Implement always `true` or always `false` `if` and `elseif` conditions

[manhunto]

Implement always `true` or always `false` `if` and `elseif` conditions

[maks-rafalko]

In my todo list there was an idea to implement if removal mutator, so IMO removing any if/elseif conditions completely would be a great addition to the mutator set.

and could be another mutator that set it to true

need to think more about it with some examples, probably worth trying to implement and check what it brings

[manhunto]

Yes, I also thought about if/else removal. What do you think is a better idea? Removal or setting it to false? IMO it would bring the same value. It seems to me that. removing if will be hard to implement.

I also thought about boolean conditions (&& and ||) that are not in if/elseif statements. It could be done in the next PR. But I need to think about it or try to implement POC.

-$var = $a && $b;
+$var = false;

[manhunto]

It seems to me that. removing if will be hard to implement.

Ok, I have to take those words back.. I thought that there is no empty node, but I found it in repo Node\Stmt\Nop. So let me know which approach, in your opinion, would be better ;)

EDIT: but there is still problem when there is elseif statement. It will produce syntax error.

[maks-rafalko]

let's try with changing condition to false then rather than removing. But feel free to change the strategy if removing will work better for some reason

[sanmai]

I'm afraid I'm going to upset you all. I think came up before, and the biggest argument against these mutations was that they are very easy to kill (or downright impossible), very obvious from the coverage report, and most of the time they are going to waste CPU cycles, skewing the metrics.

Here are all mutations this example currently generates: https://infection-php.dev/r/3w26

If I had a branch like this in my code, it is probably because it was added as an optimization to avoid costly code paths. So if I have to kill the mutation proposed here, I'm supposed to remove an optimization because there's no way to test it. How can we argue this is a good thing?

[manhunto]

@sanmai you are right, I agree with you. Most of these mutation would not be useful but there will be a lot of it

[maks-rafalko]

@manhunto did you do some analysis and have examples?

[manhunto]

No, It is just my assumption and when I thought about it may be true. I will do more research after two weeks because I'm on holidays now