The `toml` crate doesn't support TOML 1.0, but `toml_edit` does. While
there is a plan to [migrate `toml` to be on
`toml_edit`](https://github.com/toml-rs/toml/issues/340), it's not ready
yet and it's very easy to switch back to `toml` when it's ready.
We populate this buffer ourselves, so I believe it's fine for us to use
an unchecked UTF-8 cast here. It _dramatically_ simplifies so much
downstream code.
This PR adds some guardrails to avoid common false positives in our
`pandas-vet` rules. Specifically, we now avoid triggering `pandas-vet`
rules if the target of the call or attribute (i.e., the `x` in
`x.stack(...)`) is unbound, or bound to something that couldn't be a
DataFrame (like an import that _isn't_ `pandas`, or a class definition).
This lets us avoid common false positives like `np.stack(...)`.
Resolves#1659.
Imagine a .py file containing the following comment:
# TODO: Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed
# do eiusmod tempor incididunt ut labore et dolore magna aliqua.
Since `git grep` only matches individual lines `git grep TODO` would
only output the first line of the comment, cutting off potentially
important information. (git grep currently doesn't support multiline
grepping). Projects using such a workflow therefore probably format
the comment in a single line instead:
# TODO: Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua.
This commit introduces a setting to accomdate this workflow by making
the line-length checks (`E501`) optionally ignore overlong lines
if they start with a recognized task tag.
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
Programmers often leave comments to themselves and others such as:
# TODO: Use a faster algorithm?
The keywords used to prefix such comments are just a convention and vary
from project to project. Other common keywords include FIXME and HACK.
The keywords in use for the codebase are of interest to ruff because
ruff does also lint comments. For example the ERA lint detects
commented-out code but ignores comments starting with such a keyword.
Previously the ERA lint simply hardcoded the regular expression
TODO|FIXME|XXX to achieve that. This commit introduces a new `task-tags`
setting to make this configurable (and to allow other comment lints to
recognize the same set of keywords).
The term "task tags" has probably been popularized by the Eclipse
IDE.[1] For Python there has been the proposal PEP 350[2], which
referred to such keywords as "codetags". That proposal however has been
rejected. We are choosing the term "task tags" over "code tags" because
the former is more descriptive: a task tag describes a task.
While according to the PEP 350 such keywords are also sometimes used for
non-tasks e.g. NOBUG to describe a well-known problem that will never be
addressed due to design problems or domain limitations, such keywords
are so rare that we are neglecting them here in favor of more
descriptive terminology. The vast majority of such keywords does
describe tasks, so naming the setting "task-tags" is apt.
[1]: https://www.eclipse.org/pdt/help/html/task_tags.htm
[2]: https://peps.python.org/pep-0350/
Co-authored-by: Charlie Marsh <charlie.r.marsh@gmail.com>
