Type,
TypeVar,
Union,
+ cast,
)
from appdirs import user_cache_dir
from blib2to3.pgen2.parse import ParseError
-__version__ = "18.4a6"
+__version__ = "18.5b0"
DEFAULT_LINE_LENGTH = 88
+CACHE_DIR = Path(user_cache_dir("black", version=__version__))
+
# types
-syms = pygram.python_symbols
FileContent = str
Encoding = str
Depth = int
out = partial(click.secho, bold=True, err=True)
err = partial(click.secho, fg="red", err=True)
+pygram.initialize(CACHE_DIR)
+syms = pygram.python_symbols
+
class NothingChanged(UserWarning):
"""Raised by :func:`format_file` when reformatted code is the same as source."""
yield from self.line()
yield from self.visit(child)
- def visit_import_from(self, node: Node) -> Iterator[Line]:
- """Visit import_from and maybe put invisible parentheses.
-
- This is separate from `visit_stmt` because import statements don't
- support arbitrary atoms and thus handling of parentheses is custom.
- """
- check_lpar = False
- for index, child in enumerate(node.children):
- if check_lpar:
- if child.type == token.LPAR:
- # make parentheses invisible
- child.value = "" # type: ignore
- node.children[-1].value = "" # type: ignore
- else:
- # insert invisible parentheses
- node.insert_child(index, Leaf(token.LPAR, ""))
- node.append_child(Leaf(token.RPAR, ""))
- break
-
- check_lpar = (
- child.type == token.NAME and child.value == "import" # type: ignore
- )
-
- for child in node.children:
- yield from self.visit(child)
-
def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]:
"""Remove a semicolon and put the other statement on a separate line."""
yield from self.line()
self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
self.visit_return_stmt = partial(v, keywords={"return"}, parens={"return"})
+ self.visit_import_from = partial(v, keywords=Ø, parens={"import"})
self.visit_async_funcdef = self.visit_async_stmt
self.visit_decorated = self.visit_decorators
leaf.type == token.DOT
and leaf.parent
and leaf.parent.type not in {syms.import_from, syms.dotted_name}
- and (previous is None or previous.type != token.NAME)
+ and (previous is None or previous.type in CLOSING_BRACKETS)
):
return DOT_PRIORITY
return 0
-def generate_comments(leaf: Leaf) -> Iterator[Leaf]:
+def generate_comments(leaf: LN) -> Iterator[Leaf]:
"""Clean the prefix of the `leaf` and generate comments from it, if any.
Comments in lib2to3 are shoved into the whitespace prefix. This happens
def rhs(line: Line, py36: bool = False) -> Iterator[Line]:
for omit in generate_trailers_to_omit(line, line_length):
- lines = list(right_hand_split(line, py36, omit=omit))
+ lines = list(right_hand_split(line, line_length, py36, omit=omit))
if is_line_short_enough(lines[0], line_length=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.
yield from right_hand_split(line, py36)
if line.inside_brackets:
def right_hand_split(
- line: Line, py36: bool = False, omit: Collection[LeafID] = ()
+ line: Line, line_length: int, py36: bool = False, omit: Collection[LeafID] = ()
) -> Iterator[Line]:
"""Split line into many lines, starting with the last matching bracket pair.
and not line.is_import
):
omit = {id(closing_bracket), *omit}
- delimiter_count = body.bracket_tracker.delimiter_count_with_priority()
- if (
- delimiter_count == 0
- or delimiter_count == 1
- and (
- body.leaves[0].type in OPENING_BRACKETS
- or body.leaves[-1].type in CLOSING_BRACKETS
- )
- ):
+ if can_omit_invisible_parens(body, line_length):
try:
- yield from right_hand_split(line, py36=py36, omit=omit)
+ yield from right_hand_split(line, line_length, py36=py36, omit=omit)
return
except CannotSplit:
pass
Standardizes on visible parentheses for single-element tuples, and keeps
existing visible parentheses for other tuples and generator expressions.
"""
+ try:
+ list(generate_comments(node))
+ except FormatOff:
+ return # This `node` has a prefix with `# fmt: off`, don't mess with parens.
+
check_lpar = False
- for child in list(node.children):
+ for index, child in enumerate(list(node.children)):
if check_lpar:
if child.type == syms.atom:
maybe_make_parens_invisible_in_atom(child)
# wrap child in visible parentheses
lpar = Leaf(token.LPAR, "(")
rpar = Leaf(token.RPAR, ")")
- index = child.remove() or 0
+ child.remove()
node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+ elif node.type == syms.import_from:
+ # "import from" nodes store parentheses directly as part of
+ # the statement
+ if child.type == token.LPAR:
+ # make parentheses invisible
+ child.value = "" # type: ignore
+ node.children[-1].value = "" # type: ignore
+ elif child.type != token.STAR:
+ # insert invisible parentheses
+ node.insert_child(index, Leaf(token.LPAR, ""))
+ node.append_child(Leaf(token.RPAR, ""))
+ break
+
elif not (isinstance(child, Leaf) and is_multiline_string(child)):
# wrap child in invisible parentheses
lpar = Leaf(token.LPAR, "")
def should_explode(line: Line, opening_bracket: Leaf) -> bool:
"""Should `line` immediately be split with `delimiter_split()` after RHS?"""
- return bool(
+ if not (
opening_bracket.parent
and opening_bracket.parent.type in {syms.atom, syms.import_from}
and opening_bracket.value in "[{("
- and line.bracket_tracker.delimiters
- and line.bracket_tracker.max_delimiter_priority() == COMMA_PRIORITY
- )
+ ):
+ return False
+
+ try:
+ last_leaf = line.leaves[-1]
+ exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set()
+ max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude)
+ except (IndexError, ValueError):
+ return False
+
+ return max_priority == COMMA_PRIORITY
def is_python36(node: Node) -> bool:
closing_bracket = None
optional_brackets: Set[LeafID] = set()
inner_brackets: Set[LeafID] = set()
- for index, leaf in enumerate_reversed(line.leaves):
- length += len(leaf.prefix) + len(leaf.value)
+ for index, leaf, leaf_length in enumerate_with_length(line, reversed=True):
+ length += leaf_length
if length > line_length:
break
- comment: Optional[Leaf]
- for comment in line.comments_after(leaf, index):
- if "\n" in comment.prefix:
- break # Oops, standalone comment!
-
- length += len(comment.value)
- else:
- comment = None
- if comment is not None:
- break # There was a standalone comment, we can't continue.
+ has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix)
+ if leaf.type == STANDALONE_COMMENT or has_inline_comment:
+ break
optional_brackets.discard(id(leaf))
if opening_bracket:
index -= 1
+def enumerate_with_length(
+ line: Line, reversed: bool = False
+) -> Iterator[Tuple[Index, Leaf, int]]:
+ """Return an enumeration of leaves with their length.
+
+ Stops prematurely on multiline strings and standalone comments.
+ """
+ op = cast(
+ Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]],
+ enumerate_reversed if reversed else enumerate,
+ )
+ for index, leaf in op(line.leaves):
+ length = len(leaf.prefix) + len(leaf.value)
+ if "\n" in leaf.value:
+ return # Multiline strings, we can't continue.
+
+ comment: Optional[Leaf]
+ for comment in line.comments_after(leaf, index):
+ length += len(comment.value)
+
+ yield index, leaf, length
+
+
def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
"""Return True if `line` is no longer than `line_length`.
)
-CACHE_DIR = Path(user_cache_dir("black", version=__version__))
+def can_omit_invisible_parens(line: Line, line_length: int) -> 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
+ the point is to not return false positives that end up producing lines that
+ are too long.
+ """
+ bt = line.bracket_tracker
+ if not bt.delimiters:
+ # Without delimiters the optional parentheses are useless.
+ return True
+
+ max_priority = bt.max_delimiter_priority()
+ if bt.delimiter_count_with_priority(max_priority) > 1:
+ # With more than one delimiter of a kind the optional parentheses read better.
+ return False
+
+ if max_priority == DOT_PRIORITY:
+ # A single stranded method call doesn't require optional parentheses.
+ return True
+
+ 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.
+ 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
+
+ # 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.
+
+ if (
+ last.type == token.RPAR
+ or last.type == token.RBRACE
+ or (
+ # don't use indexing for omitting optional parentheses;
+ # it looks weird
+ last.type == token.RSQB
+ and last.parent
+ and last.parent.type != syms.trailer
+ )
+ ):
+ if penultimate.type in OPENING_BRACKETS:
+ # Empty brackets don't help.
+ return False
+
+ if is_multiline_string(first):
+ # Additional wrapping of a multiline string in this situation is
+ # unnecessary.
+ return True
+
+ 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 get_cache_file(line_length: int) -> Path: