All patches and comments are welcome. Please squash your changes to logical
commits before using git-format-patch and git-send-email to
patches@git.madduck.net.
If you'd read over the Git project's submission guidelines and adhered to them,
I'd be especially grateful.
3 from concurrent.futures import Executor, ProcessPoolExecutor
4 from contextlib import contextmanager
5 from datetime import datetime
7 from functools import lru_cache, partial, wraps
11 from multiprocessing import Manager, freeze_support
13 from pathlib import Path
40 from typing_extensions import Final
41 from mypy_extensions import mypyc_attr
43 from appdirs import user_cache_dir
44 from dataclasses import dataclass, field, replace
47 from typed_ast import ast3, ast27
48 from pathspec import PathSpec
51 from blib2to3.pytree import Node, Leaf, type_repr
52 from blib2to3 import pygram, pytree
53 from blib2to3.pgen2 import driver, token
54 from blib2to3.pgen2.grammar import Grammar
55 from blib2to3.pgen2.parse import ParseError
57 from _black_version import version as __version__
59 DEFAULT_LINE_LENGTH = 88
60 DEFAULT_EXCLUDES = r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|\.svn|_build|buck-out|build|dist)/" # noqa: B950
61 DEFAULT_INCLUDES = r"\.pyi?$"
62 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
74 LN = Union[Leaf, Node]
75 SplitFunc = Callable[["Line", Collection["Feature"]], Iterator["Line"]]
78 CacheInfo = Tuple[Timestamp, FileSize]
79 Cache = Dict[Path, CacheInfo]
80 out = partial(click.secho, bold=True, err=True)
81 err = partial(click.secho, fg="red", err=True)
83 pygram.initialize(CACHE_DIR)
84 syms = pygram.python_symbols
87 class NothingChanged(UserWarning):
88 """Raised when reformatted code is the same as source."""
91 class CannotSplit(Exception):
92 """A readable split that fits the allotted line length is impossible."""
95 class InvalidInput(ValueError):
96 """Raised when input source code fails all parse attempts."""
99 class WriteBack(Enum):
106 def from_configuration(cls, *, check: bool, diff: bool) -> "WriteBack":
107 if check and not diff:
110 return cls.DIFF if diff else cls.YES
119 class TargetVersion(Enum):
128 def is_python2(self) -> bool:
129 return self is TargetVersion.PY27
132 PY36_VERSIONS = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38}
136 # All string literals are unicode
139 NUMERIC_UNDERSCORES = 3
140 TRAILING_COMMA_IN_CALL = 4
141 TRAILING_COMMA_IN_DEF = 5
142 # The following two feature-flags are mutually exclusive, and exactly one should be
143 # set for every version of python.
144 ASYNC_IDENTIFIERS = 6
146 ASSIGNMENT_EXPRESSIONS = 8
147 POS_ONLY_ARGUMENTS = 9
150 VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = {
151 TargetVersion.PY27: {Feature.ASYNC_IDENTIFIERS},
152 TargetVersion.PY33: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
153 TargetVersion.PY34: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS},
154 TargetVersion.PY35: {
155 Feature.UNICODE_LITERALS,
156 Feature.TRAILING_COMMA_IN_CALL,
157 Feature.ASYNC_IDENTIFIERS,
159 TargetVersion.PY36: {
160 Feature.UNICODE_LITERALS,
162 Feature.NUMERIC_UNDERSCORES,
163 Feature.TRAILING_COMMA_IN_CALL,
164 Feature.TRAILING_COMMA_IN_DEF,
165 Feature.ASYNC_IDENTIFIERS,
167 TargetVersion.PY37: {
168 Feature.UNICODE_LITERALS,
170 Feature.NUMERIC_UNDERSCORES,
171 Feature.TRAILING_COMMA_IN_CALL,
172 Feature.TRAILING_COMMA_IN_DEF,
173 Feature.ASYNC_KEYWORDS,
175 TargetVersion.PY38: {
176 Feature.UNICODE_LITERALS,
178 Feature.NUMERIC_UNDERSCORES,
179 Feature.TRAILING_COMMA_IN_CALL,
180 Feature.TRAILING_COMMA_IN_DEF,
181 Feature.ASYNC_KEYWORDS,
182 Feature.ASSIGNMENT_EXPRESSIONS,
183 Feature.POS_ONLY_ARGUMENTS,
190 target_versions: Set[TargetVersion] = field(default_factory=set)
191 line_length: int = DEFAULT_LINE_LENGTH
192 string_normalization: bool = True
195 def get_cache_key(self) -> str:
196 if self.target_versions:
197 version_str = ",".join(
199 for version in sorted(self.target_versions, key=lambda v: v.value)
205 str(self.line_length),
206 str(int(self.string_normalization)),
207 str(int(self.is_pyi)),
209 return ".".join(parts)
212 def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool:
213 return all(feature in VERSION_TO_FEATURES[version] for version in target_versions)
216 def find_pyproject_toml(path_search_start: str) -> Optional[str]:
217 """Find the absolute filepath to a pyproject.toml if it exists"""
218 path_project_root = find_project_root(path_search_start)
219 path_pyproject_toml = path_project_root / "pyproject.toml"
220 return str(path_pyproject_toml) if path_pyproject_toml.is_file() else None
223 def parse_pyproject_toml(path_config: str) -> Dict[str, Any]:
224 """Parse a pyproject toml file, pulling out relevant parts for Black
226 If parsing fails, will raise a toml.TomlDecodeError
228 pyproject_toml = toml.load(path_config)
229 config = pyproject_toml.get("tool", {}).get("black", {})
230 return {k.replace("--", "").replace("-", "_"): v for k, v in config.items()}
233 def read_pyproject_toml(
234 ctx: click.Context, param: click.Parameter, value: Union[str, int, bool, None]
236 """Inject Black configuration from "pyproject.toml" into defaults in `ctx`.
238 Returns the path to a successfully found and read configuration file, None
241 assert not isinstance(value, (int, bool)), "Invalid parameter type passed"
243 value = find_pyproject_toml(ctx.params.get("src", ()))
248 config = parse_pyproject_toml(value)
249 except (toml.TomlDecodeError, OSError) as e:
250 raise click.FileError(
251 filename=value, hint=f"Error reading configuration file: {e}"
257 if ctx.default_map is None:
259 ctx.default_map.update(config) # type: ignore # bad types in .pyi
263 def target_version_option_callback(
264 c: click.Context, p: Union[click.Option, click.Parameter], v: Tuple[str, ...]
265 ) -> List[TargetVersion]:
266 """Compute the target versions from a --target-version flag.
268 This is its own function because mypy couldn't infer the type correctly
269 when it was a lambda, causing mypyc trouble.
271 return [TargetVersion[val.upper()] for val in v]
274 @click.command(context_settings=dict(help_option_names=["-h", "--help"]))
275 @click.option("-c", "--code", type=str, help="Format the code passed in as a string.")
280 default=DEFAULT_LINE_LENGTH,
281 help="How many characters per line to allow.",
287 type=click.Choice([v.name.lower() for v in TargetVersion]),
288 callback=target_version_option_callback,
291 "Python versions that should be supported by Black's output. [default: "
292 "per-file auto-detection]"
299 "Allow using Python 3.6-only syntax on all input files. This will put "
300 "trailing commas in function signatures and calls also after *args and "
301 "**kwargs. Deprecated; use --target-version instead. "
302 "[default: per-file auto-detection]"
309 "Format all input files like typing stubs regardless of file extension "
310 "(useful when piping source on standard input)."
315 "--skip-string-normalization",
317 help="Don't normalize string quotes or prefixes.",
323 "Don't write the files back, just return the status. Return code 0 "
324 "means nothing would change. Return code 1 means some files would be "
325 "reformatted. Return code 123 means there was an internal error."
331 help="Don't write the files back, just output a diff for each file on stdout.",
336 help="If --fast given, skip temporary sanity checks. [default: --safe]",
341 default=DEFAULT_INCLUDES,
343 "A regular expression that matches files and directories that should be "
344 "included on recursive searches. An empty value means all files are "
345 "included regardless of the name. Use forward slashes for directories on "
346 "all platforms (Windows, too). Exclusions are calculated first, inclusions "
354 default=DEFAULT_EXCLUDES,
356 "A regular expression that matches files and directories that should be "
357 "excluded on recursive searches. An empty value means no paths are excluded. "
358 "Use forward slashes for directories on all platforms (Windows, too). "
359 "Exclusions are calculated first, inclusions later."
368 "Don't emit non-error messages to stderr. Errors are still emitted; "
369 "silence those with 2>/dev/null."
377 "Also emit messages to stderr about files that were not changed or were "
378 "ignored due to --exclude=."
381 @click.version_option(version=__version__)
386 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
393 exists=False, file_okay=True, dir_okay=False, readable=True, allow_dash=False
396 callback=read_pyproject_toml,
397 help="Read configuration from PATH.",
404 target_version: List[TargetVersion],
410 skip_string_normalization: bool,
415 src: Tuple[str, ...],
416 config: Optional[str],
418 """The uncompromising code formatter."""
419 write_back = WriteBack.from_configuration(check=check, diff=diff)
422 err(f"Cannot use both --target-version and --py36")
425 versions = set(target_version)
428 "--py36 is deprecated and will be removed in a future version. "
429 "Use --target-version py36 instead."
431 versions = PY36_VERSIONS
433 # We'll autodetect later.
436 target_versions=versions,
437 line_length=line_length,
439 string_normalization=not skip_string_normalization,
441 if config and verbose:
442 out(f"Using configuration from {config}.", bold=False, fg="blue")
444 print(format_str(code, mode=mode))
447 include_regex = re_compile_maybe_verbose(include)
449 err(f"Invalid regular expression for include given: {include!r}")
452 exclude_regex = re_compile_maybe_verbose(exclude)
454 err(f"Invalid regular expression for exclude given: {exclude!r}")
456 report = Report(check=check, diff=diff, quiet=quiet, verbose=verbose)
457 root = find_project_root(src)
458 sources: Set[Path] = set()
459 path_empty(src, quiet, verbose, ctx)
464 gen_python_files_in_dir(
465 p, root, include_regex, exclude_regex, report, get_gitignore(root)
468 elif p.is_file() or s == "-":
469 # if a file was explicitly given, we don't care about its extension
472 err(f"invalid path: {s}")
473 if len(sources) == 0:
474 if verbose or not quiet:
475 out("No Python files are present to be formatted. Nothing to do 😴")
478 if len(sources) == 1:
482 write_back=write_back,
488 sources=sources, fast=fast, write_back=write_back, mode=mode, report=report
491 if verbose or not quiet:
492 out("Oh no! 💥 💔 💥" if report.return_code else "All done! ✨ 🍰 ✨")
493 click.secho(str(report), err=True)
494 ctx.exit(report.return_code)
498 src: Tuple[str, ...], quiet: bool, verbose: bool, ctx: click.Context
501 Exit if there is no `src` provided for formatting
504 if verbose or not quiet:
505 out("No Path provided. Nothing to do 😴")
510 src: Path, fast: bool, write_back: WriteBack, mode: FileMode, report: "Report"
512 """Reformat a single file under `src` without spawning child processes.
514 `fast`, `write_back`, and `mode` options are passed to
515 :func:`format_file_in_place` or :func:`format_stdin_to_stdout`.
519 if not src.is_file() and str(src) == "-":
520 if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode):
521 changed = Changed.YES
524 if write_back != WriteBack.DIFF:
525 cache = read_cache(mode)
526 res_src = src.resolve()
527 if res_src in cache and cache[res_src] == get_cache_info(res_src):
528 changed = Changed.CACHED
529 if changed is not Changed.CACHED and format_file_in_place(
530 src, fast=fast, write_back=write_back, mode=mode
532 changed = Changed.YES
533 if (write_back is WriteBack.YES and changed is not Changed.CACHED) or (
534 write_back is WriteBack.CHECK and changed is Changed.NO
536 write_cache(cache, [src], mode)
537 report.done(src, changed)
538 except Exception as exc:
539 report.failed(src, str(exc))
545 write_back: WriteBack,
549 """Reformat multiple files using a ProcessPoolExecutor."""
550 loop = asyncio.get_event_loop()
551 worker_count = os.cpu_count()
552 if sys.platform == "win32":
553 # Work around https://bugs.python.org/issue26903
554 worker_count = min(worker_count, 61)
555 executor = ProcessPoolExecutor(max_workers=worker_count)
557 loop.run_until_complete(
561 write_back=write_back,
573 async def schedule_formatting(
576 write_back: WriteBack,
579 loop: asyncio.AbstractEventLoop,
582 """Run formatting of `sources` in parallel using the provided `executor`.
584 (Use ProcessPoolExecutors for actual parallelism.)
586 `write_back`, `fast`, and `mode` options are passed to
587 :func:`format_file_in_place`.
590 if write_back != WriteBack.DIFF:
591 cache = read_cache(mode)
592 sources, cached = filter_cached(cache, sources)
593 for src in sorted(cached):
594 report.done(src, Changed.CACHED)
599 sources_to_cache = []
601 if write_back == WriteBack.DIFF:
602 # For diff output, we need locks to ensure we don't interleave output
603 # from different processes.
605 lock = manager.Lock()
607 asyncio.ensure_future(
608 loop.run_in_executor(
609 executor, format_file_in_place, src, fast, mode, write_back, lock
612 for src in sorted(sources)
614 pending: Iterable["asyncio.Future[bool]"] = tasks.keys()
616 loop.add_signal_handler(signal.SIGINT, cancel, pending)
617 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
618 except NotImplementedError:
619 # There are no good alternatives for these on Windows.
622 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
624 src = tasks.pop(task)
626 cancelled.append(task)
627 elif task.exception():
628 report.failed(src, str(task.exception()))
630 changed = Changed.YES if task.result() else Changed.NO
631 # If the file was written back or was successfully checked as
632 # well-formatted, store this information in the cache.
633 if write_back is WriteBack.YES or (
634 write_back is WriteBack.CHECK and changed is Changed.NO
636 sources_to_cache.append(src)
637 report.done(src, changed)
639 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
641 write_cache(cache, sources_to_cache, mode)
644 def format_file_in_place(
648 write_back: WriteBack = WriteBack.NO,
649 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
651 """Format file under `src` path. Return True if changed.
653 If `write_back` is DIFF, write a diff to stdout. If it is YES, write reformatted
655 `mode` and `fast` options are passed to :func:`format_file_contents`.
657 if src.suffix == ".pyi":
658 mode = replace(mode, is_pyi=True)
660 then = datetime.utcfromtimestamp(src.stat().st_mtime)
661 with open(src, "rb") as buf:
662 src_contents, encoding, newline = decode_bytes(buf.read())
664 dst_contents = format_file_contents(src_contents, fast=fast, mode=mode)
665 except NothingChanged:
668 if write_back == WriteBack.YES:
669 with open(src, "w", encoding=encoding, newline=newline) as f:
670 f.write(dst_contents)
671 elif write_back == WriteBack.DIFF:
672 now = datetime.utcnow()
673 src_name = f"{src}\t{then} +0000"
674 dst_name = f"{src}\t{now} +0000"
675 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
677 with lock or nullcontext():
678 f = io.TextIOWrapper(
684 f.write(diff_contents)
690 def format_stdin_to_stdout(
691 fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: FileMode
693 """Format file on stdin. Return True if changed.
695 If `write_back` is YES, write reformatted code back to stdout. If it is DIFF,
696 write a diff to stdout. The `mode` argument is passed to
697 :func:`format_file_contents`.
699 then = datetime.utcnow()
700 src, encoding, newline = decode_bytes(sys.stdin.buffer.read())
703 dst = format_file_contents(src, fast=fast, mode=mode)
706 except NothingChanged:
710 f = io.TextIOWrapper(
711 sys.stdout.buffer, encoding=encoding, newline=newline, write_through=True
713 if write_back == WriteBack.YES:
715 elif write_back == WriteBack.DIFF:
716 now = datetime.utcnow()
717 src_name = f"STDIN\t{then} +0000"
718 dst_name = f"STDOUT\t{now} +0000"
719 f.write(diff(src, dst, src_name, dst_name))
723 def format_file_contents(
724 src_contents: str, *, fast: bool, mode: FileMode
726 """Reformat contents a file and return new contents.
728 If `fast` is False, additionally confirm that the reformatted code is
729 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
730 `mode` is passed to :func:`format_str`.
732 if src_contents.strip() == "":
735 dst_contents = format_str(src_contents, mode=mode)
736 if src_contents == dst_contents:
740 assert_equivalent(src_contents, dst_contents)
741 assert_stable(src_contents, dst_contents, mode=mode)
745 def format_str(src_contents: str, *, mode: FileMode) -> FileContent:
746 """Reformat a string and return new contents.
748 `mode` determines formatting options, such as how many characters per line are
751 src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
753 future_imports = get_future_imports(src_node)
754 if mode.target_versions:
755 versions = mode.target_versions
757 versions = detect_target_versions(src_node)
758 normalize_fmt_off(src_node)
759 lines = LineGenerator(
760 remove_u_prefix="unicode_literals" in future_imports
761 or supports_feature(versions, Feature.UNICODE_LITERALS),
763 normalize_strings=mode.string_normalization,
765 elt = EmptyLineTracker(is_pyi=mode.is_pyi)
768 split_line_features = {
770 for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF}
771 if supports_feature(versions, feature)
773 for current_line in lines.visit(src_node):
774 dst_contents.append(str(empty_line) * after)
775 before, after = elt.maybe_empty_lines(current_line)
776 dst_contents.append(str(empty_line) * before)
777 for line in split_line(
778 current_line, line_length=mode.line_length, features=split_line_features
780 dst_contents.append(str(line))
781 return "".join(dst_contents)
784 def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]:
785 """Return a tuple of (decoded_contents, encoding, newline).
787 `newline` is either CRLF or LF but `decoded_contents` is decoded with
788 universal newlines (i.e. only contains LF).
790 srcbuf = io.BytesIO(src)
791 encoding, lines = tokenize.detect_encoding(srcbuf.readline)
793 return "", encoding, "\n"
795 newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n"
797 with io.TextIOWrapper(srcbuf, encoding) as tiow:
798 return tiow.read(), encoding, newline
801 def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]:
802 if not target_versions:
803 # No target_version specified, so try all grammars.
806 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords,
808 pygram.python_grammar_no_print_statement_no_exec_statement,
809 # Python 2.7 with future print_function import
810 pygram.python_grammar_no_print_statement,
812 pygram.python_grammar,
815 if all(version.is_python2() for version in target_versions):
816 # Python 2-only code, so try Python 2 grammars.
818 # Python 2.7 with future print_function import
819 pygram.python_grammar_no_print_statement,
821 pygram.python_grammar,
824 # Python 3-compatible code, so only try Python 3 grammar.
826 # If we have to parse both, try to parse async as a keyword first
827 if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
830 pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords
832 if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
834 grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
835 # At least one of the above branches must have been taken, because every Python
836 # version has exactly one of the two 'ASYNC_*' flags
840 def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
841 """Given a string with source, return the lib2to3 Node."""
842 if src_txt[-1:] != "\n":
845 for grammar in get_grammars(set(target_versions)):
846 drv = driver.Driver(grammar, pytree.convert)
848 result = drv.parse_string(src_txt, True)
851 except ParseError as pe:
852 lineno, column = pe.context[1]
853 lines = src_txt.splitlines()
855 faulty_line = lines[lineno - 1]
857 faulty_line = "<line number missing in source>"
858 exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}")
862 if isinstance(result, Leaf):
863 result = Node(syms.file_input, [result])
867 def lib2to3_unparse(node: Node) -> str:
868 """Given a lib2to3 node, return its string representation."""
876 class Visitor(Generic[T]):
877 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
879 def visit(self, node: LN) -> Iterator[T]:
880 """Main method to visit `node` and its children.
882 It tries to find a `visit_*()` method for the given `node.type`, like
883 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
884 If no dedicated `visit_*()` method is found, chooses `visit_default()`
887 Then yields objects of type `T` from the selected visitor.
890 name = token.tok_name[node.type]
892 name = str(type_repr(node.type))
893 # We explicitly branch on whether a visitor exists (instead of
894 # using self.visit_default as the default arg to getattr) in order
895 # to save needing to create a bound method object and so mypyc can
896 # generate a native call to visit_default.
897 visitf = getattr(self, f"visit_{name}", None)
899 yield from visitf(node)
901 yield from self.visit_default(node)
903 def visit_default(self, node: LN) -> Iterator[T]:
904 """Default `visit_*()` implementation. Recurses to children of `node`."""
905 if isinstance(node, Node):
906 for child in node.children:
907 yield from self.visit(child)
911 class DebugVisitor(Visitor[T]):
914 def visit_default(self, node: LN) -> Iterator[T]:
915 indent = " " * (2 * self.tree_depth)
916 if isinstance(node, Node):
917 _type = type_repr(node.type)
918 out(f"{indent}{_type}", fg="yellow")
920 for child in node.children:
921 yield from self.visit(child)
924 out(f"{indent}/{_type}", fg="yellow", bold=False)
926 _type = token.tok_name.get(node.type, str(node.type))
927 out(f"{indent}{_type}", fg="blue", nl=False)
929 # We don't have to handle prefixes for `Node` objects since
930 # that delegates to the first child anyway.
931 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
932 out(f" {node.value!r}", fg="blue", bold=False)
935 def show(cls, code: Union[str, Leaf, Node]) -> None:
936 """Pretty-print the lib2to3 AST of a given string of `code`.
938 Convenience method for debugging.
940 v: DebugVisitor[None] = DebugVisitor()
941 if isinstance(code, str):
942 code = lib2to3_parse(code)
946 WHITESPACE: Final = {token.DEDENT, token.INDENT, token.NEWLINE}
957 STANDALONE_COMMENT: Final = 153
958 token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT"
959 LOGIC_OPERATORS: Final = {"and", "or"}
960 COMPARATORS: Final = {
968 MATH_OPERATORS: Final = {
984 STARS: Final = {token.STAR, token.DOUBLESTAR}
985 VARARGS_SPECIALS: Final = STARS | {token.SLASH}
986 VARARGS_PARENTS: Final = {
988 syms.argument, # double star in arglist
989 syms.trailer, # single argument to call
991 syms.varargslist, # lambdas
993 UNPACKING_PARENTS: Final = {
994 syms.atom, # single element of a list or set literal
998 syms.testlist_star_expr,
1000 TEST_DESCENDANTS: Final = {
1017 ASSIGNMENTS: Final = {
1033 COMPREHENSION_PRIORITY: Final = 20
1034 COMMA_PRIORITY: Final = 18
1035 TERNARY_PRIORITY: Final = 16
1036 LOGIC_PRIORITY: Final = 14
1037 STRING_PRIORITY: Final = 12
1038 COMPARATOR_PRIORITY: Final = 10
1039 MATH_PRIORITIES: Final = {
1041 token.CIRCUMFLEX: 8,
1044 token.RIGHTSHIFT: 6,
1049 token.DOUBLESLASH: 4,
1053 token.DOUBLESTAR: 2,
1055 DOT_PRIORITY: Final = 1
1059 class BracketTracker:
1060 """Keeps track of brackets on a line."""
1063 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = field(default_factory=dict)
1064 delimiters: Dict[LeafID, Priority] = field(default_factory=dict)
1065 previous: Optional[Leaf] = None
1066 _for_loop_depths: List[int] = field(default_factory=list)
1067 _lambda_argument_depths: List[int] = field(default_factory=list)
1069 def mark(self, leaf: Leaf) -> None:
1070 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
1072 All leaves receive an int `bracket_depth` field that stores how deep
1073 within brackets a given leaf is. 0 means there are no enclosing brackets
1074 that started on this line.
1076 If a leaf is itself a closing bracket, it receives an `opening_bracket`
1077 field that it forms a pair with. This is a one-directional link to
1078 avoid reference cycles.
1080 If a leaf is a delimiter (a token on which Black can split the line if
1081 needed) and it's on depth 0, its `id()` is stored in the tracker's
1084 if leaf.type == token.COMMENT:
1087 self.maybe_decrement_after_for_loop_variable(leaf)
1088 self.maybe_decrement_after_lambda_arguments(leaf)
1089 if leaf.type in CLOSING_BRACKETS:
1091 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
1092 leaf.opening_bracket = opening_bracket
1093 leaf.bracket_depth = self.depth
1095 delim = is_split_before_delimiter(leaf, self.previous)
1096 if delim and self.previous is not None:
1097 self.delimiters[id(self.previous)] = delim
1099 delim = is_split_after_delimiter(leaf, self.previous)
1101 self.delimiters[id(leaf)] = delim
1102 if leaf.type in OPENING_BRACKETS:
1103 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
1105 self.previous = leaf
1106 self.maybe_increment_lambda_arguments(leaf)
1107 self.maybe_increment_for_loop_variable(leaf)
1109 def any_open_brackets(self) -> bool:
1110 """Return True if there is an yet unmatched open bracket on the line."""
1111 return bool(self.bracket_match)
1113 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> Priority:
1114 """Return the highest priority of a delimiter found on the line.
1116 Values are consistent with what `is_split_*_delimiter()` return.
1117 Raises ValueError on no delimiters.
1119 return max(v for k, v in self.delimiters.items() if k not in exclude)
1121 def delimiter_count_with_priority(self, priority: Priority = 0) -> int:
1122 """Return the number of delimiters with the given `priority`.
1124 If no `priority` is passed, defaults to max priority on the line.
1126 if not self.delimiters:
1129 priority = priority or self.max_delimiter_priority()
1130 return sum(1 for p in self.delimiters.values() if p == priority)
1132 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1133 """In a for loop, or comprehension, the variables are often unpacks.
1135 To avoid splitting on the comma in this situation, increase the depth of
1136 tokens between `for` and `in`.
1138 if leaf.type == token.NAME and leaf.value == "for":
1140 self._for_loop_depths.append(self.depth)
1145 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
1146 """See `maybe_increment_for_loop_variable` above for explanation."""
1148 self._for_loop_depths
1149 and self._for_loop_depths[-1] == self.depth
1150 and leaf.type == token.NAME
1151 and leaf.value == "in"
1154 self._for_loop_depths.pop()
1159 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
1160 """In a lambda expression, there might be more than one argument.
1162 To avoid splitting on the comma in this situation, increase the depth of
1163 tokens between `lambda` and `:`.
1165 if leaf.type == token.NAME and leaf.value == "lambda":
1167 self._lambda_argument_depths.append(self.depth)
1172 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
1173 """See `maybe_increment_lambda_arguments` above for explanation."""
1175 self._lambda_argument_depths
1176 and self._lambda_argument_depths[-1] == self.depth
1177 and leaf.type == token.COLON
1180 self._lambda_argument_depths.pop()
1185 def get_open_lsqb(self) -> Optional[Leaf]:
1186 """Return the most recent opening square bracket (if any)."""
1187 return self.bracket_match.get((self.depth - 1, token.RSQB))
1192 """Holds leaves and comments. Can be printed with `str(line)`."""
1195 leaves: List[Leaf] = field(default_factory=list)
1196 # keys ordered like `leaves`
1197 comments: Dict[LeafID, List[Leaf]] = field(default_factory=dict)
1198 bracket_tracker: BracketTracker = field(default_factory=BracketTracker)
1199 inside_brackets: bool = False
1200 should_explode: bool = False
1202 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
1203 """Add a new `leaf` to the end of the line.
1205 Unless `preformatted` is True, the `leaf` will receive a new consistent
1206 whitespace prefix and metadata applied by :class:`BracketTracker`.
1207 Trailing commas are maybe removed, unpacked for loop variables are
1208 demoted from being delimiters.
1210 Inline comments are put aside.
1212 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
1216 if token.COLON == leaf.type and self.is_class_paren_empty:
1217 del self.leaves[-2:]
1218 if self.leaves and not preformatted:
1219 # Note: at this point leaf.prefix should be empty except for
1220 # imports, for which we only preserve newlines.
1221 leaf.prefix += whitespace(
1222 leaf, complex_subscript=self.is_complex_subscript(leaf)
1224 if self.inside_brackets or not preformatted:
1225 self.bracket_tracker.mark(leaf)
1226 self.maybe_remove_trailing_comma(leaf)
1227 if not self.append_comment(leaf):
1228 self.leaves.append(leaf)
1230 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
1231 """Like :func:`append()` but disallow invalid standalone comment structure.
1233 Raises ValueError when any `leaf` is appended after a standalone comment
1234 or when a standalone comment is not the first leaf on the line.
1236 if self.bracket_tracker.depth == 0:
1238 raise ValueError("cannot append to standalone comments")
1240 if self.leaves and leaf.type == STANDALONE_COMMENT:
1242 "cannot append standalone comments to a populated line"
1245 self.append(leaf, preformatted=preformatted)
1248 def is_comment(self) -> bool:
1249 """Is this line a standalone comment?"""
1250 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
1253 def is_decorator(self) -> bool:
1254 """Is this line a decorator?"""
1255 return bool(self) and self.leaves[0].type == token.AT
1258 def is_import(self) -> bool:
1259 """Is this an import line?"""
1260 return bool(self) and is_import(self.leaves[0])
1263 def is_class(self) -> bool:
1264 """Is this line a class definition?"""
1267 and self.leaves[0].type == token.NAME
1268 and self.leaves[0].value == "class"
1272 def is_stub_class(self) -> bool:
1273 """Is this line a class definition with a body consisting only of "..."?"""
1274 return self.is_class and self.leaves[-3:] == [
1275 Leaf(token.DOT, ".") for _ in range(3)
1279 def is_collection_with_optional_trailing_comma(self) -> bool:
1280 """Is this line a collection literal with a trailing comma that's optional?
1282 Note that the trailing comma in a 1-tuple is not optional.
1284 if not self.leaves or len(self.leaves) < 4:
1287 # Look for and address a trailing colon.
1288 if self.leaves[-1].type == token.COLON:
1289 closer = self.leaves[-2]
1292 closer = self.leaves[-1]
1294 if closer.type not in CLOSING_BRACKETS or self.inside_brackets:
1297 if closer.type == token.RPAR:
1298 # Tuples require an extra check, because if there's only
1299 # one element in the tuple removing the comma unmakes the
1302 # We also check for parens before looking for the trailing
1303 # comma because in some cases (eg assigning a dict
1304 # literal) the literal gets wrapped in temporary parens
1305 # during parsing. This case is covered by the
1306 # collections.py test data.
1307 opener = closer.opening_bracket
1308 for _open_index, leaf in enumerate(self.leaves):
1313 # Couldn't find the matching opening paren, play it safe.
1317 comma_depth = self.leaves[close_index - 1].bracket_depth
1318 for leaf in self.leaves[_open_index + 1 : close_index]:
1319 if leaf.bracket_depth == comma_depth and leaf.type == token.COMMA:
1322 # We haven't looked yet for the trailing comma because
1323 # we might also have caught noop parens.
1324 return self.leaves[close_index - 1].type == token.COMMA
1327 return False # it's either a one-tuple or didn't have a trailing comma
1329 if self.leaves[close_index - 1].type in CLOSING_BRACKETS:
1331 closer = self.leaves[close_index]
1332 if closer.type == token.RPAR:
1333 # TODO: this is a gut feeling. Will we ever see this?
1336 if self.leaves[close_index - 1].type != token.COMMA:
1342 def is_def(self) -> bool:
1343 """Is this a function definition? (Also returns True for async defs.)"""
1345 first_leaf = self.leaves[0]
1350 second_leaf: Optional[Leaf] = self.leaves[1]
1353 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
1354 first_leaf.type == token.ASYNC
1355 and second_leaf is not None
1356 and second_leaf.type == token.NAME
1357 and second_leaf.value == "def"
1361 def is_class_paren_empty(self) -> bool:
1362 """Is this a class with no base classes but using parentheses?
1364 Those are unnecessary and should be removed.
1368 and len(self.leaves) == 4
1370 and self.leaves[2].type == token.LPAR
1371 and self.leaves[2].value == "("
1372 and self.leaves[3].type == token.RPAR
1373 and self.leaves[3].value == ")"
1377 def is_triple_quoted_string(self) -> bool:
1378 """Is the line a triple quoted string?"""
1381 and self.leaves[0].type == token.STRING
1382 and self.leaves[0].value.startswith(('"""', "'''"))
1385 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
1386 """If so, needs to be split before emitting."""
1387 for leaf in self.leaves:
1388 if leaf.type == STANDALONE_COMMENT and leaf.bracket_depth <= depth_limit:
1393 def contains_uncollapsable_type_comments(self) -> bool:
1396 last_leaf = self.leaves[-1]
1397 ignored_ids.add(id(last_leaf))
1398 if last_leaf.type == token.COMMA or (
1399 last_leaf.type == token.RPAR and not last_leaf.value
1401 # When trailing commas or optional parens are inserted by Black for
1402 # consistency, comments after the previous last element are not moved
1403 # (they don't have to, rendering will still be correct). So we ignore
1404 # trailing commas and invisible.
1405 last_leaf = self.leaves[-2]
1406 ignored_ids.add(id(last_leaf))
1410 # A type comment is uncollapsable if it is attached to a leaf
1411 # that isn't at the end of the line (since that could cause it
1412 # to get associated to a different argument) or if there are
1413 # comments before it (since that could cause it to get hidden
1415 comment_seen = False
1416 for leaf_id, comments in self.comments.items():
1417 for comment in comments:
1418 if is_type_comment(comment):
1419 if comment_seen or (
1420 not is_type_comment(comment, " ignore")
1421 and leaf_id not in ignored_ids
1429 def contains_unsplittable_type_ignore(self) -> bool:
1433 # If a 'type: ignore' is attached to the end of a line, we
1434 # can't split the line, because we can't know which of the
1435 # subexpressions the ignore was meant to apply to.
1437 # We only want this to apply to actual physical lines from the
1438 # original source, though: we don't want the presence of a
1439 # 'type: ignore' at the end of a multiline expression to
1440 # justify pushing it all onto one line. Thus we
1441 # (unfortunately) need to check the actual source lines and
1442 # only report an unsplittable 'type: ignore' if this line was
1443 # one line in the original code.
1445 # Grab the first and last line numbers, skipping generated leaves
1446 first_line = next((l.lineno for l in self.leaves if l.lineno != 0), 0)
1447 last_line = next((l.lineno for l in reversed(self.leaves) if l.lineno != 0), 0)
1449 if first_line == last_line:
1450 # We look at the last two leaves since a comma or an
1451 # invisible paren could have been added at the end of the
1453 for node in self.leaves[-2:]:
1454 for comment in self.comments.get(id(node), []):
1455 if is_type_comment(comment, " ignore"):
1460 def contains_multiline_strings(self) -> bool:
1461 return any(is_multiline_string(leaf) for leaf in self.leaves)
1463 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1464 """Remove trailing comma if there is one and it's safe."""
1465 if not (self.leaves and self.leaves[-1].type == token.COMMA):
1468 # We remove trailing commas only in the case of importing a
1469 # single name from a module.
1473 and len(self.leaves) > 4
1474 and self.leaves[-1].type == token.COMMA
1475 and closing.type in CLOSING_BRACKETS
1476 and self.leaves[-4].type == token.NAME
1478 # regular `from foo import bar,`
1479 self.leaves[-4].value == "import"
1480 # `from foo import (bar as baz,)
1482 len(self.leaves) > 6
1483 and self.leaves[-6].value == "import"
1484 and self.leaves[-3].value == "as"
1486 # `from foo import bar as baz,`
1488 len(self.leaves) > 5
1489 and self.leaves[-5].value == "import"
1490 and self.leaves[-3].value == "as"
1493 and closing.type == token.RPAR
1497 self.remove_trailing_comma()
1500 def append_comment(self, comment: Leaf) -> bool:
1501 """Add an inline or standalone comment to the line."""
1503 comment.type == STANDALONE_COMMENT
1504 and self.bracket_tracker.any_open_brackets()
1509 if comment.type != token.COMMENT:
1513 comment.type = STANDALONE_COMMENT
1517 last_leaf = self.leaves[-1]
1519 last_leaf.type == token.RPAR
1520 and not last_leaf.value
1521 and last_leaf.parent
1522 and len(list(last_leaf.parent.leaves())) <= 3
1523 and not is_type_comment(comment)
1525 # Comments on an optional parens wrapping a single leaf should belong to
1526 # the wrapped node except if it's a type comment. Pinning the comment like
1527 # this avoids unstable formatting caused by comment migration.
1528 if len(self.leaves) < 2:
1529 comment.type = STANDALONE_COMMENT
1533 last_leaf = self.leaves[-2]
1534 self.comments.setdefault(id(last_leaf), []).append(comment)
1537 def comments_after(self, leaf: Leaf) -> List[Leaf]:
1538 """Generate comments that should appear directly after `leaf`."""
1539 return self.comments.get(id(leaf), [])
1541 def remove_trailing_comma(self) -> None:
1542 """Remove the trailing comma and moves the comments attached to it."""
1543 trailing_comma = self.leaves.pop()
1544 trailing_comma_comments = self.comments.pop(id(trailing_comma), [])
1545 self.comments.setdefault(id(self.leaves[-1]), []).extend(
1546 trailing_comma_comments
1549 def is_complex_subscript(self, leaf: Leaf) -> bool:
1550 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1551 open_lsqb = self.bracket_tracker.get_open_lsqb()
1552 if open_lsqb is None:
1555 subscript_start = open_lsqb.next_sibling
1557 if isinstance(subscript_start, Node):
1558 if subscript_start.type == syms.listmaker:
1561 if subscript_start.type == syms.subscriptlist:
1562 subscript_start = child_towards(subscript_start, leaf)
1563 return subscript_start is not None and any(
1564 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1567 def __str__(self) -> str:
1568 """Render the line."""
1572 indent = " " * self.depth
1573 leaves = iter(self.leaves)
1574 first = next(leaves)
1575 res = f"{first.prefix}{indent}{first.value}"
1578 for comment in itertools.chain.from_iterable(self.comments.values()):
1582 def __bool__(self) -> bool:
1583 """Return True if the line has leaves or comments."""
1584 return bool(self.leaves or self.comments)
1588 class EmptyLineTracker:
1589 """Provides a stateful method that returns the number of potential extra
1590 empty lines needed before and after the currently processed line.
1592 Note: this tracker works on lines that haven't been split yet. It assumes
1593 the prefix of the first leaf consists of optional newlines. Those newlines
1594 are consumed by `maybe_empty_lines()` and included in the computation.
1597 is_pyi: bool = False
1598 previous_line: Optional[Line] = None
1599 previous_after: int = 0
1600 previous_defs: List[int] = field(default_factory=list)
1602 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1603 """Return the number of extra empty lines before and after the `current_line`.
1605 This is for separating `def`, `async def` and `class` with extra empty
1606 lines (two on module-level).
1608 before, after = self._maybe_empty_lines(current_line)
1610 # Black should not insert empty lines at the beginning
1613 if self.previous_line is None
1614 else before - self.previous_after
1616 self.previous_after = after
1617 self.previous_line = current_line
1618 return before, after
1620 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1622 if current_line.depth == 0:
1623 max_allowed = 1 if self.is_pyi else 2
1624 if current_line.leaves:
1625 # Consume the first leaf's extra newlines.
1626 first_leaf = current_line.leaves[0]
1627 before = first_leaf.prefix.count("\n")
1628 before = min(before, max_allowed)
1629 first_leaf.prefix = ""
1632 depth = current_line.depth
1633 while self.previous_defs and self.previous_defs[-1] >= depth:
1634 self.previous_defs.pop()
1636 before = 0 if depth else 1
1638 before = 1 if depth else 2
1639 if current_line.is_decorator or current_line.is_def or current_line.is_class:
1640 return self._maybe_empty_lines_for_class_or_def(current_line, before)
1644 and self.previous_line.is_import
1645 and not current_line.is_import
1646 and depth == self.previous_line.depth
1648 return (before or 1), 0
1652 and self.previous_line.is_class
1653 and current_line.is_triple_quoted_string
1659 def _maybe_empty_lines_for_class_or_def(
1660 self, current_line: Line, before: int
1661 ) -> Tuple[int, int]:
1662 if not current_line.is_decorator:
1663 self.previous_defs.append(current_line.depth)
1664 if self.previous_line is None:
1665 # Don't insert empty lines before the first line in the file.
1668 if self.previous_line.is_decorator:
1671 if self.previous_line.depth < current_line.depth and (
1672 self.previous_line.is_class or self.previous_line.is_def
1677 self.previous_line.is_comment
1678 and self.previous_line.depth == current_line.depth
1684 if self.previous_line.depth > current_line.depth:
1686 elif current_line.is_class or self.previous_line.is_class:
1687 if current_line.is_stub_class and self.previous_line.is_stub_class:
1688 # No blank line between classes with an empty body
1692 elif current_line.is_def and not self.previous_line.is_def:
1693 # Blank line between a block of functions and a block of non-functions
1699 if current_line.depth and newlines:
1705 class LineGenerator(Visitor[Line]):
1706 """Generates reformatted Line objects. Empty lines are not emitted.
1708 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1709 in ways that will no longer stringify to valid Python code on the tree.
1712 is_pyi: bool = False
1713 normalize_strings: bool = True
1714 current_line: Line = field(default_factory=Line)
1715 remove_u_prefix: bool = False
1717 def line(self, indent: int = 0) -> Iterator[Line]:
1720 If the line is empty, only emit if it makes sense.
1721 If the line is too long, split it first and then generate.
1723 If any lines were generated, set up a new current_line.
1725 if not self.current_line:
1726 self.current_line.depth += indent
1727 return # Line is empty, don't emit. Creating a new one unnecessary.
1729 complete_line = self.current_line
1730 self.current_line = Line(depth=complete_line.depth + indent)
1733 def visit_default(self, node: LN) -> Iterator[Line]:
1734 """Default `visit_*()` implementation. Recurses to children of `node`."""
1735 if isinstance(node, Leaf):
1736 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1737 for comment in generate_comments(node):
1738 if any_open_brackets:
1739 # any comment within brackets is subject to splitting
1740 self.current_line.append(comment)
1741 elif comment.type == token.COMMENT:
1742 # regular trailing comment
1743 self.current_line.append(comment)
1744 yield from self.line()
1747 # regular standalone comment
1748 yield from self.line()
1750 self.current_line.append(comment)
1751 yield from self.line()
1753 normalize_prefix(node, inside_brackets=any_open_brackets)
1754 if self.normalize_strings and node.type == token.STRING:
1755 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1756 normalize_string_quotes(node)
1757 if node.type == token.NUMBER:
1758 normalize_numeric_literal(node)
1759 if node.type not in WHITESPACE:
1760 self.current_line.append(node)
1761 yield from super().visit_default(node)
1763 def visit_INDENT(self, node: Leaf) -> Iterator[Line]:
1764 """Increase indentation level, maybe yield a line."""
1765 # In blib2to3 INDENT never holds comments.
1766 yield from self.line(+1)
1767 yield from self.visit_default(node)
1769 def visit_DEDENT(self, node: Leaf) -> Iterator[Line]:
1770 """Decrease indentation level, maybe yield a line."""
1771 # The current line might still wait for trailing comments. At DEDENT time
1772 # there won't be any (they would be prefixes on the preceding NEWLINE).
1773 # Emit the line then.
1774 yield from self.line()
1776 # While DEDENT has no value, its prefix may contain standalone comments
1777 # that belong to the current indentation level. Get 'em.
1778 yield from self.visit_default(node)
1780 # Finally, emit the dedent.
1781 yield from self.line(-1)
1784 self, node: Node, keywords: Set[str], parens: Set[str]
1785 ) -> Iterator[Line]:
1786 """Visit a statement.
1788 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1789 `def`, `with`, `class`, `assert` and assignments.
1791 The relevant Python language `keywords` for a given statement will be
1792 NAME leaves within it. This methods puts those on a separate line.
1794 `parens` holds a set of string leaf values immediately after which
1795 invisible parens should be put.
1797 normalize_invisible_parens(node, parens_after=parens)
1798 for child in node.children:
1799 if child.type == token.NAME and child.value in keywords: # type: ignore
1800 yield from self.line()
1802 yield from self.visit(child)
1804 def visit_suite(self, node: Node) -> Iterator[Line]:
1805 """Visit a suite."""
1806 if self.is_pyi and is_stub_suite(node):
1807 yield from self.visit(node.children[2])
1809 yield from self.visit_default(node)
1811 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1812 """Visit a statement without nested statements."""
1813 is_suite_like = node.parent and node.parent.type in STATEMENT
1815 if self.is_pyi and is_stub_body(node):
1816 yield from self.visit_default(node)
1818 yield from self.line(+1)
1819 yield from self.visit_default(node)
1820 yield from self.line(-1)
1823 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1824 yield from self.line()
1825 yield from self.visit_default(node)
1827 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1828 """Visit `async def`, `async for`, `async with`."""
1829 yield from self.line()
1831 children = iter(node.children)
1832 for child in children:
1833 yield from self.visit(child)
1835 if child.type == token.ASYNC:
1838 internal_stmt = next(children)
1839 for child in internal_stmt.children:
1840 yield from self.visit(child)
1842 def visit_decorators(self, node: Node) -> Iterator[Line]:
1843 """Visit decorators."""
1844 for child in node.children:
1845 yield from self.line()
1846 yield from self.visit(child)
1848 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1849 """Remove a semicolon and put the other statement on a separate line."""
1850 yield from self.line()
1852 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1853 """End of file. Process outstanding comments and end with a newline."""
1854 yield from self.visit_default(leaf)
1855 yield from self.line()
1857 def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]:
1858 if not self.current_line.bracket_tracker.any_open_brackets():
1859 yield from self.line()
1860 yield from self.visit_default(leaf)
1862 def visit_factor(self, node: Node) -> Iterator[Line]:
1863 """Force parentheses between a unary op and a binary power:
1865 -2 ** 8 -> -(2 ** 8)
1867 _operator, operand = node.children
1869 operand.type == syms.power
1870 and len(operand.children) == 3
1871 and operand.children[1].type == token.DOUBLESTAR
1873 lpar = Leaf(token.LPAR, "(")
1874 rpar = Leaf(token.RPAR, ")")
1875 index = operand.remove() or 0
1876 node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
1877 yield from self.visit_default(node)
1879 def __post_init__(self) -> None:
1880 """You are in a twisty little maze of passages."""
1883 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1884 self.visit_if_stmt = partial(
1885 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1887 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1888 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1889 self.visit_try_stmt = partial(
1890 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1892 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1893 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1894 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1895 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1896 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1897 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1898 self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
1899 self.visit_del_stmt = partial(v, keywords=Ø, parens={"del"})
1900 self.visit_async_funcdef = self.visit_async_stmt
1901 self.visit_decorated = self.visit_decorators
1904 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1905 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1906 OPENING_BRACKETS = set(BRACKET.keys())
1907 CLOSING_BRACKETS = set(BRACKET.values())
1908 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1909 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1912 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901
1913 """Return whitespace prefix if needed for the given `leaf`.
1915 `complex_subscript` signals whether the given leaf is part of a subscription
1916 which has non-trivial arguments, like arithmetic expressions or function calls.
1924 if t in ALWAYS_NO_SPACE:
1927 if t == token.COMMENT:
1930 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1931 if t == token.COLON and p.type not in {
1938 prev = leaf.prev_sibling
1940 prevp = preceding_leaf(p)
1941 if not prevp or prevp.type in OPENING_BRACKETS:
1944 if t == token.COLON:
1945 if prevp.type == token.COLON:
1948 elif prevp.type != token.COMMA and not complex_subscript:
1953 if prevp.type == token.EQUAL:
1955 if prevp.parent.type in {
1963 elif prevp.parent.type == syms.typedargslist:
1964 # A bit hacky: if the equal sign has whitespace, it means we
1965 # previously found it's a typed argument. So, we're using
1969 elif prevp.type in VARARGS_SPECIALS:
1970 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1973 elif prevp.type == token.COLON:
1974 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1975 return SPACE if complex_subscript else NO
1979 and prevp.parent.type == syms.factor
1980 and prevp.type in MATH_OPERATORS
1985 prevp.type == token.RIGHTSHIFT
1987 and prevp.parent.type == syms.shift_expr
1988 and prevp.prev_sibling
1989 and prevp.prev_sibling.type == token.NAME
1990 and prevp.prev_sibling.value == "print" # type: ignore
1992 # Python 2 print chevron
1995 elif prev.type in OPENING_BRACKETS:
1998 if p.type in {syms.parameters, syms.arglist}:
1999 # untyped function signatures or calls
2000 if not prev or prev.type != token.COMMA:
2003 elif p.type == syms.varargslist:
2005 if prev and prev.type != token.COMMA:
2008 elif p.type == syms.typedargslist:
2009 # typed function signatures
2013 if t == token.EQUAL:
2014 if prev.type != syms.tname:
2017 elif prev.type == token.EQUAL:
2018 # A bit hacky: if the equal sign has whitespace, it means we
2019 # previously found it's a typed argument. So, we're using that, too.
2022 elif prev.type != token.COMMA:
2025 elif p.type == syms.tname:
2028 prevp = preceding_leaf(p)
2029 if not prevp or prevp.type != token.COMMA:
2032 elif p.type == syms.trailer:
2033 # attributes and calls
2034 if t == token.LPAR or t == token.RPAR:
2039 prevp = preceding_leaf(p)
2040 if not prevp or prevp.type != token.NUMBER:
2043 elif t == token.LSQB:
2046 elif prev.type != token.COMMA:
2049 elif p.type == syms.argument:
2051 if t == token.EQUAL:
2055 prevp = preceding_leaf(p)
2056 if not prevp or prevp.type == token.LPAR:
2059 elif prev.type in {token.EQUAL} | VARARGS_SPECIALS:
2062 elif p.type == syms.decorator:
2066 elif p.type == syms.dotted_name:
2070 prevp = preceding_leaf(p)
2071 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
2074 elif p.type == syms.classdef:
2078 if prev and prev.type == token.LPAR:
2081 elif p.type in {syms.subscript, syms.sliceop}:
2084 assert p.parent is not None, "subscripts are always parented"
2085 if p.parent.type == syms.subscriptlist:
2090 elif not complex_subscript:
2093 elif p.type == syms.atom:
2094 if prev and t == token.DOT:
2095 # dots, but not the first one.
2098 elif p.type == syms.dictsetmaker:
2100 if prev and prev.type == token.DOUBLESTAR:
2103 elif p.type in {syms.factor, syms.star_expr}:
2106 prevp = preceding_leaf(p)
2107 if not prevp or prevp.type in OPENING_BRACKETS:
2110 prevp_parent = prevp.parent
2111 assert prevp_parent is not None
2112 if prevp.type == token.COLON and prevp_parent.type in {
2118 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
2121 elif t in {token.NAME, token.NUMBER, token.STRING}:
2124 elif p.type == syms.import_from:
2126 if prev and prev.type == token.DOT:
2129 elif t == token.NAME:
2133 if prev and prev.type == token.DOT:
2136 elif p.type == syms.sliceop:
2142 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
2143 """Return the first leaf that precedes `node`, if any."""
2145 res = node.prev_sibling
2147 if isinstance(res, Leaf):
2151 return list(res.leaves())[-1]
2160 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
2161 """Return the child of `ancestor` that contains `descendant`."""
2162 node: Optional[LN] = descendant
2163 while node and node.parent != ancestor:
2168 def container_of(leaf: Leaf) -> LN:
2169 """Return `leaf` or one of its ancestors that is the topmost container of it.
2171 By "container" we mean a node where `leaf` is the very first child.
2173 same_prefix = leaf.prefix
2174 container: LN = leaf
2176 parent = container.parent
2180 if parent.children[0].prefix != same_prefix:
2183 if parent.type == syms.file_input:
2186 if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS:
2193 def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2194 """Return the priority of the `leaf` delimiter, given a line break after it.
2196 The delimiter priorities returned here are from those delimiters that would
2197 cause a line break after themselves.
2199 Higher numbers are higher priority.
2201 if leaf.type == token.COMMA:
2202 return COMMA_PRIORITY
2207 def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority:
2208 """Return the priority of the `leaf` delimiter, given a line break before it.
2210 The delimiter priorities returned here are from those delimiters that would
2211 cause a line break before themselves.
2213 Higher numbers are higher priority.
2215 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
2216 # * and ** might also be MATH_OPERATORS but in this case they are not.
2217 # Don't treat them as a delimiter.
2221 leaf.type == token.DOT
2223 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
2224 and (previous is None or previous.type in CLOSING_BRACKETS)
2229 leaf.type in MATH_OPERATORS
2231 and leaf.parent.type not in {syms.factor, syms.star_expr}
2233 return MATH_PRIORITIES[leaf.type]
2235 if leaf.type in COMPARATORS:
2236 return COMPARATOR_PRIORITY
2239 leaf.type == token.STRING
2240 and previous is not None
2241 and previous.type == token.STRING
2243 return STRING_PRIORITY
2245 if leaf.type not in {token.NAME, token.ASYNC}:
2251 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
2252 or leaf.type == token.ASYNC
2255 not isinstance(leaf.prev_sibling, Leaf)
2256 or leaf.prev_sibling.value != "async"
2258 return COMPREHENSION_PRIORITY
2263 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
2265 return COMPREHENSION_PRIORITY
2267 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
2268 return TERNARY_PRIORITY
2270 if leaf.value == "is":
2271 return COMPARATOR_PRIORITY
2276 and leaf.parent.type in {syms.comp_op, syms.comparison}
2278 previous is not None
2279 and previous.type == token.NAME
2280 and previous.value == "not"
2283 return COMPARATOR_PRIORITY
2288 and leaf.parent.type == syms.comp_op
2290 previous is not None
2291 and previous.type == token.NAME
2292 and previous.value == "is"
2295 return COMPARATOR_PRIORITY
2297 if leaf.value in LOGIC_OPERATORS and leaf.parent:
2298 return LOGIC_PRIORITY
2303 FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"}
2304 FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"}
2307 def generate_comments(leaf: LN) -> Iterator[Leaf]:
2308 """Clean the prefix of the `leaf` and generate comments from it, if any.
2310 Comments in lib2to3 are shoved into the whitespace prefix. This happens
2311 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
2312 move because it does away with modifying the grammar to include all the
2313 possible places in which comments can be placed.
2315 The sad consequence for us though is that comments don't "belong" anywhere.
2316 This is why this function generates simple parentless Leaf objects for
2317 comments. We simply don't know what the correct parent should be.
2319 No matter though, we can live without this. We really only need to
2320 differentiate between inline and standalone comments. The latter don't
2321 share the line with any code.
2323 Inline comments are emitted as regular token.COMMENT leaves. Standalone
2324 are emitted with a fake STANDALONE_COMMENT token identifier.
2326 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
2327 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
2332 """Describes a piece of syntax that is a comment.
2334 It's not a :class:`blib2to3.pytree.Leaf` so that:
2336 * it can be cached (`Leaf` objects should not be reused more than once as
2337 they store their lineno, column, prefix, and parent information);
2338 * `newlines` and `consumed` fields are kept separate from the `value`. This
2339 simplifies handling of special marker comments like ``# fmt: off/on``.
2342 type: int # token.COMMENT or STANDALONE_COMMENT
2343 value: str # content of the comment
2344 newlines: int # how many newlines before the comment
2345 consumed: int # how many characters of the original leaf's prefix did we consume
2348 @lru_cache(maxsize=4096)
2349 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
2350 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
2351 result: List[ProtoComment] = []
2352 if not prefix or "#" not in prefix:
2358 for index, line in enumerate(prefix.split("\n")):
2359 consumed += len(line) + 1 # adding the length of the split '\n'
2360 line = line.lstrip()
2363 if not line.startswith("#"):
2364 # Escaped newlines outside of a comment are not really newlines at
2365 # all. We treat a single-line comment following an escaped newline
2366 # as a simple trailing comment.
2367 if line.endswith("\\"):
2371 if index == ignored_lines and not is_endmarker:
2372 comment_type = token.COMMENT # simple trailing comment
2374 comment_type = STANDALONE_COMMENT
2375 comment = make_comment(line)
2378 type=comment_type, value=comment, newlines=nlines, consumed=consumed
2385 def make_comment(content: str) -> str:
2386 """Return a consistently formatted comment from the given `content` string.
2388 All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single
2389 space between the hash sign and the content.
2391 If `content` didn't start with a hash sign, one is provided.
2393 content = content.rstrip()
2397 if content[0] == "#":
2398 content = content[1:]
2399 if content and content[0] not in " !:#'%":
2400 content = " " + content
2401 return "#" + content
2407 inner: bool = False,
2408 features: Collection[Feature] = (),
2409 ) -> Iterator[Line]:
2410 """Split a `line` into potentially many lines.
2412 They should fit in the allotted `line_length` but might not be able to.
2413 `inner` signifies that there were a pair of brackets somewhere around the
2414 current `line`, possibly transitively. This means we can fallback to splitting
2415 by delimiters if the LHS/RHS don't yield any results.
2417 `features` are syntactical features that may be used in the output.
2423 line_str = str(line).strip("\n")
2426 not line.contains_uncollapsable_type_comments()
2427 and not line.should_explode
2428 and not line.is_collection_with_optional_trailing_comma
2430 is_line_short_enough(line, line_length=line_length, line_str=line_str)
2431 or line.contains_unsplittable_type_ignore()
2437 split_funcs: List[SplitFunc]
2439 split_funcs = [left_hand_split]
2442 def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]:
2443 for omit in generate_trailers_to_omit(line, line_length):
2444 lines = list(right_hand_split(line, line_length, features, omit=omit))
2445 if is_line_short_enough(lines[0], line_length=line_length):
2449 # All splits failed, best effort split with no omits.
2450 # This mostly happens to multiline strings that are by definition
2451 # reported as not fitting a single line.
2452 # line_length=1 here was historically a bug that somehow became a feature.
2453 # See #762 and #781 for the full story.
2454 yield from right_hand_split(line, line_length=1, features=features)
2456 if line.inside_brackets:
2457 split_funcs = [delimiter_split, standalone_comment_split, rhs]
2460 for split_func in split_funcs:
2461 # We are accumulating lines in `result` because we might want to abort
2462 # mission and return the original line in the end, or attempt a different
2464 result: List[Line] = []
2466 for l in split_func(line, features):
2467 if str(l).strip("\n") == line_str:
2468 raise CannotSplit("Split function returned an unchanged result")
2472 l, line_length=line_length, inner=True, features=features
2486 def left_hand_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2487 """Split line into many lines, starting with the first matching bracket pair.
2489 Note: this usually looks weird, only use this for function definitions.
2490 Prefer RHS otherwise. This is why this function is not symmetrical with
2491 :func:`right_hand_split` which also handles optional parentheses.
2493 tail_leaves: List[Leaf] = []
2494 body_leaves: List[Leaf] = []
2495 head_leaves: List[Leaf] = []
2496 current_leaves = head_leaves
2497 matching_bracket: Optional[Leaf] = None
2498 for leaf in line.leaves:
2500 current_leaves is body_leaves
2501 and leaf.type in CLOSING_BRACKETS
2502 and leaf.opening_bracket is matching_bracket
2504 current_leaves = tail_leaves if body_leaves else head_leaves
2505 current_leaves.append(leaf)
2506 if current_leaves is head_leaves:
2507 if leaf.type in OPENING_BRACKETS:
2508 matching_bracket = leaf
2509 current_leaves = body_leaves
2510 if not matching_bracket:
2511 raise CannotSplit("No brackets found")
2513 head = bracket_split_build_line(head_leaves, line, matching_bracket)
2514 body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True)
2515 tail = bracket_split_build_line(tail_leaves, line, matching_bracket)
2516 bracket_split_succeeded_or_raise(head, body, tail)
2517 for result in (head, body, tail):
2522 def right_hand_split(
2525 features: Collection[Feature] = (),
2526 omit: Collection[LeafID] = (),
2527 ) -> Iterator[Line]:
2528 """Split line into many lines, starting with the last matching bracket pair.
2530 If the split was by optional parentheses, attempt splitting without them, too.
2531 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2534 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2536 tail_leaves: List[Leaf] = []
2537 body_leaves: List[Leaf] = []
2538 head_leaves: List[Leaf] = []
2539 current_leaves = tail_leaves
2540 opening_bracket: Optional[Leaf] = None
2541 closing_bracket: Optional[Leaf] = None
2542 for leaf in reversed(line.leaves):
2543 if current_leaves is body_leaves:
2544 if leaf is opening_bracket:
2545 current_leaves = head_leaves if body_leaves else tail_leaves
2546 current_leaves.append(leaf)
2547 if current_leaves is tail_leaves:
2548 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2549 opening_bracket = leaf.opening_bracket
2550 closing_bracket = leaf
2551 current_leaves = body_leaves
2552 if not (opening_bracket and closing_bracket and head_leaves):
2553 # If there is no opening or closing_bracket that means the split failed and
2554 # all content is in the tail. Otherwise, if `head_leaves` are empty, it means
2555 # the matching `opening_bracket` wasn't available on `line` anymore.
2556 raise CannotSplit("No brackets found")
2558 tail_leaves.reverse()
2559 body_leaves.reverse()
2560 head_leaves.reverse()
2561 head = bracket_split_build_line(head_leaves, line, opening_bracket)
2562 body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True)
2563 tail = bracket_split_build_line(tail_leaves, line, opening_bracket)
2564 bracket_split_succeeded_or_raise(head, body, tail)
2566 # the body shouldn't be exploded
2567 not body.should_explode
2568 # the opening bracket is an optional paren
2569 and opening_bracket.type == token.LPAR
2570 and not opening_bracket.value
2571 # the closing bracket is an optional paren
2572 and closing_bracket.type == token.RPAR
2573 and not closing_bracket.value
2574 # it's not an import (optional parens are the only thing we can split on
2575 # in this case; attempting a split without them is a waste of time)
2576 and not line.is_import
2577 # there are no standalone comments in the body
2578 and not body.contains_standalone_comments(0)
2579 # and we can actually remove the parens
2580 and can_omit_invisible_parens(body, line_length)
2582 omit = {id(closing_bracket), *omit}
2584 yield from right_hand_split(line, line_length, features=features, omit=omit)
2590 or is_line_short_enough(body, line_length=line_length)
2593 "Splitting failed, body is still too long and can't be split."
2596 elif head.contains_multiline_strings() or tail.contains_multiline_strings():
2598 "The current optional pair of parentheses is bound to fail to "
2599 "satisfy the splitting algorithm because the head or the tail "
2600 "contains multiline strings which by definition never fit one "
2604 ensure_visible(opening_bracket)
2605 ensure_visible(closing_bracket)
2606 for result in (head, body, tail):
2611 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2612 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2614 Do nothing otherwise.
2616 A left- or right-hand split is based on a pair of brackets. Content before
2617 (and including) the opening bracket is left on one line, content inside the
2618 brackets is put on a separate line, and finally content starting with and
2619 following the closing bracket is put on a separate line.
2621 Those are called `head`, `body`, and `tail`, respectively. If the split
2622 produced the same line (all content in `head`) or ended up with an empty `body`
2623 and the `tail` is just the closing bracket, then it's considered failed.
2625 tail_len = len(str(tail).strip())
2628 raise CannotSplit("Splitting brackets produced the same line")
2632 f"Splitting brackets on an empty body to save "
2633 f"{tail_len} characters is not worth it"
2637 def bracket_split_build_line(
2638 leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False
2640 """Return a new line with given `leaves` and respective comments from `original`.
2642 If `is_body` is True, the result line is one-indented inside brackets and as such
2643 has its first leaf's prefix normalized and a trailing comma added when expected.
2645 result = Line(depth=original.depth)
2647 result.inside_brackets = True
2650 # Since body is a new indent level, remove spurious leading whitespace.
2651 normalize_prefix(leaves[0], inside_brackets=True)
2652 # Ensure a trailing comma for imports and standalone function arguments, but
2653 # be careful not to add one after any comments or within type annotations.
2656 and opening_bracket.value == "("
2657 and not any(l.type == token.COMMA for l in leaves)
2660 if original.is_import or no_commas:
2661 for i in range(len(leaves) - 1, -1, -1):
2662 if leaves[i].type == STANDALONE_COMMENT:
2665 if leaves[i].type != token.COMMA:
2666 leaves.insert(i + 1, Leaf(token.COMMA, ","))
2671 result.append(leaf, preformatted=True)
2672 for comment_after in original.comments_after(leaf):
2673 result.append(comment_after, preformatted=True)
2675 result.should_explode = should_explode(result, opening_bracket)
2679 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2680 """Normalize prefix of the first leaf in every line returned by `split_func`.
2682 This is a decorator over relevant split functions.
2686 def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2687 for l in split_func(line, features):
2688 normalize_prefix(l.leaves[0], inside_brackets=True)
2691 return split_wrapper
2694 @dont_increase_indentation
2695 def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
2696 """Split according to delimiters of the highest priority.
2698 If the appropriate Features are given, the split will add trailing commas
2699 also in function signatures and calls that contain `*` and `**`.
2702 last_leaf = line.leaves[-1]
2704 raise CannotSplit("Line empty")
2706 bt = line.bracket_tracker
2708 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2710 raise CannotSplit("No delimiters found")
2712 if delimiter_priority == DOT_PRIORITY:
2713 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2714 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2716 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2717 lowest_depth = sys.maxsize
2718 trailing_comma_safe = True
2720 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2721 """Append `leaf` to current line or to new line if appending impossible."""
2722 nonlocal current_line
2724 current_line.append_safe(leaf, preformatted=True)
2728 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2729 current_line.append(leaf)
2731 for leaf in line.leaves:
2732 yield from append_to_line(leaf)
2734 for comment_after in line.comments_after(leaf):
2735 yield from append_to_line(comment_after)
2737 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2738 if leaf.bracket_depth == lowest_depth:
2739 if is_vararg(leaf, within={syms.typedargslist}):
2740 trailing_comma_safe = (
2741 trailing_comma_safe and Feature.TRAILING_COMMA_IN_DEF in features
2743 elif is_vararg(leaf, within={syms.arglist, syms.argument}):
2744 trailing_comma_safe = (
2745 trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features
2748 leaf_priority = bt.delimiters.get(id(leaf))
2749 if leaf_priority == delimiter_priority:
2752 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2756 and delimiter_priority == COMMA_PRIORITY
2757 and current_line.leaves[-1].type != token.COMMA
2758 and current_line.leaves[-1].type != STANDALONE_COMMENT
2760 current_line.append(Leaf(token.COMMA, ","))
2764 @dont_increase_indentation
2765 def standalone_comment_split(
2766 line: Line, features: Collection[Feature] = ()
2767 ) -> Iterator[Line]:
2768 """Split standalone comments from the rest of the line."""
2769 if not line.contains_standalone_comments(0):
2770 raise CannotSplit("Line does not have any standalone comments")
2772 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2774 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2775 """Append `leaf` to current line or to new line if appending impossible."""
2776 nonlocal current_line
2778 current_line.append_safe(leaf, preformatted=True)
2782 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2783 current_line.append(leaf)
2785 for leaf in line.leaves:
2786 yield from append_to_line(leaf)
2788 for comment_after in line.comments_after(leaf):
2789 yield from append_to_line(comment_after)
2795 def is_import(leaf: Leaf) -> bool:
2796 """Return True if the given leaf starts an import statement."""
2803 (v == "import" and p and p.type == syms.import_name)
2804 or (v == "from" and p and p.type == syms.import_from)
2809 def is_type_comment(leaf: Leaf, suffix: str = "") -> bool:
2810 """Return True if the given leaf is a special comment.
2811 Only returns true for type comments for now."""
2814 return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix)
2817 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2818 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2821 Note: don't use backslashes for formatting or you'll lose your voting rights.
2823 if not inside_brackets:
2824 spl = leaf.prefix.split("#")
2825 if "\\" not in spl[0]:
2826 nl_count = spl[-1].count("\n")
2829 leaf.prefix = "\n" * nl_count
2835 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2836 """Make all string prefixes lowercase.
2838 If remove_u_prefix is given, also removes any u prefix from the string.
2840 Note: Mutates its argument.
2842 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2843 assert match is not None, f"failed to match string {leaf.value!r}"
2844 orig_prefix = match.group(1)
2845 new_prefix = orig_prefix.replace("F", "f").replace("B", "b").replace("U", "u")
2847 new_prefix = new_prefix.replace("u", "")
2848 leaf.value = f"{new_prefix}{match.group(2)}"
2851 def normalize_string_quotes(leaf: Leaf) -> None:
2852 """Prefer double quotes but only if it doesn't cause more escaping.
2854 Adds or removes backslashes as appropriate. Doesn't parse and fix
2855 strings nested in f-strings (yet).
2857 Note: Mutates its argument.
2859 value = leaf.value.lstrip("furbFURB")
2860 if value[:3] == '"""':
2863 elif value[:3] == "'''":
2866 elif value[0] == '"':
2872 first_quote_pos = leaf.value.find(orig_quote)
2873 if first_quote_pos == -1:
2874 return # There's an internal error
2876 prefix = leaf.value[:first_quote_pos]
2877 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2878 escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}")
2879 escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}")
2880 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2881 if "r" in prefix.casefold():
2882 if unescaped_new_quote.search(body):
2883 # There's at least one unescaped new_quote in this raw string
2884 # so converting is impossible
2887 # Do not introduce or remove backslashes in raw strings
2890 # remove unnecessary escapes
2891 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2892 if body != new_body:
2893 # Consider the string without unnecessary escapes as the original
2895 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2896 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2897 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2898 if "f" in prefix.casefold():
2899 matches = re.findall(
2901 (?:[^{]|^)\{ # start of the string or a non-{ followed by a single {
2902 ([^{].*?) # contents of the brackets except if begins with {{
2903 \}(?:[^}]|$) # A } followed by end of the string or a non-}
2910 # Do not introduce backslashes in interpolated expressions
2913 if new_quote == '"""' and new_body[-1:] == '"':
2915 new_body = new_body[:-1] + '\\"'
2916 orig_escape_count = body.count("\\")
2917 new_escape_count = new_body.count("\\")
2918 if new_escape_count > orig_escape_count:
2919 return # Do not introduce more escaping
2921 if new_escape_count == orig_escape_count and orig_quote == '"':
2922 return # Prefer double quotes
2924 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2927 def normalize_numeric_literal(leaf: Leaf) -> None:
2928 """Normalizes numeric (float, int, and complex) literals.
2930 All letters used in the representation are normalized to lowercase (except
2931 in Python 2 long literals).
2933 text = leaf.value.lower()
2934 if text.startswith(("0o", "0b")):
2935 # Leave octal and binary literals alone.
2937 elif text.startswith("0x"):
2938 # Change hex literals to upper case.
2939 before, after = text[:2], text[2:]
2940 text = f"{before}{after.upper()}"
2942 before, after = text.split("e")
2944 if after.startswith("-"):
2947 elif after.startswith("+"):
2949 before = format_float_or_int_string(before)
2950 text = f"{before}e{sign}{after}"
2951 elif text.endswith(("j", "l")):
2954 # Capitalize in "2L" because "l" looks too similar to "1".
2957 text = f"{format_float_or_int_string(number)}{suffix}"
2959 text = format_float_or_int_string(text)
2963 def format_float_or_int_string(text: str) -> str:
2964 """Formats a float string like "1.0"."""
2968 before, after = text.split(".")
2969 return f"{before or 0}.{after or 0}"
2972 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2973 """Make existing optional parentheses invisible or create new ones.
2975 `parens_after` is a set of string leaf values immediately after which parens
2978 Standardizes on visible parentheses for single-element tuples, and keeps
2979 existing visible parentheses for other tuples and generator expressions.
2981 for pc in list_comments(node.prefix, is_endmarker=False):
2982 if pc.value in FMT_OFF:
2983 # This `node` has a prefix with `# fmt: off`, don't mess with parens.
2986 for index, child in enumerate(list(node.children)):
2987 # Add parentheses around long tuple unpacking in assignments.
2990 and isinstance(child, Node)
2991 and child.type == syms.testlist_star_expr
2996 if is_walrus_assignment(child):
2999 if child.type == syms.atom:
3000 if maybe_make_parens_invisible_in_atom(child, parent=node):
3001 wrap_in_parentheses(node, child, visible=False)
3002 elif is_one_tuple(child):
3003 wrap_in_parentheses(node, child, visible=True)
3004 elif node.type == syms.import_from:
3005 # "import from" nodes store parentheses directly as part of
3007 if child.type == token.LPAR:
3008 # make parentheses invisible
3009 child.value = "" # type: ignore
3010 node.children[-1].value = "" # type: ignore
3011 elif child.type != token.STAR:
3012 # insert invisible parentheses
3013 node.insert_child(index, Leaf(token.LPAR, ""))
3014 node.append_child(Leaf(token.RPAR, ""))
3017 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
3018 wrap_in_parentheses(node, child, visible=False)
3020 check_lpar = isinstance(child, Leaf) and child.value in parens_after
3023 def normalize_fmt_off(node: Node) -> None:
3024 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
3027 try_again = convert_one_fmt_off_pair(node)
3030 def convert_one_fmt_off_pair(node: Node) -> bool:
3031 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
3033 Returns True if a pair was converted.
3035 for leaf in node.leaves():
3036 previous_consumed = 0
3037 for comment in list_comments(leaf.prefix, is_endmarker=False):
3038 if comment.value in FMT_OFF:
3039 # We only want standalone comments. If there's no previous leaf or
3040 # the previous leaf is indentation, it's a standalone comment in
3042 if comment.type != STANDALONE_COMMENT:
3043 prev = preceding_leaf(leaf)
3044 if prev and prev.type not in WHITESPACE:
3047 ignored_nodes = list(generate_ignored_nodes(leaf))
3048 if not ignored_nodes:
3051 first = ignored_nodes[0] # Can be a container node with the `leaf`.
3052 parent = first.parent
3053 prefix = first.prefix
3054 first.prefix = prefix[comment.consumed :]
3056 comment.value + "\n" + "".join(str(n) for n in ignored_nodes)
3058 if hidden_value.endswith("\n"):
3059 # That happens when one of the `ignored_nodes` ended with a NEWLINE
3060 # leaf (possibly followed by a DEDENT).
3061 hidden_value = hidden_value[:-1]
3062 first_idx: Optional[int] = None
3063 for ignored in ignored_nodes:
3064 index = ignored.remove()
3065 if first_idx is None:
3067 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
3068 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
3069 parent.insert_child(
3074 prefix=prefix[:previous_consumed] + "\n" * comment.newlines,
3079 previous_consumed = comment.consumed
3084 def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]:
3085 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
3087 Stops at the end of the block.
3089 container: Optional[LN] = container_of(leaf)
3090 while container is not None and container.type != token.ENDMARKER:
3092 for comment in list_comments(container.prefix, is_endmarker=False):
3093 if comment.value in FMT_ON:
3095 elif comment.value in FMT_OFF:
3102 container = container.next_sibling
3105 def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool:
3106 """If it's safe, make the parens in the atom `node` invisible, recursively.
3107 Additionally, remove repeated, adjacent invisible parens from the atom `node`
3108 as they are redundant.
3110 Returns whether the node should itself be wrapped in invisible parentheses.
3114 node.type != syms.atom
3115 or is_empty_tuple(node)
3116 or is_one_tuple(node)
3117 or (is_yield(node) and parent.type != syms.expr_stmt)
3118 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
3122 first = node.children[0]
3123 last = node.children[-1]
3124 if first.type == token.LPAR and last.type == token.RPAR:
3125 middle = node.children[1]
3126 # make parentheses invisible
3127 first.value = "" # type: ignore
3128 last.value = "" # type: ignore
3129 maybe_make_parens_invisible_in_atom(middle, parent=parent)
3131 if is_atom_with_invisible_parens(middle):
3132 # Strip the invisible parens from `middle` by replacing
3133 # it with the child in-between the invisible parens
3134 middle.replace(middle.children[1])
3141 def is_atom_with_invisible_parens(node: LN) -> bool:
3142 """Given a `LN`, determines whether it's an atom `node` with invisible
3143 parens. Useful in dedupe-ing and normalizing parens.
3145 if isinstance(node, Leaf) or node.type != syms.atom:
3148 first, last = node.children[0], node.children[-1]
3150 isinstance(first, Leaf)
3151 and first.type == token.LPAR
3152 and first.value == ""
3153 and isinstance(last, Leaf)
3154 and last.type == token.RPAR
3155 and last.value == ""
3159 def is_empty_tuple(node: LN) -> bool:
3160 """Return True if `node` holds an empty tuple."""
3162 node.type == syms.atom
3163 and len(node.children) == 2
3164 and node.children[0].type == token.LPAR
3165 and node.children[1].type == token.RPAR
3169 def unwrap_singleton_parenthesis(node: LN) -> Optional[LN]:
3170 """Returns `wrapped` if `node` is of the shape ( wrapped ).
3172 Parenthesis can be optional. Returns None otherwise"""
3173 if len(node.children) != 3:
3176 lpar, wrapped, rpar = node.children
3177 if not (lpar.type == token.LPAR and rpar.type == token.RPAR):
3183 def wrap_in_parentheses(parent: Node, child: LN, *, visible: bool = True) -> None:
3184 """Wrap `child` in parentheses.
3186 This replaces `child` with an atom holding the parentheses and the old
3187 child. That requires moving the prefix.
3189 If `visible` is False, the leaves will be valueless (and thus invisible).
3191 lpar = Leaf(token.LPAR, "(" if visible else "")
3192 rpar = Leaf(token.RPAR, ")" if visible else "")
3193 prefix = child.prefix
3195 index = child.remove() or 0
3196 new_child = Node(syms.atom, [lpar, child, rpar])
3197 new_child.prefix = prefix
3198 parent.insert_child(index, new_child)
3201 def is_one_tuple(node: LN) -> bool:
3202 """Return True if `node` holds a tuple with one element, with or without parens."""
3203 if node.type == syms.atom:
3204 gexp = unwrap_singleton_parenthesis(node)
3205 if gexp is None or gexp.type != syms.testlist_gexp:
3208 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
3211 node.type in IMPLICIT_TUPLE
3212 and len(node.children) == 2
3213 and node.children[1].type == token.COMMA
3217 def is_walrus_assignment(node: LN) -> bool:
3218 """Return True iff `node` is of the shape ( test := test )"""
3219 inner = unwrap_singleton_parenthesis(node)
3220 return inner is not None and inner.type == syms.namedexpr_test
3223 def is_yield(node: LN) -> bool:
3224 """Return True if `node` holds a `yield` or `yield from` expression."""
3225 if node.type == syms.yield_expr:
3228 if node.type == token.NAME and node.value == "yield": # type: ignore
3231 if node.type != syms.atom:
3234 if len(node.children) != 3:
3237 lpar, expr, rpar = node.children
3238 if lpar.type == token.LPAR and rpar.type == token.RPAR:
3239 return is_yield(expr)
3244 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
3245 """Return True if `leaf` is a star or double star in a vararg or kwarg.
3247 If `within` includes VARARGS_PARENTS, this applies to function signatures.
3248 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
3249 extended iterable unpacking (PEP 3132) and additional unpacking
3250 generalizations (PEP 448).
3252 if leaf.type not in VARARGS_SPECIALS or not leaf.parent:
3256 if p.type == syms.star_expr:
3257 # Star expressions are also used as assignment targets in extended
3258 # iterable unpacking (PEP 3132). See what its parent is instead.
3264 return p.type in within
3267 def is_multiline_string(leaf: Leaf) -> bool:
3268 """Return True if `leaf` is a multiline string that actually spans many lines."""
3269 value = leaf.value.lstrip("furbFURB")
3270 return value[:3] in {'"""', "'''"} and "\n" in value
3273 def is_stub_suite(node: Node) -> bool:
3274 """Return True if `node` is a suite with a stub body."""
3276 len(node.children) != 4
3277 or node.children[0].type != token.NEWLINE
3278 or node.children[1].type != token.INDENT
3279 or node.children[3].type != token.DEDENT
3283 return is_stub_body(node.children[2])
3286 def is_stub_body(node: LN) -> bool:
3287 """Return True if `node` is a simple statement containing an ellipsis."""
3288 if not isinstance(node, Node) or node.type != syms.simple_stmt:
3291 if len(node.children) != 2:
3294 child = node.children[0]
3296 child.type == syms.atom
3297 and len(child.children) == 3
3298 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
3302 def max_delimiter_priority_in_atom(node: LN) -> Priority:
3303 """Return maximum delimiter priority inside `node`.
3305 This is specific to atoms with contents contained in a pair of parentheses.
3306 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
3308 if node.type != syms.atom:
3311 first = node.children[0]
3312 last = node.children[-1]
3313 if not (first.type == token.LPAR and last.type == token.RPAR):
3316 bt = BracketTracker()
3317 for c in node.children[1:-1]:
3318 if isinstance(c, Leaf):
3321 for leaf in c.leaves():
3324 return bt.max_delimiter_priority()
3330 def ensure_visible(leaf: Leaf) -> None:
3331 """Make sure parentheses are visible.
3333 They could be invisible as part of some statements (see
3334 :func:`normalize_invisible_parens` and :func:`visit_import_from`).
3336 if leaf.type == token.LPAR:
3338 elif leaf.type == token.RPAR:
3342 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
3343 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
3346 opening_bracket.parent
3347 and opening_bracket.parent.type in {syms.atom, syms.import_from}
3348 and opening_bracket.value in "[{("
3353 last_leaf = line.leaves[-1]
3354 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
3355 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
3356 except (IndexError, ValueError):
3359 return max_priority == COMMA_PRIORITY
3362 def get_features_used(node: Node) -> Set[Feature]:
3363 """Return a set of (relatively) new Python features used in this file.
3365 Currently looking for:
3367 - underscores in numeric literals;
3368 - trailing commas after * or ** in function signatures and calls;
3369 - positional only arguments in function signatures and lambdas;
3371 features: Set[Feature] = set()
3372 for n in node.pre_order():
3373 if n.type == token.STRING:
3374 value_head = n.value[:2] # type: ignore
3375 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
3376 features.add(Feature.F_STRINGS)
3378 elif n.type == token.NUMBER:
3379 if "_" in n.value: # type: ignore
3380 features.add(Feature.NUMERIC_UNDERSCORES)
3382 elif n.type == token.SLASH:
3383 if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}:
3384 features.add(Feature.POS_ONLY_ARGUMENTS)
3386 elif n.type == token.COLONEQUAL:
3387 features.add(Feature.ASSIGNMENT_EXPRESSIONS)
3390 n.type in {syms.typedargslist, syms.arglist}
3392 and n.children[-1].type == token.COMMA
3394 if n.type == syms.typedargslist:
3395 feature = Feature.TRAILING_COMMA_IN_DEF
3397 feature = Feature.TRAILING_COMMA_IN_CALL
3399 for ch in n.children:
3400 if ch.type in STARS:
3401 features.add(feature)
3403 if ch.type == syms.argument:
3404 for argch in ch.children:
3405 if argch.type in STARS:
3406 features.add(feature)
3411 def detect_target_versions(node: Node) -> Set[TargetVersion]:
3412 """Detect the version to target based on the nodes used."""
3413 features = get_features_used(node)
3415 version for version in TargetVersion if features <= VERSION_TO_FEATURES[version]
3419 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
3420 """Generate sets of closing bracket IDs that should be omitted in a RHS.
3422 Brackets can be omitted if the entire trailer up to and including
3423 a preceding closing bracket fits in one line.
3425 Yielded sets are cumulative (contain results of previous yields, too). First
3429 omit: Set[LeafID] = set()
3432 length = 4 * line.depth
3433 opening_bracket: Optional[Leaf] = None
3434 closing_bracket: Optional[Leaf] = None
3435 inner_brackets: Set[LeafID] = set()
3436 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
3437 length += leaf_length
3438 if length > line_length:
3441 has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
3442 if leaf.type == STANDALONE_COMMENT or has_inline_comment:
3446 if leaf is opening_bracket:
3447 opening_bracket = None
3448 elif leaf.type in CLOSING_BRACKETS:
3449 inner_brackets.add(id(leaf))
3450 elif leaf.type in CLOSING_BRACKETS:
3451 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
3452 # Empty brackets would fail a split so treat them as "inner"
3453 # brackets (e.g. only add them to the `omit` set if another
3454 # pair of brackets was good enough.
3455 inner_brackets.add(id(leaf))
3459 omit.add(id(closing_bracket))
3460 omit.update(inner_brackets)
3461 inner_brackets.clear()
3465 opening_bracket = leaf.opening_bracket
3466 closing_bracket = leaf
3469 def get_future_imports(node: Node) -> Set[str]:
3470 """Return a set of __future__ imports in the file."""
3471 imports: Set[str] = set()
3473 def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]:
3474 for child in children:
3475 if isinstance(child, Leaf):
3476 if child.type == token.NAME:
3479 elif child.type == syms.import_as_name:
3480 orig_name = child.children[0]
3481 assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
3482 assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
3483 yield orig_name.value
3485 elif child.type == syms.import_as_names:
3486 yield from get_imports_from_children(child.children)
3489 raise AssertionError("Invalid syntax parsing imports")
3491 for child in node.children:
3492 if child.type != syms.simple_stmt:
3495 first_child = child.children[0]
3496 if isinstance(first_child, Leaf):
3497 # Continue looking if we see a docstring; otherwise stop.
3499 len(child.children) == 2
3500 and first_child.type == token.STRING
3501 and child.children[1].type == token.NEWLINE
3507 elif first_child.type == syms.import_from:
3508 module_name = first_child.children[1]
3509 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
3512 imports |= set(get_imports_from_children(first_child.children[3:]))
3520 def get_gitignore(root: Path) -> PathSpec:
3521 """ Return a PathSpec matching gitignore content if present."""
3522 gitignore = root / ".gitignore"
3523 lines: List[str] = []
3524 if gitignore.is_file():
3525 with gitignore.open() as gf:
3526 lines = gf.readlines()
3527 return PathSpec.from_lines("gitwildmatch", lines)
3530 def gen_python_files_in_dir(
3533 include: Pattern[str],
3534 exclude: Pattern[str],
3536 gitignore: PathSpec,
3537 ) -> Iterator[Path]:
3538 """Generate all files under `path` whose paths are not excluded by the
3539 `exclude` regex, but are included by the `include` regex.
3541 Symbolic links pointing outside of the `root` directory are ignored.
3543 `report` is where output about exclusions goes.
3545 assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
3546 for child in path.iterdir():
3547 # First ignore files matching .gitignore
3548 if gitignore.match_file(child.as_posix()):
3549 report.path_ignored(child, f"matches the .gitignore file content")
3552 # Then ignore with `exclude` option.
3554 normalized_path = "/" + child.resolve().relative_to(root).as_posix()
3555 except OSError as e:
3556 report.path_ignored(child, f"cannot be read because {e}")
3560 if child.is_symlink():
3561 report.path_ignored(
3562 child, f"is a symbolic link that points outside {root}"
3569 normalized_path += "/"
3571 exclude_match = exclude.search(normalized_path)
3572 if exclude_match and exclude_match.group(0):
3573 report.path_ignored(child, f"matches the --exclude regular expression")
3577 yield from gen_python_files_in_dir(
3578 child, root, include, exclude, report, gitignore
3581 elif child.is_file():
3582 include_match = include.search(normalized_path)
3588 def find_project_root(srcs: Iterable[str]) -> Path:
3589 """Return a directory containing .git, .hg, or pyproject.toml.
3591 That directory can be one of the directories passed in `srcs` or their
3594 If no directory in the tree contains a marker that would specify it's the
3595 project root, the root of the file system is returned.
3598 return Path("/").resolve()
3600 common_base = min(Path(src).resolve() for src in srcs)
3601 if common_base.is_dir():
3602 # Append a fake file so `parents` below returns `common_base_dir`, too.
3603 common_base /= "fake-file"
3604 for directory in common_base.parents:
3605 if (directory / ".git").exists():
3608 if (directory / ".hg").is_dir():
3611 if (directory / "pyproject.toml").is_file():
3619 """Provides a reformatting counter. Can be rendered with `str(report)`."""
3624 verbose: bool = False
3625 change_count: int = 0
3627 failure_count: int = 0
3629 def done(self, src: Path, changed: Changed) -> None:
3630 """Increment the counter for successful reformatting. Write out a message."""
3631 if changed is Changed.YES:
3632 reformatted = "would reformat" if self.check or self.diff else "reformatted"
3633 if self.verbose or not self.quiet:
3634 out(f"{reformatted} {src}")
3635 self.change_count += 1
3638 if changed is Changed.NO:
3639 msg = f"{src} already well formatted, good job."
3641 msg = f"{src} wasn't modified on disk since last run."
3642 out(msg, bold=False)
3643 self.same_count += 1
3645 def failed(self, src: Path, message: str) -> None:
3646 """Increment the counter for failed reformatting. Write out a message."""
3647 err(f"error: cannot format {src}: {message}")
3648 self.failure_count += 1
3650 def path_ignored(self, path: Path, message: str) -> None:
3652 out(f"{path} ignored: {message}", bold=False)
3655 def return_code(self) -> int:
3656 """Return the exit code that the app should use.
3658 This considers the current state of changed files and failures:
3659 - if there were any failures, return 123;
3660 - if any files were changed and --check is being used, return 1;
3661 - otherwise return 0.
3663 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
3664 # 126 we have special return codes reserved by the shell.
3665 if self.failure_count:
3668 elif self.change_count and self.check:
3673 def __str__(self) -> str:
3674 """Render a color report of the current state.
3676 Use `click.unstyle` to remove colors.
3678 if self.check or self.diff:
3679 reformatted = "would be reformatted"
3680 unchanged = "would be left unchanged"
3681 failed = "would fail to reformat"
3683 reformatted = "reformatted"
3684 unchanged = "left unchanged"
3685 failed = "failed to reformat"
3687 if self.change_count:
3688 s = "s" if self.change_count > 1 else ""
3690 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
3693 s = "s" if self.same_count > 1 else ""
3694 report.append(f"{self.same_count} file{s} {unchanged}")
3695 if self.failure_count:
3696 s = "s" if self.failure_count > 1 else ""
3698 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
3700 return ", ".join(report) + "."
3703 def parse_ast(src: str) -> Union[ast.AST, ast3.AST, ast27.AST]:
3704 filename = "<unknown>"
3705 if sys.version_info >= (3, 8):
3706 # TODO: support Python 4+ ;)
3707 for minor_version in range(sys.version_info[1], 4, -1):
3709 return ast.parse(src, filename, feature_version=(3, minor_version))
3713 for feature_version in (7, 6):
3715 return ast3.parse(src, filename, feature_version=feature_version)
3719 return ast27.parse(src)
3722 def _fixup_ast_constants(
3723 node: Union[ast.AST, ast3.AST, ast27.AST]
3724 ) -> Union[ast.AST, ast3.AST, ast27.AST]:
3725 """Map ast nodes deprecated in 3.8 to Constant."""
3726 if isinstance(node, (ast.Str, ast3.Str, ast27.Str, ast.Bytes, ast3.Bytes)):
3727 return ast.Constant(value=node.s)
3729 if isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
3730 return ast.Constant(value=node.n)
3732 if isinstance(node, (ast.NameConstant, ast3.NameConstant)):
3733 return ast.Constant(value=node.value)
3738 def assert_equivalent(src: str, dst: str) -> None:
3739 """Raise AssertionError if `src` and `dst` aren't equivalent."""
3741 def _v(node: Union[ast.AST, ast3.AST, ast27.AST], depth: int = 0) -> Iterator[str]:
3742 """Simple visitor generating strings to compare ASTs by content."""
3744 node = _fixup_ast_constants(node)
3746 yield f"{' ' * depth}{node.__class__.__name__}("
3748 for field in sorted(node._fields): # noqa: F402
3749 # TypeIgnore has only one field 'lineno' which breaks this comparison
3750 type_ignore_classes = (ast3.TypeIgnore, ast27.TypeIgnore)
3751 if sys.version_info >= (3, 8):
3752 type_ignore_classes += (ast.TypeIgnore,)
3753 if isinstance(node, type_ignore_classes):
3757 value = getattr(node, field)
3758 except AttributeError:
3761 yield f"{' ' * (depth+1)}{field}="
3763 if isinstance(value, list):
3765 # Ignore nested tuples within del statements, because we may insert
3766 # parentheses and they change the AST.
3769 and isinstance(node, (ast.Delete, ast3.Delete, ast27.Delete))
3770 and isinstance(item, (ast.Tuple, ast3.Tuple, ast27.Tuple))
3772 for item in item.elts:
3773 yield from _v(item, depth + 2)
3775 elif isinstance(item, (ast.AST, ast3.AST, ast27.AST)):
3776 yield from _v(item, depth + 2)
3778 elif isinstance(value, (ast.AST, ast3.AST, ast27.AST)):
3779 yield from _v(value, depth + 2)
3782 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
3784 yield f"{' ' * depth}) # /{node.__class__.__name__}"
3787 src_ast = parse_ast(src)
3788 except Exception as exc:
3789 raise AssertionError(
3790 f"cannot use --safe with this file; failed to parse source file. "
3791 f"AST error message: {exc}"
3795 dst_ast = parse_ast(dst)
3796 except Exception as exc:
3797 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
3798 raise AssertionError(
3799 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
3800 f"Please report a bug on https://github.com/psf/black/issues. "
3801 f"This invalid output might be helpful: {log}"
3804 src_ast_str = "\n".join(_v(src_ast))
3805 dst_ast_str = "\n".join(_v(dst_ast))
3806 if src_ast_str != dst_ast_str:
3807 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
3808 raise AssertionError(
3809 f"INTERNAL ERROR: Black produced code that is not equivalent to "
3811 f"Please report a bug on https://github.com/psf/black/issues. "
3812 f"This diff might be helpful: {log}"
3816 def assert_stable(src: str, dst: str, mode: FileMode) -> None:
3817 """Raise AssertionError if `dst` reformats differently the second time."""
3818 newdst = format_str(dst, mode=mode)
3821 diff(src, dst, "source", "first pass"),
3822 diff(dst, newdst, "first pass", "second pass"),
3824 raise AssertionError(
3825 f"INTERNAL ERROR: Black produced different code on the second pass "
3826 f"of the formatter. "
3827 f"Please report a bug on https://github.com/psf/black/issues. "
3828 f"This diff might be helpful: {log}"
3832 @mypyc_attr(patchable=True)
3833 def dump_to_file(*output: str) -> str:
3834 """Dump `output` to a temporary file. Return path to the file."""
3835 with tempfile.NamedTemporaryFile(
3836 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
3838 for lines in output:
3840 if lines and lines[-1] != "\n":
3846 def nullcontext() -> Iterator[None]:
3847 """Return an empty context manager.
3849 To be used like `nullcontext` in Python 3.7.
3854 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
3855 """Return a unified diff string between strings `a` and `b`."""
3858 a_lines = [line + "\n" for line in a.split("\n")]
3859 b_lines = [line + "\n" for line in b.split("\n")]
3861 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
3865 def cancel(tasks: Iterable["asyncio.Task[Any]"]) -> None:
3866 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
3872 def shutdown(loop: asyncio.AbstractEventLoop) -> None:
3873 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
3875 if sys.version_info[:2] >= (3, 7):
3876 all_tasks = asyncio.all_tasks
3878 all_tasks = asyncio.Task.all_tasks
3879 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
3880 to_cancel = [task for task in all_tasks(loop) if not task.done()]
3884 for task in to_cancel:
3886 loop.run_until_complete(
3887 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
3890 # `concurrent.futures.Future` objects cannot be cancelled once they
3891 # are already running. There might be some when the `shutdown()` happened.
3892 # Silence their logger's spew about the event loop being closed.
3893 cf_logger = logging.getLogger("concurrent.futures")
3894 cf_logger.setLevel(logging.CRITICAL)
3898 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
3899 """Replace `regex` with `replacement` twice on `original`.
3901 This is used by string normalization to perform replaces on
3902 overlapping matches.
3904 return regex.sub(replacement, regex.sub(replacement, original))
3907 def re_compile_maybe_verbose(regex: str) -> Pattern[str]:
3908 """Compile a regular expression string in `regex`.
3910 If it contains newlines, use verbose mode.
3913 regex = "(?x)" + regex
3914 compiled: Pattern[str] = re.compile(regex)
3918 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
3919 """Like `reversed(enumerate(sequence))` if that were possible."""
3920 index = len(sequence) - 1
3921 for element in reversed(sequence):
3922 yield (index, element)
3926 def enumerate_with_length(
3927 line: Line, reversed: bool = False
3928 ) -> Iterator[Tuple[Index, Leaf, int]]:
3929 """Return an enumeration of leaves with their length.
3931 Stops prematurely on multiline strings and standalone comments.
3934 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
3935 enumerate_reversed if reversed else enumerate,
3937 for index, leaf in op(line.leaves):
3938 length = len(leaf.prefix) + len(leaf.value)
3939 if "\n" in leaf.value:
3940 return # Multiline strings, we can't continue.
3942 for comment in line.comments_after(leaf):
3943 length += len(comment.value)
3945 yield index, leaf, length
3948 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
3949 """Return True if `line` is no longer than `line_length`.
3951 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
3954 line_str = str(line).strip("\n")
3956 len(line_str) <= line_length
3957 and "\n" not in line_str # multiline strings
3958 and not line.contains_standalone_comments()
3962 def can_be_split(line: Line) -> bool:
3963 """Return False if the line cannot be split *for sure*.
3965 This is not an exhaustive search but a cheap heuristic that we can use to
3966 avoid some unfortunate formattings (mostly around wrapping unsplittable code
3967 in unnecessary parentheses).
3969 leaves = line.leaves
3973 if leaves[0].type == token.STRING and leaves[1].type == token.DOT:
3977 for leaf in leaves[-2::-1]:
3978 if leaf.type in OPENING_BRACKETS:
3979 if next.type not in CLOSING_BRACKETS:
3983 elif leaf.type == token.DOT:
3985 elif leaf.type == token.NAME:
3986 if not (next.type == token.DOT or next.type in OPENING_BRACKETS):
3989 elif leaf.type not in CLOSING_BRACKETS:
3992 if dot_count > 1 and call_count > 1:
3998 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
3999 """Does `line` have a shape safe to reformat without optional parens around it?
4001 Returns True for only a subset of potentially nice looking formattings but
4002 the point is to not return false positives that end up producing lines that
4005 bt = line.bracket_tracker
4006 if not bt.delimiters:
4007 # Without delimiters the optional parentheses are useless.
4010 max_priority = bt.max_delimiter_priority()
4011 if bt.delimiter_count_with_priority(max_priority) > 1:
4012 # With more than one delimiter of a kind the optional parentheses read better.
4015 if max_priority == DOT_PRIORITY:
4016 # A single stranded method call doesn't require optional parentheses.
4019 assert len(line.leaves) >= 2, "Stranded delimiter"
4021 first = line.leaves[0]
4022 second = line.leaves[1]
4023 penultimate = line.leaves[-2]
4024 last = line.leaves[-1]
4026 # With a single delimiter, omit if the expression starts or ends with
4028 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
4030 length = 4 * line.depth
4031 for _index, leaf, leaf_length in enumerate_with_length(line):
4032 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
4035 length += leaf_length
4036 if length > line_length:
4039 if leaf.type in OPENING_BRACKETS:
4040 # There are brackets we can further split on.
4044 # checked the entire string and line length wasn't exceeded
4045 if len(line.leaves) == _index + 1:
4048 # Note: we are not returning False here because a line might have *both*
4049 # a leading opening bracket and a trailing closing bracket. If the
4050 # opening bracket doesn't match our rule, maybe the closing will.
4053 last.type == token.RPAR
4054 or last.type == token.RBRACE
4056 # don't use indexing for omitting optional parentheses;
4058 last.type == token.RSQB
4060 and last.parent.type != syms.trailer
4063 if penultimate.type in OPENING_BRACKETS:
4064 # Empty brackets don't help.
4067 if is_multiline_string(first):
4068 # Additional wrapping of a multiline string in this situation is
4072 length = 4 * line.depth
4073 seen_other_brackets = False
4074 for _index, leaf, leaf_length in enumerate_with_length(line):
4075 length += leaf_length
4076 if leaf is last.opening_bracket:
4077 if seen_other_brackets or length <= line_length:
4080 elif leaf.type in OPENING_BRACKETS:
4081 # There are brackets we can further split on.
4082 seen_other_brackets = True
4087 def get_cache_file(mode: FileMode) -> Path:
4088 return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle"
4091 def read_cache(mode: FileMode) -> Cache:
4092 """Read the cache if it exists and is well formed.
4094 If it is not well formed, the call to write_cache later should resolve the issue.
4096 cache_file = get_cache_file(mode)
4097 if not cache_file.exists():
4100 with cache_file.open("rb") as fobj:
4102 cache: Cache = pickle.load(fobj)
4103 except (pickle.UnpicklingError, ValueError):
4109 def get_cache_info(path: Path) -> CacheInfo:
4110 """Return the information used to check if a file is already formatted or not."""
4112 return stat.st_mtime, stat.st_size
4115 def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]:
4116 """Split an iterable of paths in `sources` into two sets.
4118 The first contains paths of files that modified on disk or are not in the
4119 cache. The other contains paths to non-modified files.
4121 todo, done = set(), set()
4124 if cache.get(src) != get_cache_info(src):
4131 def write_cache(cache: Cache, sources: Iterable[Path], mode: FileMode) -> None:
4132 """Update the cache file."""
4133 cache_file = get_cache_file(mode)
4135 CACHE_DIR.mkdir(parents=True, exist_ok=True)
4136 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
4137 with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
4138 pickle.dump(new_cache, f, protocol=4)
4139 os.replace(f.name, cache_file)
4144 def patch_click() -> None:
4145 """Make Click not crash.
4147 On certain misconfigured environments, Python 3 selects the ASCII encoding as the
4148 default which restricts paths that it can access during the lifetime of the
4149 application. Click refuses to work in this scenario by raising a RuntimeError.
4151 In case of Black the likelihood that non-ASCII characters are going to be used in
4152 file paths is minimal since it's Python source code. Moreover, this crash was
4153 spurious on Python 3.7 thanks to PEP 538 and PEP 540.
4156 from click import core
4157 from click import _unicodefun # type: ignore
4158 except ModuleNotFoundError:
4161 for module in (core, _unicodefun):
4162 if hasattr(module, "_verify_python3_env"):
4163 module._verify_python3_env = lambda: None
4166 def patched_main() -> None:
4172 if __name__ == "__main__":