Allow users to rebind expect!
#26
Conversation
|
This doesn't test that |
|
👍 to general idea of allowing renaming Yeah, I like this approach more! I am a bit surprised that we look for first Actually, what unit the column is in? This seems to not be documented. |
src/lib.rs
Outdated
| //! | ||
| //! ``` | ||
| //! use expect_test::expect; | ||
| //! use expect_test::expect as ex; // Note the renaming |
There was a problem hiding this comment.
Let's not add this to the docs: this is a niche thing you shouldn't generally be doing. Documenting this helps for few people that do need this, but harms everyone else. Docs are like code -- the fewer lines you spend, the better!
I filed rust-lang/rust#92301 while implementing this. I believe it's code points.
Part of the complexity is because I split this out from #27. You could do |
|
bors r+ |
|
Build succeeded: |
Allow tests to be updated automatically Uses the `expect-test` crate (thanks for the recommendation `@lnicola!)` to implement snapshot testing. `expect-test` is a bit more flexible than `insta`, so it seems like a good fit. This gives nicer diffs as well, since we don't strip whitespace prior to the comparison. Setting `UPDATE_EXPECT=1` before running `cargo test` will update all tests automatically. ## Blockers - [x] Omit empty lists from `Solution`'s `Display` impl. * `chalk` currently compares prefixes to make things less noisy, but that seems error prone anyways. * Harder than it seems at first, since you need to unintern to check whether a list is empty. - [x] Handle recursive/SLG comparisons better. * Many tests use `yields`, which compares the results of the two solvers against a known output. Unfortunately, the test macro will duplicate the `expect` invocation, so if neither output matches the expected one, `expect-test` will overwrite it *twice*, almost assuredly corrupting the test file. * There's many ways to fix this. ~~I've hacked around it here by adding a method to `Expect` (which is not yet upstream).~~ Most robust is to collect the various engine results first and only compare them against the output if they agree. *This is implemented now*. - [x] rust-analyzer/expect-test#20 ## Nice to haves - [x] Don't use raw strings with a hash unconditionally when blessing (rust-analyzer/expect-test#28). - [x] Shorter alias for `expect![[`? It's currently hard-coded into the `expect-test` crate (rust-analyzer/expect-test#26). - [ ] Remove newline escapes from existing tests.
27: Support single delimiter version of `expect![]` r=matklad a=ecstatic-morse Pairs of braces are no longer necessary that we parse string literals. Allow users to omit the inner `[]`. This represents a significant change to the public API, and you might have other tooling that depends on the paired braces, so I understand if you don't want it upstream. I think it reads a bit nicer, though. Depends on #26 (although it could be separated), and is similarly lacking in integration tests for now. Co-authored-by: Dylan MacKenzie <ecstaticmorse@gmail.com>
Currently,
expect-testchecks for the stringexpect![[when updating test cases. As a result, settingUPDATE_EXPECT=1will cause a panic ifexpectis imported under a different name.We have the
columnof the macro invocation for eachExpect. Use that as a starting point instead.