fix help text for rust-analyzer.check.invocation{Strategy,Location}
I highly doubt that `check.invocationLocation` only has an effect if `cargo.buildScripts.overrideCommand` is set -- looks like a copy-paste mistake from `buildScripts.invocationLocation` to me.
feat: Implement extern crate completion
Hi, this is a draft PR for #13002.
I have basic completion working as well as a filter for existing extern crate imports in the same file. This is based on the tests, I have not actually tried this in an editor. Before going further I think this is a good point to stop and get feedback on the
structure and approach I have taken so far. Let me know what you think :)
I will make sure to add more tests, rebase commits and align with the code style guidelines before submitting a final version.
A few specific questions :
1. Is there a better way to check for matching suggestions? right now I just test if an extern crate name starts with the current
user input.
2. Am I creating the `CompletionItem` correctly? I noticed that `use_.rs` invokes a builder where as I do not.
3. When checking for existing extern crate imports the current implementation only looks at the current source file, is that sufficient?
fix: avoid problematic serde release
serde 1.0.172 and up rely on opaque non-reproducible binary blobs to function, explicitly not providing a library-level opt-out.
This is problematic for two reasons:
- directly, unauditable binary blobs are a security issue.
- indirectly, it becomes much harder to predict future behaviors of the crate.
As such, I am willing to go on a limb here and forbid building rust-analyzer with those versions of serde. Normally, my philosophy is to defer the choice to the end user, but it's also a design constraint of rust-analyzer that we don't run random binaries downloaded from the internet without explicit user's concent.
Concretely, this upper-bounds serde for both rust-analyzer workspace, as well as the lsp-server lib.
See https://github.com/serde-rs/serde/issues/2538 for wider context.
serde 1.0.172 and up rely on opaque non-reproducible binary blobs to
function, explicitly not providing a library-level opt-out.
This is problematic for two reasons:
- directly, unauditable binary blobs are a security issue.
- indirectly, it becomes much harder to predict future behaviors of the
crate.
As such, I am willing to go on a limb here and forbid building
rust-analyzer with those versions of serde. Normally, my philosophy is
to defer the choice to the end user, but it's also a design constraint
of rust-analyzer that we don't run random binaries downloaded from the
internet without explicit user's concent.
Concretely, this upper-bounds serde for both rust-analyzer workspace, as
well as the lsp-server lib.
See https://github.com/serde-rs/serde/issues/2538 for wider context.
minor : Deunwrap generate_derive
#15398 subtask 1. Since the editing closure has arms, I did something *experimental* ( in this case just a clever term for bad code ) to bypass creating an `Option` but I am ready to change this.
internal: Record import origins in ItemScope and PerNS
This records the import items definitions come from in the module scope (as well as what an import resolves to in an ItemScope). It does ignore glob imports as thats a lot more work for little to no gain, glob imports act as if the importing items are "inlined" into the scope which suffices for almost all use cases I believe (to my knowledge, attributes on them have little effect).
There is still a lot of work needed to make this available to the IDE layer, but this lays out the ground work for havin IDE layer support.
cc https://github.com/rust-lang/rust-analyzer/issues/14079
internal: Add offset param to token descending API
The offset is unused for now as we can't map by spans yet but it will be required for https://github.com/rust-lang/rust-analyzer/issues/11260 to work once the token map has been changed to record spans
fix: start hovering default values of generic constants
It's just a kind of a postscriptum for [my last PR](https://github.com/rust-lang/rust-analyzer/pull/15179) adding default values of const generics to `hir::ConstParamData`. Here I patch other pieces of code which used to ignore const default values and which I managed to find (you're welcome to show me more)
Increase the buffer size for discover project command
The default value for maxBuffer is 1 MiB[1]. If the discover project command returns stdout or stderr that is greater than 1 MiB, the extension would error with "RangeError: stderr maxBuffer length exceeded".
Set the default value for maxBuffer to 10 MiB for project discovery.
[1] https://nodejs.org/api/child_process.html#child_processexeccommand-options-callback
the "add missing members" assists: implemented substitution of default values of const params
To achieve this, I've made `hir::ConstParamData` store the default values
internal : rewrite DeMorgan assist
fixes#15239 , #15240 . This PR is a rewrite of the DeMorgan assist that essentially rids of all the string manipulation and modifies syntax trees to apply demorgan on a binary expr. The main reason for the rewrite is that I wanted to use `Expr::needs_parens_in` method to see if the expr on which the assist is applied would still need the parens it had once the parent expression's operator had equal precedence with that of the expression. I used `.clone_(subtree|for_update)` left and right and probably more than I should have, so I would also be happy to hear how I could have prevented redundant cloning.
Suggest type completions for type arguments and constant completions for constant arguments
When determining completions for generic arguments, suggest only types or only constants if the corresponding generic parameter is a type parameter or constant parameter.
Closes#12568
Fix signature help of methods from macros
Currently the receiver type is copied from AST instead re-formatting through `HirDisplay`. Macro generated functions seem to have no spaces and their signature help are rendered like `fn foo(&'amutself)` instead of `fn foo(&'a mut self)`.
The default value for maxBuffer is 1 MiB[1]. If the discover project
command returns stdout or stderr that is greater than 1 MiB, the
extension would error with "RangeError: stderr maxBuffer length
exceeded".
Set the default value for maxBuffer to 10 MiB for project discovery.
[1] https://nodejs.org/api/child_process.html#child_processexeccommand-options-callback
