X-Git-Url: https://git.madduck.net/etc/vim.git/blobdiff_plain/a20a3eeb0f738d3434efe3be8932db11722757a4..1580477615c3d14b029fc40904a10dfd31cc99a6:/black.py?ds=sidebyside diff --git a/black.py b/black.py index c2c118a..2850ae1 100644 --- a/black.py +++ b/black.py @@ -1,49 +1,67 @@ -#!/usr/bin/env python3 - import asyncio from asyncio.base_events import BaseEventLoop from concurrent.futures import Executor, ProcessPoolExecutor +from datetime import datetime from enum import Enum -from functools import partial, wraps -import keyword +from functools import lru_cache, partial, wraps +import io +import itertools import logging -from multiprocessing import Manager +from multiprocessing import Manager, freeze_support import os from pathlib import Path -import tokenize +import pickle +import re import signal import sys +import tempfile +import tokenize from typing import ( Any, Callable, + Collection, Dict, + Generator, Generic, Iterable, Iterator, List, Optional, + Pattern, + Sequence, Set, Tuple, - Type, TypeVar, Union, + cast, ) -from attr import dataclass, Factory +from appdirs import user_cache_dir +from attr import dataclass, evolve, Factory import click +import toml # 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 -__version__ = "18.3a4" + +__version__ = "18.9b0" DEFAULT_LINE_LENGTH = 88 +DEFAULT_EXCLUDES = ( + r"/(\.eggs|\.git|\.hg|\.mypy_cache|\.nox|\.tox|\.venv|_build|buck-out|build|dist)/" +) +DEFAULT_INCLUDES = r"\.pyi?$" +CACHE_DIR = Path(user_cache_dir("black", version=__version__)) + + # types -syms = pygram.python_symbols FileContent = str Encoding = str +NewLine = str Depth = int NodeType = int LeafID = int @@ -51,65 +69,199 @@ Priority = int Index = int LN = Union[Leaf, Node] SplitFunc = Callable[["Line", bool], Iterator["Line"]] +Timestamp = float +FileSize = int +CacheInfo = Tuple[Timestamp, FileSize] +Cache = Dict[Path, CacheInfo] out = partial(click.secho, bold=True, err=True) err = partial(click.secho, fg="red", err=True) +pygram.initialize(CACHE_DIR) +syms = pygram.python_symbols + class NothingChanged(UserWarning): - """Raised by :func:`format_file` when reformatted code is the same as source.""" + """Raised when reformatted code is the same as source.""" class CannotSplit(Exception): - """A readable split that fits the allotted line length is impossible. + """A readable split that fits the allotted line length is impossible.""" - Raised by :func:`left_hand_split`, :func:`right_hand_split`, and - :func:`delimiter_split`. - """ +class InvalidInput(ValueError): + """Raised when input source code fails all parse attempts.""" -class FormatError(Exception): - """Base exception for `# fmt: on` and `# fmt: off` handling. - It holds the number of bytes of the prefix consumed before the format - control comment appeared. - """ +class WriteBack(Enum): + NO = 0 + YES = 1 + DIFF = 2 + CHECK = 3 - def __init__(self, consumed: int) -> None: - super().__init__(consumed) - self.consumed = consumed + @classmethod + def from_configuration(cls, *, check: bool, diff: bool) -> "WriteBack": + if check and not diff: + return cls.CHECK - def trim_prefix(self, leaf: Leaf) -> None: - leaf.prefix = leaf.prefix[self.consumed:] + return cls.DIFF if diff else cls.YES - def leaf_from_consumed(self, leaf: Leaf) -> Leaf: - """Returns a new Leaf from the consumed part of the prefix.""" - unformatted_prefix = leaf.prefix[:self.consumed] - return Leaf(token.NEWLINE, unformatted_prefix) +class Changed(Enum): + NO = 0 + CACHED = 1 + YES = 2 + + +class TargetVersion(Enum): + PYPY35 = 1 + CPY27 = 2 + CPY33 = 3 + CPY34 = 4 + CPY35 = 5 + CPY36 = 6 + CPY37 = 7 + CPY38 = 8 + + def is_python2(self) -> bool: + return self is TargetVersion.CPY27 + + +PY36_VERSIONS = {TargetVersion.CPY36, TargetVersion.CPY37, TargetVersion.CPY38} + + +class Feature(Enum): + # All string literals are unicode + UNICODE_LITERALS = 1 + F_STRINGS = 2 + NUMERIC_UNDERSCORES = 3 + TRAILING_COMMA = 4 + + +VERSION_TO_FEATURES: Dict[TargetVersion, Set[Feature]] = { + TargetVersion.CPY27: set(), + TargetVersion.PYPY35: {Feature.UNICODE_LITERALS, Feature.F_STRINGS}, + TargetVersion.CPY33: {Feature.UNICODE_LITERALS}, + TargetVersion.CPY34: {Feature.UNICODE_LITERALS}, + TargetVersion.CPY35: {Feature.UNICODE_LITERALS, Feature.TRAILING_COMMA}, + TargetVersion.CPY36: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA, + }, + TargetVersion.CPY37: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA, + }, + TargetVersion.CPY38: { + Feature.UNICODE_LITERALS, + Feature.F_STRINGS, + Feature.NUMERIC_UNDERSCORES, + Feature.TRAILING_COMMA, + }, +} -class FormatOn(FormatError): - """Found a comment like `# fmt: on` in the file.""" +@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) -class FormatOff(FormatError): - """Found a comment like `# fmt: off` in the file.""" +def supports_feature(target_versions: Set[TargetVersion], feature: Feature) -> bool: + return all(feature in VERSION_TO_FEATURES[version] for version in target_versions) -class WriteBack(Enum): - NO = 0 - YES = 1 - DIFF = 2 + +def read_pyproject_toml( + ctx: click.Context, param: click.Parameter, value: Union[str, int, bool, None] +) -> Optional[str]: + """Inject Black configuration from "pyproject.toml" into defaults in `ctx`. + + Returns the path to a successfully found and read configuration file, None + otherwise. + """ + assert not isinstance(value, (int, bool)), "Invalid parameter type passed" + if not value: + root = find_project_root(ctx.params.get("src", ())) + path = root / "pyproject.toml" + if path.is_file(): + value = str(path) + else: + return None + + try: + pyproject_toml = toml.load(value) + config = pyproject_toml.get("tool", {}).get("black", {}) + except (toml.TomlDecodeError, OSError) as e: + raise click.FileError( + filename=value, hint=f"Error reading configuration file: {e}" + ) + + if not config: + return None + + if ctx.default_map is None: + ctx.default_map = {} + ctx.default_map.update( # type: ignore # bad types in .pyi + {k.replace("--", "").replace("-", "_"): v for k, v in config.items()} + ) + return value -@click.command() +@click.command(context_settings=dict(help_option_names=["-h", "--help"])) @click.option( "-l", "--line-length", type=int, default=DEFAULT_LINE_LENGTH, - help="How many character per line to allow.", + 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( + "--pyi", + is_flag=True, + help=( + "Format all input files like typing stubs regardless of file extension " + "(useful when piping source on standard input)." + ), +) +@click.option( + "-S", + "--skip-string-normalization", + is_flag=True, + help="Don't normalize string quotes or prefixes.", +) @click.option( "--check", is_flag=True, @@ -129,6 +281,49 @@ class WriteBack(Enum): is_flag=True, help="If --fast given, skip temporary sanity checks. [default: --safe]", ) +@click.option( + "--include", + type=str, + default=DEFAULT_INCLUDES, + help=( + "A regular expression that matches files and directories that should be " + "included on recursive searches. An empty value means all files are " + "included regardless of the name. Use forward slashes for directories on " + "all platforms (Windows, too). Exclusions are calculated first, inclusions " + "later." + ), + show_default=True, +) +@click.option( + "--exclude", + type=str, + default=DEFAULT_EXCLUDES, + help=( + "A regular expression that matches files and directories that should be " + "excluded on recursive searches. An empty value means no paths are excluded. " + "Use forward slashes for directories on all platforms (Windows, too). " + "Exclusions are calculated first, inclusions later." + ), + show_default=True, +) +@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.option( + "-v", + "--verbose", + is_flag=True, + help=( + "Also emit messages to stderr about files that were not changed or were " + "ignored due to --exclude=." + ), +) @click.version_option(version=__version__) @click.argument( "src", @@ -136,88 +331,171 @@ class WriteBack(Enum): type=click.Path( exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True ), + is_eager=True, +) +@click.option( + "--config", + type=click.Path( + exists=False, file_okay=True, dir_okay=False, readable=True, allow_dash=False + ), + is_eager=True, + callback=read_pyproject_toml, + help="Read configuration from PATH.", ) @click.pass_context def main( ctx: click.Context, line_length: int, + target_version: List[TargetVersion], check: bool, diff: bool, fast: bool, - src: List[str], + pyi: bool, + skip_string_normalization: bool, + quiet: bool, + verbose: bool, + include: str, + exclude: str, + src: Tuple[str], + config: Optional[str], ) -> None: """The uncompromising code formatter.""" - sources: List[Path] = [] + write_back = WriteBack.from_configuration(check=check, diff=diff) + if target_version: + versions = set(target_version) + 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") + try: + include_regex = re_compile_maybe_verbose(include) + except re.error: + err(f"Invalid regular expression for include given: {include!r}") + ctx.exit(2) + try: + exclude_regex = re_compile_maybe_verbose(exclude) + except re.error: + err(f"Invalid regular expression for exclude given: {exclude!r}") + ctx.exit(2) + report = Report(check=check, quiet=quiet, verbose=verbose) + root = find_project_root(src) + sources: Set[Path] = set() for s in src: p = Path(s) if p.is_dir(): - sources.extend(gen_python_files_in_dir(p)) - elif p.is_file(): + sources.update( + gen_python_files_in_dir(p, root, include_regex, exclude_regex, report) + ) + elif p.is_file() or s == "-": # if a file was explicitly given, we don't care about its extension - sources.append(p) - elif s == "-": - sources.append(Path("-")) + sources.add(p) 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: + if verbose or not quiet: + out("No paths given. Nothing to do 😴") ctx.exit(0) - elif len(sources) == 1: - p = sources[0] - report = Report(check=check) - try: - if not p.is_file() and str(p) == "-": - changed = format_stdin_to_stdout( - 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=write_back - ) - report.done(p, changed) - except Exception as exc: - report.failed(p, str(exc)) - ctx.exit(report.return_code) + + if len(sources) == 1: + reformat_one( + src=sources.pop(), + fast=fast, + write_back=write_back, + mode=mode, + report=report, + ) else: loop = asyncio.get_event_loop() executor = ProcessPoolExecutor(max_workers=os.cpu_count()) - return_code = 1 try: - return_code = loop.run_until_complete( + loop.run_until_complete( schedule_formatting( - sources, line_length, write_back, fast, loop, executor + sources=sources, + fast=fast, + write_back=write_back, + mode=mode, + report=report, + loop=loop, + executor=executor, ) ) finally: shutdown(loop) - ctx.exit(return_code) + if verbose or not quiet: + bang = "💥 💔 💥" if report.return_code else "✨ 🍰 ✨" + out(f"All done! {bang}") + click.secho(str(report), err=True) + ctx.exit(report.return_code) + + +def reformat_one( + 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 + :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(fast=fast, write_back=write_back, mode=mode): + changed = Changed.YES + else: + cache: Cache = {} + if write_back != WriteBack.DIFF: + 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, 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], mode) + report.done(src, changed) + except Exception as exc: + report.failed(src, str(exc)) async def schedule_formatting( - sources: List[Path], - line_length: int, - write_back: WriteBack, + sources: Set[Path], fast: bool, + write_back: WriteBack, + mode: FileMode, + report: "Report", loop: BaseEventLoop, executor: Executor, -) -> int: +) -> None: """Run formatting of `sources` in parallel using the provided `executor`. (Use ProcessPoolExecutors for actual parallelism.) - `line_length`, `write_back`, and `fast` options are passed to + `line_length`, `write_back`, `fast`, and `pyi` options are passed to :func:`format_file_in_place`. """ + cache: Cache = {} + if write_back != WriteBack.DIFF: + cache = read_cache(mode) + sources, cached = filter_cached(cache, sources) + for src in sorted(cached): + report.done(src, Changed.CACHED) + if not sources: + return + + cancelled = [] + sources_to_cache = [] lock = None if write_back == WriteBack.DIFF: # For diff output, we need locks to ensure we don't interleave output @@ -225,68 +503,84 @@ async def schedule_formatting( manager = Manager() lock = manager.Lock() tasks = { - src: loop.run_in_executor( - executor, format_file_in_place, src, line_length, fast, write_back, lock - ) - for src in sources + loop.run_in_executor( + executor, format_file_in_place, src, fast, mode, write_back, lock + ): src + for src in sorted(sources) } - _task_values = list(tasks.values()) - loop.add_signal_handler(signal.SIGINT, cancel, _task_values) - loop.add_signal_handler(signal.SIGTERM, cancel, _task_values) - await asyncio.wait(tasks.values()) - cancelled = [] - report = Report(check=not write_back) - for src, task in tasks.items(): - if not task.done(): - report.failed(src, "timed out, cancelling") - task.cancel() - cancelled.append(task) - elif task.cancelled(): - cancelled.append(task) - elif task.exception(): - report.failed(src, str(task.exception())) - else: - report.done(src, task.result()) + pending: Iterable[asyncio.Task] = tasks.keys() + try: + loop.add_signal_handler(signal.SIGINT, cancel, pending) + loop.add_signal_handler(signal.SIGTERM, cancel, pending) + except NotImplementedError: + # There are no good alternatives for these on Windows. + pass + while pending: + done, _ = await asyncio.wait(pending, return_when=asyncio.FIRST_COMPLETED) + for task in done: + src = tasks.pop(task) + if task.cancelled(): + cancelled.append(task) + elif task.exception(): + report.failed(src, str(task.exception())) + else: + changed = Changed.YES if task.result() else Changed.NO + # If the file was written back or was successfully checked as + # well-formatted, store this information in the cache. + if write_back is WriteBack.YES or ( + write_back is WriteBack.CHECK and changed is Changed.NO + ): + sources_to_cache.append(src) + report.done(src, changed) if cancelled: await asyncio.gather(*cancelled, loop=loop, return_exceptions=True) - else: - out("All done! ✨ 🍰 ✨") - click.echo(str(report)) - return report.return_code + if sources_to_cache: + 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, lock: Any = None, # multiprocessing.Manager().Lock() is some crazy proxy ) -> bool: """Format file under `src` path. Return True if changed. - If `write_back` is True, write reformatted code back to stdout. + 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`. """ - with tokenize.open(src) as src_buffer: - src_contents = src_buffer.read() + if src.suffix == ".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 - ) + dst_contents = format_file_contents(src_contents, fast=fast, mode=mode) except NothingChanged: return False if write_back == write_back.YES: - with open(src, "w", encoding=src_buffer.encoding) as f: + with open(src, "w", encoding=encoding, newline=newline) as f: f.write(dst_contents) elif write_back == write_back.DIFF: - src_name = f"{src.name} (original)" - dst_name = f"{src.name} (formatted)" + now = datetime.utcnow() + 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: - sys.stdout.write(diff_contents) + f = io.TextIOWrapper( + sys.stdout.buffer, + encoding=encoding, + newline=newline, + write_through=True, + ) + f.write(diff_contents) + f.detach() finally: if lock: lock.release() @@ -294,33 +588,40 @@ def format_file_in_place( def format_stdin_to_stdout( - line_length: int, fast: bool, write_back: WriteBack = WriteBack.NO + fast: bool, *, write_back: WriteBack = WriteBack.NO, mode: FileMode ) -> 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`. + If `write_back` is YES, write reformatted code back to stdout. If it is DIFF, + write a diff to stdout. The `mode` argument is passed to + :func:`format_file_contents`. """ - src = sys.stdin.read() + 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) + dst = format_file_contents(src, fast=fast, mode=mode) return True except NothingChanged: - dst = src return False finally: + f = io.TextIOWrapper( + sys.stdout.buffer, encoding=encoding, newline=newline, write_through=True + ) if write_back == WriteBack.YES: - sys.stdout.write(dst) + f.write(dst) elif write_back == WriteBack.DIFF: - src_name = " (original)" - dst_name = " (formatted)" - sys.stdout.write(diff(src, dst, src_name, dst_name)) + now = datetime.utcnow() + src_name = f"STDIN\t{then} +0000" + dst_name = f"STDOUT\t{now} +0000" + f.write(diff(src, dst, src_name, dst_name)) + f.detach() def format_file_contents( - src_contents: str, line_length: int, fast: bool + src_contents: str, *, fast: bool, mode: FileMode ) -> FileContent: """Reformat contents a file and return new contents. @@ -331,26 +632,36 @@ def format_file_contents( if src_contents.strip() == "": raise NothingChanged - dst_contents = format_str(src_contents, line_length=line_length) + 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) + assert_stable(src_contents, dst_contents, mode=mode) return dst_contents -def format_str(src_contents: str, line_length: int) -> 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. """ - src_node = lib2to3_parse(src_contents) + src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions) dst_contents = "" - lines = LineGenerator() - elt = EmptyLineTracker() - py36 = is_python36(src_node) + future_imports = get_future_imports(src_node) + 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="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=mode.is_pyi) empty_line = Line() after = 0 for current_line in lines.visit(src_node): @@ -359,26 +670,58 @@ def format_str(src_contents: str, line_length: int) -> FileContent: 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): + for line in split_line( + current_line, + line_length=mode.line_length, + supports_trailing_commas=supports_feature(versions, Feature.TRAILING_COMMA), + ): dst_contents += str(line) return dst_contents +def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]: + """Return a tuple of (decoded_contents, encoding, newline). + + `newline` is either CRLF or LF but `decoded_contents` is decoded with + universal newlines (i.e. only contains LF). + """ + srcbuf = io.BytesIO(src) + encoding, lines = tokenize.detect_encoding(srcbuf.readline) + if not lines: + return "", encoding, "\n" + + newline = "\r\n" if b"\r\n" == lines[0][-2:] else "\n" + srcbuf.seek(0) + with io.TextIOWrapper(srcbuf, encoding) as tiow: + return tiow.read(), encoding, newline + + GRAMMARS = [ pygram.python_grammar_no_print_statement_no_exec_statement, pygram.python_grammar_no_print_statement, - pygram.python_grammar_no_exec_statement, pygram.python_grammar, ] -def lib2to3_parse(src_txt: str) -> Node: +def get_grammars(target_versions: Set[TargetVersion]) -> List[Grammar]: + if not target_versions: + return GRAMMARS + elif all(not version.is_python2() for version in target_versions): + # Python 2-compatible code, so don't try Python 3 grammar. + return [ + pygram.python_grammar_no_print_statement_no_exec_statement, + pygram.python_grammar_no_print_statement, + ] + else: + return [pygram.python_grammar] + + +def lib2to3_parse(src_txt: str, target_versions: Iterable[TargetVersion] = ()) -> Node: """Given a string with source, return the lib2to3 Node.""" - grammar = pygram.python_grammar_no_print_statement - if src_txt[-1] != "\n": - nl = "\r\n" if "\r\n" in src_txt[:1024] else "\n" - src_txt += nl - for grammar in GRAMMARS: + if src_txt[-1:] != "\n": + src_txt += "\n" + + for grammar in get_grammars(set(target_versions)): drv = driver.Driver(grammar, pytree.convert) try: result = drv.parse_string(src_txt, True) @@ -391,7 +734,7 @@ def lib2to3_parse(src_txt: str) -> Node: faulty_line = lines[lineno - 1] except IndexError: faulty_line = "" - exc = ValueError(f"Cannot parse: {lineno}:{column}: {faulty_line}") + exc = InvalidInput(f"Cannot parse: {lineno}:{column}: {faulty_line}") else: raise exc from None @@ -460,18 +803,18 @@ class DebugVisitor(Visitor[T]): out(f" {node.value!r}", fg="blue", bold=False) @classmethod - def show(cls, code: str) -> None: + def show(cls, code: Union[str, Leaf, Node]) -> None: """Pretty-print the lib2to3 AST of a given string of `code`. Convenience method for debugging. """ v: DebugVisitor[None] = DebugVisitor() - list(v.visit(lib2to3_parse(code))) + if isinstance(code, str): + code = lib2to3_parse(code) + 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, @@ -483,6 +826,7 @@ STATEMENT = { syms.classdef, } STANDALONE_COMMENT = 153 +token.tok_name[STANDALONE_COMMENT] = "STANDALONE_COMMENT" LOGIC_OPERATORS = {"and", "or"} COMPARATORS = { token.LESS, @@ -493,27 +837,92 @@ COMPARATORS = { token.GREATEREQUAL, } MATH_OPERATORS = { + token.VBAR, + token.CIRCUMFLEX, + token.AMPER, + token.LEFTSHIFT, + token.RIGHTSHIFT, token.PLUS, token.MINUS, token.STAR, token.SLASH, - token.VBAR, - token.AMPER, + token.DOUBLESLASH, token.PERCENT, - token.CIRCUMFLEX, + token.AT, token.TILDE, - token.LEFTSHIFT, - token.RIGHTSHIFT, token.DOUBLESTAR, - token.DOUBLESLASH, } -VARARGS = {token.STAR, token.DOUBLESTAR} +STARS = {token.STAR, token.DOUBLESTAR} +VARARGS_PARENTS = { + syms.arglist, + syms.argument, # double star in arglist + syms.trailer, # single argument to call + syms.typedargslist, + syms.varargslist, # lambdas +} +UNPACKING_PARENTS = { + syms.atom, # single element of a list or set literal + syms.dictsetmaker, + syms.listmaker, + syms.testlist_gexp, + syms.testlist_star_expr, +} +TEST_DESCENDANTS = { + syms.test, + syms.lambdef, + syms.or_test, + syms.and_test, + syms.not_test, + syms.comparison, + syms.star_expr, + syms.expr, + syms.xor_expr, + syms.and_expr, + syms.shift_expr, + syms.arith_expr, + syms.trailer, + syms.term, + syms.power, +} +ASSIGNMENTS = { + "=", + "+=", + "-=", + "*=", + "@=", + "/=", + "%=", + "&=", + "|=", + "^=", + "<<=", + ">>=", + "**=", + "//=", +} COMPREHENSION_PRIORITY = 20 -COMMA_PRIORITY = 10 -LOGIC_PRIORITY = 5 -STRING_PRIORITY = 4 -COMPARATOR_PRIORITY = 3 -MATH_PRIORITY = 1 +COMMA_PRIORITY = 18 +TERNARY_PRIORITY = 16 +LOGIC_PRIORITY = 14 +STRING_PRIORITY = 12 +COMPARATOR_PRIORITY = 10 +MATH_PRIORITIES = { + token.VBAR: 9, + token.CIRCUMFLEX: 8, + token.AMPER: 7, + token.LEFTSHIFT: 6, + token.RIGHTSHIFT: 6, + token.PLUS: 5, + token.MINUS: 5, + token.STAR: 4, + token.SLASH: 4, + token.DOUBLESLASH: 4, + token.PERCENT: 4, + token.AT: 4, + token.TILDE: 3, + token.DOUBLESTAR: 2, +} +DOT_PRIORITY = 1 @dataclass @@ -524,6 +933,8 @@ class BracketTracker: bracket_match: Dict[Tuple[Depth, NodeType], Leaf] = Factory(dict) delimiters: Dict[LeafID, Priority] = Factory(dict) previous: Optional[Leaf] = None + _for_loop_depths: List[int] = Factory(list) + _lambda_argument_depths: List[int] = Factory(list) def mark(self, leaf: Leaf) -> None: """Mark `leaf` with bracket-related metadata. Keep track of delimiters. @@ -543,22 +954,27 @@ class BracketTracker: if leaf.type == token.COMMENT: return + self.maybe_decrement_after_for_loop_variable(leaf) + self.maybe_decrement_after_lambda_arguments(leaf) if leaf.type in CLOSING_BRACKETS: self.depth -= 1 opening_bracket = self.bracket_match.pop((self.depth, leaf.type)) 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 self.previous = leaf + self.maybe_increment_lambda_arguments(leaf) + self.maybe_increment_for_loop_variable(leaf) def any_open_brackets(self) -> bool: """Return True if there is an yet unmatched open bracket on the line.""" @@ -567,10 +983,79 @@ class BracketTracker: def max_delimiter_priority(self, exclude: Iterable[LeafID] = ()) -> int: """Return the highest priority of a delimiter found on the line. - Values are consistent with what `is_delimiter()` returns. + Values are consistent with what `is_split_*_delimiter()` return. + Raises ValueError on no delimiters. """ 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: + """Return the number of delimiters with the given `priority`. + + If no `priority` is passed, defaults to max priority on the line. + """ + if not self.delimiters: + return 0 + + priority = priority or self.max_delimiter_priority() + return sum(1 for p in self.delimiters.values() if p == priority) + + 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, increase the depth of + tokens between `for` and `in`. + """ + if leaf.type == token.NAME and leaf.value == "for": + self.depth += 1 + self._for_loop_depths.append(self.depth) + 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_depths + and self._for_loop_depths[-1] == self.depth + and leaf.type == token.NAME + and leaf.value == "in" + ): + self.depth -= 1 + self._for_loop_depths.pop() + return True + + return False + + def maybe_increment_lambda_arguments(self, leaf: Leaf) -> bool: + """In a lambda expression, there might be more than one argument. + + To avoid splitting on the comma in this situation, increase the depth of + tokens between `lambda` and `:`. + """ + if leaf.type == token.NAME and leaf.value == "lambda": + self.depth += 1 + self._lambda_argument_depths.append(self.depth) + return True + + return False + + def maybe_decrement_after_lambda_arguments(self, leaf: Leaf) -> bool: + """See `maybe_increment_lambda_arguments` above for explanation.""" + if ( + self._lambda_argument_depths + and self._lambda_argument_depths[-1] == self.depth + and leaf.type == token.COLON + ): + self.depth -= 1 + self._lambda_argument_depths.pop() + return True + + return False + + def get_open_lsqb(self) -> Optional[Leaf]: + """Return the most recent opening square bracket (if any).""" + return self.bracket_match.get((self.depth - 1, token.RSQB)) + @dataclass class Line: @@ -578,11 +1063,12 @@ class Line: depth: int = 0 leaves: List[Leaf] = Factory(list) - comments: List[Tuple[Index, Leaf]] = Factory(list) + # The LeafID keys of comments must remain ordered by the corresponding leaf's index + # in leaves + comments: Dict[LeafID, List[Leaf]] = Factory(dict) bracket_tracker: BracketTracker = Factory(BracketTracker) inside_brackets: bool = False - has_for: bool = False - _for_loop_variable: bool = False + should_explode: bool = False def append(self, leaf: Leaf, preformatted: bool = False) -> None: """Add a new `leaf` to the end of the line. @@ -594,20 +1080,21 @@ class Line: 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 + if token.COLON == leaf.type and self.is_class_paren_empty: + del self.leaves[-2:] if self.leaves and not preformatted: # Note: at this point leaf.prefix should be empty except for # imports, for which we only preserve newlines. - leaf.prefix += whitespace(leaf) + leaf.prefix += whitespace( + leaf, complex_subscript=self.is_complex_subscript(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 not self.append_comment(leaf): self.leaves.append(leaf) @@ -652,6 +1139,13 @@ class Line: and self.leaves[0].value == "class" ) + @property + def is_stub_class(self) -> bool: + """Is this line a class definition with a body consisting only of "..."?""" + return self.is_class and self.leaves[-3:] == [ + Leaf(token.DOT, ".") for _ in range(3) + ] + @property def is_def(self) -> bool: """Is this a function definition? (Also returns True for async defs.)""" @@ -664,42 +1158,50 @@ class Line: second_leaf: Optional[Leaf] = self.leaves[1] except IndexError: second_leaf = None - return ( - (first_leaf.type == token.NAME and first_leaf.value == "def") - or ( - first_leaf.type == token.ASYNC - and second_leaf is not None - and second_leaf.type == token.NAME - and second_leaf.value == "def" - ) + return (first_leaf.type == token.NAME and first_leaf.value == "def") or ( + first_leaf.type == token.ASYNC + and second_leaf is not None + and second_leaf.type == token.NAME + and second_leaf.value == "def" ) @property - def is_flow_control(self) -> bool: - """Is this line a flow control statement? + def is_class_paren_empty(self) -> bool: + """Is this a class with no base classes but using parentheses? - Those are `return`, `raise`, `break`, and `continue`. + Those are unnecessary and should be removed. """ return ( bool(self) - and self.leaves[0].type == token.NAME - and self.leaves[0].value in FLOW_CONTROL + and len(self.leaves) == 4 + and self.is_class + and self.leaves[2].type == token.LPAR + and self.leaves[2].value == "(" + and self.leaves[3].type == token.RPAR + and self.leaves[3].value == ")" ) @property - def is_yield(self) -> bool: - """Is this line a yield statement?""" + def is_triple_quoted_string(self) -> bool: + """Is the line a triple quoted string?""" return ( bool(self) - and self.leaves[0].type == token.NAME - and self.leaves[0].value == "yield" + and self.leaves[0].type == token.STRING + and self.leaves[0].value.startswith(('"""', "'''")) ) - @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: + if leaf.bracket_depth <= depth_limit: + return True + + return False + + def contains_multiline_strings(self) -> bool: + for leaf in self.leaves: + if is_multiline_string(leaf): return True return False @@ -723,9 +1225,14 @@ class Line: self.remove_trailing_comma() return True - # 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. + # 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 @@ -736,7 +1243,7 @@ class Line: else: return False - for leaf in self.leaves[_opening_index + 1:]: + for leaf in self.leaves[_opening_index + 1 :]: if leaf is closing: break @@ -753,29 +1260,6 @@ class Line: 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, 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 append_comment(self, comment: Leaf) -> bool: """Add an inline or standalone comment to the line.""" if ( @@ -788,38 +1272,55 @@ 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)) + leaf_id = id(self.leaves[-1]) + if leaf_id not in self.comments: + self.comments[leaf_id] = [comment] + else: + self.comments[leaf_id].append(comment) return True - def comments_after(self, leaf: Leaf) -> Iterator[Leaf]: + def comments_after(self, leaf: Leaf) -> List[Leaf]: """Generate comments that should appear directly after `leaf`.""" - for _leaf_index, _leaf in enumerate(self.leaves): - if leaf is _leaf: - break - - else: - return - - for index, comment_after in self.comments: - if _leaf_index == index: - yield comment_after + 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) + # Remember, the LeafID keys of self.comments are ordered by the + # corresponding leaf's index in self.leaves + # If id(self.leaves[-2]) is in self.comments, the order doesn't change. + # Otherwise, we insert it into self.comments, and it becomes the last entry. + # However, since we delete id(self.leaves[-1]) from self.comments, the invariant + # is maintained + self.comments.setdefault(id(self.leaves[-2]), []).extend( + self.comments.get(id(self.leaves[-1]), []) + ) + self.comments.pop(id(self.leaves[-1]), None) self.leaves.pop() + def is_complex_subscript(self, leaf: Leaf) -> bool: + """Return True iff `leaf` is part of a slice with non-trivial exprs.""" + open_lsqb = self.bracket_tracker.get_open_lsqb() + if open_lsqb is None: + return False + + subscript_start = open_lsqb.next_sibling + + if isinstance(subscript_start, Node): + if subscript_start.type == syms.listmaker: + return False + + if subscript_start.type == syms.subscriptlist: + subscript_start = child_towards(subscript_start, leaf) + return subscript_start is not None and any( + n.type in TEST_DESCENDANTS for n in subscript_start.pre_order() + ) + def __str__(self) -> str: """Render the line.""" if not self: @@ -831,7 +1332,7 @@ 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" @@ -840,78 +1341,27 @@ class Line: return bool(self.leaves or self.comments) -class UnformattedLines(Line): - """Just like :class:`Line` but stores lines which aren't reformatted.""" - - def append(self, leaf: Leaf, preformatted: bool = True) -> None: - """Just add a new `leaf` to the end of the lines. - - The `preformatted` argument is ignored. +@dataclass +class EmptyLineTracker: + """Provides a stateful method that returns the number of potential extra + empty lines needed before and after the currently processed line. - Keeps track of indentation `depth`, which is useful when the user - says `# fmt: on`. Otherwise, doesn't do anything with the `leaf`. - """ - try: - list(generate_comments(leaf)) - except FormatOn as f_on: - self.leaves.append(f_on.leaf_from_consumed(leaf)) - raise + Note: this tracker works on lines that haven't been split yet. It assumes + the prefix of the first leaf consists of optional newlines. Those newlines + are consumed by `maybe_empty_lines()` and included in the computation. + """ - self.leaves.append(leaf) - if leaf.type == token.INDENT: - self.depth += 1 - elif leaf.type == token.DEDENT: - self.depth -= 1 + is_pyi: bool = False + previous_line: Optional[Line] = None + previous_after: int = 0 + previous_defs: List[int] = Factory(list) - def __str__(self) -> str: - """Render unformatted lines from leaves which were added with `append()`. + def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]: + """Return the number of extra empty lines before and after the `current_line`. - `depth` is not used for indentation in this case. + This is for separating `def`, `async def` and `class` with extra empty + lines (two on module-level). """ - if not self: - return "\n" - - res = "" - for leaf in self.leaves: - res += str(leaf) - return res - - def append_comment(self, comment: Leaf) -> bool: - """Not implemented in this class. Raises `NotImplementedError`.""" - raise NotImplementedError("Unformatted lines don't store comments separately.") - - def maybe_remove_trailing_comma(self, closing: Leaf) -> bool: - """Does nothing and returns False.""" - return False - - def maybe_increment_for_loop_variable(self, leaf: Leaf) -> bool: - """Does nothing and returns False.""" - return False - - -@dataclass -class EmptyLineTracker: - """Provides a stateful method that returns the number of potential extra - empty lines needed before and after the currently processed line. - - Note: this tracker works on lines that haven't been split yet. It assumes - the prefix of the first leaf consists of optional newlines. Those newlines - are consumed by `maybe_empty_lines()` and included in the computation. - """ - 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]: - """Return the number of extra empty lines before and after the `current_line`. - - This is for separating `def`, `async def` and `class` with extra empty - lines (two on module-level), as well as providing an extra empty line - after flow control keywords to make them more prominent. - """ - if isinstance(current_line, UnformattedLines): - return 0, 0 - before, after = self._maybe_empty_lines(current_line) before -= self.previous_after self.previous_after = after @@ -921,7 +1371,7 @@ class EmptyLineTracker: def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]: max_allowed = 1 if current_line.depth == 0: - max_allowed = 2 + max_allowed = 1 if self.is_pyi else 2 if current_line.leaves: # Consume the first leaf's extra newlines. first_leaf = current_line.leaves[0] @@ -933,26 +1383,12 @@ class EmptyLineTracker: depth = current_line.depth while self.previous_defs and self.previous_defs[-1] >= depth: self.previous_defs.pop() - before = 1 if depth else 2 - is_decorator = current_line.is_decorator - if is_decorator or current_line.is_def or current_line.is_class: - if not is_decorator: - self.previous_defs.append(depth) - if self.previous_line is None: - # Don't insert empty lines before the first line in the file. - return 0, 0 - - if self.previous_line and self.previous_line.is_decorator: - # Don't insert empty lines between decorators. - return 0, 0 - - newlines = 2 - if current_line.depth: - newlines -= 1 - return newlines, 0 - - if current_line.is_flow_control: - return before, 1 + if self.is_pyi: + before = 0 if depth else 1 + else: + before = 1 if depth else 2 + if current_line.is_decorator or current_line.is_def or current_line.is_class: + return self._maybe_empty_lines_for_class_or_def(current_line, before) if ( self.previous_line @@ -964,13 +1400,57 @@ class EmptyLineTracker: if ( self.previous_line - and self.previous_line.is_yield - and (not current_line.is_yield or depth != self.previous_line.depth) + and self.previous_line.is_class + and current_line.is_triple_quoted_string ): - return (before or 1), 0 + return before, 1 return before, 0 + def _maybe_empty_lines_for_class_or_def( + self, current_line: Line, before: int + ) -> Tuple[int, int]: + if not current_line.is_decorator: + self.previous_defs.append(current_line.depth) + if self.previous_line is None: + # Don't insert empty lines before the first line in the file. + return 0, 0 + + if self.previous_line.is_decorator: + return 0, 0 + + if self.previous_line.depth < current_line.depth and ( + self.previous_line.is_class or self.previous_line.is_def + ): + return 0, 0 + + if ( + self.previous_line.is_comment + and self.previous_line.depth == current_line.depth + and before == 0 + ): + return 0, 0 + + if self.is_pyi: + if self.previous_line.depth > current_line.depth: + newlines = 1 + elif current_line.is_class or self.previous_line.is_class: + if current_line.is_stub_class and self.previous_line.is_stub_class: + # No blank line between classes with an empty body + newlines = 0 + else: + newlines = 1 + elif current_line.is_def and not self.previous_line.is_def: + # Blank line between a block of functions and a block of non-functions + newlines = 1 + else: + newlines = 0 + else: + newlines = 2 + if current_line.depth and newlines: + newlines -= 1 + return newlines, 0 + @dataclass class LineGenerator(Visitor[Line]): @@ -979,9 +1459,13 @@ class LineGenerator(Visitor[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. """ + + is_pyi: bool = False + normalize_strings: bool = True current_line: Line = Factory(Line) + remove_u_prefix: bool = False - def line(self, indent: int = 0, type: Type[Line] = Line) -> Iterator[Line]: + def line(self, indent: int = 0) -> Iterator[Line]: """Generate a line. If the line is empty, only emit if it makes sense. @@ -990,66 +1474,41 @@ class LineGenerator(Visitor[Line]): If any lines were generated, set up a new current_line. """ if not self.current_line: - if self.current_line.__class__ == type: - self.current_line.depth += indent - else: - self.current_line = type(depth=self.current_line.depth + indent) + self.current_line.depth += indent return # Line is empty, don't emit. Creating a new one unnecessary. complete_line = self.current_line - self.current_line = type(depth=complete_line.depth + indent) + self.current_line = Line(depth=complete_line.depth + indent) yield complete_line - def visit(self, node: LN) -> Iterator[Line]: - """Main method to visit `node` and its children. - - Yields :class:`Line` objects. - """ - if isinstance(self.current_line, UnformattedLines): - # File contained `# fmt: off` - yield from self.visit_unformatted(node) - - else: - yield from super().visit(node) - def visit_default(self, node: LN) -> Iterator[Line]: """Default `visit_*()` implementation. Recurses to children of `node`.""" if isinstance(node, Leaf): any_open_brackets = self.current_line.bracket_tracker.any_open_brackets() - try: - for comment in generate_comments(node): - if any_open_brackets: - # any comment within brackets is subject to splitting - self.current_line.append(comment) - elif comment.type == token.COMMENT: - # regular trailing comment - self.current_line.append(comment) - yield from self.line() - - else: - # regular standalone comment - yield from self.line() - - self.current_line.append(comment) - yield from self.line() - - except FormatOff as f_off: - f_off.trim_prefix(node) - yield from self.line(type=UnformattedLines) - yield from self.visit(node) - - except FormatOn as f_on: - # This only happens here if somebody says "fmt: on" multiple - # times in a row. - f_on.trim_prefix(node) - yield from self.visit_default(node) - - else: - normalize_prefix(node, inside_brackets=any_open_brackets) - if node.type == token.STRING: - normalize_string_quotes(node) - if node.type not in WHITESPACE: - self.current_line.append(node) + for comment in generate_comments(node): + if any_open_brackets: + # any comment within brackets is subject to splitting + self.current_line.append(comment) + elif comment.type == token.COMMENT: + # regular trailing comment + self.current_line.append(comment) + yield from self.line() + + else: + # regular standalone comment + yield from self.line() + + self.current_line.append(comment) + yield from self.line() + + normalize_prefix(node, inside_brackets=any_open_brackets) + if self.normalize_strings and node.type == token.STRING: + normalize_string_prefix(node, remove_u_prefix=self.remove_u_prefix) + normalize_string_quotes(node) + if node.type == token.NUMBER: + normalize_numeric_literal(node) + if node.type not in WHITESPACE: + self.current_line.append(node) yield from super().visit_default(node) def visit_INDENT(self, node: Node) -> Iterator[Line]: @@ -1060,34 +1519,60 @@ class LineGenerator(Visitor[Line]): def visit_DEDENT(self, node: Node) -> Iterator[Line]: """Decrease indentation level, maybe yield a line.""" - # DEDENT has no value. Additionally, in blib2to3 it never holds comments. + # The current line might still wait for trailing comments. At DEDENT time + # there won't be any (they would be prefixes on the preceding NEWLINE). + # Emit the line then. + yield from self.line() + + # While DEDENT has no value, its prefix may contain standalone comments + # that belong to the current indentation level. Get 'em. + yield from self.visit_default(node) + + # Finally, emit the dedent. 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`, `assert` and assignments. + + 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 a set of string leaf values immediately after which + invisible parens 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.visit(child) + def visit_suite(self, node: Node) -> Iterator[Line]: + """Visit a suite.""" + if self.is_pyi and is_stub_suite(node): + yield from self.visit(node.children[2]) + else: + yield from self.visit_default(node) + def visit_simple_stmt(self, node: Node) -> Iterator[Line]: """Visit a statement without nested statements.""" is_suite_like = node.parent and node.parent.type in STATEMENT if is_suite_like: - yield from self.line(+1) - yield from self.visit_default(node) - yield from self.line(-1) + if self.is_pyi and is_stub_body(node): + yield from self.visit_default(node) + else: + yield from self.line(+1) + yield from self.visit_default(node) + yield from self.line(-1) else: - yield from self.line() + if not self.is_pyi or not node.parent or not is_stub_suite(node.parent): + yield from self.line() yield from self.visit_default(node) def visit_async_stmt(self, node: Node) -> Iterator[Line]: @@ -1120,35 +1605,37 @@ class LineGenerator(Visitor[Line]): yield from self.visit_default(leaf) yield from self.line() - def visit_unformatted(self, node: LN) -> Iterator[Line]: - """Used when file contained a `# fmt: off`.""" - if isinstance(node, Node): - for child in node.children: - yield from self.visit(child) - - else: - try: - self.current_line.append(node) - except FormatOn as f_on: - f_on.trim_prefix(node) - yield from self.line() - yield from self.visit(node) + def visit_STANDALONE_COMMENT(self, leaf: Leaf) -> Iterator[Line]: + if not self.current_line.bracket_tracker.any_open_brackets(): + yield from self.line() + yield from self.visit_default(leaf) 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", "elif"} + ) + 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_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 +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()) @@ -1156,8 +1643,12 @@ BRACKETS = OPENING_BRACKETS | CLOSING_BRACKETS ALWAYS_NO_SPACE = CLOSING_BRACKETS | {token.COMMA, STANDALONE_COMMENT} -def whitespace(leaf: Leaf) -> str: # noqa C901 - """Return whitespace prefix if needed for the given `leaf`.""" +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 + which has non-trivial arguments, like arithmetic expressions or function calls. + """ NO = "" SPACE = " " DOUBLESPACE = " " @@ -1171,7 +1662,11 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 return DOUBLESPACE assert p is not None, f"INTERNAL ERROR: hand-made leaf without parent: {leaf!r}" - if t == token.COLON and p.type not in {syms.subscript, syms.subscriptlist}: + if t == token.COLON and p.type not in { + syms.subscript, + syms.subscriptlist, + syms.sliceop, + }: return NO prev = leaf.prev_sibling @@ -1181,12 +1676,21 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 return NO if t == token.COLON: - return SPACE if prevp.type == token.COMMA else NO + if prevp.type == token.COLON: + return NO + + elif prevp.type != token.COMMA and not complex_subscript: + return NO + + return SPACE if prevp.type == token.EQUAL: if prevp.parent: if prevp.parent.type in { - syms.arglist, syms.argument, syms.parameters, syms.varargslist + syms.arglist, + syms.argument, + syms.parameters, + syms.varargslist, }: return NO @@ -1196,24 +1700,17 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 # that, too. 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, - }: + elif prevp.type in STARS: + if is_vararg(prevp, within=VARARGS_PARENTS | UNPACKING_PARENTS): return NO elif prevp.type == token.COLON: if prevp.parent and prevp.parent.type in {syms.subscript, syms.sliceop}: - return NO + return SPACE if complex_subscript else NO elif ( prevp.parent - and prevp.parent.type in {syms.factor, syms.star_expr} + and prevp.parent.type == syms.factor and prevp.type in MATH_OPERATORS ): return NO @@ -1234,17 +1731,11 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 if p.type in {syms.parameters, syms.arglist}: # untyped function signatures or calls - if t == token.RPAR: - return NO - if not prev or prev.type != token.COMMA: return NO elif p.type == syms.varargslist: # lambdas - if t == token.RPAR: - return NO - if prev and prev.type != token.COMMA: return NO @@ -1299,7 +1790,7 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 if not prevp or prevp.type == token.LPAR: return NO - elif prev.type == token.EQUAL or prev.type == token.DOUBLESTAR: + elif prev.type in {token.EQUAL} | STARS: return NO elif p.type == syms.decorator: @@ -1321,7 +1812,7 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 if prev and prev.type == token.LPAR: return NO - elif p.type == syms.subscript: + elif p.type in {syms.subscript, syms.sliceop}: # indexing if not prev: assert p.parent is not None, "subscripts are always parented" @@ -1330,7 +1821,7 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 return NO - else: + elif not complex_subscript: return NO elif p.type == syms.atom: @@ -1338,21 +1829,9 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 # dots, but not the first one. return NO - elif ( - p.type == syms.listmaker - or p.type == syms.testlist_gexp - or p.type == syms.subscriptlist - ): - # list interior, including unpacking - if not prev: - return NO - elif p.type == syms.dictsetmaker: - # dict and set interior, including unpacking - if not prev: - return NO - - if prev.type == token.DOUBLESTAR: + # dict unpacking + if prev and prev.type == token.DOUBLESTAR: return NO elif p.type in {syms.factor, syms.star_expr}: @@ -1365,14 +1844,15 @@ def whitespace(leaf: Leaf) -> str: # noqa C901 prevp_parent = prevp.parent assert prevp_parent is not None if prevp.type == token.COLON and prevp_parent.type in { - syms.subscript, syms.sliceop + syms.subscript, + syms.sliceop, }: return NO elif prevp.type == token.EQUAL and prevp_parent.type == syms.argument: return NO - elif t == token.NAME or t == token.NUMBER: + elif t in {token.NAME, token.NUMBER, token.STRING}: return NO elif p.type == syms.import_from: @@ -1411,7 +1891,40 @@ def preceding_leaf(node: Optional[LN]) -> Optional[Leaf]: return None -def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int: +def child_towards(ancestor: Node, descendant: LN) -> Optional[LN]: + """Return the child of `ancestor` that contains `descendant`.""" + node: Optional[LN] = descendant + while node and node.parent != ancestor: + node = node.parent + return node + + +def container_of(leaf: Leaf) -> LN: + """Return `leaf` or one of its ancestors that is the topmost container of it. + + By "container" we mean a node where `leaf` is the very first child. + """ + same_prefix = leaf.prefix + container: LN = leaf + while container: + parent = container.parent + if parent is None: + break + + if parent.children[0].prefix != same_prefix: + break + + if parent.type == syms.file_input: + break + + if parent.prev_sibling is not None and parent.prev_sibling.type in BRACKETS: + break + + container = parent + return container + + +def is_split_after_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int: """Return the priority of the `leaf` delimiter, given a line break after it. The delimiter priorities returned here are from those delimiters that would @@ -1422,30 +1935,36 @@ def is_split_after_delimiter(leaf: Leaf, previous: Leaf = None) -> int: 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 -def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int: - """Return the priority of the `leaf` delimiter, given a line before after it. +def is_split_before_delimiter(leaf: Leaf, previous: Optional[Leaf] = None) -> int: + """Return the priority of the `leaf` delimiter, given a line break before it. The delimiter priorities returned here are from those delimiters that would cause a line break before themselves. Higher numbers are higher priority. """ + if is_vararg(leaf, within=VARARGS_PARENTS | UNPACKING_PARENTS): + # * 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 == token.DOT + and leaf.parent + and leaf.parent.type not in {syms.import_from, syms.dotted_name} + and (previous is None or previous.type in CLOSING_BRACKETS) + ): + return DOT_PRIORITY + if ( leaf.type in MATH_OPERATORS and leaf.parent and leaf.parent.type not in {syms.factor, syms.star_expr} ): - return MATH_PRIORITY + return MATH_PRIORITIES[leaf.type] if leaf.type in COMPARATORS: return COMPARATOR_PRIORITY @@ -1457,40 +1976,69 @@ def is_split_before_delimiter(leaf: Leaf, previous: Leaf = None) -> int: ): return STRING_PRIORITY + if leaf.type not in {token.NAME, token.ASYNC}: + return 0 + if ( - leaf.type == token.NAME - and leaf.value == "for" + leaf.value == "for" and leaf.parent and leaf.parent.type in {syms.comp_for, syms.old_comp_for} + or leaf.type == token.ASYNC ): - return COMPREHENSION_PRIORITY + if ( + not isinstance(leaf.prev_sibling, Leaf) + or leaf.prev_sibling.value != "async" + ): + return COMPREHENSION_PRIORITY if ( - leaf.type == token.NAME - and leaf.value == "if" + leaf.value == "if" and leaf.parent and leaf.parent.type in {syms.comp_if, syms.old_comp_if} ): return COMPREHENSION_PRIORITY - if leaf.type == token.NAME and leaf.value in LOGIC_OPERATORS and leaf.parent: + if leaf.value in {"if", "else"} and leaf.parent and leaf.parent.type == syms.test: + return TERNARY_PRIORITY + + if leaf.value == "is": + return COMPARATOR_PRIORITY + + if ( + leaf.value == "in" + and leaf.parent + and leaf.parent.type in {syms.comp_op, syms.comparison} + and not ( + previous is not None + and previous.type == token.NAME + and previous.value == "not" + ) + ): + return COMPARATOR_PRIORITY + + if ( + leaf.value == "not" + and leaf.parent + and leaf.parent.type == syms.comp_op + and not ( + previous is not None + and previous.type == token.NAME + and previous.value == "is" + ) + ): + return COMPARATOR_PRIORITY + + if leaf.value in LOGIC_OPERATORS and leaf.parent: return LOGIC_PRIORITY return 0 -def is_delimiter(leaf: Leaf, previous: Leaf = None) -> int: - """Return the priority of the `leaf` delimiter. Return 0 if not delimiter. - - Higher numbers are higher priority. - """ - return max( - is_split_before_delimiter(leaf, previous), - is_split_after_delimiter(leaf, previous), - ) +FMT_OFF = {"# fmt: off", "# fmt:off", "# yapf: disable"} +FMT_ON = {"# fmt: on", "# fmt:on", "# yapf: enable"} -def generate_comments(leaf: Leaf) -> Iterator[Leaf]: +def generate_comments(leaf: LN) -> Iterator[Leaf]: """Clean the prefix of the `leaf` and generate comments from it, if any. Comments in lib2to3 are shoved into the whitespace prefix. This happens @@ -1509,16 +2057,38 @@ def generate_comments(leaf: Leaf) -> Iterator[Leaf]: Inline comments are emitted as regular token.COMMENT leaves. Standalone are emitted with a fake STANDALONE_COMMENT token identifier. """ - p = leaf.prefix - if not p: - return + for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER): + yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines) - if "#" not in p: - return + +@dataclass +class ProtoComment: + """Describes a piece of syntax that is a comment. + + It's not a :class:`blib2to3.pytree.Leaf` so that: + + * it can be cached (`Leaf` objects should not be reused more than once as + they store their lineno, column, prefix, and parent information); + * `newlines` and `consumed` fields are kept separate from the `value`. This + simplifies handling of special marker comments like ``# fmt: off/on``. + """ + + type: int # token.COMMENT or STANDALONE_COMMENT + value: str # content of the comment + newlines: int # how many newlines before the comment + consumed: int # how many characters of the original leaf's prefix did we consume + + +@lru_cache(maxsize=4096) +def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]: + """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`.""" + result: List[ProtoComment] = [] + if not prefix or "#" not in prefix: + return result consumed = 0 nlines = 0 - for index, line in enumerate(p.split("\n")): + 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: @@ -1526,27 +2096,25 @@ def generate_comments(leaf: Leaf) -> Iterator[Leaf]: if not line.startswith("#"): continue - if index == 0 and leaf.type != token.ENDMARKER: + if index == 0 and not is_endmarker: comment_type = token.COMMENT # simple trailing comment else: comment_type = STANDALONE_COMMENT comment = make_comment(line) - yield Leaf(comment_type, comment, prefix="\n" * nlines) - - if comment in {"# fmt: on", "# yapf: enable"}: - raise FormatOn(consumed) - - if comment in {"# fmt: off", "# yapf: disable"}: - raise FormatOff(consumed) - + result.append( + ProtoComment( + type=comment_type, value=comment, newlines=nlines, consumed=consumed + ) + ) nlines = 0 + return result def make_comment(content: str) -> str: """Return a consistently formatted comment from the given `content` string. - All comments (except for "##", "#!", "#:") should have a single space between - the hash sign and the content. + All comments (except for "##", "#!", "#:", '#'", "#%%") should have a single + space between the hash sign and the content. If `content` didn't start with a hash sign, one is provided. """ @@ -1556,13 +2124,16 @@ def make_comment(content: str) -> str: if content[0] == "#": content = content[1:] - if content and content[0] not in " !:#": + if content and content[0] not in " !:#'%": content = " " + content return "#" + content def split_line( - line: Line, line_length: int, inner: bool = False, py36: bool = False + line: Line, + line_length: int, + inner: bool = False, + supports_trailing_commas: bool = False, ) -> Iterator[Line]: """Split a `line` into potentially many lines. @@ -1571,18 +2142,26 @@ 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. + If `supports_trailing_commas` is True, splitting may use the TRAILING_COMMA feature. """ - if isinstance(line, UnformattedLines) or line.is_comment: + if line.is_comment: yield line return line_str = str(line).strip("\n") + + # we don't want to split special comments like type annotations + # https://github.com/python/typing/issues/186 + has_special_comment = False + for leaf in line.leaves: + for comment in line.comments_after(leaf): + if leaf.type == token.COMMA and is_special_comment(comment): + has_special_comment = True + if ( - len(line_str) <= line_length - and "\n" not in line_str # multiline strings - and not line.contains_standalone_comments + not has_special_comment + and not line.should_explode + and is_line_short_enough(line, line_length=line_length, line_str=line_str) ): yield line return @@ -1590,24 +2169,47 @@ def split_line( split_funcs: List[SplitFunc] if line.is_def: split_funcs = [left_hand_split] - elif line.inside_brackets: - split_funcs = [delimiter_split, standalone_comment_split, right_hand_split] else: - split_funcs = [right_hand_split] + + def rhs(line: Line, supports_trailing_commas: bool = False) -> Iterator[Line]: + for omit in generate_trailers_to_omit(line, line_length): + lines = list( + right_hand_split( + line, line_length, supports_trailing_commas, omit=omit + ) + ) + if is_line_short_enough(lines[0], line_length=line_length): + yield from lines + return + + # All splits failed, best effort split with no omits. + # This mostly happens to multiline strings that are by definition + # reported as not fitting a single line. + yield from right_hand_split(line, supports_trailing_commas) + + if line.inside_brackets: + split_funcs = [delimiter_split, standalone_comment_split, rhs] + else: + split_funcs = [rhs] for split_func in split_funcs: # We are accumulating lines in `result` because we might want to abort # mission and return the original line in the end, or attempt a different # split altogether. result: List[Line] = [] try: - for l in split_func(line, py36): + for l in split_func(line, supports_trailing_commas): 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, + supports_trailing_commas=supports_trailing_commas, + ) ) - except CannotSplit as cs: + except CannotSplit: continue else: @@ -1618,15 +2220,15 @@ def split_line( yield line -def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]: +def left_hand_split( + line: Line, supports_trailing_commas: 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. - Prefer RHS otherwise. + Prefer RHS otherwise. This is why this function is not symmetrical with + :func:`right_hand_split` which also handles optional parentheses. """ - head = Line(depth=line.depth) - body = Line(depth=line.depth + 1, inside_brackets=True) - tail = Line(depth=line.depth) tail_leaves: List[Leaf] = [] body_leaves: List[Leaf] = [] head_leaves: List[Leaf] = [] @@ -1644,57 +2246,107 @@ def left_hand_split(line: Line, py36: bool = False) -> Iterator[Line]: if leaf.type in OPENING_BRACKETS: matching_bracket = leaf current_leaves = body_leaves - # Since body is a new indent level, remove spurious leading whitespace. - 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 leaf in leaves: - result.append(leaf, preformatted=True) - for comment_after in line.comments_after(leaf): - result.append(comment_after, preformatted=True) + if not matching_bracket: + raise CannotSplit("No brackets found") + + head = bracket_split_build_line(head_leaves, line, matching_bracket) + body = bracket_split_build_line(body_leaves, line, matching_bracket, is_body=True) + tail = bracket_split_build_line(tail_leaves, line, matching_bracket) bracket_split_succeeded_or_raise(head, body, tail) for result in (head, body, tail): if result: yield result -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) - tail = Line(depth=line.depth) +def right_hand_split( + line: Line, + line_length: int, + supports_trailing_commas: bool = False, + omit: Collection[LeafID] = (), +) -> Iterator[Line]: + """Split line into many lines, starting with the last matching bracket pair. + + If the split was by optional parentheses, attempt splitting without them, too. + `omit` is a collection of closing bracket IDs that shouldn't be considered for + this split. + + Note: running this function modifies `bracket_depth` on the leaves of `line`. + """ tail_leaves: List[Leaf] = [] body_leaves: List[Leaf] = [] 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 + if not (opening_bracket and closing_bracket and head_leaves): + # If there is no opening or closing_bracket that means the split failed and + # all content is in the tail. Otherwise, if `head_leaves` are empty, it means + # the matching `opening_bracket` wasn't available on `line` anymore. + raise CannotSplit("No brackets found") + tail_leaves.reverse() body_leaves.reverse() head_leaves.reverse() - # Since body is a new indent level, remove spurious leading whitespace. - 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 leaf in leaves: - result.append(leaf, preformatted=True) - for comment_after in line.comments_after(leaf): - result.append(comment_after, preformatted=True) + head = bracket_split_build_line(head_leaves, line, opening_bracket) + body = bracket_split_build_line(body_leaves, line, opening_bracket, is_body=True) + tail = bracket_split_build_line(tail_leaves, line, opening_bracket) bracket_split_succeeded_or_raise(head, body, tail) + if ( + # the body shouldn't be exploded + not body.should_explode + # the opening bracket is an optional paren + and opening_bracket.type == token.LPAR + and not opening_bracket.value + # the closing bracket is an optional paren + and closing_bracket.type == token.RPAR + and not closing_bracket.value + # it's not an import (optional parens are the only thing we can split on + # in this case; attempting a split without them is a waste of time) + and not line.is_import + # there are no standalone comments in the body + and not body.contains_standalone_comments(0) + # and we can actually remove the parens + and can_omit_invisible_parens(body, line_length) + ): + omit = {id(closing_bracket), *omit} + try: + yield from right_hand_split( + line, + line_length, + supports_trailing_commas=supports_trailing_commas, + omit=omit, + ) + return + + except CannotSplit: + if not ( + can_be_split(body) + or is_line_short_enough(body, line_length=line_length) + ): + raise CannotSplit( + "Splitting failed, body is still too long and can't be split." + ) + + elif head.contains_multiline_strings() or tail.contains_multiline_strings(): + raise CannotSplit( + "The current optional pair of parentheses is bound to fail to " + "satisfy the splitting algorithm because the head or the tail " + "contains multiline strings which by definition never fit one " + "line." + ) + + ensure_visible(opening_bracket) + ensure_visible(closing_bracket) for result in (head, body, tail): if result: yield result @@ -1726,6 +2378,35 @@ def bracket_split_succeeded_or_raise(head: Line, body: Line, tail: Line) -> None ) +def bracket_split_build_line( + leaves: List[Leaf], original: Line, opening_bracket: Leaf, *, is_body: bool = False +) -> Line: + """Return a new line with given `leaves` and respective comments from `original`. + + If `is_body` is True, the result line is one-indented inside brackets and as such + has its first leaf's prefix normalized and a trailing comma added when expected. + """ + result = Line(depth=original.depth) + if is_body: + result.inside_brackets = True + result.depth += 1 + 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, ",")) + # Populate the line + for leaf in leaves: + result.append(leaf, preformatted=True) + for comment_after in original.comments_after(leaf): + result.append(comment_after, preformatted=True) + if is_body: + result.should_explode = should_explode(result, opening_bracket) + return result + + def dont_increase_indentation(split_func: SplitFunc) -> SplitFunc: """Normalize prefix of the first leaf in every line returned by `split_func`. @@ -1733,8 +2414,10 @@ 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, supports_trailing_commas: bool = False + ) -> Iterator[Line]: + for l in split_func(line, supports_trailing_commas): normalize_prefix(l.leaves[0], inside_brackets=True) yield l @@ -1742,25 +2425,29 @@ 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, supports_trailing_commas: bool = False +) -> 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 `supports_trailing_commas` is True, the split will add trailing commas + also in function signatures that contain `*` and `**`. """ try: last_leaf = line.leaves[-1] except IndexError: raise CannotSplit("Line empty") - delimiters = line.bracket_tracker.delimiters + bt = line.bracket_tracker try: - delimiter_priority = line.bracket_tracker.max_delimiter_priority( - exclude={id(last_leaf)} - ) + delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)}) except ValueError: raise CannotSplit("No delimiters found") + if delimiter_priority == DOT_PRIORITY: + if bt.delimiter_count_with_priority(delimiter_priority) == 1: + raise CannotSplit("Splitting a single attribute from its owner looks wrong") + current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets) lowest_depth = sys.maxsize trailing_comma_safe = True @@ -1770,7 +2457,7 @@ def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]: nonlocal current_line try: current_line.append_safe(leaf, preformatted=True) - except ValueError as ve: + except ValueError: yield current_line current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets) @@ -1783,36 +2470,32 @@ def delimiter_split(line: Line, py36: bool = False) -> Iterator[Line]: yield from append_to_line(comment_after) lowest_depth = min(lowest_depth, leaf.bracket_depth) - if ( - leaf.bracket_depth == lowest_depth - and leaf.type == token.STAR - or leaf.type == token.DOUBLESTAR + if leaf.bracket_depth == lowest_depth and is_vararg( + leaf, within=VARARGS_PARENTS ): - trailing_comma_safe = trailing_comma_safe and py36 - leaf_priority = delimiters.get(id(leaf)) + trailing_comma_safe = trailing_comma_safe and supports_trailing_commas + leaf_priority = bt.delimiters.get(id(leaf)) if leaf_priority == delimiter_priority: yield current_line 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]: +def standalone_comment_split( + line: Line, supports_trailing_commas: 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) @@ -1822,7 +2505,7 @@ def standalone_comment_split(line: Line, py36: bool = False) -> Iterator[Line]: nonlocal current_line try: current_line.append_safe(leaf, preformatted=True) - except ValueError as ve: + except ValueError: yield current_line current_line = Line(depth=line.depth, inside_brackets=line.inside_brackets) @@ -1852,6 +2535,16 @@ def is_import(leaf: Leaf) -> bool: ) +def is_special_comment(leaf: Leaf) -> 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 bool( + (t == token.COMMENT or t == STANDALONE_COMMENT) and (v.startswith("# type:")) + ) + + def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None: """Leave existing extra newlines if not `inside_brackets`. Remove everything else. @@ -1870,6 +2563,22 @@ def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None: leaf.prefix = "" +def normalize_string_prefix(leaf: Leaf, remove_u_prefix: bool = False) -> None: + """Make all string prefixes lowercase. + + If remove_u_prefix is given, also removes any u prefix from the string. + + Note: Mutates its argument. + """ + match = re.match(r"^([furbFURB]*)(.*)$", leaf.value, re.DOTALL) + assert match is not None, f"failed to match string {leaf.value!r}" + orig_prefix = match.group(1) + new_prefix = orig_prefix.lower() + if remove_u_prefix: + new_prefix = new_prefix.replace("u", "") + leaf.value = f"{new_prefix}{match.group(2)}" + + def normalize_string_quotes(leaf: Leaf) -> None: """Prefer double quotes but only if it doesn't cause more escaping. @@ -1895,11 +2604,35 @@ def normalize_string_quotes(leaf: Leaf) -> None: if first_quote_pos == -1: return # There's an internal error - 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}" - ) - if new_quote == '"""' and new_body[-1] == '"': + prefix = leaf.value[:first_quote_pos] + unescaped_new_quote = re.compile(rf"(([^\\]|^)(\\\\)*){new_quote}") + escaped_new_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){new_quote}") + escaped_orig_quote = re.compile(rf"([^\\]|^)\\((?:\\\\)*){orig_quote}") + body = leaf.value[first_quote_pos + len(orig_quote) : -len(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: + # remove unnecessary escapes + new_body = sub_twice(escaped_new_quote, rf"\1\2{new_quote}", body) + if body != new_body: + # Consider the string without unnecessary escapes as the original + body = new_body + leaf.value = f"{prefix}{orig_quote}{body}{orig_quote}" + 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) + for m in matches: + if "\\" in str(m): + # Do not introduce backslashes in interpolated expressions + return + if new_quote == '"""' and new_body[-1:] == '"': # edge case: new_body = new_body[:-1] + '\\"' orig_escape_count = body.count("\\") @@ -1910,72 +2643,618 @@ def normalize_string_quotes(leaf: Leaf) -> None: 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 is_python36(node: Node) -> bool: - """Return True if the current file is using Python 3.6+ features. +def normalize_numeric_literal(leaf: Leaf) -> None: + """Normalizes numeric (float, int, and complex) literals. - Currently looking for: - - f-strings; and - - trailing commas after * or ** in function signatures. + All letters used in the representation are normalized to lowercase (except + in Python 2 long literals). """ - 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 + text = leaf.value.lower() + if text.startswith(("0o", "0b")): + # Leave octal and binary literals alone. + pass + elif text.startswith("0x"): + # Change hex literals to upper case. + before, after = text[:2], text[2:] + text = f"{before}{after.upper()}" + elif "e" in text: + before, after = text.split("e") + sign = "" + if after.startswith("-"): + after = after[1:] + sign = "-" + elif after.startswith("+"): + after = after[1:] + before = format_float_or_int_string(before) + text = f"{before}e{sign}{after}" + elif text.endswith(("j", "l")): + number = text[:-1] + suffix = text[-1] + # Capitalize in "2L" because "l" looks too similar to "1". + if suffix == "l": + suffix = "L" + text = f"{format_float_or_int_string(number)}{suffix}" + else: + text = format_float_or_int_string(text) + leaf.value = text - 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 +def format_float_or_int_string(text: str) -> str: + """Formats a float string like "1.0".""" + if "." not in text: + return text + before, after = text.split(".") + return f"{before or 0}.{after or 0}" -PYTHON_EXTENSIONS = {".py"} -BLACKLISTED_DIRECTORIES = { - "build", "buck-out", "dist", "_build", ".git", ".hg", ".mypy_cache", ".tox", ".venv" -} +def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: + """Make existing optional parentheses invisible or create new ones. -def gen_python_files_in_dir(path: Path) -> Iterator[Path]: - """Generate all files under `path` which aren't under BLACKLISTED_DIRECTORIES - and have one of the PYTHON_EXTENSIONS. + `parens_after` is a set of string leaf values immeditely after which parens + should be put. + + Standardizes on visible parentheses for single-element tuples, and keeps + existing visible parentheses for other tuples and generator expressions. + """ + for pc in list_comments(node.prefix, is_endmarker=False): + if pc.value in FMT_OFF: + # This `node` has a prefix with `# fmt: off`, don't mess with parens. + return + + check_lpar = False + for index, child in enumerate(list(node.children)): + if check_lpar: + if child.type == syms.atom: + if maybe_make_parens_invisible_in_atom(child): + lpar = Leaf(token.LPAR, "") + rpar = Leaf(token.RPAR, "") + index = child.remove() or 0 + node.insert_child(index, Node(syms.atom, [lpar, child, rpar])) + elif is_one_tuple(child): + # wrap child in visible parentheses + lpar = Leaf(token.LPAR, "(") + rpar = Leaf(token.RPAR, ")") + child.remove() + node.insert_child(index, Node(syms.atom, [lpar, child, rpar])) + elif node.type == syms.import_from: + # "import from" nodes store parentheses directly as part of + # the statement + if child.type == token.LPAR: + # make parentheses invisible + child.value = "" # type: ignore + node.children[-1].value = "" # type: ignore + elif child.type != token.STAR: + # insert invisible parentheses + node.insert_child(index, Leaf(token.LPAR, "")) + node.append_child(Leaf(token.RPAR, "")) + break + + elif not (isinstance(child, Leaf) and is_multiline_string(child)): + # wrap child in invisible parentheses + lpar = Leaf(token.LPAR, "") + 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 normalize_fmt_off(node: Node) -> None: + """Convert content between `# fmt: off`/`# fmt: on` into standalone comments.""" + try_again = True + while try_again: + try_again = convert_one_fmt_off_pair(node) + + +def convert_one_fmt_off_pair(node: Node) -> bool: + """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment. + + Returns True if a pair was converted. + """ + for leaf in node.leaves(): + previous_consumed = 0 + for comment in list_comments(leaf.prefix, is_endmarker=False): + if comment.value in FMT_OFF: + # We only want standalone comments. If there's no previous leaf or + # the previous leaf is indentation, it's a standalone comment in + # disguise. + if comment.type != STANDALONE_COMMENT: + prev = preceding_leaf(leaf) + if prev and prev.type not in WHITESPACE: + continue + + ignored_nodes = list(generate_ignored_nodes(leaf)) + if not ignored_nodes: + continue + + first = ignored_nodes[0] # Can be a container node with the `leaf`. + parent = first.parent + prefix = first.prefix + first.prefix = prefix[comment.consumed :] + hidden_value = ( + comment.value + "\n" + "".join(str(n) for n in ignored_nodes) + ) + if hidden_value.endswith("\n"): + # That happens when one of the `ignored_nodes` ended with a NEWLINE + # leaf (possibly followed by a DEDENT). + hidden_value = hidden_value[:-1] + first_idx = None + for ignored in ignored_nodes: + index = ignored.remove() + if first_idx is None: + first_idx = index + assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)" + assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)" + parent.insert_child( + first_idx, + Leaf( + STANDALONE_COMMENT, + hidden_value, + prefix=prefix[:previous_consumed] + "\n" * comment.newlines, + ), + ) + return True + + previous_consumed = comment.consumed + + return False + + +def generate_ignored_nodes(leaf: Leaf) -> Iterator[LN]: + """Starting from the container of `leaf`, generate all leaves until `# fmt: on`. + + Stops at the end of the block. + """ + container: Optional[LN] = container_of(leaf) + while container is not None and container.type != token.ENDMARKER: + for comment in list_comments(container.prefix, is_endmarker=False): + if comment.value in FMT_ON: + return + + yield container + + container = container.next_sibling + + +def maybe_make_parens_invisible_in_atom(node: LN) -> bool: + """If it's safe, make the parens in the atom `node` invisible, recursively. + + Returns whether the node should itself be wrapped in invisible parentheses. + + """ + if ( + node.type != syms.atom + or is_empty_tuple(node) + or is_one_tuple(node) + or is_yield(node) + or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY + ): + return False + + first = node.children[0] + last = node.children[-1] + if first.type == token.LPAR and last.type == token.RPAR: + # 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]) + return False + + return True + + +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 is_yield(node: LN) -> bool: + """Return True if `node` holds a `yield` or `yield from` expression.""" + if node.type == syms.yield_expr: + return True + + if node.type == token.NAME and node.value == "yield": # type: ignore + return True + + if node.type != syms.atom: + return False + + if len(node.children) != 3: + return False + + lpar, expr, rpar = node.children + if lpar.type == token.LPAR and rpar.type == token.RPAR: + return is_yield(expr) + + return False + + +def is_vararg(leaf: Leaf, within: Set[NodeType]) -> bool: + """Return True if `leaf` is a star or double star in a vararg or kwarg. + + If `within` includes VARARGS_PARENTS, this applies to function signatures. + If `within` includes UNPACKING_PARENTS, it applies to right hand-side + extended iterable unpacking (PEP 3132) and additional unpacking + generalizations (PEP 448). """ + if leaf.type not in STARS or not leaf.parent: + return False + + p = leaf.parent + if p.type == syms.star_expr: + # Star expressions are also used as assignment targets in extended + # iterable unpacking (PEP 3132). See what its parent is instead. + if not p.parent: + return False + + p = p.parent + + return p.type in within + + +def is_multiline_string(leaf: Leaf) -> bool: + """Return True if `leaf` is a multiline string that actually spans many lines.""" + value = leaf.value.lstrip("furbFURB") + return value[:3] in {'"""', "'''"} and "\n" in value + + +def is_stub_suite(node: Node) -> bool: + """Return True if `node` is a suite with a stub body.""" + if ( + len(node.children) != 4 + or node.children[0].type != token.NEWLINE + or node.children[1].type != token.INDENT + or node.children[3].type != token.DEDENT + ): + return False + + return is_stub_body(node.children[2]) + + +def is_stub_body(node: LN) -> bool: + """Return True if `node` is a simple statement containing an ellipsis.""" + if not isinstance(node, Node) or node.type != syms.simple_stmt: + return False + + if len(node.children) != 2: + return False + + child = node.children[0] + return ( + child.type == syms.atom + and len(child.children) == 3 + and all(leaf == Leaf(token.DOT, ".") for leaf in child.children) + ) + + +def max_delimiter_priority_in_atom(node: LN) -> int: + """Return maximum delimiter priority inside `node`. + + This is specific to atoms with contents contained in a pair of parentheses. + If `node` isn't an atom or there are no enclosing parentheses, returns 0. + """ + 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 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} + and opening_bracket.value in "[{(" + ): + return False + + try: + last_leaf = line.leaves[-1] + exclude = {id(last_leaf)} if last_leaf.type == token.COMMA else set() + max_priority = line.bracket_tracker.max_delimiter_priority(exclude=exclude) + except (IndexError, ValueError): + return False + + return max_priority == COMMA_PRIORITY + + +def 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. + """ + 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"}: + features.add(Feature.F_STRINGS) + + elif n.type == token.NUMBER: + if "_" in n.value: # type: ignore + features.add(Feature.NUMERIC_UNDERSCORES) + + elif ( + n.type in {syms.typedargslist, syms.arglist} + and n.children + and n.children[-1].type == token.COMMA + ): + for ch in n.children: + if ch.type in STARS: + features.add(Feature.TRAILING_COMMA) + + if ch.type == syms.argument: + for argch in ch.children: + if argch.type in STARS: + features.add(Feature.TRAILING_COMMA) + + return features + + +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]]: + """Generate sets of closing bracket IDs that should be omitted in a RHS. + + Brackets can be omitted if the entire trailer up to and including + a preceding closing bracket fits in one line. + + Yielded sets are cumulative (contain results of previous yields, too). First + set is empty. + """ + + omit: Set[LeafID] = set() + yield omit + + length = 4 * line.depth + opening_bracket = None + closing_bracket = None + inner_brackets: Set[LeafID] = set() + for index, leaf, leaf_length in enumerate_with_length(line, reversed=True): + length += leaf_length + if length > line_length: + break + + has_inline_comment = leaf_length > len(leaf.value) + len(leaf.prefix) + if leaf.type == STANDALONE_COMMENT or has_inline_comment: + break + + if opening_bracket: + if leaf is opening_bracket: + opening_bracket = None + elif leaf.type in CLOSING_BRACKETS: + inner_brackets.add(id(leaf)) + elif leaf.type in CLOSING_BRACKETS: + if index > 0 and line.leaves[index - 1].type in OPENING_BRACKETS: + # Empty brackets would fail a split so treat them as "inner" + # brackets (e.g. only add them to the `omit` set if another + # pair of brackets was good enough. + inner_brackets.add(id(leaf)) + continue + + if closing_bracket: + omit.add(id(closing_bracket)) + omit.update(inner_brackets) + inner_brackets.clear() + yield omit + + if leaf.value: + opening_bracket = leaf.opening_bracket + closing_bracket = leaf + + +def get_future_imports(node: Node) -> Set[str]: + """Return a set of __future__ imports in the file.""" + imports: Set[str] = set() + + def get_imports_from_children(children: List[LN]) -> Generator[str, None, None]: + for child in children: + if isinstance(child, Leaf): + if child.type == token.NAME: + yield child.value + elif child.type == syms.import_as_name: + orig_name = child.children[0] + assert isinstance(orig_name, Leaf), "Invalid syntax parsing imports" + assert orig_name.type == token.NAME, "Invalid syntax parsing imports" + yield orig_name.value + elif child.type == syms.import_as_names: + yield from get_imports_from_children(child.children) + else: + assert False, "Invalid syntax parsing imports" + + for child in node.children: + if child.type != syms.simple_stmt: + break + first_child = child.children[0] + if isinstance(first_child, Leaf): + # Continue looking if we see a docstring; otherwise stop. + if ( + len(child.children) == 2 + and first_child.type == token.STRING + and child.children[1].type == token.NEWLINE + ): + continue + else: + break + elif first_child.type == syms.import_from: + module_name = first_child.children[1] + if not isinstance(module_name, Leaf) or module_name.value != "__future__": + break + imports |= set(get_imports_from_children(first_child.children[3:])) + else: + break + return imports + + +def gen_python_files_in_dir( + path: Path, + root: Path, + include: Pattern[str], + exclude: Pattern[str], + report: "Report", +) -> Iterator[Path]: + """Generate all files under `path` whose paths are not excluded by the + `exclude` regex, but are included by the `include` regex. + + Symbolic links pointing outside of the `root` directory are ignored. + + `report` is where output about exclusions goes. + """ + assert root.is_absolute(), f"INTERNAL ERROR: `root` must be absolute but is {root}" for child in path.iterdir(): - if child.is_dir(): - if child.name in BLACKLISTED_DIRECTORIES: + try: + normalized_path = "/" + child.resolve().relative_to(root).as_posix() + except ValueError: + if child.is_symlink(): + report.path_ignored( + child, f"is a symbolic link that points outside {root}" + ) continue - yield from gen_python_files_in_dir(child) + raise + + 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) - elif child.suffix in PYTHON_EXTENSIONS: - yield child + elif child.is_file(): + include_match = include.search(normalized_path) + if include_match: + yield child + + +@lru_cache() +def find_project_root(srcs: Iterable[str]) -> Path: + """Return a directory containing .git, .hg, or pyproject.toml. + + That directory can be one of the directories passed in `srcs` or their + common parent. + + If no directory in the tree contains a marker that would specify it's the + project root, the root of the file system is returned. + """ + if not srcs: + return Path("/").resolve() + + common_base = min(Path(src).resolve() for src in srcs) + if common_base.is_dir(): + # Append a fake file so `parents` below returns `common_base_dir`, too. + common_base /= "fake-file" + for directory in common_base.parents: + if (directory / ".git").is_dir(): + return directory + + if (directory / ".hg").is_dir(): + return directory + + if (directory / "pyproject.toml").is_file(): + return directory + + return directory @dataclass class Report: """Provides a reformatting counter. Can be rendered with `str(report)`.""" + check: bool = False + quiet: bool = False + verbose: bool = False change_count: int = 0 same_count: int = 0 failure_count: int = 0 - def done(self, src: Path, changed: bool) -> None: + def done(self, src: Path, changed: Changed) -> None: """Increment the counter for successful reformatting. Write out a message.""" - if changed: + if changed is Changed.YES: reformatted = "would reformat" if self.check else "reformatted" - out(f"{reformatted} {src}") + if self.verbose or not self.quiet: + out(f"{reformatted} {src}") self.change_count += 1 else: - out(f"{src} already well formatted, good job.", bold=False) + if self.verbose: + if changed is Changed.NO: + msg = f"{src} already well formatted, good job." + else: + msg = f"{src} wasn't modified on disk since last run." + out(msg, bold=False) self.same_count += 1 def failed(self, src: Path, message: str) -> None: @@ -1983,6 +3262,10 @@ class Report: err(f"error: cannot format {src}: {message}") self.failure_count += 1 + def path_ignored(self, path: Path, message: str) -> None: + if self.verbose: + out(f"{path} ignored: {message}", bold=False) + @property def return_code(self) -> int: """Return the exit code that the app should use. @@ -1993,7 +3276,7 @@ class Report: - otherwise return 0. """ # According to http://tldp.org/LDP/abs/html/exitcodes.html starting with - # 126 we have special returncodes reserved by the shell. + # 126 we have special return codes reserved by the shell. if self.failure_count: return 123 @@ -2052,7 +3335,16 @@ 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) + and isinstance(item, ast.Tuple) + ): + for item in item.elts: + yield from _v(item, depth + 2) + elif isinstance(item, ast.AST): yield from _v(item, depth + 2) elif isinstance(value, ast.AST): @@ -2095,9 +3387,9 @@ def assert_equivalent(src: str, dst: str) -> None: ) from None -def assert_stable(src: str, dst: str, line_length: int) -> 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) + newdst = format_str(dst, mode=mode) if dst != newdst: log = dump_to_file( diff(src, dst, "source", "first pass"), @@ -2116,7 +3408,7 @@ def dump_to_file(*output: str) -> str: import tempfile with tempfile.NamedTemporaryFile( - mode="w", prefix="blk_", suffix=".log", delete=False + mode="w", prefix="blk_", suffix=".log", delete=False, encoding="utf8" ) as f: for lines in output: f.write(lines) @@ -2136,7 +3428,7 @@ def diff(a: str, b: str, a_name: str, b_name: str) -> str: ) -def cancel(tasks: List[asyncio.Task]) -> None: +def cancel(tasks: Iterable[asyncio.Task]) -> None: """asyncio signal handler that cancels all `tasks` and reports to stderr.""" err("Aborted!") for task in tasks: @@ -2165,5 +3457,279 @@ def shutdown(loop: BaseEventLoop) -> None: loop.close() -if __name__ == "__main__": +def sub_twice(regex: Pattern[str], replacement: str, original: str) -> str: + """Replace `regex` with `replacement` twice on `original`. + + This is used by string normalization to perform replaces on + overlapping matches. + """ + return regex.sub(replacement, regex.sub(replacement, original)) + + +def re_compile_maybe_verbose(regex: str) -> Pattern[str]: + """Compile a regular expression string in `regex`. + + If it contains newlines, use verbose mode. + """ + if "\n" in regex: + regex = "(?x)" + regex + return re.compile(regex) + + +def enumerate_reversed(sequence: Sequence[T]) -> Iterator[Tuple[Index, T]]: + """Like `reversed(enumerate(sequence))` if that were possible.""" + index = len(sequence) - 1 + for element in reversed(sequence): + yield (index, element) + index -= 1 + + +def enumerate_with_length( + line: Line, reversed: bool = False +) -> Iterator[Tuple[Index, Leaf, int]]: + """Return an enumeration of leaves with their length. + + Stops prematurely on multiline strings and standalone comments. + """ + op = cast( + Callable[[Sequence[Leaf]], Iterator[Tuple[Index, Leaf]]], + enumerate_reversed if reversed else enumerate, + ) + for index, leaf in op(line.leaves): + length = len(leaf.prefix) + len(leaf.value) + if "\n" in leaf.value: + return # Multiline strings, we can't continue. + + comment: Optional[Leaf] + for comment in line.comments_after(leaf): + length += len(comment.value) + + yield index, leaf, length + + +def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool: + """Return True if `line` is no longer than `line_length`. + + Uses the provided `line_str` rendering, if any, otherwise computes a new one. + """ + if not line_str: + line_str = str(line).strip("\n") + return ( + len(line_str) <= line_length + and "\n" not in line_str # multiline strings + and not line.contains_standalone_comments() + ) + + +def can_be_split(line: Line) -> bool: + """Return False if the line cannot be split *for sure*. + + This is not an exhaustive search but a cheap heuristic that we can use to + avoid some unfortunate formattings (mostly around wrapping unsplittable code + in unnecessary parentheses). + """ + leaves = line.leaves + if len(leaves) < 2: + return False + + if leaves[0].type == token.STRING and leaves[1].type == token.DOT: + call_count = 0 + dot_count = 0 + next = leaves[-1] + for leaf in leaves[-2::-1]: + if leaf.type in OPENING_BRACKETS: + if next.type not in CLOSING_BRACKETS: + return False + + call_count += 1 + elif leaf.type == token.DOT: + dot_count += 1 + elif leaf.type == token.NAME: + if not (next.type == token.DOT or next.type in OPENING_BRACKETS): + return False + + elif leaf.type not in CLOSING_BRACKETS: + return False + + if dot_count > 1 and call_count > 1: + return False + + return True + + +def can_omit_invisible_parens(line: Line, line_length: int) -> bool: + """Does `line` have a shape safe to reformat without optional parens around it? + + Returns True for only a subset of potentially nice looking formattings but + the point is to not return false positives that end up producing lines that + are too long. + """ + bt = line.bracket_tracker + if not bt.delimiters: + # Without delimiters the optional parentheses are useless. + return True + + max_priority = bt.max_delimiter_priority() + if bt.delimiter_count_with_priority(max_priority) > 1: + # With more than one delimiter of a kind the optional parentheses read better. + return False + + if max_priority == DOT_PRIORITY: + # A single stranded method call doesn't require optional parentheses. + return True + + assert len(line.leaves) >= 2, "Stranded delimiter" + + first = line.leaves[0] + second = line.leaves[1] + penultimate = line.leaves[-2] + last = line.leaves[-1] + + # With a single delimiter, omit if the expression starts or ends with + # a bracket. + if first.type in OPENING_BRACKETS and second.type not in CLOSING_BRACKETS: + remainder = False + length = 4 * line.depth + for _index, leaf, leaf_length in enumerate_with_length(line): + if leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is first: + remainder = True + if remainder: + length += leaf_length + if length > line_length: + break + + if leaf.type in OPENING_BRACKETS: + # There are brackets we can further split on. + remainder = False + + else: + # checked the entire string and line length wasn't exceeded + if len(line.leaves) == _index + 1: + return True + + # Note: we are not returning False here because a line might have *both* + # a leading opening bracket and a trailing closing bracket. If the + # opening bracket doesn't match our rule, maybe the closing will. + + if ( + last.type == token.RPAR + or last.type == token.RBRACE + or ( + # don't use indexing for omitting optional parentheses; + # it looks weird + last.type == token.RSQB + and last.parent + and last.parent.type != syms.trailer + ) + ): + if penultimate.type in OPENING_BRACKETS: + # Empty brackets don't help. + return False + + if is_multiline_string(first): + # Additional wrapping of a multiline string in this situation is + # unnecessary. + return True + + length = 4 * line.depth + seen_other_brackets = False + for _index, leaf, leaf_length in enumerate_with_length(line): + length += leaf_length + if leaf is last.opening_bracket: + if seen_other_brackets or length <= line_length: + return True + + elif leaf.type in OPENING_BRACKETS: + # There are brackets we can further split on. + seen_other_brackets = True + + return False + + +def get_cache_file(mode: FileMode) -> Path: + return CACHE_DIR / f"cache.{mode.get_cache_key()}.pickle" + + +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(mode) + if not cache_file.exists(): + return {} + + with cache_file.open("rb") as fobj: + try: + cache: Cache = pickle.load(fobj) + except pickle.UnpicklingError: + return {} + + return cache + + +def get_cache_info(path: Path) -> CacheInfo: + """Return the information used to check if a file is already formatted or not.""" + stat = path.stat() + return stat.st_mtime, stat.st_size + + +def filter_cached(cache: Cache, sources: Iterable[Path]) -> Tuple[Set[Path], Set[Path]]: + """Split an iterable of paths in `sources` into two sets. + + The first contains paths of files that modified on disk or are not in the + cache. The other contains paths to non-modified files. + """ + todo, done = set(), set() + for src in sources: + src = src.resolve() + if cache.get(src) != get_cache_info(src): + todo.add(src) + else: + done.add(src) + return todo, done + + +def write_cache(cache: Cache, sources: Iterable[Path], mode: FileMode) -> None: + """Update the cache file.""" + cache_file = get_cache_file(mode) + try: + CACHE_DIR.mkdir(parents=True, exist_ok=True) + new_cache = {**cache, **{src.resolve(): get_cache_info(src) for src in sources}} + with tempfile.NamedTemporaryFile(dir=str(cache_file.parent), delete=False) as f: + pickle.dump(new_cache, f, protocol=pickle.HIGHEST_PROTOCOL) + os.replace(f.name, cache_file) + except OSError: + pass + + +def patch_click() -> None: + """Make Click not crash. + + On certain misconfigured environments, Python 3 selects the ASCII encoding as the + default which restricts paths that it can access during the lifetime of the + application. Click refuses to work in this scenario by raising a RuntimeError. + + In case of Black the likelihood that non-ASCII characters are going to be used in + file paths is minimal since it's Python source code. Moreover, this crash was + spurious on Python 3.7 thanks to PEP 538 and PEP 540. + """ + try: + from click import core + from click import _unicodefun # type: ignore + except ModuleNotFoundError: + return + + for module in (core, _unicodefun): + if hasattr(module, "_verify_python3_env"): + module._verify_python3_env = lambda: None + + +def patched_main() -> None: + freeze_support() + patch_click() main() + + +if __name__ == "__main__": + patched_main()