Move diff rendering to ruff_db

[ntBre]

Summary

This is a preparatory PR in support of #19919. It moves our Diff rendering code from ruff_linter to ruff_db, where we have direct access to the DiagnosticStylesheet used by our other diagnostic rendering code. As shown by the tests, this shouldn't cause any visible changes. The colors aren't exactly the same, as I note in a TODO comment, but I don't think there's any existing way to see those, even in tests.

The Diff implementation is mostly unchanged. I just switched from a Ruff-specific SourceFile to a DiagnosticSource (removing an expect_ruff_source_file call) and updated the LineStyle struct and other styling calls to use fmt_styled and our existing stylesheet.

In support of these changes, I added three styles to our stylesheet: insertion and deletion for the corresponding diff operations, and underline, which apparently we can use, as I hoped on Discord. This isn't supported in all terminals, though. It worked in ghostty but not in st for me.

I moved the calculate_print_width function from the now-deleted diff.rs to a method on OneIndexed, where it was available everywhere we needed it. I'm not sure if that's desirable, or if my other changes to the function are either (using ilog10 instead of a loop). This does make it const and slightly simplifies things in my opinion, but I'm happy to revert it if preferred.

I also inlined a version of show_nonprinting from the ShowNonprinting trait in ruff_linter:

ruff/crates/ruff_linter/src/text_helpers.rs

Lines 3 to 5 in f4be05a

	pub(crate) trait ShowNonprinting {
	fn show_nonprinting(&self) -> Cow<'_, str>;
	}

This trait is now only used in source_kind.rs, so I'm not sure it's worth having the trait or the macro-generated implementation (which is only called once). This is obviously closely related to our unprintable character handling in diagnostic rendering, but the usage seems different enough not to try to combine them.

ruff/crates/ruff_db/src/diagnostic/render.rs

Lines 990 to 998 in f4be05a

	let unprintable_replacement = |c: char| -> Option<char> {
	match c {
	'\x07' => Some('␇'),
	'\x08' => Some('␈'),
	'\x1b' => Some('␛'),
	'\x7f' => Some('␡'),
	_ => None,
	}
	};

We could also move the trait to another crate where we can use it in ruff_db instead of inlining here, of course.

Finally, this PR makes TextEmitter a very thin wrapper around a DisplayDiagnosticsConfig. It's still used in a few places, though, unlike the other emitters we've replaced, so I figured it was worth keeping around. It's a pretty nice API for setting all of the options on the config and then passing that along to a DisplayDiagnostics.

Test Plan

Existing snapshot tests with diffs

[github-actions]

Diagnostic diff on typing conformance tests

No changes detected when running ty on typing conformance tests ✅

[github-actions]

mypy_primer results

No ecosystem changes detected ✅
No memory usage changes detected ✅

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

[MichaReiser]

This is a bit tough to review because the moves the code and changes the code are in the same commit but I don't think it's worth the work to split the commits now.

[MichaReiser]

I don't think I understand what this method does :sweat. I'd find it helpful if it had a few doctests to demonstrate what it does.

The name width is also somewhat misleading because it suggests to me that it computes the unicode width, which it doesn't (but, as I said, I don't understand what it does).

If I understand this correctly, this computes the number of digits? Maybe digits or digits_len would be better name.

[ntBre]

This sounds like a good reason to put it back to the old version:

ruff/crates/ruff_linter/src/message/diff.rs

Lines 190 to 202 in 1a38831

	/// Calculate the length of the string representation of `value`
	pub(super) fn calculate_print_width(mut value: OneIndexed) -> NonZeroUsize {
	const TEN: OneIndexed = OneIndexed::from_zero_indexed(9);
	let mut width = OneIndexed::ONE;
	while value >= TEN {
	value = OneIndexed::new(value.get() / 10).unwrap_or(OneIndexed::MIN);
	width = width.checked_add(1).unwrap();
	}
	width
	}

But yes, it just computes the number of digits. I can add doctests and rename it either way.

Sorry about not splitting the commits!

[MichaReiser]

No worries. I think it's fine on OneIndexed. It's just the name that I found confusing. I think digits would be easier to understand.

Could we use this also in

ruff/crates/ty_server/src/server/api/requests/completion.rs

Line 60 in 7dfde3b

let max_index_len = completions.len().saturating_sub(1).to_string().len();

[ntBre]

I meant the old implementation in the new method, but either works for me. I renamed and expanded the docs.

It looks like we can use it there in ty too! Is it okay to unwrap since we know completions is non-empty?

        let max_index_len = OneIndexed::new(completions.len()).unwrap().digits().get();

Otherwise we can use OneIndexed::from_zero_indexed, but that adds 1 after saturating_sub subtracted one, so it seems a little silly. We can also unwrap_or_default.