feat: Remove support for 1.58 proc-macro abi
This seems old enough that we can drop the support for it now, the less ABIs we have the less work it is adjusting our span implementation.
Extracted from https://github.com/rust-lang/rust-analyzer/pull/14061, will rebase that over this once merged.
feat: Improve "match to let else" assist
Closes https://github.com/rust-lang/rust-analyzer/issues/13540
Handles complex `let` patterns (rather than just idents), and diverging block expressions have their `{`/`}` stripped to create nicer code.
Update documentation for emacs and split it for LSP-mode and Eglot
Emacs has now two LSP clients, the more minimalistic and lightweight [Eglot](https://joaotavora.github.io/eglot) and the extensive though a bit bloated [LSP Mode](https://github.com/emacs-lsp/lsp-mode). Eglot will soon be shipped with Emacs29. Both have rust-analyzer enabled by default and require no further setup other then being installed and enabled. `lsp-rust.el` is not required anymore.
The base-installation for each of those modes is so easy now that I don't think an enumerated list is necessary, both packages can be installed via the standard `M-x package-install` and the installation is a one-liner that I provide.
Configuration is only necessary for supporting the rust-analyzer extensions to the LSP protocol, which are built into LSP mode and require an [extension-package](https://github.com/nemethf/eglot-x) for Eglot.
But for further documentation, including the LSP extensions, I link against official documentation where possible to avoid duplicating efforts having to continually update this to stay up-to-date.
I rewrote most of the original emacs documentation, but the [linked blog](https://robert.kra.hn/posts/2021-02-07_rust-with-emacs/) post by `@rksm` seems still being actively updated with updates to LSP mode, so I kept the link. That blog post is opinionated, I personally use different packages which achieve similar end results (Eglot instead of LSP-mode, corfu instead of capf, vertico instead of helm etc.). But if someone doesn't already have an extensive Emacs configuration, I think this is not a bad starting point.
Disclaimer: I'm a Rust beginner, which is why I read the rust-analyzer setup docs. So I necessarily know how most Rust experts use Emacs. But I'm an experienced Emacs user who uses several other programming languages via LSP-mode support in Emacs. I used both, initially LSP-mode and recently migrated to Eglot.
Also I'm not an experienced in writing asciidoc and I didn't do a local test-built, hopefully the html builds in the way I imagine it. So I recommend to check that aspect of the PR. Maybe the documentation is in the CI build-artifacts?
This is a duplicate of a PR to the old rust-analyzer project https://github.com/rust-analyzer/rust-analyzer.github.io/pull/197, which I made because I didn't know that the documentation now lives here.
Emacs has now two LSP clients, the more minimalistic and lightweight Eglot and
the extensive though a bit bloated LSP-Mode. Eglot will soon be
shipped with Emacs29. Both have rust-analyzer enabled by default and require
no further setup then just being installed and enabled. `lsp-rust.el` is not
required anymore.
The base-installation for each of those modes is so easy now that I don't think
an enumerated list is necessary, both package can be installed via the standard
`M-x package-install` and the installation is a one-liner that I provide.
Configuration mostly comes into play for support the rust-analyzer extensions to
the LSP protocol, which are built into LSP mode and require an extension-package
for Eglot.
But for the configuration beyond the base configuration I link against official
documentation, quickstart guides and documentation for the lsp extensions, to
avoid showing outdated information here.
This commit is mostly a duplicate of a PR [1] that I made against the
rust-analyzer github project.
[1]: https://github.com/rust-analyzer/rust-analyzer.github.io/pull/197,
7ff0113006
Spelling: Space before version number in Emacs 29 in manual
Co-authored-by: Laurențiu Nicola <lnicola@users.noreply.github.com>
internal: remove `TypeWalk`
Because less code is better!
`hir_ty::TypeWalk` is only used in analysis-stats and its usage can be replaced by checking `TypeFlags` (which is precomputed upon `TyKind` interning so it should make analysis-stats a bit faster, though it was really negligible in my local environment).
We should just use chalk's `TypeVisitor` or `TypeFolder` instead even if we come to need it again.
Don't escape non-snippets in assist
I was misunderstanding that we're always sending snippets as response to assist request. For assists that never return snippets like `move_const_to_impl` we don't need to escape, and I don't think `utils::escape_non_snippet()` is useful at the moment since we guarantee that only a single edit will have `InsertTextFormat.Snippet` and we have `utils::render_snippet()` for that.
