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.
5 from asyncio.base_events import BaseEventLoop
6 from concurrent.futures import Executor, ProcessPoolExecutor
8 from functools import partial, wraps
11 from multiprocessing import Manager
13 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
46 __version__ = "18.4a2"
47 DEFAULT_LINE_LENGTH = 88
49 syms = pygram.python_symbols
57 LN = Union[Leaf, Node]
58 SplitFunc = Callable[["Line", bool], Iterator["Line"]]
61 CacheInfo = Tuple[Timestamp, FileSize]
62 Cache = Dict[Path, CacheInfo]
63 out = partial(click.secho, bold=True, err=True)
64 err = partial(click.secho, fg="red", err=True)
67 class NothingChanged(UserWarning):
68 """Raised by :func:`format_file` when reformatted code is the same as source."""
71 class CannotSplit(Exception):
72 """A readable split that fits the allotted line length is impossible.
74 Raised by :func:`left_hand_split`, :func:`right_hand_split`, and
75 :func:`delimiter_split`.
79 class FormatError(Exception):
80 """Base exception for `# fmt: on` and `# fmt: off` handling.
82 It holds the number of bytes of the prefix consumed before the format
83 control comment appeared.
86 def __init__(self, consumed: int) -> None:
87 super().__init__(consumed)
88 self.consumed = consumed
90 def trim_prefix(self, leaf: Leaf) -> None:
91 leaf.prefix = leaf.prefix[self.consumed:]
93 def leaf_from_consumed(self, leaf: Leaf) -> Leaf:
94 """Returns a new Leaf from the consumed part of the prefix."""
95 unformatted_prefix = leaf.prefix[:self.consumed]
96 return Leaf(token.NEWLINE, unformatted_prefix)
99 class FormatOn(FormatError):
100 """Found a comment like `# fmt: on` in the file."""
103 class FormatOff(FormatError):
104 """Found a comment like `# fmt: off` in the file."""
107 class WriteBack(Enum):
124 default=DEFAULT_LINE_LENGTH,
125 help="How many character per line to allow.",
132 "Don't write the files back, just return the status. Return code 0 "
133 "means nothing would change. Return code 1 means some files would be "
134 "reformatted. Return code 123 means there was an internal error."
140 help="Don't write the files back, just output a diff for each file on stdout.",
145 help="If --fast given, skip temporary sanity checks. [default: --safe]",
152 "Don't emit non-error messages to stderr. Errors are still emitted, "
153 "silence those with 2>/dev/null."
156 @click.version_option(version=__version__)
161 exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True
174 """The uncompromising code formatter."""
175 sources: List[Path] = []
179 sources.extend(gen_python_files_in_dir(p))
181 # if a file was explicitly given, we don't care about its extension
184 sources.append(Path("-"))
186 err(f"invalid path: {s}")
188 if check and not diff:
189 write_back = WriteBack.NO
191 write_back = WriteBack.DIFF
193 write_back = WriteBack.YES
194 report = Report(check=check, quiet=quiet)
195 if len(sources) == 0:
199 elif len(sources) == 1:
200 reformat_one(sources[0], line_length, fast, write_back, report)
202 loop = asyncio.get_event_loop()
203 executor = ProcessPoolExecutor(max_workers=os.cpu_count())
205 loop.run_until_complete(
207 sources, line_length, fast, write_back, report, loop, executor
213 out("All done! ✨ 🍰 ✨")
214 click.echo(str(report))
215 ctx.exit(report.return_code)
219 src: Path, line_length: int, fast: bool, write_back: WriteBack, report: "Report"
221 """Reformat a single file under `src` without spawning child processes.
223 If `quiet` is True, non-error messages are not output. `line_length`,
224 `write_back`, and `fast` options are passed to :func:`format_file_in_place`.
228 if not src.is_file() and str(src) == "-":
229 if format_stdin_to_stdout(
230 line_length=line_length, fast=fast, write_back=write_back
232 changed = Changed.YES
235 if write_back != WriteBack.DIFF:
238 if src in cache and cache[src] == get_cache_info(src):
239 changed = Changed.CACHED
241 changed is not Changed.CACHED
242 and format_file_in_place(
243 src, line_length=line_length, fast=fast, write_back=write_back
246 changed = Changed.YES
247 if write_back != WriteBack.DIFF and changed is not Changed.NO:
248 write_cache(cache, [src])
249 report.done(src, changed)
250 except Exception as exc:
251 report.failed(src, str(exc))
254 async def schedule_formatting(
258 write_back: WriteBack,
263 """Run formatting of `sources` in parallel using the provided `executor`.
265 (Use ProcessPoolExecutors for actual parallelism.)
267 `line_length`, `write_back`, and `fast` options are passed to
268 :func:`format_file_in_place`.
271 if write_back != WriteBack.DIFF:
273 sources, cached = filter_cached(cache, sources)
275 report.done(src, Changed.CACHED)
280 if write_back == WriteBack.DIFF:
281 # For diff output, we need locks to ensure we don't interleave output
282 # from different processes.
284 lock = manager.Lock()
286 src: loop.run_in_executor(
287 executor, format_file_in_place, src, line_length, fast, write_back, lock
291 _task_values = list(tasks.values())
293 loop.add_signal_handler(signal.SIGINT, cancel, _task_values)
294 loop.add_signal_handler(signal.SIGTERM, cancel, _task_values)
295 except NotImplementedError:
296 # There are no good alternatives for these on Windows
298 await asyncio.wait(_task_values)
299 for src, task in tasks.items():
301 report.failed(src, "timed out, cancelling")
303 cancelled.append(task)
304 elif task.cancelled():
305 cancelled.append(task)
306 elif task.exception():
307 report.failed(src, str(task.exception()))
309 formatted.append(src)
310 report.done(src, Changed.YES if task.result() else Changed.NO)
313 await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
314 if write_back != WriteBack.DIFF and formatted:
315 write_cache(cache, formatted)
318 def format_file_in_place(
322 write_back: WriteBack = WriteBack.NO,
323 lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
325 """Format file under `src` path. Return True if changed.
327 If `write_back` is True, write reformatted code back to stdout.
328 `line_length` and `fast` options are passed to :func:`format_file_contents`.
331 with tokenize.open(src) as src_buffer:
332 src_contents = src_buffer.read()
334 dst_contents = format_file_contents(
335 src_contents, line_length=line_length, fast=fast
337 except NothingChanged:
340 if write_back == write_back.YES:
341 with open(src, "w", encoding=src_buffer.encoding) as f:
342 f.write(dst_contents)
343 elif write_back == write_back.DIFF:
344 src_name = f"{src} (original)"
345 dst_name = f"{src} (formatted)"
346 diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
350 sys.stdout.write(diff_contents)
357 def format_stdin_to_stdout(
358 line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO
360 """Format file on stdin. Return True if changed.
362 If `write_back` is True, write reformatted code back to stdout.
363 `line_length` and `fast` arguments are passed to :func:`format_file_contents`.
365 src = sys.stdin.read()
368 dst = format_file_contents(src, line_length=line_length, fast=fast)
371 except NothingChanged:
375 if write_back == WriteBack.YES:
376 sys.stdout.write(dst)
377 elif write_back == WriteBack.DIFF:
378 src_name = "<stdin> (original)"
379 dst_name = "<stdin> (formatted)"
380 sys.stdout.write(diff(src, dst, src_name, dst_name))
383 def format_file_contents(
384 src_contents: str, line_length: int, fast: bool
386 """Reformat contents a file and return new contents.
388 If `fast` is False, additionally confirm that the reformatted code is
389 valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
390 `line_length` is passed to :func:`format_str`.
392 if src_contents.strip() == "":
395 dst_contents = format_str(src_contents, line_length=line_length)
396 if src_contents == dst_contents:
400 assert_equivalent(src_contents, dst_contents)
401 assert_stable(src_contents, dst_contents, line_length=line_length)
405 def format_str(src_contents: str, line_length: int) -> FileContent:
406 """Reformat a string and return new contents.
408 `line_length` determines how many characters per line are allowed.
410 src_node = lib2to3_parse(src_contents)
412 lines = LineGenerator()
413 elt = EmptyLineTracker()
414 py36 = is_python36(src_node)
417 for current_line in lines.visit(src_node):
418 for _ in range(after):
419 dst_contents += str(empty_line)
420 before, after = elt.maybe_empty_lines(current_line)
421 for _ in range(before):
422 dst_contents += str(empty_line)
423 for line in split_line(current_line, line_length=line_length, py36=py36):
424 dst_contents += str(line)
429 pygram.python_grammar_no_print_statement_no_exec_statement,
430 pygram.python_grammar_no_print_statement,
431 pygram.python_grammar,
435 def lib2to3_parse(src_txt: str) -> Node:
436 """Given a string with source, return the lib2to3 Node."""
437 grammar = pygram.python_grammar_no_print_statement
438 if src_txt[-1] != "\n":
439 nl = "\r\n" if "\r\n" in src_txt[:1024] else "\n"
441 for grammar in GRAMMARS:
442 drv = driver.Driver(grammar, pytree.convert)
444 result = drv.parse_string(src_txt, True)
447 except ParseError as pe:
448 lineno, column = pe.context[1]
449 lines = src_txt.splitlines()
451 faulty_line = lines[lineno - 1]
453 faulty_line = "<line number missing in source>"
454 exc = ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}")
458 if isinstance(result, Leaf):
459 result = Node(syms.file_input, [result])
463 def lib2to3_unparse(node: Node) -> str:
464 """Given a lib2to3 node, return its string representation."""
472 class Visitor(Generic[T]):
473 """Basic lib2to3 visitor that yields things of type `T` on `visit()`."""
475 def visit(self, node: LN) -> Iterator[T]:
476 """Main method to visit `node` and its children.
478 It tries to find a `visit_*()` method for the given `node.type`, like
479 `visit_simple_stmt` for Node objects or `visit_INDENT` for Leaf objects.
480 If no dedicated `visit_*()` method is found, chooses `visit_default()`
483 Then yields objects of type `T` from the selected visitor.
486 name = token.tok_name[node.type]
488 name = type_repr(node.type)
489 yield from getattr(self, f"visit_{name}", self.visit_default)(node)
491 def visit_default(self, node: LN) -> Iterator[T]:
492 """Default `visit_*()` implementation. Recurses to children of `node`."""
493 if isinstance(node, Node):
494 for child in node.children:
495 yield from self.visit(child)
499 class DebugVisitor(Visitor[T]):
502 def visit_default(self, node: LN) -> Iterator[T]:
503 indent = " " * (2 * self.tree_depth)
504 if isinstance(node, Node):
505 _type = type_repr(node.type)
506 out(f"{indent}{_type}", fg="yellow")
508 for child in node.children:
509 yield from self.visit(child)
512 out(f"{indent}/{_type}", fg="yellow", bold=False)
514 _type = token.tok_name.get(node.type, str(node.type))
515 out(f"{indent}{_type}", fg="blue", nl=False)
517 # We don't have to handle prefixes for `Node` objects since
518 # that delegates to the first child anyway.
519 out(f" {node.prefix!r}", fg="green", bold=False, nl=False)
520 out(f" {node.value!r}", fg="blue", bold=False)
523 def show(cls, code: str) -> None:
524 """Pretty-print the lib2to3 AST of a given string of `code`.
526 Convenience method for debugging.
528 v: DebugVisitor[None] = DebugVisitor()
529 list(v.visit(lib2to3_parse(code)))
532 KEYWORDS = set(keyword.kwlist)
533 WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE}
534 FLOW_CONTROL = {"return", "raise", "break", "continue"}
545 STANDALONE_COMMENT = 153
546 LOGIC_OPERATORS = {"and", "or"}
570 STARS = {token.STAR, token.DOUBLESTAR}
573 syms.argument, # double star in arglist
574 syms.trailer, # single argument to call
576 syms.varargslist, # lambdas
578 UNPACKING_PARENTS = {
579 syms.atom, # single element of a list or set literal
584 COMPREHENSION_PRIORITY = 20
588 COMPARATOR_PRIORITY = 3
593 class BracketTracker:
594 """Keeps track of brackets on a line."""
597 bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
598 delimiters: Dict[LeafID, Priority] = Factory(dict)
599 previous: Optional[Leaf] = None
600 _for_loop_variable: bool = False
601 _lambda_arguments: bool = False
603 def mark(self, leaf: Leaf) -> None:
604 """Mark `leaf` with bracket-related metadata. Keep track of delimiters.
606 All leaves receive an int `bracket_depth` field that stores how deep
607 within brackets a given leaf is. 0 means there are no enclosing brackets
608 that started on this line.
610 If a leaf is itself a closing bracket, it receives an `opening_bracket`
611 field that it forms a pair with. This is a one-directional link to
612 avoid reference cycles.
614 If a leaf is a delimiter (a token on which Black can split the line if
615 needed) and it's on depth 0, its `id()` is stored in the tracker's
618 if leaf.type == token.COMMENT:
621 self.maybe_decrement_after_for_loop_variable(leaf)
622 self.maybe_decrement_after_lambda_arguments(leaf)
623 if leaf.type in CLOSING_BRACKETS:
625 opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
626 leaf.opening_bracket = opening_bracket
627 leaf.bracket_depth = self.depth
629 delim = is_split_before_delimiter(leaf, self.previous)
630 if delim and self.previous is not None:
631 self.delimiters[id(self.previous)] = delim
633 delim = is_split_after_delimiter(leaf, self.previous)
635 self.delimiters[id(leaf)] = delim
636 if leaf.type in OPENING_BRACKETS:
637 self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
640 self.maybe_increment_lambda_arguments(leaf)
641 self.maybe_increment_for_loop_variable(leaf)
643 def any_open_brackets(self) -> bool:
644 """Return True if there is an yet unmatched open bracket on the line."""
645 return bool(self.bracket_match)
647 def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int:
648 """Return the highest priority of a delimiter found on the line.
650 Values are consistent with what `is_split_*_delimiter()` return.
651 Raises ValueError on no delimiters.
653 return max(v for k, v in self.delimiters.items() if k not in exclude)
655 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
656 """In a for loop, or comprehension, the variables are often unpacks.
658 To avoid splitting on the comma in this situation, increase the depth of
659 tokens between `for` and `in`.
661 if leaf.type == token.NAME and leaf.value == "for":
663 self._for_loop_variable = True
668 def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
669 """See `maybe_increment_for_loop_variable` above for explanation."""
670 if self._for_loop_variable and leaf.type == token.NAME and leaf.value == "in":
672 self._for_loop_variable = False
677 def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool:
678 """In a lambda expression, there might be more than one argument.
680 To avoid splitting on the comma in this situation, increase the depth of
681 tokens between `lambda` and `:`.
683 if leaf.type == token.NAME and leaf.value == "lambda":
685 self._lambda_arguments = True
690 def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool:
691 """See `maybe_increment_lambda_arguments` above for explanation."""
692 if self._lambda_arguments and leaf.type == token.COLON:
694 self._lambda_arguments = False
702 """Holds leaves and comments. Can be printed with `str(line)`."""
705 leaves: List[Leaf] = Factory(list)
706 comments: List[Tuple[Index, Leaf]] = Factory(list)
707 bracket_tracker: BracketTracker = Factory(BracketTracker)
708 inside_brackets: bool = False
710 def append(self, leaf: Leaf, preformatted: bool = False) -> None:
711 """Add a new `leaf` to the end of the line.
713 Unless `preformatted` is True, the `leaf` will receive a new consistent
714 whitespace prefix and metadata applied by :class:`BracketTracker`.
715 Trailing commas are maybe removed, unpacked for loop variables are
716 demoted from being delimiters.
718 Inline comments are put aside.
720 has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
724 if self.leaves and not preformatted:
725 # Note: at this point leaf.prefix should be empty except for
726 # imports, for which we only preserve newlines.
727 leaf.prefix += whitespace(leaf)
728 if self.inside_brackets or not preformatted:
729 self.bracket_tracker.mark(leaf)
730 self.maybe_remove_trailing_comma(leaf)
732 if not self.append_comment(leaf):
733 self.leaves.append(leaf)
735 def append_safe(self, leaf: Leaf, preformatted: bool = False) -> None:
736 """Like :func:`append()` but disallow invalid standalone comment structure.
738 Raises ValueError when any `leaf` is appended after a standalone comment
739 or when a standalone comment is not the first leaf on the line.
741 if self.bracket_tracker.depth == 0:
743 raise ValueError("cannot append to standalone comments")
745 if self.leaves and leaf.type == STANDALONE_COMMENT:
747 "cannot append standalone comments to a populated line"
750 self.append(leaf, preformatted=preformatted)
753 def is_comment(self) -> bool:
754 """Is this line a standalone comment?"""
755 return len(self.leaves) == 1 and self.leaves[0].type == STANDALONE_COMMENT
758 def is_decorator(self) -> bool:
759 """Is this line a decorator?"""
760 return bool(self) and self.leaves[0].type == token.AT
763 def is_import(self) -> bool:
764 """Is this an import line?"""
765 return bool(self) and is_import(self.leaves[0])
768 def is_class(self) -> bool:
769 """Is this line a class definition?"""
772 and self.leaves[0].type == token.NAME
773 and self.leaves[0].value == "class"
777 def is_def(self) -> bool:
778 """Is this a function definition? (Also returns True for async defs.)"""
780 first_leaf = self.leaves[0]
785 second_leaf: Optional[Leaf] = self.leaves[1]
789 (first_leaf.type == token.NAME and first_leaf.value == "def")
791 first_leaf.type == token.ASYNC
792 and second_leaf is not None
793 and second_leaf.type == token.NAME
794 and second_leaf.value == "def"
799 def is_flow_control(self) -> bool:
800 """Is this line a flow control statement?
802 Those are `return`, `raise`, `break`, and `continue`.
806 and self.leaves[0].type == token.NAME
807 and self.leaves[0].value in FLOW_CONTROL
811 def is_yield(self) -> bool:
812 """Is this line a yield statement?"""
815 and self.leaves[0].type == token.NAME
816 and self.leaves[0].value == "yield"
819 def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
820 """If so, needs to be split before emitting."""
821 for leaf in self.leaves:
822 if leaf.type == STANDALONE_COMMENT:
823 if leaf.bracket_depth <= depth_limit:
828 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
829 """Remove trailing comma if there is one and it's safe."""
832 and self.leaves[-1].type == token.COMMA
833 and closing.type in CLOSING_BRACKETS
837 if closing.type == token.RBRACE:
838 self.remove_trailing_comma()
841 if closing.type == token.RSQB:
842 comma = self.leaves[-1]
843 if comma.parent and comma.parent.type == syms.listmaker:
844 self.remove_trailing_comma()
847 # For parens let's check if it's safe to remove the comma. If the
848 # trailing one is the only one, we might mistakenly change a tuple
849 # into a different type by removing the comma.
850 depth = closing.bracket_depth + 1
852 opening = closing.opening_bracket
853 for _opening_index, leaf in enumerate(self.leaves):
860 for leaf in self.leaves[_opening_index + 1:]:
864 bracket_depth = leaf.bracket_depth
865 if bracket_depth == depth and leaf.type == token.COMMA:
867 if leaf.parent and leaf.parent.type == syms.arglist:
872 self.remove_trailing_comma()
877 def append_comment(self, comment: Leaf) -> bool:
878 """Add an inline or standalone comment to the line."""
880 comment.type == STANDALONE_COMMENT
881 and self.bracket_tracker.any_open_brackets()
886 if comment.type != token.COMMENT:
889 after = len(self.leaves) - 1
891 comment.type = STANDALONE_COMMENT
896 self.comments.append((after, comment))
899 def comments_after(self, leaf: Leaf) -> Iterator[Leaf]:
900 """Generate comments that should appear directly after `leaf`."""
901 for _leaf_index, _leaf in enumerate(self.leaves):
908 for index, comment_after in self.comments:
909 if _leaf_index == index:
912 def remove_trailing_comma(self) -> None:
913 """Remove the trailing comma and moves the comments attached to it."""
914 comma_index = len(self.leaves) - 1
915 for i in range(len(self.comments)):
916 comment_index, comment = self.comments[i]
917 if comment_index == comma_index:
918 self.comments[i] = (comma_index - 1, comment)
921 def __str__(self) -> str:
922 """Render the line."""
926 indent = " " * self.depth
927 leaves = iter(self.leaves)
929 res = f"{first.prefix}{indent}{first.value}"
932 for _, comment in self.comments:
936 def __bool__(self) -> bool:
937 """Return True if the line has leaves or comments."""
938 return bool(self.leaves or self.comments)
941 class UnformattedLines(Line):
942 """Just like :class:`Line` but stores lines which aren't reformatted."""
944 def append(self, leaf: Leaf, preformatted: bool = True) -> None:
945 """Just add a new `leaf` to the end of the lines.
947 The `preformatted` argument is ignored.
949 Keeps track of indentation `depth`, which is useful when the user
950 says `# fmt: on`. Otherwise, doesn't do anything with the `leaf`.
953 list(generate_comments(leaf))
954 except FormatOn as f_on:
955 self.leaves.append(f_on.leaf_from_consumed(leaf))
958 self.leaves.append(leaf)
959 if leaf.type == token.INDENT:
961 elif leaf.type == token.DEDENT:
964 def __str__(self) -> str:
965 """Render unformatted lines from leaves which were added with `append()`.
967 `depth` is not used for indentation in this case.
973 for leaf in self.leaves:
977 def append_comment(self, comment: Leaf) -> bool:
978 """Not implemented in this class. Raises `NotImplementedError`."""
979 raise NotImplementedError("Unformatted lines don't store comments separately.")
981 def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
982 """Does nothing and returns False."""
985 def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
986 """Does nothing and returns False."""
991 class EmptyLineTracker:
992 """Provides a stateful method that returns the number of potential extra
993 empty lines needed before and after the currently processed line.
995 Note: this tracker works on lines that haven't been split yet. It assumes
996 the prefix of the first leaf consists of optional newlines. Those newlines
997 are consumed by `maybe_empty_lines()` and included in the computation.
999 previous_line: Optional[Line] = None
1000 previous_after: int = 0
1001 previous_defs: List[int] = Factory(list)
1003 def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1004 """Return the number of extra empty lines before and after the `current_line`.
1006 This is for separating `def`, `async def` and `class` with extra empty
1007 lines (two on module-level), as well as providing an extra empty line
1008 after flow control keywords to make them more prominent.
1010 if isinstance(current_line, UnformattedLines):
1013 before, after = self._maybe_empty_lines(current_line)
1014 before -= self.previous_after
1015 self.previous_after = after
1016 self.previous_line = current_line
1017 return before, after
1019 def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
1021 if current_line.depth == 0:
1023 if current_line.leaves:
1024 # Consume the first leaf's extra newlines.
1025 first_leaf = current_line.leaves[0]
1026 before = first_leaf.prefix.count("\n")
1027 before = min(before, max_allowed)
1028 first_leaf.prefix = ""
1031 depth = current_line.depth
1032 while self.previous_defs and self.previous_defs[-1] >= depth:
1033 self.previous_defs.pop()
1034 before = 1 if depth else 2
1035 is_decorator = current_line.is_decorator
1036 if is_decorator or current_line.is_def or current_line.is_class:
1037 if not is_decorator:
1038 self.previous_defs.append(depth)
1039 if self.previous_line is None:
1040 # Don't insert empty lines before the first line in the file.
1043 if self.previous_line and self.previous_line.is_decorator:
1044 # Don't insert empty lines between decorators.
1048 if current_line.depth:
1052 if current_line.is_flow_control:
1057 and self.previous_line.is_import
1058 and not current_line.is_import
1059 and depth == self.previous_line.depth
1061 return (before or 1), 0
1065 and self.previous_line.is_yield
1066 and (not current_line.is_yield or depth != self.previous_line.depth)
1068 return (before or 1), 0
1074 class LineGenerator(Visitor[Line]):
1075 """Generates reformatted Line objects. Empty lines are not emitted.
1077 Note: destroys the tree it's visiting by mutating prefixes of its leaves
1078 in ways that will no longer stringify to valid Python code on the tree.
1080 current_line: Line = Factory(Line)
1082 def line(self, indent: int = 0, type: Type[Line] = Line) -> Iterator[Line]:
1085 If the line is empty, only emit if it makes sense.
1086 If the line is too long, split it first and then generate.
1088 If any lines were generated, set up a new current_line.
1090 if not self.current_line:
1091 if self.current_line.__class__ == type:
1092 self.current_line.depth += indent
1094 self.current_line = type(depth=self.current_line.depth + indent)
1095 return # Line is empty, don't emit. Creating a new one unnecessary.
1097 complete_line = self.current_line
1098 self.current_line = type(depth=complete_line.depth + indent)
1101 def visit(self, node: LN) -> Iterator[Line]:
1102 """Main method to visit `node` and its children.
1104 Yields :class:`Line` objects.
1106 if isinstance(self.current_line, UnformattedLines):
1107 # File contained `# fmt: off`
1108 yield from self.visit_unformatted(node)
1111 yield from super().visit(node)
1113 def visit_default(self, node: LN) -> Iterator[Line]:
1114 """Default `visit_*()` implementation. Recurses to children of `node`."""
1115 if isinstance(node, Leaf):
1116 any_open_brackets = self.current_line.bracket_tracker.any_open_brackets()
1118 for comment in generate_comments(node):
1119 if any_open_brackets:
1120 # any comment within brackets is subject to splitting
1121 self.current_line.append(comment)
1122 elif comment.type == token.COMMENT:
1123 # regular trailing comment
1124 self.current_line.append(comment)
1125 yield from self.line()
1128 # regular standalone comment
1129 yield from self.line()
1131 self.current_line.append(comment)
1132 yield from self.line()
1134 except FormatOff as f_off:
1135 f_off.trim_prefix(node)
1136 yield from self.line(type=UnformattedLines)
1137 yield from self.visit(node)
1139 except FormatOn as f_on:
1140 # This only happens here if somebody says "fmt: on" multiple
1142 f_on.trim_prefix(node)
1143 yield from self.visit_default(node)
1146 normalize_prefix(node, inside_brackets=any_open_brackets)
1147 if node.type == token.STRING:
1148 normalize_string_quotes(node)
1149 if node.type not in WHITESPACE:
1150 self.current_line.append(node)
1151 yield from super().visit_default(node)
1153 def visit_INDENT(self, node: Node) -> Iterator[Line]:
1154 """Increase indentation level, maybe yield a line."""
1155 # In blib2to3 INDENT never holds comments.
1156 yield from self.line(+1)
1157 yield from self.visit_default(node)
1159 def visit_DEDENT(self, node: Node) -> Iterator[Line]:
1160 """Decrease indentation level, maybe yield a line."""
1161 # DEDENT has no value. Additionally, in blib2to3 it never holds comments.
1162 yield from self.line(-1)
1165 self, node: Node, keywords: Set[str], parens: Set[str]
1166 ) -> Iterator[Line]:
1167 """Visit a statement.
1169 This implementation is shared for `if`, `while`, `for`, `try`, `except`,
1170 `def`, `with`, `class`, and `assert`.
1172 The relevant Python language `keywords` for a given statement will be
1173 NAME leaves within it. This methods puts those on a separate line.
1175 `parens` holds pairs of nodes where invisible parentheses should be put.
1176 Keys hold nodes after which opening parentheses should be put, values
1177 hold nodes before which closing parentheses should be put.
1179 normalize_invisible_parens(node, parens_after=parens)
1180 for child in node.children:
1181 if child.type == token.NAME and child.value in keywords: # type: ignore
1182 yield from self.line()
1184 yield from self.visit(child)
1186 def visit_simple_stmt(self, node: Node) -> Iterator[Line]:
1187 """Visit a statement without nested statements."""
1188 is_suite_like = node.parent and node.parent.type in STATEMENT
1190 yield from self.line(+1)
1191 yield from self.visit_default(node)
1192 yield from self.line(-1)
1195 yield from self.line()
1196 yield from self.visit_default(node)
1198 def visit_async_stmt(self, node: Node) -> Iterator[Line]:
1199 """Visit `async def`, `async for`, `async with`."""
1200 yield from self.line()
1202 children = iter(node.children)
1203 for child in children:
1204 yield from self.visit(child)
1206 if child.type == token.ASYNC:
1209 internal_stmt = next(children)
1210 for child in internal_stmt.children:
1211 yield from self.visit(child)
1213 def visit_decorators(self, node: Node) -> Iterator[Line]:
1214 """Visit decorators."""
1215 for child in node.children:
1216 yield from self.line()
1217 yield from self.visit(child)
1219 def visit_import_from(self, node: Node) -> Iterator[Line]:
1220 """Visit import_from and maybe put invisible parentheses.
1222 This is separate from `visit_stmt` because import statements don't
1223 support arbitrary atoms and thus handling of parentheses is custom.
1226 for index, child in enumerate(node.children):
1228 if child.type == token.LPAR:
1229 # make parentheses invisible
1230 child.value = "" # type: ignore
1231 node.children[-1].value = "" # type: ignore
1233 # insert invisible parentheses
1234 node.insert_child(index, Leaf(token.LPAR, ""))
1235 node.append_child(Leaf(token.RPAR, ""))
1239 child.type == token.NAME and child.value == "import" # type: ignore
1242 for child in node.children:
1243 yield from self.visit(child)
1245 def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
1246 """Remove a semicolon and put the other statement on a separate line."""
1247 yield from self.line()
1249 def visit_ENDMARKER(self, leaf: Leaf) -> Iterator[Line]:
1250 """End of file. Process outstanding comments and end with a newline."""
1251 yield from self.visit_default(leaf)
1252 yield from self.line()
1254 def visit_unformatted(self, node: LN) -> Iterator[Line]:
1255 """Used when file contained a `# fmt: off`."""
1256 if isinstance(node, Node):
1257 for child in node.children:
1258 yield from self.visit(child)
1262 self.current_line.append(node)
1263 except FormatOn as f_on:
1264 f_on.trim_prefix(node)
1265 yield from self.line()
1266 yield from self.visit(node)
1268 if node.type == token.ENDMARKER:
1269 # somebody decided not to put a final `# fmt: on`
1270 yield from self.line()
1272 def __attrs_post_init__(self) -> None:
1273 """You are in a twisty little maze of passages."""
1276 self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
1277 self.visit_if_stmt = partial(v, keywords={"if", "else", "elif"}, parens={"if"})
1278 self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
1279 self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
1280 self.visit_try_stmt = partial(
1281 v, keywords={"try", "except", "else", "finally"}, parens=Ø
1283 self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
1284 self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
1285 self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
1286 self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
1287 self.visit_async_funcdef = self.visit_async_stmt
1288 self.visit_decorated = self.visit_decorators
1291 IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
1292 BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
1293 OPENING_BRACKETS = set(BRACKET.keys())
1294 CLOSING_BRACKETS = set(BRACKET.values())
1295 BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
1296 ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT}
1299 def whitespace(leaf: Leaf) -> str: # noqa C901
1300 """Return whitespace prefix if needed for the given `leaf`."""
1307 if t in ALWAYS_NO_SPACE:
1310 if t == token.COMMENT:
1313 assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
1314 if t == token.COLON and p.type not in {syms.subscript, syms.subscriptlist}:
1317 prev = leaf.prev_sibling
1319 prevp = preceding_leaf(p)
1320 if not prevp or prevp.type in OPENING_BRACKETS:
1323 if t == token.COLON:
1324 return SPACE if prevp.type == token.COMMA else NO
1326 if prevp.type == token.EQUAL:
1328 if prevp.parent.type in {
1329 syms.arglist, syms.argument, syms.parameters, syms.varargslist
1333 elif prevp.parent.type == syms.typedargslist:
1334 # A bit hacky: if the equal sign has whitespace, it means we
1335 # previously found it's a typed argument. So, we're using
1339 elif prevp.type in STARS:
1340 if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1343 elif prevp.type == token.COLON:
1344 if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}:
1349 and prevp.parent.type == syms.factor
1350 and prevp.type in MATH_OPERATORS
1355 prevp.type == token.RIGHTSHIFT
1357 and prevp.parent.type == syms.shift_expr
1358 and prevp.prev_sibling
1359 and prevp.prev_sibling.type == token.NAME
1360 and prevp.prev_sibling.value == "print" # type: ignore
1362 # Python 2 print chevron
1365 elif prev.type in OPENING_BRACKETS:
1368 if p.type in {syms.parameters, syms.arglist}:
1369 # untyped function signatures or calls
1370 if not prev or prev.type != token.COMMA:
1373 elif p.type == syms.varargslist:
1375 if prev and prev.type != token.COMMA:
1378 elif p.type == syms.typedargslist:
1379 # typed function signatures
1383 if t == token.EQUAL:
1384 if prev.type != syms.tname:
1387 elif prev.type == token.EQUAL:
1388 # A bit hacky: if the equal sign has whitespace, it means we
1389 # previously found it's a typed argument. So, we're using that, too.
1392 elif prev.type != token.COMMA:
1395 elif p.type == syms.tname:
1398 prevp = preceding_leaf(p)
1399 if not prevp or prevp.type != token.COMMA:
1402 elif p.type == syms.trailer:
1403 # attributes and calls
1404 if t == token.LPAR or t == token.RPAR:
1409 prevp = preceding_leaf(p)
1410 if not prevp or prevp.type != token.NUMBER:
1413 elif t == token.LSQB:
1416 elif prev.type != token.COMMA:
1419 elif p.type == syms.argument:
1421 if t == token.EQUAL:
1425 prevp = preceding_leaf(p)
1426 if not prevp or prevp.type == token.LPAR:
1429 elif prev.type in {token.EQUAL} | STARS:
1432 elif p.type == syms.decorator:
1436 elif p.type == syms.dotted_name:
1440 prevp = preceding_leaf(p)
1441 if not prevp or prevp.type == token.AT or prevp.type == token.DOT:
1444 elif p.type == syms.classdef:
1448 if prev and prev.type == token.LPAR:
1451 elif p.type == syms.subscript:
1454 assert p.parent is not None, "subscripts are always parented"
1455 if p.parent.type == syms.subscriptlist:
1463 elif p.type == syms.atom:
1464 if prev and t == token.DOT:
1465 # dots, but not the first one.
1468 elif p.type == syms.dictsetmaker:
1470 if prev and prev.type == token.DOUBLESTAR:
1473 elif p.type in {syms.factor, syms.star_expr}:
1476 prevp = preceding_leaf(p)
1477 if not prevp or prevp.type in OPENING_BRACKETS:
1480 prevp_parent = prevp.parent
1481 assert prevp_parent is not None
1483 prevp.type == token.COLON
1484 and prevp_parent.type in {syms.subscript, syms.sliceop}
1488 elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
1491 elif t == token.NAME or t == token.NUMBER:
1494 elif p.type == syms.import_from:
1496 if prev and prev.type == token.DOT:
1499 elif t == token.NAME:
1503 if prev and prev.type == token.DOT:
1506 elif p.type == syms.sliceop:
1512 def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]:
1513 """Return the first leaf that precedes `node`, if any."""
1515 res = node.prev_sibling
1517 if isinstance(res, Leaf):
1521 return list(res.leaves())[-1]
1530 def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1531 """Return the priority of the `leaf` delimiter, given a line break after it.
1533 The delimiter priorities returned here are from those delimiters that would
1534 cause a line break after themselves.
1536 Higher numbers are higher priority.
1538 if leaf.type == token.COMMA:
1539 return COMMA_PRIORITY
1544 def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int:
1545 """Return the priority of the `leaf` delimiter, given a line before after it.
1547 The delimiter priorities returned here are from those delimiters that would
1548 cause a line break before themselves.
1550 Higher numbers are higher priority.
1552 if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS):
1553 # * and ** might also be MATH_OPERATORS but in this case they are not.
1554 # Don't treat them as a delimiter.
1558 leaf.type in MATH_OPERATORS
1560 and leaf.parent.type not in {syms.factor, syms.star_expr}
1562 return MATH_PRIORITY
1564 if leaf.type in COMPARATORS:
1565 return COMPARATOR_PRIORITY
1568 leaf.type == token.STRING
1569 and previous is not None
1570 and previous.type == token.STRING
1572 return STRING_PRIORITY
1575 leaf.type == token.NAME
1576 and leaf.value == "for"
1578 and leaf.parent.type in {syms.comp_for, syms.old_comp_for}
1580 return COMPREHENSION_PRIORITY
1583 leaf.type == token.NAME
1584 and leaf.value == "if"
1586 and leaf.parent.type in {syms.comp_if, syms.old_comp_if}
1588 return COMPREHENSION_PRIORITY
1590 if leaf.type == token.NAME and leaf.value in LOGIC_OPERATORS and leaf.parent:
1591 return LOGIC_PRIORITY
1596 def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
1597 """Clean the prefix of the `leaf` and generate comments from it, if any.
1599 Comments in lib2to3 are shoved into the whitespace prefix. This happens
1600 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
1601 move because it does away with modifying the grammar to include all the
1602 possible places in which comments can be placed.
1604 The sad consequence for us though is that comments don't "belong" anywhere.
1605 This is why this function generates simple parentless Leaf objects for
1606 comments. We simply don't know what the correct parent should be.
1608 No matter though, we can live without this. We really only need to
1609 differentiate between inline and standalone comments. The latter don't
1610 share the line with any code.
1612 Inline comments are emitted as regular token.COMMENT leaves. Standalone
1613 are emitted with a fake STANDALONE_COMMENT token identifier.
1624 for index, line in enumerate(p.split("\n")):
1625 consumed += len(line) + 1 # adding the length of the split '\n'
1626 line = line.lstrip()
1629 if not line.startswith("#"):
1632 if index == 0 and leaf.type != token.ENDMARKER:
1633 comment_type = token.COMMENT # simple trailing comment
1635 comment_type = STANDALONE_COMMENT
1636 comment = make_comment(line)
1637 yield Leaf(comment_type, comment, prefix="\n" * nlines)
1639 if comment in {"# fmt: on", "# yapf: enable"}:
1640 raise FormatOn(consumed)
1642 if comment in {"# fmt: off", "# yapf: disable"}:
1643 if comment_type == STANDALONE_COMMENT:
1644 raise FormatOff(consumed)
1646 prev = preceding_leaf(leaf)
1647 if not prev or prev.type in WHITESPACE: # standalone comment in disguise
1648 raise FormatOff(consumed)
1653 def make_comment(content: str) -> str:
1654 """Return a consistently formatted comment from the given `content` string.
1656 All comments (except for "##", "#!", "#:") should have a single space between
1657 the hash sign and the content.
1659 If `content` didn't start with a hash sign, one is provided.
1661 content = content.rstrip()
1665 if content[0] == "#":
1666 content = content[1:]
1667 if content and content[0] not in " !:#":
1668 content = " " + content
1669 return "#" + content
1673 line: Line, line_length: int, inner: bool = False, py36: bool = False
1674 ) -> Iterator[Line]:
1675 """Split a `line` into potentially many lines.
1677 They should fit in the allotted `line_length` but might not be able to.
1678 `inner` signifies that there were a pair of brackets somewhere around the
1679 current `line`, possibly transitively. This means we can fallback to splitting
1680 by delimiters if the LHS/RHS don't yield any results.
1682 If `py36` is True, splitting may generate syntax that is only compatible
1683 with Python 3.6 and later.
1685 if isinstance(line, UnformattedLines) or line.is_comment:
1689 line_str = str(line).strip("\n")
1691 len(line_str) <= line_length
1692 and "\n" not in line_str # multiline strings
1693 and not line.contains_standalone_comments()
1698 split_funcs: List[SplitFunc]
1700 split_funcs = [left_hand_split]
1701 elif line.inside_brackets:
1702 split_funcs = [delimiter_split, standalone_comment_split, right_hand_split]
1704 split_funcs = [right_hand_split]
1705 for split_func in split_funcs:
1706 # We are accumulating lines in `result` because we might want to abort
1707 # mission and return the original line in the end, or attempt a different
1709 result: List[Line] = []
1711 for l in split_func(line, py36):
1712 if str(l).strip("\n") == line_str:
1713 raise CannotSplit("Split function returned an unchanged result")
1716 split_line(l, line_length=line_length, inner=True, py36=py36)
1718 except CannotSplit as cs:
1729 def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
1730 """Split line into many lines, starting with the first matching bracket pair.
1732 Note: this usually looks weird, only use this for function definitions.
1733 Prefer RHS otherwise.
1735 head = Line(depth=line.depth)
1736 body = Line(depth=line.depth + 1, inside_brackets=True)
1737 tail = Line(depth=line.depth)
1738 tail_leaves: List[Leaf] = []
1739 body_leaves: List[Leaf] = []
1740 head_leaves: List[Leaf] = []
1741 current_leaves = head_leaves
1742 matching_bracket = None
1743 for leaf in line.leaves:
1745 current_leaves is body_leaves
1746 and leaf.type in CLOSING_BRACKETS
1747 and leaf.opening_bracket is matching_bracket
1749 current_leaves = tail_leaves if body_leaves else head_leaves
1750 current_leaves.append(leaf)
1751 if current_leaves is head_leaves:
1752 if leaf.type in OPENING_BRACKETS:
1753 matching_bracket = leaf
1754 current_leaves = body_leaves
1755 # Since body is a new indent level, remove spurious leading whitespace.
1757 normalize_prefix(body_leaves[0], inside_brackets=True)
1758 # Build the new lines.
1759 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
1761 result.append(leaf, preformatted=True)
1762 for comment_after in line.comments_after(leaf):
1763 result.append(comment_after, preformatted=True)
1764 bracket_split_succeeded_or_raise(head, body, tail)
1765 for result in (head, body, tail):
1770 def right_hand_split(
1771 line: Line, py36: bool = False, omit: Collection[LeafID] = ()
1772 ) -> Iterator[Line]:
1773 """Split line into many lines, starting with the last matching bracket pair."""
1774 head = Line(depth=line.depth)
1775 body = Line(depth=line.depth + 1, inside_brackets=True)
1776 tail = Line(depth=line.depth)
1777 tail_leaves: List[Leaf] = []
1778 body_leaves: List[Leaf] = []
1779 head_leaves: List[Leaf] = []
1780 current_leaves = tail_leaves
1781 opening_bracket = None
1782 closing_bracket = None
1783 for leaf in reversed(line.leaves):
1784 if current_leaves is body_leaves:
1785 if leaf is opening_bracket:
1786 current_leaves = head_leaves if body_leaves else tail_leaves
1787 current_leaves.append(leaf)
1788 if current_leaves is tail_leaves:
1789 if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
1790 opening_bracket = leaf.opening_bracket
1791 closing_bracket = leaf
1792 current_leaves = body_leaves
1793 tail_leaves.reverse()
1794 body_leaves.reverse()
1795 head_leaves.reverse()
1796 # Since body is a new indent level, remove spurious leading whitespace.
1798 normalize_prefix(body_leaves[0], inside_brackets=True)
1799 elif not head_leaves:
1800 # No `head` and no `body` means the split failed. `tail` has all content.
1801 raise CannotSplit("No brackets found")
1803 # Build the new lines.
1804 for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
1806 result.append(leaf, preformatted=True)
1807 for comment_after in line.comments_after(leaf):
1808 result.append(comment_after, preformatted=True)
1809 bracket_split_succeeded_or_raise(head, body, tail)
1810 assert opening_bracket and closing_bracket
1812 opening_bracket.type == token.LPAR
1813 and not opening_bracket.value
1814 and closing_bracket.type == token.RPAR
1815 and not closing_bracket.value
1817 # These parens were optional. If there aren't any delimiters or standalone
1818 # comments in the body, they were unnecessary and another split without
1819 # them should be attempted.
1821 body.bracket_tracker.delimiters or line.contains_standalone_comments(0)
1823 omit = {id(closing_bracket), *omit}
1824 yield from right_hand_split(line, py36=py36, omit=omit)
1827 ensure_visible(opening_bracket)
1828 ensure_visible(closing_bracket)
1829 for result in (head, body, tail):
1834 def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
1835 """Raise :exc:`CannotSplit` if the last left- or right-hand split failed.
1837 Do nothing otherwise.
1839 A left- or right-hand split is based on a pair of brackets. Content before
1840 (and including) the opening bracket is left on one line, content inside the
1841 brackets is put on a separate line, and finally content starting with and
1842 following the closing bracket is put on a separate line.
1844 Those are called `head`, `body`, and `tail`, respectively. If the split
1845 produced the same line (all content in `head`) or ended up with an empty `body`
1846 and the `tail` is just the closing bracket, then it's considered failed.
1848 tail_len = len(str(tail).strip())
1851 raise CannotSplit("Splitting brackets produced the same line")
1855 f"Splitting brackets on an empty body to save "
1856 f"{tail_len} characters is not worth it"
1860 def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc:
1861 """Normalize prefix of the first leaf in every line returned by `split_func`.
1863 This is a decorator over relevant split functions.
1867 def split_wrapper(line: Line, py36: bool = False) -> Iterator[Line]:
1868 for l in split_func(line, py36):
1869 normalize_prefix(l.leaves[0], inside_brackets=True)
1872 return split_wrapper
1875 @dont_increase_indentation
1876 def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]:
1877 """Split according to delimiters of the highest priority.
1879 If `py36` is True, the split will add trailing commas also in function
1880 signatures that contain `*` and `**`.
1883 last_leaf = line.leaves[-1]
1885 raise CannotSplit("Line empty")
1887 delimiters = line.bracket_tracker.delimiters
1889 delimiter_priority = line.bracket_tracker.max_delimiter_priority(
1890 exclude={id(last_leaf)}
1893 raise CannotSplit("No delimiters found")
1895 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
1896 lowest_depth = sys.maxsize
1897 trailing_comma_safe = True
1899 def append_to_line(leaf: Leaf) -> Iterator[Line]:
1900 """Append `leaf` to current line or to new line if appending impossible."""
1901 nonlocal current_line
1903 current_line.append_safe(leaf, preformatted=True)
1904 except ValueError as ve:
1907 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
1908 current_line.append(leaf)
1910 for leaf in line.leaves:
1911 yield from append_to_line(leaf)
1913 for comment_after in line.comments_after(leaf):
1914 yield from append_to_line(comment_after)
1916 lowest_depth = min(lowest_depth, leaf.bracket_depth)
1918 leaf.bracket_depth == lowest_depth
1919 and is_vararg(leaf, within=VARARGS_PARENTS)
1921 trailing_comma_safe = trailing_comma_safe and py36
1922 leaf_priority = delimiters.get(id(leaf))
1923 if leaf_priority == delimiter_priority:
1926 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
1930 and delimiter_priority == COMMA_PRIORITY
1931 and current_line.leaves[-1].type != token.COMMA
1932 and current_line.leaves[-1].type != STANDALONE_COMMENT
1934 current_line.append(Leaf(token.COMMA, ","))
1938 @dont_increase_indentation
1939 def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]:
1940 """Split standalone comments from the rest of the line."""
1941 if not line.contains_standalone_comments(0):
1942 raise CannotSplit("Line does not have any standalone comments")
1944 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
1946 def append_to_line(leaf: Leaf) -> Iterator[Line]:
1947 """Append `leaf` to current line or to new line if appending impossible."""
1948 nonlocal current_line
1950 current_line.append_safe(leaf, preformatted=True)
1951 except ValueError as ve:
1954 current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
1955 current_line.append(leaf)
1957 for leaf in line.leaves:
1958 yield from append_to_line(leaf)
1960 for comment_after in line.comments_after(leaf):
1961 yield from append_to_line(comment_after)
1967 def is_import(leaf: Leaf) -> bool:
1968 """Return True if the given leaf starts an import statement."""
1975 (v == "import" and p and p.type == syms.import_name)
1976 or (v == "from" and p and p.type == syms.import_from)
1981 def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
1982 """Leave existing extra newlines if not `inside_brackets`. Remove everything
1985 Note: don't use backslashes for formatting or you'll lose your voting rights.
1987 if not inside_brackets:
1988 spl = leaf.prefix.split("#")
1989 if "\\" not in spl[0]:
1990 nl_count = spl[-1].count("\n")
1993 leaf.prefix = "\n" * nl_count
1999 def normalize_string_quotes(leaf: Leaf) -> None:
2000 """Prefer double quotes but only if it doesn't cause more escaping.
2002 Adds or removes backslashes as appropriate. Doesn't parse and fix
2003 strings nested in f-strings (yet).
2005 Note: Mutates its argument.
2007 value = leaf.value.lstrip("furbFURB")
2008 if value[:3] == '"""':
2011 elif value[:3] == "'''":
2014 elif value[0] == '"':
2020 first_quote_pos = leaf.value.find(orig_quote)
2021 if first_quote_pos == -1:
2022 return # There's an internal error
2024 prefix = leaf.value[:first_quote_pos]
2025 unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
2026 escaped_new_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{new_quote}")
2027 escaped_orig_quote = re.compile(rf"([^\\]|^)\\(\\\\)*{orig_quote}")
2028 body = leaf.value[first_quote_pos + len(orig_quote):-len(orig_quote)]
2029 if "r" in prefix.casefold():
2030 if unescaped_new_quote.search(body):
2031 # There's at least one unescaped new_quote in this raw string
2032 # so converting is impossible
2035 # Do not introduce or remove backslashes in raw strings
2038 # remove unnecessary quotes
2039 new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body)
2040 if body != new_body:
2041 # Consider the string without unnecessary quotes as the original
2043 leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}"
2044 new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body)
2045 new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body)
2046 if new_quote == '"""' and new_body[-1] == '"':
2048 new_body = new_body[:-1] + '\\"'
2049 orig_escape_count = body.count("\\")
2050 new_escape_count = new_body.count("\\")
2051 if new_escape_count > orig_escape_count:
2052 return # Do not introduce more escaping
2054 if new_escape_count == orig_escape_count and orig_quote == '"':
2055 return # Prefer double quotes
2057 leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
2060 def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
2061 """Make existing optional parentheses invisible or create new ones.
2063 Standardizes on visible parentheses for single-element tuples, and keeps
2064 existing visible parentheses for other tuples and generator expressions.
2067 for child in list(node.children):
2069 if child.type == syms.atom:
2071 is_empty_tuple(child)
2072 or is_one_tuple(child)
2073 or max_delimiter_priority_in_atom(child) >= COMMA_PRIORITY
2075 first = child.children[0]
2076 last = child.children[-1]
2077 if first.type == token.LPAR and last.type == token.RPAR:
2078 # make parentheses invisible
2079 first.value = "" # type: ignore
2080 last.value = "" # type: ignore
2081 elif is_one_tuple(child):
2082 # wrap child in visible parentheses
2083 lpar = Leaf(token.LPAR, "(")
2084 rpar = Leaf(token.RPAR, ")")
2085 index = child.remove() or 0
2086 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2088 # wrap child in invisible parentheses
2089 lpar = Leaf(token.LPAR, "")
2090 rpar = Leaf(token.RPAR, "")
2091 index = child.remove() or 0
2092 node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
2094 check_lpar = isinstance(child, Leaf) and child.value in parens_after
2097 def is_empty_tuple(node: LN) -> bool:
2098 """Return True if `node` holds an empty tuple."""
2100 node.type == syms.atom
2101 and len(node.children) == 2
2102 and node.children[0].type == token.LPAR
2103 and node.children[1].type == token.RPAR
2107 def is_one_tuple(node: LN) -> bool:
2108 """Return True if `node` holds a tuple with one element, with or without parens."""
2109 if node.type == syms.atom:
2110 if len(node.children) != 3:
2113 lpar, gexp, rpar = node.children
2115 lpar.type == token.LPAR
2116 and gexp.type == syms.testlist_gexp
2117 and rpar.type == token.RPAR
2121 return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
2124 node.type in IMPLICIT_TUPLE
2125 and len(node.children) == 2
2126 and node.children[1].type == token.COMMA
2130 def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool:
2131 """Return True if `leaf` is a star or double star in a vararg or kwarg.
2133 If `within` includes VARARGS_PARENTS, this applies to function signatures.
2134 If `within` includes COLLECTION_LIBERALS_PARENTS, it applies to right
2135 hand-side extended iterable unpacking (PEP 3132) and additional unpacking
2136 generalizations (PEP 448).
2138 if leaf.type not in STARS or not leaf.parent:
2142 if p.type == syms.star_expr:
2143 # Star expressions are also used as assignment targets in extended
2144 # iterable unpacking (PEP 3132). See what its parent is instead.
2150 return p.type in within
2153 def max_delimiter_priority_in_atom(node: LN) -> int:
2154 """Return maximum delimiter priority inside `node`.
2156 This is specific to atoms with contents contained in a pair of parentheses.
2157 If `node` isn't an atom or there are no enclosing parentheses, returns 0.
2159 if node.type != syms.atom:
2162 first = node.children[0]
2163 last = node.children[-1]
2164 if not (first.type == token.LPAR and last.type == token.RPAR):
2167 bt = BracketTracker()
2168 for c in node.children[1:-1]:
2169 if isinstance(c, Leaf):
2172 for leaf in c.leaves():
2175 return bt.max_delimiter_priority()
2181 def ensure_visible(leaf: Leaf) -> None:
2182 """Make sure parentheses are visible.
2184 They could be invisible as part of some statements (see
2185 :func:`normalize_invible_parens` and :func:`visit_import_from`).
2187 if leaf.type == token.LPAR:
2189 elif leaf.type == token.RPAR:
2193 def is_python36(node: Node) -> bool:
2194 """Return True if the current file is using Python 3.6+ features.
2196 Currently looking for:
2198 - trailing commas after * or ** in function signatures.
2200 for n in node.pre_order():
2201 if n.type == token.STRING:
2202 value_head = n.value[:2] # type: ignore
2203 if value_head in {'f"', 'F"', "f'", "F'", "rf", "fr", "RF", "FR"}:
2207 n.type == syms.typedargslist
2209 and n.children[-1].type == token.COMMA
2211 for ch in n.children:
2212 if ch.type in STARS:
2218 PYTHON_EXTENSIONS = {".py"}
2219 BLACKLISTED_DIRECTORIES = {
2220 "build", "buck-out", "dist", "_build", ".git", ".hg", ".mypy_cache", ".tox", ".venv"
2224 def gen_python_files_in_dir(path: Path) -> Iterator[Path]:
2225 """Generate all files under `path` which aren't under BLACKLISTED_DIRECTORIES
2226 and have one of the PYTHON_EXTENSIONS.
2228 for child in path.iterdir():
2230 if child.name in BLACKLISTED_DIRECTORIES:
2233 yield from gen_python_files_in_dir(child)
2235 elif child.suffix in PYTHON_EXTENSIONS:
2241 """Provides a reformatting counter. Can be rendered with `str(report)`."""
2244 change_count: int = 0
2246 failure_count: int = 0
2248 def done(self, src: Path, changed: Changed) -> None:
2249 """Increment the counter for successful reformatting. Write out a message."""
2250 if changed is Changed.YES:
2251 reformatted = "would reformat" if self.check else "reformatted"
2253 out(f"{reformatted} {src}")
2254 self.change_count += 1
2257 if changed is Changed.NO:
2258 msg = f"{src} already well formatted, good job."
2260 msg = f"{src} wasn't modified on disk since last run."
2261 out(msg, bold=False)
2262 self.same_count += 1
2264 def failed(self, src: Path, message: str) -> None:
2265 """Increment the counter for failed reformatting. Write out a message."""
2266 err(f"error: cannot format {src}: {message}")
2267 self.failure_count += 1
2270 def return_code(self) -> int:
2271 """Return the exit code that the app should use.
2273 This considers the current state of changed files and failures:
2274 - if there were any failures, return 123;
2275 - if any files were changed and --check is being used, return 1;
2276 - otherwise return 0.
2278 # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
2279 # 126 we have special returncodes reserved by the shell.
2280 if self.failure_count:
2283 elif self.change_count and self.check:
2288 def __str__(self) -> str:
2289 """Render a color report of the current state.
2291 Use `click.unstyle` to remove colors.
2294 reformatted = "would be reformatted"
2295 unchanged = "would be left unchanged"
2296 failed = "would fail to reformat"
2298 reformatted = "reformatted"
2299 unchanged = "left unchanged"
2300 failed = "failed to reformat"
2302 if self.change_count:
2303 s = "s" if self.change_count > 1 else ""
2305 click.style(f"{self.change_count} file{s} {reformatted}", bold=True)
2308 s = "s" if self.same_count > 1 else ""
2309 report.append(f"{self.same_count} file{s} {unchanged}")
2310 if self.failure_count:
2311 s = "s" if self.failure_count > 1 else ""
2313 click.style(f"{self.failure_count} file{s} {failed}", fg="red")
2315 return ", ".join(report) + "."
2318 def assert_equivalent(src: str, dst: str) -> None:
2319 """Raise AssertionError if `src` and `dst` aren't equivalent."""
2324 def _v(node: ast.AST, depth: int = 0) -> Iterator[str]:
2325 """Simple visitor generating strings to compare ASTs by content."""
2326 yield f"{' ' * depth}{node.__class__.__name__}("
2328 for field in sorted(node._fields):
2330 value = getattr(node, field)
2331 except AttributeError:
2334 yield f"{' ' * (depth+1)}{field}="
2336 if isinstance(value, list):
2338 if isinstance(item, ast.AST):
2339 yield from _v(item, depth + 2)
2341 elif isinstance(value, ast.AST):
2342 yield from _v(value, depth + 2)
2345 yield f"{' ' * (depth+2)}{value!r}, # {value.__class__.__name__}"
2347 yield f"{' ' * depth}) # /{node.__class__.__name__}"
2350 src_ast = ast.parse(src)
2351 except Exception as exc:
2352 major, minor = sys.version_info[:2]
2353 raise AssertionError(
2354 f"cannot use --safe with this file; failed to parse source file "
2355 f"with Python {major}.{minor}'s builtin AST. Re-run with --fast "
2356 f"or stop using deprecated Python 2 syntax. AST error message: {exc}"
2360 dst_ast = ast.parse(dst)
2361 except Exception as exc:
2362 log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst)
2363 raise AssertionError(
2364 f"INTERNAL ERROR: Black produced invalid code: {exc}. "
2365 f"Please report a bug on https://github.com/ambv/black/issues. "
2366 f"This invalid output might be helpful: {log}"
2369 src_ast_str = "\n".join(_v(src_ast))
2370 dst_ast_str = "\n".join(_v(dst_ast))
2371 if src_ast_str != dst_ast_str:
2372 log = dump_to_file(diff(src_ast_str, dst_ast_str, "src", "dst"))
2373 raise AssertionError(
2374 f"INTERNAL ERROR: Black produced code that is not equivalent to "
2376 f"Please report a bug on https://github.com/ambv/black/issues. "
2377 f"This diff might be helpful: {log}"
2381 def assert_stable(src: str, dst: str, line_length: int) -> None:
2382 """Raise AssertionError if `dst` reformats differently the second time."""
2383 newdst = format_str(dst, line_length=line_length)
2386 diff(src, dst, "source", "first pass"),
2387 diff(dst, newdst, "first pass", "second pass"),
2389 raise AssertionError(
2390 f"INTERNAL ERROR: Black produced different code on the second pass "
2391 f"of the formatter. "
2392 f"Please report a bug on https://github.com/ambv/black/issues. "
2393 f"This diff might be helpful: {log}"
2397 def dump_to_file(*output: str) -> str:
2398 """Dump `output` to a temporary file. Return path to the file."""
2401 with tempfile.NamedTemporaryFile(
2402 mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8"
2404 for lines in output:
2406 if lines and lines[-1] != "\n":
2411 def diff(a: str, b: str, a_name: str, b_name: str) -> str:
2412 """Return a unified diff string between strings `a` and `b`."""
2415 a_lines = [line + "\n" for line in a.split("\n")]
2416 b_lines = [line + "\n" for line in b.split("\n")]
2418 difflib.unified_diff(a_lines, b_lines, fromfile=a_name, tofile=b_name, n=5)
2422 def cancel(tasks: List[asyncio.Task]) -> None:
2423 """asyncio signal handler that cancels all `tasks` and reports to stderr."""
2429 def shutdown(loop: BaseEventLoop) -> None:
2430 """Cancel all pending tasks on `loop`, wait for them, and close the loop."""
2432 # This part is borrowed from asyncio/runners.py in Python 3.7b2.
2433 to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()]
2437 for task in to_cancel:
2439 loop.run_until_complete(
2440 asyncio.gather(*to_cancel, loop=loop, return_exceptions=True)
2443 # `concurrent.futures.Future` objects cannot be cancelled once they
2444 # are already running. There might be some when the `shutdown()` happened.
2445 # Silence their logger's spew about the event loop being closed.
2446 cf_logger = logging.getLogger("concurrent.futures")
2447 cf_logger.setLevel(logging.CRITICAL)
2451 def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str:
2452 """Replace `regex` with `replacement` twice on `original`.
2454 This is used by string normalization to perform replaces on
2455 overlapping matches.
2457 return regex.sub(replacement, regex.sub(replacement, original))
2460 CACHE_DIR = Path(user_cache_dir("black", version=__version__))
2461 CACHE_FILE = CACHE_DIR / "cache.pickle"
2464 def read_cache() -> Cache:
2465 """Read the cache if it exists and is well formed.
2467 If it is not well formed, the call to write_cache later should resolve the issue.
2469 if not CACHE_FILE.exists():
2472 with CACHE_FILE.open("rb") as fobj:
2474 cache: Cache = pickle.load(fobj)
2475 except pickle.UnpicklingError:
2481 def get_cache_info(path: Path) -> CacheInfo:
2482 """Return the information used to check if a file is already formatted or not."""
2484 return stat.st_mtime, stat.st_size
2488 cache: Cache, sources: Iterable[Path]
2489 ) -> Tuple[List[Path], List[Path]]:
2490 """Split a list of paths into two.
2492 The first list contains paths of files that modified on disk or are not in the
2493 cache. The other list contains paths to non-modified files.
2498 if cache.get(src) != get_cache_info(src):
2505 def write_cache(cache: Cache, sources: List[Path]) -> None:
2506 """Update the cache file."""
2508 if not CACHE_DIR.exists():
2509 CACHE_DIR.mkdir(parents=True)
2510 new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
2511 with CACHE_FILE.open("wb") as fobj:
2512 pickle.dump(new_cache, fobj, protocol=pickle.HIGHEST_PROTOCOL)
2517 if __name__ == "__main__":