fix: crystal next compatibility

[veelenga]

refs crystal-lang/crystal#11597
fixes #406

cc @straight-shoota

[Sija]

Lint/UselessAssign rule is using transformed?, just fyi

[veelenga]

Lint/UselessAssign rule is using transformed?, just fyi

Exactly, I left it for BC. It is not needed anymore for Crystal master and above.

[Sija]

@veelenga Could you please explain what do you mean by BC in this particular case? And why it's not longer needed?

[veelenga]

@Sija, sure.

In this PR, we've introduced support for Crystal's master version and subsequent versions. The term "BC" refers to the backward compatibility of Crystal (1.9.x) and earlier versions.

We implemented the transformed? method in the past to determine if a node represents an assignment transformed by a compiler. To illustrate, in Crystal 1.9.x, the following code:

collection.each do |(a, b)|
  puts b
end

is transformed before the AST phase to:

collection.each do |__arg0|
  a = __arg0[0]
  b = __arg0[1]
  puts(b)
end

This transformation leads to false positives since the assignment to the variable a becomes redundant. So we had to handle that case "manually" via #transformed? method.

However, in the Crystal master, the block mentioned above remains unaltered, and the resulting AST mirrors the original code:

collection.each do |(a, b)|
  puts(b)
end

Consequently, for the Crystal master and later versions, the transformed? method becomes unnecessary. However, I suggest retaining this method until the next release for the sake of backward compatibility. This ensures that users using both Ameba's master and Crystal 1.9.x will still not encounter these false positives.

[Sija]

@veelenga Thanks for the thorough explanation ❤️ It all makes sense (re-reading referenced PR helped too 😅), let's have this shipped then!