"""
String transformers that can split and merge strings.
"""
+import re
+import sys
from abc import ABC, abstractmethod
from collections import defaultdict
from dataclasses import dataclass
-import regex as re # We need recursive patterns here (?R)
from typing import (
Any,
Callable,
TypeVar,
Union,
)
-import sys
if sys.version_info < (3, 8):
- from typing_extensions import Final
+ from typing_extensions import Final, Literal
else:
- from typing import Final
+ from typing import Literal, Final
from mypy_extensions import trait
-from black.rusty import Result, Ok, Err
-
-from black.mode import Feature
-from black.nodes import syms, replace_child, parent_type
-from black.nodes import is_empty_par, is_empty_lpar, is_empty_rpar
-from black.nodes import OPENING_BRACKETS, CLOSING_BRACKETS, STANDALONE_COMMENT
-from black.lines import Line, append_leaves
-from black.brackets import BracketMatchError
from black.comments import contains_pragma_comment
-from black.strings import has_triple_quotes, get_string_prefix, assert_is_leaf_string
-from black.strings import normalize_string_quotes
-
-from blib2to3.pytree import Leaf, Node
+from black.lines import Line, append_leaves
+from black.mode import Feature
+from black.nodes import (
+ CLOSING_BRACKETS,
+ OPENING_BRACKETS,
+ STANDALONE_COMMENT,
+ 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,
+ get_string_prefix,
+ has_triple_quotes,
+ normalize_string_quotes,
+)
from blib2to3.pgen2 import token
+from blib2to3.pytree import Leaf, Node
class CannotTransform(Exception):
return Err(cant_transform)
+def hug_power_op(line: Line, features: Collection[Feature]) -> Iterator[Line]:
+ """A transformer which normalizes spacing around power operators."""
+
+ # Performance optimization to avoid unnecessary Leaf clones and other ops.
+ for leaf in line.leaves:
+ if leaf.type == token.DOUBLESTAR:
+ break
+ else:
+ raise CannotTransform("No doublestar token was found in the line.")
+
+ def is_simple_lookup(index: int, step: Literal[1, -1]) -> bool:
+ # Brackets and parentheses indicate calls, subscripts, etc. ...
+ # basically stuff that doesn't count as "simple". Only a NAME lookup
+ # or dotted lookup (eg. NAME.NAME) is OK.
+ if step == -1:
+ disallowed = {token.RPAR, token.RSQB}
+ else:
+ disallowed = {token.LPAR, token.LSQB}
+
+ while 0 <= index < len(line.leaves):
+ current = line.leaves[index]
+ if current.type in disallowed:
+ return False
+ if current.type not in {token.NAME, token.DOT} or current.value == "for":
+ # If the current token isn't disallowed, we'll assume this is simple as
+ # only the disallowed tokens are semantically attached to this lookup
+ # expression we're checking. Also, stop early if we hit the 'for' bit
+ # of a comprehension.
+ return True
+
+ index += step
+
+ return True
+
+ def is_simple_operand(index: int, kind: Literal["base", "exponent"]) -> bool:
+ # An operand is considered "simple" if's a NAME, a numeric CONSTANT, a simple
+ # lookup (see above), with or without a preceding unary operator.
+ start = line.leaves[index]
+ if start.type in {token.NAME, token.NUMBER}:
+ return is_simple_lookup(index, step=(1 if kind == "exponent" else -1))
+
+ if start.type in {token.PLUS, token.MINUS, token.TILDE}:
+ if line.leaves[index + 1].type in {token.NAME, token.NUMBER}:
+ # step is always one as bases with a preceding unary op will be checked
+ # for simplicity starting from the next token (so it'll hit the check
+ # above).
+ return is_simple_lookup(index + 1, step=1)
+
+ return False
+
+ new_line = line.clone()
+ should_hug = False
+ for idx, leaf in enumerate(line.leaves):
+ new_leaf = leaf.clone()
+ if should_hug:
+ new_leaf.prefix = ""
+ should_hug = False
+
+ should_hug = (
+ (0 < idx < len(line.leaves) - 1)
+ and leaf.type == token.DOUBLESTAR
+ and is_simple_operand(idx - 1, kind="base")
+ and line.leaves[idx - 1].value != "lambda"
+ and is_simple_operand(idx + 1, kind="exponent")
+ )
+ if should_hug:
+ new_leaf.prefix = ""
+
+ # We have to be careful to make a new line properly:
+ # - bracket related metadata must be maintained (handled by Line.append)
+ # - comments need to copied over, updating the leaf IDs they're attached to
+ new_line.append(new_leaf, preformatted=True)
+ for comment_leaf in line.comments_after(leaf):
+ new_line.append(comment_leaf, preformatted=True)
+
+ yield new_line
+
+
class StringTransformer(ABC):
"""
An implementation of the Transformer protocol that relies on its
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):
+ for i, leaf in enumerate(LL):
if (
leaf.type == token.STRING
and is_valid_index(i + 1)
and LL[i + 1].type == token.STRING
):
+ if is_part_of_annotation(leaf):
+ return TErr("String is part of type annotation.")
return Ok(i)
if leaf.type == token.STRING and "\\\n" in leaf.value:
Returns:
Ok(new_line), if ALL of the validation checks found in
- __validate_msg(...) pass.
+ _validate_msg(...) pass.
OR
Err(CannotTransform), otherwise.
"""
# with 'f'...
if "f" in prefix and "f" not in next_prefix:
# Then we must escape any braces contained in this substring.
- SS = re.subf(r"(\{|\})", "{1}{1}", SS)
+ SS = re.sub(r"(\{|\})", r"\1\1", SS)
NSS = make_naked(SS, next_prefix)
next_str_idx += 1
+ # Take a note on the index of the non-STRING leaf.
+ non_string_idx = next_str_idx
+
S_leaf = Leaf(token.STRING, S)
if self.normalize_strings:
S_leaf.value = normalize_string_quotes(S_leaf.value)
string_leaf = Leaf(token.STRING, S_leaf.value.replace(BREAK_MARK, ""))
if atom_node is not None:
- replace_child(atom_node, string_leaf)
+ # If not all children of the atom node are merged (this can happen
+ # when there is a standalone comment in the middle) ...
+ if non_string_idx - string_idx < len(atom_node.children):
+ # We need to replace the old STRING leaves with the new string leaf.
+ first_child_idx = LL[string_idx].remove()
+ for idx in range(string_idx + 1, non_string_idx):
+ LL[idx].remove()
+ if first_child_idx is not None:
+ atom_node.insert_child(first_child_idx, string_leaf)
+ else:
+ # 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):
+ for i, leaf in enumerate(LL):
if i == string_idx:
new_line.append(string_leaf)
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):
+ for idx, leaf in enumerate(LL):
# Should be a string...
if leaf.type != token.STRING:
continue
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)
+ append_leaves(new_line, line, LL[: string_idx - 1])
string_leaf = Leaf(token.STRING, LL[string_idx].value)
LL[string_idx - 1].remove()
max_string_length = self.line_length - offset
return max_string_length
+ @staticmethod
+ def _prefer_paren_wrap_match(LL: List[Leaf]) -> Optional[int]:
+ """
+ Returns:
+ string_idx such that @LL[string_idx] is equal to our target (i.e.
+ matched) string, if this line matches the "prefer paren wrap" statement
+ requirements listed in the 'Requirements' section of the StringParenWrapper
+ class's docstring.
+ OR
+ None, otherwise.
+ """
+ # The line must start with a string.
+ 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
+ ):
+ return 0
+
+ return None
+
+
+def iter_fexpr_spans(s: str) -> Iterator[Tuple[int, int]]:
+ """
+ Yields spans corresponding to expressions in a given f-string.
+ Spans are half-open ranges (left inclusive, right exclusive).
+ Assumes the input string is a valid f-string, but will not crash if the input
+ string is invalid.
+ """
+ stack: List[int] = [] # our curly paren stack
+ i = 0
+ while i < len(s):
+ if s[i] == "{":
+ # if we're in a string part of the f-string, ignore escaped curly braces
+ if not stack and i + 1 < len(s) and s[i + 1] == "{":
+ i += 2
+ continue
+ stack.append(i)
+ i += 1
+ continue
+
+ if s[i] == "}":
+ if not stack:
+ i += 1
+ continue
+ j = stack.pop()
+ # we've made it back out of the expression! yield the span
+ if not stack:
+ yield (j, i + 1)
+ i += 1
+ continue
+
+ # if we're in an expression part of the f-string, fast forward through strings
+ # note that backslashes are not legal in the expression portion of f-strings
+ if stack:
+ delim = None
+ if s[i : i + 3] in ("'''", '"""'):
+ delim = s[i : i + 3]
+ elif s[i] in ("'", '"'):
+ delim = s[i]
+ if delim:
+ i += len(delim)
+ while i < len(s) and s[i : i + len(delim)] != delim:
+ i += 1
+ i += len(delim)
+ continue
+ i += 1
+
+
+def fstring_contains_expr(s: str) -> bool:
+ return any(iter_fexpr_spans(s))
+
class StringSplitter(BaseStringSplitter, CustomSplitMapMixin):
"""
"""
MIN_SUBSTR_SIZE: Final = 6
- # Matches an "f-expression" (e.g. {var}) that might be found in an f-string.
- RE_FEXPR: Final = r"""
- (?<!\{) (?:\{\{)* \{ (?!\{)
- (?:
- [^\{\}]
- | \{\{
- | \}\}
- | (?R)
- )+
- \}
- """
def do_splitter_match(self, line: Line) -> TMatchResult:
LL = line.leaves
+ if self._prefer_paren_wrap_match(LL) is not None:
+ return TErr("Line needs to be wrapped in parens first.")
+
is_valid_index = is_valid_index_factory(LL)
idx = 0
# contain any f-expressions, but ONLY if the original f-string
# contains at least one f-expression. Otherwise, we will alter the AST
# of the program.
- drop_pointless_f_prefix = ("f" in prefix) and re.search(
- self.RE_FEXPR, LL[string_idx].value, re.VERBOSE
+ drop_pointless_f_prefix = ("f" in prefix) and fstring_contains_expr(
+ LL[string_idx].value
)
first_string_line = True
string_op_leaves = self._get_string_operator_leaves(LL)
string_op_leaves_length = (
- sum([len(str(prefix_leaf)) for prefix_leaf in string_op_leaves]) + 1
+ sum(len(str(prefix_leaf)) for prefix_leaf in string_op_leaves) + 1
if string_op_leaves
else 0
)
# 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.
"""
if "f" not in get_string_prefix(string).lower():
return
-
- for match in re.finditer(self.RE_FEXPR, string, re.VERBOSE):
- yield match.span()
+ yield from iter_fexpr_spans(string)
def _get_illegal_split_indices(self, string: str) -> Set[Index]:
illegal_indices: Set[Index] = set()
"""
assert_is_leaf_string(string)
- if "f" in prefix and not re.search(self.RE_FEXPR, string, re.VERBOSE):
+ if "f" in prefix and not fstring_contains_expr(string):
new_prefix = prefix.replace("f", "")
temp = string[len(prefix) :]
class StringParenWrapper(BaseStringSplitter, CustomSplitMapMixin):
"""
- StringTransformer that splits non-"atom" strings (i.e. strings that do not
- exist on lines by themselves).
+ StringTransformer that wraps strings in parens and then splits at the LPAR.
Requirements:
All of the requirements listed in BaseStringSplitter's docstring in
OR
* The line is a dictionary key assignment where some valid key is being
assigned the value of some string.
+ OR
+ * 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).
Transformations:
The chosen string is wrapped in parentheses and then split at the LPAR.
changed such that it no longer needs to be given its own line,
StringParenWrapper relies on StringParenStripper to clean up the
parentheses it created.
+
+ For "atom" strings that prefers to be wrapped in parens, it requires
+ StringSplitter to hold the split until the string is wrapped in parens.
"""
def do_splitter_match(self, line: Line) -> TMatchResult:
or self._else_match(LL)
or self._assert_match(LL)
or self._assign_match(LL)
- or self._dict_match(LL)
+ or self._dict_or_lambda_match(LL)
+ or self._prefer_paren_wrap_match(LL)
)
if string_idx is not None:
if parent_type(LL[0]) == syms.assert_stmt and LL[0].value == "assert":
is_valid_index = is_valid_index_factory(LL)
- for (i, leaf) in enumerate(LL):
+ for i, leaf in enumerate(LL):
# We MUST find a comma...
if leaf.type == token.COMMA:
idx = i + 2 if is_empty_par(LL[i + 1]) else i + 1
):
is_valid_index = is_valid_index_factory(LL)
- for (i, leaf) in enumerate(LL):
+ for i, leaf in enumerate(LL):
# We MUST find either an '=' or '+=' symbol...
if leaf.type in [token.EQUAL, token.PLUSEQUAL]:
idx = i + 2 if is_empty_par(LL[i + 1]) else i + 1
return None
@staticmethod
- def _dict_match(LL: List[Leaf]) -> Optional[int]:
+ def _dict_or_lambda_match(LL: List[Leaf]) -> Optional[int]:
"""
Returns:
string_idx such that @LL[string_idx] is equal to our target (i.e.
matched) string, if this line matches the dictionary key assignment
- statement requirements listed in the 'Requirements' section of this
- classes' docstring.
+ statement or lambda expression requirements listed in the
+ 'Requirements' section of this classes' docstring.
OR
None, otherwise.
"""
- # If this line is apart of a dictionary key assignment...
- if syms.dictsetmaker in [parent_type(LL[0]), parent_type(LL[0].parent)]:
+ # If this line is a part of a dictionary key assignment or lambda expression...
+ parent_types = [parent_type(LL[0]), parent_type(LL[0].parent)]
+ if syms.dictsetmaker in parent_types or syms.lambdef in parent_types:
is_valid_index = is_valid_index_factory(LL)
- for (i, leaf) in enumerate(LL):
- # We MUST find a colon...
- if leaf.type == token.COLON:
+ 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 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...
f" (left_leaves={left_leaves}, right_leaves={right_leaves})"
)
old_rpar_leaf = right_leaves.pop()
+ elif right_leaves and right_leaves[-1].type == token.RPAR:
+ # Special case for lambda expressions as dict's value, e.g.:
+ # my_dict = {
+ # "key": lambda x: f"formatted: {x},
+ # }
+ # After wrapping the dict's value with parentheses, the string is
+ # followed by a RPAR but its opening bracket is lambda's, not
+ # the string's:
+ # "key": (lambda x: f"formatted: {x}),
+ opening_bracket = right_leaves[-1].opening_bracket
+ if opening_bracket is not None and opening_bracket in left_leaves:
+ index = left_leaves.index(opening_bracket)
+ if (
+ index > 0
+ and index < len(left_leaves) - 1
+ and left_leaves[index - 1].type == token.COLON
+ and left_leaves[index + 1].value == "lambda"
+ ):
+ right_leaves.pop()
append_leaves(string_line, line, right_leaves)
projects / etc / vim.git / blobdiff