feat: add the ability to limit the number of threads launched by `main_loop`
## Motivation
`main_loop` defaults to launch as many threads as cpus in one machine. When developing on multi-core remote servers on multiple projects, this will lead to thousands of idle threads being created. This is very annoying when one wants check whether his program under developing is running correctly via `htop`.
<img width="756" alt="image" src="https://user-images.githubusercontent.com/41831480/206656419-fa3f0dd2-e554-4f36-be1b-29d54739930c.png">
## Contribution
This patch introduce the configuration option `rust-analyzer.numThreads` to set the desired thread number used by the main thread pool.
This should have no effects on the performance as not all threads are actually used.
<img width="1325" alt="image" src="https://user-images.githubusercontent.com/41831480/206656834-fe625c4c-b993-4771-8a82-7427c297fd41.png">
## Demonstration
The following is a snippet of `lunarvim` configuration using my own build.
```lua
vim.list_extend(lvim.lsp.automatic_configuration.skipped_servers, { "rust_analyzer" })
require("lvim.lsp.manager").setup("rust_analyzer", {
cmd = { "env", "RA_LOG=debug", "RA_LOG_FILE=/tmp/ra-test.log",
"/home/jlhu/Projects/rust-analyzer/target/debug/rust-analyzer",
},
init_options = {
numThreads = 4,
},
settings = {
cachePriming = {
numThreads = 8,
},
},
})
```
## Limitations
The `numThreads` can only be modified via `initializationOptions` in early initialisation because everything has to wait until the thread pool starts including the dynamic settings modification support.
The `numThreads` also does not reflect the end results of how many threads is actually created, because I have not yet tracked down everything that spawns threads.
This makes code more readale and concise,
moving all format arguments like `format!("{}", foo)`
into the more compact `format!("{foo}")` form.
The change was automatically created with, so there are far less change
of an accidental typo.
```
cargo clippy --fix -- -A clippy::all -W clippy::uninlined_format_args
```
The reason for that was that we were calculating the crate defmaps
of the file we are saving by accident causing us to get stuck waiting
on their expensive computation, while we only need the relevant crate
id.
When receiving multiple change events for a single file id where the
last change is a delete the server panics, as it tries to access the
file contents of a deleted file. This occurs due to the VFS changes and
the in memory file contents being updated immediately, while
`process_changes` processes the events afterwards in sequence which no
longer works as it will only observe the final file contents. By
folding these events together, we will no longer try to process these
intermediate changes, as they aren't relevant anyways.
Potentially fixes https://github.com/rust-lang/rust-analyzer/issues/13236
When r-a starts up, it starts switching the workspace before all vfs
events have been processed which causes us to switch workspace multiple
times until all vfs changes have been processed. This scales with the
size of the project and its dependencies. If workspace files from
dependencies as well as the sysroot get loaded, we shouldn't switch
the workspace as those have no impact on the project workspace.
feat: Only flycheck workspace that belongs to saved file
Supercedes https://github.com/rust-lang/rust-analyzer/pull/11038
There is still the problem that all the diagnostics are cleared, only clearing diagnostics of the relevant workspace isn't easily doable though I think, will have to dig into that
It's good that rust-analyzer doesn't belly-up on a panic in some random
assist.
It is less good that rust-analyzer devs only know that the assists are
buggy when they are actively looking at the logs.
I don't think there's anything wrong with project_model depending on
proc_macro_api directly -- fundamentally, both are about gluing our pure
data model to the messy outside world.
However, it's easy enough to avoid the dependency, so why not.
As an additional consideration, `proc_macro_api` now pulls in `base_db`.
project_model should definitely not depend on that!
From the dawn of time, when dinosaurs roamed the and we didn't have
hierarchical profiling, there was the `latest_requests` infra we used to
track the time of ten last requests.
Today, no one is actually using it and, what's more, it itself became
pretty useless -- LSP grew way more chatty, and 10 requests don't really
paint any kind of picture.
Personally, it's been years since I last looked at latest requests in
the status output.
So, let's remove a tiny bit of state from the big ball of complexity
that is `GlobalState` and `main_loop`!
