mirror of
https://github.com/astral-sh/ruff.git
synced 2025-07-08 13:45:07 +00:00
76619b96e5
It turns out that `annotate-snippets` doesn't do a great job of consistently handling tabs. The intent of the implementation is clearly to expand tabs into 4 ASCII whitespace characters. But there are a few places where the column computation wasn't taking this expansion into account. In particular, the `unicode-width` crate returns `None` for a `\t` input, and `annotate-snippets` would in turn treat this as either zero columns or one column. Both are wrong. In patching this, it caused one of the existing `annotate-snippets` tests to fail. I spent a fair bit of time on it trying to fix it before coming to the conclusion that the test itself was wrong. In particular, the annotation ranges are 4 bytes off. However, when the range was wrong, the buggy code was rendering the example as intended since `\t` characters were treated as taking up zero columns of space. Now that they are correctly computed as taking up 4 columns of space, the offsets of the test needed to be adjusted. Fixes #670 |
||
|---|---|---|
| .. | ||
| examples | ||
| src | ||
| tests | ||
| Cargo.toml | ||
| LICENSE-APACHE | ||
| LICENSE-MIT | ||
| README.md | ||
This is a fork of the annotate-snippets crate. The principle motivation for
this fork, at the time of writing, is issue #167. Specifically, we wanted to
upgrade our version of annotate-snippets, but do so without changing our
diagnostic message format.
This copy of annotate-snippets is basically identical to upstream, but with
an extra Level::None variant that permits skipping over a new non-optional
header emitted by annotate-snippets.
More generally, it seems plausible that we may want to tweak other aspects of the output format in the future, so it might make sense to stick with our own copy so that we can be masters of our own destiny.