X-Git-Url: https://git.madduck.net/etc/vim.git/blobdiff_plain/4938cc9e9abf2581cb154c6a8d1ae66eb18e0d65..4ca92ac91c944b252caf8c2265e5854f839a66f4:/src/black/__init__.py?ds=sidebyside diff --git a/src/black/__init__.py b/src/black/__init__.py index 2250943..96dd0e4 100644 --- a/src/black/__init__.py +++ b/src/black/__init__.py @@ -195,6 +195,7 @@ class Feature(Enum): ASYNC_KEYWORDS = 7 ASSIGNMENT_EXPRESSIONS = 8 POS_ONLY_ARGUMENTS = 9 + FORCE_OPTIONAL_PARENTHESES = 50 VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = { @@ -640,10 +641,9 @@ def path_empty( """ Exit if there is no `src` provided for formatting """ - if len(src) == 0: - if verbose or not quiet: - out(msg) - ctx.exit(0) + if not src and (verbose or not quiet): + out(msg) + ctx.exit(0) def reformat_one( @@ -927,7 +927,7 @@ def format_file_contents(src_contents: str, *, fast: bool, mode: Mode) -> FileCo valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it. `mode` is passed to :func:`format_str`. """ - if src_contents.strip() == "": + if not src_contents.strip(): raise NothingChanged dst_contents = format_str(src_contents, mode=mode) @@ -1061,7 +1061,7 @@ def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]: def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node: """Given a string with source, return the lib2to3 Node.""" - if src_txt[-1:] != "\n": + if not src_txt.endswith("\n"): src_txt += "\n" for grammar in get_grammars(set(target_versions)): @@ -1284,6 +1284,7 @@ class BracketTracker: previous: Optional[Leaf] = None _for_loop_depths: List[int] = field(default_factory=list) _lambda_argument_depths: List[int] = field(default_factory=list) + invisible: List[Leaf] = field(default_factory=list) def mark(self, leaf: Leaf) -> None: """Mark `leaf` with bracket-related metadata. Keep track of delimiters. @@ -1309,6 +1310,8 @@ class BracketTracker: self.depth -= 1 opening_bracket = self.bracket_match.pop((self.depth, leaf.type)) leaf.opening_bracket = opening_bracket + if not leaf.value: + self.invisible.append(leaf) leaf.bracket_depth = self.depth if self.depth == 0: delim = is_split_before_delimiter(leaf, self.previous) @@ -1321,6 +1324,8 @@ class BracketTracker: if leaf.type in OPENING_BRACKETS: self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf self.depth += 1 + if not leaf.value: + self.invisible.append(leaf) self.previous = leaf self.maybe_increment_lambda_arguments(leaf) self.maybe_increment_for_loop_variable(leaf) @@ -1442,7 +1447,8 @@ class Line: ) if self.inside_brackets or not preformatted: self.bracket_tracker.mark(leaf) - self.maybe_remove_trailing_comma(leaf) + if self.maybe_should_explode(leaf): + self.should_explode = True if not self.append_comment(leaf): self.leaves.append(leaf) @@ -1618,59 +1624,25 @@ class Line: def contains_multiline_strings(self) -> bool: return any(is_multiline_string(leaf) for leaf in self.leaves) - def maybe_remove_trailing_comma(self, closing: Leaf) -> bool: - """Remove trailing comma if there is one and it's safe.""" + def maybe_should_explode(self, closing: Leaf) -> bool: + """Return True if this line should explode (always be split), that is when: + - there's a trailing comma here; and + - it's not a one-tuple. + """ if not ( - self.leaves + closing.type in CLOSING_BRACKETS + and self.leaves and self.leaves[-1].type == token.COMMA - and closing.type in CLOSING_BRACKETS ): return False - if closing.type == token.RBRACE: - self.remove_trailing_comma() + if closing.type in {token.RBRACE, token.RSQB}: return True - if closing.type == token.RSQB: - comma = self.leaves[-1] - if comma.parent and comma.parent.type == syms.listmaker: - self.remove_trailing_comma() - return True - - # For parens let's check if it's safe to remove the comma. - # Imports are always safe. if self.is_import: - self.remove_trailing_comma() return True - # Otherwise, if the trailing one is the only one, we might mistakenly - # change a tuple into a different type by removing the comma. - depth = closing.bracket_depth + 1 - commas = 0 - opening = closing.opening_bracket - for _opening_index, leaf in enumerate(self.leaves): - if leaf is opening: - break - - else: - return False - - for leaf in self.leaves[_opening_index + 1 :]: - if leaf is closing: - break - - bracket_depth = leaf.bracket_depth - if bracket_depth == depth and leaf.type == token.COMMA: - commas += 1 - if leaf.parent and leaf.parent.type in { - syms.arglist, - syms.typedargslist, - }: - commas += 1 - break - - if commas > 1: - self.remove_trailing_comma() + if not is_one_tuple_between(closing.opening_bracket, closing, self.leaves): return True return False @@ -2063,13 +2035,20 @@ class LineGenerator(Visitor[Line]): yield from self.visit_default(node) def visit_STRING(self, leaf: Leaf) -> Iterator[Line]: - # Check if it's a docstring - if prev_siblings_are( - leaf.parent, [None, token.NEWLINE, token.INDENT, syms.simple_stmt] - ) and is_multiline_string(leaf): - prefix = " " * self.current_line.depth - docstring = fix_docstring(leaf.value[3:-3], prefix) - leaf.value = leaf.value[0:3] + docstring + leaf.value[-3:] + if is_docstring(leaf) and "\\\n" not in leaf.value: + # We're ignoring docstrings with backslash newline escapes because changing + # indentation of those changes the AST representation of the code. + prefix = get_string_prefix(leaf.value) + lead_len = len(prefix) + 3 + tail_len = -3 + indent = " " * 4 * self.current_line.depth + docstring = fix_docstring(leaf.value[lead_len:tail_len], indent) + if docstring: + if leaf.value[lead_len - 1] == docstring[0]: + docstring = " " + docstring + if leaf.value[tail_len + 1] == docstring[-1]: + docstring = docstring + " " + leaf.value = leaf.value[0:lead_len] + docstring + leaf.value[tail_len:] normalize_string_quotes(leaf) yield from self.visit_default(leaf) @@ -2647,7 +2626,7 @@ def transform_line( is_line_short_enough(line, line_length=mode.line_length, line_str=line_str) or line.contains_unsplittable_type_ignore() ) - and not (line.contains_standalone_comments() and line.inside_brackets) + and not (line.inside_brackets and line.contains_standalone_comments()) ): # Only apply basic string preprocessing, since lines shouldn't be split here. if mode.experimental_string_processing: @@ -2659,20 +2638,31 @@ def transform_line( else: def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]: + """Wraps calls to `right_hand_split`. + + The calls increasingly `omit` right-hand trailers (bracket pairs with + content), meaning the trailers get glued together to split on another + bracket pair instead. + """ for omit in generate_trailers_to_omit(line, mode.line_length): lines = list( right_hand_split(line, mode.line_length, features, omit=omit) ) + # Note: this check is only able to figure out if the first line of the + # *current* transformation fits in the line length. This is true only + # for simple cases. All others require running more transforms via + # `transform_line()`. This check doesn't know if those would succeed. if is_line_short_enough(lines[0], line_length=mode.line_length): yield from lines return # All splits failed, best effort split with no omits. # This mostly happens to multiline strings that are by definition - # reported as not fitting a single line. - # line_length=1 here was historically a bug that somehow became a feature. - # See #762 and #781 for the full story. - yield from right_hand_split(line, line_length=1, features=features) + # reported as not fitting a single line, as well as lines that contain + # trailing commas (those have to be exploded). + yield from right_hand_split( + line, line_length=mode.line_length, features=features + ) if mode.experimental_string_processing: if line.inside_brackets: @@ -2703,17 +2693,8 @@ def transform_line( # We are accumulating lines in `result` because we might want to abort # mission and return the original line in the end, or attempt a different # split altogether. - result: List[Line] = [] try: - for transformed_line in transform(line, features): - if str(transformed_line).strip("\n") == line_str: - raise CannotTransform( - "Line transformer returned an unchanged result" - ) - - result.extend( - transform_line(transformed_line, mode=mode, features=features) - ) + result = run_transformer(line, transform, mode, features, line_str=line_str) except CannotTransform: continue else: @@ -2754,6 +2735,7 @@ class StringTransformer(ABC): line_length: int normalize_strings: bool + __name__ = "StringTransformer" @abstractmethod def do_match(self, line: Line) -> TMatchResult: @@ -3000,7 +2982,7 @@ class StringMerger(CustomSplitMapMixin, StringTransformer): ) new_line = line.clone() - new_line.comments = line.comments + new_line.comments = line.comments.copy() append_leaves(new_line, line, LL) new_string_leaf = new_line.leaves[string_idx] @@ -3328,7 +3310,6 @@ class StringParenStripper(StringTransformer): new_line = line.clone() new_line.comments = line.comments.copy() - append_leaves(new_line, line, LL[: string_idx - 1]) string_leaf = Leaf(token.STRING, LL[string_idx].value) @@ -4010,12 +3991,13 @@ class StringParenWrapper(CustomSplitMapMixin, BaseStringSplitter): def do_splitter_match(self, line: Line) -> TMatchResult: LL = line.leaves - string_idx = None - string_idx = string_idx or self._return_match(LL) - string_idx = string_idx or self._else_match(LL) - string_idx = string_idx or self._assert_match(LL) - string_idx = string_idx or self._assign_match(LL) - string_idx = string_idx or self._dict_match(LL) + string_idx = ( + self._return_match(LL) + or self._else_match(LL) + or self._assert_match(LL) + or self._assign_match(LL) + or self._dict_match(LL) + ) if string_idx is not None: string_value = line.leaves[string_idx].value @@ -4214,7 +4196,7 @@ class StringParenWrapper(CustomSplitMapMixin, BaseStringSplitter): is_valid_index = is_valid_index_factory(LL) insert_str_child = insert_str_child_factory(LL[string_idx]) - comma_idx = len(LL) - 1 + comma_idx = -1 ends_with_comma = False if LL[comma_idx].type == token.COMMA: ends_with_comma = True @@ -4462,11 +4444,10 @@ def contains_pragma_comment(comment_list: List[Leaf]) -> bool: of the more common static analysis tools for python (e.g. mypy, flake8, pylint). """ - for comment in comment_list: - if comment.value.startswith(("# type:", "# noqa", "# pylint:")): - return True - - return False + return any( + comment.value.startswith(("# type:", "# noqa", "# pylint:")) + for comment in comment_list + ) def insert_str_child_factory(string_leaf: Leaf) -> Callable[[LN], None]: @@ -4772,8 +4753,7 @@ def right_hand_split( tail = bracket_split_build_line(tail_leaves, line, opening_bracket) bracket_split_succeeded_or_raise(head, body, tail) if ( - # the body shouldn't be exploded - not body.should_explode + Feature.FORCE_OPTIONAL_PARENTHESES not in features # the opening bracket is an optional paren and opening_bracket.type == token.LPAR and not opening_bracket.value @@ -4786,7 +4766,7 @@ def right_hand_split( # there are no standalone comments in the body and not body.contains_standalone_comments(0) # and we can actually remove the parens - and can_omit_invisible_parens(body, line_length) + and can_omit_invisible_parens(body, line_length, omit_on_explode=omit) ): omit = {id(closing_bracket), *omit} try: @@ -4872,7 +4852,8 @@ def bracket_split_build_line( continue if leaves[i].type != token.COMMA: - leaves.insert(i + 1, Leaf(token.COMMA, ",")) + new_comma = Leaf(token.COMMA, ",") + leaves.insert(i + 1, new_comma) break # Populate the line @@ -4880,8 +4861,8 @@ def bracket_split_build_line( result.append(leaf, preformatted=True) for comment_after in original.comments_after(leaf): result.append(comment_after, preformatted=True) - if is_body: - result.should_explode = should_explode(result, opening_bracket) + if is_body and should_split_body_explode(result, opening_bracket): + result.should_explode = True return result @@ -4966,7 +4947,8 @@ def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[ and current_line.leaves[-1].type != token.COMMA and current_line.leaves[-1].type != STANDALONE_COMMENT ): - current_line.append(Leaf(token.COMMA, ",")) + new_comma = Leaf(token.COMMA, ",") + current_line.append(new_comma) yield current_line @@ -5588,24 +5570,63 @@ def ensure_visible(leaf: Leaf) -> None: leaf.value = ")" -def should_explode(line: Line, opening_bracket: Leaf) -> bool: - """Should `line` immediately be split with `delimiter_split()` after RHS?""" +def should_split_body_explode(line: Line, opening_bracket: Leaf) -> bool: + """Should `line` be immediately split with `delimiter_split()` after RHS?""" - if not ( - opening_bracket.parent - and opening_bracket.parent.type in {syms.atom, syms.import_from} - and opening_bracket.value in "[{(" - ): + if not (opening_bracket.parent and opening_bracket.value in "[{("): return False + # We're essentially checking if the body is delimited by commas and there's more + # than one of them (we're excluding the trailing comma and if the delimiter priority + # is still commas, that means there's more). + exclude = set() + trailing_comma = False try: last_leaf = line.leaves[-1] - exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set() + if last_leaf.type == token.COMMA: + trailing_comma = True + exclude.add(id(last_leaf)) max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude) except (IndexError, ValueError): return False - return max_priority == COMMA_PRIORITY + return max_priority == COMMA_PRIORITY and ( + trailing_comma + # always explode imports + or opening_bracket.parent.type in {syms.atom, syms.import_from} + ) + + +def is_one_tuple_between(opening: Leaf, closing: Leaf, leaves: List[Leaf]) -> bool: + """Return True if content between `opening` and `closing` looks like a one-tuple.""" + if opening.type != token.LPAR and closing.type != token.RPAR: + return False + + depth = closing.bracket_depth + 1 + for _opening_index, leaf in enumerate(leaves): + if leaf is opening: + break + + else: + raise LookupError("Opening paren not found in `leaves`") + + commas = 0 + _opening_index += 1 + for leaf in leaves[_opening_index:]: + if leaf is closing: + break + + bracket_depth = leaf.bracket_depth + if bracket_depth == depth and leaf.type == token.COMMA: + commas += 1 + if leaf.parent and leaf.parent.type in { + syms.arglist, + syms.typedargslist, + }: + commas += 1 + break + + return commas < 2 def get_features_used(node: Node) -> Set[Feature]: @@ -5672,11 +5693,13 @@ def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[Leaf a preceding closing bracket fits in one line. Yielded sets are cumulative (contain results of previous yields, too). First - set is empty. + set is empty, unless the line should explode, in which case bracket pairs until + the one that needs to explode are omitted. """ omit: Set[LeafID] = set() - yield omit + if not line.should_explode: + yield omit length = 4 * line.depth opening_bracket: Optional[Leaf] = None @@ -5695,9 +5718,23 @@ def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[Leaf if leaf is opening_bracket: opening_bracket = None elif leaf.type in CLOSING_BRACKETS: + prev = line.leaves[index - 1] if index > 0 else None + if ( + line.should_explode + and prev + and prev.type == token.COMMA + and not is_one_tuple_between( + leaf.opening_bracket, leaf, line.leaves + ) + ): + # Never omit bracket pairs with trailing commas. + # We need to explode on those. + break + inner_brackets.add(id(leaf)) elif leaf.type in CLOSING_BRACKETS: - if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS: + prev = line.leaves[index - 1] if index > 0 else None + if prev and prev.type in OPENING_BRACKETS: # Empty brackets would fail a split so treat them as "inner" # brackets (e.g. only add them to the `omit` set if another # pair of brackets was good enough. @@ -5710,6 +5747,16 @@ def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[Leaf inner_brackets.clear() yield omit + if ( + line.should_explode + and prev + and prev.type == token.COMMA + and not is_one_tuple_between(leaf.opening_bracket, leaf, line.leaves) + ): + # Never omit bracket pairs with trailing commas. + # We need to explode on those. + break + if leaf.value: opening_bracket = leaf.opening_bracket closing_bracket = leaf @@ -6115,6 +6162,7 @@ def assert_stable(src: str, dst: str, mode: Mode) -> None: newdst = format_str(dst, mode=mode) if dst != newdst: log = dump_to_file( + str(mode), diff(src, dst, "source", "first pass"), diff(dst, newdst, "first pass", "second pass"), ) @@ -6291,7 +6339,11 @@ def can_be_split(line: Line) -> bool: return True -def can_omit_invisible_parens(line: Line, line_length: int) -> bool: +def can_omit_invisible_parens( + line: Line, + line_length: int, + omit_on_explode: Collection[LeafID] = (), +) -> bool: """Does `line` have a shape safe to reformat without optional parens around it? Returns True for only a subset of potentially nice looking formattings but @@ -6314,37 +6366,27 @@ def can_omit_invisible_parens(line: Line, line_length: int) -> bool: assert len(line.leaves) >= 2, "Stranded delimiter" - first = line.leaves[0] - second = line.leaves[1] - penultimate = line.leaves[-2] - last = line.leaves[-1] - # With a single delimiter, omit if the expression starts or ends with # a bracket. + first = line.leaves[0] + second = line.leaves[1] if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS: - remainder = False - length = 4 * line.depth - for _index, leaf, leaf_length in enumerate_with_length(line): - if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first: - remainder = True - if remainder: - length += leaf_length - if length > line_length: - break - - if leaf.type in OPENING_BRACKETS: - # There are brackets we can further split on. - remainder = False - - else: - # checked the entire string and line length wasn't exceeded - if len(line.leaves) == _index + 1: - return True + if _can_omit_opening_paren(line, first=first, line_length=line_length): + return True # Note: we are not returning False here because a line might have *both* # a leading opening bracket and a trailing closing bracket. If the # opening bracket doesn't match our rule, maybe the closing will. + penultimate = line.leaves[-2] + last = line.leaves[-1] + if line.should_explode: + try: + penultimate, last = last_two_except(line.leaves, omit=omit_on_explode) + except LookupError: + # Turns out we'd omit everything. We cannot skip the optional parentheses. + return False + if ( last.type == token.RPAR or last.type == token.RBRACE @@ -6365,21 +6407,120 @@ def can_omit_invisible_parens(line: Line, line_length: int) -> bool: # unnecessary. return True - length = 4 * line.depth - seen_other_brackets = False - for _index, leaf, leaf_length in enumerate_with_length(line): + if line.should_explode and penultimate.type == token.COMMA: + # The rightmost non-omitted bracket pair is the one we want to explode on. + return True + + if _can_omit_closing_paren(line, last=last, line_length=line_length): + return True + + return False + + +def _can_omit_opening_paren(line: Line, *, first: Leaf, line_length: int) -> bool: + """See `can_omit_invisible_parens`.""" + remainder = False + length = 4 * line.depth + _index = -1 + for _index, leaf, leaf_length in enumerate_with_length(line): + if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first: + remainder = True + if remainder: length += leaf_length - if leaf is last.opening_bracket: - if seen_other_brackets or length <= line_length: - return True + if length > line_length: + break - elif leaf.type in OPENING_BRACKETS: + if leaf.type in OPENING_BRACKETS: # There are brackets we can further split on. - seen_other_brackets = True + remainder = False + + else: + # checked the entire string and line length wasn't exceeded + if len(line.leaves) == _index + 1: + return True + + return False + + +def _can_omit_closing_paren(line: Line, *, last: Leaf, line_length: int) -> bool: + """See `can_omit_invisible_parens`.""" + length = 4 * line.depth + seen_other_brackets = False + for _index, leaf, leaf_length in enumerate_with_length(line): + length += leaf_length + if leaf is last.opening_bracket: + if seen_other_brackets or length <= line_length: + return True + + elif leaf.type in OPENING_BRACKETS: + # There are brackets we can further split on. + seen_other_brackets = True return False +def last_two_except(leaves: List[Leaf], omit: Collection[LeafID]) -> Tuple[Leaf, Leaf]: + """Return (penultimate, last) leaves skipping brackets in `omit` and contents.""" + stop_after = None + last = None + for leaf in reversed(leaves): + if stop_after: + if leaf is stop_after: + stop_after = None + continue + + if last: + return leaf, last + + if id(leaf) in omit: + stop_after = leaf.opening_bracket + else: + last = leaf + else: + raise LookupError("Last two leaves were also skipped") + + +def run_transformer( + line: Line, + transform: Transformer, + mode: Mode, + features: Collection[Feature], + *, + line_str: str = "", +) -> List[Line]: + if not line_str: + line_str = line_to_string(line) + result: List[Line] = [] + for transformed_line in transform(line, features): + if str(transformed_line).strip("\n") == line_str: + raise CannotTransform("Line transformer returned an unchanged result") + + result.extend(transform_line(transformed_line, mode=mode, features=features)) + + if not ( + transform.__name__ == "rhs" + and line.bracket_tracker.invisible + and not any(bracket.value for bracket in line.bracket_tracker.invisible) + and not line.contains_multiline_strings() + and not result[0].contains_uncollapsable_type_comments() + and not result[0].contains_unsplittable_type_ignore() + and not is_line_short_enough(result[0], line_length=mode.line_length) + ): + return result + + line_copy = line.clone() + append_leaves(line_copy, line, line.leaves) + features_fop = set(features) | {Feature.FORCE_OPTIONAL_PARENTHESES} + second_opinion = run_transformer( + line_copy, transform, mode, features_fop, line_str=line_str + ) + if all( + is_line_short_enough(ln, line_length=mode.line_length) for ln in second_opinion + ): + result = second_opinion + return result + + def get_cache_file(mode: Mode) -> Path: return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle" @@ -6465,6 +6606,26 @@ def patched_main() -> None: main() +def is_docstring(leaf: Leaf) -> bool: + if not is_multiline_string(leaf): + # For the purposes of docstring re-indentation, we don't need to do anything + # with single-line docstrings. + return False + + if prev_siblings_are( + leaf.parent, [None, token.NEWLINE, token.INDENT, syms.simple_stmt] + ): + return True + + # Multiline docstring on the same line as the `def`. + if prev_siblings_are(leaf.parent, [syms.parameters, token.COLON, syms.simple_stmt]): + # `syms.parameters` is only used in funcdefs and async_funcdefs in the Python + # grammar. We're safe to return True without further checks. + return True + + return False + + def fix_docstring(docstring: str, prefix: str) -> str: # https://www.python.org/dev/peps/pep-0257/#handling-docstring-indentation if not docstring: @@ -6488,7 +6649,6 @@ def fix_docstring(docstring: str, prefix: str) -> str: trimmed.append(prefix + stripped_line) else: trimmed.append("") - # Return a single string: return "\n".join(trimmed)