New lint: replace iter.filter(x).count() > 0 with iter.any(x)

[ebroto]

What it does

It suggests replacing iter.filter(x).count() > 0 with iter.any(x). The suggestion should be MachineApplicable.

Categories (optional)

• Kind: complexity or performance

What is the advantage of the recommended code over the original code

It's more concise and it should be faster since any is allowed to return early while filter().count() has to walk through all items.

Drawbacks

None.

Example

segments.iter().filter(|ps| ps.ident.as_str() == "prelude").count() > 0

Could be written as:

segments.iter().any(|ps| ps.ident.as_str() == "prelude")

[matthiaskrgr]

I think this might as well fit into the perf category since .filter().count() has to walk through all iterator items but .any() can return early?

[matthiaskrgr]

Hmm, but that might also cause false positives for the case that the .filter() call has side effects and the .any() call reduces the number of side-effects that happen :/

[ebroto]

Indeed! That's similar to #3351

[flip1995]

The lint idea is good, but we have to think about side-effects. I think this is hard to do/to get right. Cases where we have side effects would be:

• mutating an UpVar
• calling a function, which has side effects (hiding this inside an iterator is a bad idea IMO, but suggesting replacing this with any would still change the semantics)
• Any kind of I/O
• ???

[camsteffen]

This could be generalized to suggest iter.count() > 0 -> iter.next().is_some() in some cases.

[pickfire]

all() could also be applied here as well as != 0 rather than > 0.