X-Git-Url: https://git.madduck.net/etc/vim.git/blobdiff_plain/f2ea461e9e9fa5c47bb61fd72d512c748928badc..40053b522ebe8f33328c1c7e29780fc37d17911e:/src/black/linegen.py diff --git a/src/black/linegen.py b/src/black/linegen.py index 2e16b6f..caffbab 100644 --- a/src/black/linegen.py +++ b/src/black/linegen.py @@ -3,14 +3,18 @@ Generating lines of code. """ from functools import partial, wraps import sys -from typing import Collection, Iterator, List, Optional, Set, Union +from typing import Collection, Iterator, List, Optional, Set, Union, cast -from dataclasses import dataclass, field - -from black.nodes import WHITESPACE, STATEMENT, STANDALONE_COMMENT +from black.nodes import WHITESPACE, RARROW, STATEMENT, STANDALONE_COMMENT from black.nodes import ASSIGNMENTS, OPENING_BRACKETS, CLOSING_BRACKETS -from black.nodes import Visitor, syms, first_child_is_arith, ensure_visible -from black.nodes import is_docstring, is_empty_tuple, is_one_tuple, is_one_tuple_between +from black.nodes import Visitor, syms, is_arith_like, ensure_visible +from black.nodes import ( + is_docstring, + is_empty_tuple, + is_one_tuple, + is_one_sequence_between, +) +from black.nodes import is_name_token, is_lpar_token, is_rpar_token from black.nodes import is_walrus_assignment, is_yield, is_vararg, is_multiline_string from black.nodes import is_stub_suite, is_stub_body, is_atom_with_invisible_parens from black.nodes import wrap_in_parentheses @@ -22,10 +26,9 @@ from black.comments import generate_comments, list_comments, FMT_OFF from black.numerics import normalize_numeric_literal from black.strings import get_string_prefix, fix_docstring from black.strings import normalize_string_prefix, normalize_string_quotes -from black.trans import Transformer, CannotTransform, StringMerger -from black.trans import StringSplitter, StringParenWrapper, StringParenStripper -from black.mode import Mode -from black.mode import Feature +from black.trans import Transformer, CannotTransform, StringMerger, StringSplitter +from black.trans import StringParenWrapper, StringParenStripper, hug_power_op +from black.mode import Mode, Feature, Preview from blib2to3.pytree import Node, Leaf from blib2to3.pgen2 import token @@ -40,7 +43,8 @@ class CannotSplit(CannotTransform): """A readable split that fits the allotted line length is impossible.""" -@dataclass +# This isn't a dataclass because @dataclass + Generic breaks mypyc. +# See also https://github.com/mypyc/mypyc/issues/827. class LineGenerator(Visitor[Line]): """Generates reformatted Line objects. Empty lines are not emitted. @@ -48,9 +52,10 @@ class LineGenerator(Visitor[Line]): in ways that will no longer stringify to valid Python code on the tree. """ - mode: Mode - remove_u_prefix: bool = False - current_line: Line = field(init=False) + def __init__(self, mode: Mode) -> None: + self.mode = mode + self.current_line: Line + self.__post_init__() def line(self, indent: int = 0) -> Iterator[Line]: """Generate a line. @@ -72,7 +77,7 @@ class LineGenerator(Visitor[Line]): """Default `visit_*()` implementation. Recurses to children of `node`.""" if isinstance(node, Leaf): any_open_brackets = self.current_line.bracket_tracker.any_open_brackets() - for comment in generate_comments(node): + for comment in generate_comments(node, preview=self.mode.preview): if any_open_brackets: # any comment within brackets is subject to splitting self.current_line.append(comment) @@ -90,9 +95,7 @@ class LineGenerator(Visitor[Line]): normalize_prefix(node, inside_brackets=any_open_brackets) if self.mode.string_normalization and node.type == token.STRING: - node.value = normalize_string_prefix( - node.value, remove_u_prefix=self.remove_u_prefix - ) + node.value = normalize_string_prefix(node.value) node.value = normalize_string_quotes(node.value) if node.type == token.NUMBER: normalize_numeric_literal(node) @@ -126,7 +129,7 @@ class LineGenerator(Visitor[Line]): """Visit a statement. This implementation is shared for `if`, `while`, `for`, `try`, `except`, - `def`, `with`, `class`, `assert` and assignments. + `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. @@ -134,13 +137,48 @@ class LineGenerator(Visitor[Line]): `parens` holds a set of string leaf values immediately after which invisible parens should be put. """ - normalize_invisible_parens(node, parens_after=parens) + normalize_invisible_parens(node, parens_after=parens, preview=self.mode.preview) for child in node.children: - if child.type == token.NAME and child.value in keywords: # type: ignore + if is_name_token(child) and child.value in keywords: yield from self.line() yield from self.visit(child) + def visit_funcdef(self, node: Node) -> Iterator[Line]: + """Visit function definition.""" + if Preview.annotation_parens not in self.mode: + yield from self.visit_stmt(node, keywords={"def"}, parens=set()) + else: + yield from self.line() + + # Remove redundant brackets around return type annotation. + is_return_annotation = False + for child in node.children: + if child.type == token.RARROW: + is_return_annotation = True + elif is_return_annotation: + if child.type == syms.atom and child.children[0].type == token.LPAR: + if maybe_make_parens_invisible_in_atom( + child, + parent=node, + remove_brackets_around_comma=False, + ): + wrap_in_parentheses(node, child, visible=False) + else: + wrap_in_parentheses(node, child, visible=False) + is_return_annotation = False + + for child in node.children: + yield from self.visit(child) + + def visit_match_case(self, node: Node) -> Iterator[Line]: + """Visit either a match or case statement.""" + normalize_invisible_parens(node, parens_after=set(), preview=self.mode.preview) + + yield from self.line() + for child in node.children: + yield from self.visit(child) + def visit_suite(self, node: Node) -> Iterator[Line]: """Visit a suite.""" if self.mode.is_pyi and is_stub_suite(node): @@ -150,8 +188,12 @@ class LineGenerator(Visitor[Line]): def visit_simple_stmt(self, node: Node) -> Iterator[Line]: """Visit a statement without nested statements.""" - if first_child_is_arith(node): - wrap_in_parentheses(node, node.children[0], visible=False) + prev_type: Optional[int] = None + for child in node.children: + if (prev_type is None or prev_type == token.SEMI) and is_arith_like(child): + wrap_in_parentheses(node, child, visible=False) + prev_type = child.type + is_suite_like = node.parent and node.parent.type in STATEMENT if is_suite_like: if self.mode.is_pyi and is_stub_body(node): @@ -191,6 +233,31 @@ class LineGenerator(Visitor[Line]): yield from self.line() yield from self.visit(child) + def visit_power(self, node: Node) -> Iterator[Line]: + for idx, leaf in enumerate(node.children[:-1]): + next_leaf = node.children[idx + 1] + + if not isinstance(leaf, Leaf): + continue + + value = leaf.value.lower() + if ( + leaf.type == token.NUMBER + and next_leaf.type == syms.trailer + # Ensure that we are in an attribute trailer + and next_leaf.children[0].type == token.DOT + # It shouldn't wrap hexadecimal, binary and octal literals + and not value.startswith(("0x", "0b", "0o")) + # It shouldn't wrap complex literals + and "j" not in value + ): + wrap_in_parentheses(node, leaf) + + if Preview.remove_redundant_parens in self.mode: + remove_await_parens(node) + + yield from self.visit_default(node) + def visit_SEMI(self, leaf: Leaf) -> Iterator[Line]: """Remove a semicolon and put the other statement on a separate line.""" yield from self.line() @@ -226,8 +293,9 @@ class LineGenerator(Visitor[Line]): if is_docstring(leaf) and "\\\n" not in leaf.value: # We're ignoring docstrings with backslash newline escapes because changing # indentation of those changes the AST representation of the code. - prefix = get_string_prefix(leaf.value) - docstring = leaf.value[len(prefix) :] # Remove the prefix + docstring = normalize_string_prefix(leaf.value) + prefix = get_string_prefix(docstring) + docstring = docstring[len(prefix) :] # Remove the prefix quote_char = docstring[0] # A natural way to remove the outer quotes is to do: # docstring = docstring.strip(quote_char) @@ -236,6 +304,7 @@ class LineGenerator(Visitor[Line]): # characters but only if they are the same as the first. quote_len = 1 if docstring[1] != quote_char else 3 docstring = docstring[quote_len:-quote_len] + docstring_started_empty = not docstring if is_multiline_string(leaf): indent = " " * 4 * self.current_line.depth @@ -255,8 +324,7 @@ class LineGenerator(Visitor[Line]): # Odd number of tailing backslashes, add some padding to # avoid escaping the closing string quote. docstring += " " - else: - # Add some padding if the docstring is empty. + elif not docstring_started_empty: docstring = " " # We could enforce triple quotes at this point. @@ -280,9 +348,14 @@ class LineGenerator(Visitor[Line]): 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) + if self.mode.preview: + self.visit_except_clause = partial( + v, keywords={"except"}, parens={"except"} + ) + self.visit_with_stmt = partial(v, keywords={"with"}, parens={"with"}) + else: + self.visit_except_clause = partial(v, keywords={"except"}, parens=Ø) + self.visit_with_stmt = partial(v, keywords={"with"}, 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"}) @@ -291,6 +364,10 @@ class LineGenerator(Visitor[Line]): self.visit_async_funcdef = self.visit_async_stmt self.visit_decorated = self.visit_decorators + # PEP 634 + self.visit_match_stmt = self.visit_match_case + self.visit_case_block = self.visit_match_case + def transform_line( line: Line, mode: Mode, features: Collection[Feature] = () @@ -326,7 +403,7 @@ def transform_line( and not (line.inside_brackets and line.contains_standalone_comments()) ): # Only apply basic string preprocessing, since lines shouldn't be split here. - if mode.experimental_string_processing: + if Preview.string_processing in mode: transformers = [string_merge, string_paren_strip] else: transformers = [] @@ -334,7 +411,9 @@ def transform_line( transformers = [left_hand_split] else: - def rhs(line: Line, features: Collection[Feature]) -> Iterator[Line]: + def _rhs( + self: object, line: Line, features: Collection[Feature] + ) -> Iterator[Line]: """Wraps calls to `right_hand_split`. The calls increasingly `omit` right-hand trailers (bracket pairs with @@ -361,7 +440,13 @@ def transform_line( line, line_length=mode.line_length, features=features ) - if mode.experimental_string_processing: + # HACK: nested functions (like _rhs) compiled by mypyc don't retain their + # __name__ attribute which is needed in `run_transformer` further down. + # Unfortunately a nested class breaks mypyc too. So a class must be created + # via type ... https://github.com/mypyc/mypyc/issues/884 + rhs = type("rhs", (), {"__call__": _rhs})() + + if Preview.string_processing in mode: if line.inside_brackets: transformers = [ string_merge, @@ -385,6 +470,9 @@ def transform_line( transformers = [delimiter_split, standalone_comment_split, rhs] else: transformers = [rhs] + # It's always safe to attempt hugging of power operations and pretty much every line + # could match. + transformers.append(hug_power_op) for transform in transformers: # We are accumulating lines in `result` because we might want to abort @@ -419,7 +507,10 @@ def left_hand_split(line: Line, _features: Collection[Feature] = ()) -> Iterator current_leaves is body_leaves and leaf.type in CLOSING_BRACKETS and leaf.opening_bracket is matching_bracket + and isinstance(matching_bracket, Leaf) ): + ensure_visible(leaf) + ensure_visible(matching_bracket) current_leaves = tail_leaves if body_leaves else head_leaves current_leaves.append(leaf) if current_leaves is head_leaves: @@ -495,21 +586,21 @@ def right_hand_split( # 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_on_explode=omit) + and can_omit_invisible_parens(body, line_length) ): omit = {id(closing_bracket), *omit} try: yield from right_hand_split(line, line_length, features=features, omit=omit) return - except CannotSplit: + except CannotSplit as e: 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." - ) + ) from e elif head.contains_multiline_strings() or tail.contains_multiline_strings(): raise CannotSplit( @@ -517,7 +608,7 @@ def right_hand_split( " satisfy the splitting algorithm because the head or the tail" " contains multiline strings which by definition never fit one" " line." - ) + ) from e ensure_visible(opening_bracket) ensure_visible(closing_bracket) @@ -573,6 +664,20 @@ def bracket_split_build_line( original.is_def and opening_bracket.value == "(" and not any(leaf.type == token.COMMA for leaf in leaves) + # In particular, don't add one within a parenthesized return annotation. + # Unfortunately the indicator we're in a return annotation (RARROW) may + # be defined directly in the parent node, the parent of the parent ... + # and so on depending on how complex the return annotation is. + # This isn't perfect and there's some false negatives but they are in + # contexts were a comma is actually fine. + and not any( + node.prev_sibling.type == RARROW + for node in ( + leaves[0].parent, + getattr(leaves[0].parent, "parent", None), + ) + if isinstance(node, Node) and isinstance(node.prev_sibling, Leaf) + ) ) if original.is_import or no_commas: @@ -603,9 +708,9 @@ def dont_increase_indentation(split_func: Transformer) -> Transformer: @wraps(split_func) def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]: - for line in split_func(line, features): - normalize_prefix(line.leaves[0], inside_brackets=True) - yield line + for split_line in split_func(line, features): + normalize_prefix(split_line.leaves[0], inside_brackets=True) + yield split_line return split_wrapper @@ -620,13 +725,13 @@ def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[ try: last_leaf = line.leaves[-1] except IndexError: - raise CannotSplit("Line empty") + raise CannotSplit("Line empty") from None bt = line.bracket_tracker try: delimiter_priority = bt.max_delimiter_priority(exclude={id(last_leaf)}) except ValueError: - raise CannotSplit("No delimiters found") + raise CannotSplit("No delimiters found") from None if delimiter_priority == DOT_PRIORITY: if bt.delimiter_count_with_priority(delimiter_priority) == 1: @@ -740,7 +845,9 @@ def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None: leaf.prefix = "" -def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: +def normalize_invisible_parens( + node: Node, parens_after: Set[str], *, preview: bool +) -> None: """Make existing optional parentheses invisible or create new ones. `parens_after` is a set of string leaf values immediately after which parens @@ -749,7 +856,7 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: 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): + for pc in list_comments(node.prefix, is_endmarker=False, preview=preview): if pc.value in FMT_OFF: # This `node` has a prefix with `# fmt: off`, don't mess with parens. return @@ -758,7 +865,9 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: # Fixes a bug where invisible parens are not properly stripped from # assignment statements that contain type annotations. if isinstance(child, Node) and child.type == syms.annassign: - normalize_invisible_parens(child, parens_after=parens_after) + normalize_invisible_parens( + child, parens_after=parens_after, preview=preview + ) # Add parentheses around long tuple unpacking in assignments. if ( @@ -769,18 +878,38 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: check_lpar = True if check_lpar: - if child.type == syms.atom: - if maybe_make_parens_invisible_in_atom(child, parent=node): + if ( + preview + and child.type == syms.atom + and node.type == syms.for_stmt + and isinstance(child.prev_sibling, Leaf) + and child.prev_sibling.type == token.NAME + and child.prev_sibling.value == "for" + ): + if maybe_make_parens_invisible_in_atom( + child, + parent=node, + remove_brackets_around_comma=True, + ): + wrap_in_parentheses(node, child, visible=False) + elif preview and isinstance(child, Node) and node.type == syms.with_stmt: + remove_with_parens(child, node) + elif child.type == syms.atom: + if maybe_make_parens_invisible_in_atom( + child, + parent=node, + ): wrap_in_parentheses(node, child, visible=False) elif is_one_tuple(child): wrap_in_parentheses(node, child, visible=True) elif node.type == syms.import_from: # "import from" nodes store parentheses directly as part of # the statement - if child.type == token.LPAR: + if is_lpar_token(child): + assert is_rpar_token(node.children[-1]) # make parentheses invisible - child.value = "" # type: ignore - node.children[-1].value = "" # type: ignore + child.value = "" + node.children[-1].value = "" elif child.type != token.STAR: # insert invisible parentheses node.insert_child(index, Leaf(token.LPAR, "")) @@ -790,24 +919,114 @@ def normalize_invisible_parens(node: Node, parens_after: Set[str]) -> None: elif not (isinstance(child, Leaf) and is_multiline_string(child)): wrap_in_parentheses(node, child, visible=False) - check_lpar = isinstance(child, Leaf) and child.value in parens_after + comma_check = child.type == token.COMMA if preview else False + + check_lpar = isinstance(child, Leaf) and ( + child.value in parens_after or comma_check + ) + + +def remove_await_parens(node: Node) -> None: + if node.children[0].type == token.AWAIT and len(node.children) > 1: + if ( + node.children[1].type == syms.atom + and node.children[1].children[0].type == token.LPAR + ): + if maybe_make_parens_invisible_in_atom( + node.children[1], + parent=node, + remove_brackets_around_comma=True, + ): + wrap_in_parentheses(node, node.children[1], visible=False) + + # Since await is an expression we shouldn't remove + # brackets in cases where this would change + # the AST due to operator precedence. + # Therefore we only aim to remove brackets around + # power nodes that aren't also await expressions themselves. + # https://peps.python.org/pep-0492/#updated-operator-precedence-table + # N.B. We've still removed any redundant nested brackets though :) + opening_bracket = cast(Leaf, node.children[1].children[0]) + closing_bracket = cast(Leaf, node.children[1].children[-1]) + bracket_contents = cast(Node, node.children[1].children[1]) + if bracket_contents.type != syms.power: + ensure_visible(opening_bracket) + ensure_visible(closing_bracket) + elif ( + bracket_contents.type == syms.power + and bracket_contents.children[0].type == token.AWAIT + ): + ensure_visible(opening_bracket) + ensure_visible(closing_bracket) + # If we are in a nested await then recurse down. + remove_await_parens(bracket_contents) + + +def remove_with_parens(node: Node, parent: Node) -> None: + """Recursively hide optional parens in `with` statements.""" + # Removing all unnecessary parentheses in with statements in one pass is a tad + # complex as different variations of bracketed statements result in pretty + # different parse trees: + # + # with (open("file")) as f: # this is an asexpr_test + # ... + # + # with (open("file") as f): # this is an atom containing an + # ... # asexpr_test + # + # with (open("file")) as f, (open("file")) as f: # this is asexpr_test, COMMA, + # ... # asexpr_test + # + # with (open("file") as f, open("file") as f): # an atom containing a + # ... # testlist_gexp which then + # # contains multiple asexpr_test(s) + if node.type == syms.atom: + if maybe_make_parens_invisible_in_atom( + node, + parent=parent, + remove_brackets_around_comma=True, + ): + wrap_in_parentheses(parent, node, visible=False) + if isinstance(node.children[1], Node): + remove_with_parens(node.children[1], node) + elif node.type == syms.testlist_gexp: + for child in node.children: + if isinstance(child, Node): + remove_with_parens(child, node) + elif node.type == syms.asexpr_test and not any( + leaf.type == token.COLONEQUAL for leaf in node.leaves() + ): + if maybe_make_parens_invisible_in_atom( + node.children[0], + parent=node, + remove_brackets_around_comma=True, + ): + wrap_in_parentheses(node, node.children[0], visible=False) -def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool: +def maybe_make_parens_invisible_in_atom( + node: LN, + parent: LN, + remove_brackets_around_comma: bool = False, +) -> bool: """If it's safe, make the parens in the atom `node` invisible, recursively. Additionally, remove repeated, adjacent invisible parens from the atom `node` as they are redundant. Returns whether the node should itself be wrapped in invisible parentheses. - """ - if ( node.type != syms.atom or is_empty_tuple(node) or is_one_tuple(node) or (is_yield(node) and parent.type != syms.expr_stmt) - or max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY + or ( + # This condition tries to prevent removing non-optional brackets + # around a tuple, however, can be a bit overzealous so we provide + # and option to skip this check for `for` and `with` statements. + not remove_brackets_around_comma + and max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY + ) ): return False @@ -825,12 +1044,16 @@ def maybe_make_parens_invisible_in_atom(node: LN, parent: LN) -> bool: first = node.children[0] last = node.children[-1] - if first.type == token.LPAR and last.type == token.RPAR: + if is_lpar_token(first) and is_rpar_token(last): middle = node.children[1] # make parentheses invisible - first.value = "" # type: ignore - last.value = "" # type: ignore - maybe_make_parens_invisible_in_atom(middle, parent=parent) + first.value = "" + last.value = "" + maybe_make_parens_invisible_in_atom( + middle, + parent=parent, + remove_brackets_around_comma=remove_brackets_around_comma, + ) if is_atom_with_invisible_parens(middle): # Strip the invisible parens from `middle` by replacing @@ -905,7 +1128,8 @@ def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[Leaf if ( prev and prev.type == token.COMMA - and not is_one_tuple_between( + and leaf.opening_bracket is not None + and not is_one_sequence_between( leaf.opening_bracket, leaf, line.leaves ) ): @@ -932,7 +1156,8 @@ def generate_trailers_to_omit(line: Line, line_length: int) -> Iterator[Set[Leaf if ( prev and prev.type == token.COMMA - and not is_one_tuple_between(leaf.opening_bracket, leaf, line.leaves) + and leaf.opening_bracket is not None + and not is_one_sequence_between(leaf.opening_bracket, leaf, line.leaves) ): # Never omit bracket pairs with trailing commas. # We need to explode on those. @@ -960,14 +1185,19 @@ def run_transformer( result.extend(transform_line(transformed_line, mode=mode, features=features)) - if not ( - transform.__name__ == "rhs" - and line.bracket_tracker.invisible - and not any(bracket.value for bracket in line.bracket_tracker.invisible) - and not line.contains_multiline_strings() - and not result[0].contains_uncollapsable_type_comments() - and not result[0].contains_unsplittable_type_ignore() - and not is_line_short_enough(result[0], line_length=mode.line_length) + if ( + transform.__class__.__name__ != "rhs" + or not line.bracket_tracker.invisible + or any(bracket.value for bracket in line.bracket_tracker.invisible) + or line.contains_multiline_strings() + or result[0].contains_uncollapsable_type_comments() + or result[0].contains_unsplittable_type_ignore() + or is_line_short_enough(result[0], line_length=mode.line_length) + # If any leaves have no parents (which _can_ occur since + # `transform(line)` potentially destroys the line's underlying node + # structure), then we can't proceed. Doing so would cause the below + # call to `append_leaves()` to fail. + or any(leaf.parent is None for leaf in line.leaves) ): return result