import asyncio
from asyncio.base_events import BaseEventLoop
from concurrent.futures import Executor, ProcessPoolExecutor
+from enum import Enum
from functools import partial, wraps
import keyword
import logging
+from multiprocessing import Manager
import os
from pathlib import Path
+import re
import tokenize
import signal
import sys
from typing import (
+ Any,
Callable,
+ Collection,
Dict,
Generic,
Iterable,
from blib2to3.pgen2 import driver, token
from blib2to3.pgen2.parse import ParseError
-__version__ = "18.3a4"
+__version__ = "18.4a0"
DEFAULT_LINE_LENGTH = 88
# types
syms = pygram.python_symbols
"""Found a comment like `# fmt: off` in the file."""
+class WriteBack(Enum):
+ NO = 0
+ YES = 1
+ DIFF = 2
+
+
@click.command()
@click.option(
"-l",
"--check",
is_flag=True,
help=(
- "Don't write back the files, just return the status. Return code 0 "
+ "Don't write the files back, just return the status. Return code 0 "
"means nothing would change. Return code 1 means some files would be "
"reformatted. Return code 123 means there was an internal error."
),
)
+@click.option(
+ "--diff",
+ is_flag=True,
+ help="Don't write the files back, just output a diff for each file on stdout.",
+)
@click.option(
"--fast/--safe",
is_flag=True,
help="If --fast given, skip temporary sanity checks. [default: --safe]",
)
+@click.option(
+ "-q",
+ "--quiet",
+ is_flag=True,
+ help=(
+ "Don't emit non-error messages to stderr. Errors are still emitted, "
+ "silence those with 2>/dev/null."
+ ),
+)
@click.version_option(version=__version__)
@click.argument(
"src",
)
@click.pass_context
def main(
- ctx: click.Context, line_length: int, check: bool, fast: bool, src: List[str]
+ ctx: click.Context,
+ line_length: int,
+ check: bool,
+ diff: bool,
+ fast: bool,
+ quiet: bool,
+ src: List[str],
) -> None:
"""The uncompromising code formatter."""
sources: List[Path] = []
sources.append(Path("-"))
else:
err(f"invalid path: {s}")
+ if check and diff:
+ exc = click.ClickException("Options --check and --diff are mutually exclusive")
+ exc.exit_code = 2
+ raise exc
+
+ if check:
+ write_back = WriteBack.NO
+ elif diff:
+ write_back = WriteBack.DIFF
+ else:
+ write_back = WriteBack.YES
if len(sources) == 0:
ctx.exit(0)
elif len(sources) == 1:
p = sources[0]
- report = Report(check=check)
+ report = Report(check=check, quiet=quiet)
try:
if not p.is_file() and str(p) == "-":
changed = format_stdin_to_stdout(
- line_length=line_length, fast=fast, write_back=not check
+ line_length=line_length, fast=fast, write_back=write_back
)
else:
changed = format_file_in_place(
- p, line_length=line_length, fast=fast, write_back=not check
+ p, line_length=line_length, fast=fast, write_back=write_back
)
report.done(p, changed)
except Exception as exc:
try:
return_code = loop.run_until_complete(
schedule_formatting(
- sources, line_length, not check, fast, loop, executor
+ sources, line_length, write_back, fast, quiet, loop, executor
)
)
finally:
async def schedule_formatting(
sources: List[Path],
line_length: int,
- write_back: bool,
+ write_back: WriteBack,
fast: bool,
+ quiet: bool,
loop: BaseEventLoop,
executor: Executor,
) -> int:
`line_length`, `write_back`, and `fast` options are passed to
:func:`format_file_in_place`.
"""
+ lock = None
+ if write_back == WriteBack.DIFF:
+ # For diff output, we need locks to ensure we don't interleave output
+ # from different processes.
+ manager = Manager()
+ lock = manager.Lock()
tasks = {
src: loop.run_in_executor(
- executor, format_file_in_place, src, line_length, fast, write_back
+ executor, format_file_in_place, src, line_length, fast, write_back, lock
)
for src in sources
}
loop.add_signal_handler(signal.SIGTERM, cancel, _task_values)
await asyncio.wait(tasks.values())
cancelled = []
- report = Report(check=not write_back)
+ report = Report(check=write_back is WriteBack.NO, quiet=quiet)
for src, task in tasks.items():
if not task.done():
report.failed(src, "timed out, cancelling")
report.done(src, task.result())
if cancelled:
await asyncio.gather(*cancelled, loop=loop, return_exceptions=True)
- else:
+ elif not quiet:
out("All done! ✨ 🍰 ✨")
- click.echo(str(report))
+ if not quiet:
+ click.echo(str(report))
return report.return_code
def format_file_in_place(
- src: Path, line_length: int, fast: bool, write_back: bool = False
+ src: Path,
+ line_length: int,
+ fast: bool,
+ write_back: WriteBack = WriteBack.NO,
+ lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy
) -> bool:
"""Format file under `src` path. Return True if changed.
with tokenize.open(src) as src_buffer:
src_contents = src_buffer.read()
try:
- contents = format_file_contents(
+ dst_contents = format_file_contents(
src_contents, line_length=line_length, fast=fast
)
except NothingChanged:
return False
- if write_back:
+ if write_back == write_back.YES:
with open(src, "w", encoding=src_buffer.encoding) as f:
- f.write(contents)
+ f.write(dst_contents)
+ elif write_back == write_back.DIFF:
+ src_name = f"{src.name} (original)"
+ dst_name = f"{src.name} (formatted)"
+ diff_contents = diff(src_contents, dst_contents, src_name, dst_name)
+ if lock:
+ lock.acquire()
+ try:
+ sys.stdout.write(diff_contents)
+ finally:
+ if lock:
+ lock.release()
return True
def format_stdin_to_stdout(
- line_length: int, fast: bool, write_back: bool = False
+ line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO
) -> bool:
"""Format file on stdin. Return True if changed.
If `write_back` is True, write reformatted code back to stdout.
`line_length` and `fast` arguments are passed to :func:`format_file_contents`.
"""
- contents = sys.stdin.read()
+ src = sys.stdin.read()
+ dst = src
try:
- contents = format_file_contents(contents, line_length=line_length, fast=fast)
+ dst = format_file_contents(src, line_length=line_length, fast=fast)
return True
except NothingChanged:
return False
finally:
- if write_back:
- sys.stdout.write(contents)
+ if write_back == WriteBack.YES:
+ sys.stdout.write(dst)
+ elif write_back == WriteBack.DIFF:
+ src_name = "<stdin> (original)"
+ dst_name = "<stdin> (formatted)"
+ sys.stdout.write(diff(src, dst, src_name, dst_name))
def format_file_contents(
leaf.opening_bracket = opening_bracket
leaf.bracket_depth = self.depth
if self.depth == 0:
- after_delim = is_split_after_delimiter(leaf, self.previous)
- before_delim = is_split_before_delimiter(leaf, self.previous)
- if after_delim > before_delim:
- self.delimiters[id(leaf)] = after_delim
- elif before_delim > after_delim and self.previous is not None:
- self.delimiters[id(self.previous)] = before_delim
+ delim = is_split_before_delimiter(leaf, self.previous)
+ if delim and self.previous is not None:
+ self.delimiters[id(self.previous)] = delim
+ else:
+ delim = is_split_after_delimiter(leaf, self.previous)
+ if delim:
+ self.delimiters[id(leaf)] = delim
if leaf.type in OPENING_BRACKETS:
self.bracket_match[self.depth, BRACKET[leaf.type]] = leaf
self.depth += 1
"""Return the highest priority of a delimiter found on the line.
Values are consistent with what `is_delimiter()` returns.
+ Raises ValueError on no delimiters.
"""
return max(v for k, v in self.delimiters.items() if k not in exclude)
Inline comments are put aside.
"""
- has_value = leaf.value.strip()
+ has_value = leaf.type in BRACKETS or bool(leaf.value.strip())
if not has_value:
return
and self.leaves[0].value == "yield"
)
- @property
- def contains_standalone_comments(self) -> bool:
+ 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:
- return True
+ if leaf.bracket_depth <= depth_limit:
+ return True
return False
# DEDENT has no value. Additionally, in blib2to3 it never holds comments.
yield from self.line(-1)
- def visit_stmt(self, node: Node, keywords: Set[str]) -> Iterator[Line]:
+ def visit_stmt(
+ self, node: Node, keywords: Set[str], parens: Set[str]
+ ) -> Iterator[Line]:
"""Visit a statement.
This implementation is shared for `if`, `while`, `for`, `try`, `except`,
- `def`, `with`, and `class`.
+ `def`, `with`, `class`, and `assert`.
+
+ The relevant Python language `keywords` for a given statement will be
+ NAME leaves within it. This methods puts those on a separate line.
- The relevant Python language `keywords` for a given statement will be NAME
- leaves within it. This methods puts those on a separate line.
+ `parens` holds pairs of nodes where invisible parentheses should be put.
+ Keys hold nodes after which opening parentheses should be put, values
+ hold nodes before which closing parentheses should be put.
"""
+ normalize_invisible_parens(node, parens_after=parens)
for child in node.children:
if child.type == token.NAME and child.value in keywords: # type: ignore
yield from self.line()
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()
yield from self.line()
yield from self.visit(node)
+ if node.type == token.ENDMARKER:
+ # somebody decided not to put a final `# fmt: on`
+ yield from self.line()
+
def __attrs_post_init__(self) -> None:
"""You are in a twisty little maze of passages."""
v = self.visit_stmt
- self.visit_if_stmt = partial(v, keywords={"if", "else", "elif"})
- self.visit_while_stmt = partial(v, keywords={"while", "else"})
- self.visit_for_stmt = partial(v, keywords={"for", "else"})
- self.visit_try_stmt = partial(v, keywords={"try", "except", "else", "finally"})
- self.visit_except_clause = partial(v, keywords={"except"})
- self.visit_funcdef = partial(v, keywords={"def"})
- self.visit_with_stmt = partial(v, keywords={"with"})
- self.visit_classdef = partial(v, keywords={"class"})
+ Ø: Set[str] = set()
+ self.visit_assert_stmt = partial(v, keywords={"assert"}, parens={"assert", ","})
+ self.visit_if_stmt = partial(v, keywords={"if", "else", "elif"}, parens={"if"})
+ self.visit_while_stmt = partial(v, keywords={"while", "else"}, parens={"while"})
+ self.visit_for_stmt = partial(v, keywords={"for", "else"}, parens={"for", "in"})
+ self.visit_try_stmt = partial(
+ v, keywords={"try", "except", "else", "finally"}, parens=Ø
+ )
+ self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø)
+ self.visit_with_stmt = partial(v, keywords={"with"}, parens=Ø)
+ self.visit_funcdef = partial(v, keywords={"def"}, parens=Ø)
+ self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
self.visit_async_funcdef = self.visit_async_stmt
self.visit_decorated = self.visit_decorators
+IMPLICIT_TUPLE = {syms.testlist, syms.testlist_star_expr, syms.exprlist}
BRACKET = {token.LPAR: token.RPAR, token.LSQB: token.RSQB, token.LBRACE: token.RBRACE}
OPENING_BRACKETS = set(BRACKET.keys())
CLOSING_BRACKETS = set(BRACKET.values())
return prevp.prefix
elif prevp.type == token.DOUBLESTAR:
- if prevp.parent and prevp.parent.type in {
- syms.arglist,
- syms.argument,
- syms.dictsetmaker,
- syms.parameters,
- syms.typedargslist,
- syms.varargslist,
- }:
+ if (
+ prevp.parent
+ and prevp.parent.type in {
+ syms.arglist,
+ syms.argument,
+ syms.dictsetmaker,
+ syms.parameters,
+ syms.typedargslist,
+ syms.varargslist,
+ }
+ ):
return NO
elif prevp.type == token.COLON:
prevp_parent = prevp.parent
assert prevp_parent is not None
- if prevp.type == token.COLON and prevp_parent.type in {
- syms.subscript, syms.sliceop
- }:
+ if (
+ prevp.type == token.COLON
+ and prevp_parent.type in {syms.subscript, syms.sliceop}
+ ):
return NO
elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument:
if leaf.type == token.COMMA:
return COMMA_PRIORITY
- if (
- leaf.type in VARARGS
- and leaf.parent
- and leaf.parent.type in {syms.argument, syms.typedargslist}
- ):
- return MATH_PRIORITY
-
return 0
Higher numbers are higher priority.
"""
+ if (
+ leaf.type in VARARGS
+ and leaf.parent
+ and leaf.parent.type in {syms.argument, syms.typedargslist}
+ ):
+ # * and ** might also be MATH_OPERATORS but in this case they are not.
+ # Don't treat them as a delimiter.
+ return 0
+
if (
leaf.type in MATH_OPERATORS
and leaf.parent
raise FormatOn(consumed)
if comment in {"# fmt: off", "# yapf: disable"}:
- raise FormatOff(consumed)
+ if comment_type == STANDALONE_COMMENT:
+ raise FormatOff(consumed)
+
+ prev = preceding_leaf(leaf)
+ if not prev or prev.type in WHITESPACE: # standalone comment in disguise
+ raise FormatOff(consumed)
nlines = 0
if (
len(line_str) <= line_length
and "\n" not in line_str # multiline strings
- and not line.contains_standalone_comments
+ and not line.contains_standalone_comments()
):
yield line
return
if body_leaves:
normalize_prefix(body_leaves[0], inside_brackets=True)
# Build the new lines.
- for result, leaves in (
- (head, head_leaves), (body, body_leaves), (tail, tail_leaves)
- ):
+ for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
for leaf in leaves:
result.append(leaf, preformatted=True)
for comment_after in line.comments_after(leaf):
yield result
-def right_hand_split(line: Line, py36: bool = False) -> Iterator[Line]:
+def right_hand_split(
+ line: Line, py36: bool = False, omit: Collection[LeafID] = ()
+) -> 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)
head_leaves: List[Leaf] = []
current_leaves = tail_leaves
opening_bracket = None
+ closing_bracket = None
for leaf in reversed(line.leaves):
if current_leaves is body_leaves:
if leaf is opening_bracket:
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:
+ if leaf.type in CLOSING_BRACKETS and id(leaf) not in omit:
opening_bracket = leaf.opening_bracket
+ closing_bracket = leaf
current_leaves = body_leaves
tail_leaves.reverse()
body_leaves.reverse()
# Since body is a new indent level, remove spurious leading whitespace.
if body_leaves:
normalize_prefix(body_leaves[0], inside_brackets=True)
+ elif not head_leaves:
+ # No `head` and no `body` means the split failed. `tail` has all content.
+ raise CannotSplit("No brackets found")
+
# Build the new lines.
- for result, leaves in (
- (head, head_leaves), (body, body_leaves), (tail, tail_leaves)
- ):
+ for result, leaves in (head, head_leaves), (body, body_leaves), (tail, tail_leaves):
for leaf in leaves:
result.append(leaf, preformatted=True)
for comment_after in line.comments_after(leaf):
result.append(comment_after, preformatted=True)
bracket_split_succeeded_or_raise(head, body, tail)
+ assert opening_bracket and closing_bracket
+ if (
+ opening_bracket.type == token.LPAR
+ and not opening_bracket.value
+ and closing_bracket.type == token.RPAR
+ and not closing_bracket.value
+ ):
+ # These parens were optional. If there aren't any delimiters or standalone
+ # comments in the body, they were unnecessary and another split without
+ # them should be attempted.
+ if not (
+ body.bracket_tracker.delimiters or line.contains_standalone_comments(0)
+ ):
+ omit = {id(closing_bracket), *omit}
+ yield from right_hand_split(line, py36=py36, omit=omit)
+ return
+
+ ensure_visible(opening_bracket)
+ ensure_visible(closing_bracket)
for result in (head, body, tail):
if result:
yield result
current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
if current_line:
if (
- delimiter_priority == COMMA_PRIORITY
+ trailing_comma_safe
+ and delimiter_priority == COMMA_PRIORITY
and current_line.leaves[-1].type != token.COMMA
- and trailing_comma_safe
+ and current_line.leaves[-1].type != STANDALONE_COMMENT
):
current_line.append(Leaf(token.COMMA, ","))
yield current_line
@dont_increase_indentation
def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]:
"""Split standalone comments from the rest of the line."""
- for leaf in line.leaves:
- if leaf.type == STANDALONE_COMMENT:
- if leaf.bracket_depth == 0:
- break
-
- else:
+ if not line.contains_standalone_comments(0):
raise CannotSplit("Line does not have any standalone comments")
current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets)
if first_quote_pos == -1:
return # There's an internal error
+ prefix = leaf.value[:first_quote_pos]
body = leaf.value[first_quote_pos + len(orig_quote):-len(orig_quote)]
- new_body = body.replace(f"\\{orig_quote}", orig_quote).replace(
- new_quote, f"\\{new_quote}"
- )
+ unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}")
+ escaped_orig_quote = re.compile(rf"\\(\\\\)*{orig_quote}")
+ if "r" in prefix.casefold():
+ if unescaped_new_quote.search(body):
+ # There's at least one unescaped new_quote in this raw string
+ # so converting is impossible
+ return
+
+ # Do not introduce or remove backslashes in raw strings
+ new_body = body
+ else:
+ new_body = escaped_orig_quote.sub(rf"\1{orig_quote}", body)
+ new_body = unescaped_new_quote.sub(rf"\1\\{new_quote}", new_body)
+ # Add escapes again for consecutive occurences of new_quote (sub
+ # doesn't match overlapping substrings).
+ new_body = unescaped_new_quote.sub(rf"\1\\{new_quote}", new_body)
if new_quote == '"""' and new_body[-1] == '"':
# edge case:
new_body = new_body[:-1] + '\\"'
if new_escape_count == orig_escape_count and orig_quote == '"':
return # Prefer double quotes
- prefix = leaf.value[:first_quote_pos]
leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}"
+def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None:
+ """Make existing optional parentheses invisible or create new ones.
+
+ Standardizes on visible parentheses for single-element tuples, and keeps
+ existing visible parentheses for other tuples and generator expressions.
+ """
+ check_lpar = False
+ for child in list(node.children):
+ if check_lpar:
+ if child.type == syms.atom:
+ if not (
+ is_empty_tuple(child)
+ or is_one_tuple(child)
+ or max_delimiter_priority_in_atom(child) >= COMMA_PRIORITY
+ ):
+ first = child.children[0]
+ last = child.children[-1]
+ if first.type == token.LPAR and last.type == token.RPAR:
+ # make parentheses invisible
+ first.value = "" # type: ignore
+ last.value = "" # type: ignore
+ elif is_one_tuple(child):
+ # wrap child in visible parentheses
+ lpar = Leaf(token.LPAR, "(")
+ rpar = Leaf(token.RPAR, ")")
+ index = child.remove() or 0
+ node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+ else:
+ # wrap child in invisible parentheses
+ lpar = Leaf(token.LPAR, "")
+ rpar = Leaf(token.RPAR, "")
+ index = child.remove() or 0
+ node.insert_child(index, Node(syms.atom, [lpar, child, rpar]))
+
+ check_lpar = isinstance(child, Leaf) and child.value in parens_after
+
+
+def is_empty_tuple(node: LN) -> bool:
+ """Return True if `node` holds an empty tuple."""
+ return (
+ node.type == syms.atom
+ and len(node.children) == 2
+ and node.children[0].type == token.LPAR
+ and node.children[1].type == token.RPAR
+ )
+
+
+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 len(node.children) != 3:
+ return False
+
+ lpar, gexp, rpar = node.children
+ if not (
+ lpar.type == token.LPAR
+ and gexp.type == syms.testlist_gexp
+ and rpar.type == token.RPAR
+ ):
+ return False
+
+ return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA
+
+ return (
+ node.type in IMPLICIT_TUPLE
+ and len(node.children) == 2
+ and node.children[1].type == token.COMMA
+ )
+
+
+def max_delimiter_priority_in_atom(node: LN) -> int:
+ if node.type != syms.atom:
+ return 0
+
+ first = node.children[0]
+ last = node.children[-1]
+ if not (first.type == token.LPAR and last.type == token.RPAR):
+ return 0
+
+ bt = BracketTracker()
+ for c in node.children[1:-1]:
+ if isinstance(c, Leaf):
+ bt.mark(c)
+ else:
+ for leaf in c.leaves():
+ bt.mark(leaf)
+ try:
+ return bt.max_delimiter_priority()
+
+ except ValueError:
+ return 0
+
+
+def ensure_visible(leaf: Leaf) -> None:
+ """Make sure parentheses are visible.
+
+ They could be invisible as part of some statements (see
+ :func:`normalize_invible_parens` and :func:`visit_import_from`).
+ """
+ if leaf.type == token.LPAR:
+ leaf.value = "("
+ elif leaf.type == token.RPAR:
+ leaf.value = ")"
+
+
def is_python36(node: Node) -> bool:
"""Return True if the current file is using Python 3.6+ features.
class Report:
"""Provides a reformatting counter. Can be rendered with `str(report)`."""
check: bool = False
+ quiet: bool = False
change_count: int = 0
same_count: int = 0
failure_count: int = 0
"""Increment the counter for successful reformatting. Write out a message."""
if changed:
reformatted = "would reformat" if self.check else "reformatted"
- out(f"{reformatted} {src}")
+ if not self.quiet:
+ out(f"{reformatted} {src}")
self.change_count += 1
else:
- out(f"{src} already well formatted, good job.", bold=False)
+ if not self.quiet:
+ out(f"{src} already well formatted, good job.", bold=False)
self.same_count += 1
def failed(self, src: Path, message: str) -> None:
) as f:
for lines in output:
f.write(lines)
- f.write("\n")
+ if lines and lines[-1] != "\n":
+ f.write("\n")
return f.name
projects / etc / vim.git / blobdiff