from mypy_extensions import trait
-from black.brackets import BracketMatchError
from black.comments import contains_pragma_comment
from black.lines import Line, append_leaves
-from black.mode import Feature
+from black.mode import Feature, Mode
from black.nodes import (
CLOSING_BRACKETS,
OPENING_BRACKETS,
is_empty_lpar,
is_empty_par,
is_empty_rpar,
+ is_part_of_annotation,
parent_type,
replace_child,
syms,
from black.rusty import Err, Ok, Result
from black.strings import (
assert_is_leaf_string,
+ count_chars_in_width,
get_string_prefix,
has_triple_quotes,
normalize_string_quotes,
+ str_width,
)
from blib2to3.pgen2 import token
from blib2to3.pytree import Leaf, Node
# types
T = TypeVar("T")
LN = Union[Leaf, Node]
-Transformer = Callable[[Line, Collection[Feature]], Iterator[Line]]
+Transformer = Callable[[Line, Collection[Feature], Mode], Iterator[Line]]
Index = int
NodeType = int
ParserState = int
StringID = int
TResult = Result[T, CannotTransform] # (T)ransform Result
-TMatchResult = TResult[Index]
+TMatchResult = TResult[List[Index]]
+
+SPLIT_SAFE_CHARS = frozenset(["\u3001", "\u3002", "\uff0c"]) # East Asian stops
def TErr(err_msg: str) -> Err[CannotTransform]:
return Err(cant_transform)
-def hug_power_op(line: Line, features: Collection[Feature]) -> Iterator[Line]:
+def hug_power_op(
+ line: Line, features: Collection[Feature], mode: Mode
+) -> Iterator[Line]:
"""A transformer which normalizes spacing around power operators."""
# Performance optimization to avoid unnecessary Leaf clones and other ops.
def do_match(self, line: Line) -> TMatchResult:
"""
Returns:
- * Ok(string_idx) such that `line.leaves[string_idx]` is our target
- string, if a match was able to be made.
+ * Ok(string_indices) such that for each index, `line.leaves[index]`
+ is our target string if a match was able to be made. For
+ transformers that don't result in more lines (e.g. StringMerger,
+ StringParenStripper), multiple matches and transforms are done at
+ once to reduce the complexity.
OR
- * Err(CannotTransform), if a match was not able to be made.
+ * Err(CannotTransform), if no match could be made.
"""
@abstractmethod
- def do_transform(self, line: Line, string_idx: int) -> Iterator[TResult[Line]]:
+ def do_transform(
+ self, line: Line, string_indices: List[int]
+ ) -> Iterator[TResult[Line]]:
"""
Yields:
* Ok(new_line) where new_line is the new transformed line.
yield an CannotTransform after that point.)
"""
- def __call__(self, line: Line, _features: Collection[Feature]) -> Iterator[Line]:
+ def __call__(
+ self, line: Line, _features: Collection[Feature], _mode: Mode
+ ) -> Iterator[Line]:
"""
StringTransformer instances have a call signature that mirrors that of
the Transformer type.
" this line as one that it can transform."
) from cant_transform
- string_idx = match_result.ok()
+ string_indices = match_result.ok()
- for line_result in self.do_transform(line, string_idx):
+ for line_result in self.do_transform(line, string_indices):
if isinstance(line_result, Err):
cant_transform = line_result.err()
raise CannotTransform(
Requirements:
(A) The line contains adjacent strings such that ALL of the validation checks
- listed in StringMerger.__validate_msg(...)'s docstring pass.
+ listed in StringMerger._validate_msg(...)'s docstring pass.
OR
(B) The line contains a string which uses line continuation backslashes.
is_valid_index = is_valid_index_factory(LL)
- for i, leaf in enumerate(LL):
+ string_indices = []
+ idx = 0
+ while is_valid_index(idx):
+ leaf = LL[idx]
if (
leaf.type == token.STRING
- and is_valid_index(i + 1)
- and LL[i + 1].type == token.STRING
+ and is_valid_index(idx + 1)
+ and LL[idx + 1].type == token.STRING
):
- return Ok(i)
+ if not is_part_of_annotation(leaf):
+ string_indices.append(idx)
- if leaf.type == token.STRING and "\\\n" in leaf.value:
- return Ok(i)
+ # Advance to the next non-STRING leaf.
+ idx += 2
+ while is_valid_index(idx) and LL[idx].type == token.STRING:
+ idx += 1
- return TErr("This line has no strings that need merging.")
+ elif leaf.type == token.STRING and "\\\n" in leaf.value:
+ string_indices.append(idx)
+ # Advance to the next non-STRING leaf.
+ idx += 1
+ while is_valid_index(idx) and LL[idx].type == token.STRING:
+ idx += 1
- def do_transform(self, line: Line, string_idx: int) -> Iterator[TResult[Line]]:
+ else:
+ idx += 1
+
+ if string_indices:
+ return Ok(string_indices)
+ else:
+ return TErr("This line has no strings that need merging.")
+
+ def do_transform(
+ self, line: Line, string_indices: List[int]
+ ) -> Iterator[TResult[Line]]:
new_line = line
+
rblc_result = self._remove_backslash_line_continuation_chars(
- new_line, string_idx
+ new_line, string_indices
)
if isinstance(rblc_result, Ok):
new_line = rblc_result.ok()
- msg_result = self._merge_string_group(new_line, string_idx)
+ msg_result = self._merge_string_group(new_line, string_indices)
if isinstance(msg_result, Ok):
new_line = msg_result.ok()
@staticmethod
def _remove_backslash_line_continuation_chars(
- line: Line, string_idx: int
+ line: Line, string_indices: List[int]
) -> TResult[Line]:
"""
Merge strings that were split across multiple lines using
"""
LL = line.leaves
- string_leaf = LL[string_idx]
- if not (
- string_leaf.type == token.STRING
- and "\\\n" in string_leaf.value
- and not has_triple_quotes(string_leaf.value)
- ):
+ indices_to_transform = []
+ for string_idx in string_indices:
+ string_leaf = LL[string_idx]
+ if (
+ string_leaf.type == token.STRING
+ and "\\\n" in string_leaf.value
+ and not has_triple_quotes(string_leaf.value)
+ ):
+ indices_to_transform.append(string_idx)
+
+ if not indices_to_transform:
return TErr(
- f"String leaf {string_leaf} does not contain any backslash line"
- " continuation characters."
+ "Found no string leaves that contain backslash line continuation"
+ " characters."
)
new_line = line.clone()
new_line.comments = line.comments.copy()
append_leaves(new_line, line, LL)
- new_string_leaf = new_line.leaves[string_idx]
- new_string_leaf.value = new_string_leaf.value.replace("\\\n", "")
+ for string_idx in indices_to_transform:
+ new_string_leaf = new_line.leaves[string_idx]
+ new_string_leaf.value = new_string_leaf.value.replace("\\\n", "")
return Ok(new_line)
- def _merge_string_group(self, line: Line, string_idx: int) -> TResult[Line]:
+ def _merge_string_group(
+ self, line: Line, string_indices: List[int]
+ ) -> TResult[Line]:
"""
- Merges string group (i.e. set of adjacent strings) where the first
- string in the group is `line.leaves[string_idx]`.
+ Merges string groups (i.e. set of adjacent strings).
+
+ Each index from `string_indices` designates one string group's first
+ leaf in `line.leaves`.
Returns:
Ok(new_line), if ALL of the validation checks found in
- __validate_msg(...) pass.
+ _validate_msg(...) pass.
OR
Err(CannotTransform), otherwise.
"""
is_valid_index = is_valid_index_factory(LL)
- vresult = self._validate_msg(line, string_idx)
- if isinstance(vresult, Err):
- return vresult
+ # A dict of {string_idx: tuple[num_of_strings, string_leaf]}.
+ merged_string_idx_dict: Dict[int, Tuple[int, Leaf]] = {}
+ for string_idx in string_indices:
+ vresult = self._validate_msg(line, string_idx)
+ if isinstance(vresult, Err):
+ continue
+ merged_string_idx_dict[string_idx] = self._merge_one_string_group(
+ LL, string_idx, is_valid_index
+ )
+
+ if not merged_string_idx_dict:
+ return TErr("No string group is merged")
+ # Build the final line ('new_line') that this method will later return.
+ new_line = line.clone()
+ previous_merged_string_idx = -1
+ previous_merged_num_of_strings = -1
+ for i, leaf in enumerate(LL):
+ if i in merged_string_idx_dict:
+ previous_merged_string_idx = i
+ previous_merged_num_of_strings, string_leaf = merged_string_idx_dict[i]
+ new_line.append(string_leaf)
+
+ if (
+ previous_merged_string_idx
+ <= i
+ < previous_merged_string_idx + previous_merged_num_of_strings
+ ):
+ for comment_leaf in line.comments_after(LL[i]):
+ new_line.append(comment_leaf, preformatted=True)
+ continue
+
+ append_leaves(new_line, line, [leaf])
+
+ return Ok(new_line)
+
+ def _merge_one_string_group(
+ self, LL: List[Leaf], string_idx: int, is_valid_index: Callable[[int], bool]
+ ) -> Tuple[int, Leaf]:
+ """
+ Merges one string group where the first string in the group is
+ `LL[string_idx]`.
+
+ Returns:
+ A tuple of `(num_of_strings, leaf)` where `num_of_strings` is the
+ number of strings merged and `leaf` is the newly merged string
+ to be replaced in the new line.
+ """
# If the string group is wrapped inside an Atom node, we must make sure
# to later replace that Atom with our new (merged) string leaf.
atom_node = LL[string_idx].parent
characters have been escaped.
"""
assert_is_leaf_string(string)
+ if "f" in string_prefix:
+ string = _toggle_fexpr_quotes(string, QUOTE)
+ # After quotes toggling, quotes in expressions won't be escaped
+ # because quotes can't be reused in f-strings. So we can simply
+ # let the escaping logic below run without knowing f-string
+ # expressions.
RE_EVEN_BACKSLASHES = r"(?:(?<!\\)(?:\\\\)*)"
naked_string = string[len(string_prefix) + 1 : -1]
# Else replace the atom node with the new string leaf.
replace_child(atom_node, string_leaf)
- # Build the final line ('new_line') that this method will later return.
- new_line = line.clone()
- for i, leaf in enumerate(LL):
- if i == string_idx:
- new_line.append(string_leaf)
-
- if string_idx <= i < string_idx + num_of_strings:
- for comment_leaf in line.comments_after(LL[i]):
- new_line.append(comment_leaf, preformatted=True)
- continue
-
- append_leaves(new_line, line, [leaf])
-
self.add_custom_splits(string_leaf.value, custom_splits)
- return Ok(new_line)
+ return num_of_strings, string_leaf
@staticmethod
def _validate_msg(line: Line, string_idx: int) -> TResult[None]:
"""Validate (M)erge (S)tring (G)roup
- Transform-time string validation logic for __merge_string_group(...).
+ Transform-time string validation logic for _merge_string_group(...).
Returns:
* Ok(None), if ALL validation checks (listed below) pass.
- The set of all string prefixes in the string group is of
length greater than one and is not equal to {"", "f"}.
- The string group consists of raw strings.
+ - The string group is stringified type annotations. We don't want to
+ process stringified type annotations since pyright doesn't support
+ them spanning multiple string values. (NOTE: mypy, pytype, pyre do
+ support them, so we can change if pyright also gains support in the
+ future. See https://github.com/microsoft/pyright/issues/4359.)
"""
# We first check for "inner" stand-alone comments (i.e. stand-alone
# comments that have a string leaf before them AND after them).
is_valid_index = is_valid_index_factory(LL)
- for idx, leaf in enumerate(LL):
+ string_indices = []
+
+ idx = -1
+ while True:
+ idx += 1
+ if idx >= len(LL):
+ break
+ leaf = LL[idx]
+
# Should be a string...
if leaf.type != token.STRING:
continue
}:
continue
- return Ok(string_idx)
+ string_indices.append(string_idx)
+ idx = string_idx
+ while idx < len(LL) - 1 and LL[idx + 1].type == token.STRING:
+ idx += 1
+ if string_indices:
+ return Ok(string_indices)
return TErr("This line has no strings wrapped in parens.")
- def do_transform(self, line: Line, string_idx: int) -> Iterator[TResult[Line]]:
+ def do_transform(
+ self, line: Line, string_indices: List[int]
+ ) -> Iterator[TResult[Line]]:
LL = line.leaves
- string_parser = StringParser()
- rpar_idx = string_parser.parse(LL, string_idx)
+ string_and_rpar_indices: List[int] = []
+ for string_idx in string_indices:
+ string_parser = StringParser()
+ rpar_idx = string_parser.parse(LL, string_idx)
+
+ should_transform = True
+ for leaf in (LL[string_idx - 1], LL[rpar_idx]):
+ if line.comments_after(leaf):
+ # Should not strip parentheses which have comments attached
+ # to them.
+ should_transform = False
+ break
+ if should_transform:
+ string_and_rpar_indices.extend((string_idx, rpar_idx))
- for leaf in (LL[string_idx - 1], LL[rpar_idx]):
- if line.comments_after(leaf):
- yield TErr(
- "Will not strip parentheses which have comments attached to them."
- )
- return
+ if string_and_rpar_indices:
+ yield Ok(self._transform_to_new_line(line, string_and_rpar_indices))
+ else:
+ yield Err(
+ CannotTransform("All string groups have comments attached to them.")
+ )
+
+ def _transform_to_new_line(
+ self, line: Line, string_and_rpar_indices: List[int]
+ ) -> Line:
+ LL = line.leaves
new_line = line.clone()
new_line.comments = line.comments.copy()
- try:
- append_leaves(new_line, line, LL[: string_idx - 1])
- except BracketMatchError:
- # HACK: I believe there is currently a bug somewhere in
- # right_hand_split() that is causing brackets to not be tracked
- # properly by a shared BracketTracker.
- append_leaves(new_line, line, LL[: string_idx - 1], preformatted=True)
-
- string_leaf = Leaf(token.STRING, LL[string_idx].value)
- LL[string_idx - 1].remove()
- replace_child(LL[string_idx], string_leaf)
- new_line.append(string_leaf)
-
- append_leaves(
- new_line, line, LL[string_idx + 1 : rpar_idx] + LL[rpar_idx + 1 :]
- )
- LL[rpar_idx].remove()
+ previous_idx = -1
+ # We need to sort the indices, since string_idx and its matching
+ # rpar_idx may not come in order, e.g. in
+ # `("outer" % ("inner".join(items)))`, the "inner" string's
+ # string_idx is smaller than "outer" string's rpar_idx.
+ for idx in sorted(string_and_rpar_indices):
+ leaf = LL[idx]
+ lpar_or_rpar_idx = idx - 1 if leaf.type == token.STRING else idx
+ append_leaves(new_line, line, LL[previous_idx + 1 : lpar_or_rpar_idx])
+ if leaf.type == token.STRING:
+ string_leaf = Leaf(token.STRING, LL[idx].value)
+ LL[lpar_or_rpar_idx].remove() # Remove lpar.
+ replace_child(LL[idx], string_leaf)
+ new_line.append(string_leaf)
+ else:
+ LL[lpar_or_rpar_idx].remove() # This is a rpar.
+
+ previous_idx = idx
- yield Ok(new_line)
+ # Append the leaves after the last idx:
+ append_leaves(new_line, line, LL[idx + 1 :])
+
+ return new_line
class BaseStringSplitter(StringTransformer):
if isinstance(match_result, Err):
return match_result
- string_idx = match_result.ok()
+ string_indices = match_result.ok()
+ assert len(string_indices) == 1, (
+ f"{self.__class__.__name__} should only find one match at a time, found"
+ f" {len(string_indices)}"
+ )
+ string_idx = string_indices[0]
vresult = self._validate(line, string_idx)
if isinstance(vresult, Err):
return vresult
# WMA4 the length of the inline comment.
offset += len(comment_leaf.value)
- max_string_length = self.line_length - offset
+ max_string_length = count_chars_in_width(str(line), self.line_length - offset)
return max_string_length
@staticmethod
if LL[0].type != token.STRING:
return None
- # If the string is surrounded by commas (or is the first/last child)...
- prev_sibling = LL[0].prev_sibling
- next_sibling = LL[0].next_sibling
- if not prev_sibling and not next_sibling and parent_type(LL[0]) == syms.atom:
- # If it's an atom string, we need to check the parent atom's siblings.
- parent = LL[0].parent
- assert parent is not None # For type checkers.
- prev_sibling = parent.prev_sibling
- next_sibling = parent.next_sibling
- if (not prev_sibling or prev_sibling.type == token.COMMA) and (
- not next_sibling or next_sibling.type == token.COMMA
+ matching_nodes = [
+ syms.listmaker,
+ syms.dictsetmaker,
+ syms.testlist_gexp,
+ ]
+ # If the string is an immediate child of a list/set/tuple literal...
+ if (
+ parent_type(LL[0]) in matching_nodes
+ or parent_type(LL[0].parent) in matching_nodes
):
- return 0
+ # And the string is surrounded by commas (or is the first/last child)...
+ prev_sibling = LL[0].prev_sibling
+ next_sibling = LL[0].next_sibling
+ if (
+ not prev_sibling
+ and not next_sibling
+ and parent_type(LL[0]) == syms.atom
+ ):
+ # If it's an atom string, we need to check the parent atom's siblings.
+ parent = LL[0].parent
+ assert parent is not None # For type checkers.
+ prev_sibling = parent.prev_sibling
+ next_sibling = parent.next_sibling
+ if (not prev_sibling or prev_sibling.type == token.COMMA) and (
+ not next_sibling or next_sibling.type == token.COMMA
+ ):
+ return 0
return None
return any(iter_fexpr_spans(s))
+def _toggle_fexpr_quotes(fstring: str, old_quote: str) -> str:
+ """
+ Toggles quotes used in f-string expressions that are `old_quote`.
+
+ f-string expressions can't contain backslashes, so we need to toggle the
+ quotes if the f-string itself will end up using the same quote. We can
+ simply toggle without escaping because, quotes can't be reused in f-string
+ expressions. They will fail to parse.
+
+ NOTE: If PEP 701 is accepted, above statement will no longer be true.
+ Though if quotes can be reused, we can simply reuse them without updates or
+ escaping, once Black figures out how to parse the new grammar.
+ """
+ new_quote = "'" if old_quote == '"' else '"'
+ parts = []
+ previous_index = 0
+ for start, end in iter_fexpr_spans(fstring):
+ parts.append(fstring[previous_index:start])
+ parts.append(fstring[start:end].replace(old_quote, new_quote))
+ previous_index = end
+ parts.append(fstring[previous_index:])
+ return "".join(parts)
+
+
class StringSplitter(BaseStringSplitter, CustomSplitMapMixin):
"""
StringTransformer that splits "atom" strings (i.e. strings which exist on
if is_valid_index(idx):
return TErr("This line does not end with a string.")
- return Ok(string_idx)
+ return Ok([string_idx])
- def do_transform(self, line: Line, string_idx: int) -> Iterator[TResult[Line]]:
+ def do_transform(
+ self, line: Line, string_indices: List[int]
+ ) -> Iterator[TResult[Line]]:
LL = line.leaves
+ assert len(string_indices) == 1, (
+ f"{self.__class__.__name__} should only find one match at a time, found"
+ f" {len(string_indices)}"
+ )
+ string_idx = string_indices[0]
QUOTE = LL[string_idx].value[-1]
is_valid_index(string_idx + 1) and LL[string_idx + 1].type == token.COMMA
)
- def max_last_string() -> int:
+ def max_last_string_column() -> int:
"""
Returns:
- The max allowed length of the string value used for the last
- line we will construct.
+ The max allowed width of the string value used for the last
+ line we will construct. Note that this value means the width
+ rather than the number of characters (e.g., many East Asian
+ characters expand to two columns).
"""
result = self.line_length
result -= line.depth * 4
result -= string_op_leaves_length
return result
- # --- Calculate Max Break Index (for string value)
+ # --- Calculate Max Break Width (for string value)
# We start with the line length limit
- max_break_idx = self.line_length
+ max_break_width = self.line_length
# The last index of a string of length N is N-1.
- max_break_idx -= 1
+ max_break_width -= 1
# Leading whitespace is not present in the string value (e.g. Leaf.value).
- max_break_idx -= line.depth * 4
- if max_break_idx < 0:
+ max_break_width -= line.depth * 4
+ if max_break_width < 0:
yield TErr(
f"Unable to split {LL[string_idx].value} at such high of a line depth:"
f" {line.depth}"
# line limit.
use_custom_breakpoints = bool(
custom_splits
- and all(csplit.break_idx <= max_break_idx for csplit in custom_splits)
+ and all(csplit.break_idx <= max_break_width for csplit in custom_splits)
)
# Temporary storage for the remaining chunk of the string line that
if use_custom_breakpoints:
return len(custom_splits) > 1
else:
- return len(rest_value) > max_last_string()
+ return str_width(rest_value) > max_last_string_column()
string_line_results: List[Ok[Line]] = []
while more_splits_should_be_made():
break_idx = csplit.break_idx
else:
# Algorithmic Split (automatic)
- max_bidx = max_break_idx - string_op_leaves_length
+ max_bidx = (
+ count_chars_in_width(rest_value, max_break_width)
+ - string_op_leaves_length
+ )
maybe_break_idx = self._get_break_idx(rest_value, max_bidx)
if maybe_break_idx is None:
# If we are unable to algorithmically determine a good split
# prefix, and the current custom split did NOT originally use a
# prefix...
if (
- next_value != self._normalize_f_string(next_value, prefix)
- and use_custom_breakpoints
+ use_custom_breakpoints
and not csplit.has_prefix
+ and (
+ # `next_value == prefix + QUOTE` happens when the custom
+ # split is an empty string.
+ next_value == prefix + QUOTE
+ or next_value != self._normalize_f_string(next_value, prefix)
+ )
):
# Then `csplit.break_idx` will be off by one after removing
# the 'f' prefix.
# Try to fit them all on the same line with the last substring...
if (
- len(temp_value) <= max_last_string()
+ str_width(temp_value) <= max_last_string_column()
or LL[string_idx + 1].type == token.COMMA
):
last_line.append(rest_leaf)
section of this classes' docstring would be be met by returning @i.
"""
is_space = string[i] == " "
+ is_split_safe = is_valid_index(i - 1) and string[i - 1] in SPLIT_SAFE_CHARS
is_not_escaped = True
j = i - 1
and len(string[:i]) >= self.MIN_SUBSTR_SIZE
)
return (
- is_space
+ (is_space or is_split_safe)
and is_not_escaped
and is_big_enough
and not breaks_unsplittable_expression(i)
* The line is an lambda expression and the value is a string.
OR
* The line starts with an "atom" string that prefers to be wrapped in
- parens. It's preferred to be wrapped when the string is surrounded by
- commas (or is the first/last child).
+ parens. It's preferred to be wrapped when it's is an immediate child of
+ a list/set/tuple literal, AND the string is surrounded by commas (or is
+ the first/last child).
Transformations:
The chosen string is wrapped in parentheses and then split at the LPAR.
if string_idx is not None:
string_value = line.leaves[string_idx].value
- # If the string has no spaces...
- if " " not in string_value:
+ # If the string has neither spaces nor East Asian stops...
+ if not any(
+ char == " " or char in SPLIT_SAFE_CHARS for char in string_value
+ ):
# And will still violate the line length limit when split...
- max_string_length = self.line_length - ((line.depth + 1) * 4)
- if len(string_value) > max_string_length:
+ max_string_width = self.line_length - ((line.depth + 1) * 4)
+ if str_width(string_value) > max_string_width:
# And has no associated custom splits...
if not self.has_custom_splits(string_value):
# Then we should NOT put this string on its own line.
" resultant line would still be over the specified line"
" length and can't be split further by StringSplitter."
)
- return Ok(string_idx)
+ return Ok([string_idx])
return TErr("This line does not contain any non-atomic strings.")
for i, leaf in enumerate(LL):
# We MUST find a colon, it can either be dict's or lambda's colon...
- if leaf.type == token.COLON:
+ if leaf.type == token.COLON and i < len(LL) - 1:
idx = i + 2 if is_empty_par(LL[i + 1]) else i + 1
# That colon MUST be followed by a string...
return None
- def do_transform(self, line: Line, string_idx: int) -> Iterator[TResult[Line]]:
+ def do_transform(
+ self, line: Line, string_indices: List[int]
+ ) -> Iterator[TResult[Line]]:
LL = line.leaves
+ assert len(string_indices) == 1, (
+ f"{self.__class__.__name__} should only find one match at a time, found"
+ f" {len(string_indices)}"
+ )
+ string_idx = string_indices[0]
is_valid_index = is_valid_index_factory(LL)
insert_str_child = insert_str_child_factory(LL[string_idx])
projects / etc / vim.git / blobdiff