import click
import toml
from typed_ast import ast3, ast27
+from pathspec import PathSpec
# lib2to3 fork
from blib2to3.pytree import Node, Leaf, type_repr
from blib2to3.pgen2.grammar import Grammar
from blib2to3.pgen2.parse import ParseError
-from _version import version as __version__
+from _black_version import version as __version__
DEFAULT_LINE_LENGTH = 88
-DEFAULT_EXCLUDES = (
- r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|_build|buck-out|build|dist)/"
-)
+DEFAULT_EXCLUDES = r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|\.svn|_build|buck-out|build|dist)/" # noqa: B950
DEFAULT_INCLUDES = r"\.pyi?$"
CACHE_DIR = Path(user_cache_dir("black", version=__version__))
p = Path(s)
if p.is_dir():
sources.update(
- gen_python_files_in_dir(p, root, include_regex, exclude_regex, report)
+ gen_python_files_in_dir(
+ p, root, include_regex, exclude_regex, report, get_gitignore(root)
+ )
)
elif p.is_file() or s == "-":
# if a file was explicitly given, we don't care about its extension
# Python 2.7
pygram.python_grammar,
]
- elif all(version.is_python2() for version in target_versions):
+
+ if all(version.is_python2() for version in target_versions):
# Python 2-only code, so try Python 2 grammars.
return [
# Python 2.7 with future print_function import
# Python 2.7
pygram.python_grammar,
]
- else:
- # Python 3-compatible code, so only try Python 3 grammar.
- grammars = []
- # If we have to parse both, try to parse async as a keyword first
- if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
- # Python 3.7+
- grammars.append(
- pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords # noqa: B950
- )
- if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
- # Python 3.0-3.6
- grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
- # At least one of the above branches must have been taken, because every Python
- # version has exactly one of the two 'ASYNC_*' flags
- return grammars
+
+ # Python 3-compatible code, so only try Python 3 grammar.
+ grammars = []
+ # If we have to parse both, try to parse async as a keyword first
+ if not supports_feature(target_versions, Feature.ASYNC_IDENTIFIERS):
+ # Python 3.7+
+ grammars.append(
+ pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords
+ )
+ if not supports_feature(target_versions, Feature.ASYNC_KEYWORDS):
+ # Python 3.0-3.6
+ grammars.append(pygram.python_grammar_no_print_statement_no_exec_statement)
+ # At least one of the above branches must have been taken, because every Python
+ # version has exactly one of the two 'ASYNC_*' flags
+ return grammars
def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node:
Leaf(token.DOT, ".") for _ in range(3)
]
+ @property
+ def is_collection_with_optional_trailing_comma(self) -> bool:
+ """Is this line a collection literal with a trailing comma that's optional?
+
+ Note that the trailing comma in a 1-tuple is not optional.
+ """
+ if not self.leaves or len(self.leaves) < 4:
+ return False
+
+ # Look for and address a trailing colon.
+ if self.leaves[-1].type == token.COLON:
+ closer = self.leaves[-2]
+ close_index = -2
+ else:
+ closer = self.leaves[-1]
+ close_index = -1
+ if closer.type not in CLOSING_BRACKETS or self.inside_brackets:
+ return False
+
+ if closer.type == token.RPAR:
+ # Tuples require an extra check, because if there's only
+ # one element in the tuple removing the comma unmakes the
+ # tuple.
+ #
+ # We also check for parens before looking for the trailing
+ # comma because in some cases (eg assigning a dict
+ # literal) the literal gets wrapped in temporary parens
+ # during parsing. This case is covered by the
+ # collections.py test data.
+ opener = closer.opening_bracket
+ for _open_index, leaf in enumerate(self.leaves):
+ if leaf is opener:
+ break
+
+ else:
+ # Couldn't find the matching opening paren, play it safe.
+ return False
+
+ commas = 0
+ comma_depth = self.leaves[close_index - 1].bracket_depth
+ for leaf in self.leaves[_open_index + 1 : close_index]:
+ if leaf.bracket_depth == comma_depth and leaf.type == token.COMMA:
+ commas += 1
+ if commas > 1:
+ # We haven't looked yet for the trailing comma because
+ # we might also have caught noop parens.
+ return self.leaves[close_index - 1].type == token.COMMA
+
+ elif commas == 1:
+ return False # it's either a one-tuple or didn't have a trailing comma
+
+ if self.leaves[close_index - 1].type in CLOSING_BRACKETS:
+ close_index -= 1
+ closer = self.leaves[close_index]
+ if closer.type == token.RPAR:
+ # TODO: this is a gut feeling. Will we ever see this?
+ return False
+
+ if self.leaves[close_index - 1].type != token.COMMA:
+ return False
+
+ return True
+
@property
def is_def(self) -> bool:
"""Is this a function definition? (Also returns True for async defs.)"""
def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
"""If so, needs to be split before emitting."""
for leaf in self.leaves:
- if leaf.type == STANDALONE_COMMENT:
- if leaf.bracket_depth <= depth_limit:
- return True
+ if leaf.type == STANDALONE_COMMENT and leaf.bracket_depth <= depth_limit:
+ return True
+
return False
def contains_uncollapsable_type_comments(self) -> bool:
# only report an unsplittable 'type: ignore' if this line was
# one line in the original code.
- # Like in the type comment check above, we need to skip a black added
- # trailing comma or invisible paren, since it will be the original leaf
- # before it that has the original line number.
- last_idx = -1
- last_leaf = self.leaves[-1]
- if len(self.leaves) > 2 and (
- last_leaf.type == token.COMMA
- or (last_leaf.type == token.RPAR and not last_leaf.value)
- ):
- last_idx = -2
+ # Grab the first and last line numbers, skipping generated leaves
+ first_line = next((l.lineno for l in self.leaves if l.lineno != 0), 0)
+ last_line = next((l.lineno for l in reversed(self.leaves) if l.lineno != 0), 0)
- if self.leaves[0].lineno == self.leaves[last_idx].lineno:
- for node in self.leaves[last_idx:]:
+ if first_line == last_line:
+ # We look at the last two leaves since a comma or an
+ # invisible paren could have been added at the end of the
+ # line.
+ for node in self.leaves[-2:]:
for comment in self.comments.get(id(node), []):
if is_type_comment(comment, " ignore"):
return True
def maybe_remove_trailing_comma(self, closing: Leaf) -> bool:
"""Remove trailing comma if there is one and it's safe."""
+ if not (self.leaves and self.leaves[-1].type == token.COMMA):
+ return False
+
+ # We remove trailing commas only in the case of importing a
+ # single name from a module.
if not (
self.leaves
+ and self.is_import
+ and len(self.leaves) > 4
and self.leaves[-1].type == token.COMMA
and closing.type in CLOSING_BRACKETS
+ and self.leaves[-4].type == token.NAME
+ and (
+ # regular `from foo import bar,`
+ self.leaves[-4].value == "import"
+ # `from foo import (bar as baz,)
+ or (
+ len(self.leaves) > 6
+ and self.leaves[-6].value == "import"
+ and self.leaves[-3].value == "as"
+ )
+ # `from foo import bar as baz,`
+ or (
+ len(self.leaves) > 5
+ and self.leaves[-5].value == "import"
+ and self.leaves[-3].value == "as"
+ )
+ )
+ and closing.type == token.RPAR
):
return False
- if closing.type == token.RBRACE:
- self.remove_trailing_comma()
- 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()
- return True
-
- return False
+ self.remove_trailing_comma()
+ return True
def append_comment(self, comment: Leaf) -> bool:
"""Add an inline or standalone comment to the line."""
comment.type = STANDALONE_COMMENT
comment.prefix = ""
return False
+
last_leaf = self.leaves[-2]
self.comments.setdefault(id(last_leaf), []).append(comment)
return True
self.current_line.append(node)
yield from super().visit_default(node)
- def visit_atom(self, node: Node) -> Iterator[Line]:
- # Always make parentheses invisible around a single node, because it should
- # not be needed (except in the case of yield, where removing the parentheses
- # produces a SyntaxError).
- if (
- len(node.children) == 3
- and isinstance(node.children[0], Leaf)
- and node.children[0].type == token.LPAR
- and isinstance(node.children[2], Leaf)
- and node.children[2].type == token.RPAR
- and isinstance(node.children[1], Leaf)
- and not (
- node.children[1].type == token.NAME
- and node.children[1].value == "yield"
- )
- ):
- node.children[0].value = ""
- node.children[2].value = ""
- yield from super().visit_default(node)
-
- def visit_factor(self, node: Node) -> Iterator[Line]:
- """Force parentheses between a unary op and a binary power:
-
- -2 ** 8 -> -(2 ** 8)
- """
- child = node.children[1]
- if child.type == syms.power and len(child.children) == 3:
- lpar = Leaf(token.LPAR, "(")
- rpar = Leaf(token.RPAR, ")")
- index = child.remove() or 0
- node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
- yield from self.visit_default(node)
-
def visit_INDENT(self, node: Node) -> Iterator[Line]:
"""Increase indentation level, maybe yield a line."""
# In blib2to3 INDENT never holds comments.
yield from self.line()
yield from self.visit_default(leaf)
+ def visit_factor(self, node: Node) -> Iterator[Line]:
+ """Force parentheses between a unary op and a binary power:
+
+ -2 ** 8 -> -(2 ** 8)
+ """
+ _operator, operand = node.children
+ if (
+ operand.type == syms.power
+ and len(operand.children) == 3
+ and operand.children[1].type == token.DOUBLESTAR
+ ):
+ lpar = Leaf(token.LPAR, "(")
+ rpar = Leaf(token.RPAR, ")")
+ index = operand.remove() or 0
+ node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
+ yield from self.visit_default(node)
+
def __attrs_post_init__(self) -> None:
"""You are in a twisty little maze of passages."""
v = self.visit_stmt
if (
not line.contains_uncollapsable_type_comments()
and not line.should_explode
+ and not line.is_collection_with_optional_trailing_comma
and (
is_line_short_enough(line, line_length=line_length, line_str=line_str)
or line.contains_unsplittable_type_ignore()
# 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, line_length, features=features)
+ # 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)
if line.inside_brackets:
split_funcs = [delimiter_split, standalone_comment_split, rhs]
# Since body is a new indent level, remove spurious leading whitespace.
normalize_prefix(leaves[0], inside_brackets=True)
# Ensure a trailing comma for imports and standalone function arguments, but
- # be careful not to add one after any comments.
- no_commas = original.is_def and not any(
- l.type == token.COMMA for l in leaves
+ # be careful not to add one after any comments or within type annotations.
+ no_commas = (
+ original.is_def
+ and opening_bracket.value == "("
+ and not any(l.type == token.COMMA for l in leaves)
)
if original.is_import or no_commas:
for i in range(len(leaves) - 1, -1, -1):
if leaves[i].type == STANDALONE_COMMENT:
continue
- elif leaves[i].type == token.COMMA:
- break
- else:
+
+ if leaves[i].type != token.COMMA:
leaves.insert(i + 1, Leaf(token.COMMA, ","))
- break
+ break
+
# Populate the line
for leaf in leaves:
result.append(leaf, preformatted=True)
Only returns true for type comments for now."""
t = leaf.type
v = leaf.value
- return t in {token.COMMENT, t == STANDALONE_COMMENT} and v.startswith(
- "# type:" + suffix
- )
+ return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix)
def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
if "\\" in str(m):
# Do not introduce backslashes in interpolated expressions
return
+
if new_quote == '"""' and new_body[-1:] == '"':
# edge case:
new_body = new_body[:-1] + '\\"'
if pc.value in FMT_OFF:
# This `node` has a prefix with `# fmt: off`, don't mess with parens.
return
-
check_lpar = False
for index, child in enumerate(list(node.children)):
# Add parentheses around long tuple unpacking in assignments.
if check_lpar:
if is_walrus_assignment(child):
continue
- if child.type == syms.atom:
- # Determines if the underlying atom should be surrounded with
- # invisible params - also makes parens invisible recursively
- # within the atom and removes repeated invisible parens within
- # the atom
- should_surround_with_parens = maybe_make_parens_invisible_in_atom(
- child, parent=node
- )
- if should_surround_with_parens:
- lpar = Leaf(token.LPAR, "")
- rpar = Leaf(token.RPAR, "")
- index = child.remove() or 0
- node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+ if child.type == syms.atom:
+ if maybe_make_parens_invisible_in_atom(child, parent=node):
+ wrap_in_parentheses(node, child, visible=False)
elif is_one_tuple(child):
- # wrap child in visible parentheses
- lpar = Leaf(token.LPAR, "(")
- rpar = Leaf(token.RPAR, ")")
- child.remove()
- node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+ wrap_in_parentheses(node, child, visible=True)
elif node.type == syms.import_from:
# "import from" nodes store parentheses directly as part of
# the statement
break
elif not (isinstance(child, Leaf) and is_multiline_string(child)):
- # wrap child in invisible parentheses
- lpar = Leaf(token.LPAR, "")
- rpar = Leaf(token.RPAR, "")
- index = child.remove() or 0
- prefix = child.prefix
- child.prefix = ""
- new_child = Node(syms.atom, [lpar, child, rpar])
- new_child.prefix = prefix
- node.insert_child(index, new_child)
+ wrap_in_parentheses(node, child, visible=False)
check_lpar = isinstance(child, Leaf) and child.value in parens_after
"""
container: Optional[LN] = container_of(leaf)
while container is not None and container.type != token.ENDMARKER:
+ is_fmt_on = False
for comment in list_comments(container.prefix, is_endmarker=False):
if comment.value in FMT_ON:
- return
+ is_fmt_on = True
+ elif comment.value in FMT_OFF:
+ is_fmt_on = False
+ if is_fmt_on:
+ return
yield container
Parenthesis can be optional. Returns None otherwise"""
if len(node.children) != 3:
return None
+
lpar, wrapped, rpar = node.children
if not (lpar.type == token.LPAR and rpar.type == token.RPAR):
return None
return wrapped
+def wrap_in_parentheses(parent: Node, child: LN, *, visible: bool = True) -> None:
+ """Wrap `child` in parentheses.
+
+ This replaces `child` with an atom holding the parentheses and the old
+ child. That requires moving the prefix.
+
+ If `visible` is False, the leaves will be valueless (and thus invisible).
+ """
+ lpar = Leaf(token.LPAR, "(" if visible else "")
+ rpar = Leaf(token.RPAR, ")" if visible else "")
+ prefix = child.prefix
+ child.prefix = ""
+ index = child.remove() or 0
+ new_child = Node(syms.atom, [lpar, child, rpar])
+ new_child.prefix = prefix
+ parent.insert_child(index, new_child)
+
+
def is_one_tuple(node: LN) -> bool:
"""Return True if `node` holds a tuple with one element, with or without parens."""
if node.type == syms.atom:
if isinstance(child, Leaf):
if child.type == token.NAME:
yield child.value
+
elif child.type == syms.import_as_name:
orig_name = child.children[0]
assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports"
assert orig_name.type == token.NAME, "Invalid syntax parsing imports"
yield orig_name.value
+
elif child.type == syms.import_as_names:
yield from get_imports_from_children(child.children)
+
else:
raise AssertionError("Invalid syntax parsing imports")
for child in node.children:
if child.type != syms.simple_stmt:
break
+
first_child = child.children[0]
if isinstance(first_child, Leaf):
# Continue looking if we see a docstring; otherwise stop.
and child.children[1].type == token.NEWLINE
):
continue
- else:
- break
+
+ break
+
elif first_child.type == syms.import_from:
module_name = first_child.children[1]
if not isinstance(module_name, Leaf) or module_name.value != "__future__":
break
+
imports |= set(get_imports_from_children(first_child.children[3:]))
else:
break
+
return imports
+@lru_cache()
+def get_gitignore(root: Path) -> PathSpec:
+ """ Return a PathSpec matching gitignore content if present."""
+ gitignore = root / ".gitignore"
+ lines: List[str] = []
+ if gitignore.is_file():
+ with gitignore.open() as gf:
+ lines = gf.readlines()
+ return PathSpec.from_lines("gitwildmatch", lines)
+
+
def gen_python_files_in_dir(
path: Path,
root: Path,
include: Pattern[str],
exclude: Pattern[str],
report: "Report",
+ gitignore: PathSpec,
) -> Iterator[Path]:
"""Generate all files under `path` whose paths are not excluded by the
`exclude` regex, but are included by the `include` regex.
"""
assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}"
for child in path.iterdir():
+ # First ignore files matching .gitignore
+ if gitignore.match_file(child.as_posix()):
+ report.path_ignored(child, f"matches the .gitignore file content")
+ continue
+
+ # Then ignore with `exclude` option.
try:
normalized_path = "/" + child.resolve().relative_to(root).as_posix()
+ except OSError as e:
+ report.path_ignored(child, f"cannot be read because {e}")
+ continue
+
except ValueError:
if child.is_symlink():
report.path_ignored(
if child.is_dir():
normalized_path += "/"
+
exclude_match = exclude.search(normalized_path)
if exclude_match and exclude_match.group(0):
report.path_ignored(child, f"matches the --exclude regular expression")
continue
if child.is_dir():
- yield from gen_python_files_in_dir(child, root, include, exclude, report)
+ yield from gen_python_files_in_dir(
+ child, root, include, exclude, report, gitignore
+ )
elif child.is_file():
include_match = include.search(normalized_path)
node: Union[ast.AST, ast3.AST, ast27.AST]
) -> Union[ast.AST, ast3.AST, ast27.AST]:
"""Map ast nodes deprecated in 3.8 to Constant."""
- # casts are required until this is released:
- # https://github.com/python/typeshed/pull/3142
if isinstance(node, (ast.Str, ast3.Str, ast27.Str, ast.Bytes, ast3.Bytes)):
- return cast(ast.AST, ast.Constant(value=node.s))
- elif isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
- return cast(ast.AST, ast.Constant(value=node.n))
- elif isinstance(node, (ast.NameConstant, ast3.NameConstant)):
- return cast(ast.AST, ast.Constant(value=node.value))
+ return ast.Constant(value=node.s)
+
+ if isinstance(node, (ast.Num, ast3.Num, ast27.Num)):
+ return ast.Constant(value=node.n)
+
+ if isinstance(node, (ast.NameConstant, ast3.NameConstant)):
+ return ast.Constant(value=node.value)
+
return node
):
for item in item.elts:
yield from _v(item, depth + 2)
+
elif isinstance(item, (ast.AST, ast3.AST, ast27.AST)):
yield from _v(item, depth + 2)
@contextmanager
def nullcontext() -> Iterator[None]:
- """Return context manager that does nothing.
- Similar to `nullcontext` from python 3.7"""
+ """Return an empty context manager.
+
+ To be used like `nullcontext` in Python 3.7.
+ """
yield
with cache_file.open("rb") as fobj:
try:
cache: Cache = pickle.load(fobj)
- except pickle.UnpicklingError:
+ except (pickle.UnpicklingError, ValueError):
return {}
return cache
CACHE_DIR.mkdir(parents=True, exist_ok=True)
new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}}
with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f:
- pickle.dump(new_cache, f, protocol=pickle.HIGHEST_PROTOCOL)
+ pickle.dump(new_cache, f, protocol=4)
os.replace(f.name, cache_file)
except OSError:
pass
projects / etc / vim.git / blobdiff