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 asyncio.base_events import BaseEventLoop
4 from concurrent.futures import Executor, ProcessPoolExecutor
6 from functools import partial, wraps
9 from multiprocessing import Manager
11 from pathlib import Path
36 from appdirs import user_cache_dir
37 from attr import dataclass, Factory
41 from blib2to3.pytree import Node, Leaf, type_repr
42 from blib2to3 import pygram, pytree
43 from blib2to3.pgen2 import driver, token
44 from blib2to3.pgen2.parse import ParseError
47 __version__ = "18.5b0"
48 DEFAULT_LINE_LENGTH = 88
49 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
60 LN = Union[Leaf, Node]
61 SplitFunc = Callable[["Line", bool], Iterator["Line"]]
64 CacheInfo = Tuple[Timestamp, FileSize]
65 Cache = Dict[Path, CacheInfo]
66 out = partial(click.secho, bold=True, err=True)
67 err = partial(click.secho, fg="red", err=True)
69 pygram.initialize(CACHE_DIR)
70 syms = pygram.python_symbols
73 class NothingChanged(UserWarning):
74 """Raised by :func:`format_file` when reformatted code is the same as source."""
77 class CannotSplit(Exception):
78 """A readable split that fits the allotted line length is impossible.
80 Raised by :func:`left_hand_split`, :func:`right_hand_split`, and
81 :func:`delimiter_split`.
85 class FormatError(Exception):
86 """Base exception for `# fmt: on` and `# fmt: off` handling.
88 It holds the number of bytes of the prefix consumed before the format
89 control comment appeared.
92 def __init__(self, consumed: int) -> None:
93 super().__init__(consumed)
94 self.consumed = consumed
96 def trim_prefix(self, leaf: Leaf) -> None:
97 leaf.prefix = leaf.prefix[self.consumed :]
99 def leaf_from_consumed(self, leaf: Leaf) -> Leaf:
100 """Returns a new Leaf from the consumed part of the prefix."""
101 unformatted_prefix = leaf.prefix[: self.consumed]
102 return Leaf(token.NEWLINE, unformatted_prefix)
105 class FormatOn(FormatError):
106 """Found a comment like `# fmt: on` in the file."""
109 class FormatOff(FormatError):
110 """Found a comment like `# fmt: off` in the file."""
113 class WriteBack(Enum):
130 default=DEFAULT_LINE_LENGTH,
131 help="How many character per line to allow.",
138 "Don't write the files back, just return the status. Return code 0 "
139 "means nothing would change. Return code 1 means some files would be "
140 "reformatted. Return code 123 means there was an internal error."
146 help="Don't write the files back, just output a diff for each file on stdout.",
151 help="If --fast given, skip temporary sanity checks. [default: --safe]",
158 "Don't emit non-error messages to stderr. Errors are still emitted, "
159 "silence those with 2>/dev/null."
162 @click.version_option(version=__version__)
167 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
180 """The uncompromising code formatter."""
181 sources: List[Path] = []
185 sources.extend(gen_python_files_in_dir(p))
187 # if a file was explicitly given, we don't care about its extension
190 sources.append(Path("-"))
192 err(f"invalid path: {s}")
194 if check and not diff:
195 write_back = WriteBack.NO
197 write_back = WriteBack.DIFF
199 write_back = WriteBack.YES
200 report = Report(check=check, quiet=quiet)
201 if len(sources) == 0:
202 out("No paths given. Nothing to do 😴")
206 elif len(sources) == 1:
207 reformat_one(sources[0], line_length, fast, write_back, report)
209 loop = asyncio.get_event_loop()
210 executor = ProcessPoolExecutor(max_workers=os.cpu_count())
212 loop.run_until_complete(
214 sources, line_length, fast, write_back, report, loop, executor
220 out("All done! ✨ 🍰 ✨")
221 click.echo(str(report))
222 ctx.exit(report.return_code)
226 src: Path, line_length: int, fast: bool, write_back: WriteBack, report: "Report"
228 """Reformat a single file under `src` without spawning child processes.
230 If `quiet` is True, non-error messages are not output. `line_length`,
231 `write_back`, and `fast` options are passed to :func:`format_file_in_place`.
235 if not src.is_file() and str(src) == "-":
236 if format_stdin_to_stdout(
237 line_length=line_length, fast=fast, write_back=write_back
239 changed = Changed.YES
242 if write_back != WriteBack.DIFF:
243 cache = read_cache(line_length)
245 if src in cache and cache[src] == get_cache_info(src):
246 changed = Changed.CACHED
247 if changed is not Changed.CACHED and format_file_in_place(
248 src, line_length=line_length, fast=fast, write_back=write_back
250 changed = Changed.YES
251 if write_back == WriteBack.YES and changed is not Changed.NO:
252 write_cache(cache, [src], line_length)
253 report.done(src, changed)
254 except Exception as exc:
255 report.failed(src, str(exc))
258 async def schedule_formatting(
262 write_back: WriteBack,
267 """Run formatting of `sources` in parallel using the provided `executor`.
269 (Use ProcessPoolExecutors for actual parallelism.)
271 `line_length`, `write_back`, and `fast` options are passed to
272 :func:`format_file_in_place`.
275 if write_back != WriteBack.DIFF:
276 cache = read_cache(line_length)
277 sources, cached = filter_cached(cache, sources)
279 report.done(src, Changed.CACHED)
284 if write_back == WriteBack.DIFF:
285 # For diff output, we need locks to ensure we don't interleave output
286 # from different processes.
288 lock = manager.Lock()
290 loop.run_in_executor(
291 executor, format_file_in_place, src, line_length, fast, write_back, lock
293 for src in sorted(sources)
295 pending: Iterable[asyncio.Task] = tasks.keys()
297 loop.add_signal_handler(signal.SIGINT, cancel, pending)
298 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
299 except NotImplementedError:
300 # There are no good alternatives for these on Windows
303 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
305 src = tasks.pop(task)
307 cancelled.append(task)
308 elif task.exception():
309 report.failed(src, str(task.exception()))
311 formatted.append(src)
312 report.done(src, Changed.YES if task.result() else Changed.NO)
314 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
315 if write_back == WriteBack.YES and formatted:
316 write_cache(cache, formatted, line_length)
319 def format_file_in_place(
323 write_back: WriteBack = WriteBack.NO,
324 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
326 """Format file under `src` path. Return True if changed.
328 If `write_back` is True, write reformatted code back to stdout.
329 `line_length` and `fast` options are passed to :func:`format_file_contents`.
331 is_pyi = src.suffix == ".pyi"
333 with tokenize.open(src) as src_buffer:
334 src_contents = src_buffer.read()
336 dst_contents = format_file_contents(
337 src_contents, line_length=line_length, fast=fast, is_pyi=is_pyi
339 except NothingChanged:
342 if write_back == write_back.YES:
343 with open(src, "w", encoding=src_buffer.encoding) as f:
344 f.write(dst_contents)
345 elif write_back == write_back.DIFF:
346 src_name = f"{src} (original)"
347 dst_name = f"{src} (formatted)"
348 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
352 sys.stdout.write(diff_contents)
359 def format_stdin_to_stdout(
360 line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO
362 """Format file on stdin. Return True if changed.
364 If `write_back` is True, write reformatted code back to stdout.
365 `line_length` and `fast` arguments are passed to :func:`format_file_contents`.
367 src = sys.stdin.read()
370 dst = format_file_contents(src, line_length=line_length, fast=fast)
373 except NothingChanged:
377 if write_back == WriteBack.YES:
378 sys.stdout.write(dst)
379 elif write_back == WriteBack.DIFF:
380 src_name = "<stdin> (original)"
381 dst_name = "<stdin> (formatted)"
382 sys.stdout.write(diff(src, dst, src_name, dst_name))
385 def format_file_contents(
386 src_contents: str, *, line_length: int, fast: bool, is_pyi: bool = False
388 """Reformat contents a file and return new contents.
390 If `fast` is False, additionally confirm that the reformatted code is
391 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
392 `line_length` is passed to :func:`format_str`.
394 if src_contents.strip() == "":
397 dst_contents = format_str(src_contents, line_length=line_length, is_pyi=is_pyi)
398 if src_contents == dst_contents:
402 assert_equivalent(src_contents, dst_contents)
404 src_contents, dst_contents, line_length=line_length, is_pyi=is_pyi
410 src_contents: str, line_length: int, *, is_pyi: bool = False
412 """Reformat a string and return new contents.
414 `line_length` determines how many characters per line are allowed.
416 src_node = lib2to3_parse(src_contents)
418 future_imports = get_future_imports(src_node)
419 elt = EmptyLineTracker(is_pyi=is_pyi)
420 py36 = is_python36(src_node)
421 lines = LineGenerator(
422 remove_u_prefix=py36 or "unicode_literals" in future_imports, is_pyi=is_pyi
426 for current_line in lines.visit(src_node):
427 for _ in range(after):
428 dst_contents += str(empty_line)
429 before, after = elt.maybe_empty_lines(current_line)
430 for _ in range(before):
431 dst_contents += str(empty_line)
432 for line in split_line(current_line, line_length=line_length, py36=py36):
433 dst_contents += str(line)
438 pygram.python_grammar_no_print_statement_no_exec_statement,
439 pygram.python_grammar_no_print_statement,
440 pygram.python_grammar,
444 def lib2to3_parse(src_txt: str) -> Node:
445 """Given a string with source, return the lib2to3 Node."""
446 grammar = pygram.python_grammar_no_print_statement
447 if src_txt[-1] != "\n":
448 nl = "\r\n" if "\r\n" in src_txt[:1024] else "\n"
450 for grammar in GRAMMARS:
451 drv = driver.Driver(grammar, pytree.convert)
453 result = drv.parse_string(src_txt, True)
456 except ParseError as pe:
457 lineno, column = pe.context[1]
458 lines = src_txt.splitlines()
460 faulty_line = lines[lineno - 1]
462 faulty_line = "<line number missing in source>"
463 exc = ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}")
467 if isinstance(result, Leaf):
468 result = Node(syms.file_input, [result])
472 def lib2to3_unparse(node: Node) -> str:
473 """Given a lib2to3 node, return its string representation."""
481 class Visitor(Generic[T]):
482 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
484 def visit(self, node: LN) -> Iterator[T]:
485 """Main method to visit `node` and its children.
487 It tries to find a `visit_*()` method for the given `node.type`, like
488 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
489 If no dedicated `visit_*()` method is found, chooses `visit_default()`
492 Then yields objects of type `T` from the selected visitor.
495 name = token.tok_name[node.type]
497 name = type_repr(node.type)
498 yield from getattr(self, f"visit_{name}", self.visit_default)(node)
500 def visit_default(self, node: LN) -> Iterator[T]:
501 """Default `visit_*()` implementation. Recurses to children of `node`."""
502 if isinstance(node, Node):
503 for child in node.children:
504 yield from self.visit(child)
508 class DebugVisitor(Visitor[T]):
511 def visit_default(self, node: LN) -> Iterator[T]:
512 indent = " " * (2 * self.tree_depth)
513 if isinstance(node, Node):
514 _type = type_repr(node.type)
515 out(f"{indent}{_type}", fg="yellow")
517 for child in node.children:
518 yield from self.visit(child)
521 out(f"{indent}/{_type}", fg="yellow", bold=False)
523 _type = token.tok_name.get(node.type, str(node.type))
524 out(f"{indent}{_type}", fg="blue", nl=False)
526 # We don't have to handle prefixes for `Node` objects since
527 # that delegates to the first child anyway.
528 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
529 out(f" {node.value!r}", fg="blue", bold=False)
532 def show(cls, code: str) -> None:
533 """Pretty-print the lib2to3 AST of a given string of `code`.
535 Convenience method for debugging.
537 v: DebugVisitor[None] = DebugVisitor()
538 list(v.visit(lib2to3_parse(code)))
541 KEYWORDS = set(keyword.kwlist)
542 WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE}
543 FLOW_CONTROL = {"return", "raise", "break", "continue"}
554 STANDALONE_COMMENT = 153
555 LOGIC_OPERATORS = {"and", "or"}
580 STARS = {token.STAR, token.DOUBLESTAR}
583 syms.argument, # double star in arglist
584 syms.trailer, # single argument to call
586 syms.varargslist, # lambdas
588 UNPACKING_PARENTS = {
589 syms.atom, # single element of a list or set literal
627 COMPREHENSION_PRIORITY = 20
629 TERNARY_PRIORITY = 16
632 COMPARATOR_PRIORITY = 10
643 token.DOUBLESLASH: 4,
653 class BracketTracker:
654 """Keeps track of brackets on a line."""
657 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
658 delimiters: Dict[LeafID, Priority] = Factory(dict)
659 previous: Optional[Leaf] = None
660 _for_loop_variable: int = 0
661 _lambda_arguments: int = 0
663 def mark(self, leaf: Leaf) -> None:
664 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
666 All leaves receive an int `bracket_depth` field that stores how deep
667 within brackets a given leaf is. 0 means there are no enclosing brackets
668 that started on this line.
670 If a leaf is itself a closing bracket, it receives an `opening_bracket`
671 field that it forms a pair with. This is a one-directional link to
672 avoid reference cycles.
674 If a leaf is a delimiter (a token on which Black can split the line if
675 needed) and it's on depth 0, its `id()` is stored in the tracker's
678 if leaf.type == token.COMMENT:
681 self.maybe_decrement_after_for_loop_variable(leaf)
682 self.maybe_decrement_after_lambda_arguments(leaf)
683 if leaf.type in CLOSING_BRACKETS:
685 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
686 leaf.opening_bracket = opening_bracket
687 leaf.bracket_depth = self.depth
689 delim = is_split_before_delimiter(leaf, self.previous)
690 if delim and self.previous is not None:
691 self.delimiters[id(self.previous)] = delim
693 delim = is_split_after_delimiter(leaf, self.previous)
695 self.delimiters[id(leaf)] = delim
696 if leaf.type in OPENING_BRACKETS:
697 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
700 self.maybe_increment_lambda_arguments(leaf)
701 self.maybe_increment_for_loop_variable(leaf)
703 def any_open_brackets(self) -> bool:
704 """Return True if there is an yet unmatched open bracket on the line."""
705 return bool(self.bracket_match)
707 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int:
708 """Return the highest priority of a delimiter found on the line.
710 Values are consistent with what `is_split_*_delimiter()` return.
711 Raises ValueError on no delimiters.
713 return max(v for k, v in self.delimiters.items() if k not in exclude)
715 def delimiter_count_with_priority(self, priority: int = 0) -> int:
716 """Return the number of delimiters with the given `priority`.
718 If no `priority` is passed, defaults to max priority on the line.
720 if not self.delimiters:
723 priority = priority or self.max_delimiter_priority()
724 return sum(1 for p in self.delimiters.values() if p == priority)
726 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
727 """In a for loop, or comprehension, the variables are often unpacks.
729 To avoid splitting on the comma in this situation, increase the depth of
730 tokens between `for` and `in`.
732 if leaf.type == token.NAME and leaf.value == "for":
734 self._for_loop_variable += 1
739 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
740 """See `maybe_increment_for_loop_variable` above for explanation."""
741 if self._for_loop_variable and leaf.type == token.NAME and leaf.value == "in":
743 self._for_loop_variable -= 1
748 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
749 """In a lambda expression, there might be more than one argument.
751 To avoid splitting on the comma in this situation, increase the depth of
752 tokens between `lambda` and `:`.
754 if leaf.type == token.NAME and leaf.value == "lambda":
756 self._lambda_arguments += 1
761 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
762 """See `maybe_increment_lambda_arguments` above for explanation."""
763 if self._lambda_arguments and leaf.type == token.COLON:
765 self._lambda_arguments -= 1
770 def get_open_lsqb(self) -> Optional[Leaf]:
771 """Return the most recent opening square bracket (if any)."""
772 return self.bracket_match.get((self.depth - 1, token.RSQB))
777 """Holds leaves and comments. Can be printed with `str(line)`."""
780 leaves: List[Leaf] = Factory(list)
781 comments: List[Tuple[Index, Leaf]] = Factory(list)
782 bracket_tracker: BracketTracker = Factory(BracketTracker)
783 inside_brackets: bool = False
784 should_explode: bool = False
786 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
787 """Add a new `leaf` to the end of the line.
789 Unless `preformatted` is True, the `leaf` will receive a new consistent
790 whitespace prefix and metadata applied by :class:`BracketTracker`.
791 Trailing commas are maybe removed, unpacked for loop variables are
792 demoted from being delimiters.
794 Inline comments are put aside.
796 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
800 if token.COLON == leaf.type and self.is_class_paren_empty:
802 if self.leaves and not preformatted:
803 # Note: at this point leaf.prefix should be empty except for
804 # imports, for which we only preserve newlines.
805 leaf.prefix += whitespace(
806 leaf, complex_subscript=self.is_complex_subscript(leaf)
808 if self.inside_brackets or not preformatted:
809 self.bracket_tracker.mark(leaf)
810 self.maybe_remove_trailing_comma(leaf)
811 if not self.append_comment(leaf):
812 self.leaves.append(leaf)
814 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
815 """Like :func:`append()` but disallow invalid standalone comment structure.
817 Raises ValueError when any `leaf` is appended after a standalone comment
818 or when a standalone comment is not the first leaf on the line.
820 if self.bracket_tracker.depth == 0:
822 raise ValueError("cannot append to standalone comments")
824 if self.leaves and leaf.type == STANDALONE_COMMENT:
826 "cannot append standalone comments to a populated line"
829 self.append(leaf, preformatted=preformatted)
832 def is_comment(self) -> bool:
833 """Is this line a standalone comment?"""
834 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
837 def is_decorator(self) -> bool:
838 """Is this line a decorator?"""
839 return bool(self) and self.leaves[0].type == token.AT
842 def is_import(self) -> bool:
843 """Is this an import line?"""
844 return bool(self) and is_import(self.leaves[0])
847 def is_class(self) -> bool:
848 """Is this line a class definition?"""
851 and self.leaves[0].type == token.NAME
852 and self.leaves[0].value == "class"
856 def is_stub_class(self) -> bool:
857 """Is this line a class definition with a body consisting only of "..."?"""
858 return self.is_class and self.leaves[-3:] == [
859 Leaf(token.DOT, ".") for _ in range(3)
863 def is_def(self) -> bool:
864 """Is this a function definition? (Also returns True for async defs.)"""
866 first_leaf = self.leaves[0]
871 second_leaf: Optional[Leaf] = self.leaves[1]
874 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
875 first_leaf.type == token.ASYNC
876 and second_leaf is not None
877 and second_leaf.type == token.NAME
878 and second_leaf.value == "def"
882 def is_flow_control(self) -> bool:
883 """Is this line a flow control statement?
885 Those are `return`, `raise`, `break`, and `continue`.
889 and self.leaves[0].type == token.NAME
890 and self.leaves[0].value in FLOW_CONTROL
894 def is_yield(self) -> bool:
895 """Is this line a yield statement?"""
898 and self.leaves[0].type == token.NAME
899 and self.leaves[0].value == "yield"
903 def is_class_paren_empty(self) -> bool:
904 """Is this a class with no base classes but using parentheses?
906 Those are unnecessary and should be removed.
910 and len(self.leaves) == 4
912 and self.leaves[2].type == token.LPAR
913 and self.leaves[2].value == "("
914 and self.leaves[3].type == token.RPAR
915 and self.leaves[3].value == ")"
918 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
919 """If so, needs to be split before emitting."""
920 for leaf in self.leaves:
921 if leaf.type == STANDALONE_COMMENT:
922 if leaf.bracket_depth <= depth_limit:
927 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
928 """Remove trailing comma if there is one and it's safe."""
931 and self.leaves[-1].type == token.COMMA
932 and closing.type in CLOSING_BRACKETS
936 if closing.type == token.RBRACE:
937 self.remove_trailing_comma()
940 if closing.type == token.RSQB:
941 comma = self.leaves[-1]
942 if comma.parent and comma.parent.type == syms.listmaker:
943 self.remove_trailing_comma()
946 # For parens let's check if it's safe to remove the comma.
947 # Imports are always safe.
949 self.remove_trailing_comma()
952 # Otheriwsse, if the trailing one is the only one, we might mistakenly
953 # change a tuple into a different type by removing the comma.
954 depth = closing.bracket_depth + 1
956 opening = closing.opening_bracket
957 for _opening_index, leaf in enumerate(self.leaves):
964 for leaf in self.leaves[_opening_index + 1 :]:
968 bracket_depth = leaf.bracket_depth
969 if bracket_depth == depth and leaf.type == token.COMMA:
971 if leaf.parent and leaf.parent.type == syms.arglist:
976 self.remove_trailing_comma()
981 def append_comment(self, comment: Leaf) -> bool:
982 """Add an inline or standalone comment to the line."""
984 comment.type == STANDALONE_COMMENT
985 and self.bracket_tracker.any_open_brackets()
990 if comment.type != token.COMMENT:
993 after = len(self.leaves) - 1
995 comment.type = STANDALONE_COMMENT
1000 self.comments.append((after, comment))
1003 def comments_after(self, leaf: Leaf, _index: int = -1) -> Iterator[Leaf]:
1004 """Generate comments that should appear directly after `leaf`.
1006 Provide a non-negative leaf `_index` to speed up the function.
1009 for _index, _leaf in enumerate(self.leaves):
1016 for index, comment_after in self.comments:
1020 def remove_trailing_comma(self) -> None:
1021 """Remove the trailing comma and moves the comments attached to it."""
1022 comma_index = len(self.leaves) - 1
1023 for i in range(len(self.comments)):
1024 comment_index, comment = self.comments[i]
1025 if comment_index == comma_index:
1026 self.comments[i] = (comma_index - 1, comment)
1029 def is_complex_subscript(self, leaf: Leaf) -> bool:
1030 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1032 leaf if leaf.type == token.LSQB else self.bracket_tracker.get_open_lsqb()
1034 if open_lsqb is None:
1037 subscript_start = open_lsqb.next_sibling
1039 isinstance(subscript_start, Node)
1040 and subscript_start.type == syms.subscriptlist
1042 subscript_start = child_towards(subscript_start, leaf)
1043 return subscript_start is not None and any(
1044 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1047 def __str__(self) -> str:
1048 """Render the line."""
1052 indent = " " * self.depth
1053 leaves = iter(self.leaves)
1054 first = next(leaves)
1055 res = f"{first.prefix}{indent}{first.value}"
1058 for _, comment in self.comments:
1062 def __bool__(self) -> bool:
1063 """Return True if the line has leaves or comments."""
1064 return bool(self.leaves or self.comments)
1067 class UnformattedLines(Line):
1068 """Just like :class:`Line` but stores lines which aren't reformatted."""
1070 def append(self, leaf: Leaf, preformatted: bool = True) -> None:
1071 """Just add a new `leaf` to the end of the lines.
1073 The `preformatted` argument is ignored.
1075 Keeps track of indentation `depth`, which is useful when the user
1076 says `# fmt: on`. Otherwise, doesn't do anything with the `leaf`.
1079 list(generate_comments(leaf))
1080 except FormatOn as f_on:
1081 self.leaves.append(f_on.leaf_from_consumed(leaf))
1084 self.leaves.append(leaf)
1085 if leaf.type == token.INDENT:
1087 elif leaf.type == token.DEDENT:
1090 def __str__(self) -> str:
1091 """Render unformatted lines from leaves which were added with `append()`.
1093 `depth` is not used for indentation in this case.
1099 for leaf in self.leaves:
1103 def append_comment(self, comment: Leaf) -> bool:
1104 """Not implemented in this class. Raises `NotImplementedError`."""
1105 raise NotImplementedError("Unformatted lines don't store comments separately.")
1107 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1108 """Does nothing and returns False."""
1111 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1112 """Does nothing and returns False."""
1117 class EmptyLineTracker:
1118 """Provides a stateful method that returns the number of potential extra
1119 empty lines needed before and after the currently processed line.
1121 Note: this tracker works on lines that haven't been split yet. It assumes
1122 the prefix of the first leaf consists of optional newlines. Those newlines
1123 are consumed by `maybe_empty_lines()` and included in the computation.
1125 is_pyi: bool = False
1126 previous_line: Optional[Line] = None
1127 previous_after: int = 0
1128 previous_defs: List[int] = Factory(list)
1130 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1131 """Return the number of extra empty lines before and after the `current_line`.
1133 This is for separating `def`, `async def` and `class` with extra empty
1134 lines (two on module-level), as well as providing an extra empty line
1135 after flow control keywords to make them more prominent.
1137 if isinstance(current_line, UnformattedLines):
1140 before, after = self._maybe_empty_lines(current_line)
1141 before -= self.previous_after
1142 self.previous_after = after
1143 self.previous_line = current_line
1144 return before, after
1146 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1148 if current_line.depth == 0:
1149 max_allowed = 1 if self.is_pyi else 2
1150 if current_line.leaves:
1151 # Consume the first leaf's extra newlines.
1152 first_leaf = current_line.leaves[0]
1153 before = first_leaf.prefix.count("\n")
1154 before = min(before, max_allowed)
1155 first_leaf.prefix = ""
1158 depth = current_line.depth
1159 while self.previous_defs and self.previous_defs[-1] >= depth:
1160 self.previous_defs.pop()
1162 before = 0 if depth else 1
1164 before = 1 if depth else 2
1165 is_decorator = current_line.is_decorator
1166 if is_decorator or current_line.is_def or current_line.is_class:
1167 if not is_decorator:
1168 self.previous_defs.append(depth)
1169 if self.previous_line is None:
1170 # Don't insert empty lines before the first line in the file.
1173 if self.previous_line.is_decorator:
1177 self.previous_line.is_comment
1178 and self.previous_line.depth == current_line.depth
1184 if self.previous_line.depth > current_line.depth:
1186 elif current_line.is_class or self.previous_line.is_class:
1187 if current_line.is_stub_class and self.previous_line.is_stub_class:
1195 if current_line.depth and newlines:
1201 and self.previous_line.is_import
1202 and not current_line.is_import
1203 and depth == self.previous_line.depth
1205 return (before or 1), 0
1211 class LineGenerator(Visitor[Line]):
1212 """Generates reformatted Line objects. Empty lines are not emitted.
1214 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1215 in ways that will no longer stringify to valid Python code on the tree.
1217 is_pyi: bool = False
1218 current_line: Line = Factory(Line)
1219 remove_u_prefix: bool = False
1221 def line(self, indent: int = 0, type: Type[Line] = Line) -> Iterator[Line]:
1224 If the line is empty, only emit if it makes sense.
1225 If the line is too long, split it first and then generate.
1227 If any lines were generated, set up a new current_line.
1229 if not self.current_line:
1230 if self.current_line.__class__ == type:
1231 self.current_line.depth += indent
1233 self.current_line = type(depth=self.current_line.depth + indent)
1234 return # Line is empty, don't emit. Creating a new one unnecessary.
1236 complete_line = self.current_line
1237 self.current_line = type(depth=complete_line.depth + indent)
1240 def visit(self, node: LN) -> Iterator[Line]:
1241 """Main method to visit `node` and its children.
1243 Yields :class:`Line` objects.
1245 if isinstance(self.current_line, UnformattedLines):
1246 # File contained `# fmt: off`
1247 yield from self.visit_unformatted(node)
1250 yield from super().visit(node)
1252 def visit_default(self, node: LN) -> Iterator[Line]:
1253 """Default `visit_*()` implementation. Recurses to children of `node`."""
1254 if isinstance(node, Leaf):
1255 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1257 for comment in generate_comments(node):
1258 if any_open_brackets:
1259 # any comment within brackets is subject to splitting
1260 self.current_line.append(comment)
1261 elif comment.type == token.COMMENT:
1262 # regular trailing comment
1263 self.current_line.append(comment)
1264 yield from self.line()
1267 # regular standalone comment
1268 yield from self.line()
1270 self.current_line.append(comment)
1271 yield from self.line()
1273 except FormatOff as f_off:
1274 f_off.trim_prefix(node)
1275 yield from self.line(type=UnformattedLines)
1276 yield from self.visit(node)
1278 except FormatOn as f_on:
1279 # This only happens here if somebody says "fmt: on" multiple
1281 f_on.trim_prefix(node)
1282 yield from self.visit_default(node)
1285 normalize_prefix(node, inside_brackets=any_open_brackets)
1286 if node.type == token.STRING:
1287 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1288 normalize_string_quotes(node)
1289 if node.type not in WHITESPACE:
1290 self.current_line.append(node)
1291 yield from super().visit_default(node)
1293 def visit_INDENT(self, node: Node) -> Iterator[Line]:
1294 """Increase indentation level, maybe yield a line."""
1295 # In blib2to3 INDENT never holds comments.
1296 yield from self.line(+1)
1297 yield from self.visit_default(node)
1299 def visit_DEDENT(self, node: Node) -> Iterator[Line]:
1300 """Decrease indentation level, maybe yield a line."""
1301 # The current line might still wait for trailing comments. At DEDENT time
1302 # there won't be any (they would be prefixes on the preceding NEWLINE).
1303 # Emit the line then.
1304 yield from self.line()
1306 # While DEDENT has no value, its prefix may contain standalone comments
1307 # that belong to the current indentation level. Get 'em.
1308 yield from self.visit_default(node)
1310 # Finally, emit the dedent.
1311 yield from self.line(-1)
1314 self, node: Node, keywords: Set[str], parens: Set[str]
1315 ) -> Iterator[Line]:
1316 """Visit a statement.
1318 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1319 `def`, `with`, `class`, `assert` and assignments.
1321 The relevant Python language `keywords` for a given statement will be
1322 NAME leaves within it. This methods puts those on a separate line.
1324 `parens` holds a set of string leaf values immediately after which
1325 invisible parens should be put.
1327 normalize_invisible_parens(node, parens_after=parens)
1328 for child in node.children:
1329 if child.type == token.NAME and child.value in keywords: # type: ignore
1330 yield from self.line()
1332 yield from self.visit(child)
1334 def visit_suite(self, node: Node) -> Iterator[Line]:
1335 """Visit a suite."""
1336 if self.is_pyi and is_stub_suite(node):
1337 yield from self.visit(node.children[2])
1339 yield from self.visit_default(node)
1341 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1342 """Visit a statement without nested statements."""
1343 is_suite_like = node.parent and node.parent.type in STATEMENT
1345 if self.is_pyi and is_stub_body(node):
1346 yield from self.visit_default(node)
1348 yield from self.line(+1)
1349 yield from self.visit_default(node)
1350 yield from self.line(-1)
1353 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1354 yield from self.line()
1355 yield from self.visit_default(node)
1357 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1358 """Visit `async def`, `async for`, `async with`."""
1359 yield from self.line()
1361 children = iter(node.children)
1362 for child in children:
1363 yield from self.visit(child)
1365 if child.type == token.ASYNC:
1368 internal_stmt = next(children)
1369 for child in internal_stmt.children:
1370 yield from self.visit(child)
1372 def visit_decorators(self, node: Node) -> Iterator[Line]:
1373 """Visit decorators."""
1374 for child in node.children:
1375 yield from self.line()
1376 yield from self.visit(child)
1378 def visit_import_from(self, node: Node) -> Iterator[Line]:
1379 """Visit import_from and maybe put invisible parentheses.
1381 This is separate from `visit_stmt` because import statements don't
1382 support arbitrary atoms and thus handling of parentheses is custom.
1385 for index, child in enumerate(node.children):
1387 if child.type == token.LPAR:
1388 # make parentheses invisible
1389 child.value = "" # type: ignore
1390 node.children[-1].value = "" # type: ignore
1392 # insert invisible parentheses
1393 node.insert_child(index, Leaf(token.LPAR, ""))
1394 node.append_child(Leaf(token.RPAR, ""))
1398 child.type == token.NAME and child.value == "import" # type: ignore
1401 for child in node.children:
1402 yield from self.visit(child)
1404 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1405 """Remove a semicolon and put the other statement on a separate line."""
1406 yield from self.line()
1408 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1409 """End of file. Process outstanding comments and end with a newline."""
1410 yield from self.visit_default(leaf)
1411 yield from self.line()
1413 def visit_unformatted(self, node: LN) -> Iterator[Line]:
1414 """Used when file contained a `# fmt: off`."""
1415 if isinstance(node, Node):
1416 for child in node.children:
1417 yield from self.visit(child)
1421 self.current_line.append(node)
1422 except FormatOn as f_on:
1423 f_on.trim_prefix(node)
1424 yield from self.line()
1425 yield from self.visit(node)
1427 if node.type == token.ENDMARKER:
1428 # somebody decided not to put a final `# fmt: on`
1429 yield from self.line()
1431 def __attrs_post_init__(self) -> None:
1432 """You are in a twisty little maze of passages."""
1435 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1436 self.visit_if_stmt = partial(
1437 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1439 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1440 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1441 self.visit_try_stmt = partial(
1442 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1444 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1445 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1446 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1447 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1448 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1449 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1450 self.visit_async_funcdef = self.visit_async_stmt
1451 self.visit_decorated = self.visit_decorators
1454 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1455 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1456 OPENING_BRACKETS = set(BRACKET.keys())
1457 CLOSING_BRACKETS = set(BRACKET.values())
1458 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1459 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1462 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa C901
1463 """Return whitespace prefix if needed for the given `leaf`.
1465 `complex_subscript` signals whether the given leaf is part of a subscription
1466 which has non-trivial arguments, like arithmetic expressions or function calls.
1474 if t in ALWAYS_NO_SPACE:
1477 if t == token.COMMENT:
1480 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1481 if t == token.COLON and p.type not in {
1488 prev = leaf.prev_sibling
1490 prevp = preceding_leaf(p)
1491 if not prevp or prevp.type in OPENING_BRACKETS:
1494 if t == token.COLON:
1495 if prevp.type == token.COLON:
1498 elif prevp.type != token.COMMA and not complex_subscript:
1503 if prevp.type == token.EQUAL:
1505 if prevp.parent.type in {
1513 elif prevp.parent.type == syms.typedargslist:
1514 # A bit hacky: if the equal sign has whitespace, it means we
1515 # previously found it's a typed argument. So, we're using
1519 elif prevp.type in STARS:
1520 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1523 elif prevp.type == token.COLON:
1524 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1525 return SPACE if complex_subscript else NO
1529 and prevp.parent.type == syms.factor
1530 and prevp.type in MATH_OPERATORS
1535 prevp.type == token.RIGHTSHIFT
1537 and prevp.parent.type == syms.shift_expr
1538 and prevp.prev_sibling
1539 and prevp.prev_sibling.type == token.NAME
1540 and prevp.prev_sibling.value == "print" # type: ignore
1542 # Python 2 print chevron
1545 elif prev.type in OPENING_BRACKETS:
1548 if p.type in {syms.parameters, syms.arglist}:
1549 # untyped function signatures or calls
1550 if not prev or prev.type != token.COMMA:
1553 elif p.type == syms.varargslist:
1555 if prev and prev.type != token.COMMA:
1558 elif p.type == syms.typedargslist:
1559 # typed function signatures
1563 if t == token.EQUAL:
1564 if prev.type != syms.tname:
1567 elif prev.type == token.EQUAL:
1568 # A bit hacky: if the equal sign has whitespace, it means we
1569 # previously found it's a typed argument. So, we're using that, too.
1572 elif prev.type != token.COMMA:
1575 elif p.type == syms.tname:
1578 prevp = preceding_leaf(p)
1579 if not prevp or prevp.type != token.COMMA:
1582 elif p.type == syms.trailer:
1583 # attributes and calls
1584 if t == token.LPAR or t == token.RPAR:
1589 prevp = preceding_leaf(p)
1590 if not prevp or prevp.type != token.NUMBER:
1593 elif t == token.LSQB:
1596 elif prev.type != token.COMMA:
1599 elif p.type == syms.argument:
1601 if t == token.EQUAL:
1605 prevp = preceding_leaf(p)
1606 if not prevp or prevp.type == token.LPAR:
1609 elif prev.type in {token.EQUAL} | STARS:
1612 elif p.type == syms.decorator:
1616 elif p.type == syms.dotted_name:
1620 prevp = preceding_leaf(p)
1621 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
1624 elif p.type == syms.classdef:
1628 if prev and prev.type == token.LPAR:
1631 elif p.type in {syms.subscript, syms.sliceop}:
1634 assert p.parent is not None, "subscripts are always parented"
1635 if p.parent.type == syms.subscriptlist:
1640 elif not complex_subscript:
1643 elif p.type == syms.atom:
1644 if prev and t == token.DOT:
1645 # dots, but not the first one.
1648 elif p.type == syms.dictsetmaker:
1650 if prev and prev.type == token.DOUBLESTAR:
1653 elif p.type in {syms.factor, syms.star_expr}:
1656 prevp = preceding_leaf(p)
1657 if not prevp or prevp.type in OPENING_BRACKETS:
1660 prevp_parent = prevp.parent
1661 assert prevp_parent is not None
1662 if prevp.type == token.COLON and prevp_parent.type in {
1668 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
1671 elif t == token.NAME or t == token.NUMBER:
1674 elif p.type == syms.import_from:
1676 if prev and prev.type == token.DOT:
1679 elif t == token.NAME:
1683 if prev and prev.type == token.DOT:
1686 elif p.type == syms.sliceop:
1692 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
1693 """Return the first leaf that precedes `node`, if any."""
1695 res = node.prev_sibling
1697 if isinstance(res, Leaf):
1701 return list(res.leaves())[-1]
1710 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
1711 """Return the child of `ancestor` that contains `descendant`."""
1712 node: Optional[LN] = descendant
1713 while node and node.parent != ancestor:
1718 def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1719 """Return the priority of the `leaf` delimiter, given a line break after it.
1721 The delimiter priorities returned here are from those delimiters that would
1722 cause a line break after themselves.
1724 Higher numbers are higher priority.
1726 if leaf.type == token.COMMA:
1727 return COMMA_PRIORITY
1732 def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1733 """Return the priority of the `leaf` delimiter, given a line before after it.
1735 The delimiter priorities returned here are from those delimiters that would
1736 cause a line break before themselves.
1738 Higher numbers are higher priority.
1740 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1741 # * and ** might also be MATH_OPERATORS but in this case they are not.
1742 # Don't treat them as a delimiter.
1746 leaf.type == token.DOT
1748 and leaf.parent.type not in {syms.import_from, syms.dotted_name}
1749 and (previous is None or previous.type in CLOSING_BRACKETS)
1754 leaf.type in MATH_OPERATORS
1756 and leaf.parent.type not in {syms.factor, syms.star_expr}
1758 return MATH_PRIORITIES[leaf.type]
1760 if leaf.type in COMPARATORS:
1761 return COMPARATOR_PRIORITY
1764 leaf.type == token.STRING
1765 and previous is not None
1766 and previous.type == token.STRING
1768 return STRING_PRIORITY
1770 if leaf.type != token.NAME:
1776 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
1778 return COMPREHENSION_PRIORITY
1783 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
1785 return COMPREHENSION_PRIORITY
1787 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
1788 return TERNARY_PRIORITY
1790 if leaf.value == "is":
1791 return COMPARATOR_PRIORITY
1796 and leaf.parent.type in {syms.comp_op, syms.comparison}
1798 previous is not None
1799 and previous.type == token.NAME
1800 and previous.value == "not"
1803 return COMPARATOR_PRIORITY
1808 and leaf.parent.type == syms.comp_op
1810 previous is not None
1811 and previous.type == token.NAME
1812 and previous.value == "is"
1815 return COMPARATOR_PRIORITY
1817 if leaf.value in LOGIC_OPERATORS and leaf.parent:
1818 return LOGIC_PRIORITY
1823 def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
1824 """Clean the prefix of the `leaf` and generate comments from it, if any.
1826 Comments in lib2to3 are shoved into the whitespace prefix. This happens
1827 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
1828 move because it does away with modifying the grammar to include all the
1829 possible places in which comments can be placed.
1831 The sad consequence for us though is that comments don't "belong" anywhere.
1832 This is why this function generates simple parentless Leaf objects for
1833 comments. We simply don't know what the correct parent should be.
1835 No matter though, we can live without this. We really only need to
1836 differentiate between inline and standalone comments. The latter don't
1837 share the line with any code.
1839 Inline comments are emitted as regular token.COMMENT leaves. Standalone
1840 are emitted with a fake STANDALONE_COMMENT token identifier.
1851 for index, line in enumerate(p.split("\n")):
1852 consumed += len(line) + 1 # adding the length of the split '\n'
1853 line = line.lstrip()
1856 if not line.startswith("#"):
1859 if index == 0 and leaf.type != token.ENDMARKER:
1860 comment_type = token.COMMENT # simple trailing comment
1862 comment_type = STANDALONE_COMMENT
1863 comment = make_comment(line)
1864 yield Leaf(comment_type, comment, prefix="\n" * nlines)
1866 if comment in {"# fmt: on", "# yapf: enable"}:
1867 raise FormatOn(consumed)
1869 if comment in {"# fmt: off", "# yapf: disable"}:
1870 if comment_type == STANDALONE_COMMENT:
1871 raise FormatOff(consumed)
1873 prev = preceding_leaf(leaf)
1874 if not prev or prev.type in WHITESPACE: # standalone comment in disguise
1875 raise FormatOff(consumed)
1880 def make_comment(content: str) -> str:
1881 """Return a consistently formatted comment from the given `content` string.
1883 All comments (except for "##", "#!", "#:") should have a single space between
1884 the hash sign and the content.
1886 If `content` didn't start with a hash sign, one is provided.
1888 content = content.rstrip()
1892 if content[0] == "#":
1893 content = content[1:]
1894 if content and content[0] not in " !:#":
1895 content = " " + content
1896 return "#" + content
1900 line: Line, line_length: int, inner: bool = False, py36: bool = False
1901 ) -> Iterator[Line]:
1902 """Split a `line` into potentially many lines.
1904 They should fit in the allotted `line_length` but might not be able to.
1905 `inner` signifies that there were a pair of brackets somewhere around the
1906 current `line`, possibly transitively. This means we can fallback to splitting
1907 by delimiters if the LHS/RHS don't yield any results.
1909 If `py36` is True, splitting may generate syntax that is only compatible
1910 with Python 3.6 and later.
1912 if isinstance(line, UnformattedLines) or line.is_comment:
1916 line_str = str(line).strip("\n")
1917 if not line.should_explode and is_line_short_enough(
1918 line, line_length=line_length, line_str=line_str
1923 split_funcs: List[SplitFunc]
1925 split_funcs = [left_hand_split]
1928 def rhs(line: Line, py36: bool = False) -> Iterator[Line]:
1929 for omit in generate_trailers_to_omit(line, line_length):
1930 lines = list(right_hand_split(line, line_length, py36, omit=omit))
1931 if is_line_short_enough(lines[0], line_length=line_length):
1935 # All splits failed, best effort split with no omits.
1936 # This mostly happens to multiline strings that are by definition
1937 # reported as not fitting a single line.
1938 yield from right_hand_split(line, py36)
1940 if line.inside_brackets:
1941 split_funcs = [delimiter_split, standalone_comment_split, rhs]
1944 for split_func in split_funcs:
1945 # We are accumulating lines in `result` because we might want to abort
1946 # mission and return the original line in the end, or attempt a different
1948 result: List[Line] = []
1950 for l in split_func(line, py36):
1951 if str(l).strip("\n") == line_str:
1952 raise CannotSplit("Split function returned an unchanged result")
1955 split_line(l, line_length=line_length, inner=True, py36=py36)
1957 except CannotSplit as cs:
1968 def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
1969 """Split line into many lines, starting with the first matching bracket pair.
1971 Note: this usually looks weird, only use this for function definitions.
1972 Prefer RHS otherwise. This is why this function is not symmetrical with
1973 :func:`right_hand_split` which also handles optional parentheses.
1975 head = Line(depth=line.depth)
1976 body = Line(depth=line.depth + 1, inside_brackets=True)
1977 tail = Line(depth=line.depth)
1978 tail_leaves: List[Leaf] = []
1979 body_leaves: List[Leaf] = []
1980 head_leaves: List[Leaf] = []
1981 current_leaves = head_leaves
1982 matching_bracket = None
1983 for leaf in line.leaves:
1985 current_leaves is body_leaves
1986 and leaf.type in CLOSING_BRACKETS
1987 and leaf.opening_bracket is matching_bracket
1989 current_leaves = tail_leaves if body_leaves else head_leaves
1990 current_leaves.append(leaf)
1991 if current_leaves is head_leaves:
1992 if leaf.type in OPENING_BRACKETS:
1993 matching_bracket = leaf
1994 current_leaves = body_leaves
1995 # Since body is a new indent level, remove spurious leading whitespace.
1997 normalize_prefix(body_leaves[0], inside_brackets=True)
1998 # Build the new lines.
1999 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
2001 result.append(leaf, preformatted=True)
2002 for comment_after in line.comments_after(leaf):
2003 result.append(comment_after, preformatted=True)
2004 bracket_split_succeeded_or_raise(head, body, tail)
2005 for result in (head, body, tail):
2010 def right_hand_split(
2011 line: Line, line_length: int, py36: bool = False, omit: Collection[LeafID] = ()
2012 ) -> Iterator[Line]:
2013 """Split line into many lines, starting with the last matching bracket pair.
2015 If the split was by optional parentheses, attempt splitting without them, too.
2016 `omit` is a collection of closing bracket IDs that shouldn't be considered for
2019 Note: running this function modifies `bracket_depth` on the leaves of `line`.
2021 head = Line(depth=line.depth)
2022 body = Line(depth=line.depth + 1, inside_brackets=True)
2023 tail = Line(depth=line.depth)
2024 tail_leaves: List[Leaf] = []
2025 body_leaves: List[Leaf] = []
2026 head_leaves: List[Leaf] = []
2027 current_leaves = tail_leaves
2028 opening_bracket = None
2029 closing_bracket = None
2030 for leaf in reversed(line.leaves):
2031 if current_leaves is body_leaves:
2032 if leaf is opening_bracket:
2033 current_leaves = head_leaves if body_leaves else tail_leaves
2034 current_leaves.append(leaf)
2035 if current_leaves is tail_leaves:
2036 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2037 opening_bracket = leaf.opening_bracket
2038 closing_bracket = leaf
2039 current_leaves = body_leaves
2040 tail_leaves.reverse()
2041 body_leaves.reverse()
2042 head_leaves.reverse()
2043 # Since body is a new indent level, remove spurious leading whitespace.
2045 normalize_prefix(body_leaves[0], inside_brackets=True)
2047 # No `head` means the split failed. Either `tail` has all content or
2048 # the matching `opening_bracket` wasn't available on `line` anymore.
2049 raise CannotSplit("No brackets found")
2051 # Build the new lines.
2052 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
2054 result.append(leaf, preformatted=True)
2055 for comment_after in line.comments_after(leaf):
2056 result.append(comment_after, preformatted=True)
2057 bracket_split_succeeded_or_raise(head, body, tail)
2058 assert opening_bracket and closing_bracket
2060 # the opening bracket is an optional paren
2061 opening_bracket.type == token.LPAR
2062 and not opening_bracket.value
2063 # the closing bracket is an optional paren
2064 and closing_bracket.type == token.RPAR
2065 and not closing_bracket.value
2066 # there are no standalone comments in the body
2067 and not line.contains_standalone_comments(0)
2068 # and it's not an import (optional parens are the only thing we can split
2069 # on in this case; attempting a split without them is a waste of time)
2070 and not line.is_import
2072 omit = {id(closing_bracket), *omit}
2073 if can_omit_invisible_parens(body, line_length):
2075 yield from right_hand_split(line, line_length, py36=py36, omit=omit)
2080 ensure_visible(opening_bracket)
2081 ensure_visible(closing_bracket)
2082 body.should_explode = should_explode(body, opening_bracket)
2083 for result in (head, body, tail):
2088 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2089 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2091 Do nothing otherwise.
2093 A left- or right-hand split is based on a pair of brackets. Content before
2094 (and including) the opening bracket is left on one line, content inside the
2095 brackets is put on a separate line, and finally content starting with and
2096 following the closing bracket is put on a separate line.
2098 Those are called `head`, `body`, and `tail`, respectively. If the split
2099 produced the same line (all content in `head`) or ended up with an empty `body`
2100 and the `tail` is just the closing bracket, then it's considered failed.
2102 tail_len = len(str(tail).strip())
2105 raise CannotSplit("Splitting brackets produced the same line")
2109 f"Splitting brackets on an empty body to save "
2110 f"{tail_len} characters is not worth it"
2114 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2115 """Normalize prefix of the first leaf in every line returned by `split_func`.
2117 This is a decorator over relevant split functions.
2121 def split_wrapper(line: Line, py36: bool = False) -> Iterator[Line]:
2122 for l in split_func(line, py36):
2123 normalize_prefix(l.leaves[0], inside_brackets=True)
2126 return split_wrapper
2129 @dont_increase_indentation
2130 def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]:
2131 """Split according to delimiters of the highest priority.
2133 If `py36` is True, the split will add trailing commas also in function
2134 signatures that contain `*` and `**`.
2137 last_leaf = line.leaves[-1]
2139 raise CannotSplit("Line empty")
2141 bt = line.bracket_tracker
2143 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2145 raise CannotSplit("No delimiters found")
2147 if delimiter_priority == DOT_PRIORITY:
2148 if bt.delimiter_count_with_priority(delimiter_priority) == 1:
2149 raise CannotSplit("Splitting a single attribute from its owner looks wrong")
2151 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2152 lowest_depth = sys.maxsize
2153 trailing_comma_safe = True
2155 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2156 """Append `leaf` to current line or to new line if appending impossible."""
2157 nonlocal current_line
2159 current_line.append_safe(leaf, preformatted=True)
2160 except ValueError as ve:
2163 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2164 current_line.append(leaf)
2166 for index, leaf in enumerate(line.leaves):
2167 yield from append_to_line(leaf)
2169 for comment_after in line.comments_after(leaf, index):
2170 yield from append_to_line(comment_after)
2172 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2173 if leaf.bracket_depth == lowest_depth and is_vararg(
2174 leaf, within=VARARGS_PARENTS
2176 trailing_comma_safe = trailing_comma_safe and py36
2177 leaf_priority = bt.delimiters.get(id(leaf))
2178 if leaf_priority == delimiter_priority:
2181 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2185 and delimiter_priority == COMMA_PRIORITY
2186 and current_line.leaves[-1].type != token.COMMA
2187 and current_line.leaves[-1].type != STANDALONE_COMMENT
2189 current_line.append(Leaf(token.COMMA, ","))
2193 @dont_increase_indentation
2194 def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]:
2195 """Split standalone comments from the rest of the line."""
2196 if not line.contains_standalone_comments(0):
2197 raise CannotSplit("Line does not have any standalone comments")
2199 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2201 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2202 """Append `leaf` to current line or to new line if appending impossible."""
2203 nonlocal current_line
2205 current_line.append_safe(leaf, preformatted=True)
2206 except ValueError as ve:
2209 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2210 current_line.append(leaf)
2212 for index, leaf in enumerate(line.leaves):
2213 yield from append_to_line(leaf)
2215 for comment_after in line.comments_after(leaf, index):
2216 yield from append_to_line(comment_after)
2222 def is_import(leaf: Leaf) -> bool:
2223 """Return True if the given leaf starts an import statement."""
2230 (v == "import" and p and p.type == syms.import_name)
2231 or (v == "from" and p and p.type == syms.import_from)
2236 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
2237 """Leave existing extra newlines if not `inside_brackets`. Remove everything
2240 Note: don't use backslashes for formatting or you'll lose your voting rights.
2242 if not inside_brackets:
2243 spl = leaf.prefix.split("#")
2244 if "\\" not in spl[0]:
2245 nl_count = spl[-1].count("\n")
2248 leaf.prefix = "\n" * nl_count
2254 def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None:
2255 """Make all string prefixes lowercase.
2257 If remove_u_prefix is given, also removes any u prefix from the string.
2259 Note: Mutates its argument.
2261 match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL)
2262 assert match is not None, f"failed to match string {leaf.value!r}"
2263 orig_prefix = match.group(1)
2264 new_prefix = orig_prefix.lower()
2266 new_prefix = new_prefix.replace("u", "")
2267 leaf.value = f"{new_prefix}{match.group(2)}"
2270 def normalize_string_quotes(leaf: Leaf) -> None:
2271 """Prefer double quotes but only if it doesn't cause more escaping.
2273 Adds or removes backslashes as appropriate. Doesn't parse and fix
2274 strings nested in f-strings (yet).
2276 Note: Mutates its argument.
2278 value = leaf.value.lstrip("furbFURB")
2279 if value[:3] == '"""':
2282 elif value[:3] == "'''":
2285 elif value[0] == '"':
2291 first_quote_pos = leaf.value.find(orig_quote)
2292 if first_quote_pos == -1:
2293 return # There's an internal error
2295 prefix = leaf.value[:first_quote_pos]
2296 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2297 escaped_new_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{new_quote}")
2298 escaped_orig_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{orig_quote}")
2299 body = leaf.value[first_quote_pos + len(orig_quote) : -len(orig_quote)]
2300 if "r" in prefix.casefold():
2301 if unescaped_new_quote.search(body):
2302 # There's at least one unescaped new_quote in this raw string
2303 # so converting is impossible
2306 # Do not introduce or remove backslashes in raw strings
2309 # remove unnecessary quotes
2310 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2311 if body != new_body:
2312 # Consider the string without unnecessary quotes as the original
2314 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2315 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2316 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2317 if new_quote == '"""' and new_body[-1] == '"':
2319 new_body = new_body[:-1] + '\\"'
2320 orig_escape_count = body.count("\\")
2321 new_escape_count = new_body.count("\\")
2322 if new_escape_count > orig_escape_count:
2323 return # Do not introduce more escaping
2325 if new_escape_count == orig_escape_count and orig_quote == '"':
2326 return # Prefer double quotes
2328 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2331 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2332 """Make existing optional parentheses invisible or create new ones.
2334 `parens_after` is a set of string leaf values immeditely after which parens
2337 Standardizes on visible parentheses for single-element tuples, and keeps
2338 existing visible parentheses for other tuples and generator expressions.
2341 for child in list(node.children):
2343 if child.type == syms.atom:
2344 maybe_make_parens_invisible_in_atom(child)
2345 elif is_one_tuple(child):
2346 # wrap child in visible parentheses
2347 lpar = Leaf(token.LPAR, "(")
2348 rpar = Leaf(token.RPAR, ")")
2349 index = child.remove() or 0
2350 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2351 elif not (isinstance(child, Leaf) and is_multiline_string(child)):
2352 # wrap child in invisible parentheses
2353 lpar = Leaf(token.LPAR, "")
2354 rpar = Leaf(token.RPAR, "")
2355 index = child.remove() or 0
2356 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2358 check_lpar = isinstance(child, Leaf) and child.value in parens_after
2361 def maybe_make_parens_invisible_in_atom(node: LN) -> bool:
2362 """If it's safe, make the parens in the atom `node` invisible, recusively."""
2364 node.type != syms.atom
2365 or is_empty_tuple(node)
2366 or is_one_tuple(node)
2368 or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
2372 first = node.children[0]
2373 last = node.children[-1]
2374 if first.type == token.LPAR and last.type == token.RPAR:
2375 # make parentheses invisible
2376 first.value = "" # type: ignore
2377 last.value = "" # type: ignore
2378 if len(node.children) > 1:
2379 maybe_make_parens_invisible_in_atom(node.children[1])
2385 def is_empty_tuple(node: LN) -> bool:
2386 """Return True if `node` holds an empty tuple."""
2388 node.type == syms.atom
2389 and len(node.children) == 2
2390 and node.children[0].type == token.LPAR
2391 and node.children[1].type == token.RPAR
2395 def is_one_tuple(node: LN) -> bool:
2396 """Return True if `node` holds a tuple with one element, with or without parens."""
2397 if node.type == syms.atom:
2398 if len(node.children) != 3:
2401 lpar, gexp, rpar = node.children
2403 lpar.type == token.LPAR
2404 and gexp.type == syms.testlist_gexp
2405 and rpar.type == token.RPAR
2409 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
2412 node.type in IMPLICIT_TUPLE
2413 and len(node.children) == 2
2414 and node.children[1].type == token.COMMA
2418 def is_yield(node: LN) -> bool:
2419 """Return True if `node` holds a `yield` or `yield from` expression."""
2420 if node.type == syms.yield_expr:
2423 if node.type == token.NAME and node.value == "yield": # type: ignore
2426 if node.type != syms.atom:
2429 if len(node.children) != 3:
2432 lpar, expr, rpar = node.children
2433 if lpar.type == token.LPAR and rpar.type == token.RPAR:
2434 return is_yield(expr)
2439 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
2440 """Return True if `leaf` is a star or double star in a vararg or kwarg.
2442 If `within` includes VARARGS_PARENTS, this applies to function signatures.
2443 If `within` includes UNPACKING_PARENTS, it applies to right hand-side
2444 extended iterable unpacking (PEP 3132) and additional unpacking
2445 generalizations (PEP 448).
2447 if leaf.type not in STARS or not leaf.parent:
2451 if p.type == syms.star_expr:
2452 # Star expressions are also used as assignment targets in extended
2453 # iterable unpacking (PEP 3132). See what its parent is instead.
2459 return p.type in within
2462 def is_multiline_string(leaf: Leaf) -> bool:
2463 """Return True if `leaf` is a multiline string that actually spans many lines."""
2464 value = leaf.value.lstrip("furbFURB")
2465 return value[:3] in {'"""', "'''"} and "\n" in value
2468 def is_stub_suite(node: Node) -> bool:
2469 """Return True if `node` is a suite with a stub body."""
2471 len(node.children) != 4
2472 or node.children[0].type != token.NEWLINE
2473 or node.children[1].type != token.INDENT
2474 or node.children[3].type != token.DEDENT
2478 return is_stub_body(node.children[2])
2481 def is_stub_body(node: LN) -> bool:
2482 """Return True if `node` is a simple statement containing an ellipsis."""
2483 if not isinstance(node, Node) or node.type != syms.simple_stmt:
2486 if len(node.children) != 2:
2489 child = node.children[0]
2491 child.type == syms.atom
2492 and len(child.children) == 3
2493 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
2497 def max_delimiter_priority_in_atom(node: LN) -> int:
2498 """Return maximum delimiter priority inside `node`.
2500 This is specific to atoms with contents contained in a pair of parentheses.
2501 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
2503 if node.type != syms.atom:
2506 first = node.children[0]
2507 last = node.children[-1]
2508 if not (first.type == token.LPAR and last.type == token.RPAR):
2511 bt = BracketTracker()
2512 for c in node.children[1:-1]:
2513 if isinstance(c, Leaf):
2516 for leaf in c.leaves():
2519 return bt.max_delimiter_priority()
2525 def ensure_visible(leaf: Leaf) -> None:
2526 """Make sure parentheses are visible.
2528 They could be invisible as part of some statements (see
2529 :func:`normalize_invible_parens` and :func:`visit_import_from`).
2531 if leaf.type == token.LPAR:
2533 elif leaf.type == token.RPAR:
2537 def should_explode(line: Line, opening_bracket: Leaf) -> bool:
2538 """Should `line` immediately be split with `delimiter_split()` after RHS?"""
2540 opening_bracket.parent
2541 and opening_bracket.parent.type in {syms.atom, syms.import_from}
2542 and opening_bracket.value in "[{("
2547 last_leaf = line.leaves[-1]
2548 exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
2549 max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
2550 except (IndexError, ValueError):
2553 return max_priority == COMMA_PRIORITY
2556 def is_python36(node: Node) -> bool:
2557 """Return True if the current file is using Python 3.6+ features.
2559 Currently looking for:
2561 - trailing commas after * or ** in function signatures and calls.
2563 for n in node.pre_order():
2564 if n.type == token.STRING:
2565 value_head = n.value[:2] # type: ignore
2566 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
2570 n.type in {syms.typedargslist, syms.arglist}
2572 and n.children[-1].type == token.COMMA
2574 for ch in n.children:
2575 if ch.type in STARS:
2578 if ch.type == syms.argument:
2579 for argch in ch.children:
2580 if argch.type in STARS:
2586 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
2587 """Generate sets of closing bracket IDs that should be omitted in a RHS.
2589 Brackets can be omitted if the entire trailer up to and including
2590 a preceding closing bracket fits in one line.
2592 Yielded sets are cumulative (contain results of previous yields, too). First
2596 omit: Set[LeafID] = set()
2599 length = 4 * line.depth
2600 opening_bracket = None
2601 closing_bracket = None
2602 optional_brackets: Set[LeafID] = set()
2603 inner_brackets: Set[LeafID] = set()
2604 for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
2605 length += leaf_length
2606 if length > line_length:
2609 if leaf.type == STANDALONE_COMMENT:
2612 optional_brackets.discard(id(leaf))
2614 if leaf is opening_bracket:
2615 opening_bracket = None
2616 elif leaf.type in CLOSING_BRACKETS:
2617 inner_brackets.add(id(leaf))
2618 elif leaf.type in CLOSING_BRACKETS:
2620 optional_brackets.add(id(opening_bracket))
2623 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
2624 # Empty brackets would fail a split so treat them as "inner"
2625 # brackets (e.g. only add them to the `omit` set if another
2626 # pair of brackets was good enough.
2627 inner_brackets.add(id(leaf))
2630 opening_bracket = leaf.opening_bracket
2632 omit.add(id(closing_bracket))
2633 omit.update(inner_brackets)
2634 inner_brackets.clear()
2636 closing_bracket = leaf
2639 def get_future_imports(node: Node) -> Set[str]:
2640 """Return a set of __future__ imports in the file."""
2642 for child in node.children:
2643 if child.type != syms.simple_stmt:
2645 first_child = child.children[0]
2646 if isinstance(first_child, Leaf):
2647 # Continue looking if we see a docstring; otherwise stop.
2649 len(child.children) == 2
2650 and first_child.type == token.STRING
2651 and child.children[1].type == token.NEWLINE
2656 elif first_child.type == syms.import_from:
2657 module_name = first_child.children[1]
2658 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
2660 for import_from_child in first_child.children[3:]:
2661 if isinstance(import_from_child, Leaf):
2662 if import_from_child.type == token.NAME:
2663 imports.add(import_from_child.value)
2665 assert import_from_child.type == syms.import_as_names
2666 for leaf in import_from_child.children:
2667 if isinstance(leaf, Leaf) and leaf.type == token.NAME:
2668 imports.add(leaf.value)
2674 PYTHON_EXTENSIONS = {".py", ".pyi"}
2675 BLACKLISTED_DIRECTORIES = {
2688 def gen_python_files_in_dir(path: Path) -> Iterator[Path]:
2689 """Generate all files under `path` which aren't under BLACKLISTED_DIRECTORIES
2690 and have one of the PYTHON_EXTENSIONS.
2692 for child in path.iterdir():
2694 if child.name in BLACKLISTED_DIRECTORIES:
2697 yield from gen_python_files_in_dir(child)
2699 elif child.is_file() and child.suffix in PYTHON_EXTENSIONS:
2705 """Provides a reformatting counter. Can be rendered with `str(report)`."""
2708 change_count: int = 0
2710 failure_count: int = 0
2712 def done(self, src: Path, changed: Changed) -> None:
2713 """Increment the counter for successful reformatting. Write out a message."""
2714 if changed is Changed.YES:
2715 reformatted = "would reformat" if self.check else "reformatted"
2717 out(f"{reformatted} {src}")
2718 self.change_count += 1
2721 if changed is Changed.NO:
2722 msg = f"{src} already well formatted, good job."
2724 msg = f"{src} wasn't modified on disk since last run."
2725 out(msg, bold=False)
2726 self.same_count += 1
2728 def failed(self, src: Path, message: str) -> None:
2729 """Increment the counter for failed reformatting. Write out a message."""
2730 err(f"error: cannot format {src}: {message}")
2731 self.failure_count += 1
2734 def return_code(self) -> int:
2735 """Return the exit code that the app should use.
2737 This considers the current state of changed files and failures:
2738 - if there were any failures, return 123;
2739 - if any files were changed and --check is being used, return 1;
2740 - otherwise return 0.
2742 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
2743 # 126 we have special returncodes reserved by the shell.
2744 if self.failure_count:
2747 elif self.change_count and self.check:
2752 def __str__(self) -> str:
2753 """Render a color report of the current state.
2755 Use `click.unstyle` to remove colors.
2758 reformatted = "would be reformatted"
2759 unchanged = "would be left unchanged"
2760 failed = "would fail to reformat"
2762 reformatted = "reformatted"
2763 unchanged = "left unchanged"
2764 failed = "failed to reformat"
2766 if self.change_count:
2767 s = "s" if self.change_count > 1 else ""
2769 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
2772 s = "s" if self.same_count > 1 else ""
2773 report.append(f"{self.same_count} file{s} {unchanged}")
2774 if self.failure_count:
2775 s = "s" if self.failure_count > 1 else ""
2777 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
2779 return ", ".join(report) + "."
2782 def assert_equivalent(src: str, dst: str) -> None:
2783 """Raise AssertionError if `src` and `dst` aren't equivalent."""
2788 def _v(node: ast.AST, depth: int = 0) -> Iterator[str]:
2789 """Simple visitor generating strings to compare ASTs by content."""
2790 yield f"{' ' * depth}{node.__class__.__name__}("
2792 for field in sorted(node._fields):
2794 value = getattr(node, field)
2795 except AttributeError:
2798 yield f"{' ' * (depth+1)}{field}="
2800 if isinstance(value, list):
2802 if isinstance(item, ast.AST):
2803 yield from _v(item, depth + 2)
2805 elif isinstance(value, ast.AST):
2806 yield from _v(value, depth + 2)
2809 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
2811 yield f"{' ' * depth}) # /{node.__class__.__name__}"
2814 src_ast = ast.parse(src)
2815 except Exception as exc:
2816 major, minor = sys.version_info[:2]
2817 raise AssertionError(
2818 f"cannot use --safe with this file; failed to parse source file "
2819 f"with Python {major}.{minor}'s builtin AST. Re-run with --fast "
2820 f"or stop using deprecated Python 2 syntax. AST error message: {exc}"
2824 dst_ast = ast.parse(dst)
2825 except Exception as exc:
2826 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
2827 raise AssertionError(
2828 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
2829 f"Please report a bug on https://github.com/ambv/black/issues. "
2830 f"This invalid output might be helpful: {log}"
2833 src_ast_str = "\n".join(_v(src_ast))
2834 dst_ast_str = "\n".join(_v(dst_ast))
2835 if src_ast_str != dst_ast_str:
2836 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
2837 raise AssertionError(
2838 f"INTERNAL ERROR: Black produced code that is not equivalent to "
2840 f"Please report a bug on https://github.com/ambv/black/issues. "
2841 f"This diff might be helpful: {log}"
2845 def assert_stable(src: str, dst: str, line_length: int, is_pyi: bool = False) -> None:
2846 """Raise AssertionError if `dst` reformats differently the second time."""
2847 newdst = format_str(dst, line_length=line_length, is_pyi=is_pyi)
2850 diff(src, dst, "source", "first pass"),
2851 diff(dst, newdst, "first pass", "second pass"),
2853 raise AssertionError(
2854 f"INTERNAL ERROR: Black produced different code on the second pass "
2855 f"of the formatter. "
2856 f"Please report a bug on https://github.com/ambv/black/issues. "
2857 f"This diff might be helpful: {log}"
2861 def dump_to_file(*output: str) -> str:
2862 """Dump `output` to a temporary file. Return path to the file."""
2865 with tempfile.NamedTemporaryFile(
2866 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
2868 for lines in output:
2870 if lines and lines[-1] != "\n":
2875 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
2876 """Return a unified diff string between strings `a` and `b`."""
2879 a_lines = [line + "\n" for line in a.split("\n")]
2880 b_lines = [line + "\n" for line in b.split("\n")]
2882 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
2886 def cancel(tasks: Iterable[asyncio.Task]) -> None:
2887 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
2893 def shutdown(loop: BaseEventLoop) -> None:
2894 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
2896 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
2897 to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()]
2901 for task in to_cancel:
2903 loop.run_until_complete(
2904 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
2907 # `concurrent.futures.Future` objects cannot be cancelled once they
2908 # are already running. There might be some when the `shutdown()` happened.
2909 # Silence their logger's spew about the event loop being closed.
2910 cf_logger = logging.getLogger("concurrent.futures")
2911 cf_logger.setLevel(logging.CRITICAL)
2915 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
2916 """Replace `regex` with `replacement` twice on `original`.
2918 This is used by string normalization to perform replaces on
2919 overlapping matches.
2921 return regex.sub(replacement, regex.sub(replacement, original))
2924 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
2925 """Like `reversed(enumerate(sequence))` if that were possible."""
2926 index = len(sequence) - 1
2927 for element in reversed(sequence):
2928 yield (index, element)
2932 def enumerate_with_length(
2933 line: Line, reversed: bool = False
2934 ) -> Iterator[Tuple[Index, Leaf, int]]:
2935 """Return an enumeration of leaves with their length.
2937 Stops prematurely on multiline strings and standalone comments.
2940 Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
2941 enumerate_reversed if reversed else enumerate,
2943 for index, leaf in op(line.leaves):
2944 length = len(leaf.prefix) + len(leaf.value)
2945 if "\n" in leaf.value:
2946 return # Multiline strings, we can't continue.
2948 comment: Optional[Leaf]
2949 for comment in line.comments_after(leaf, index):
2950 if "\n" in comment.prefix:
2951 return # Oops, standalone comment!
2953 length += len(comment.value)
2955 yield index, leaf, length
2958 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
2959 """Return True if `line` is no longer than `line_length`.
2961 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
2964 line_str = str(line).strip("\n")
2966 len(line_str) <= line_length
2967 and "\n" not in line_str # multiline strings
2968 and not line.contains_standalone_comments()
2972 def can_omit_invisible_parens(line: Line, line_length: int) -> bool:
2973 """Does `line` have a shape safe to reformat without optional parens around it?
2975 Returns True for only a subset of potentially nice looking formattings but
2976 the point is to not return false positives that end up producing lines that
2979 bt = line.bracket_tracker
2980 if not bt.delimiters:
2981 # Without delimiters the optional parentheses are useless.
2984 max_priority = bt.max_delimiter_priority()
2985 if bt.delimiter_count_with_priority(max_priority) > 1:
2986 # With more than one delimiter of a kind the optional parentheses read better.
2989 if max_priority == DOT_PRIORITY:
2990 # A single stranded method call doesn't require optional parentheses.
2993 assert len(line.leaves) >= 2, "Stranded delimiter"
2995 first = line.leaves[0]
2996 second = line.leaves[1]
2997 penultimate = line.leaves[-2]
2998 last = line.leaves[-1]
3000 # With a single delimiter, omit if the expression starts or ends with
3002 if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS:
3004 length = 4 * line.depth
3005 for _index, leaf, leaf_length in enumerate_with_length(line):
3006 if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first:
3009 length += leaf_length
3010 if length > line_length:
3013 if leaf.type in OPENING_BRACKETS:
3014 # There are brackets we can further split on.
3018 # checked the entire string and line length wasn't exceeded
3019 if len(line.leaves) == _index + 1:
3022 # Note: we are not returning False here because a line might have *both*
3023 # a leading opening bracket and a trailing closing bracket. If the
3024 # opening bracket doesn't match our rule, maybe the closing will.
3027 last.type == token.RPAR
3028 or last.type == token.RBRACE
3030 # don't use indexing for omitting optional parentheses;
3032 last.type == token.RSQB
3034 and last.parent.type != syms.trailer
3037 if penultimate.type in OPENING_BRACKETS:
3038 # Empty brackets don't help.
3041 if is_multiline_string(first):
3042 # Additional wrapping of a multiline string in this situation is
3046 length = 4 * line.depth
3047 seen_other_brackets = False
3048 for _index, leaf, leaf_length in enumerate_with_length(line):
3049 length += leaf_length
3050 if leaf is last.opening_bracket:
3051 if seen_other_brackets or length <= line_length:
3054 elif leaf.type in OPENING_BRACKETS:
3055 # There are brackets we can further split on.
3056 seen_other_brackets = True
3061 def get_cache_file(line_length: int) -> Path:
3062 return CACHE_DIR / f"cache.{line_length}.pickle"
3065 def read_cache(line_length: int) -> Cache:
3066 """Read the cache if it exists and is well formed.
3068 If it is not well formed, the call to write_cache later should resolve the issue.
3070 cache_file = get_cache_file(line_length)
3071 if not cache_file.exists():
3074 with cache_file.open("rb") as fobj:
3076 cache: Cache = pickle.load(fobj)
3077 except pickle.UnpicklingError:
3083 def get_cache_info(path: Path) -> CacheInfo:
3084 """Return the information used to check if a file is already formatted or not."""
3086 return stat.st_mtime, stat.st_size
3090 cache: Cache, sources: Iterable[Path]
3091 ) -> Tuple[List[Path], List[Path]]:
3092 """Split a list of paths into two.
3094 The first list contains paths of files that modified on disk or are not in the
3095 cache. The other list contains paths to non-modified files.
3100 if cache.get(src) != get_cache_info(src):
3107 def write_cache(cache: Cache, sources: List[Path], line_length: int) -> None:
3108 """Update the cache file."""
3109 cache_file = get_cache_file(line_length)
3111 if not CACHE_DIR.exists():
3112 CACHE_DIR.mkdir(parents=True)
3113 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
3114 with cache_file.open("wb") as fobj:
3115 pickle.dump(new_cache, fobj, protocol=pickle.HIGHEST_PROTOCOL)
3120 if __name__ == "__main__":