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
35 from appdirs import user_cache_dir
36 from attr import dataclass, Factory
40 from blib2to3.pytree import Node, Leaf, type_repr
41 from blib2to3 import pygram, pytree
42 from blib2to3.pgen2 import driver, token
43 from blib2to3.pgen2.parse import ParseError
46 __version__ = "18.4a6"
47 DEFAULT_LINE_LENGTH = 88
50 syms = pygram.python_symbols
58 LN = Union[Leaf, Node]
59 SplitFunc = Callable[["Line", bool], Iterator["Line"]]
62 CacheInfo = Tuple[Timestamp, FileSize]
63 Cache = Dict[Path, CacheInfo]
64 out = partial(click.secho, bold=True, err=True)
65 err = partial(click.secho, fg="red", err=True)
68 class NothingChanged(UserWarning):
69 """Raised by :func:`format_file` when reformatted code is the same as source."""
72 class CannotSplit(Exception):
73 """A readable split that fits the allotted line length is impossible.
75 Raised by :func:`left_hand_split`, :func:`right_hand_split`, and
76 :func:`delimiter_split`.
80 class FormatError(Exception):
81 """Base exception for `# fmt: on` and `# fmt: off` handling.
83 It holds the number of bytes of the prefix consumed before the format
84 control comment appeared.
87 def __init__(self, consumed: int) -> None:
88 super().__init__(consumed)
89 self.consumed = consumed
91 def trim_prefix(self, leaf: Leaf) -> None:
92 leaf.prefix = leaf.prefix[self.consumed :]
94 def leaf_from_consumed(self, leaf: Leaf) -> Leaf:
95 """Returns a new Leaf from the consumed part of the prefix."""
96 unformatted_prefix = leaf.prefix[: self.consumed]
97 return Leaf(token.NEWLINE, unformatted_prefix)
100 class FormatOn(FormatError):
101 """Found a comment like `# fmt: on` in the file."""
104 class FormatOff(FormatError):
105 """Found a comment like `# fmt: off` in the file."""
108 class WriteBack(Enum):
125 default=DEFAULT_LINE_LENGTH,
126 help="How many character per line to allow.",
133 "Don't write the files back, just return the status. Return code 0 "
134 "means nothing would change. Return code 1 means some files would be "
135 "reformatted. Return code 123 means there was an internal error."
141 help="Don't write the files back, just output a diff for each file on stdout.",
146 help="If --fast given, skip temporary sanity checks. [default: --safe]",
153 "Don't emit non-error messages to stderr. Errors are still emitted, "
154 "silence those with 2>/dev/null."
157 @click.version_option(version=__version__)
162 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
175 """The uncompromising code formatter."""
176 sources: List[Path] = []
180 sources.extend(gen_python_files_in_dir(p))
182 # if a file was explicitly given, we don't care about its extension
185 sources.append(Path("-"))
187 err(f"invalid path: {s}")
189 if check and not diff:
190 write_back = WriteBack.NO
192 write_back = WriteBack.DIFF
194 write_back = WriteBack.YES
195 report = Report(check=check, quiet=quiet)
196 if len(sources) == 0:
197 out("No paths given. Nothing to do 😴")
201 elif len(sources) == 1:
202 reformat_one(sources[0], line_length, fast, write_back, report)
204 loop = asyncio.get_event_loop()
205 executor = ProcessPoolExecutor(max_workers=os.cpu_count())
207 loop.run_until_complete(
209 sources, line_length, fast, write_back, report, loop, executor
215 out("All done! ✨ 🍰 ✨")
216 click.echo(str(report))
217 ctx.exit(report.return_code)
221 src: Path, line_length: int, fast: bool, write_back: WriteBack, report: "Report"
223 """Reformat a single file under `src` without spawning child processes.
225 If `quiet` is True, non-error messages are not output. `line_length`,
226 `write_back`, and `fast` options are passed to :func:`format_file_in_place`.
230 if not src.is_file() and str(src) == "-":
231 if format_stdin_to_stdout(
232 line_length=line_length, fast=fast, write_back=write_back
234 changed = Changed.YES
237 if write_back != WriteBack.DIFF:
238 cache = read_cache(line_length)
240 if src in cache and cache[src] == get_cache_info(src):
241 changed = Changed.CACHED
242 if changed is not Changed.CACHED and format_file_in_place(
243 src, line_length=line_length, fast=fast, write_back=write_back
245 changed = Changed.YES
246 if write_back == WriteBack.YES and changed is not Changed.NO:
247 write_cache(cache, [src], line_length)
248 report.done(src, changed)
249 except Exception as exc:
250 report.failed(src, str(exc))
253 async def schedule_formatting(
257 write_back: WriteBack,
262 """Run formatting of `sources` in parallel using the provided `executor`.
264 (Use ProcessPoolExecutors for actual parallelism.)
266 `line_length`, `write_back`, and `fast` options are passed to
267 :func:`format_file_in_place`.
270 if write_back != WriteBack.DIFF:
271 cache = read_cache(line_length)
272 sources, cached = filter_cached(cache, sources)
274 report.done(src, Changed.CACHED)
279 if write_back == WriteBack.DIFF:
280 # For diff output, we need locks to ensure we don't interleave output
281 # from different processes.
283 lock = manager.Lock()
285 loop.run_in_executor(
286 executor, format_file_in_place, src, line_length, fast, write_back, lock
288 for src in sorted(sources)
290 pending: Iterable[asyncio.Task] = tasks.keys()
292 loop.add_signal_handler(signal.SIGINT, cancel, pending)
293 loop.add_signal_handler(signal.SIGTERM, cancel, pending)
294 except NotImplementedError:
295 # There are no good alternatives for these on Windows
298 done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED)
300 src = tasks.pop(task)
302 cancelled.append(task)
303 elif task.exception():
304 report.failed(src, str(task.exception()))
306 formatted.append(src)
307 report.done(src, Changed.YES if task.result() else Changed.NO)
309 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
310 if write_back == WriteBack.YES and formatted:
311 write_cache(cache, formatted, line_length)
314 def format_file_in_place(
318 write_back: WriteBack = WriteBack.NO,
319 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
321 """Format file under `src` path. Return True if changed.
323 If `write_back` is True, write reformatted code back to stdout.
324 `line_length` and `fast` options are passed to :func:`format_file_contents`.
326 is_pyi = src.suffix == ".pyi"
328 with tokenize.open(src) as src_buffer:
329 src_contents = src_buffer.read()
331 dst_contents = format_file_contents(
332 src_contents, line_length=line_length, fast=fast, is_pyi=is_pyi
334 except NothingChanged:
337 if write_back == write_back.YES:
338 with open(src, "w", encoding=src_buffer.encoding) as f:
339 f.write(dst_contents)
340 elif write_back == write_back.DIFF:
341 src_name = f"{src} (original)"
342 dst_name = f"{src} (formatted)"
343 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
347 sys.stdout.write(diff_contents)
354 def format_stdin_to_stdout(
355 line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO
357 """Format file on stdin. Return True if changed.
359 If `write_back` is True, write reformatted code back to stdout.
360 `line_length` and `fast` arguments are passed to :func:`format_file_contents`.
362 src = sys.stdin.read()
365 dst = format_file_contents(src, line_length=line_length, fast=fast)
368 except NothingChanged:
372 if write_back == WriteBack.YES:
373 sys.stdout.write(dst)
374 elif write_back == WriteBack.DIFF:
375 src_name = "<stdin> (original)"
376 dst_name = "<stdin> (formatted)"
377 sys.stdout.write(diff(src, dst, src_name, dst_name))
380 def format_file_contents(
381 src_contents: str, *, line_length: int, fast: bool, is_pyi: bool = False
383 """Reformat contents a file and return new contents.
385 If `fast` is False, additionally confirm that the reformatted code is
386 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
387 `line_length` is passed to :func:`format_str`.
389 if src_contents.strip() == "":
392 dst_contents = format_str(src_contents, line_length=line_length, is_pyi=is_pyi)
393 if src_contents == dst_contents:
397 assert_equivalent(src_contents, dst_contents)
399 src_contents, dst_contents, line_length=line_length, is_pyi=is_pyi
405 src_contents: str, line_length: int, *, is_pyi: bool = False
407 """Reformat a string and return new contents.
409 `line_length` determines how many characters per line are allowed.
411 src_node = lib2to3_parse(src_contents)
413 future_imports = get_future_imports(src_node)
414 elt = EmptyLineTracker(is_pyi=is_pyi)
415 py36 = is_python36(src_node)
416 lines = LineGenerator(
417 remove_u_prefix=py36 or "unicode_literals" in future_imports, is_pyi=is_pyi
421 for current_line in lines.visit(src_node):
422 for _ in range(after):
423 dst_contents += str(empty_line)
424 before, after = elt.maybe_empty_lines(current_line)
425 for _ in range(before):
426 dst_contents += str(empty_line)
427 for line in split_line(current_line, line_length=line_length, py36=py36):
428 dst_contents += str(line)
433 pygram.python_grammar_no_print_statement_no_exec_statement,
434 pygram.python_grammar_no_print_statement,
435 pygram.python_grammar,
439 def lib2to3_parse(src_txt: str) -> Node:
440 """Given a string with source, return the lib2to3 Node."""
441 grammar = pygram.python_grammar_no_print_statement
442 if src_txt[-1] != "\n":
443 nl = "\r\n" if "\r\n" in src_txt[:1024] else "\n"
445 for grammar in GRAMMARS:
446 drv = driver.Driver(grammar, pytree.convert)
448 result = drv.parse_string(src_txt, True)
451 except ParseError as pe:
452 lineno, column = pe.context[1]
453 lines = src_txt.splitlines()
455 faulty_line = lines[lineno - 1]
457 faulty_line = "<line number missing in source>"
458 exc = ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}")
462 if isinstance(result, Leaf):
463 result = Node(syms.file_input, [result])
467 def lib2to3_unparse(node: Node) -> str:
468 """Given a lib2to3 node, return its string representation."""
476 class Visitor(Generic[T]):
477 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
479 def visit(self, node: LN) -> Iterator[T]:
480 """Main method to visit `node` and its children.
482 It tries to find a `visit_*()` method for the given `node.type`, like
483 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
484 If no dedicated `visit_*()` method is found, chooses `visit_default()`
487 Then yields objects of type `T` from the selected visitor.
490 name = token.tok_name[node.type]
492 name = type_repr(node.type)
493 yield from getattr(self, f"visit_{name}", self.visit_default)(node)
495 def visit_default(self, node: LN) -> Iterator[T]:
496 """Default `visit_*()` implementation. Recurses to children of `node`."""
497 if isinstance(node, Node):
498 for child in node.children:
499 yield from self.visit(child)
503 class DebugVisitor(Visitor[T]):
506 def visit_default(self, node: LN) -> Iterator[T]:
507 indent = " " * (2 * self.tree_depth)
508 if isinstance(node, Node):
509 _type = type_repr(node.type)
510 out(f"{indent}{_type}", fg="yellow")
512 for child in node.children:
513 yield from self.visit(child)
516 out(f"{indent}/{_type}", fg="yellow", bold=False)
518 _type = token.tok_name.get(node.type, str(node.type))
519 out(f"{indent}{_type}", fg="blue", nl=False)
521 # We don't have to handle prefixes for `Node` objects since
522 # that delegates to the first child anyway.
523 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
524 out(f" {node.value!r}", fg="blue", bold=False)
527 def show(cls, code: str) -> None:
528 """Pretty-print the lib2to3 AST of a given string of `code`.
530 Convenience method for debugging.
532 v: DebugVisitor[None] = DebugVisitor()
533 list(v.visit(lib2to3_parse(code)))
536 KEYWORDS = set(keyword.kwlist)
537 WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE}
538 FLOW_CONTROL = {"return", "raise", "break", "continue"}
549 STANDALONE_COMMENT = 153
550 LOGIC_OPERATORS = {"and", "or"}
575 STARS = {token.STAR, token.DOUBLESTAR}
578 syms.argument, # double star in arglist
579 syms.trailer, # single argument to call
581 syms.varargslist, # lambdas
583 UNPACKING_PARENTS = {
584 syms.atom, # single element of a list or set literal
622 COMPREHENSION_PRIORITY = 20
624 TERNARY_PRIORITY = 16
627 COMPARATOR_PRIORITY = 10
638 token.DOUBLESLASH: 3,
647 class BracketTracker:
648 """Keeps track of brackets on a line."""
651 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
652 delimiters: Dict[LeafID, Priority] = Factory(dict)
653 previous: Optional[Leaf] = None
654 _for_loop_variable: int = 0
655 _lambda_arguments: int = 0
657 def mark(self, leaf: Leaf) -> None:
658 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
660 All leaves receive an int `bracket_depth` field that stores how deep
661 within brackets a given leaf is. 0 means there are no enclosing brackets
662 that started on this line.
664 If a leaf is itself a closing bracket, it receives an `opening_bracket`
665 field that it forms a pair with. This is a one-directional link to
666 avoid reference cycles.
668 If a leaf is a delimiter (a token on which Black can split the line if
669 needed) and it's on depth 0, its `id()` is stored in the tracker's
672 if leaf.type == token.COMMENT:
675 self.maybe_decrement_after_for_loop_variable(leaf)
676 self.maybe_decrement_after_lambda_arguments(leaf)
677 if leaf.type in CLOSING_BRACKETS:
679 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
680 leaf.opening_bracket = opening_bracket
681 leaf.bracket_depth = self.depth
683 delim = is_split_before_delimiter(leaf, self.previous)
684 if delim and self.previous is not None:
685 self.delimiters[id(self.previous)] = delim
687 delim = is_split_after_delimiter(leaf, self.previous)
689 self.delimiters[id(leaf)] = delim
690 if leaf.type in OPENING_BRACKETS:
691 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
694 self.maybe_increment_lambda_arguments(leaf)
695 self.maybe_increment_for_loop_variable(leaf)
697 def any_open_brackets(self) -> bool:
698 """Return True if there is an yet unmatched open bracket on the line."""
699 return bool(self.bracket_match)
701 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int:
702 """Return the highest priority of a delimiter found on the line.
704 Values are consistent with what `is_split_*_delimiter()` return.
705 Raises ValueError on no delimiters.
707 return max(v for k, v in self.delimiters.items() if k not in exclude)
709 def delimiter_count_with_priority(self, priority: int = 0) -> int:
710 """Return the number of delimiters with the given `priority`.
712 If no `priority` is passed, defaults to max priority on the line.
714 if not self.delimiters:
717 priority = priority or self.max_delimiter_priority()
718 return sum(1 for p in self.delimiters.values() if p == priority)
720 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
721 """In a for loop, or comprehension, the variables are often unpacks.
723 To avoid splitting on the comma in this situation, increase the depth of
724 tokens between `for` and `in`.
726 if leaf.type == token.NAME and leaf.value == "for":
728 self._for_loop_variable += 1
733 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
734 """See `maybe_increment_for_loop_variable` above for explanation."""
735 if self._for_loop_variable and leaf.type == token.NAME and leaf.value == "in":
737 self._for_loop_variable -= 1
742 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
743 """In a lambda expression, there might be more than one argument.
745 To avoid splitting on the comma in this situation, increase the depth of
746 tokens between `lambda` and `:`.
748 if leaf.type == token.NAME and leaf.value == "lambda":
750 self._lambda_arguments += 1
755 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
756 """See `maybe_increment_lambda_arguments` above for explanation."""
757 if self._lambda_arguments and leaf.type == token.COLON:
759 self._lambda_arguments -= 1
764 def get_open_lsqb(self) -> Optional[Leaf]:
765 """Return the most recent opening square bracket (if any)."""
766 return self.bracket_match.get((self.depth - 1, token.RSQB))
771 """Holds leaves and comments. Can be printed with `str(line)`."""
774 leaves: List[Leaf] = Factory(list)
775 comments: List[Tuple[Index, Leaf]] = Factory(list)
776 bracket_tracker: BracketTracker = Factory(BracketTracker)
777 inside_brackets: bool = False
779 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
780 """Add a new `leaf` to the end of the line.
782 Unless `preformatted` is True, the `leaf` will receive a new consistent
783 whitespace prefix and metadata applied by :class:`BracketTracker`.
784 Trailing commas are maybe removed, unpacked for loop variables are
785 demoted from being delimiters.
787 Inline comments are put aside.
789 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
793 if token.COLON == leaf.type and self.is_class_paren_empty:
795 if self.leaves and not preformatted:
796 # Note: at this point leaf.prefix should be empty except for
797 # imports, for which we only preserve newlines.
798 leaf.prefix += whitespace(
799 leaf, complex_subscript=self.is_complex_subscript(leaf)
801 if self.inside_brackets or not preformatted:
802 self.bracket_tracker.mark(leaf)
803 self.maybe_remove_trailing_comma(leaf)
804 if not self.append_comment(leaf):
805 self.leaves.append(leaf)
807 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
808 """Like :func:`append()` but disallow invalid standalone comment structure.
810 Raises ValueError when any `leaf` is appended after a standalone comment
811 or when a standalone comment is not the first leaf on the line.
813 if self.bracket_tracker.depth == 0:
815 raise ValueError("cannot append to standalone comments")
817 if self.leaves and leaf.type == STANDALONE_COMMENT:
819 "cannot append standalone comments to a populated line"
822 self.append(leaf, preformatted=preformatted)
825 def is_comment(self) -> bool:
826 """Is this line a standalone comment?"""
827 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
830 def is_decorator(self) -> bool:
831 """Is this line a decorator?"""
832 return bool(self) and self.leaves[0].type == token.AT
835 def is_import(self) -> bool:
836 """Is this an import line?"""
837 return bool(self) and is_import(self.leaves[0])
840 def is_class(self) -> bool:
841 """Is this line a class definition?"""
844 and self.leaves[0].type == token.NAME
845 and self.leaves[0].value == "class"
849 def is_stub_class(self) -> bool:
850 """Is this line a class definition with a body consisting only of "..."?"""
851 return self.is_class and self.leaves[-3:] == [
852 Leaf(token.DOT, ".") for _ in range(3)
856 def is_def(self) -> bool:
857 """Is this a function definition? (Also returns True for async defs.)"""
859 first_leaf = self.leaves[0]
864 second_leaf: Optional[Leaf] = self.leaves[1]
867 return (first_leaf.type == token.NAME and first_leaf.value == "def") or (
868 first_leaf.type == token.ASYNC
869 and second_leaf is not None
870 and second_leaf.type == token.NAME
871 and second_leaf.value == "def"
875 def is_flow_control(self) -> bool:
876 """Is this line a flow control statement?
878 Those are `return`, `raise`, `break`, and `continue`.
882 and self.leaves[0].type == token.NAME
883 and self.leaves[0].value in FLOW_CONTROL
887 def is_yield(self) -> bool:
888 """Is this line a yield statement?"""
891 and self.leaves[0].type == token.NAME
892 and self.leaves[0].value == "yield"
896 def is_class_paren_empty(self) -> bool:
897 """Is this a class with no base classes but using parentheses?
899 Those are unnecessary and should be removed.
903 and len(self.leaves) == 4
905 and self.leaves[2].type == token.LPAR
906 and self.leaves[2].value == "("
907 and self.leaves[3].type == token.RPAR
908 and self.leaves[3].value == ")"
911 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
912 """If so, needs to be split before emitting."""
913 for leaf in self.leaves:
914 if leaf.type == STANDALONE_COMMENT:
915 if leaf.bracket_depth <= depth_limit:
920 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
921 """Remove trailing comma if there is one and it's safe."""
924 and self.leaves[-1].type == token.COMMA
925 and closing.type in CLOSING_BRACKETS
929 if closing.type == token.RBRACE:
930 self.remove_trailing_comma()
933 if closing.type == token.RSQB:
934 comma = self.leaves[-1]
935 if comma.parent and comma.parent.type == syms.listmaker:
936 self.remove_trailing_comma()
939 # For parens let's check if it's safe to remove the comma.
940 # Imports are always safe.
942 self.remove_trailing_comma()
945 # Otheriwsse, if the trailing one is the only one, we might mistakenly
946 # change a tuple into a different type by removing the comma.
947 depth = closing.bracket_depth + 1
949 opening = closing.opening_bracket
950 for _opening_index, leaf in enumerate(self.leaves):
957 for leaf in self.leaves[_opening_index + 1 :]:
961 bracket_depth = leaf.bracket_depth
962 if bracket_depth == depth and leaf.type == token.COMMA:
964 if leaf.parent and leaf.parent.type == syms.arglist:
969 self.remove_trailing_comma()
974 def append_comment(self, comment: Leaf) -> bool:
975 """Add an inline or standalone comment to the line."""
977 comment.type == STANDALONE_COMMENT
978 and self.bracket_tracker.any_open_brackets()
983 if comment.type != token.COMMENT:
986 after = len(self.leaves) - 1
988 comment.type = STANDALONE_COMMENT
993 self.comments.append((after, comment))
996 def comments_after(self, leaf: Leaf, _index: int = -1) -> Iterator[Leaf]:
997 """Generate comments that should appear directly after `leaf`.
999 Provide a non-negative leaf `_index` to speed up the function.
1002 for _index, _leaf in enumerate(self.leaves):
1009 for index, comment_after in self.comments:
1013 def remove_trailing_comma(self) -> None:
1014 """Remove the trailing comma and moves the comments attached to it."""
1015 comma_index = len(self.leaves) - 1
1016 for i in range(len(self.comments)):
1017 comment_index, comment = self.comments[i]
1018 if comment_index == comma_index:
1019 self.comments[i] = (comma_index - 1, comment)
1022 def is_complex_subscript(self, leaf: Leaf) -> bool:
1023 """Return True iff `leaf` is part of a slice with non-trivial exprs."""
1025 leaf if leaf.type == token.LSQB else self.bracket_tracker.get_open_lsqb()
1027 if open_lsqb is None:
1030 subscript_start = open_lsqb.next_sibling
1032 isinstance(subscript_start, Node)
1033 and subscript_start.type == syms.subscriptlist
1035 subscript_start = child_towards(subscript_start, leaf)
1036 return subscript_start is not None and any(
1037 n.type in TEST_DESCENDANTS for n in subscript_start.pre_order()
1040 def __str__(self) -> str:
1041 """Render the line."""
1045 indent = " " * self.depth
1046 leaves = iter(self.leaves)
1047 first = next(leaves)
1048 res = f"{first.prefix}{indent}{first.value}"
1051 for _, comment in self.comments:
1055 def __bool__(self) -> bool:
1056 """Return True if the line has leaves or comments."""
1057 return bool(self.leaves or self.comments)
1060 class UnformattedLines(Line):
1061 """Just like :class:`Line` but stores lines which aren't reformatted."""
1063 def append(self, leaf: Leaf, preformatted: bool = True) -> None:
1064 """Just add a new `leaf` to the end of the lines.
1066 The `preformatted` argument is ignored.
1068 Keeps track of indentation `depth`, which is useful when the user
1069 says `# fmt: on`. Otherwise, doesn't do anything with the `leaf`.
1072 list(generate_comments(leaf))
1073 except FormatOn as f_on:
1074 self.leaves.append(f_on.leaf_from_consumed(leaf))
1077 self.leaves.append(leaf)
1078 if leaf.type == token.INDENT:
1080 elif leaf.type == token.DEDENT:
1083 def __str__(self) -> str:
1084 """Render unformatted lines from leaves which were added with `append()`.
1086 `depth` is not used for indentation in this case.
1092 for leaf in self.leaves:
1096 def append_comment(self, comment: Leaf) -> bool:
1097 """Not implemented in this class. Raises `NotImplementedError`."""
1098 raise NotImplementedError("Unformatted lines don't store comments separately.")
1100 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
1101 """Does nothing and returns False."""
1104 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
1105 """Does nothing and returns False."""
1110 class EmptyLineTracker:
1111 """Provides a stateful method that returns the number of potential extra
1112 empty lines needed before and after the currently processed line.
1114 Note: this tracker works on lines that haven't been split yet. It assumes
1115 the prefix of the first leaf consists of optional newlines. Those newlines
1116 are consumed by `maybe_empty_lines()` and included in the computation.
1118 is_pyi: bool = False
1119 previous_line: Optional[Line] = None
1120 previous_after: int = 0
1121 previous_defs: List[int] = Factory(list)
1123 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1124 """Return the number of extra empty lines before and after the `current_line`.
1126 This is for separating `def`, `async def` and `class` with extra empty
1127 lines (two on module-level), as well as providing an extra empty line
1128 after flow control keywords to make them more prominent.
1130 if isinstance(current_line, UnformattedLines):
1133 before, after = self._maybe_empty_lines(current_line)
1134 before -= self.previous_after
1135 self.previous_after = after
1136 self.previous_line = current_line
1137 return before, after
1139 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1141 if current_line.depth == 0:
1142 max_allowed = 1 if self.is_pyi else 2
1143 if current_line.leaves:
1144 # Consume the first leaf's extra newlines.
1145 first_leaf = current_line.leaves[0]
1146 before = first_leaf.prefix.count("\n")
1147 before = min(before, max_allowed)
1148 first_leaf.prefix = ""
1151 depth = current_line.depth
1152 while self.previous_defs and self.previous_defs[-1] >= depth:
1153 self.previous_defs.pop()
1155 before = 0 if depth else 1
1157 before = 1 if depth else 2
1158 is_decorator = current_line.is_decorator
1159 if is_decorator or current_line.is_def or current_line.is_class:
1160 if not is_decorator:
1161 self.previous_defs.append(depth)
1162 if self.previous_line is None:
1163 # Don't insert empty lines before the first line in the file.
1166 if self.previous_line.is_decorator:
1170 self.previous_line.is_comment
1171 and self.previous_line.depth == current_line.depth
1177 if self.previous_line.depth > current_line.depth:
1179 elif current_line.is_class or self.previous_line.is_class:
1180 if current_line.is_stub_class and self.previous_line.is_stub_class:
1188 if current_line.depth and newlines:
1194 and self.previous_line.is_import
1195 and not current_line.is_import
1196 and depth == self.previous_line.depth
1198 return (before or 1), 0
1204 class LineGenerator(Visitor[Line]):
1205 """Generates reformatted Line objects. Empty lines are not emitted.
1207 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1208 in ways that will no longer stringify to valid Python code on the tree.
1210 is_pyi: bool = False
1211 current_line: Line = Factory(Line)
1212 remove_u_prefix: bool = False
1214 def line(self, indent: int = 0, type: Type[Line] = Line) -> Iterator[Line]:
1217 If the line is empty, only emit if it makes sense.
1218 If the line is too long, split it first and then generate.
1220 If any lines were generated, set up a new current_line.
1222 if not self.current_line:
1223 if self.current_line.__class__ == type:
1224 self.current_line.depth += indent
1226 self.current_line = type(depth=self.current_line.depth + indent)
1227 return # Line is empty, don't emit. Creating a new one unnecessary.
1229 complete_line = self.current_line
1230 self.current_line = type(depth=complete_line.depth + indent)
1233 def visit(self, node: LN) -> Iterator[Line]:
1234 """Main method to visit `node` and its children.
1236 Yields :class:`Line` objects.
1238 if isinstance(self.current_line, UnformattedLines):
1239 # File contained `# fmt: off`
1240 yield from self.visit_unformatted(node)
1243 yield from super().visit(node)
1245 def visit_default(self, node: LN) -> Iterator[Line]:
1246 """Default `visit_*()` implementation. Recurses to children of `node`."""
1247 if isinstance(node, Leaf):
1248 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1250 for comment in generate_comments(node):
1251 if any_open_brackets:
1252 # any comment within brackets is subject to splitting
1253 self.current_line.append(comment)
1254 elif comment.type == token.COMMENT:
1255 # regular trailing comment
1256 self.current_line.append(comment)
1257 yield from self.line()
1260 # regular standalone comment
1261 yield from self.line()
1263 self.current_line.append(comment)
1264 yield from self.line()
1266 except FormatOff as f_off:
1267 f_off.trim_prefix(node)
1268 yield from self.line(type=UnformattedLines)
1269 yield from self.visit(node)
1271 except FormatOn as f_on:
1272 # This only happens here if somebody says "fmt: on" multiple
1274 f_on.trim_prefix(node)
1275 yield from self.visit_default(node)
1278 normalize_prefix(node, inside_brackets=any_open_brackets)
1279 if node.type == token.STRING:
1280 normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix)
1281 normalize_string_quotes(node)
1282 if node.type not in WHITESPACE:
1283 self.current_line.append(node)
1284 yield from super().visit_default(node)
1286 def visit_INDENT(self, node: Node) -> Iterator[Line]:
1287 """Increase indentation level, maybe yield a line."""
1288 # In blib2to3 INDENT never holds comments.
1289 yield from self.line(+1)
1290 yield from self.visit_default(node)
1292 def visit_DEDENT(self, node: Node) -> Iterator[Line]:
1293 """Decrease indentation level, maybe yield a line."""
1294 # The current line might still wait for trailing comments. At DEDENT time
1295 # there won't be any (they would be prefixes on the preceding NEWLINE).
1296 # Emit the line then.
1297 yield from self.line()
1299 # While DEDENT has no value, its prefix may contain standalone comments
1300 # that belong to the current indentation level. Get 'em.
1301 yield from self.visit_default(node)
1303 # Finally, emit the dedent.
1304 yield from self.line(-1)
1307 self, node: Node, keywords: Set[str], parens: Set[str]
1308 ) -> Iterator[Line]:
1309 """Visit a statement.
1311 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1312 `def`, `with`, `class`, `assert` and assignments.
1314 The relevant Python language `keywords` for a given statement will be
1315 NAME leaves within it. This methods puts those on a separate line.
1317 `parens` holds a set of string leaf values immeditely after which
1318 invisible parens should be put.
1320 normalize_invisible_parens(node, parens_after=parens)
1321 for child in node.children:
1322 if child.type == token.NAME and child.value in keywords: # type: ignore
1323 yield from self.line()
1325 yield from self.visit(child)
1327 def visit_suite(self, node: Node) -> Iterator[Line]:
1328 """Visit a suite."""
1329 if self.is_pyi and is_stub_suite(node):
1330 yield from self.visit(node.children[2])
1332 yield from self.visit_default(node)
1334 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1335 """Visit a statement without nested statements."""
1336 is_suite_like = node.parent and node.parent.type in STATEMENT
1338 if self.is_pyi and is_stub_body(node):
1339 yield from self.visit_default(node)
1341 yield from self.line(+1)
1342 yield from self.visit_default(node)
1343 yield from self.line(-1)
1346 if not self.is_pyi or not node.parent or not is_stub_suite(node.parent):
1347 yield from self.line()
1348 yield from self.visit_default(node)
1350 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1351 """Visit `async def`, `async for`, `async with`."""
1352 yield from self.line()
1354 children = iter(node.children)
1355 for child in children:
1356 yield from self.visit(child)
1358 if child.type == token.ASYNC:
1361 internal_stmt = next(children)
1362 for child in internal_stmt.children:
1363 yield from self.visit(child)
1365 def visit_decorators(self, node: Node) -> Iterator[Line]:
1366 """Visit decorators."""
1367 for child in node.children:
1368 yield from self.line()
1369 yield from self.visit(child)
1371 def visit_import_from(self, node: Node) -> Iterator[Line]:
1372 """Visit import_from and maybe put invisible parentheses.
1374 This is separate from `visit_stmt` because import statements don't
1375 support arbitrary atoms and thus handling of parentheses is custom.
1378 for index, child in enumerate(node.children):
1380 if child.type == token.LPAR:
1381 # make parentheses invisible
1382 child.value = "" # type: ignore
1383 node.children[-1].value = "" # type: ignore
1385 # insert invisible parentheses
1386 node.insert_child(index, Leaf(token.LPAR, ""))
1387 node.append_child(Leaf(token.RPAR, ""))
1391 child.type == token.NAME and child.value == "import" # type: ignore
1394 for child in node.children:
1395 yield from self.visit(child)
1397 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1398 """Remove a semicolon and put the other statement on a separate line."""
1399 yield from self.line()
1401 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1402 """End of file. Process outstanding comments and end with a newline."""
1403 yield from self.visit_default(leaf)
1404 yield from self.line()
1406 def visit_unformatted(self, node: LN) -> Iterator[Line]:
1407 """Used when file contained a `# fmt: off`."""
1408 if isinstance(node, Node):
1409 for child in node.children:
1410 yield from self.visit(child)
1414 self.current_line.append(node)
1415 except FormatOn as f_on:
1416 f_on.trim_prefix(node)
1417 yield from self.line()
1418 yield from self.visit(node)
1420 if node.type == token.ENDMARKER:
1421 # somebody decided not to put a final `# fmt: on`
1422 yield from self.line()
1424 def __attrs_post_init__(self) -> None:
1425 """You are in a twisty little maze of passages."""
1428 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1429 self.visit_if_stmt = partial(
1430 v, keywords={"if", "else", "elif"}, parens={"if", "elif"}
1432 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1433 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1434 self.visit_try_stmt = partial(
1435 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1437 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1438 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1439 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1440 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1441 self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
1442 self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
1443 self.visit_async_funcdef = self.visit_async_stmt
1444 self.visit_decorated = self.visit_decorators
1447 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1448 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1449 OPENING_BRACKETS = set(BRACKET.keys())
1450 CLOSING_BRACKETS = set(BRACKET.values())
1451 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1452 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1455 def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa C901
1456 """Return whitespace prefix if needed for the given `leaf`.
1458 `complex_subscript` signals whether the given leaf is part of a subscription
1459 which has non-trivial arguments, like arithmetic expressions or function calls.
1467 if t in ALWAYS_NO_SPACE:
1470 if t == token.COMMENT:
1473 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1474 if t == token.COLON and p.type not in {
1475 syms.subscript, syms.subscriptlist, syms.sliceop
1479 prev = leaf.prev_sibling
1481 prevp = preceding_leaf(p)
1482 if not prevp or prevp.type in OPENING_BRACKETS:
1485 if t == token.COLON:
1486 if prevp.type == token.COLON:
1489 elif prevp.type != token.COMMA and not complex_subscript:
1494 if prevp.type == token.EQUAL:
1496 if prevp.parent.type in {
1497 syms.arglist, syms.argument, syms.parameters, syms.varargslist
1501 elif prevp.parent.type == syms.typedargslist:
1502 # A bit hacky: if the equal sign has whitespace, it means we
1503 # previously found it's a typed argument. So, we're using
1507 elif prevp.type in STARS:
1508 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1511 elif prevp.type == token.COLON:
1512 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1513 return SPACE if complex_subscript else NO
1517 and prevp.parent.type == syms.factor
1518 and prevp.type in MATH_OPERATORS
1523 prevp.type == token.RIGHTSHIFT
1525 and prevp.parent.type == syms.shift_expr
1526 and prevp.prev_sibling
1527 and prevp.prev_sibling.type == token.NAME
1528 and prevp.prev_sibling.value == "print" # type: ignore
1530 # Python 2 print chevron
1533 elif prev.type in OPENING_BRACKETS:
1536 if p.type in {syms.parameters, syms.arglist}:
1537 # untyped function signatures or calls
1538 if not prev or prev.type != token.COMMA:
1541 elif p.type == syms.varargslist:
1543 if prev and prev.type != token.COMMA:
1546 elif p.type == syms.typedargslist:
1547 # typed function signatures
1551 if t == token.EQUAL:
1552 if prev.type != syms.tname:
1555 elif prev.type == token.EQUAL:
1556 # A bit hacky: if the equal sign has whitespace, it means we
1557 # previously found it's a typed argument. So, we're using that, too.
1560 elif prev.type != token.COMMA:
1563 elif p.type == syms.tname:
1566 prevp = preceding_leaf(p)
1567 if not prevp or prevp.type != token.COMMA:
1570 elif p.type == syms.trailer:
1571 # attributes and calls
1572 if t == token.LPAR or t == token.RPAR:
1577 prevp = preceding_leaf(p)
1578 if not prevp or prevp.type != token.NUMBER:
1581 elif t == token.LSQB:
1584 elif prev.type != token.COMMA:
1587 elif p.type == syms.argument:
1589 if t == token.EQUAL:
1593 prevp = preceding_leaf(p)
1594 if not prevp or prevp.type == token.LPAR:
1597 elif prev.type in {token.EQUAL} | STARS:
1600 elif p.type == syms.decorator:
1604 elif p.type == syms.dotted_name:
1608 prevp = preceding_leaf(p)
1609 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
1612 elif p.type == syms.classdef:
1616 if prev and prev.type == token.LPAR:
1619 elif p.type in {syms.subscript, syms.sliceop}:
1622 assert p.parent is not None, "subscripts are always parented"
1623 if p.parent.type == syms.subscriptlist:
1628 elif not complex_subscript:
1631 elif p.type == syms.atom:
1632 if prev and t == token.DOT:
1633 # dots, but not the first one.
1636 elif p.type == syms.dictsetmaker:
1638 if prev and prev.type == token.DOUBLESTAR:
1641 elif p.type in {syms.factor, syms.star_expr}:
1644 prevp = preceding_leaf(p)
1645 if not prevp or prevp.type in OPENING_BRACKETS:
1648 prevp_parent = prevp.parent
1649 assert prevp_parent is not None
1650 if prevp.type == token.COLON and prevp_parent.type in {
1651 syms.subscript, syms.sliceop
1655 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
1658 elif t == token.NAME or t == token.NUMBER:
1661 elif p.type == syms.import_from:
1663 if prev and prev.type == token.DOT:
1666 elif t == token.NAME:
1670 if prev and prev.type == token.DOT:
1673 elif p.type == syms.sliceop:
1679 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
1680 """Return the first leaf that precedes `node`, if any."""
1682 res = node.prev_sibling
1684 if isinstance(res, Leaf):
1688 return list(res.leaves())[-1]
1697 def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]:
1698 """Return the child of `ancestor` that contains `descendant`."""
1699 node: Optional[LN] = descendant
1700 while node and node.parent != ancestor:
1705 def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1706 """Return the priority of the `leaf` delimiter, given a line break after it.
1708 The delimiter priorities returned here are from those delimiters that would
1709 cause a line break after themselves.
1711 Higher numbers are higher priority.
1713 if leaf.type == token.COMMA:
1714 return COMMA_PRIORITY
1719 def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1720 """Return the priority of the `leaf` delimiter, given a line before after it.
1722 The delimiter priorities returned here are from those delimiters that would
1723 cause a line break before themselves.
1725 Higher numbers are higher priority.
1727 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1728 # * and ** might also be MATH_OPERATORS but in this case they are not.
1729 # Don't treat them as a delimiter.
1733 leaf.type in MATH_OPERATORS
1735 and leaf.parent.type not in {syms.factor, syms.star_expr}
1737 return MATH_PRIORITIES[leaf.type]
1739 if leaf.type in COMPARATORS:
1740 return COMPARATOR_PRIORITY
1743 leaf.type == token.STRING
1744 and previous is not None
1745 and previous.type == token.STRING
1747 return STRING_PRIORITY
1749 if leaf.type != token.NAME:
1755 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
1757 return COMPREHENSION_PRIORITY
1762 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
1764 return COMPREHENSION_PRIORITY
1766 if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test:
1767 return TERNARY_PRIORITY
1769 if leaf.value == "is":
1770 return COMPARATOR_PRIORITY
1775 and leaf.parent.type in {syms.comp_op, syms.comparison}
1777 previous is not None
1778 and previous.type == token.NAME
1779 and previous.value == "not"
1782 return COMPARATOR_PRIORITY
1787 and leaf.parent.type == syms.comp_op
1789 previous is not None
1790 and previous.type == token.NAME
1791 and previous.value == "is"
1794 return COMPARATOR_PRIORITY
1796 if leaf.value in LOGIC_OPERATORS and leaf.parent:
1797 return LOGIC_PRIORITY
1802 def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
1803 """Clean the prefix of the `leaf` and generate comments from it, if any.
1805 Comments in lib2to3 are shoved into the whitespace prefix. This happens
1806 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
1807 move because it does away with modifying the grammar to include all the
1808 possible places in which comments can be placed.
1810 The sad consequence for us though is that comments don't "belong" anywhere.
1811 This is why this function generates simple parentless Leaf objects for
1812 comments. We simply don't know what the correct parent should be.
1814 No matter though, we can live without this. We really only need to
1815 differentiate between inline and standalone comments. The latter don't
1816 share the line with any code.
1818 Inline comments are emitted as regular token.COMMENT leaves. Standalone
1819 are emitted with a fake STANDALONE_COMMENT token identifier.
1830 for index, line in enumerate(p.split("\n")):
1831 consumed += len(line) + 1 # adding the length of the split '\n'
1832 line = line.lstrip()
1835 if not line.startswith("#"):
1838 if index == 0 and leaf.type != token.ENDMARKER:
1839 comment_type = token.COMMENT # simple trailing comment
1841 comment_type = STANDALONE_COMMENT
1842 comment = make_comment(line)
1843 yield Leaf(comment_type, comment, prefix="\n" * nlines)
1845 if comment in {"# fmt: on", "# yapf: enable"}:
1846 raise FormatOn(consumed)
1848 if comment in {"# fmt: off", "# yapf: disable"}:
1849 if comment_type == STANDALONE_COMMENT:
1850 raise FormatOff(consumed)
1852 prev = preceding_leaf(leaf)
1853 if not prev or prev.type in WHITESPACE: # standalone comment in disguise
1854 raise FormatOff(consumed)
1859 def make_comment(content: str) -> str:
1860 """Return a consistently formatted comment from the given `content` string.
1862 All comments (except for "##", "#!", "#:") should have a single space between
1863 the hash sign and the content.
1865 If `content` didn't start with a hash sign, one is provided.
1867 content = content.rstrip()
1871 if content[0] == "#":
1872 content = content[1:]
1873 if content and content[0] not in " !:#":
1874 content = " " + content
1875 return "#" + content
1879 line: Line, line_length: int, inner: bool = False, py36: bool = False
1880 ) -> Iterator[Line]:
1881 """Split a `line` into potentially many lines.
1883 They should fit in the allotted `line_length` but might not be able to.
1884 `inner` signifies that there were a pair of brackets somewhere around the
1885 current `line`, possibly transitively. This means we can fallback to splitting
1886 by delimiters if the LHS/RHS don't yield any results.
1888 If `py36` is True, splitting may generate syntax that is only compatible
1889 with Python 3.6 and later.
1891 if isinstance(line, UnformattedLines) or line.is_comment:
1895 line_str = str(line).strip("\n")
1896 if is_line_short_enough(line, line_length=line_length, line_str=line_str):
1900 split_funcs: List[SplitFunc]
1902 split_funcs = [left_hand_split]
1903 elif line.is_import:
1904 split_funcs = [explode_split]
1907 def rhs(line: Line, py36: bool = False) -> Iterator[Line]:
1908 for omit in generate_trailers_to_omit(line, line_length):
1909 lines = list(right_hand_split(line, py36, omit=omit))
1910 if is_line_short_enough(lines[0], line_length=line_length):
1914 # All splits failed, best effort split with no omits.
1915 yield from right_hand_split(line, py36)
1917 if line.inside_brackets:
1918 split_funcs = [delimiter_split, standalone_comment_split, rhs]
1921 for split_func in split_funcs:
1922 # We are accumulating lines in `result` because we might want to abort
1923 # mission and return the original line in the end, or attempt a different
1925 result: List[Line] = []
1927 for l in split_func(line, py36):
1928 if str(l).strip("\n") == line_str:
1929 raise CannotSplit("Split function returned an unchanged result")
1932 split_line(l, line_length=line_length, inner=True, py36=py36)
1934 except CannotSplit as cs:
1945 def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
1946 """Split line into many lines, starting with the first matching bracket pair.
1948 Note: this usually looks weird, only use this for function definitions.
1949 Prefer RHS otherwise. This is why this function is not symmetrical with
1950 :func:`right_hand_split` which also handles optional parentheses.
1952 head = Line(depth=line.depth)
1953 body = Line(depth=line.depth + 1, inside_brackets=True)
1954 tail = Line(depth=line.depth)
1955 tail_leaves: List[Leaf] = []
1956 body_leaves: List[Leaf] = []
1957 head_leaves: List[Leaf] = []
1958 current_leaves = head_leaves
1959 matching_bracket = None
1960 for leaf in line.leaves:
1962 current_leaves is body_leaves
1963 and leaf.type in CLOSING_BRACKETS
1964 and leaf.opening_bracket is matching_bracket
1966 current_leaves = tail_leaves if body_leaves else head_leaves
1967 current_leaves.append(leaf)
1968 if current_leaves is head_leaves:
1969 if leaf.type in OPENING_BRACKETS:
1970 matching_bracket = leaf
1971 current_leaves = body_leaves
1972 # Since body is a new indent level, remove spurious leading whitespace.
1974 normalize_prefix(body_leaves[0], inside_brackets=True)
1975 # Build the new lines.
1976 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
1978 result.append(leaf, preformatted=True)
1979 for comment_after in line.comments_after(leaf):
1980 result.append(comment_after, preformatted=True)
1981 bracket_split_succeeded_or_raise(head, body, tail)
1982 for result in (head, body, tail):
1987 def right_hand_split(
1988 line: Line, py36: bool = False, omit: Collection[LeafID] = ()
1989 ) -> Iterator[Line]:
1990 """Split line into many lines, starting with the last matching bracket pair.
1992 If the split was by optional parentheses, attempt splitting without them, too.
1993 `omit` is a collection of closing bracket IDs that shouldn't be considered for
1996 Note: running this function modifies `bracket_depth` on the leaves of `line`.
1998 head = Line(depth=line.depth)
1999 body = Line(depth=line.depth + 1, inside_brackets=True)
2000 tail = Line(depth=line.depth)
2001 tail_leaves: List[Leaf] = []
2002 body_leaves: List[Leaf] = []
2003 head_leaves: List[Leaf] = []
2004 current_leaves = tail_leaves
2005 opening_bracket = None
2006 closing_bracket = None
2007 for leaf in reversed(line.leaves):
2008 if current_leaves is body_leaves:
2009 if leaf is opening_bracket:
2010 current_leaves = head_leaves if body_leaves else tail_leaves
2011 current_leaves.append(leaf)
2012 if current_leaves is tail_leaves:
2013 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
2014 opening_bracket = leaf.opening_bracket
2015 closing_bracket = leaf
2016 current_leaves = body_leaves
2017 tail_leaves.reverse()
2018 body_leaves.reverse()
2019 head_leaves.reverse()
2020 # Since body is a new indent level, remove spurious leading whitespace.
2022 normalize_prefix(body_leaves[0], inside_brackets=True)
2024 # No `head` means the split failed. Either `tail` has all content or
2025 # the matching `opening_bracket` wasn't available on `line` anymore.
2026 raise CannotSplit("No brackets found")
2028 # Build the new lines.
2029 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
2031 result.append(leaf, preformatted=True)
2032 for comment_after in line.comments_after(leaf):
2033 result.append(comment_after, preformatted=True)
2034 bracket_split_succeeded_or_raise(head, body, tail)
2035 assert opening_bracket and closing_bracket
2037 # the opening bracket is an optional paren
2038 opening_bracket.type == token.LPAR
2039 and not opening_bracket.value
2040 # the closing bracket is an optional paren
2041 and closing_bracket.type == token.RPAR
2042 and not closing_bracket.value
2043 # there are no standalone comments in the body
2044 and not line.contains_standalone_comments(0)
2045 # and it's not an import (optional parens are the only thing we can split
2046 # on in this case; attempting a split without them is a waste of time)
2047 and not line.is_import
2049 omit = {id(closing_bracket), *omit}
2050 delimiter_count = body.bracket_tracker.delimiter_count_with_priority()
2052 delimiter_count == 0
2053 or delimiter_count == 1
2055 body.leaves[0].type in OPENING_BRACKETS
2056 or body.leaves[-1].type in CLOSING_BRACKETS
2060 yield from right_hand_split(line, py36=py36, omit=omit)
2065 ensure_visible(opening_bracket)
2066 ensure_visible(closing_bracket)
2067 for result in (head, body, tail):
2072 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
2073 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
2075 Do nothing otherwise.
2077 A left- or right-hand split is based on a pair of brackets. Content before
2078 (and including) the opening bracket is left on one line, content inside the
2079 brackets is put on a separate line, and finally content starting with and
2080 following the closing bracket is put on a separate line.
2082 Those are called `head`, `body`, and `tail`, respectively. If the split
2083 produced the same line (all content in `head`) or ended up with an empty `body`
2084 and the `tail` is just the closing bracket, then it's considered failed.
2086 tail_len = len(str(tail).strip())
2089 raise CannotSplit("Splitting brackets produced the same line")
2093 f"Splitting brackets on an empty body to save "
2094 f"{tail_len} characters is not worth it"
2098 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
2099 """Normalize prefix of the first leaf in every line returned by `split_func`.
2101 This is a decorator over relevant split functions.
2105 def split_wrapper(line: Line, py36: bool = False) -> Iterator[Line]:
2106 for l in split_func(line, py36):
2107 normalize_prefix(l.leaves[0], inside_brackets=True)
2110 return split_wrapper
2113 @dont_increase_indentation
2114 def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]:
2115 """Split according to delimiters of the highest priority.
2117 If `py36` is True, the split will add trailing commas also in function
2118 signatures that contain `*` and `**`.
2121 last_leaf = line.leaves[-1]
2123 raise CannotSplit("Line empty")
2125 bt = line.bracket_tracker
2127 delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)})
2129 raise CannotSplit("No delimiters found")
2131 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2132 lowest_depth = sys.maxsize
2133 trailing_comma_safe = True
2135 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2136 """Append `leaf` to current line or to new line if appending impossible."""
2137 nonlocal current_line
2139 current_line.append_safe(leaf, preformatted=True)
2140 except ValueError as ve:
2143 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2144 current_line.append(leaf)
2146 for index, leaf in enumerate(line.leaves):
2147 yield from append_to_line(leaf)
2149 for comment_after in line.comments_after(leaf, index):
2150 yield from append_to_line(comment_after)
2152 lowest_depth = min(lowest_depth, leaf.bracket_depth)
2153 if leaf.bracket_depth == lowest_depth and is_vararg(
2154 leaf, within=VARARGS_PARENTS
2156 trailing_comma_safe = trailing_comma_safe and py36
2157 leaf_priority = bt.delimiters.get(id(leaf))
2158 if leaf_priority == delimiter_priority:
2161 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2165 and delimiter_priority == COMMA_PRIORITY
2166 and current_line.leaves[-1].type != token.COMMA
2167 and current_line.leaves[-1].type != STANDALONE_COMMENT
2169 current_line.append(Leaf(token.COMMA, ","))
2173 @dont_increase_indentation
2174 def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]:
2175 """Split standalone comments from the rest of the line."""
2176 if not line.contains_standalone_comments(0):
2177 raise CannotSplit("Line does not have any standalone comments")
2179 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2181 def append_to_line(leaf: Leaf) -> Iterator[Line]:
2182 """Append `leaf` to current line or to new line if appending impossible."""
2183 nonlocal current_line
2185 current_line.append_safe(leaf, preformatted=True)
2186 except ValueError as ve:
2189 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
2190 current_line.append(leaf)
2192 for index, leaf in enumerate(line.leaves):
2193 yield from append_to_line(leaf)
2195 for comment_after in line.comments_after(leaf, index):
2196 yield from append_to_line(comment_after)
2203 line: Line, py36: bool = False, omit: Collection[LeafID] = ()
2204 ) -> Iterator[Line]:
2205 """Split by rightmost bracket and immediately split contents by a delimiter."""
2206 new_lines = list(right_hand_split(line, py36, omit))
2207 if len(new_lines) != 3:
2208 yield from new_lines
2214 yield from delimiter_split(new_lines[1], py36)
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]))
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_stub_suite(node: Node) -> bool:
2463 """Return True if `node` is a suite with a stub body."""
2465 len(node.children) != 4
2466 or node.children[0].type != token.NEWLINE
2467 or node.children[1].type != token.INDENT
2468 or node.children[3].type != token.DEDENT
2472 return is_stub_body(node.children[2])
2475 def is_stub_body(node: LN) -> bool:
2476 """Return True if `node` is a simple statement containing an ellipsis."""
2477 if not isinstance(node, Node) or node.type != syms.simple_stmt:
2480 if len(node.children) != 2:
2483 child = node.children[0]
2485 child.type == syms.atom
2486 and len(child.children) == 3
2487 and all(leaf == Leaf(token.DOT, ".") for leaf in child.children)
2491 def max_delimiter_priority_in_atom(node: LN) -> int:
2492 """Return maximum delimiter priority inside `node`.
2494 This is specific to atoms with contents contained in a pair of parentheses.
2495 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
2497 if node.type != syms.atom:
2500 first = node.children[0]
2501 last = node.children[-1]
2502 if not (first.type == token.LPAR and last.type == token.RPAR):
2505 bt = BracketTracker()
2506 for c in node.children[1:-1]:
2507 if isinstance(c, Leaf):
2510 for leaf in c.leaves():
2513 return bt.max_delimiter_priority()
2519 def ensure_visible(leaf: Leaf) -> None:
2520 """Make sure parentheses are visible.
2522 They could be invisible as part of some statements (see
2523 :func:`normalize_invible_parens` and :func:`visit_import_from`).
2525 if leaf.type == token.LPAR:
2527 elif leaf.type == token.RPAR:
2531 def is_python36(node: Node) -> bool:
2532 """Return True if the current file is using Python 3.6+ features.
2534 Currently looking for:
2536 - trailing commas after * or ** in function signatures and calls.
2538 for n in node.pre_order():
2539 if n.type == token.STRING:
2540 value_head = n.value[:2] # type: ignore
2541 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
2545 n.type in {syms.typedargslist, syms.arglist}
2547 and n.children[-1].type == token.COMMA
2549 for ch in n.children:
2550 if ch.type in STARS:
2553 if ch.type == syms.argument:
2554 for argch in ch.children:
2555 if argch.type in STARS:
2561 def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]:
2562 """Generate sets of closing bracket IDs that should be omitted in a RHS.
2564 Brackets can be omitted if the entire trailer up to and including
2565 a preceding closing bracket fits in one line.
2567 Yielded sets are cumulative (contain results of previous yields, too). First
2571 omit: Set[LeafID] = set()
2574 length = 4 * line.depth
2575 opening_bracket = None
2576 closing_bracket = None
2577 optional_brackets: Set[LeafID] = set()
2578 inner_brackets: Set[LeafID] = set()
2579 for index, leaf in enumerate_reversed(line.leaves):
2580 length += len(leaf.prefix) + len(leaf.value)
2581 if length > line_length:
2584 comment: Optional[Leaf]
2585 for comment in line.comments_after(leaf, index):
2586 if "\n" in comment.prefix:
2587 break # Oops, standalone comment!
2589 length += len(comment.value)
2592 if comment is not None:
2593 break # There was a standalone comment, we can't continue.
2595 optional_brackets.discard(id(leaf))
2597 if leaf is opening_bracket:
2598 opening_bracket = None
2599 elif leaf.type in CLOSING_BRACKETS:
2600 inner_brackets.add(id(leaf))
2601 elif leaf.type in CLOSING_BRACKETS:
2603 optional_brackets.add(id(opening_bracket))
2606 if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS:
2607 # Empty brackets would fail a split so treat them as "inner"
2608 # brackets (e.g. only add them to the `omit` set if another
2609 # pair of brackets was good enough.
2610 inner_brackets.add(id(leaf))
2613 opening_bracket = leaf.opening_bracket
2615 omit.add(id(closing_bracket))
2616 omit.update(inner_brackets)
2617 inner_brackets.clear()
2619 closing_bracket = leaf
2622 def get_future_imports(node: Node) -> Set[str]:
2623 """Return a set of __future__ imports in the file."""
2625 for child in node.children:
2626 if child.type != syms.simple_stmt:
2628 first_child = child.children[0]
2629 if isinstance(first_child, Leaf):
2630 # Continue looking if we see a docstring; otherwise stop.
2632 len(child.children) == 2
2633 and first_child.type == token.STRING
2634 and child.children[1].type == token.NEWLINE
2639 elif first_child.type == syms.import_from:
2640 module_name = first_child.children[1]
2641 if not isinstance(module_name, Leaf) or module_name.value != "__future__":
2643 for import_from_child in first_child.children[3:]:
2644 if isinstance(import_from_child, Leaf):
2645 if import_from_child.type == token.NAME:
2646 imports.add(import_from_child.value)
2648 assert import_from_child.type == syms.import_as_names
2649 for leaf in import_from_child.children:
2650 if isinstance(leaf, Leaf) and leaf.type == token.NAME:
2651 imports.add(leaf.value)
2657 PYTHON_EXTENSIONS = {".py", ".pyi"}
2658 BLACKLISTED_DIRECTORIES = {
2659 "build", "buck-out", "dist", "_build", ".git", ".hg", ".mypy_cache", ".tox", ".venv"
2663 def gen_python_files_in_dir(path: Path) -> Iterator[Path]:
2664 """Generate all files under `path` which aren't under BLACKLISTED_DIRECTORIES
2665 and have one of the PYTHON_EXTENSIONS.
2667 for child in path.iterdir():
2669 if child.name in BLACKLISTED_DIRECTORIES:
2672 yield from gen_python_files_in_dir(child)
2674 elif child.is_file() and child.suffix in PYTHON_EXTENSIONS:
2680 """Provides a reformatting counter. Can be rendered with `str(report)`."""
2683 change_count: int = 0
2685 failure_count: int = 0
2687 def done(self, src: Path, changed: Changed) -> None:
2688 """Increment the counter for successful reformatting. Write out a message."""
2689 if changed is Changed.YES:
2690 reformatted = "would reformat" if self.check else "reformatted"
2692 out(f"{reformatted} {src}")
2693 self.change_count += 1
2696 if changed is Changed.NO:
2697 msg = f"{src} already well formatted, good job."
2699 msg = f"{src} wasn't modified on disk since last run."
2700 out(msg, bold=False)
2701 self.same_count += 1
2703 def failed(self, src: Path, message: str) -> None:
2704 """Increment the counter for failed reformatting. Write out a message."""
2705 err(f"error: cannot format {src}: {message}")
2706 self.failure_count += 1
2709 def return_code(self) -> int:
2710 """Return the exit code that the app should use.
2712 This considers the current state of changed files and failures:
2713 - if there were any failures, return 123;
2714 - if any files were changed and --check is being used, return 1;
2715 - otherwise return 0.
2717 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
2718 # 126 we have special returncodes reserved by the shell.
2719 if self.failure_count:
2722 elif self.change_count and self.check:
2727 def __str__(self) -> str:
2728 """Render a color report of the current state.
2730 Use `click.unstyle` to remove colors.
2733 reformatted = "would be reformatted"
2734 unchanged = "would be left unchanged"
2735 failed = "would fail to reformat"
2737 reformatted = "reformatted"
2738 unchanged = "left unchanged"
2739 failed = "failed to reformat"
2741 if self.change_count:
2742 s = "s" if self.change_count > 1 else ""
2744 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
2747 s = "s" if self.same_count > 1 else ""
2748 report.append(f"{self.same_count} file{s} {unchanged}")
2749 if self.failure_count:
2750 s = "s" if self.failure_count > 1 else ""
2752 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
2754 return ", ".join(report) + "."
2757 def assert_equivalent(src: str, dst: str) -> None:
2758 """Raise AssertionError if `src` and `dst` aren't equivalent."""
2763 def _v(node: ast.AST, depth: int = 0) -> Iterator[str]:
2764 """Simple visitor generating strings to compare ASTs by content."""
2765 yield f"{' ' * depth}{node.__class__.__name__}("
2767 for field in sorted(node._fields):
2769 value = getattr(node, field)
2770 except AttributeError:
2773 yield f"{' ' * (depth+1)}{field}="
2775 if isinstance(value, list):
2777 if isinstance(item, ast.AST):
2778 yield from _v(item, depth + 2)
2780 elif isinstance(value, ast.AST):
2781 yield from _v(value, depth + 2)
2784 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
2786 yield f"{' ' * depth}) # /{node.__class__.__name__}"
2789 src_ast = ast.parse(src)
2790 except Exception as exc:
2791 major, minor = sys.version_info[:2]
2792 raise AssertionError(
2793 f"cannot use --safe with this file; failed to parse source file "
2794 f"with Python {major}.{minor}'s builtin AST. Re-run with --fast "
2795 f"or stop using deprecated Python 2 syntax. AST error message: {exc}"
2799 dst_ast = ast.parse(dst)
2800 except Exception as exc:
2801 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
2802 raise AssertionError(
2803 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
2804 f"Please report a bug on https://github.com/ambv/black/issues. "
2805 f"This invalid output might be helpful: {log}"
2808 src_ast_str = "\n".join(_v(src_ast))
2809 dst_ast_str = "\n".join(_v(dst_ast))
2810 if src_ast_str != dst_ast_str:
2811 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
2812 raise AssertionError(
2813 f"INTERNAL ERROR: Black produced code that is not equivalent to "
2815 f"Please report a bug on https://github.com/ambv/black/issues. "
2816 f"This diff might be helpful: {log}"
2820 def assert_stable(src: str, dst: str, line_length: int, is_pyi: bool = False) -> None:
2821 """Raise AssertionError if `dst` reformats differently the second time."""
2822 newdst = format_str(dst, line_length=line_length, is_pyi=is_pyi)
2825 diff(src, dst, "source", "first pass"),
2826 diff(dst, newdst, "first pass", "second pass"),
2828 raise AssertionError(
2829 f"INTERNAL ERROR: Black produced different code on the second pass "
2830 f"of the formatter. "
2831 f"Please report a bug on https://github.com/ambv/black/issues. "
2832 f"This diff might be helpful: {log}"
2836 def dump_to_file(*output: str) -> str:
2837 """Dump `output` to a temporary file. Return path to the file."""
2840 with tempfile.NamedTemporaryFile(
2841 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
2843 for lines in output:
2845 if lines and lines[-1] != "\n":
2850 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
2851 """Return a unified diff string between strings `a` and `b`."""
2854 a_lines = [line + "\n" for line in a.split("\n")]
2855 b_lines = [line + "\n" for line in b.split("\n")]
2857 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
2861 def cancel(tasks: Iterable[asyncio.Task]) -> None:
2862 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
2868 def shutdown(loop: BaseEventLoop) -> None:
2869 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
2871 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
2872 to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()]
2876 for task in to_cancel:
2878 loop.run_until_complete(
2879 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
2882 # `concurrent.futures.Future` objects cannot be cancelled once they
2883 # are already running. There might be some when the `shutdown()` happened.
2884 # Silence their logger's spew about the event loop being closed.
2885 cf_logger = logging.getLogger("concurrent.futures")
2886 cf_logger.setLevel(logging.CRITICAL)
2890 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
2891 """Replace `regex` with `replacement` twice on `original`.
2893 This is used by string normalization to perform replaces on
2894 overlapping matches.
2896 return regex.sub(replacement, regex.sub(replacement, original))
2899 def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]:
2900 """Like `reversed(enumerate(sequence))` if that were possible."""
2901 index = len(sequence) - 1
2902 for element in reversed(sequence):
2903 yield (index, element)
2907 def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
2908 """Return True if `line` is no longer than `line_length`.
2910 Uses the provided `line_str` rendering, if any, otherwise computes a new one.
2913 line_str = str(line).strip("\n")
2915 len(line_str) <= line_length
2916 and "\n" not in line_str # multiline strings
2917 and not line.contains_standalone_comments()
2921 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
2924 def get_cache_file(line_length: int) -> Path:
2925 return CACHE_DIR / f"cache.{line_length}.pickle"
2928 def read_cache(line_length: int) -> Cache:
2929 """Read the cache if it exists and is well formed.
2931 If it is not well formed, the call to write_cache later should resolve the issue.
2933 cache_file = get_cache_file(line_length)
2934 if not cache_file.exists():
2937 with cache_file.open("rb") as fobj:
2939 cache: Cache = pickle.load(fobj)
2940 except pickle.UnpicklingError:
2946 def get_cache_info(path: Path) -> CacheInfo:
2947 """Return the information used to check if a file is already formatted or not."""
2949 return stat.st_mtime, stat.st_size
2953 cache: Cache, sources: Iterable[Path]
2954 ) -> Tuple[List[Path], List[Path]]:
2955 """Split a list of paths into two.
2957 The first list contains paths of files that modified on disk or are not in the
2958 cache. The other list contains paths to non-modified files.
2963 if cache.get(src) != get_cache_info(src):
2970 def write_cache(cache: Cache, sources: List[Path], line_length: int) -> None:
2971 """Update the cache file."""
2972 cache_file = get_cache_file(line_length)
2974 if not CACHE_DIR.exists():
2975 CACHE_DIR.mkdir(parents=True)
2976 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
2977 with cache_file.open("wb") as fobj:
2978 pickle.dump(new_cache, fobj, protocol=pickle.HIGHEST_PROTOCOL)
2983 if __name__ == "__main__":