Sukera

It's just a different suggestion to avoid conflicts with the ecosystem :shrug: In [Hypothesis](https://hypothesis.readthedocs.io/en/latest/details.html#making-assumptions) (and my soon-to-be-released port of it), `assume` is a verb to designate conditions that MUST be...

A `@fastmath` version is IMO not a solution to the non-`@fastmath` version being slower than necessary - the assumptions violated by `@fastmath` need to hold for the regular versions after...

> Could one always use @fastmath for sum and prod?. When I tried it out, the fast versions dealt with NaN correctly. If that were possible, we'd already do that....

I've modified your benchmarking script a bit, to ensure each evaluation starts on fresh & newly generated data, so that there's no chance of caching effects arbitrarily inflating some numbers....

> The speedup for prod(v; init=T(1)) makes it as fast as prod(v) on my machine. Do you think one can get faster than that? I'm not at the computer I...

I'm almost positive that it's correct, but I must admit, once it looked half-way decent I didn't actually check :thinking: Here's the case I was looking at, which seems to...

I've encountered the same problem when using `c++`

Come to think of it, the argument checking out so allow raw symbols as well :thinking: So this additional version would be needed too: ``` macro constfield(ex::Symbol) ex = esc(ex)...

Yeah, that's why I ultimately ended up with the `@const`-per-field macro that's a no-op on

There's also these error codes that seem to overlap considerably: https://github.com/julia-vscode/LanguageServer.jl/blob/e67befb27eb0783a7dcbef11eca8513d15e3a0a5/src/requests/features.jl#L155-L167 This error code is seemingly used for a formatting failure, but the [LSP spec](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#errorCodes) says this specific code should...