X-Git-Url: https://git.madduck.net/etc/vim.git/blobdiff_plain/5bc62a4a82a4e59ad3bb8e9078b9992396ebfd2f..95fd5cb6485c21cf0294899a4b2cb99c453fa01d:/black.py diff --git a/black.py b/black.py index 74cf650..953d532 100644 --- a/black.py +++ b/black.py @@ -1,20 +1,23 @@ +import ast import asyncio -from asyncio.base_events import BaseEventLoop from concurrent.futures import Executor, ProcessPoolExecutor +from contextlib import contextmanager from datetime import datetime -from enum import Enum, Flag +from enum import Enum from functools import lru_cache, partial, wraps import io -import keyword +import itertools import logging -from multiprocessing import Manager +from multiprocessing import Manager, freeze_support import os from pathlib import Path import pickle -import re +import regex as re import signal import sys +import tempfile import tokenize +import traceback from typing import ( Any, Callable, @@ -36,22 +39,23 @@ from typing import ( ) from appdirs import user_cache_dir -from attr import dataclass, Factory +from attr import dataclass, evolve, Factory 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 import pygram, pytree from blib2to3.pgen2 import driver, token +from blib2to3.pgen2.grammar import Grammar from blib2to3.pgen2.parse import ParseError +from _black_version import version as __version__ -__version__ = "18.9b0" DEFAULT_LINE_LENGTH = 88 -DEFAULT_EXCLUDES = ( - r"/(\.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__)) @@ -66,7 +70,7 @@ LeafID = int Priority = int Index = int LN = Union[Leaf, Node] -SplitFunc = Callable[["Line", bool], Iterator["Line"]] +SplitFunc = Callable[["Line", Collection["Feature"]], Iterator["Line"]] Timestamp = float FileSize = int CacheInfo = Tuple[Timestamp, FileSize] @@ -110,32 +114,101 @@ class Changed(Enum): YES = 2 -class FileMode(Flag): - AUTO_DETECT = 0 - PYTHON36 = 1 - PYI = 2 - NO_STRING_NORMALIZATION = 4 - NO_NUMERIC_UNDERSCORE_NORMALIZATION = 8 +class TargetVersion(Enum): + PY27 = 2 + PY33 = 3 + PY34 = 4 + PY35 = 5 + PY36 = 6 + PY37 = 7 + PY38 = 8 + + def is_python2(self) -> bool: + return self is TargetVersion.PY27 + + +PY36_VERSIONS = {TargetVersion.PY36, TargetVersion.PY37, TargetVersion.PY38} + + +class Feature(Enum): + # All string literals are unicode + UNICODE_LITERALS = 1 + F_STRINGS = 2 + NUMERIC_UNDERSCORES = 3 + TRAILING_COMMA_IN_CALL = 4 + TRAILING_COMMA_IN_DEF = 5 + # The following two feature-flags are mutually exclusive, and exactly one should be + # set for every version of python. + ASYNC_IDENTIFIERS = 6 + ASYNC_KEYWORDS = 7 + ASSIGNMENT_EXPRESSIONS = 8 + POS_ONLY_ARGUMENTS = 9 + + +VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = { + TargetVersion.PY27: {Feature.ASYNC_IDENTIFIERS}, + TargetVersion.PY33: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS}, + TargetVersion.PY34: {Feature.UNICODE_LITERALS, Feature.ASYNC_IDENTIFIERS}, + TargetVersion.PY35: { + Feature.UNICODE_LITERALS, + Feature.TRAILING_COMMA_IN_CALL, + Feature.ASYNC_IDENTIFIERS, + }, + TargetVersion.PY36: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA_IN_CALL, + Feature.TRAILING_COMMA_IN_DEF, + Feature.ASYNC_IDENTIFIERS, + }, + TargetVersion.PY37: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA_IN_CALL, + Feature.TRAILING_COMMA_IN_DEF, + Feature.ASYNC_KEYWORDS, + }, + TargetVersion.PY38: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA_IN_CALL, + Feature.TRAILING_COMMA_IN_DEF, + Feature.ASYNC_KEYWORDS, + Feature.ASSIGNMENT_EXPRESSIONS, + Feature.POS_ONLY_ARGUMENTS, + }, +} - @classmethod - def from_configuration( - cls, - *, - py36: bool, - pyi: bool, - skip_string_normalization: bool, - skip_numeric_underscore_normalization: bool, - ) -> "FileMode": - mode = cls.AUTO_DETECT - if py36: - mode |= cls.PYTHON36 - if pyi: - mode |= cls.PYI - if skip_string_normalization: - mode |= cls.NO_STRING_NORMALIZATION - if skip_numeric_underscore_normalization: - mode |= cls.NO_NUMERIC_UNDERSCORE_NORMALIZATION - return mode + +@dataclass +class FileMode: + target_versions: Set[TargetVersion] = Factory(set) + line_length: int = DEFAULT_LINE_LENGTH + string_normalization: bool = True + is_pyi: bool = False + + def get_cache_key(self) -> str: + if self.target_versions: + version_str = ",".join( + str(version.value) + for version in sorted(self.target_versions, key=lambda v: v.value) + ) + else: + version_str = "-" + parts = [ + version_str, + str(self.line_length), + str(int(self.string_normalization)), + str(int(self.is_pyi)), + ] + return ".".join(parts) + + +def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool: + return all(feature in VERSION_TO_FEATURES[version] for version in target_versions) def read_pyproject_toml( @@ -159,7 +232,9 @@ def read_pyproject_toml( pyproject_toml = toml.load(value) config = pyproject_toml.get("tool", {}).get("black", {}) except (toml.TomlDecodeError, OSError) as e: - raise click.BadOptionUsage(f"Error reading configuration file: {e}", ctx) + raise click.FileError( + filename=value, hint=f"Error reading configuration file: {e}" + ) if not config: return None @@ -173,6 +248,7 @@ def read_pyproject_toml( @click.command(context_settings=dict(help_option_names=["-h", "--help"])) +@click.option("-c", "--code", type=str, help="Format the code passed in as a string.") @click.option( "-l", "--line-length", @@ -181,13 +257,25 @@ def read_pyproject_toml( help="How many characters per line to allow.", show_default=True, ) +@click.option( + "-t", + "--target-version", + type=click.Choice([v.name.lower() for v in TargetVersion]), + callback=lambda c, p, v: [TargetVersion[val.upper()] for val in v], + multiple=True, + help=( + "Python versions that should be supported by Black's output. [default: " + "per-file auto-detection]" + ), +) @click.option( "--py36", is_flag=True, help=( "Allow using Python 3.6-only syntax on all input files. This will put " "trailing commas in function signatures and calls also after *args and " - "**kwargs. [default: per-file auto-detection]" + "**kwargs. Deprecated; use --target-version instead. " + "[default: per-file auto-detection]" ), ) @click.option( @@ -204,12 +292,6 @@ def read_pyproject_toml( is_flag=True, help="Don't normalize string quotes or prefixes.", ) -@click.option( - "-N", - "--skip-numeric-underscore-normalization", - is_flag=True, - help="Don't normalize underscores in numeric literals.", -) @click.option( "--check", is_flag=True, @@ -259,7 +341,7 @@ def read_pyproject_toml( "--quiet", is_flag=True, help=( - "Don't emit non-error messages to stderr. Errors are still emitted, " + "Don't emit non-error messages to stderr. Errors are still emitted; " "silence those with 2>/dev/null." ), ) @@ -293,14 +375,15 @@ def read_pyproject_toml( @click.pass_context def main( ctx: click.Context, + code: Optional[str], line_length: int, + target_version: List[TargetVersion], check: bool, diff: bool, fast: bool, pyi: bool, py36: bool, skip_string_normalization: bool, - skip_numeric_underscore_normalization: bool, quiet: bool, verbose: bool, include: str, @@ -310,14 +393,32 @@ def main( ) -> None: """The uncompromising code formatter.""" write_back = WriteBack.from_configuration(check=check, diff=diff) - mode = FileMode.from_configuration( - py36=py36, - pyi=pyi, - skip_string_normalization=skip_string_normalization, - skip_numeric_underscore_normalization=skip_numeric_underscore_normalization, + if target_version: + if py36: + err(f"Cannot use both --target-version and --py36") + ctx.exit(2) + else: + versions = set(target_version) + elif py36: + err( + "--py36 is deprecated and will be removed in a future version. " + "Use --target-version py36 instead." + ) + versions = PY36_VERSIONS + else: + # We'll autodetect later. + versions = set() + mode = FileMode( + target_versions=versions, + line_length=line_length, + is_pyi=pyi, + string_normalization=not skip_string_normalization, ) if config and verbose: out(f"Using configuration from {config}.", bold=False, fg="blue") + if code is not None: + print(format_str(code, mode=mode)) + ctx.exit(0) try: include_regex = re_compile_maybe_verbose(include) except re.error: @@ -331,11 +432,14 @@ def main( report = Report(check=check, quiet=quiet, verbose=verbose) root = find_project_root(src) sources: Set[Path] = set() + path_empty(src, quiet, verbose, ctx) for s in src: 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 @@ -344,108 +448,121 @@ def main( err(f"invalid path: {s}") if len(sources) == 0: if verbose or not quiet: - out("No paths given. Nothing to do 😴") + out("No Python files are present to be formatted. Nothing to do 😴") ctx.exit(0) if len(sources) == 1: reformat_one( src=sources.pop(), - line_length=line_length, fast=fast, write_back=write_back, mode=mode, report=report, ) else: - loop = asyncio.get_event_loop() - executor = ProcessPoolExecutor(max_workers=os.cpu_count()) - try: - loop.run_until_complete( - schedule_formatting( - sources=sources, - line_length=line_length, - fast=fast, - write_back=write_back, - mode=mode, - report=report, - loop=loop, - executor=executor, - ) - ) - finally: - shutdown(loop) + reformat_many( + sources=sources, fast=fast, write_back=write_back, mode=mode, report=report + ) + if verbose or not quiet: - bang = "💥 💔 💥" if report.return_code else "✨ 🍰 ✨" - out(f"All done! {bang}") + out("Oh no! 💥 💔 💥" if report.return_code else "All done! ✨ 🍰 ✨") click.secho(str(report), err=True) ctx.exit(report.return_code) +def path_empty(src: Tuple[str], quiet: bool, verbose: bool, ctx: click.Context) -> None: + """ + Exit if there is no `src` provided for formatting + """ + if not src: + if verbose or not quiet: + out("No Path provided. Nothing to do 😴") + ctx.exit(0) + + def reformat_one( - src: Path, - line_length: int, - fast: bool, - write_back: WriteBack, - mode: FileMode, - report: "Report", + src: Path, fast: bool, write_back: WriteBack, mode: FileMode, report: "Report" ) -> None: """Reformat a single file under `src` without spawning child processes. - If `quiet` is True, non-error messages are not output. `line_length`, - `write_back`, `fast` and `pyi` options are passed to + `fast`, `write_back`, and `mode` options are passed to :func:`format_file_in_place` or :func:`format_stdin_to_stdout`. """ try: changed = Changed.NO if not src.is_file() and str(src) == "-": - if format_stdin_to_stdout( - line_length=line_length, fast=fast, write_back=write_back, mode=mode - ): + if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode): changed = Changed.YES else: cache: Cache = {} if write_back != WriteBack.DIFF: - cache = read_cache(line_length, mode) + cache = read_cache(mode) res_src = src.resolve() if res_src in cache and cache[res_src] == get_cache_info(res_src): changed = Changed.CACHED if changed is not Changed.CACHED and format_file_in_place( - src, - line_length=line_length, - fast=fast, - write_back=write_back, - mode=mode, + src, fast=fast, write_back=write_back, mode=mode ): changed = Changed.YES if (write_back is WriteBack.YES and changed is not Changed.CACHED) or ( write_back is WriteBack.CHECK and changed is Changed.NO ): - write_cache(cache, [src], line_length, mode) + write_cache(cache, [src], mode) report.done(src, changed) except Exception as exc: report.failed(src, str(exc)) +def reformat_many( + sources: Set[Path], + fast: bool, + write_back: WriteBack, + mode: FileMode, + report: "Report", +) -> None: + """Reformat multiple files using a ProcessPoolExecutor.""" + loop = asyncio.get_event_loop() + worker_count = os.cpu_count() + if sys.platform == "win32": + # Work around https://bugs.python.org/issue26903 + worker_count = min(worker_count, 61) + executor = ProcessPoolExecutor(max_workers=worker_count) + try: + loop.run_until_complete( + schedule_formatting( + sources=sources, + fast=fast, + write_back=write_back, + mode=mode, + report=report, + loop=loop, + executor=executor, + ) + ) + finally: + shutdown(loop) + executor.shutdown() + + async def schedule_formatting( sources: Set[Path], - line_length: int, fast: bool, write_back: WriteBack, mode: FileMode, report: "Report", - loop: BaseEventLoop, + loop: asyncio.AbstractEventLoop, executor: Executor, ) -> None: """Run formatting of `sources` in parallel using the provided `executor`. (Use ProcessPoolExecutors for actual parallelism.) - `line_length`, `write_back`, `fast`, and `pyi` options are passed to + `write_back`, `fast`, and `mode` options are passed to :func:`format_file_in_place`. """ cache: Cache = {} if write_back != WriteBack.DIFF: - cache = read_cache(line_length, mode) + cache = read_cache(mode) sources, cached = filter_cached(cache, sources) for src in sorted(cached): report.done(src, Changed.CACHED) @@ -461,19 +578,14 @@ async def schedule_formatting( manager = Manager() lock = manager.Lock() tasks = { - loop.run_in_executor( - executor, - format_file_in_place, - src, - line_length, - fast, - write_back, - mode, - lock, + asyncio.ensure_future( + loop.run_in_executor( + executor, format_file_in_place, src, fast, mode, write_back, lock + ) ): src for src in sorted(sources) } - pending: Iterable[asyncio.Task] = tasks.keys() + pending: Iterable[asyncio.Future] = tasks.keys() try: loop.add_signal_handler(signal.SIGINT, cancel, pending) loop.add_signal_handler(signal.SIGTERM, cancel, pending) @@ -500,33 +612,30 @@ async def schedule_formatting( if cancelled: await asyncio.gather(*cancelled, loop=loop, return_exceptions=True) if sources_to_cache: - write_cache(cache, sources_to_cache, line_length, mode) + write_cache(cache, sources_to_cache, mode) def format_file_in_place( src: Path, - line_length: int, fast: bool, + mode: FileMode, write_back: WriteBack = WriteBack.NO, - mode: FileMode = FileMode.AUTO_DETECT, lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy ) -> bool: """Format file under `src` path. Return True if changed. If `write_back` is DIFF, write a diff to stdout. If it is YES, write reformatted code to the file. - `line_length` and `fast` options are passed to :func:`format_file_contents`. + `mode` and `fast` options are passed to :func:`format_file_contents`. """ if src.suffix == ".pyi": - mode |= FileMode.PYI + mode = evolve(mode, is_pyi=True) then = datetime.utcfromtimestamp(src.stat().st_mtime) with open(src, "rb") as buf: src_contents, encoding, newline = decode_bytes(buf.read()) try: - dst_contents = format_file_contents( - src_contents, line_length=line_length, fast=fast, mode=mode - ) + dst_contents = format_file_contents(src_contents, fast=fast, mode=mode) except NothingChanged: return False @@ -538,9 +647,8 @@ def format_file_in_place( src_name = f"{src}\t{then} +0000" dst_name = f"{src}\t{now} +0000" diff_contents = diff(src_contents, dst_contents, src_name, dst_name) - if lock: - lock.acquire() - try: + + with lock or nullcontext(): f = io.TextIOWrapper( sys.stdout.buffer, encoding=encoding, @@ -549,30 +657,24 @@ def format_file_in_place( ) f.write(diff_contents) f.detach() - finally: - if lock: - lock.release() + return True def format_stdin_to_stdout( - line_length: int, - fast: bool, - write_back: WriteBack = WriteBack.NO, - mode: FileMode = FileMode.AUTO_DETECT, + fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: FileMode ) -> bool: """Format file on stdin. Return True if changed. If `write_back` is YES, write reformatted code back to stdout. If it is DIFF, - write a diff to stdout. - `line_length`, `fast`, `is_pyi`, and `force_py36` arguments are passed to + write a diff to stdout. The `mode` argument is passed to :func:`format_file_contents`. """ then = datetime.utcnow() src, encoding, newline = decode_bytes(sys.stdin.buffer.read()) dst = src try: - dst = format_file_contents(src, line_length=line_length, fast=fast, mode=mode) + dst = format_file_contents(src, fast=fast, mode=mode) return True except NothingChanged: @@ -593,64 +695,66 @@ def format_stdin_to_stdout( def format_file_contents( - src_contents: str, - *, - line_length: int, - fast: bool, - mode: FileMode = FileMode.AUTO_DETECT, + src_contents: str, *, fast: bool, mode: FileMode ) -> FileContent: """Reformat contents a file and return new contents. If `fast` is False, additionally confirm that the reformatted code is valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it. - `line_length` is passed to :func:`format_str`. + `mode` is passed to :func:`format_str`. """ if src_contents.strip() == "": raise NothingChanged - dst_contents = format_str(src_contents, line_length=line_length, mode=mode) + dst_contents = format_str(src_contents, mode=mode) if src_contents == dst_contents: raise NothingChanged if not fast: assert_equivalent(src_contents, dst_contents) - assert_stable(src_contents, dst_contents, line_length=line_length, mode=mode) + assert_stable(src_contents, dst_contents, mode=mode) return dst_contents -def format_str( - src_contents: str, line_length: int, *, mode: FileMode = FileMode.AUTO_DETECT -) -> FileContent: +def format_str(src_contents: str, *, mode: FileMode) -> FileContent: """Reformat a string and return new contents. - `line_length` determines how many characters per line are allowed. + `mode` determines formatting options, such as how many characters per line are + allowed. """ - src_node = lib2to3_parse(src_contents.lstrip()) - dst_contents = "" + src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions) + dst_contents = [] future_imports = get_future_imports(src_node) - is_pyi = bool(mode & FileMode.PYI) - py36 = bool(mode & FileMode.PYTHON36) or is_python36(src_node) - normalize_strings = not bool(mode & FileMode.NO_STRING_NORMALIZATION) + if mode.target_versions: + versions = mode.target_versions + else: + versions = detect_target_versions(src_node) normalize_fmt_off(src_node) lines = LineGenerator( - remove_u_prefix=py36 or "unicode_literals" in future_imports, - is_pyi=is_pyi, - normalize_strings=normalize_strings, - allow_underscores=py36 - and not bool(mode & FileMode.NO_NUMERIC_UNDERSCORE_NORMALIZATION), + remove_u_prefix="unicode_literals" in future_imports + or supports_feature(versions, Feature.UNICODE_LITERALS), + is_pyi=mode.is_pyi, + normalize_strings=mode.string_normalization, ) - elt = EmptyLineTracker(is_pyi=is_pyi) + elt = EmptyLineTracker(is_pyi=mode.is_pyi) empty_line = Line() after = 0 + split_line_features = { + feature + for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF} + if supports_feature(versions, feature) + } for current_line in lines.visit(src_node): for _ in range(after): - dst_contents += str(empty_line) + dst_contents.append(str(empty_line)) before, after = elt.maybe_empty_lines(current_line) for _ in range(before): - dst_contents += str(empty_line) - for line in split_line(current_line, line_length=line_length, py36=py36): - dst_contents += str(line) - return dst_contents + dst_contents.append(str(empty_line)) + for line in split_line( + current_line, line_length=mode.line_length, features=split_line_features + ): + dst_contents.append(str(line)) + return "".join(dst_contents) def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]: @@ -670,18 +774,51 @@ def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]: return tiow.read(), encoding, newline -GRAMMARS = [ - pygram.python_grammar_no_print_statement_no_exec_statement, - pygram.python_grammar_no_print_statement, - pygram.python_grammar, -] +def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]: + if not target_versions: + # No target_version specified, so try all grammars. + return [ + # Python 3.7+ + pygram.python_grammar_no_print_statement_no_exec_statement_async_keywords, + # Python 3.0-3.6 + pygram.python_grammar_no_print_statement_no_exec_statement, + # Python 2.7 with future print_function import + pygram.python_grammar_no_print_statement, + # Python 2.7 + pygram.python_grammar, + ] + 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 + pygram.python_grammar_no_print_statement, + # Python 2.7 + pygram.python_grammar, + ] -def lib2to3_parse(src_txt: str) -> Node: + # 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: """Given a string with source, return the lib2to3 Node.""" if src_txt[-1:] != "\n": src_txt += "\n" - for grammar in GRAMMARS: + + for grammar in get_grammars(set(target_versions)): drv = driver.Driver(grammar, pytree.convert) try: result = drv.parse_string(src_txt, True) @@ -774,9 +911,7 @@ class DebugVisitor(Visitor[T]): list(v.visit(code)) -KEYWORDS = set(keyword.kwlist) WHITESPACE = {token.DEDENT, token.INDENT, token.NEWLINE} -FLOW_CONTROL = {"return", "raise", "break", "continue"} STATEMENT = { syms.if_stmt, syms.while_stmt, @@ -815,6 +950,7 @@ MATH_OPERATORS = { token.DOUBLESTAR, } STARS = {token.STAR, token.DOUBLESTAR} +VARARGS_SPECIALS = STARS | {token.SLASH} VARARGS_PARENTS = { syms.arglist, syms.argument, # double star in arglist @@ -942,7 +1078,7 @@ class BracketTracker: """Return True if there is an yet unmatched open bracket on the line.""" return bool(self.bracket_match) - def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int: + def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> Priority: """Return the highest priority of a delimiter found on the line. Values are consistent with what `is_split_*_delimiter()` return. @@ -950,7 +1086,7 @@ class BracketTracker: """ return max(v for k, v in self.delimiters.items() if k not in exclude) - def delimiter_count_with_priority(self, priority: int = 0) -> int: + def delimiter_count_with_priority(self, priority: Priority = 0) -> int: """Return the number of delimiters with the given `priority`. If no `priority` is passed, defaults to max priority on the line. @@ -1025,7 +1161,7 @@ class Line: depth: int = 0 leaves: List[Leaf] = Factory(list) - comments: List[Tuple[Index, Leaf]] = Factory(list) + comments: Dict[LeafID, List[Leaf]] = Factory(dict) # keys ordered like `leaves` bracket_tracker: BracketTracker = Factory(BracketTracker) inside_brackets: bool = False should_explode: bool = False @@ -1106,6 +1242,61 @@ class Line: 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.)""" @@ -1156,6 +1347,69 @@ class Line: if leaf.type == STANDALONE_COMMENT: if leaf.bracket_depth <= depth_limit: return True + return False + + def contains_uncollapsable_type_comments(self) -> bool: + ignored_ids = set() + try: + last_leaf = self.leaves[-1] + ignored_ids.add(id(last_leaf)) + if last_leaf.type == token.COMMA or ( + last_leaf.type == token.RPAR and not last_leaf.value + ): + # When trailing commas or optional parens are inserted by Black for + # consistency, comments after the previous last element are not moved + # (they don't have to, rendering will still be correct). So we ignore + # trailing commas and invisible. + last_leaf = self.leaves[-2] + ignored_ids.add(id(last_leaf)) + except IndexError: + return False + + # A type comment is uncollapsable if it is attached to a leaf + # that isn't at the end of the line (since that could cause it + # to get associated to a different argument) or if there are + # comments before it (since that could cause it to get hidden + # behind a comment. + comment_seen = False + for leaf_id, comments in self.comments.items(): + for comment in comments: + if is_type_comment(comment): + if leaf_id not in ignored_ids or comment_seen: + return True + + comment_seen = True + + return False + + def contains_unsplittable_type_ignore(self) -> bool: + if not self.leaves: + return False + + # If a 'type: ignore' is attached to the end of a line, we + # can't split the line, because we can't know which of the + # subexpressions the ignore was meant to apply to. + # + # We only want this to apply to actual physical lines from the + # original source, though: we don't want the presence of a + # 'type: ignore' at the end of a multiline expression to + # justify pushing it all onto one line. Thus we + # (unfortunately) need to check the actual source lines and + # only report an unsplittable 'type: ignore' if this line was + # one line in the original code. + + # 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 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 return False @@ -1168,57 +1422,39 @@ class Line: 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 == syms.arglist: - 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.""" @@ -1232,44 +1468,41 @@ class Line: if comment.type != token.COMMENT: return False - after = len(self.leaves) - 1 - if after == -1: + if not self.leaves: comment.type = STANDALONE_COMMENT comment.prefix = "" return False - else: - self.comments.append((after, comment)) - return True - - def comments_after(self, leaf: Leaf, _index: int = -1) -> Iterator[Leaf]: - """Generate comments that should appear directly after `leaf`. - - Provide a non-negative leaf `_index` to speed up the function. - """ - if not self.comments: - return - - if _index == -1: - for _index, _leaf in enumerate(self.leaves): - if leaf is _leaf: - break - - else: - return + last_leaf = self.leaves[-1] + if ( + last_leaf.type == token.RPAR + and not last_leaf.value + and last_leaf.parent + and len(list(last_leaf.parent.leaves())) <= 3 + and not is_type_comment(comment) + ): + # Comments on an optional parens wrapping a single leaf should belong to + # the wrapped node except if it's a type comment. Pinning the comment like + # this avoids unstable formatting caused by comment migration. + if len(self.leaves) < 2: + comment.type = STANDALONE_COMMENT + comment.prefix = "" + return False + last_leaf = self.leaves[-2] + self.comments.setdefault(id(last_leaf), []).append(comment) + return True - for index, comment_after in self.comments: - if _index == index: - yield comment_after + def comments_after(self, leaf: Leaf) -> List[Leaf]: + """Generate comments that should appear directly after `leaf`.""" + return self.comments.get(id(leaf), []) def remove_trailing_comma(self) -> None: """Remove the trailing comma and moves the comments attached to it.""" - comma_index = len(self.leaves) - 1 - for i in range(len(self.comments)): - comment_index, comment = self.comments[i] - if comment_index == comma_index: - self.comments[i] = (comma_index - 1, comment) - self.leaves.pop() + trailing_comma = self.leaves.pop() + trailing_comma_comments = self.comments.pop(id(trailing_comma), []) + self.comments.setdefault(id(self.leaves[-1]), []).extend( + trailing_comma_comments + ) def is_complex_subscript(self, leaf: Leaf) -> bool: """Return True iff `leaf` is part of a slice with non-trivial exprs.""" @@ -1300,7 +1533,7 @@ class Line: res = f"{first.prefix}{indent}{first.value}" for leaf in leaves: res += str(leaf) - for _, comment in self.comments: + for comment in itertools.chain.from_iterable(self.comments.values()): res += str(comment) return res + "\n" @@ -1331,7 +1564,13 @@ class EmptyLineTracker: lines (two on module-level). """ before, after = self._maybe_empty_lines(current_line) - before -= self.previous_after + before = ( + # Black should not insert empty lines at the beginning + # of the file + 0 + if self.previous_line is None + else before - self.previous_after + ) self.previous_after = after self.previous_line = current_line return before, after @@ -1432,7 +1671,6 @@ class LineGenerator(Visitor[Line]): normalize_strings: bool = True current_line: Line = Factory(Line) remove_u_prefix: bool = False - allow_underscores: bool = False def line(self, indent: int = 0) -> Iterator[Line]: """Generate a line. @@ -1475,7 +1713,7 @@ class LineGenerator(Visitor[Line]): normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix) normalize_string_quotes(node) if node.type == token.NUMBER: - normalize_numeric_literal(node, self.allow_underscores) + normalize_numeric_literal(node) if node.type not in WHITESPACE: self.current_line.append(node) yield from super().visit_default(node) @@ -1579,6 +1817,23 @@ class LineGenerator(Visitor[Line]): 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 @@ -1599,6 +1854,7 @@ class LineGenerator(Visitor[Line]): 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_del_stmt = partial(v, keywords=Ø, parens={"del"}) self.visit_async_funcdef = self.visit_async_stmt self.visit_decorated = self.visit_decorators @@ -1611,7 +1867,7 @@ BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT} -def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa C901 +def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa: C901 """Return whitespace prefix if needed for the given `leaf`. `complex_subscript` signals whether the given leaf is part of a subscription @@ -1668,7 +1924,7 @@ def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa C901 # that, too. return prevp.prefix - elif prevp.type in STARS: + elif prevp.type in VARARGS_SPECIALS: if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS): return NO @@ -1758,7 +2014,7 @@ def whitespace(leaf: Leaf, *, complex_subscript: bool) -> str: # noqa C901 if not prevp or prevp.type == token.LPAR: return NO - elif prev.type in {token.EQUAL} | STARS: + elif prev.type in {token.EQUAL} | VARARGS_SPECIALS: return NO elif p.type == syms.decorator: @@ -1892,7 +2148,7 @@ def container_of(leaf: Leaf) -> LN: return container -def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int: +def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority: """Return the priority of the `leaf` delimiter, given a line break after it. The delimiter priorities returned here are from those delimiters that would @@ -1906,7 +2162,7 @@ def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int return 0 -def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int: +def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> Priority: """Return the priority of the `leaf` delimiter, given a line break before it. The delimiter priorities returned here are from those delimiters that would @@ -2056,15 +2312,21 @@ def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]: consumed = 0 nlines = 0 + ignored_lines = 0 for index, line in enumerate(prefix.split("\n")): consumed += len(line) + 1 # adding the length of the split '\n' line = line.lstrip() if not line: nlines += 1 if not line.startswith("#"): + # Escaped newlines outside of a comment are not really newlines at + # all. We treat a single-line comment following an escaped newline + # as a simple trailing comment. + if line.endswith("\\"): + ignored_lines += 1 continue - if index == 0 and not is_endmarker: + if index == ignored_lines and not is_endmarker: comment_type = token.COMMENT # simple trailing comment else: comment_type = STANDALONE_COMMENT @@ -2098,7 +2360,10 @@ def make_comment(content: str) -> str: def split_line( - line: Line, line_length: int, inner: bool = False, py36: bool = False + line: Line, + line_length: int, + inner: bool = False, + features: Collection[Feature] = (), ) -> Iterator[Line]: """Split a `line` into potentially many lines. @@ -2107,16 +2372,22 @@ def split_line( 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. + `features` are syntactical features that may be used in the output. """ if line.is_comment: yield line return line_str = str(line).strip("\n") - if not line.should_explode and is_line_short_enough( - line, line_length=line_length, line_str=line_str + + 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() + ) ): yield line return @@ -2126,9 +2397,9 @@ def split_line( split_funcs = [left_hand_split] else: - def rhs(line: Line, py36: bool = False) -> Iterator[Line]: + def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]: for omit in generate_trailers_to_omit(line, line_length): - lines = list(right_hand_split(line, line_length, py36, omit=omit)) + lines = list(right_hand_split(line, line_length, features, omit=omit)) if is_line_short_enough(lines[0], line_length=line_length): yield from lines return @@ -2136,7 +2407,9 @@ def split_line( # 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) + # 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] @@ -2148,12 +2421,14 @@ def split_line( # split altogether. result: List[Line] = [] try: - for l in split_func(line, py36): + for l in split_func(line, features): 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, py36=py36) + split_line( + l, line_length=line_length, inner=True, features=features + ) ) except CannotSplit: continue @@ -2166,7 +2441,7 @@ def split_line( yield line -def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]: +def left_hand_split(line: Line, features: Collection[Feature] = ()) -> 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. @@ -2203,7 +2478,10 @@ def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]: def right_hand_split( - line: Line, line_length: int, py36: bool = False, omit: Collection[LeafID] = () + line: Line, + line_length: int, + features: Collection[Feature] = (), + omit: Collection[LeafID] = (), ) -> Iterator[Line]: """Split line into many lines, starting with the last matching bracket pair. @@ -2261,7 +2539,7 @@ def right_hand_split( ): omit = {id(closing_bracket), *omit} try: - yield from right_hand_split(line, line_length, py36=py36, omit=omit) + yield from right_hand_split(line, line_length, features=features, omit=omit) return except CannotSplit: @@ -2329,10 +2607,23 @@ def bracket_split_build_line( if leaves: # Since body is a new indent level, remove spurious leading whitespace. normalize_prefix(leaves[0], inside_brackets=True) - # Ensure a trailing comma when expected. - if original.is_import: - if leaves[-1].type != token.COMMA: - leaves.append(Leaf(token.COMMA, ",")) + # Ensure a trailing comma for imports and standalone function arguments, but + # 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: + leaves.insert(i + 1, Leaf(token.COMMA, ",")) + break # Populate the line for leaf in leaves: result.append(leaf, preformatted=True) @@ -2350,8 +2641,8 @@ def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc: """ @wraps(split_func) - def split_wrapper(line: Line, py36: bool = False) -> Iterator[Line]: - for l in split_func(line, py36): + def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]: + for l in split_func(line, features): normalize_prefix(l.leaves[0], inside_brackets=True) yield l @@ -2359,11 +2650,11 @@ def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc: @dont_increase_indentation -def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]: +def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]: """Split according to delimiters of the highest priority. - If `py36` is True, the split will add trailing commas also in function - signatures that contain `*` and `**`. + If the appropriate Features are given, the split will add trailing commas + also in function signatures and calls that contain `*` and `**`. """ try: last_leaf = line.leaves[-1] @@ -2395,17 +2686,23 @@ def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]: current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets) current_line.append(leaf) - for index, leaf in enumerate(line.leaves): + for leaf in line.leaves: yield from append_to_line(leaf) - for comment_after in line.comments_after(leaf, index): + for comment_after in line.comments_after(leaf): yield from append_to_line(comment_after) lowest_depth = min(lowest_depth, leaf.bracket_depth) - if leaf.bracket_depth == lowest_depth and is_vararg( - leaf, within=VARARGS_PARENTS - ): - trailing_comma_safe = trailing_comma_safe and py36 + if leaf.bracket_depth == lowest_depth: + if is_vararg(leaf, within={syms.typedargslist}): + trailing_comma_safe = ( + trailing_comma_safe and Feature.TRAILING_COMMA_IN_DEF in features + ) + elif is_vararg(leaf, within={syms.arglist, syms.argument}): + trailing_comma_safe = ( + trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features + ) + leaf_priority = bt.delimiters.get(id(leaf)) if leaf_priority == delimiter_priority: yield current_line @@ -2423,7 +2720,9 @@ def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]: @dont_increase_indentation -def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]: +def standalone_comment_split( + line: Line, features: Collection[Feature] = () +) -> Iterator[Line]: """Split standalone comments from the rest of the line.""" if not line.contains_standalone_comments(0): raise CannotSplit("Line does not have any standalone comments") @@ -2441,10 +2740,10 @@ def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]: current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets) current_line.append(leaf) - for index, leaf in enumerate(line.leaves): + for leaf in line.leaves: yield from append_to_line(leaf) - for comment_after in line.comments_after(leaf, index): + for comment_after in line.comments_after(leaf): yield from append_to_line(comment_after) if current_line: @@ -2465,6 +2764,14 @@ def is_import(leaf: Leaf) -> bool: ) +def is_type_comment(leaf: Leaf, suffix: str = "") -> bool: + """Return True if the given leaf is a special comment. + Only returns true for type comments for now.""" + t = leaf.type + v = leaf.value + return t in {token.COMMENT, STANDALONE_COMMENT} and v.startswith("# type:" + suffix) + + def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None: """Leave existing extra newlines if not `inside_brackets`. Remove everything else. @@ -2547,7 +2854,15 @@ def normalize_string_quotes(leaf: Leaf) -> None: new_body = sub_twice(escaped_orig_quote, rf"\1\2{orig_quote}", new_body) new_body = sub_twice(unescaped_new_quote, rf"\1\\{new_quote}", new_body) if "f" in prefix.casefold(): - matches = re.findall(r"[^{]\{(.*?)\}[^}]", new_body) + matches = re.findall( + r""" + (?:[^{]|^)\{ # start of the string or a non-{ followed by a single { + ([^{].*?) # contents of the brackets except if begins with {{ + \}(?:[^}]|$) # A } followed by end of the string or a non-} + """, + new_body, + re.VERBOSE, + ) for m in matches: if "\\" in str(m): # Do not introduce backslashes in interpolated expressions @@ -2566,11 +2881,11 @@ def normalize_string_quotes(leaf: Leaf) -> None: leaf.value = f"{prefix}{new_quote}{new_body}{new_quote}" -def normalize_numeric_literal(leaf: Leaf, allow_underscores: bool) -> None: +def normalize_numeric_literal(leaf: Leaf) -> None: """Normalizes numeric (float, int, and complex) literals. All letters used in the representation are normalized to lowercase (except - in Python 2 long literals), and long number literals are split using underscores. + in Python 2 long literals). """ text = leaf.value.lower() if text.startswith(("0o", "0b")): @@ -2588,8 +2903,7 @@ def normalize_numeric_literal(leaf: Leaf, allow_underscores: bool) -> None: sign = "-" elif after.startswith("+"): after = after[1:] - before = format_float_or_int_string(before, allow_underscores) - after = format_int_string(after, allow_underscores) + before = format_float_or_int_string(before) text = f"{before}e{sign}{after}" elif text.endswith(("j", "l")): number = text[:-1] @@ -2597,56 +2911,25 @@ def normalize_numeric_literal(leaf: Leaf, allow_underscores: bool) -> None: # Capitalize in "2L" because "l" looks too similar to "1". if suffix == "l": suffix = "L" - text = f"{format_float_or_int_string(number, allow_underscores)}{suffix}" + text = f"{format_float_or_int_string(number)}{suffix}" else: - text = format_float_or_int_string(text, allow_underscores) + text = format_float_or_int_string(text) leaf.value = text -def format_float_or_int_string(text: str, allow_underscores: bool) -> str: +def format_float_or_int_string(text: str) -> str: """Formats a float string like "1.0".""" if "." not in text: - return format_int_string(text, allow_underscores) - - before, after = text.split(".") - before = format_int_string(before, allow_underscores) if before else "0" - if after: - after = format_int_string(after, allow_underscores, count_from_end=False) - else: - after = "0" - return f"{before}.{after}" - - -def format_int_string( - text: str, allow_underscores: bool, count_from_end: bool = True -) -> str: - """Normalizes underscores in a string to e.g. 1_000_000. - - Input must be a string of digits and optional underscores. - If count_from_end is False, we add underscores after groups of three digits - counting from the beginning instead of the end of the strings. This is used - for the fractional part of float literals. - """ - if not allow_underscores: - return text - - text = text.replace("_", "") - if len(text) <= 5: - # No underscores for numbers <= 5 digits long. return text - if count_from_end: - # Avoid removing leading zeros, which are important if we're formatting - # part of a number like "0.001". - return format(int("1" + text), "3_")[1:].lstrip("_") - else: - return "_".join(text[i : i + 3] for i in range(0, len(text), 3)) + before, after = text.split(".") + return f"{before or 0}.{after or 0}" def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: """Make existing optional parentheses invisible or create new ones. - `parens_after` is a set of string leaf values immeditely after which parens + `parens_after` is a set of string leaf values immediately after which parens should be put. Standardizes on visible parentheses for single-element tuples, and keeps @@ -2659,9 +2942,27 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: check_lpar = False for index, child in enumerate(list(node.children)): + # Add parentheses around long tuple unpacking in assignments. + if ( + index == 0 + and isinstance(child, Node) + and child.type == syms.testlist_star_expr + ): + check_lpar = True + if check_lpar: + if is_walrus_assignment(child): + continue if child.type == syms.atom: - if maybe_make_parens_invisible_in_atom(child): + # 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 @@ -2690,7 +2991,11 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: lpar = Leaf(token.LPAR, "") rpar = Leaf(token.RPAR, "") index = child.remove() or 0 - node.insert_child(index, Node(syms.atom, [lpar, child, rpar])) + prefix = child.prefix + child.prefix = "" + new_child = Node(syms.atom, [lpar, child, rpar]) + new_child.prefix = prefix + node.insert_child(index, new_child) check_lpar = isinstance(child, Leaf) and child.value in parens_after @@ -2772,8 +3077,10 @@ def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]: container = container.next_sibling -def maybe_make_parens_invisible_in_atom(node: LN) -> bool: +def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool: """If it's safe, make the parens in the atom `node` invisible, recursively. + Additionally, remove repeated, adjacent invisible parens from the atom `node` + as they are redundant. Returns whether the node should itself be wrapped in invisible parentheses. @@ -2782,7 +3089,7 @@ def maybe_make_parens_invisible_in_atom(node: LN) -> bool: node.type != syms.atom or is_empty_tuple(node) or is_one_tuple(node) - or is_yield(node) + or (is_yield(node) and parent.type != syms.expr_stmt) or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY ): return False @@ -2790,16 +3097,40 @@ def maybe_make_parens_invisible_in_atom(node: LN) -> bool: first = node.children[0] last = node.children[-1] if first.type == token.LPAR and last.type == token.RPAR: + middle = node.children[1] # make parentheses invisible first.value = "" # type: ignore last.value = "" # type: ignore - if len(node.children) > 1: - maybe_make_parens_invisible_in_atom(node.children[1]) + maybe_make_parens_invisible_in_atom(middle, parent=parent) + + if is_atom_with_invisible_parens(middle): + # Strip the invisible parens from `middle` by replacing + # it with the child in-between the invisible parens + middle.replace(middle.children[1]) + return False return True +def is_atom_with_invisible_parens(node: LN) -> bool: + """Given a `LN`, determines whether it's an atom `node` with invisible + parens. Useful in dedupe-ing and normalizing parens. + """ + if isinstance(node, Leaf) or node.type != syms.atom: + return False + + first, last = node.children[0], node.children[-1] + return ( + isinstance(first, Leaf) + and first.type == token.LPAR + and first.value == "" + and isinstance(last, Leaf) + and last.type == token.RPAR + and last.value == "" + ) + + def is_empty_tuple(node: LN) -> bool: """Return True if `node` holds an empty tuple.""" return ( @@ -2810,18 +3141,24 @@ def is_empty_tuple(node: LN) -> bool: ) +def unwrap_singleton_parenthesis(node: LN) -> Optional[LN]: + """Returns `wrapped` if `node` is of the shape ( wrapped ). + + 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 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 - ): + gexp = unwrap_singleton_parenthesis(node) + if gexp is None or gexp.type != syms.testlist_gexp: return False return len(gexp.children) == 2 and gexp.children[1].type == token.COMMA @@ -2833,6 +3170,12 @@ def is_one_tuple(node: LN) -> bool: ) +def is_walrus_assignment(node: LN) -> bool: + """Return True iff `node` is of the shape ( test := test )""" + inner = unwrap_singleton_parenthesis(node) + return inner is not None and inner.type == syms.namedexpr_test + + def is_yield(node: LN) -> bool: """Return True if `node` holds a `yield` or `yield from` expression.""" if node.type == syms.yield_expr: @@ -2862,7 +3205,7 @@ def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool: extended iterable unpacking (PEP 3132) and additional unpacking generalizations (PEP 448). """ - if leaf.type not in STARS or not leaf.parent: + if leaf.type not in VARARGS_SPECIALS or not leaf.parent: return False p = leaf.parent @@ -2912,7 +3255,7 @@ def is_stub_body(node: LN) -> bool: ) -def max_delimiter_priority_in_atom(node: LN) -> int: +def max_delimiter_priority_in_atom(node: LN) -> Priority: """Return maximum delimiter priority inside `node`. This is specific to atoms with contents contained in a pair of parentheses. @@ -2944,7 +3287,7 @@ 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`). + :func:`normalize_invisible_parens` and :func:`visit_import_from`). """ if leaf.type == token.LPAR: leaf.value = "(" @@ -2954,6 +3297,7 @@ def ensure_visible(leaf: Leaf) -> None: def should_explode(line: Line, opening_bracket: Leaf) -> bool: """Should `line` immediately be split with `delimiter_split()` after RHS?""" + if not ( opening_bracket.parent and opening_bracket.parent.type in {syms.atom, syms.import_from} @@ -2971,39 +3315,61 @@ def should_explode(line: Line, opening_bracket: Leaf) -> bool: return max_priority == COMMA_PRIORITY -def is_python36(node: Node) -> bool: - """Return True if the current file is using Python 3.6+ features. +def get_features_used(node: Node) -> Set[Feature]: + """Return a set of (relatively) new Python features used in this file. Currently looking for: - f-strings; - - underscores in numeric literals; and - - trailing commas after * or ** in function signatures and calls. + - underscores in numeric literals; + - trailing commas after * or ** in function signatures and calls; + - positional only arguments in function signatures and lambdas; """ + features: Set[Feature] = set() 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 + features.add(Feature.F_STRINGS) elif n.type == token.NUMBER: if "_" in n.value: # type: ignore - return True + features.add(Feature.NUMERIC_UNDERSCORES) + + elif n.type == token.SLASH: + if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}: + features.add(Feature.POS_ONLY_ARGUMENTS) + + elif n.type == token.COLONEQUAL: + features.add(Feature.ASSIGNMENT_EXPRESSIONS) elif ( n.type in {syms.typedargslist, syms.arglist} and n.children and n.children[-1].type == token.COMMA ): + if n.type == syms.typedargslist: + feature = Feature.TRAILING_COMMA_IN_DEF + else: + feature = Feature.TRAILING_COMMA_IN_CALL + for ch in n.children: if ch.type in STARS: - return True + features.add(feature) if ch.type == syms.argument: for argch in ch.children: if argch.type in STARS: - return True + features.add(feature) + + return features - return False + +def detect_target_versions(node: Node) -> Set[TargetVersion]: + """Detect the version to target based on the nodes used.""" + features = get_features_used(node) + return { + version for version in TargetVersion if features <= VERSION_TO_FEATURES[version] + } def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[LeafID]]: @@ -3073,7 +3439,7 @@ def get_future_imports(node: Node) -> Set[str]: elif child.type == syms.import_as_names: yield from get_imports_from_children(child.children) else: - assert False, "Invalid syntax parsing imports" + raise AssertionError("Invalid syntax parsing imports") for child in node.children: if child.type != syms.simple_stmt: @@ -3099,12 +3465,23 @@ def get_future_imports(node: Node) -> Set[str]: return imports +@lru_cache() +def get_gitignore(root: Path) -> PathSpec: + """ Return a PathSpec matching gitignore content if present.""" + gitignore = root / ".gitignore" + if not gitignore.is_file(): + return PathSpec.from_lines("gitwildmatch", []) + else: + return PathSpec.from_lines("gitwildmatch", gitignore.open()) + + 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. @@ -3115,8 +3492,17 @@ def gen_python_files_in_dir( """ 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( @@ -3128,13 +3514,16 @@ def gen_python_files_in_dir( 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) @@ -3257,17 +3646,58 @@ class Report: return ", ".join(report) + "." +def parse_ast(src: str) -> Union[ast.AST, ast3.AST, ast27.AST]: + filename = "" + if sys.version_info >= (3, 8): + # TODO: support Python 4+ ;) + for minor_version in range(sys.version_info[1], 4, -1): + try: + return ast.parse(src, filename, feature_version=(3, minor_version)) + except SyntaxError: + continue + else: + for feature_version in (7, 6): + try: + return ast3.parse(src, filename, feature_version=feature_version) + except SyntaxError: + continue + + return ast27.parse(src) + + +def _fixup_ast_constants( + 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 node + + def assert_equivalent(src: str, dst: str) -> None: """Raise AssertionError if `src` and `dst` aren't equivalent.""" - import ast - import traceback - - def _v(node: ast.AST, depth: int = 0) -> Iterator[str]: + def _v(node: Union[ast.AST, ast3.AST, ast27.AST], depth: int = 0) -> Iterator[str]: """Simple visitor generating strings to compare ASTs by content.""" + + node = _fixup_ast_constants(node) + yield f"{' ' * depth}{node.__class__.__name__}(" for field in sorted(node._fields): + # TypeIgnore has only one field 'lineno' which breaks this comparison + type_ignore_classes = (ast3.TypeIgnore, ast27.TypeIgnore) + if sys.version_info >= (3, 8): + type_ignore_classes += (ast.TypeIgnore,) + if isinstance(node, type_ignore_classes): + break + try: value = getattr(node, field) except AttributeError: @@ -3277,10 +3707,19 @@ def assert_equivalent(src: str, dst: str) -> None: if isinstance(value, list): for item in value: - if isinstance(item, ast.AST): + # Ignore nested tuples within del statements, because we may insert + # parentheses and they change the AST. + if ( + field == "targets" + and isinstance(node, (ast.Delete, ast3.Delete, ast27.Delete)) + and isinstance(item, (ast.Tuple, ast3.Tuple, ast27.Tuple)) + ): + 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) - elif isinstance(value, ast.AST): + elif isinstance(value, (ast.AST, ast3.AST, ast27.AST)): yield from _v(value, depth + 2) else: @@ -3289,22 +3728,20 @@ def assert_equivalent(src: str, dst: str) -> None: yield f"{' ' * depth}) # /{node.__class__.__name__}" try: - src_ast = ast.parse(src) + src_ast = parse_ast(src) except Exception as exc: - major, minor = sys.version_info[:2] raise AssertionError( - f"cannot use --safe with this file; failed to parse source file " - f"with Python {major}.{minor}'s builtin AST. Re-run with --fast " - f"or stop using deprecated Python 2 syntax. AST error message: {exc}" + f"cannot use --safe with this file; failed to parse source file. " + f"AST error message: {exc}" ) try: - dst_ast = ast.parse(dst) + dst_ast = parse_ast(dst) except Exception as exc: log = dump_to_file("".join(traceback.format_tb(exc.__traceback__)), dst) raise AssertionError( f"INTERNAL ERROR: Black produced invalid code: {exc}. " - f"Please report a bug on https://github.com/ambv/black/issues. " + f"Please report a bug on https://github.com/psf/black/issues. " f"This invalid output might be helpful: {log}" ) from None @@ -3315,16 +3752,14 @@ def assert_equivalent(src: str, dst: str) -> None: raise AssertionError( 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"Please report a bug on https://github.com/psf/black/issues. " f"This diff might be helpful: {log}" ) from None -def assert_stable( - src: str, dst: str, line_length: int, mode: FileMode = FileMode.AUTO_DETECT -) -> None: +def assert_stable(src: str, dst: str, mode: FileMode) -> None: """Raise AssertionError if `dst` reformats differently the second time.""" - newdst = format_str(dst, line_length=line_length, mode=mode) + newdst = format_str(dst, mode=mode) if dst != newdst: log = dump_to_file( diff(src, dst, "source", "first pass"), @@ -3333,15 +3768,13 @@ def assert_stable( raise AssertionError( 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"Please report a bug on https://github.com/psf/black/issues. " f"This diff might be helpful: {log}" ) from None def dump_to_file(*output: str) -> str: """Dump `output` to a temporary file. Return path to the file.""" - import tempfile - with tempfile.NamedTemporaryFile( mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8" ) as f: @@ -3352,6 +3785,15 @@ def dump_to_file(*output: str) -> str: return f.name +@contextmanager +def nullcontext() -> Iterator[None]: + """Return an empty context manager. + + To be used like `nullcontext` in Python 3.7. + """ + yield + + def diff(a: str, b: str, a_name: str, b_name: str) -> str: """Return a unified diff string between strings `a` and `b`.""" import difflib @@ -3370,11 +3812,15 @@ def cancel(tasks: Iterable[asyncio.Task]) -> None: task.cancel() -def shutdown(loop: BaseEventLoop) -> None: +def shutdown(loop: asyncio.AbstractEventLoop) -> None: """Cancel all pending tasks on `loop`, wait for them, and close the loop.""" try: + if sys.version_info[:2] >= (3, 7): + all_tasks = asyncio.all_tasks + else: + all_tasks = asyncio.Task.all_tasks # This part is borrowed from asyncio/runners.py in Python 3.7b2. - to_cancel = [task for task in asyncio.Task.all_tasks(loop) if not task.done()] + to_cancel = [task for task in all_tasks(loop) if not task.done()] if not to_cancel: return @@ -3408,7 +3854,8 @@ def re_compile_maybe_verbose(regex: str) -> Pattern[str]: """ if "\n" in regex: regex = "(?x)" + regex - return re.compile(regex) + compiled: Pattern[str] = re.compile(regex) + return compiled def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]: @@ -3435,8 +3882,7 @@ def enumerate_with_length( if "\n" in leaf.value: return # Multiline strings, we can't continue. - comment: Optional[Leaf] - for comment in line.comments_after(leaf, index): + for comment in line.comments_after(leaf): length += len(comment.value) yield index, leaf, length @@ -3581,23 +4027,23 @@ def can_omit_invisible_parens(line: Line, line_length: int) -> bool: return False -def get_cache_file(line_length: int, mode: FileMode) -> Path: - return CACHE_DIR / f"cache.{line_length}.{mode.value}.pickle" +def get_cache_file(mode: FileMode) -> Path: + return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle" -def read_cache(line_length: int, mode: FileMode) -> Cache: +def read_cache(mode: FileMode) -> Cache: """Read the cache if it exists and is well formed. If it is not well formed, the call to write_cache later should resolve the issue. """ - cache_file = get_cache_file(line_length, mode) + cache_file = get_cache_file(mode) if not cache_file.exists(): return {} with cache_file.open("rb") as fobj: try: cache: Cache = pickle.load(fobj) - except pickle.UnpicklingError: + except (pickle.UnpicklingError, ValueError): return {} return cache @@ -3625,17 +4071,15 @@ def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set return todo, done -def write_cache( - cache: Cache, sources: Iterable[Path], line_length: int, mode: FileMode -) -> None: +def write_cache(cache: Cache, sources: Iterable[Path], mode: FileMode) -> None: """Update the cache file.""" - cache_file = get_cache_file(line_length, mode) + cache_file = get_cache_file(mode) try: - if not CACHE_DIR.exists(): - CACHE_DIR.mkdir(parents=True) + CACHE_DIR.mkdir(parents=True, exist_ok=True) new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}} - with cache_file.open("wb") as fobj: - pickle.dump(new_cache, fobj, protocol=pickle.HIGHEST_PROTOCOL) + with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f: + pickle.dump(new_cache, f, protocol=4) + os.replace(f.name, cache_file) except OSError: pass @@ -3663,6 +4107,7 @@ def patch_click() -> None: def patched_main() -> None: + freeze_support() patch_click() main()