]> git.madduck.net Git - etc/vim.git/blobdiff - src/black/__init__.py

madduck's git repository

Every one of the projects in this repository is available at the canonical URL git://git.madduck.net/madduck/pub/<projectpath> — see each project's metadata for the exact URL.

All patches and comments are welcome. Please squash your changes to logical commits before using git-format-patch and git-send-email to patches@git.madduck.net. If you'd read over the Git project's submission guidelines and adhered to them, I'd be especially grateful.

SSH access, as well as push access can be individually arranged.

If you use my repositories frequently, consider adding the following snippet to ~/.gitconfig and using the third clone URL listed for each project:

[url "git://git.madduck.net/madduck/"]
  insteadOf = madduck:

Bump pypa/cibuildwheel from 2.11.3 to 2.11.4 (#3475)
[etc/vim.git] / src / black / __init__.py
index afc76e1fa0cb5e7dda1a51544718ef778e846344..f00749aaed81146dbadc7778cf49bad23af6e3c5 100644 (file)
@@ -1,6 +1,5 @@
 import io
 import json
-import os
 import platform
 import re
 import sys
@@ -28,14 +27,10 @@ from typing import (
     Union,
 )
 
-if sys.version_info >= (3, 8):
-    from typing import Final
-else:
-    from typing_extensions import Final
-
 import click
 from click.core import ParameterSource
 from mypy_extensions import mypyc_attr
+from pathspec import PathSpec
 from pathspec.patterns.gitwildmatch import GitWildMatchPatternError
 
 from _black_version import version as __version__
@@ -67,7 +62,7 @@ from black.handle_ipynb_magics import (
     unmask_cell,
 )
 from black.linegen import LN, LineGenerator, transform_line
-from black.lines import EmptyLineTracker, Line
+from black.lines import EmptyLineTracker, LinesBlock
 from black.mode import (
     FUTURE_FLAG_TO_FEATURE,
     VERSION_TO_FEATURES,
@@ -92,7 +87,6 @@ from blib2to3.pgen2 import token
 from blib2to3.pytree import Leaf, Node
 
 COMPILED = Path(__file__).suffix in (".pyd", ".so")
-DEFAULT_WORKERS: Final = os.cpu_count()
 
 # types
 FileContent = str
@@ -255,6 +249,12 @@ def validate_regex(
     ),
     default=[],
 )
+@click.option(
+    "-x",
+    "--skip-source-first-line",
+    is_flag=True,
+    help="Skip the first line of the source code.",
+)
 @click.option(
     "-S",
     "--skip-string-normalization",
@@ -371,9 +371,8 @@ def validate_regex(
     "-W",
     "--workers",
     type=click.IntRange(min=1),
-    default=DEFAULT_WORKERS,
-    show_default=True,
-    help="Number of parallel workers",
+    default=None,
+    help="Number of parallel workers [default: number of CPUs in the system]",
 )
 @click.option(
     "-q",
@@ -436,6 +435,7 @@ def main(  # noqa: C901
     pyi: bool,
     ipynb: bool,
     python_cell_magics: Sequence[str],
+    skip_source_first_line: bool,
     skip_string_normalization: bool,
     skip_magic_trailing_comma: bool,
     experimental_string_processing: bool,
@@ -448,7 +448,7 @@ def main(  # noqa: C901
     extend_exclude: Optional[Pattern[str]],
     force_exclude: Optional[Pattern[str]],
     stdin_filename: Optional[str],
-    workers: int,
+    workers: Optional[int],
     src: Tuple[str, ...],
     config: Optional[str],
 ) -> None:
@@ -498,8 +498,10 @@ def main(  # noqa: C901
             user_level_config = str(find_user_pyproject_toml())
             if config == user_level_config:
                 out(
-                    "Using configuration from user-level config at "
-                    f"'{user_level_config}'.",
+                    (
+                        "Using configuration from user-level config at "
+                        f"'{user_level_config}'."
+                    ),
                     fg="blue",
                 )
             elif config_source in (
@@ -509,6 +511,9 @@ def main(  # noqa: C901
                 out("Using configuration from project root.", fg="blue")
             else:
                 out(f"Using configuration in '{config}'.", fg="blue")
+            if ctx.default_map:
+                for param, value in ctx.default_map.items():
+                    out(f"{param}: {value}")
 
     error_msg = "Oh no! 💥 💔 💥"
     if (
@@ -536,6 +541,7 @@ def main(  # noqa: C901
         line_length=line_length,
         is_pyi=pyi,
         is_ipynb=ipynb,
+        skip_source_first_line=skip_source_first_line,
         string_normalization=not skip_string_normalization,
         magic_trailing_comma=not skip_magic_trailing_comma,
         experimental_string_processing=experimental_string_processing,
@@ -623,12 +629,12 @@ def get_sources(
 ) -> Set[Path]:
     """Compute the set of files to be formatted."""
     sources: Set[Path] = set()
+    root = ctx.obj["root"]
 
-    if exclude is None:
-        exclude = re_compile_maybe_verbose(DEFAULT_EXCLUDES)
-        gitignore = get_gitignore(ctx.obj["root"])
-    else:
-        gitignore = None
+    using_default_exclude = exclude is None
+    exclude = re_compile_maybe_verbose(DEFAULT_EXCLUDES) if exclude is None else exclude
+    gitignore: Optional[Dict[Path, PathSpec]] = None
+    root_gitignore = get_gitignore(root)
 
     for s in src:
         if s == "-" and stdin_filename:
@@ -663,6 +669,11 @@ def get_sources(
 
             sources.add(p)
         elif p.is_dir():
+            if using_default_exclude:
+                gitignore = {
+                    root: root_gitignore,
+                    root / p: get_gitignore(p),
+                }
             sources.update(
                 gen_python_files(
                     p.iterdir(),
@@ -793,7 +804,10 @@ def format_file_in_place(
         mode = replace(mode, is_ipynb=True)
 
     then = datetime.utcfromtimestamp(src.stat().st_mtime)
+    header = b""
     with open(src, "rb") as buf:
+        if mode.skip_source_first_line:
+            header = buf.readline()
         src_contents, encoding, newline = decode_bytes(buf.read())
     try:
         dst_contents = format_file_contents(src_contents, fast=fast, mode=mode)
@@ -803,6 +817,8 @@ def format_file_in_place(
         raise ValueError(
             f"File '{src}' cannot be parsed as valid Jupyter notebook."
         ) from None
+    src_contents = header.decode(encoding) + src_contents
+    dst_contents = header.decode(encoding) + dst_contents
 
     if write_back == WriteBack.YES:
         with open(src, "w", encoding=encoding, newline=newline) as f:
@@ -904,7 +920,7 @@ def format_file_contents(src_contents: str, *, fast: bool, mode: Mode) -> FileCo
     valid by calling :func:`assert_equivalent` and :func:`assert_stable` on it.
     `mode` is passed to :func:`format_str`.
     """
-    if not src_contents.strip():
+    if not mode.preview and not src_contents.strip():
         raise NothingChanged
 
     if mode.is_ipynb:
@@ -1001,6 +1017,9 @@ def format_ipynb_string(src_contents: str, *, fast: bool, mode: Mode) -> FileCon
     Operate cell-by-cell, only on code cells, only for Python notebooks.
     If the ``.ipynb`` originally had a trailing newline, it'll be preserved.
     """
+    if mode.preview and not src_contents:
+        raise NothingChanged
+
     trailing_newline = src_contents[-1] == "\n"
     modified = False
     nb = json.loads(src_contents)
@@ -1065,7 +1084,7 @@ def format_str(src_contents: str, *, mode: Mode) -> str:
 
 def _format_str_once(src_contents: str, *, mode: Mode) -> str:
     src_node = lib2to3_parse(src_contents.lstrip(), mode.target_versions)
-    dst_contents = []
+    dst_blocks: List[LinesBlock] = []
     if mode.target_versions:
         versions = mode.target_versions
     else:
@@ -1074,22 +1093,32 @@ def _format_str_once(src_contents: str, *, mode: Mode) -> str:
 
     normalize_fmt_off(src_node, preview=mode.preview)
     lines = LineGenerator(mode=mode)
-    elt = EmptyLineTracker(is_pyi=mode.is_pyi)
-    empty_line = Line(mode=mode)
-    after = 0
+    elt = EmptyLineTracker(mode=mode)
     split_line_features = {
         feature
         for feature in {Feature.TRAILING_COMMA_IN_CALL, Feature.TRAILING_COMMA_IN_DEF}
         if supports_feature(versions, feature)
     }
+    block: Optional[LinesBlock] = None
     for current_line in lines.visit(src_node):
-        dst_contents.append(str(empty_line) * after)
-        before, after = elt.maybe_empty_lines(current_line)
-        dst_contents.append(str(empty_line) * before)
+        block = elt.maybe_empty_lines(current_line)
+        dst_blocks.append(block)
         for line in transform_line(
             current_line, mode=mode, features=split_line_features
         ):
-            dst_contents.append(str(line))
+            block.content_lines.append(str(line))
+    if dst_blocks:
+        dst_blocks[-1].after = 0
+    dst_contents = []
+    for block in dst_blocks:
+        dst_contents.extend(block.all_lines())
+    if mode.preview and not dst_contents:
+        # Use decode_bytes to retrieve the correct source newline (CRLF or LF),
+        # and check if normalized_content has more than one line
+        normalized_content, _, newline = decode_bytes(src_contents.encode("utf-8"))
+        if "\n" in normalized_content:
+            return newline
+        return ""
     return "".join(dst_contents)
 
 
@@ -1372,13 +1401,15 @@ def patch_click() -> None:
 
     for module in modules:
         if hasattr(module, "_verify_python3_env"):
-            module._verify_python3_env = lambda: None  # type: ignore
+            module._verify_python3_env = lambda: None
         if hasattr(module, "_verify_python_env"):
-            module._verify_python_env = lambda: None  # type: ignore
+            module._verify_python_env = lambda: None
 
 
 def patched_main() -> None:
-    if sys.platform == "win32" and getattr(sys, "frozen", False):
+    # PyInstaller patches multiprocessing to need freeze_support() even in non-Windows
+    # environments so just assume we always need to call it if frozen.
+    if getattr(sys, "frozen", False):
         from multiprocessing import freeze_support
 
         freeze_support()