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

CI Test: Deprecating 'set-output' command (#3757)
[etc/vim.git] / src / black / lines.py
index 08281bcf370a1a987fcffa9602d54a7fe3d71593..ea8fe520756d3a61702cf9e81453416bf534b632 100644 (file)
@@ -1,4 +1,5 @@
 import itertools
+import math
 import sys
 from dataclasses import dataclass, field
 from typing import (
@@ -10,10 +11,11 @@ from typing import (
     Sequence,
     Tuple,
     TypeVar,
+    Union,
     cast,
 )
 
-from black.brackets import DOT_PRIORITY, BracketTracker
+from black.brackets import COMMA_PRIORITY, DOT_PRIORITY, BracketTracker
 from black.mode import Mode, Preview
 from black.nodes import (
     BRACKETS,
@@ -26,10 +28,13 @@ from black.nodes import (
     is_multiline_string,
     is_one_sequence_between,
     is_type_comment,
+    is_type_ignore_comment,
+    is_with_or_async_with_stmt,
     replace_child,
     syms,
     whitespace,
 )
+from black.strings import str_width
 from blib2to3.pgen2 import token
 from blib2to3.pytree import Leaf, Node
 
@@ -37,6 +42,7 @@ from blib2to3.pytree import Leaf, Node
 T = TypeVar("T")
 Index = int
 LeafID = int
+LN = Union[Leaf, Node]
 
 
 @dataclass
@@ -119,6 +125,11 @@ class Line:
         """Is this an import line?"""
         return bool(self) and is_import(self.leaves[0])
 
+    @property
+    def is_with_or_async_with_stmt(self) -> bool:
+        """Is this a with_stmt line?"""
+        return bool(self) and is_with_or_async_with_stmt(self.leaves[0])
+
     @property
     def is_class(self) -> bool:
         """Is this line a class definition?"""
@@ -186,6 +197,26 @@ class Line:
             return False
         return self.leaves[-1].type == token.COLON
 
+    def is_fmt_pass_converted(
+        self, *, first_leaf_matches: Optional[Callable[[Leaf], bool]] = None
+    ) -> bool:
+        """Is this line converted from fmt off/skip code?
+
+        If first_leaf_matches is not None, it only returns True if the first
+        leaf of converted code matches.
+        """
+        if len(self.leaves) != 1:
+            return False
+        leaf = self.leaves[0]
+        if (
+            leaf.type != STANDALONE_COMMENT
+            or leaf.fmt_pass_converted_first_leaf is None
+        ):
+            return False
+        return first_leaf_matches is None or first_leaf_matches(
+            leaf.fmt_pass_converted_first_leaf
+        )
+
     def contains_standalone_comments(self, depth_limit: int = sys.maxsize) -> bool:
         """If so, needs to be split before emitting."""
         for leaf in self.leaves:
@@ -221,7 +252,7 @@ class Line:
             for comment in comments:
                 if is_type_comment(comment):
                     if comment_seen or (
-                        not is_type_comment(comment, " ignore")
+                        not is_type_ignore_comment(comment)
                         and leaf_id not in ignored_ids
                     ):
                         return True
@@ -258,7 +289,7 @@ class Line:
             # line.
             for node in self.leaves[-2:]:
                 for comment in self.comments.get(id(node), []):
-                    if is_type_comment(comment, " ignore"):
+                    if is_type_ignore_comment(comment):
                         return True
 
         return False
@@ -275,8 +306,7 @@ class Line:
         - it's not a single-element subscript
         Additionally, if ensure_removable:
         - it's not from square bracket indexing
-        (specifically, single-element square bracket indexing with
-        Preview.skip_magic_trailing_comma_in_subscript)
+        (specifically, single-element square bracket indexing)
         """
         if not (
             closing.type in CLOSING_BRACKETS
@@ -290,8 +320,7 @@ class Line:
 
         if closing.type == token.RSQB:
             if (
-                Preview.one_element_subscript in self.mode
-                and closing.parent
+                closing.parent
                 and closing.parent.type == syms.trailer
                 and closing.opening_bracket
                 and is_one_sequence_between(
@@ -309,18 +338,16 @@ class Line:
             comma = self.leaves[-1]
             if comma.parent is None:
                 return False
-            if Preview.skip_magic_trailing_comma_in_subscript in self.mode:
-                return (
-                    comma.parent.type != syms.subscriptlist
-                    or closing.opening_bracket is None
-                    or not is_one_sequence_between(
-                        closing.opening_bracket,
-                        closing,
-                        self.leaves,
-                        brackets=(token.LSQB, token.RSQB),
-                    )
+            return (
+                comma.parent.type != syms.subscriptlist
+                or closing.opening_bracket is None
+                or not is_one_sequence_between(
+                    closing.opening_bracket,
+                    closing,
+                    self.leaves,
+                    brackets=(token.LSQB, token.RSQB),
                 )
-            return comma.parent.type == syms.listmaker
+            )
 
         if self.is_import:
             return True
@@ -450,6 +477,17 @@ class Line:
         return bool(self.leaves or self.comments)
 
 
+@dataclass
+class RHSResult:
+    """Intermediate split result from a right hand split."""
+
+    head: Line
+    body: Line
+    tail: Line
+    opening_bracket: Leaf
+    closing_bracket: Leaf
+
+
 @dataclass
 class LinesBlock:
     """Class that holds information about a block of formatted lines.
@@ -485,7 +523,7 @@ class EmptyLineTracker:
     mode: Mode
     previous_line: Optional[Line] = None
     previous_block: Optional[LinesBlock] = None
-    previous_defs: List[int] = field(default_factory=list)
+    previous_defs: List[Line] = field(default_factory=list)
     semantic_leading_comment: Optional[LinesBlock] = None
 
     def maybe_empty_lines(self, current_line: Line) -> LinesBlock:
@@ -520,7 +558,8 @@ class EmptyLineTracker:
                 and (self.semantic_leading_comment is None or before)
             ):
                 self.semantic_leading_comment = block
-        elif not current_line.is_decorator:
+        # `or before` means this decorator already has an empty line before
+        elif not current_line.is_decorator or before:
             self.semantic_leading_comment = None
 
         self.previous_line = current_line
@@ -540,12 +579,18 @@ class EmptyLineTracker:
         else:
             before = 0
         depth = current_line.depth
-        while self.previous_defs and self.previous_defs[-1] >= depth:
+        while self.previous_defs and self.previous_defs[-1].depth >= depth:
             if self.mode.is_pyi:
                 assert self.previous_line is not None
                 if depth and not current_line.is_def and self.previous_line.is_def:
                     # Empty lines between attributes and methods should be preserved.
                     before = min(1, before)
+                elif (
+                    Preview.blank_line_after_nested_stub_class in self.mode
+                    and self.previous_defs[-1].is_class
+                    and not self.previous_defs[-1].is_stub_class
+                ):
+                    before = 1
                 elif depth:
                     before = 0
                 else:
@@ -555,7 +600,7 @@ class EmptyLineTracker:
                     before = 1
                 elif (
                     not depth
-                    and self.previous_defs[-1]
+                    and self.previous_defs[-1].depth
                     and current_line.leaves[-1].type == token.COLON
                     and (
                         current_line.leaves[0].value
@@ -580,6 +625,7 @@ class EmptyLineTracker:
             self.previous_line
             and self.previous_line.is_import
             and not current_line.is_import
+            and not current_line.is_fmt_pass_converted(first_leaf_matches=is_import)
             and depth == self.previous_line.depth
         ):
             return (before or 1), 0
@@ -589,13 +635,11 @@ class EmptyLineTracker:
             and self.previous_line.is_class
             and current_line.is_triple_quoted_string
         ):
+            if Preview.no_blank_line_before_class_docstring in current_line.mode:
+                return 0, 1
             return before, 1
 
-        if (
-            Preview.remove_block_trailing_newline in current_line.mode
-            and self.previous_line
-            and self.previous_line.opens_block
-        ):
+        if self.previous_line and self.previous_line.opens_block:
             return 0, 0
         return before, 0
 
@@ -603,7 +647,7 @@ class EmptyLineTracker:
         self, current_line: Line, before: int
     ) -> Tuple[int, int]:
         if not current_line.is_decorator:
-            self.previous_defs.append(current_line.depth)
+            self.previous_defs.append(current_line)
         if self.previous_line is None:
             # Don't insert empty lines before the first line in the file.
             return 0, 0
@@ -628,9 +672,7 @@ class EmptyLineTracker:
         ):
             slc = self.semantic_leading_comment
             if (
-                Preview.empty_lines_before_class_or_def_with_leading_comments
-                in current_line.mode
-                and slc is not None
+                slc is not None
                 and slc.previous_block is not None
                 and not slc.previous_block.original_line.is_class
                 and not slc.previous_block.original_line.opens_block
@@ -710,18 +752,95 @@ def append_leaves(
             new_line.append(comment_leaf, preformatted=True)
 
 
-def is_line_short_enough(line: Line, *, line_length: int, line_str: str = "") -> bool:
-    """Return True if `line` is no longer than `line_length`.
-
+def is_line_short_enough(  # noqa: C901
+    line: Line, *, mode: Mode, line_str: str = ""
+) -> bool:
+    """For non-multiline strings, return True if `line` is no longer than `line_length`.
+    For multiline strings, looks at the context around `line` to determine
+    if it should be inlined or split up.
     Uses the provided `line_str` rendering, if any, otherwise computes a new one.
     """
     if not line_str:
         line_str = line_to_string(line)
-    return (
-        len(line_str) <= line_length
-        and "\n" not in line_str  # multiline strings
-        and not line.contains_standalone_comments()
-    )
+
+    width = str_width if mode.preview else len
+
+    if Preview.multiline_string_handling not in mode:
+        return (
+            width(line_str) <= mode.line_length
+            and "\n" not in line_str  # multiline strings
+            and not line.contains_standalone_comments()
+        )
+
+    if line.contains_standalone_comments():
+        return False
+    if "\n" not in line_str:
+        # No multiline strings (MLS) present
+        return width(line_str) <= mode.line_length
+
+    first, *_, last = line_str.split("\n")
+    if width(first) > mode.line_length or width(last) > mode.line_length:
+        return False
+
+    # Traverse the AST to examine the context of the multiline string (MLS),
+    # tracking aspects such as depth and comma existence,
+    # to determine whether to split the MLS or keep it together.
+    # Depth (which is based on the existing bracket_depth concept)
+    # is needed to determine nesting level of the MLS.
+    # Includes special case for trailing commas.
+    commas: List[int] = []  # tracks number of commas per depth level
+    multiline_string: Optional[Leaf] = None
+    # store the leaves that contain parts of the MLS
+    multiline_string_contexts: List[LN] = []
+
+    max_level_to_update: Union[int, float] = math.inf  # track the depth of the MLS
+    for i, leaf in enumerate(line.leaves):
+        if max_level_to_update == math.inf:
+            had_comma: Optional[int] = None
+            if leaf.bracket_depth + 1 > len(commas):
+                commas.append(0)
+            elif leaf.bracket_depth + 1 < len(commas):
+                had_comma = commas.pop()
+            if (
+                had_comma is not None
+                and multiline_string is not None
+                and multiline_string.bracket_depth == leaf.bracket_depth + 1
+            ):
+                # Have left the level with the MLS, stop tracking commas
+                max_level_to_update = leaf.bracket_depth
+                if had_comma > 0:
+                    # MLS was in parens with at least one comma - force split
+                    return False
+
+        if leaf.bracket_depth <= max_level_to_update and leaf.type == token.COMMA:
+            # Ignore non-nested trailing comma
+            # directly after MLS/MLS-containing expression
+            ignore_ctxs: List[Optional[LN]] = [None]
+            ignore_ctxs += multiline_string_contexts
+            if not (leaf.prev_sibling in ignore_ctxs and i == len(line.leaves) - 1):
+                commas[leaf.bracket_depth] += 1
+        if max_level_to_update != math.inf:
+            max_level_to_update = min(max_level_to_update, leaf.bracket_depth)
+
+        if is_multiline_string(leaf):
+            if len(multiline_string_contexts) > 0:
+                # >1 multiline string cannot fit on a single line - force split
+                return False
+            multiline_string = leaf
+            ctx: LN = leaf
+            # fetch the leaf components of the MLS in the AST
+            while str(ctx) in line_str:
+                multiline_string_contexts.append(ctx)
+                if ctx.parent is None:
+                    break
+                ctx = ctx.parent
+
+    # May not have a triple-quoted multiline string at all,
+    # in case of a regular string with embedded newlines and line continuations
+    if len(multiline_string_contexts) == 0:
+        return True
+
+    return all(val == 0 for val in commas)
 
 
 def can_be_split(line: Line) -> bool:
@@ -761,25 +880,42 @@ def can_be_split(line: Line) -> bool:
 
 
 def can_omit_invisible_parens(
-    line: Line,
+    rhs: RHSResult,
     line_length: int,
 ) -> bool:
-    """Does `line` have a shape safe to reformat without optional parens around it?
+    """Does `rhs.body` have a shape safe to reformat without optional parens around it?
 
     Returns True for only a subset of potentially nice looking formattings but
     the point is to not return false positives that end up producing lines that
     are too long.
     """
+    line = rhs.body
     bt = line.bracket_tracker
     if not bt.delimiters:
         # Without delimiters the optional parentheses are useless.
         return True
 
     max_priority = bt.max_delimiter_priority()
-    if bt.delimiter_count_with_priority(max_priority) > 1:
+    delimiter_count = bt.delimiter_count_with_priority(max_priority)
+    if delimiter_count > 1:
         # With more than one delimiter of a kind the optional parentheses read better.
         return False
 
+    if delimiter_count == 1:
+        if (
+            Preview.wrap_multiple_context_managers_in_parens in line.mode
+            and max_priority == COMMA_PRIORITY
+            and rhs.head.is_with_or_async_with_stmt
+        ):
+            # For two context manager with statements, the optional parentheses read
+            # better. In this case, `rhs.body` is the context managers part of
+            # the with statement. `rhs.head` is the `with (` part on the previous
+            # line.
+            return False
+        # Otherwise it may also read better, but we don't do it today and requires
+        # careful considerations for all possible cases. See
+        # https://github.com/psf/black/issues/2156.
+
     if max_priority == DOT_PRIORITY:
         # A single stranded method call doesn't require optional parentheses.
         return True