X-Git-Url: https://git.madduck.net/etc/vim.git/blobdiff_plain/a2b5ba2a3a97eb3a0b5130a5b317bb750c1624cf..3e731527e4418b0b6d9791d6e32caee9227ba69d:/src/black/__init__.py?ds=inline diff --git a/src/black/__init__.py b/src/black/__init__.py index a985926..cfa2c76 100644 --- a/src/black/__init__.py +++ b/src/black/__init__.py @@ -1,4 +1,6 @@ import asyncio +from json.decoder import JSONDecodeError +import json from concurrent.futures import Executor, ThreadPoolExecutor, ProcessPoolExecutor from contextlib import contextmanager from datetime import datetime @@ -7,7 +9,8 @@ import io from multiprocessing import Manager, freeze_support import os from pathlib import Path -import regex as re +from pathspec.patterns.gitwildmatch import GitWildMatchPatternError +import re import signal import sys import tokenize @@ -18,6 +21,7 @@ from typing import ( Generator, Iterator, List, + MutableMapping, Optional, Pattern, Set, @@ -26,26 +30,38 @@ from typing import ( Union, ) -from dataclasses import replace import click +from click.core import ParameterSource +from dataclasses import replace +from mypy_extensions import mypyc_attr from black.const import DEFAULT_LINE_LENGTH, DEFAULT_INCLUDES, DEFAULT_EXCLUDES from black.const import STDIN_PLACEHOLDER from black.nodes import STARS, syms, is_simple_decorator_expression +from black.nodes import is_string_token from black.lines import Line, EmptyLineTracker from black.linegen import transform_line, LineGenerator, LN from black.comments import normalize_fmt_off -from black.mode import Mode, TargetVersion +from black.mode import FUTURE_FLAG_TO_FEATURE, Mode, TargetVersion from black.mode import Feature, supports_feature, VERSION_TO_FEATURES from black.cache import read_cache, write_cache, get_cache_info, filter_cached, Cache from black.concurrency import cancel, shutdown, maybe_install_uvloop -from black.output import dump_to_file, diff, color_diff, out, err -from black.report import Report, Changed +from black.output import dump_to_file, ipynb_diff, diff, color_diff, out, err +from black.report import Report, Changed, NothingChanged from black.files import find_project_root, find_pyproject_toml, parse_pyproject_toml from black.files import gen_python_files, get_gitignore, normalize_path_maybe_ignore from black.files import wrap_stream_for_windows from black.parsing import InvalidInput # noqa F401 from black.parsing import lib2to3_parse, parse_ast, stringify_ast +from black.handle_ipynb_magics import ( + mask_cell, + unmask_cell, + remove_trailing_semicolon, + put_trailing_semicolon_back, + TRANSFORMED_MAGICS, + PYTHON_CELL_MAGICS, + jupyter_dependencies_are_installed, +) # lib2to3 fork @@ -54,16 +70,14 @@ from blib2to3.pgen2 import token from _black_version import version as __version__ +COMPILED = Path(__file__).suffix in (".pyd", ".so") + # types FileContent = str Encoding = str NewLine = str -class NothingChanged(UserWarning): - """Raised when reformatted code is the same as source.""" - - class WriteBack(Enum): NO = 0 YES = 1 @@ -87,6 +101,8 @@ class WriteBack(Enum): # Legacy name, left for integrations. FileMode = Mode +DEFAULT_WORKERS = os.cpu_count() + def read_pyproject_toml( ctx: click.Context, param: click.Parameter, value: Optional[str] @@ -106,7 +122,7 @@ def read_pyproject_toml( except (OSError, ValueError) as e: raise click.FileError( filename=value, hint=f"Error reading configuration file: {e}" - ) + ) from None if not config: return None @@ -160,14 +176,19 @@ def validate_regex( ctx: click.Context, param: click.Parameter, value: Optional[str], -) -> Optional[Pattern]: +) -> Optional[Pattern[str]]: try: return re_compile_maybe_verbose(value) if value is not None else None - except re.error: - raise click.BadParameter("Not a valid regular expression") + except re.error as e: + raise click.BadParameter(f"Not a valid regular expression: {e}") from None -@click.command(context_settings=dict(help_option_names=["-h", "--help"])) +@click.command( + context_settings={"help_option_names": ["-h", "--help"]}, + # While Click does set this field automatically using the docstring, mypyc + # (annoyingly) strips 'em so we need to set it here too. + help="The uncompromising code formatter.", +) @click.option("-c", "--code", type=str, help="Format the code passed in as a string.") @click.option( "-l", @@ -196,6 +217,14 @@ def validate_regex( " when piping source on standard input)." ), ) +@click.option( + "--ipynb", + is_flag=True, + help=( + "Format all input files like Jupyter Notebooks regardless of file extension " + "(useful when piping source on standard input)." + ), +) @click.option( "-S", "--skip-string-normalization", @@ -302,6 +331,14 @@ def validate_regex( "editors that rely on using stdin." ), ) +@click.option( + "-W", + "--workers", + type=click.IntRange(min=1), + default=DEFAULT_WORKERS, + show_default=True, + help="Number of parallel workers", +) @click.option( "-q", "--quiet", @@ -320,7 +357,10 @@ def validate_regex( " due to exclusion patterns." ), ) -@click.version_option(version=__version__) +@click.version_option( + version=__version__, + message=f"%(prog)s, %(version)s (compiled: {'yes' if COMPILED else 'no'})", +) @click.argument( "src", nargs=-1, @@ -328,6 +368,7 @@ def validate_regex( exists=True, file_okay=True, dir_okay=True, readable=True, allow_dash=True ), is_eager=True, + metavar="SRC ...", ) @click.option( "--config", @@ -354,23 +395,54 @@ def main( color: bool, fast: bool, pyi: bool, + ipynb: bool, skip_string_normalization: bool, skip_magic_trailing_comma: bool, experimental_string_processing: bool, quiet: bool, verbose: bool, - required_version: str, - include: Pattern, - exclude: Optional[Pattern], - extend_exclude: Optional[Pattern], - force_exclude: Optional[Pattern], + required_version: Optional[str], + include: Pattern[str], + exclude: Optional[Pattern[str]], + extend_exclude: Optional[Pattern[str]], + force_exclude: Optional[Pattern[str]], stdin_filename: Optional[str], + workers: int, src: Tuple[str, ...], config: Optional[str], ) -> None: """The uncompromising code formatter.""" - if config and verbose: - out(f"Using configuration from {config}.", bold=False, fg="blue") + ctx.ensure_object(dict) + root, method = find_project_root(src) if code is None else (None, None) + ctx.obj["root"] = root + + if verbose: + if root: + out( + f"Identified `{root}` as project root containing a {method}.", + fg="blue", + ) + + normalized = [ + (normalize_path_maybe_ignore(Path(source), root), source) + for source in src + ] + srcs_string = ", ".join( + [ + f'"{_norm}"' + if _norm + else f'\033[31m"{source} (skipping - invalid)"\033[34m' + for _norm, source in normalized + ] + ) + out(f"Sources to be formatted: {srcs_string}", fg="blue") + + if config: + config_source = ctx.get_parameter_source("config") + if config_source in (ParameterSource.DEFAULT, ParameterSource.DEFAULT_MAP): + out("Using configuration from project root.", fg="blue") + else: + out(f"Using configuration in '{config}'.", fg="blue") error_msg = "Oh no! 💥 💔 💥" if required_version and required_version != __version__: @@ -379,6 +451,9 @@ def main( f" the running version `{__version__}`!" ) ctx.exit(1) + if ipynb and pyi: + err("Cannot pass both `pyi` and `ipynb` flags!") + ctx.exit(1) write_back = WriteBack.from_configuration(check=check, diff=diff, color=color) if target_version: @@ -390,6 +465,7 @@ def main( target_versions=versions, line_length=line_length, is_pyi=pyi, + is_ipynb=ipynb, string_normalization=not skip_string_normalization, magic_trailing_comma=not skip_magic_trailing_comma, experimental_string_processing=experimental_string_processing, @@ -407,18 +483,21 @@ def main( content=code, fast=fast, write_back=write_back, mode=mode, report=report ) else: - sources = get_sources( - ctx=ctx, - src=src, - quiet=quiet, - verbose=verbose, - include=include, - exclude=exclude, - extend_exclude=extend_exclude, - force_exclude=force_exclude, - report=report, - stdin_filename=stdin_filename, - ) + try: + sources = get_sources( + ctx=ctx, + src=src, + quiet=quiet, + verbose=verbose, + include=include, + exclude=exclude, + extend_exclude=extend_exclude, + force_exclude=force_exclude, + report=report, + stdin_filename=stdin_filename, + ) + except GitWildMatchPatternError: + ctx.exit(1) path_empty( sources, @@ -443,6 +522,7 @@ def main( write_back=write_back, mode=mode, report=report, + workers=workers, ) if verbose or not quiet: @@ -466,14 +546,12 @@ def get_sources( stdin_filename: Optional[str], ) -> Set[Path]: """Compute the set of files to be formatted.""" - - root = find_project_root(src) sources: Set[Path] = set() path_empty(src, "No Path provided. Nothing to do 😴", quiet, verbose, ctx) if exclude is None: exclude = re_compile_maybe_verbose(DEFAULT_EXCLUDES) - gitignore = get_gitignore(root) + gitignore = get_gitignore(ctx.obj["root"]) else: gitignore = None @@ -486,7 +564,7 @@ def get_sources( is_stdin = False if is_stdin or p.is_file(): - normalized_path = normalize_path_maybe_ignore(p, root, report) + normalized_path = normalize_path_maybe_ignore(p, ctx.obj["root"], report) if normalized_path is None: continue @@ -503,18 +581,25 @@ def get_sources( if is_stdin: p = Path(f"{STDIN_PLACEHOLDER}{str(p)}") + if p.suffix == ".ipynb" and not jupyter_dependencies_are_installed( + verbose=verbose, quiet=quiet + ): + continue + sources.add(p) elif p.is_dir(): sources.update( gen_python_files( p.iterdir(), - root, + ctx.obj["root"], include, exclude, extend_exclude, force_exclude, report, gitignore, + verbose=verbose, + quiet=quiet, ) ) elif s == "-": @@ -584,6 +669,8 @@ def reformat_one( if is_stdin: if src.suffix == ".pyi": mode = replace(mode, is_pyi=True) + elif src.suffix == ".ipynb": + mode = replace(mode, is_ipynb=True) if format_stdin_to_stdout(fast=fast, write_back=write_back, mode=mode): changed = Changed.YES else: @@ -609,19 +696,28 @@ def reformat_one( report.failed(src, str(exc)) +# diff-shades depends on being to monkeypatch this function to operate. I know it's +# not ideal, but this shouldn't cause any issues ... hopefully. ~ichard26 +@mypyc_attr(patchable=True) def reformat_many( - sources: Set[Path], fast: bool, write_back: WriteBack, mode: Mode, report: "Report" + sources: Set[Path], + fast: bool, + write_back: WriteBack, + mode: Mode, + report: "Report", + workers: Optional[int], ) -> None: """Reformat multiple files using a ProcessPoolExecutor.""" executor: Executor loop = asyncio.get_event_loop() - worker_count = os.cpu_count() + worker_count = workers if workers is not None else DEFAULT_WORKERS if sys.platform == "win32": # Work around https://bugs.python.org/issue26903 + assert worker_count is not None worker_count = min(worker_count, 60) try: executor = ProcessPoolExecutor(max_workers=worker_count) - except (ImportError, OSError): + except (ImportError, NotImplementedError, OSError): # we arrive here if the underlying system does not support multi-processing # like in AWS Lambda or Termux, in which case we gracefully fallback to # a ThreadPoolExecutor with just a single worker (more workers would not do us @@ -712,7 +808,10 @@ async def schedule_formatting( sources_to_cache.append(src) report.done(src, changed) if cancelled: - await asyncio.gather(*cancelled, loop=loop, return_exceptions=True) + if sys.version_info >= (3, 7): + await asyncio.gather(*cancelled, return_exceptions=True) + else: + await asyncio.gather(*cancelled, loop=loop, return_exceptions=True) if sources_to_cache: write_cache(cache, sources_to_cache, mode) @@ -732,6 +831,8 @@ def format_file_in_place( """ if src.suffix == ".pyi": mode = replace(mode, is_pyi=True) + elif src.suffix == ".ipynb": + mode = replace(mode, is_ipynb=True) then = datetime.utcfromtimestamp(src.stat().st_mtime) with open(src, "rb") as buf: @@ -740,6 +841,10 @@ def format_file_in_place( dst_contents = format_file_contents(src_contents, fast=fast, mode=mode) except NothingChanged: return False + except JSONDecodeError: + raise ValueError( + f"File '{src}' cannot be parsed as valid Jupyter notebook." + ) from None if write_back == WriteBack.YES: with open(src, "w", encoding=encoding, newline=newline) as f: @@ -748,7 +853,10 @@ def format_file_in_place( 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 mode.is_ipynb: + diff_contents = ipynb_diff(src_contents, dst_contents, src_name, dst_name) + else: + diff_contents = diff(src_contents, dst_contents, src_name, dst_name) if write_back == WriteBack.COLOR_DIFF: diff_contents = color_diff(diff_contents) @@ -803,7 +911,8 @@ def format_stdin_to_stdout( ) if write_back == WriteBack.YES: # Make sure there's a newline after the content - dst += "" if dst[-1] == "\n" else "\n" + if dst and dst[-1] != "\n": + dst += "\n" f.write(dst) elif write_back in (WriteBack.DIFF, WriteBack.COLOR_DIFF): now = datetime.utcnow() @@ -817,6 +926,29 @@ def format_stdin_to_stdout( f.detach() +def check_stability_and_equivalence( + src_contents: str, dst_contents: str, *, mode: Mode +) -> None: + """Perform stability and equivalence checks. + + Raise AssertionError if source and destination contents are not + equivalent, or if a second pass of the formatter would format the + content differently. + """ + assert_equivalent(src_contents, dst_contents) + + # Forced second pass to work around optional trailing commas (becoming + # forced trailing commas on pass 2) interacting differently with optional + # parentheses. Admittedly ugly. + dst_contents_pass2 = format_str(dst_contents, mode=mode) + if dst_contents != dst_contents_pass2: + dst_contents = dst_contents_pass2 + assert_equivalent(src_contents, dst_contents, pass_num=2) + assert_stable(src_contents, dst_contents, mode=mode) + # Note: no need to explicitly call `assert_stable` if `dst_contents` was + # the same as `dst_contents_pass2`. + + def format_file_contents(src_contents: str, *, fast: bool, mode: Mode) -> FileContent: """Reformat contents of a file and return new contents. @@ -827,26 +959,120 @@ def format_file_contents(src_contents: str, *, fast: bool, mode: Mode) -> FileCo if not src_contents.strip(): raise NothingChanged - dst_contents = format_str(src_contents, mode=mode) + if mode.is_ipynb: + dst_contents = format_ipynb_string(src_contents, fast=fast, mode=mode) + else: + dst_contents = format_str(src_contents, mode=mode) if src_contents == dst_contents: raise NothingChanged - if not fast: - assert_equivalent(src_contents, dst_contents) - - # Forced second pass to work around optional trailing commas (becoming - # forced trailing commas on pass 2) interacting differently with optional - # parentheses. Admittedly ugly. - dst_contents_pass2 = format_str(dst_contents, mode=mode) - if dst_contents != dst_contents_pass2: - dst_contents = dst_contents_pass2 - assert_equivalent(src_contents, dst_contents, pass_num=2) - assert_stable(src_contents, dst_contents, mode=mode) - # Note: no need to explicitly call `assert_stable` if `dst_contents` was - # the same as `dst_contents_pass2`. + if not fast and not mode.is_ipynb: + # Jupyter notebooks will already have been checked above. + check_stability_and_equivalence(src_contents, dst_contents, mode=mode) return dst_contents +def validate_cell(src: str) -> None: + """Check that cell does not already contain TransformerManager transformations, + or non-Python cell magics, which might cause tokenizer_rt to break because of + indentations. + + If a cell contains ``!ls``, then it'll be transformed to + ``get_ipython().system('ls')``. However, if the cell originally contained + ``get_ipython().system('ls')``, then it would get transformed in the same way: + + >>> TransformerManager().transform_cell("get_ipython().system('ls')") + "get_ipython().system('ls')\n" + >>> TransformerManager().transform_cell("!ls") + "get_ipython().system('ls')\n" + + Due to the impossibility of safely roundtripping in such situations, cells + containing transformed magics will be ignored. + """ + if any(transformed_magic in src for transformed_magic in TRANSFORMED_MAGICS): + raise NothingChanged + if src[:2] == "%%" and src.split()[0][2:] not in PYTHON_CELL_MAGICS: + raise NothingChanged + + +def format_cell(src: str, *, fast: bool, mode: Mode) -> str: + """Format code in given cell of Jupyter notebook. + + General idea is: + + - if cell has trailing semicolon, remove it; + - if cell has IPython magics, mask them; + - format cell; + - reinstate IPython magics; + - reinstate trailing semicolon (if originally present); + - strip trailing newlines. + + Cells with syntax errors will not be processed, as they + could potentially be automagics or multi-line magics, which + are currently not supported. + """ + validate_cell(src) + src_without_trailing_semicolon, has_trailing_semicolon = remove_trailing_semicolon( + src + ) + try: + masked_src, replacements = mask_cell(src_without_trailing_semicolon) + except SyntaxError: + raise NothingChanged from None + masked_dst = format_str(masked_src, mode=mode) + if not fast: + check_stability_and_equivalence(masked_src, masked_dst, mode=mode) + dst_without_trailing_semicolon = unmask_cell(masked_dst, replacements) + dst = put_trailing_semicolon_back( + dst_without_trailing_semicolon, has_trailing_semicolon + ) + dst = dst.rstrip("\n") + if dst == src: + raise NothingChanged from None + return dst + + +def validate_metadata(nb: MutableMapping[str, Any]) -> None: + """If notebook is marked as non-Python, don't format it. + + All notebook metadata fields are optional, see + https://nbformat.readthedocs.io/en/latest/format_description.html. So + if a notebook has empty metadata, we will try to parse it anyway. + """ + language = nb.get("metadata", {}).get("language_info", {}).get("name", None) + if language is not None and language != "python": + raise NothingChanged from None + + +def format_ipynb_string(src_contents: str, *, fast: bool, mode: Mode) -> FileContent: + """Format Jupyter notebook. + + Operate cell-by-cell, only on code cells, only for Python notebooks. + If the ``.ipynb`` originally had a trailing newline, it'll be preserved. + """ + trailing_newline = src_contents[-1] == "\n" + modified = False + nb = json.loads(src_contents) + validate_metadata(nb) + for cell in nb["cells"]: + if cell.get("cell_type", None) == "code": + try: + src = "".join(cell["source"]) + dst = format_cell(src, fast=fast, mode=mode) + except NothingChanged: + pass + else: + cell["source"] = dst.splitlines(keepends=True) + modified = True + if modified: + dst_contents = json.dumps(nb, indent=1, ensure_ascii=False) + if trailing_newline: + dst_contents = dst_contents + "\n" + return dst_contents + else: + raise NothingChanged + + def format_str(src_contents: str, *, mode: Mode) -> FileContent: """Reformat a string and return new contents. @@ -883,13 +1109,10 @@ def format_str(src_contents: str, *, mode: Mode) -> FileContent: if mode.target_versions: versions = mode.target_versions else: - versions = detect_target_versions(src_node) + versions = detect_target_versions(src_node, future_imports=future_imports) + normalize_fmt_off(src_node) - lines = LineGenerator( - mode=mode, - remove_u_prefix="unicode_literals" in future_imports - or supports_feature(versions, Feature.UNICODE_LITERALS), - ) + lines = LineGenerator(mode=mode) elt = EmptyLineTracker(is_pyi=mode.is_pyi) empty_line = Line(mode=mode) after = 0 @@ -926,7 +1149,9 @@ def decode_bytes(src: bytes) -> Tuple[FileContent, Encoding, NewLine]: return tiow.read(), encoding, newline -def get_features_used(node: Node) -> Set[Feature]: +def get_features_used( # noqa: C901 + node: Node, *, future_imports: Optional[Set[str]] = None +) -> Set[Feature]: """Return a set of (relatively) new Python features used in this file. Currently looking for: @@ -936,20 +1161,34 @@ def get_features_used(node: Node) -> Set[Feature]: - positional only arguments in function signatures and lambdas; - assignment expression; - relaxed decorator syntax; + - usage of __future__ flags (annotations); + - print / exec statements; """ features: Set[Feature] = set() + if future_imports: + features |= { + FUTURE_FLAG_TO_FEATURE[future_import] + for future_import in future_imports + if future_import in FUTURE_FLAG_TO_FEATURE + } + for n in node.pre_order(): - if n.type == token.STRING: - value_head = n.value[:2] # type: ignore + if is_string_token(n): + value_head = n.value[:2] 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 + assert isinstance(n, Leaf) + if "_" in n.value: features.add(Feature.NUMERIC_UNDERSCORES) elif n.type == token.SLASH: - if n.parent and n.parent.type in {syms.typedargslist, syms.arglist}: + if n.parent and n.parent.type in { + syms.typedargslist, + syms.arglist, + syms.varargslist, + }: features.add(Feature.POS_ONLY_ARGUMENTS) elif n.type == token.COLONEQUAL: @@ -980,12 +1219,29 @@ def get_features_used(node: Node) -> Set[Feature]: if argch.type in STARS: features.add(feature) + elif ( + n.type in {syms.return_stmt, syms.yield_expr} + and len(n.children) >= 2 + and n.children[1].type == syms.testlist_star_expr + and any(child.type == syms.star_expr for child in n.children[1].children) + ): + features.add(Feature.UNPACKING_ON_FLOW) + + elif ( + n.type == syms.annassign + and len(n.children) >= 4 + and n.children[3].type == syms.testlist_star_expr + ): + features.add(Feature.ANN_ASSIGN_EXTENDED_RHS) + return features -def detect_target_versions(node: Node) -> Set[TargetVersion]: +def detect_target_versions( + node: Node, *, future_imports: Optional[Set[str]] = None +) -> Set[TargetVersion]: """Detect the version to target based on the nodes used.""" - features = get_features_used(node) + features = get_features_used(node, future_imports=future_imports) return { version for version in TargetVersion if features <= VERSION_TO_FEATURES[version] } @@ -1047,9 +1303,8 @@ def assert_equivalent(src: str, dst: str, *, pass_num: int = 1) -> None: src_ast = parse_ast(src) except Exception as exc: raise AssertionError( - "cannot use --safe with this file; failed to parse source file. AST" - f" error message: {exc}" - ) + f"cannot use --safe with this file; failed to parse source file: {exc}" + ) from exc try: dst_ast = parse_ast(dst) @@ -1110,7 +1365,7 @@ def patch_click() -> None: """ try: from click import core - from click import _unicodefun # type: ignore + from click import _unicodefun except ModuleNotFoundError: return