]> git.madduck.net Git - etc/vim.git/blobdiff - src/black/linegen.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:

Report all stacktraces in verbose mode (#3938)
[etc/vim.git] / src / black / linegen.py
index fe6ea11c501abba9f2116f15a08c40d8880ac1af..bdc4ee54ab285351588659f458f804aa212573c2 100644 (file)
@@ -2,7 +2,7 @@
 Generating lines of code.
 """
 import sys
-from dataclasses import dataclass
+from dataclasses import replace
 from enum import Enum, auto
 from functools import partial, wraps
 from typing import Collection, Iterator, List, Optional, Set, Union, cast
@@ -16,6 +16,7 @@ from black.brackets import (
 from black.comments import FMT_OFF, generate_comments, list_comments
 from black.lines import (
     Line,
+    RHSResult,
     append_leaves,
     can_be_split,
     can_omit_invisible_parens,
@@ -35,6 +36,7 @@ from black.nodes import (
     Visitor,
     ensure_visible,
     is_arith_like,
+    is_async_stmt_or_funcdef,
     is_atom_with_invisible_parens,
     is_docstring,
     is_empty_tuple,
@@ -46,6 +48,8 @@ from black.nodes import (
     is_rpar_token,
     is_stub_body,
     is_stub_suite,
+    is_tuple_containing_walrus,
+    is_type_ignore_comment_string,
     is_vararg,
     is_walrus_assignment,
     is_yield,
@@ -58,6 +62,7 @@ from black.strings import (
     get_string_prefix,
     normalize_string_prefix,
     normalize_string_quotes,
+    normalize_unicode_escape_sequences,
 )
 from black.trans import (
     CannotTransform,
@@ -89,8 +94,9 @@ class LineGenerator(Visitor[Line]):
     in ways that will no longer stringify to valid Python code on the tree.
     """
 
-    def __init__(self, mode: Mode) -> None:
+    def __init__(self, mode: Mode, features: Collection[Feature]) -> None:
         self.mode = mode
+        self.features = features
         self.current_line: Line
         self.__post_init__()
 
@@ -106,6 +112,17 @@ class LineGenerator(Visitor[Line]):
             self.current_line.depth += indent
             return  # Line is empty, don't emit. Creating a new one unnecessary.
 
+        if (
+            Preview.improved_async_statements_handling in self.mode
+            and len(self.current_line.leaves) == 1
+            and is_async_stmt_or_funcdef(self.current_line.leaves[0])
+        ):
+            # Special case for async def/for/with statements. `visit_async_stmt`
+            # adds an `ASYNC` leaf then visits the child def/for/with statement
+            # nodes. Line yields from those nodes shouldn't treat the former
+            # `ASYNC` leaf as a complete line.
+            return
+
         complete_line = self.current_line
         self.current_line = Line(mode=self.mode, depth=complete_line.depth + indent)
         yield complete_line
@@ -114,7 +131,7 @@ 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, preview=self.mode.preview):
+            for comment in generate_comments(node):
                 if any_open_brackets:
                     # any comment within brackets is subject to splitting
                     self.current_line.append(comment)
@@ -140,6 +157,22 @@ class LineGenerator(Visitor[Line]):
                 self.current_line.append(node)
         yield from super().visit_default(node)
 
+    def visit_test(self, node: Node) -> Iterator[Line]:
+        """Visit an `x if y else z` test"""
+
+        if Preview.parenthesize_conditional_expressions in self.mode:
+            already_parenthesized = (
+                node.prev_sibling and node.prev_sibling.type == token.LPAR
+            )
+
+            if not already_parenthesized:
+                lpar = Leaf(token.LPAR, "")
+                rpar = Leaf(token.RPAR, "")
+                node.insert_child(0, lpar)
+                node.append_child(rpar)
+
+        yield from self.visit_default(node)
+
     def visit_INDENT(self, node: Leaf) -> Iterator[Line]:
         """Increase indentation level, maybe yield a line."""
         # In blib2to3 INDENT never holds comments.
@@ -174,13 +207,27 @@ 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, preview=self.mode.preview)
+        normalize_invisible_parens(
+            node, parens_after=parens, mode=self.mode, features=self.features
+        )
         for child in node.children:
             if is_name_token(child) and child.value in keywords:
                 yield from self.line()
 
             yield from self.visit(child)
 
+    def visit_typeparams(self, node: Node) -> Iterator[Line]:
+        yield from self.visit_default(node)
+        node.children[0].prefix = ""
+
+    def visit_typevartuple(self, node: Node) -> Iterator[Line]:
+        yield from self.visit_default(node)
+        node.children[1].prefix = ""
+
+    def visit_paramspec(self, node: Node) -> Iterator[Line]:
+        yield from self.visit_default(node)
+        node.children[1].prefix = ""
+
     def visit_dictsetmaker(self, node: Node) -> Iterator[Line]:
         if Preview.wrap_long_dict_values_in_parens in self.mode:
             for i, child in enumerate(node.children):
@@ -200,34 +247,33 @@ class LineGenerator(Visitor[Line]):
 
     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()
+        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:
+        # 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)
-                    is_return_annotation = False
+                else:
+                    wrap_in_parentheses(node, child, visible=False)
+                is_return_annotation = False
 
-            for child in node.children:
-                yield from self.visit(child)
+        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)
+        normalize_invisible_parens(
+            node, parens_after=set(), mode=self.mode, features=self.features
+        )
 
         yield from self.line()
         for child in node.children:
@@ -235,7 +281,9 @@ class LineGenerator(Visitor[Line]):
 
     def visit_suite(self, node: Node) -> Iterator[Line]:
         """Visit a suite."""
-        if self.mode.is_pyi and is_stub_suite(node):
+        if (
+            self.mode.is_pyi or Preview.dummy_implementations in self.mode
+        ) and is_stub_suite(node):
             yield from self.visit(node.children[2])
         else:
             yield from self.visit_default(node)
@@ -250,7 +298,9 @@ class LineGenerator(Visitor[Line]):
 
         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):
+            if (
+                self.mode.is_pyi or Preview.dummy_implementations in self.mode
+            ) and is_stub_body(node):
                 yield from self.visit_default(node)
             else:
                 yield from self.line(+1)
@@ -259,7 +309,7 @@ class LineGenerator(Visitor[Line]):
 
         else:
             if (
-                not self.mode.is_pyi
+                not (self.mode.is_pyi or Preview.dummy_implementations in self.mode)
                 or not node.parent
                 or not is_stub_suite(node.parent)
             ):
@@ -280,8 +330,11 @@ class LineGenerator(Visitor[Line]):
                 break
 
         internal_stmt = next(children)
-        for child in internal_stmt.children:
-            yield from self.visit(child)
+        if Preview.improved_async_statements_handling in self.mode:
+            yield from self.visit(internal_stmt)
+        else:
+            for child in internal_stmt.children:
+                yield from self.visit(child)
 
     def visit_decorators(self, node: Node) -> Iterator[Line]:
         """Visit decorators."""
@@ -309,8 +362,7 @@ class LineGenerator(Visitor[Line]):
             ):
                 wrap_in_parentheses(node, leaf)
 
-        if Preview.remove_redundant_parens in self.mode:
-            remove_await_parens(node)
+        remove_await_parens(node)
 
         yield from self.visit_default(node)
 
@@ -345,28 +397,42 @@ class LineGenerator(Visitor[Line]):
             node.insert_child(index, Node(syms.atom, [lpar, operand, rpar]))
         yield from self.visit_default(node)
 
+    def visit_tname(self, node: Node) -> Iterator[Line]:
+        """
+        Add potential parentheses around types in function parameter lists to be made
+        into real parentheses in case the type hint is too long to fit on a line
+        Examples:
+        def foo(a: int, b: float = 7): ...
+
+        ->
+
+        def foo(a: (int), b: (float) = 7): ...
+        """
+        if Preview.parenthesize_long_type_hints in self.mode:
+            assert len(node.children) == 3
+            if maybe_make_parens_invisible_in_atom(node.children[2], parent=node):
+                wrap_in_parentheses(node, node.children[2], visible=False)
+
+        yield from self.visit_default(node)
+
     def visit_STRING(self, leaf: Leaf) -> Iterator[Line]:
+        if Preview.hex_codes_in_unicode_sequences in self.mode:
+            normalize_unicode_escape_sequences(leaf)
+
         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.
-            if Preview.normalize_docstring_quotes_and_prefixes_properly in self.mode:
-                # There was a bug where --skip-string-normalization wouldn't stop us
-                # from normalizing docstring prefixes. To maintain stability, we can
-                # only address this buggy behaviour while the preview style is enabled.
-                if self.mode.string_normalization:
-                    docstring = normalize_string_prefix(leaf.value)
-                    # visit_default() does handle string normalization for us, but
-                    # since this method acts differently depending on quote style (ex.
-                    # see padding logic below), there's a possibility for unstable
-                    # formatting as visit_default() is called *after*. To avoid a
-                    # situation where this function formats a docstring differently on
-                    # the second pass, normalize it early.
-                    docstring = normalize_string_quotes(docstring)
-                else:
-                    docstring = leaf.value
-            else:
-                # ... otherwise, we'll keep the buggy behaviour >.<
+            if self.mode.string_normalization:
                 docstring = normalize_string_prefix(leaf.value)
+                # visit_default() does handle string normalization for us, but
+                # since this method acts differently depending on quote style (ex.
+                # see padding logic below), there's a possibility for unstable
+                # formatting as visit_default() is called *after*. To avoid a
+                # situation where this function formats a docstring differently on
+                # the second pass, normalize it early.
+                docstring = normalize_string_quotes(docstring)
+            else:
+                docstring = leaf.value
             prefix = get_string_prefix(docstring)
             docstring = docstring[len(prefix) :]  # Remove the prefix
             quote_char = docstring[0]
@@ -385,6 +451,7 @@ class LineGenerator(Visitor[Line]):
             else:
                 docstring = docstring.strip()
 
+            has_trailing_backslash = False
             if docstring:
                 # Add some padding if the docstring starts / ends with a quote mark.
                 if docstring[0] == quote_char:
@@ -397,6 +464,7 @@ class LineGenerator(Visitor[Line]):
                         # Odd number of tailing backslashes, add some padding to
                         # avoid escaping the closing string quote.
                         docstring += " "
+                        has_trailing_backslash = True
             elif not docstring_started_empty:
                 docstring = " "
 
@@ -404,7 +472,7 @@ class LineGenerator(Visitor[Line]):
             quote = quote_char * quote_len
 
             # It's invalid to put closing single-character quotes on a new line.
-            if Preview.long_docstring_quotes_on_newline in self.mode and quote_len == 3:
+            if self.mode and quote_len == 3:
                 # We need to find the length of the last line of the docstring
                 # to find if we can add the closing quotes to the line without
                 # exceeding the maximum line length.
@@ -419,6 +487,8 @@ class LineGenerator(Visitor[Line]):
                 if (
                     len(lines) > 1
                     and last_line_length + quote_len > self.mode.line_length
+                    and len(indent) + quote_len <= self.mode.line_length
+                    and not has_trailing_backslash
                 ):
                     leaf.value = prefix + quote + docstring + "\n" + indent + quote
                 else:
@@ -443,16 +513,17 @@ class LineGenerator(Visitor[Line]):
         self.visit_try_stmt = partial(
             v, keywords={"try", "except", "else", "finally"}, 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_except_clause = partial(v, keywords={"except"}, parens={"except"})
+        self.visit_with_stmt = partial(v, keywords={"with"}, parens={"with"})
         self.visit_classdef = partial(v, keywords={"class"}, parens=Ø)
-        self.visit_expr_stmt = partial(v, keywords=Ø, parens=ASSIGNMENTS)
+
+        # When this is moved out of preview, add ":" directly to ASSIGNMENTS in nodes.py
+        if Preview.parenthesize_long_type_hints in self.mode:
+            assignments = ASSIGNMENTS | {":"}
+        else:
+            assignments = ASSIGNMENTS
+        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"})
@@ -492,7 +563,7 @@ def transform_line(
         and not line.should_split_rhs
         and not line.magic_trailing_comma
         and (
-            is_line_short_enough(line, line_length=mode.line_length, line_str=line_str)
+            is_line_short_enough(line, mode=mode, line_str=line_str)
             or line.contains_unsplittable_type_ignore()
         )
         and not (line.inside_brackets and line.contains_standalone_comments())
@@ -502,12 +573,12 @@ def transform_line(
             transformers = [string_merge, string_paren_strip]
         else:
             transformers = []
-    elif line.is_def:
+    elif line.is_def and not should_split_funcdef_with_rhs(line, mode):
         transformers = [left_hand_split]
     else:
 
         def _rhs(
-            self: object, line: Line, features: Collection[Feature]
+            self: object, line: Line, features: Collection[Feature], mode: Mode
         ) -> Iterator[Line]:
             """Wraps calls to `right_hand_split`.
 
@@ -516,14 +587,12 @@ def transform_line(
             bracket pair instead.
             """
             for omit in generate_trailers_to_omit(line, mode.line_length):
-                lines = list(
-                    right_hand_split(line, mode.line_length, features, omit=omit)
-                )
+                lines = list(right_hand_split(line, mode, features, omit=omit))
                 # Note: this check is only able to figure out if the first line of the
                 # *current* transformation fits in the line length.  This is true only
                 # for simple cases.  All others require running more transforms via
                 # `transform_line()`.  This check doesn't know if those would succeed.
-                if is_line_short_enough(lines[0], line_length=mode.line_length):
+                if is_line_short_enough(lines[0], mode=mode):
                     yield from lines
                     return
 
@@ -531,9 +600,7 @@ def transform_line(
             # This mostly happens to multiline strings that are by definition
             # reported as not fitting a single line, as well as lines that contain
             # trailing commas (those have to be exploded).
-            yield from right_hand_split(
-                line, line_length=mode.line_length, features=features
-            )
+            yield from right_hand_split(line, mode, features=features)
 
         # HACK: nested functions (like _rhs) compiled by mypyc don't retain their
         # __name__ attribute which is needed in `run_transformer` further down.
@@ -585,13 +652,49 @@ def transform_line(
         yield line
 
 
+def should_split_funcdef_with_rhs(line: Line, mode: Mode) -> bool:
+    """If a funcdef has a magic trailing comma in the return type, then we should first
+    split the line with rhs to respect the comma.
+    """
+    if Preview.respect_magic_trailing_comma_in_return_type not in mode:
+        return False
+
+    return_type_leaves: List[Leaf] = []
+    in_return_type = False
+
+    for leaf in line.leaves:
+        if leaf.type == token.COLON:
+            in_return_type = False
+        if in_return_type:
+            return_type_leaves.append(leaf)
+        if leaf.type == token.RARROW:
+            in_return_type = True
+
+    # using `bracket_split_build_line` will mess with whitespace, so we duplicate a
+    # couple lines from it.
+    result = Line(mode=line.mode, depth=line.depth)
+    leaves_to_track = get_leaves_inside_matching_brackets(return_type_leaves)
+    for leaf in return_type_leaves:
+        result.append(
+            leaf,
+            preformatted=True,
+            track_bracket=id(leaf) in leaves_to_track,
+        )
+
+    # we could also return true if the line is too long, and the return type is longer
+    # than the param list. Or if `should_split_rhs` returns True.
+    return result.magic_trailing_comma is not None
+
+
 class _BracketSplitComponent(Enum):
     head = auto()
     body = auto()
     tail = auto()
 
 
-def left_hand_split(line: Line, _features: Collection[Feature] = ()) -> Iterator[Line]:
+def left_hand_split(
+    line: Line, _features: Collection[Feature], mode: Mode
+) -> 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.
@@ -636,20 +739,9 @@ def left_hand_split(line: Line, _features: Collection[Feature] = ()) -> Iterator
             yield result
 
 
-@dataclass
-class _RHSResult:
-    """Intermediate split result from a right hand split."""
-
-    head: Line
-    body: Line
-    tail: Line
-    opening_bracket: Leaf
-    closing_bracket: Leaf
-
-
 def right_hand_split(
     line: Line,
-    line_length: int,
+    mode: Mode,
     features: Collection[Feature] = (),
     omit: Collection[LeafID] = (),
 ) -> Iterator[Line]:
@@ -663,14 +755,14 @@ def right_hand_split(
     """
     rhs_result = _first_right_hand_split(line, omit=omit)
     yield from _maybe_split_omitting_optional_parens(
-        rhs_result, line, line_length, features=features, omit=omit
+        rhs_result, line, mode, features=features, omit=omit
     )
 
 
 def _first_right_hand_split(
     line: Line,
     omit: Collection[LeafID] = (),
-) -> _RHSResult:
+) -> RHSResult:
     """Split the line into head, body, tail starting with the last bracket pair.
 
     Note: this function should not have side effects. It's relied upon by
@@ -712,13 +804,13 @@ def _first_right_hand_split(
         tail_leaves, line, opening_bracket, component=_BracketSplitComponent.tail
     )
     bracket_split_succeeded_or_raise(head, body, tail)
-    return _RHSResult(head, body, tail, opening_bracket, closing_bracket)
+    return RHSResult(head, body, tail, opening_bracket, closing_bracket)
 
 
 def _maybe_split_omitting_optional_parens(
-    rhs: _RHSResult,
+    rhs: RHSResult,
     line: Line,
-    line_length: int,
+    mode: Mode,
     features: Collection[Feature] = (),
     omit: Collection[LeafID] = (),
 ) -> Iterator[Line]:
@@ -736,38 +828,39 @@ def _maybe_split_omitting_optional_parens(
         # there are no standalone comments in the body
         and not rhs.body.contains_standalone_comments(0)
         # and we can actually remove the parens
-        and can_omit_invisible_parens(rhs.body, line_length)
+        and can_omit_invisible_parens(rhs, mode.line_length)
     ):
         omit = {id(rhs.closing_bracket), *omit}
         try:
-            # The _RHSResult Omitting Optional Parens.
+            # The RHSResult Omitting Optional Parens.
             rhs_oop = _first_right_hand_split(line, omit=omit)
             if not (
                 Preview.prefer_splitting_right_hand_side_of_assignments in line.mode
                 # the split is right after `=`
                 and len(rhs.head.leaves) >= 2
                 and rhs.head.leaves[-2].type == token.EQUAL
-                # the left side of assignement contains brackets
+                # the left side of assignment contains brackets
                 and any(leaf.type in BRACKETS for leaf in rhs.head.leaves[:-1])
                 # the left side of assignment is short enough (the -1 is for the ending
                 # optional paren)
-                and is_line_short_enough(rhs.head, line_length=line_length - 1)
+                and is_line_short_enough(
+                    rhs.head, mode=replace(mode, line_length=mode.line_length - 1)
+                )
                 # the left side of assignment won't explode further because of magic
                 # trailing comma
                 and rhs.head.magic_trailing_comma is None
                 # the split by omitting optional parens isn't preferred by some other
                 # reason
-                and not _prefer_split_rhs_oop(rhs_oop, line_length=line_length)
+                and not _prefer_split_rhs_oop(rhs_oop, mode)
             ):
                 yield from _maybe_split_omitting_optional_parens(
-                    rhs_oop, line, line_length, features=features, omit=omit
+                    rhs_oop, line, mode, features=features, omit=omit
                 )
                 return
 
         except CannotSplit as e:
             if not (
-                can_be_split(rhs.body)
-                or is_line_short_enough(rhs.body, line_length=line_length)
+                can_be_split(rhs.body) or is_line_short_enough(rhs.body, mode=mode)
             ):
                 raise CannotSplit(
                     "Splitting failed, body is still too long and can't be split."
@@ -791,7 +884,7 @@ def _maybe_split_omitting_optional_parens(
             yield result
 
 
-def _prefer_split_rhs_oop(rhs_oop: _RHSResult, line_length: int) -> bool:
+def _prefer_split_rhs_oop(rhs_oop: RHSResult, mode: Mode) -> bool:
     """
     Returns whether we should prefer the result from a split omitting optional parens.
     """
@@ -811,7 +904,7 @@ def _prefer_split_rhs_oop(rhs_oop: _RHSResult, line_length: int) -> bool:
             # the first line still contains the `=`)
             any(leaf.type == token.EQUAL for leaf in rhs_oop.head.leaves)
             # the first line is short enough
-            and is_line_short_enough(rhs_oop.head, line_length=line_length)
+            and is_line_short_enough(rhs_oop.head, mode=mode)
         )
         # contains unsplittable type ignore
         or rhs_oop.head.contains_unsplittable_type_ignore()
@@ -889,6 +982,13 @@ def bracket_split_build_line(
                     )
                     if isinstance(node, Node) and isinstance(node.prev_sibling, Leaf)
                 )
+                # Except the false negatives above for PEP 604 unions where we
+                # can't add the comma.
+                and not (
+                    leaves[0].parent
+                    and leaves[0].parent.next_sibling
+                    and leaves[0].parent.next_sibling.type == token.VBAR
+                )
             )
 
             if original.is_import or no_commas:
@@ -902,10 +1002,7 @@ def bracket_split_build_line(
                     break
 
     leaves_to_track: Set[LeafID] = set()
-    if (
-        Preview.handle_trailing_commas_in_head in original.mode
-        and component is _BracketSplitComponent.head
-    ):
+    if component is _BracketSplitComponent.head:
         leaves_to_track = get_leaves_inside_matching_brackets(leaves)
     # Populate the line
     for leaf in leaves:
@@ -930,16 +1027,39 @@ def dont_increase_indentation(split_func: Transformer) -> Transformer:
     """
 
     @wraps(split_func)
-    def split_wrapper(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
-        for split_line in split_func(line, features):
+    def split_wrapper(
+        line: Line, features: Collection[Feature], mode: Mode
+    ) -> Iterator[Line]:
+        for split_line in split_func(line, features, mode):
             normalize_prefix(split_line.leaves[0], inside_brackets=True)
             yield split_line
 
     return split_wrapper
 
 
+def _get_last_non_comment_leaf(line: Line) -> Optional[int]:
+    for leaf_idx in range(len(line.leaves) - 1, 0, -1):
+        if line.leaves[leaf_idx].type != STANDALONE_COMMENT:
+            return leaf_idx
+    return None
+
+
+def _safe_add_trailing_comma(safe: bool, delimiter_priority: int, line: Line) -> Line:
+    if (
+        safe
+        and delimiter_priority == COMMA_PRIORITY
+        and line.leaves[-1].type != token.COMMA
+        and line.leaves[-1].type != STANDALONE_COMMENT
+    ):
+        new_comma = Leaf(token.COMMA, ",")
+        line.append(new_comma)
+    return line
+
+
 @dont_increase_indentation
-def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[Line]:
+def delimiter_split(
+    line: Line, features: Collection[Feature], mode: Mode
+) -> Iterator[Line]:
     """Split according to delimiters of the highest priority.
 
     If the appropriate Features are given, the split will add trailing commas
@@ -979,7 +1099,8 @@ def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[
             )
             current_line.append(leaf)
 
-    for leaf in line.leaves:
+    last_non_comment_leaf = _get_last_non_comment_leaf(line)
+    for leaf_idx, leaf in enumerate(line.leaves):
         yield from append_to_line(leaf)
 
         for comment_after in line.comments_after(leaf):
@@ -996,6 +1117,15 @@ def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[
                     trailing_comma_safe and Feature.TRAILING_COMMA_IN_CALL in features
                 )
 
+        if (
+            Preview.add_trailing_comma_consistently in mode
+            and last_leaf.type == STANDALONE_COMMENT
+            and leaf_idx == last_non_comment_leaf
+        ):
+            current_line = _safe_add_trailing_comma(
+                trailing_comma_safe, delimiter_priority, current_line
+            )
+
         leaf_priority = bt.delimiters.get(id(leaf))
         if leaf_priority == delimiter_priority:
             yield current_line
@@ -1004,20 +1134,15 @@ def delimiter_split(line: Line, features: Collection[Feature] = ()) -> Iterator[
                 mode=line.mode, depth=line.depth, inside_brackets=line.inside_brackets
             )
     if current_line:
-        if (
-            trailing_comma_safe
-            and delimiter_priority == COMMA_PRIORITY
-            and current_line.leaves[-1].type != token.COMMA
-            and current_line.leaves[-1].type != STANDALONE_COMMENT
-        ):
-            new_comma = Leaf(token.COMMA, ",")
-            current_line.append(new_comma)
+        current_line = _safe_add_trailing_comma(
+            trailing_comma_safe, delimiter_priority, current_line
+        )
         yield current_line
 
 
 @dont_increase_indentation
 def standalone_comment_split(
-    line: Line, features: Collection[Feature] = ()
+    line: Line, features: Collection[Feature], mode: Mode
 ) -> Iterator[Line]:
     """Split standalone comments from the rest of the line."""
     if not line.contains_standalone_comments(0):
@@ -1069,7 +1194,7 @@ def normalize_prefix(leaf: Leaf, *, inside_brackets: bool) -> None:
 
 
 def normalize_invisible_parens(
-    node: Node, parens_after: Set[str], *, preview: bool
+    node: Node, parens_after: Set[str], *, mode: Mode, features: Collection[Feature]
 ) -> None:
     """Make existing optional parentheses invisible or create new ones.
 
@@ -1079,17 +1204,24 @@ def normalize_invisible_parens(
     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, preview=preview):
+    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
+
+    # The multiple context managers grammar has a different pattern, thus this is
+    # separate from the for-loop below. This possibly wraps them in invisible parens,
+    # and later will be removed in remove_with_parens when needed.
+    if node.type == syms.with_stmt:
+        _maybe_wrap_cms_in_parens(node, mode, features)
+
     check_lpar = False
     for index, child in enumerate(list(node.children)):
         # 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, preview=preview
+                child, parens_after=parens_after, mode=mode, features=features
             )
 
         # Add parentheses around long tuple unpacking in assignments.
@@ -1102,8 +1234,7 @@ def normalize_invisible_parens(
 
         if check_lpar:
             if (
-                preview
-                and child.type == syms.atom
+                child.type == syms.atom
                 and node.type == syms.for_stmt
                 and isinstance(child.prev_sibling, Leaf)
                 and child.prev_sibling.type == token.NAME
@@ -1115,7 +1246,7 @@ def normalize_invisible_parens(
                     remove_brackets_around_comma=True,
                 ):
                     wrap_in_parentheses(node, child, visible=False)
-            elif preview and isinstance(child, Node) and node.type == syms.with_stmt:
+            elif 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(
@@ -1126,17 +1257,7 @@ def normalize_invisible_parens(
             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 is_lpar_token(child):
-                    assert is_rpar_token(node.children[-1])
-                    # make parentheses invisible
-                    child.value = ""
-                    node.children[-1].value = ""
-                elif child.type != token.STAR:
-                    # insert invisible parentheses
-                    node.insert_child(index, Leaf(token.LPAR, ""))
-                    node.append_child(Leaf(token.RPAR, ""))
+                _normalize_import_from(node, child, index)
                 break
             elif (
                 index == 1
@@ -1151,13 +1272,27 @@ def normalize_invisible_parens(
             elif not (isinstance(child, Leaf) and is_multiline_string(child)):
                 wrap_in_parentheses(node, child, visible=False)
 
-        comma_check = child.type == token.COMMA if preview else False
+        comma_check = child.type == token.COMMA
 
         check_lpar = isinstance(child, Leaf) and (
             child.value in parens_after or comma_check
         )
 
 
+def _normalize_import_from(parent: Node, child: LN, index: int) -> None:
+    # "import from" nodes store parentheses directly as part of
+    # the statement
+    if is_lpar_token(child):
+        assert is_rpar_token(parent.children[-1])
+        # make parentheses invisible
+        child.value = ""
+        parent.children[-1].value = ""
+    elif child.type != token.STAR:
+        # insert invisible parentheses
+        parent.insert_child(index, Leaf(token.LPAR, ""))
+        parent.append_child(Leaf(token.RPAR, ""))
+
+
 def remove_await_parens(node: Node) -> None:
     if node.children[0].type == token.AWAIT and len(node.children) > 1:
         if (
@@ -1180,18 +1315,62 @@ def remove_await_parens(node: Node) -> None:
             # 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)
+            bracket_contents = node.children[1].children[1]
+            if isinstance(bracket_contents, Node):
+                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 _maybe_wrap_cms_in_parens(
+    node: Node, mode: Mode, features: Collection[Feature]
+) -> None:
+    """When enabled and safe, wrap the multiple context managers in invisible parens.
+
+    It is only safe when `features` contain Feature.PARENTHESIZED_CONTEXT_MANAGERS.
+    """
+    if (
+        Feature.PARENTHESIZED_CONTEXT_MANAGERS not in features
+        or Preview.wrap_multiple_context_managers_in_parens not in mode
+        or len(node.children) <= 2
+        # If it's an atom, it's already wrapped in parens.
+        or node.children[1].type == syms.atom
+    ):
+        return
+    colon_index: Optional[int] = None
+    for i in range(2, len(node.children)):
+        if node.children[i].type == token.COLON:
+            colon_index = i
+            break
+    if colon_index is not None:
+        lpar = Leaf(token.LPAR, "")
+        rpar = Leaf(token.RPAR, "")
+        context_managers = node.children[1:colon_index]
+        for child in context_managers:
+            child.remove()
+        # After wrapping, the with_stmt will look like this:
+        #   with_stmt
+        #     NAME 'with'
+        #     atom
+        #       LPAR ''
+        #       testlist_gexp
+        #         ... <-- context_managers
+        #       /testlist_gexp
+        #       RPAR ''
+        #     /atom
+        #     COLON ':'
+        new_child = Node(
+            syms.atom, [lpar, Node(syms.testlist_gexp, context_managers), rpar]
+        )
+        node.insert_child(1, new_child)
 
 
 def remove_with_parens(node: Node, parent: Node) -> None:
@@ -1248,7 +1427,7 @@ def maybe_make_parens_invisible_in_atom(
     Returns whether the node should itself be wrapped in invisible parentheses.
     """
     if (
-        node.type != syms.atom
+        node.type not in (syms.atom, syms.expr)
         or is_empty_tuple(node)
         or is_one_tuple(node)
         or (is_yield(node) and parent.type != syms.expr_stmt)
@@ -1259,6 +1438,7 @@ def maybe_make_parens_invisible_in_atom(
             not remove_brackets_around_comma
             and max_delimiter_priority_in_atom(node) >= COMMA_PRIORITY
         )
+        or is_tuple_containing_walrus(node)
     ):
         return False
 
@@ -1268,9 +1448,14 @@ def maybe_make_parens_invisible_in_atom(
             syms.expr_stmt,
             syms.assert_stmt,
             syms.return_stmt,
+            syms.except_clause,
+            syms.funcdef,
+            syms.with_stmt,
+            syms.tname,
             # these ones aren't useful to end users, but they do please fuzzers
             syms.for_stmt,
             syms.del_stmt,
+            syms.for_stmt,
         ]:
             return False
 
@@ -1279,8 +1464,13 @@ def maybe_make_parens_invisible_in_atom(
     if is_lpar_token(first) and is_rpar_token(last):
         middle = node.children[1]
         # make parentheses invisible
-        first.value = ""
-        last.value = ""
+        if (
+            # If the prefix of `middle` includes a type comment with
+            # ignore annotation, then we do not remove the parentheses
+            not is_type_ignore_comment_string(middle.prefix.strip())
+        ):
+            first.value = ""
+            last.value = ""
         maybe_make_parens_invisible_in_atom(
             middle,
             parent=parent,
@@ -1411,7 +1601,7 @@ def run_transformer(
     if not line_str:
         line_str = line_to_string(line)
     result: List[Line] = []
-    for transformed_line in transform(line, features):
+    for transformed_line in transform(line, features, mode):
         if str(transformed_line).strip("\n") == line_str:
             raise CannotTransform("Line transformer returned an unchanged result")
 
@@ -1426,7 +1616,7 @@ def run_transformer(
         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)
+        or is_line_short_enough(result[0], mode=mode)
         # 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
@@ -1441,8 +1631,6 @@ def run_transformer(
     second_opinion = run_transformer(
         line_copy, transform, mode, features_fop, line_str=line_str
     )
-    if all(
-        is_line_short_enough(ln, line_length=mode.line_length) for ln in second_opinion
-    ):
+    if all(is_line_short_enough(ln, mode=mode) for ln in second_opinion):
         result = second_opinion
     return result