]> 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:

blib2to3: add a few annotations (#3675)
[etc/vim.git] / src / black / lines.py
index 63225c0e6d31288bd7c68004e751eb05ca537382..9d33bfa10b434e4075e2124cd41f586512762ffe 100644 (file)
@@ -1,9 +1,9 @@
-from dataclasses import dataclass, field
 import itertools
+import math
 import sys
+from dataclasses import dataclass, field
 from typing import (
     Callable,
-    Collection,
     Dict,
     Iterator,
     List,
@@ -11,24 +11,37 @@ from typing import (
     Sequence,
     Tuple,
     TypeVar,
+    Union,
     cast,
 )
 
-from blib2to3.pytree import Node, Leaf
+from black.brackets import COMMA_PRIORITY, DOT_PRIORITY, BracketTracker
+from black.mode import Mode, Preview
+from black.nodes import (
+    BRACKETS,
+    CLOSING_BRACKETS,
+    OPENING_BRACKETS,
+    STANDALONE_COMMENT,
+    TEST_DESCENDANTS,
+    child_towards,
+    is_import,
+    is_multiline_string,
+    is_one_sequence_between,
+    is_type_comment,
+    is_with_or_async_with_stmt,
+    replace_child,
+    syms,
+    whitespace,
+)
+from black.strings import str_width
 from blib2to3.pgen2 import token
-
-from black.brackets import BracketTracker, DOT_PRIORITY
-from black.mode import Mode
-from black.nodes import STANDALONE_COMMENT, TEST_DESCENDANTS
-from black.nodes import BRACKETS, OPENING_BRACKETS, CLOSING_BRACKETS
-from black.nodes import syms, whitespace, replace_child, child_towards
-from black.nodes import is_multiline_string, is_import, is_type_comment, last_two_except
-from black.nodes import is_one_tuple_between
+from blib2to3.pytree import Leaf, Node
 
 # types
 T = TypeVar("T")
 Index = int
 LeafID = int
+LN = Union[Leaf, Node]
 
 
 @dataclass
@@ -45,7 +58,9 @@ class Line:
     should_split_rhs: bool = False
     magic_trailing_comma: Optional[Leaf] = None
 
-    def append(self, leaf: Leaf, preformatted: bool = False) -> None:
+    def append(
+        self, leaf: Leaf, preformatted: bool = False, track_bracket: bool = False
+    ) -> None:
         """Add a new `leaf` to the end of the line.
 
         Unless `preformatted` is True, the `leaf` will receive a new consistent
@@ -67,7 +82,7 @@ class Line:
             leaf.prefix += whitespace(
                 leaf, complex_subscript=self.is_complex_subscript(leaf)
             )
-        if self.inside_brackets or not preformatted:
+        if self.inside_brackets or not preformatted or track_bracket:
             self.bracket_tracker.mark(leaf)
             if self.mode.magic_trailing_comma:
                 if self.has_magic_trailing_comma(leaf):
@@ -109,6 +124,11 @@ 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?"""
@@ -169,6 +189,33 @@ class Line:
             and self.leaves[0].value.startswith(('"""', "'''"))
         )
 
+    @property
+    def opens_block(self) -> bool:
+        """Does this line open a new level of indentation."""
+        if len(self.leaves) == 0:
+            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:
@@ -255,8 +302,10 @@ class Line:
         """Return True if we have a magic trailing comma, that is when:
         - there's a trailing comma here
         - it's not a one-tuple
+        - it's not a single-element subscript
         Additionally, if ensure_removable:
         - it's not from square bracket indexing
+        (specifically, single-element square bracket indexing)
         """
         if not (
             closing.type in CLOSING_BRACKETS
@@ -269,15 +318,42 @@ class Line:
             return True
 
         if closing.type == token.RSQB:
+            if (
+                closing.parent
+                and closing.parent.type == syms.trailer
+                and closing.opening_bracket
+                and is_one_sequence_between(
+                    closing.opening_bracket,
+                    closing,
+                    self.leaves,
+                    brackets=(token.LSQB, token.RSQB),
+                )
+            ):
+                return False
+
             if not ensure_removable:
                 return True
+
             comma = self.leaves[-1]
-            return bool(comma.parent and comma.parent.type == syms.listmaker)
+            if comma.parent is None:
+                return False
+            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),
+                )
+            )
 
         if self.is_import:
             return True
 
-        if not is_one_tuple_between(closing.opening_bracket, closing, self.leaves):
+        if closing.opening_bracket is not None and not is_one_sequence_between(
+            closing.opening_bracket, closing, self.leaves
+        ):
             return True
 
         return False
@@ -400,6 +476,39 @@ 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.
+
+    This is introduced so that the EmptyLineTracker can look behind the standalone
+    comments and adjust their empty lines for class or def lines.
+    """
+
+    mode: Mode
+    previous_block: Optional["LinesBlock"]
+    original_line: Line
+    before: int = 0
+    content_lines: List[str] = field(default_factory=list)
+    after: int = 0
+
+    def all_lines(self) -> List[str]:
+        empty_line = str(Line(mode=self.mode))
+        return (
+            [empty_line * self.before] + self.content_lines + [empty_line * self.after]
+        )
+
+
 @dataclass
 class EmptyLineTracker:
     """Provides a stateful method that returns the number of potential extra
@@ -410,33 +519,56 @@ class EmptyLineTracker:
     are consumed by `maybe_empty_lines()` and included in the computation.
     """
 
-    is_pyi: bool = False
+    mode: Mode
     previous_line: Optional[Line] = None
-    previous_after: int = 0
-    previous_defs: List[int] = field(default_factory=list)
+    previous_block: Optional[LinesBlock] = None
+    previous_defs: List[Line] = field(default_factory=list)
+    semantic_leading_comment: Optional[LinesBlock] = None
 
-    def maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
+    def maybe_empty_lines(self, current_line: Line) -> LinesBlock:
         """Return the number of extra empty lines before and after the `current_line`.
 
         This is for separating `def`, `async def` and `class` with extra empty
         lines (two on module-level).
         """
         before, after = self._maybe_empty_lines(current_line)
+        previous_after = self.previous_block.after if self.previous_block else 0
         before = (
             # Black should not insert empty lines at the beginning
             # of the file
             0
             if self.previous_line is None
-            else before - self.previous_after
+            else before - previous_after
         )
-        self.previous_after = after
+        block = LinesBlock(
+            mode=self.mode,
+            previous_block=self.previous_block,
+            original_line=current_line,
+            before=before,
+            after=after,
+        )
+
+        # Maintain the semantic_leading_comment state.
+        if current_line.is_comment:
+            if self.previous_line is None or (
+                not self.previous_line.is_decorator
+                # `or before` means this comment already has an empty line before
+                and (not self.previous_line.is_comment or before)
+                and (self.semantic_leading_comment is None or before)
+            ):
+                self.semantic_leading_comment = block
+        # `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
-        return before, after
+        self.previous_block = block
+        return block
 
     def _maybe_empty_lines(self, current_line: Line) -> Tuple[int, int]:
         max_allowed = 1
         if current_line.depth == 0:
-            max_allowed = 1 if self.is_pyi else 2
+            max_allowed = 1 if self.mode.is_pyi else 2
         if current_line.leaves:
             # Consume the first leaf's extra newlines.
             first_leaf = current_line.leaves[0]
@@ -446,12 +578,45 @@ class EmptyLineTracker:
         else:
             before = 0
         depth = current_line.depth
-        while self.previous_defs and self.previous_defs[-1] >= depth:
-            self.previous_defs.pop()
-            if self.is_pyi:
-                before = 0 if depth else 1
+        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:
+                    before = 1
             else:
-                before = 1 if depth else 2
+                if depth:
+                    before = 1
+                elif (
+                    not depth
+                    and self.previous_defs[-1].depth
+                    and current_line.leaves[-1].type == token.COLON
+                    and (
+                        current_line.leaves[0].value
+                        not in ("with", "try", "for", "while", "if", "match")
+                    )
+                ):
+                    # We shouldn't add two newlines between an indented function and
+                    # a dependent non-indented clause. This is to avoid issues with
+                    # conditional function definitions that are technically top-level
+                    # and therefore get two trailing newlines, but look weird and
+                    # inconsistent when they're followed by elif, else, etc. This is
+                    # worse because these functions only get *one* preceding newline
+                    # already.
+                    before = 1
+                else:
+                    before = 2
+            self.previous_defs.pop()
         if current_line.is_decorator or current_line.is_def or current_line.is_class:
             return self._maybe_empty_lines_for_class_or_def(current_line, before)
 
@@ -459,6 +624,7 @@ 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
@@ -470,19 +636,21 @@ class EmptyLineTracker:
         ):
             return before, 1
 
+        if self.previous_line and self.previous_line.opens_block:
+            return 0, 0
         return before, 0
 
     def _maybe_empty_lines_for_class_or_def(
         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
 
         if self.previous_line.is_decorator:
-            if self.is_pyi and current_line.is_stub_class:
+            if self.mode.is_pyi and current_line.is_stub_class:
                 # Insert an empty line after a decorated stub class
                 return 0, 1
 
@@ -493,18 +661,31 @@ class EmptyLineTracker:
         ):
             return 0, 0
 
+        comment_to_add_newlines: Optional[LinesBlock] = None
         if (
             self.previous_line.is_comment
             and self.previous_line.depth == current_line.depth
             and before == 0
         ):
-            return 0, 0
+            slc = self.semantic_leading_comment
+            if (
+                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
+                and slc.before <= 1
+            ):
+                comment_to_add_newlines = slc
+            else:
+                return 0, 0
 
-        if self.is_pyi:
-            if self.previous_line.depth > current_line.depth:
-                newlines = 1
-            elif current_line.is_class or self.previous_line.is_class:
-                if current_line.is_stub_class and self.previous_line.is_stub_class:
+        if self.mode.is_pyi:
+            if current_line.is_class or self.previous_line.is_class:
+                if self.previous_line.depth < current_line.depth:
+                    newlines = 0
+                elif self.previous_line.depth > current_line.depth:
+                    newlines = 1
+                elif current_line.is_stub_class and self.previous_line.is_stub_class:
                     # No blank line between classes with an empty body
                     newlines = 0
                 else:
@@ -512,15 +693,27 @@ class EmptyLineTracker:
             elif (
                 current_line.is_def or current_line.is_decorator
             ) and not self.previous_line.is_def:
-                # Blank line between a block of functions (maybe with preceding
-                # decorators) and a block of non-functions
+                if current_line.depth:
+                    # In classes empty lines between attributes and methods should
+                    # be preserved.
+                    newlines = min(1, before)
+                else:
+                    # Blank line between a block of functions (maybe with preceding
+                    # decorators) and a block of non-functions
+                    newlines = 1
+            elif self.previous_line.depth > current_line.depth:
                 newlines = 1
             else:
                 newlines = 0
         else:
-            newlines = 2
-        if current_line.depth and newlines:
-            newlines -= 1
+            newlines = 1 if current_line.depth else 2
+        if comment_to_add_newlines is not None:
+            previous_block = comment_to_add_newlines.previous_block
+            if previous_block is not None:
+                comment_to_add_newlines.before = (
+                    max(comment_to_add_newlines.before, newlines) - previous_block.after
+                )
+                newlines = 0
         return newlines, 0
 
 
@@ -556,18 +749,95 @@ 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:
@@ -607,26 +877,42 @@ def can_be_split(line: Line) -> bool:
 
 
 def can_omit_invisible_parens(
-    line: Line,
+    rhs: RHSResult,
     line_length: int,
-    omit_on_explode: Collection[LeafID] = (),
 ) -> 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
@@ -647,12 +933,6 @@ def can_omit_invisible_parens(
 
     penultimate = line.leaves[-2]
     last = line.leaves[-1]
-    if line.magic_trailing_comma:
-        try:
-            penultimate, last = last_two_except(line.leaves, omit=omit_on_explode)
-        except LookupError:
-            # Turns out we'd omit everything.  We cannot skip the optional parentheses.
-            return False
 
     if (
         last.type == token.RPAR
@@ -674,10 +954,6 @@ def can_omit_invisible_parens(
             # unnecessary.
             return True
 
-        if line.magic_trailing_comma and penultimate.type == token.COMMA:
-            # The rightmost non-omitted bracket pair is the one we want to explode on.
-            return True
-
         if _can_omit_closing_paren(line, last=last, line_length=line_length):
             return True