diesel tables have unresolved methods

[Veykril]

          Hi! Just updated to the latest version of rust-analyzer and now all my diesel models are red

Originally posted by @lenstr in #14243 (comment)

[Elrendio]

Hi!

I'm also impacted by a similar bug. All diesel table usage is broken, currently I see the error:

no method `inner_join` on type `<table as FindDsl<OrderLineId>>::Output`

Thanks 🙏

[Veykril]

You can put unresolved-method in the rust-analyzer.diagnostics.disabled setting to silence the warning

[Elrendio]

You can put unresolved-method in the rust-analyzer.diagnostics.disabled setting to silence the warning

Thanks for the temporary fix, do you know if rust-analyzer plans to provide a long term fix any time soon? Can we provide some help?

[krypt0nn]

Same issue happens for Relm4 library, and in addition to unresolved-method also requires unresolved-field. I think it can be related to this view! macro

[edwardlee4948]

You can put unresolved-method in the rust-analyzer.diagnostics.disabled setting to silence the warning

I understand that it may come across as demanding, but if it's okay with you, I would prefer to use this as the default. I just want to avoid any unexpected redlines.

[flodiebold]

@Veykril maybe we should make these diagnostics experimental for now?

[Veykril]

Judging from the feedback I think it would be better to do so yes, I didn't expect them to trigger incorrectly that often (and they don't trigger on r-a itself at all, feels like rust-analyzer is a bad testing bed for itself ...)

[Ten0]

I understand that it may come across as demanding, but if it's okay with you, I would prefer to use this as the default.

That does not seem very reasonable (#6975) until all methods are properly resolved. (#9881 and possibly this with diesel if it's not the same issue? It seems that rust-lang/chalk#780 may fix this)

What are the upsides?

If the purpose is to get quicker feedback on a typo than rustc would give, personally I've been using:

	"editor.semanticTokenColorCustomizations": {
		"[Default Dark+]": {
			"enabled": true,
			"rules": {
				"unresolvedReference": {
					"foreground": "#fd5c5c"
				},

for more than a year, and that worked well for me. It's less intrusive but may help out as you're typing, and also highlights places where resolution doesn't work properly and a type annotation may help.

they don't trigger on r-a itself at all, feels like rust-analyzer is a bad testing bed for itself

Failure to resolve seems to largely happen when dealing with complex trait systems. I'm guessing Diesel would be a great testing bed for the general topic of #6975

[lukesneeringer]

I'm impacted by this too (I expect every Diesel user is), and I'm having trouble understanding the fix. Obviously I want rust-analyzer to tell me if there's an actual unresolved method; it just seems like it's a false positive here.

If I disable the warning, am I going to lose important feedback elsewhere?

[lnicola]

The warning did not exist until the last release. You will still get compiler errors when you save a file, assuming you didn't disable that.

[defiled]

You can use an earlier version of Rust Analyzer (v0.4.1405) as a temporary fix.

[lnicola]

@defiled the 0.4 versions are nightlies, 0.3s are releases. You can run a nightly, but beware of #14305.

[lukesneeringer]

Okay great, thank you!

[matty-r]

I switched my rust-analyzer version to v0.4.1432 and this seems to have resolved it for me.

[Veykril]

This is fixed on diesel master, the cause was the diesel proc-macro using 2015 edition name resolution rules which r-a just not supports atm.

[Ten0]

On the latest diesel master I still have quite a lot of type inference failing:
diesel-rs/diesel#3564 (comment)

I suspect #9881 may be the cause.