feature.NUMERIC_UNDERSCORES)
3403 elif n.type == token.SLASH:
3404 if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}:
3405 features.add(Feature.POS_ONLY_ARGUMENTS)
3407 elif n.type == token.COLONEQUAL:
3408 features.add(Feature.ASSIGNMENT_EXPRESSIONS)
3411 n.type in {syms.typedargslist, syms.arglist}
3413 and n.children[-1].type == token.COMMA
3415 if n.type == syms.typedargslist:
3416 feature = Feature.TRAILING_COMMA_IN_DEF
3418 feature = Feature.TRAILING_COMMA_IN_CALL
3420 for ch in n.children:
3421 if ch.type in STARS:
3422 features.add(feature)
3424 if ch.type == syms.argument:
3425 for argch in ch.children:
3426 if argch.type in STARS:
3427 features.add(feature)
3432 def detect_target_versions(node: Node) -> Set[TargetVersion]:
3433 """Detect the version to target based on the nodes used."""
3434 features = get_features_used(node)
3436 version for version in TargetVersion if features <= VERSION_TO_FEATURES[version]
3440 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
3441 """Generate sets of closing bracket IDs that should be omitted in a RHS.
3443 Brackets can be omitted if the entire trailer up to and including
3444 a preceding closing bracket fits in one line.
3446 Yielded sets are cumulative (contain results of previous yields, too). First
3450 omit: Set[LeafID] = set()
3453 length = 4 * line.depth
3454 opening_bracket: Optional[Leaf] = None
3455 closing_bracket: Optional[Leaf] = None
3456 inner_brackets: Set[LeafID] = set()
3457 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
3458 length += leaf_length
3459 if length > line_length:
3462 has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
3463 if leaf.type == STANDALONE_COMMENT or has_inline_comment:
3467 if leaf is opening_bracket:
3468 opening_bracket = None
3469 elif leaf.type in CLOSING_BRACKETS:
3470 inner_brackets.add(id(leaf))
3471 elif leaf.type in CLOSING_BRACKETS:
3472 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
3473 # Empty brackets would fail a split so treat them as "inner"
3474 # brackets (e.g. only add them to the `omit` set if another
3475 # pair of brackets was good enough.
3476 inner_brackets.add(id(leaf))
3480 omit.add(id(closing_bracket))
3481 omit.update(inner_brackets)
3482 inner_brackets.clear()
3486 opening_bracket = leaf.opening_bracket
3487 closing_bracket = leaf
3490 def get_future_imports(node: Node) -> Set[str]:
3491 """Return a set of __future__ imports in the file."""
3492 imports: Set[str] = set()
3494 def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]:
3495 for child in children:
3496 if isinstance(child, Leaf):
3497 if child.type == token.NAME:
3500 elif child.type == syms.import_as_name:
3501 orig_name = child.children[0]
3502 assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
3503 assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
3504 yield orig_name.value
3506 elif child.type == syms.import_as_names:
3507 yield from get_imports_from_children(child.children)
3510 raise AssertionError("Invalid syntax parsing imports")
3512 for child in node.children:
3513 if child.type != syms.simple_stmt:
3516 first_child = child.children[0]
3517 if isinstance(first_child, Leaf):
3518 # Continue looking if we see a docstring; otherwise stop.
3520 len(child.children) == 2
3521 and first_child.type == token.STRING
3522 and child.children[1].type == token.NEWLINE
3528 elif first_child.type == syms.import_from:
3529 module_name = first_child.children[1]
3530 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
3533 imports |= set(get_imports_from_children(first_child.children[3:]))
3541 def get_gitignore(root: Path) -> PathSpec:
3542 """ Return a PathSpec matching gitignore content if present."""
3543 gitignore = root / ".gitignore"
3544 lines: List[str] = []
3545 if gitignore.is_file():
3546 with gitignore.open() as gf:
3547 lines = gf.readlines()
3548 return PathSpec.from_lines("gitwildmatch", lines)
3551 def gen_python_files_in_dir(
3554 include: Pattern[str],
3555 exclude: Pattern[str],
3557 gitignore: PathSpec,
3558 ) -> Iterator[Path]:
3559 """Generate all files under `path` whose paths are not excluded by the
3560 `exclude` regex, but are included by the `include` regex.
3562 Symbolic links pointing outside of the `root` directory are ignored.
3564 `report` is where output about exclusions goes.
3566 assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
3567 for child in path.iterdir():
3568 # First ignore files matching .gitignore
3569 if gitignore.match_file(child.as_posix()):
3570 report.path_ignored(child, "matches the .gitignore file content")
3573 # Then ignore with `exclude` option.
3575 normalized_path = "/" + child.resolve().relative_to(root).as_posix()
3576 except OSError as e:
3577 report.path_ignored(child, f"cannot be read because {e}")
3581 if child.is_symlink():
3582 report.path_ignored(
3583 child, f"is a symbolic link that points outside {root}"
3590 normalized_path += "/"
3592 exclude_match = exclude.search(normalized_path)
3593 if exclude_match and exclude_match.group(0):
3594 report.path_ignored(child, "matches the --exclude regular expression")
3598 yield from gen_python_files_in_dir(
3599 child, root, include, exclude, report, gitignore
3602 elif child.is_file():
3603 include_match = include.search(normalized_path)
3609 def find_project_root(srcs: Iterable[str]) -> Path:
3610 """Return a directory containing .git, .hg, or pyproject.toml.
3612 That directory can be one of the directories passed in `srcs` or their
3615 If no directory in the tree contains a marker that would specify it's the
3616 project root, the root of the file system is returned.
3619 return Path("/").resolve()
3621 common_base = min(Path(src).resolve() for src in srcs)
3622 if common_base.is_dir():
3623 # Append a fake file so `parents` below returns `common_base_dir`, too.
3624 common_base /= "fake-file"
3625 for directory in common_base.parents:
3626 if (directory / ".git").exists():
3629 if (directory / ".hg").is_dir():
3632 if (directory / "pyproject.toml").is_file():
3640 """Provides a reformatting counter. Can be rendered with `str(report)`."""
3645 verbose: bool = False
3646 change_count: int = 0
3648 failure_count: int = 0
3650 def done(self, src: Path, changed: Changed) -> None:
3651 """Increment the counter for successful reformatting. Write out a message."""
3652 if changed is Changed.YES:
3653 reformatted = "would reformat" if self.check or self.diff else "reformatted"
3654 if self.verbose or not self.quiet:
3655 out(f"{reformatted} {src}")
3656 self.change_count += 1
3659 if changed is Changed.NO:
3660 msg = f"{src} already well formatted, good job."
3662 msg = f"{src} wasn't modified on disk since last run."
3663 out(msg, bold=False)
3664 self.same_count += 1
3666 def failed(self, src: Path, message: str) -> None:
3667 """Increment the counter for failed reformatting. Write out a message."""
3668 err(f"error: cannot format {src}: {message}")
3669 self.failure_count += 1
3671 def path_ignored(self, path: Path, message: str) -> None:
3673 out(f"{path} ignored: {message}", bold=False)
3676 def return_code(self) -> int:
3677 """Return the exit code that the app should use.
3679 This considers the current state of changed files and failures:
3680 - if there were any failures, return 123;
3681 - if any files were changed and --check is being used, return 1;
3682 - otherwise return 0.
3684 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
3685 # 126 we have special return codes reserved by the shell.
3686 if self.failure_count:
3689 elif self.change_count and self.check:
3694 def __str__(self) -> str:
3695 """Render a color report of the current state.
3697 Use `click.unstyle` to remove colors.
3699 if self.check or self.diff:
3700 reformatted = "would be reformatted"
3701 unchanged = "would be left unchanged"
3702 failed = "would fail to reformat"
3704 reformatted = "reformatted"
3705 unchanged = "left unchanged"
3706 failed = "failed to reformat"
3708 if self.change_count:
3709 s = "s" if self.change_count > 1 else ""
3711 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
3714 s = "s" if self.same_count > 1 else ""
3715 report.append(f"{self.same_count} file{s} {unchanged}")
3716 if self.failure_count:
3717 s = "s" if self.failure_count > 1 else ""
3719 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
3721 return ", ".join(report) + "."
3724 def parse_ast(src: str) -> Union[ast.AST, ast3.AST, ast27.AST]:
3725 filename = "<unknown>"
3726 if sys.version_info >= (3, 8):
3727 # TODO: support Python 4+ ;)
3728 for minor_version in range(sys.version_info[1], 4, -1):
3730 return ast.parse(src, filename, feature_version=(3, minor_version))
3734 for feature_version in (7, 6):
3736 return ast3.parse(src, filename, feature_version=feature_version)
3740 return ast27.parse(src)
3743 def _fixup_ast_constants(
3744 node: Union[ast.AST, ast3.AST, ast27.AST]
3745 ) -> Union[ast.AST, ast3.AST, ast27.AST]:
3746 """Map ast nodes deprecated in 3.8 to Constant."""
3747 if isinstance(node, (ast.Str, ast3.Str, ast27.Str, ast.Bytes, ast3.Bytes)):
3748 return ast.Constant(value=node.s)
3750 if isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
3751 return ast.Constant(value=node.n)
3753 if isinstance(node, (ast.NameConstant, ast3.NameConstant)):
3754 return ast.Constant(value=node.value)
3759 def assert_equivalent(src: str, dst: str) -> None:
3760 """Raise AssertionError if `src` and `dst` aren't equivalent."""
3762 def _v(node: Union[ast.AST, ast3.AST, ast27.AST], depth: int = 0) -> Iterator[str]:
3763 """Simple visitor generating strings to compare ASTs by content."""
3765 node = _fixup_ast_constants(node)
3767 yield f"{' ' * depth}{node.__class__.__name__}("
3769 for field in sorted(node._fields): # noqa: F402
3770 # TypeIgnore has only one field 'lineno' which breaks this comparison
3771 type_ignore_classes = (ast3.TypeIgnore, ast27.TypeIgnore)
3772 if sys.version_info >= (3, 8):
3773 type_ignore_classes += (ast.TypeIgnore,)
3774 if isinstance(node, type_ignore_classes):
3778 value = getattr(node, field)
3779 except AttributeError:
3782 yield f"{' ' * (depth+1)}{field}="
3784 if isinstance(value, list):
3786 # Ignore nested tuples within del statements, because we may insert
3787 # parentheses and they change the AST.
3790 and isinstance(node, (ast.Delete, ast3.Delete, ast27.Delete))
3791 and isinstance(item, (ast.Tuple, ast3.Tuple, ast27.Tuple))
3793 for item in item.elts:
3794 yield from _v(item, depth + 2)
3796 elif isinstance(item, (ast.AST, ast3.AST, ast27.AST)):
3797 yield from _v(item, depth + 2)
3799 elif isinstance(value, (ast.AST, ast3.AST, ast27.AST)):
3800 yield from _v(value, depth + 2)
3803 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
3805 yield f"{' ' * depth}) # /{node.__class__.__name__}"
3808 src_ast = parse_ast(src)
3809 except Exception as exc:
3810 raise AssertionError(
3811 f"cannot use --safe with this file; failed to parse source file. "
3812 f"AST error message: {exc}"
3816 dst_ast = parse_ast(dst)
3817 except Exception as exc:
3818 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
3819 raise AssertionError(
3820 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
3821 f"Please report a bug on https://github.com/psf/black/issues. "
3822 f"This invalid output might be helpful: {log}"
3825 src_ast_str = "\n".join(_v(src_ast))
3826 dst_ast_str = "\n".join(_v(dst_ast))
3827 if src_ast_str != dst_ast_str:
3828 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
3829 raise AssertionError(
3830 f"INTERNAL ERROR: Black produced code that is not equivalent to "
3832 f"Please report a bug on https://github.com/psf/black/issues. "
3833 f"This diff might be helpful: {log}"
3837 def assert_stable(src: str, dst: str, mode: Mode) -> None:
3838 """Raise AssertionError if `dst` reformats differently the second time."""
3839 newdst = format_str(dst, mode=mode)
3842 diff(src, dst, "source", "first pass"),
3843 diff(dst, newdst, "first pass", "second pass"),
3845 raise AssertionError(
3846 f"INTERNAL ERROR: Black produced different code on the second pass "
3847 f"of the formatter. "
3848 f"Please report a bug on https://github.com/psf/black/issues. "
3849 f"This diff might be helpful: {log}"
3853 @mypyc_attr(patchable=True)
3854 def dump_to_file(*output: str) -> str:
3855 """Dump `output` to a temporary file. Return path to the file."""
3856 with tempfile.NamedTemporaryFile(
3857 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
3859 for lines in output:
3861 if lines and lines[-1] != "\n":
3867 def nullcontext() -> Iterator[None]:
3868 """Return an empty context manager.
3870 To be used like `nullcontext` in Python 3.7.
3875 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
3876 """Return a unified diff string between strings `a` and `b`."""
3879 a_lines = [line + "\n" for line in a.splitlines()]
3880 b_lines = [line + "\n" for line in b.splitlines()]
3882 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
3886 def cancel(tasks: Iterable["asyncio.Task[Any]"]) -> None:
3887 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
3893 def shutdown(loop: asyncio.AbstractEventLoop) -> None:
3894 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
3896 if sys.version_info[:2] >= (3, 7):
3897 all_tasks = asyncio.all_tasks
3899 all_tasks = asyncio.Task.all_tasks
3900 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
3901 to_cancel = [task for task in all_tasks(loop) if not task.done()]
3905 for task in to_cancel:
3907 loop.run_until_complete(
3908 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
3911 # `concurrent.futures.Future` objects cannot be cancelled once they
3912 # are already running. There might be some when the `shutdown()` happened.
3913 # Silence their logger's spew about the event loop being closed.
3914 cf_logger = logging.getLogger("concurrent.futures")
3915 cf_logger.setLevel(logging.CRITICAL)
3919 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
3920 """Replace `regex` with `replacement` twice on `original`.
3922 This is used by string normalization to perform replaces on
3923 overlapping matches.
3925 return regex.sub(replacement, regex.sub(replacement, original))
3928 def re_compile_maybe_verbose(regex: str) -> Pattern[str]:
3929 """Compile a regular expression string in `regex`.
3931 If it contains newlines, use verbose mode.
3934 regex = "(?x)" + regex
3935 compiled: Pattern[str] = re.compile(regex)
3939 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
3940 """Like `reversed(enumerate(sequence))` if that were possible."""
3941 index = len(sequence) - 1
3942 for element in reversed(sequence):
3943 yield (index, element)
3947 def enumerate_with_length(
3948 line: Line, reversed: bool = False
3949 ) -> Iterator[Tuple[Index, Leaf, int]]:
3950 """Return an enumeration of leaves with their length.
3952 Stops prematurely on multiline strings and standalone comments.
3955 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
3956 enumerate_reversed if reversed else enumerate,
3958 for index, leaf in op(line.leaves):
3959 length = len(leaf.prefix) + len(leaf.value)
3960 if "\n" in leaf.value:
3961 return # Multiline strings, we can't continue.
3963 for comment in line.comments_after(leaf):
3964 length += len(comment.value)
3966 yield index, leaf, length
3969 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
3970 """Return True if `line` is no longer than `line_length`.
3972 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
3975 line_str = str(line).strip("\n")
3977 len(line_str) <= line_length
3978 and "\n" not in line_str # multiline strings
3979 and not line.contains_standalone_comments()
3983 def can_be_split(line: Line) -> bool:
3984 """Return False if the line cannot be split *for sure*.
3986 This is not an exhaustive search but a cheap heuristic that we can use to
3987 avoid some unfortunate formattings (mostly around wrapping unsplittable code
3988 in unnecessary parentheses).
3990 leaves = line.leaves
3994 if leaves[0].type == token.STRING and leaves[1].type == token.DOT:
3998 for leaf in leaves[-2::-1]:
3999 if leaf.type in OPENING_BRACKETS:
4000 if next.type not in CLOSING_BRACKETS:
4004 elif leaf.type == token.DOT:
4006 elif leaf.type == token.NAME:
4007 if not (next.type == token.DOT or next.type in OPENING_BRACKETS):
4010 elif leaf.type not in CLOSING_BRACKETS:
4013 if dot_count > 1 and call_count > 1:
4019 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
4020 """Does `line` have a shape safe to reformat without optional parens around it?
4022 Returns True for only a subset of potentially nice looking formattings but
4023 the point is to not return false positives that end up producing lines that
4026 bt = line.bracket_tracker
4027 if not bt.delimiters:
4028 # Without delimiters the optional parentheses are useless.
4031 max_priority = bt.max_delimiter_priority()
4032 if bt.delimiter_count_with_priority(max_priority) > 1:
4033 # With more than one delimiter of a kind the optional parentheses read better.
4036 if max_priority == DOT_PRIORITY:
4037 # A single stranded method call doesn't require optional parentheses.
4040 assert len(line.leaves) >= 2, "Stranded delimiter"
4042 first = line.leaves[0]
4043 second = line.leaves[1]
4044 penultimate = line.leaves[-2]
4045 last = line.leaves[-1]
4047 # With a single delimiter, omit if the expression starts or ends with
4049 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
4051 length = 4 * line.depth
4052 for _index, leaf, leaf_length in enumerate_with_length(line):
4053 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
4056 length += leaf_length
4057 if length > line_length:
4060 if leaf.type in OPENING_BRACKETS:
4061 # There are brackets we can further split on.
4065 # checked the entire string and line length wasn't exceeded
4066 if len(line.leaves) == _index + 1:
4069 # Note: we are not returning False here because a line might have *both*
4070 # a leading opening bracket and a trailing closing bracket. If the
4071 # opening bracket doesn't match our rule, maybe the closing will.
4074 last.type == token.RPAR
4075 or last.type == token.RBRACE
4077 # don't use indexing for omitting optional parentheses;
4079 last.type == token.RSQB
4081 and last.parent.type != syms.trailer
4084 if penultimate.type in OPENING_BRACKETS:
4085 # Empty brackets don't help.
4088 if is_multiline_string(first):
4089 # Additional wrapping of a multiline string in this situation is
4093 length = 4 * line.depth
4094 seen_other_brackets = False
4095 for _index, leaf, leaf_length in enumerate_with_length(line):
4096 length += leaf_length
4097 if leaf is last.opening_bracket:
4098 if seen_other_brackets or length <= line_length:
4101 elif leaf.type in OPENING_BRACKETS:
4102 # There are brackets we can further split on.
4103 seen_other_brackets = True
4108 def get_cache_file(mode: Mode) -> Path:
4109 return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle"
4112 def read_cache(mode: Mode) -> Cache:
4113 """Read the cache if it exists and is well formed.
4115 If it is not well formed, the call to write_cache later should resolve the issue.
4117 cache_file = get_cache_file(mode)
4118 if not cache_file.exists():
4121 with cache_file.open("rb") as fobj:
4123 cache: Cache = pickle.load(fobj)
4124 except (pickle.UnpicklingError, ValueError):
4130 def get_cache_info(path: Path) -> CacheInfo:
4131 """Return the information used to check if a file is already formatted or not."""
4133 return stat.st_mtime, stat.st_size
4136 def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]:
4137 """Split an iterable of paths in `sources` into two sets.
4139 The first contains paths of files that modified on disk or are not in the
4140 cache. The other contains paths to non-modified files.
4142 todo, done = set(), set()
4145 if cache.get(src) != get_cache_info(src):
4152 def write_cache(cache: Cache, sources: Iterable[Path], mode: Mode) -> None:
4153 """Update the cache file."""
4154 cache_file = get_cache_file(mode)
4156 CACHE_DIR.mkdir(parents=True, exist_ok=True)
4157 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
4158 with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
4159 pickle.dump(new_cache, f, protocol=4)
4160 os.replace(f.name, cache_file)
4165 def patch_click() -> None:
4166 """Make Click not crash.
4168 On certain misconfigured environments, Python 3 selects the ASCII encoding as the
4169 default which restricts paths that it can access during the lifetime of the
4170 application. Click refuses to work in this scenario by raising a RuntimeError.
4172 In case of Black the likelihood that non-ASCII characters are going to be used in
4173 file paths is minimal since it's Python source code. Moreover, this crash was
4174 spurious on Python 3.7 thanks to PEP 538 and PEP 540.
4177 from click import core
4178 from click import _unicodefun # type: ignore
4179 except ModuleNotFoundError:
4182 for module in (core, _unicodefun):
4183 if hasattr(module, "_verify_python3_env"):
4184 module._verify_python3_env = lambda: None
4187 def patched_main() -> None:
4193 if __name__ == "__main__":