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.
3 from dataclasses import dataclass
4 from functools import lru_cache
5 from typing import Iterator, List, Optional, Union
7 if sys.version_info >= (3, 8):
8 from typing import Final
10 from typing_extensions import Final
12 from black.nodes import (
21 from blib2to3.pgen2 import token
22 from blib2to3.pytree import Leaf, Node
25 LN = Union[Leaf, Node]
27 FMT_OFF: Final = {"# fmt: off", "# fmt:off", "# yapf: disable"}
28 FMT_SKIP: Final = {"# fmt: skip", "# fmt:skip"}
29 FMT_PASS: Final = {*FMT_OFF, *FMT_SKIP}
30 FMT_ON: Final = {"# fmt: on", "# fmt:on", "# yapf: enable"}
32 COMMENT_EXCEPTIONS = " !:#'"
37 """Describes a piece of syntax that is a comment.
39 It's not a :class:`blib2to3.pytree.Leaf` so that:
41 * it can be cached (`Leaf` objects should not be reused more than once as
42 they store their lineno, column, prefix, and parent information);
43 * `newlines` and `consumed` fields are kept separate from the `value`. This
44 simplifies handling of special marker comments like ``# fmt: off/on``.
47 type: int # token.COMMENT or STANDALONE_COMMENT
48 value: str # content of the comment
49 newlines: int # how many newlines before the comment
50 consumed: int # how many characters of the original leaf's prefix did we consume
53 def generate_comments(leaf: LN) -> Iterator[Leaf]:
54 """Clean the prefix of the `leaf` and generate comments from it, if any.
56 Comments in lib2to3 are shoved into the whitespace prefix. This happens
57 in `pgen2/driver.py:Driver.parse_tokens()`. This was a brilliant implementation
58 move because it does away with modifying the grammar to include all the
59 possible places in which comments can be placed.
61 The sad consequence for us though is that comments don't "belong" anywhere.
62 This is why this function generates simple parentless Leaf objects for
63 comments. We simply don't know what the correct parent should be.
65 No matter though, we can live without this. We really only need to
66 differentiate between inline and standalone comments. The latter don't
67 share the line with any code.
69 Inline comments are emitted as regular token.COMMENT leaves. Standalone
70 are emitted with a fake STANDALONE_COMMENT token identifier.
72 for pc in list_comments(leaf.prefix, is_endmarker=leaf.type == token.ENDMARKER):
73 yield Leaf(pc.type, pc.value, prefix="\n" * pc.newlines)
76 @lru_cache(maxsize=4096)
77 def list_comments(prefix: str, *, is_endmarker: bool) -> List[ProtoComment]:
78 """Return a list of :class:`ProtoComment` objects parsed from the given `prefix`."""
79 result: List[ProtoComment] = []
80 if not prefix or "#" not in prefix:
86 for index, line in enumerate(re.split("\r?\n", prefix)):
87 consumed += len(line) + 1 # adding the length of the split '\n'
91 if not line.startswith("#"):
92 # Escaped newlines outside of a comment are not really newlines at
93 # all. We treat a single-line comment following an escaped newline
94 # as a simple trailing comment.
95 if line.endswith("\\"):
99 if index == ignored_lines and not is_endmarker:
100 comment_type = token.COMMENT # simple trailing comment
102 comment_type = STANDALONE_COMMENT
103 comment = make_comment(line)
106 type=comment_type, value=comment, newlines=nlines, consumed=consumed
113 def make_comment(content: str) -> str:
114 """Return a consistently formatted comment from the given `content` string.
116 All comments (except for "##", "#!", "#:", '#'") should have a single
117 space between the hash sign and the content.
119 If `content` didn't start with a hash sign, one is provided.
121 content = content.rstrip()
125 if content[0] == "#":
126 content = content[1:]
127 NON_BREAKING_SPACE = " "
130 and content[0] == NON_BREAKING_SPACE
131 and not content.lstrip().startswith("type:")
133 content = " " + content[1:] # Replace NBSP by a simple space
134 if content and content[0] not in COMMENT_EXCEPTIONS:
135 content = " " + content
139 def normalize_fmt_off(node: Node) -> None:
140 """Convert content between `# fmt: off`/`# fmt: on` into standalone comments."""
143 try_again = convert_one_fmt_off_pair(node)
146 def convert_one_fmt_off_pair(node: Node) -> bool:
147 """Convert content of a single `# fmt: off`/`# fmt: on` into a standalone comment.
149 Returns True if a pair was converted.
151 for leaf in node.leaves():
152 previous_consumed = 0
153 for comment in list_comments(leaf.prefix, is_endmarker=False):
154 if comment.value not in FMT_PASS:
155 previous_consumed = comment.consumed
157 # We only want standalone comments. If there's no previous leaf or
158 # the previous leaf is indentation, it's a standalone comment in
160 if comment.value in FMT_PASS and comment.type != STANDALONE_COMMENT:
161 prev = preceding_leaf(leaf)
163 if comment.value in FMT_OFF and prev.type not in WHITESPACE:
165 if comment.value in FMT_SKIP and prev.type in WHITESPACE:
168 ignored_nodes = list(generate_ignored_nodes(leaf, comment))
169 if not ignored_nodes:
172 first = ignored_nodes[0] # Can be a container node with the `leaf`.
173 parent = first.parent
174 prefix = first.prefix
175 if comment.value in FMT_OFF:
176 first.prefix = prefix[comment.consumed :]
177 if comment.value in FMT_SKIP:
179 standalone_comment_prefix = prefix
181 standalone_comment_prefix = (
182 prefix[:previous_consumed] + "\n" * comment.newlines
184 hidden_value = "".join(str(n) for n in ignored_nodes)
185 if comment.value in FMT_OFF:
186 hidden_value = comment.value + "\n" + hidden_value
187 if comment.value in FMT_SKIP:
188 hidden_value += " " + comment.value
189 if hidden_value.endswith("\n"):
190 # That happens when one of the `ignored_nodes` ended with a NEWLINE
191 # leaf (possibly followed by a DEDENT).
192 hidden_value = hidden_value[:-1]
193 first_idx: Optional[int] = None
194 for ignored in ignored_nodes:
195 index = ignored.remove()
196 if first_idx is None:
198 assert parent is not None, "INTERNAL ERROR: fmt: on/off handling (1)"
199 assert first_idx is not None, "INTERNAL ERROR: fmt: on/off handling (2)"
205 prefix=standalone_comment_prefix,
213 def generate_ignored_nodes(leaf: Leaf, comment: ProtoComment) -> Iterator[LN]:
214 """Starting from the container of `leaf`, generate all leaves until `# fmt: on`.
216 If comment is skip, returns leaf only.
217 Stops at the end of the block.
219 if comment.value in FMT_SKIP:
220 yield from _generate_ignored_nodes_from_fmt_skip(leaf, comment)
222 container: Optional[LN] = container_of(leaf)
223 while container is not None and container.type != token.ENDMARKER:
224 if is_fmt_on(container):
227 # fix for fmt: on in children
228 if children_contains_fmt_on(container):
229 for index, child in enumerate(container.children):
230 if isinstance(child, Leaf) and is_fmt_on(child):
231 if child.type in CLOSING_BRACKETS:
232 # This means `# fmt: on` is placed at a different bracket level
233 # than `# fmt: off`. This is an invalid use, but as a courtesy,
234 # we include this closing bracket in the ignored nodes.
235 # The alternative is to fail the formatting.
239 child.type == token.INDENT
240 and index < len(container.children) - 1
241 and children_contains_fmt_on(container.children[index + 1])
243 # This means `# fmt: on` is placed right after an indentation
244 # level, and we shouldn't swallow the previous INDENT token.
246 if children_contains_fmt_on(child):
250 if container.type == token.DEDENT and container.next_sibling is None:
251 # This can happen when there is no matching `# fmt: on` comment at the
252 # same level as `# fmt: on`. We need to keep this DEDENT.
255 container = container.next_sibling
258 def _generate_ignored_nodes_from_fmt_skip(
259 leaf: Leaf, comment: ProtoComment
261 """Generate all leaves that should be ignored by the `# fmt: skip` from `leaf`."""
262 prev_sibling = leaf.prev_sibling
264 # Need to properly format the leaf prefix to compare it to comment.value,
265 # which is also formatted
266 comments = list_comments(leaf.prefix, is_endmarker=False)
267 if not comments or comment.value != comments[0].value:
269 if prev_sibling is not None:
271 siblings = [prev_sibling]
272 while "\n" not in prev_sibling.prefix and prev_sibling.prev_sibling is not None:
273 prev_sibling = prev_sibling.prev_sibling
274 siblings.insert(0, prev_sibling)
277 parent is not None and parent.type == syms.suite and leaf.type == token.NEWLINE
279 # The `# fmt: skip` is on the colon line of the if/while/def/class/...
280 # statements. The ignored nodes should be previous siblings of the
283 ignored_nodes: List[LN] = []
284 parent_sibling = parent.prev_sibling
285 while parent_sibling is not None and parent_sibling.type != syms.suite:
286 ignored_nodes.insert(0, parent_sibling)
287 parent_sibling = parent_sibling.prev_sibling
288 # Special case for `async_stmt` where the ASYNC token is on the
290 grandparent = parent.parent
292 grandparent is not None
293 and grandparent.prev_sibling is not None
294 and grandparent.prev_sibling.type == token.ASYNC
296 ignored_nodes.insert(0, grandparent.prev_sibling)
297 yield from iter(ignored_nodes)
300 def is_fmt_on(container: LN) -> bool:
301 """Determine whether formatting is switched on within a container.
302 Determined by whether the last `# fmt:` comment is `on` or `off`.
305 for comment in list_comments(container.prefix, is_endmarker=False):
306 if comment.value in FMT_ON:
308 elif comment.value in FMT_OFF:
313 def children_contains_fmt_on(container: LN) -> bool:
314 """Determine if children have formatting switched on."""
315 for child in container.children:
316 leaf = first_leaf_of(child)
317 if leaf is not None and is_fmt_on(leaf):
323 def contains_pragma_comment(comment_list: List[Leaf]) -> bool:
326 True iff one of the comments in @comment_list is a pragma used by one
327 of the more common static analysis tools for python (e.g. mypy, flake8,
330 for comment in comment_list:
331 if comment.value.startswith(("# type:", "# noqa", "# pylint:")):