import os
from pathlib import Path
import tokenize
+import sys
from typing import (
Dict, Generic, Iterable, Iterator, List, Optional, Set, Tuple, TypeVar, Union
)
-from attr import attrib, dataclass, Factory
+from attr import dataclass, Factory
import click
# lib2to3 fork
from blib2to3.pgen2 import driver, token
from blib2to3.pgen2.parse import ParseError
-__version__ = "18.3a1"
+__version__ = "18.3a2"
DEFAULT_LINE_LENGTH = 88
# types
syms = pygram.python_symbols
help='How many character per line to allow.',
show_default=True,
)
+@click.option(
+ '--check',
+ is_flag=True,
+ help=(
+ "Don't write back the files, just return the status. Return code 0 "
+ "means nothing changed. Return code 1 means some files were "
+ "reformatted. Return code 123 means there was an internal error."
+ ),
+)
@click.option(
'--fast/--safe',
is_flag=True,
type=click.Path(exists=True, file_okay=True, dir_okay=True, readable=True),
)
@click.pass_context
-def main(ctx: click.Context, line_length: int, fast: bool, src: List[str]) -> None:
+def main(
+ ctx: click.Context, line_length: int, check: bool, fast: bool, src: List[str]
+) -> None:
"""The uncompromising code formatter."""
sources: List[Path] = []
for s in src:
p = sources[0]
report = Report()
try:
- changed = format_file_in_place(p, line_length=line_length, fast=fast)
+ changed = format_file_in_place(
+ p, line_length=line_length, fast=fast, write_back=not check
+ )
report.done(p, changed)
except Exception as exc:
report.failed(p, str(exc))
return_code = 1
try:
return_code = loop.run_until_complete(
- schedule_formatting(sources, line_length, fast, loop, executor)
+ schedule_formatting(
+ sources, line_length, not check, fast, loop, executor
+ )
)
finally:
loop.close()
async def schedule_formatting(
sources: List[Path],
line_length: int,
+ write_back: bool,
fast: bool,
loop: BaseEventLoop,
executor: Executor,
) -> int:
tasks = {
src: loop.run_in_executor(
- executor, format_file_in_place, src, line_length, fast
+ executor, format_file_in_place, src, line_length, fast, write_back
)
for src in sources
}
return report.return_code
-def format_file_in_place(src: Path, line_length: int, fast: bool) -> bool:
+def format_file_in_place(
+ src: Path, line_length: int, fast: bool, write_back: bool = False
+) -> bool:
"""Format the file and rewrite if changed. Return True if changed."""
try:
contents, encoding = format_file(src, line_length=line_length, fast=fast)
except NothingChanged:
return False
- with open(src, "w", encoding=encoding) as f:
- f.write(contents)
+ if write_back:
+ with open(src, "w", encoding=encoding) as f:
+ f.write(contents)
return True
comments: List[Line] = []
lines = LineGenerator()
elt = EmptyLineTracker()
+ py36 = is_python36(src_node)
empty_line = Line()
after = 0
for current_line in lines.visit(src_node):
for comment in comments:
dst_contents += str(comment)
comments = []
- for line in split_line(current_line, line_length=line_length):
+ for line in split_line(current_line, line_length=line_length, py36=py36):
dst_contents += str(line)
else:
comments.append(current_line)
@dataclass
class DebugVisitor(Visitor[T]):
- tree_depth: int = attrib(default=0)
+ tree_depth: int = 0
def visit_default(self, node: LN) -> Iterator[T]:
indent = ' ' * (2 * self.tree_depth)
@dataclass
class BracketTracker:
- depth: int = attrib(default=0)
- bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = attrib(default=Factory(dict))
- delimiters: Dict[LeafID, Priority] = attrib(default=Factory(dict))
- previous: Optional[Leaf] = attrib(default=None)
+ depth: int = 0
+ bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict)
+ delimiters: Dict[LeafID, Priority] = Factory(dict)
+ previous: Optional[Leaf] = None
def mark(self, leaf: Leaf) -> None:
if leaf.type == token.COMMENT:
if leaf.type in CLOSING_BRACKETS:
self.depth -= 1
opening_bracket = self.bracket_match.pop((self.depth, leaf.type))
- leaf.opening_bracket = opening_bracket # type: ignore
- leaf.bracket_depth = self.depth # type: ignore
+ leaf.opening_bracket = opening_bracket
+ leaf.bracket_depth = self.depth
if self.depth == 0:
delim = is_delimiter(leaf)
if delim:
"""Returns True if there is an yet unmatched open bracket on the line."""
return bool(self.bracket_match)
- def max_priority(self, exclude: Iterable[LeafID] = ()) -> int:
+ def max_priority(self, exclude: Iterable[LeafID] =()) -> int:
"""Returns the highest priority of a delimiter found on the line.
Values are consistent with what `is_delimiter()` returns.
@dataclass
class Line:
- depth: int = attrib(default=0)
- leaves: List[Leaf] = attrib(default=Factory(list))
- comments: Dict[LeafID, Leaf] = attrib(default=Factory(dict))
- bracket_tracker: BracketTracker = attrib(default=Factory(BracketTracker))
- inside_brackets: bool = attrib(default=False)
+ depth: int = 0
+ leaves: List[Leaf] = Factory(list)
+ comments: Dict[LeafID, Leaf] = Factory(dict)
+ bracket_tracker: BracketTracker = Factory(BracketTracker)
+ inside_brackets: bool = False
+ has_for: bool = False
+ _for_loop_variable: bool = False
def append(self, leaf: Leaf, preformatted: bool = False) -> None:
has_value = leaf.value.strip()
# imports, for which we only preserve newlines.
leaf.prefix += whitespace(leaf)
if self.inside_brackets or not preformatted:
+ self.maybe_decrement_after_for_loop_variable(leaf)
self.bracket_tracker.mark(leaf)
self.maybe_remove_trailing_comma(leaf)
+ self.maybe_increment_for_loop_variable(leaf)
if self.maybe_adapt_standalone_comment(leaf):
return
# For parens let's check if it's safe to remove the comma. 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 # type: ignore
+ depth = closing.bracket_depth + 1
commas = 0
- opening = closing.opening_bracket # type: ignore
+ opening = closing.opening_bracket
for _opening_index, leaf in enumerate(self.leaves):
if leaf is opening:
break
if leaf is closing:
break
- bracket_depth = leaf.bracket_depth # type: ignore
+ bracket_depth = leaf.bracket_depth
if bracket_depth == depth and leaf.type == token.COMMA:
commas += 1
+ if leaf.parent and leaf.parent.type == syms.arglist:
+ commas += 1
+ break
+
if commas > 1:
self.leaves.pop()
return True
return False
+ def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool:
+ """In a for loop, or comprehension, the variables are often unpacks.
+
+ To avoid splitting on the comma in this situation, we will increase
+ the depth of tokens between `for` and `in`.
+ """
+ if leaf.type == token.NAME and leaf.value == 'for':
+ self.has_for = True
+ self.bracket_tracker.depth += 1
+ self._for_loop_variable = True
+ return True
+
+ return False
+
+ def maybe_decrement_after_for_loop_variable(self, leaf: Leaf) -> bool:
+ # See `maybe_increment_for_loop_variable` above for explanation.
+ if self._for_loop_variable and leaf.type == token.NAME and leaf.value == 'in':
+ self.bracket_tracker.depth -= 1
+ self._for_loop_variable = False
+ return True
+
+ return False
+
def maybe_adapt_standalone_comment(self, comment: Leaf) -> bool:
"""Hack a standalone comment to act as a trailing comment for line splitting.
Note: this tracker works on lines that haven't been split yet.
"""
- previous_line: Optional[Line] = attrib(default=None)
- previous_after: int = attrib(default=0)
- previous_defs: List[int] = attrib(default=Factory(list))
+ previous_line: Optional[Line] = None
+ previous_after: int = 0
+ previous_defs: List[int] = Factory(list)
def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
"""Returns the number of extra empty lines before and after the `current_line`.
Note: destroys the tree it's visiting by mutating prefixes of its leaves
in ways that will no longer stringify to valid Python code on the tree.
"""
- current_line: Line = attrib(default=Factory(Line))
- standalone_comments: List[Leaf] = attrib(default=Factory(list))
+ current_line: Line = Factory(Line)
+ standalone_comments: List[Leaf] = Factory(list)
def line(self, indent: int = 0) -> Iterator[Line]:
"""Generate a line.
OPENING_BRACKETS = set(BRACKET.keys())
CLOSING_BRACKETS = set(BRACKET.values())
BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS
+ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, token.COLON, STANDALONE_COMMENT}
-def whitespace(leaf: Leaf) -> str:
+def whitespace(leaf: Leaf) -> str: # noqa C901
"""Return whitespace prefix if needed for the given `leaf`."""
NO = ''
SPACE = ' '
t = leaf.type
p = leaf.parent
v = leaf.value
- if t == token.COLON:
- return NO
-
- if t == token.COMMA:
- return NO
-
- if t == token.RPAR:
+ if t in ALWAYS_NO_SPACE:
return NO
if t == token.COMMENT:
return DOUBLESPACE
- if t == STANDALONE_COMMENT:
+ assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
+ prev = leaf.prev_sibling
+ if not prev:
+ prevp = preceding_leaf(p)
+ if not prevp or prevp.type in OPENING_BRACKETS:
+ return NO
+
+ if prevp.type == token.EQUAL:
+ if prevp.parent and prevp.parent.type in {
+ syms.typedargslist,
+ syms.varargslist,
+ syms.parameters,
+ syms.arglist,
+ syms.argument,
+ }:
+ return NO
+
+ elif prevp.type == token.DOUBLESTAR:
+ if prevp.parent and prevp.parent.type in {
+ syms.typedargslist,
+ syms.varargslist,
+ syms.parameters,
+ syms.arglist,
+ syms.dictsetmaker,
+ }:
+ return NO
+
+ elif prevp.type == token.COLON:
+ if prevp.parent and prevp.parent.type == syms.subscript:
+ return NO
+
+ elif prevp.parent and prevp.parent.type in {syms.factor, syms.star_expr}:
+ return NO
+
+ elif prev.type in OPENING_BRACKETS:
return NO
- assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}"
if p.type in {syms.parameters, syms.arglist}:
# untyped function signatures or calls
if t == token.RPAR:
return NO
- prev = leaf.prev_sibling
if not prev or prev.type != token.COMMA:
return NO
if t == token.RPAR:
return NO
- prev = leaf.prev_sibling
if prev and prev.type != token.COMMA:
return NO
elif p.type == syms.typedargslist:
# typed function signatures
- prev = leaf.prev_sibling
if not prev:
return NO
elif p.type == syms.tname:
# type names
- prev = leaf.prev_sibling
if not prev:
prevp = preceding_leaf(p)
if not prevp or prevp.type != token.COMMA:
if t == token.LPAR or t == token.RPAR:
return NO
- prev = leaf.prev_sibling
if not prev:
if t == token.DOT:
prevp = preceding_leaf(p)
if t == token.EQUAL:
return NO
- prev = leaf.prev_sibling
if not prev:
prevp = preceding_leaf(p)
if not prevp or prevp.type == token.LPAR:
return NO
elif p.type == syms.dotted_name:
- prev = leaf.prev_sibling
if prev:
return NO
if t == token.LPAR:
return NO
- prev = leaf.prev_sibling
if prev and prev.type == token.LPAR:
return NO
elif p.type == syms.subscript:
# indexing
- if t == token.COLON:
- return NO
-
- prev = leaf.prev_sibling
- if not prev or prev.type == token.COLON:
- return NO
-
- elif p.type in {
- syms.test,
- syms.not_test,
- syms.xor_expr,
- syms.or_test,
- syms.and_test,
- syms.arith_expr,
- syms.shift_expr,
- syms.yield_expr,
- syms.term,
- syms.power,
- syms.comparison,
- }:
- # various arithmetic and logic expressions
- prev = leaf.prev_sibling
if not prev:
- prevp = preceding_leaf(p)
- if not prevp or prevp.type in OPENING_BRACKETS:
- return NO
-
- if prevp.type == token.EQUAL:
- if prevp.parent and prevp.parent.type in {
- syms.varargslist, syms.parameters, syms.arglist, syms.argument
- }:
- return NO
-
- return SPACE
+ assert p.parent is not None, "subscripts are always parented"
+ if p.parent.type == syms.subscriptlist:
+ return SPACE
- elif p.type == syms.atom:
- if t in CLOSING_BRACKETS:
return NO
- prev = leaf.prev_sibling
- if not prev:
- prevp = preceding_leaf(p)
- if not prevp:
- return NO
-
- if prevp.type in OPENING_BRACKETS:
- return NO
-
- if prevp.type == token.EQUAL:
- if prevp.parent and prevp.parent.type in {
- syms.varargslist, syms.parameters, syms.arglist, syms.argument
- }:
- return NO
-
- if prevp.type == token.DOUBLESTAR:
- if prevp.parent and prevp.parent.type in {
- syms.varargslist, syms.parameters, syms.arglist, syms.dictsetmaker
- }:
- return NO
-
- elif prev.type in OPENING_BRACKETS:
+ elif prev.type == token.COLON:
return NO
- elif t == token.DOT:
+ elif p.type == syms.atom:
+ if prev and t == token.DOT:
# dots, but not the first one.
return NO
p.type == syms.subscriptlist
):
# list interior, including unpacking
- prev = leaf.prev_sibling
if not prev:
return NO
elif p.type == syms.dictsetmaker:
# dict and set interior, including unpacking
- prev = leaf.prev_sibling
if not prev:
return NO
if prev.type == token.DOUBLESTAR:
return NO
- elif p.type == syms.factor or p.type == syms.star_expr:
+ elif p.type in {syms.factor, syms.star_expr}:
# unary ops
- prev = leaf.prev_sibling
if not prev:
prevp = preceding_leaf(p)
if not prevp or prevp.type in OPENING_BRACKETS:
elif p.type == syms.import_from:
if t == token.DOT:
- prev = leaf.prev_sibling
if prev and prev.type == token.DOT:
return NO
if v == 'import':
return SPACE
- prev = leaf.prev_sibling
if prev and prev.type == token.DOT:
return NO
yield Leaf(STANDALONE_COMMENT, line)
-def split_line(line: Line, line_length: int, inner: bool = False) -> Iterator[Line]:
+def split_line(
+ line: Line, line_length: int, inner: bool = False, py36: bool = False
+) -> Iterator[Line]:
"""Splits a `line` into potentially many lines.
They should fit in the allotted `line_length` but might not be able to.
`inner` signifies that there were a pair of brackets somewhere around the
current `line`, possibly transitively. This means we can fallback to splitting
by delimiters if the LHS/RHS don't yield any results.
+
+ If `py36` is True, splitting may generate syntax that is only compatible
+ with Python 3.6 and later.
"""
line_str = str(line).strip('\n')
if len(line_str) <= line_length and '\n' not in line_str:
# split altogether.
result: List[Line] = []
try:
- for l in split_func(line):
+ for l in split_func(line, py36=py36):
if str(l).strip('\n') == line_str:
raise CannotSplit("Split function returned an unchanged result")
- result.extend(split_line(l, line_length=line_length, inner=True))
+ result.extend(
+ split_line(l, line_length=line_length, inner=True, py36=py36)
+ )
except CannotSplit as cs:
continue
yield line
-def left_hand_split(line: Line) -> Iterator[Line]:
+def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
"""Split line into many lines, starting with the first matching bracket pair.
Note: this usually looks weird, only use this for function definitions.
if (
current_leaves is body_leaves and
leaf.type in CLOSING_BRACKETS and
- leaf.opening_bracket is matching_bracket # type: ignore
+ leaf.opening_bracket is matching_bracket
):
- current_leaves = tail_leaves
+ current_leaves = tail_leaves if body_leaves else head_leaves
current_leaves.append(leaf)
if current_leaves is head_leaves:
if leaf.type in OPENING_BRACKETS:
comment_after = line.comments.get(id(leaf))
if comment_after:
result.append(comment_after, preformatted=True)
- # Check if the split succeeded.
- tail_len = len(str(tail))
- if not body:
- if tail_len == 0:
- raise CannotSplit("Splitting brackets produced the same line")
-
- elif tail_len < 3:
- raise CannotSplit(
- f"Splitting brackets on an empty body to save "
- f"{tail_len} characters is not worth it"
- )
-
+ split_succeeded_or_raise(head, body, tail)
for result in (head, body, tail):
if result:
yield result
-def right_hand_split(line: Line) -> Iterator[Line]:
+def right_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
"""Split line into many lines, starting with the last matching bracket pair."""
head = Line(depth=line.depth)
body = Line(depth=line.depth + 1, inside_brackets=True)
for leaf in reversed(line.leaves):
if current_leaves is body_leaves:
if leaf is opening_bracket:
- current_leaves = head_leaves
+ current_leaves = head_leaves if body_leaves else tail_leaves
current_leaves.append(leaf)
if current_leaves is tail_leaves:
if leaf.type in CLOSING_BRACKETS:
- opening_bracket = leaf.opening_bracket # type: ignore
+ opening_bracket = leaf.opening_bracket
current_leaves = body_leaves
tail_leaves.reverse()
body_leaves.reverse()
comment_after = line.comments.get(id(leaf))
if comment_after:
result.append(comment_after, preformatted=True)
- # Check if the split succeeded.
- tail_len = len(str(tail).strip('\n'))
+ split_succeeded_or_raise(head, body, tail)
+ for result in (head, body, tail):
+ if result:
+ yield result
+
+
+def split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None:
+ tail_len = len(str(tail).strip())
if not body:
if tail_len == 0:
raise CannotSplit("Splitting brackets produced the same line")
f"{tail_len} characters is not worth it"
)
- for result in (head, body, tail):
- if result:
- yield result
-
-def delimiter_split(line: Line) -> Iterator[Line]:
+def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]:
"""Split according to delimiters of the highest priority.
This kind of split doesn't increase indentation.
+ If `py36` is True, the split will add trailing commas also in function
+ signatures that contain * and **.
"""
try:
last_leaf = line.leaves[-1]
raise CannotSplit("No delimiters found")
current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
+ lowest_depth = sys.maxsize
+ trailing_comma_safe = True
for leaf in line.leaves:
current_line.append(leaf, preformatted=True)
comment_after = line.comments.get(id(leaf))
if comment_after:
current_line.append(comment_after, preformatted=True)
+ lowest_depth = min(lowest_depth, leaf.bracket_depth)
+ if (
+ leaf.bracket_depth == lowest_depth and
+ leaf.type == token.STAR or
+ leaf.type == token.DOUBLESTAR
+ ):
+ trailing_comma_safe = trailing_comma_safe and py36
leaf_priority = delimiters.get(id(leaf))
if leaf_priority == delimiter_priority:
normalize_prefix(current_line.leaves[0])
if current_line:
if (
delimiter_priority == COMMA_PRIORITY and
- current_line.leaves[-1].type != token.COMMA
+ current_line.leaves[-1].type != token.COMMA and
+ trailing_comma_safe
):
current_line.append(Leaf(token.COMMA, ','))
normalize_prefix(current_line.leaves[0])
leaf.prefix = ''
+def is_python36(node: Node) -> bool:
+ """Returns True if the current file is using Python 3.6+ features.
+
+ Currently looking for:
+ - f-strings; and
+ - trailing commas after * or ** in function signatures.
+ """
+ for n in node.pre_order():
+ if n.type == token.STRING:
+ value_head = n.value[:2] # type: ignore
+ if value_head in {'f"', 'F"', "f'", "F'", 'rf', 'fr', 'RF', 'FR'}:
+ return True
+
+ elif (
+ n.type == syms.typedargslist and
+ n.children and
+ n.children[-1].type == token.COMMA
+ ):
+ for ch in n.children:
+ if ch.type == token.STAR or ch.type == token.DOUBLESTAR:
+ return True
+
+ return False
+
+
PYTHON_EXTENSIONS = {'.py'}
BLACKLISTED_DIRECTORIES = {
'build', 'buck-out', 'dist', '_build', '.git', '.hg', '.mypy_cache', '.tox', '.venv'
@dataclass
class Report:
"""Provides a reformatting counter."""
- change_count: int = attrib(default=0)
- same_count: int = attrib(default=0)
- failure_count: int = attrib(default=0)
+ change_count: int = 0
+ same_count: int = 0
+ failure_count: int = 0
def done(self, src: Path, changed: bool) -> None:
"""Increment the counter for successful reformatting. Write out a message."""
@property
def return_code(self) -> int:
"""Which return code should the app use considering the current state."""
- return 1 if self.failure_count else 0
+ # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with
+ # 126 we have special returncodes reserved by the shell.
+ if self.failure_count:
+ return 123
+
+ elif self.change_count:
+ return 1
+
+ return 0
def __str__(self) -> str:
"""A color report of the current state.
raise AssertionError(
f"INTERNAL ERROR: Black produced invalid code: {exc}. "
f"Please report a bug on https://github.com/ambv/black/issues. "
- f"This invalid output might be helpful: {log}",
+ f"This invalid output might be helpful: {log}"
) from None
src_ast_str = '\n'.join(_v(src_ast))
f"INTERNAL ERROR: Black produced code that is not equivalent to "
f"the source. "
f"Please report a bug on https://github.com/ambv/black/issues. "
- f"This diff might be helpful: {log}",
+ f"This diff might be helpful: {log}"
) from None
f"INTERNAL ERROR: Black produced different code on the second pass "
f"of the formatter. "
f"Please report a bug on https://github.com/ambv/black/issues. "
- f"This diff might be helpful: {log}",
+ f"This diff might be helpful: {log}"
) from None
projects / etc / vim.git / blobdiff