Feat: Add try_maps on AccountInfo Ref/RefMut

[stegaBOB]

No description provided.

[Aursen]

Did you benchmark this kind of changes?

[sonicfromnewyoke]

@Aursen i expect small performance degradation

[stegaBOB]

I'll benchmark it, but I doubt this would have any performance impacts. The compiler should optimize the error branch out.

[stegaBOB]

@Aursen @sonicfromnewyoke https://rust.godbolt.org/z/9YeWsr1oP map with try_map infallible under the hood compiles down to the same assembly, so this should have 0 performance impact. This should be trivial for the compiler to optimize. One of the reasons Infallible exists :)

[sonicfromnewyoke]

@Aursen @sonicfromnewyoke https://rust.godbolt.org/z/9YeWsr1oP map with try_map infallible under the hood compiles down to the same assembly, so this should have 0 performance impact. This should be trivial for the compiler to optimize. One of the reasons Infallible exists :)

@stegaBOB it works even better than i expected, thanks

the last question: what do you think about inlining? Will it be better to use #[inline(always)] instead of putting the decision on the compiler??

[Aursen]

@stegaBOB it works even better than i expected, thanks

the last question: what do you think about inlining? Will it be better to use #[inline(always)] instead of putting the decision on the compiler??

I think for this kind of choice, @febo just benchmarks

[stegaBOB]

the last question: what do you think about inlining? Will it be better to use #[inline(always)] instead of putting the decision on the compiler??

I think #[inline(always)] should only really be done if you have serious benchmarks to justify it (or if the function is just empty). The compiler will probably make the right choice

[febo]

@stegaBOB it works even better than i expected, thanks
the last question: what do you think about inlining? Will it be better to use #[inline(always)] instead of putting the decision on the compiler??

I think for this kind of choice, @febo just benchmarks

In this case #[inline] seems more appropriate since we don't know what the mapping function will be – it might simple, but it might not.

[febo]

👌

It would be good to include a test.

[febo]

Great, thanks!