Make ** be right associative

[asterite]

For https://forum.crystal-lang.org/t/exponentiation-operator-has-unexpected-behaviour/2457

I think it would be better if the associativity is the same as most major languages out there. And we better do this change before 1.0

[bcardiff]

Can we add a spec to settle also the -2 ** 2 case as explained by https://codeplea.com/exponentiation-associativity-options ?

As of 0.35.1 -2 ** 2 == 4, but in Ruby is -4.

[asterite]

Oh, that's totally unexpected for me, but I guess if most other programming languages, and even Google, work this way it makes sense to do it like that.

[bcardiff]

I also find it confusing :-). Even if we choose to leave as it is and behave differently from others a spec will settle the choice at least.

[asterite]

I don't know how to make it work :'-(

[asterite]

Feel free to try to make it work. It's already tricky as it is, but it's even more tricky with things like -2 ** 4 being parsed as -(2 ** 4), where -2 is parsed as a single token.

Maybe we can leave things as they are now, but document them well.

[radha]

Feel free to try to make it work. It's already tricky as it is, but it's even more tricky with things like -2 ** 4 being parsed as -(2 ** 4), where -2 is parsed as a single token.

Maybe we can leave things as they are now, but document them well.

Unary minus has a higher precedence than exponentiation in crystal. So it is expected -2 ** 2 == 4. In ruby the order is unary plus > exponentiaion > unary minus.

I think associativity of exponentiation and precedence of unary minus are separate problems.

[asterite]

Yes, I think they are separate problems too. Should we merge this and tackle the other problem later if needed?

[bcardiff]

Ok, sure.

[bcardiff]

@asterite I'm reopening this. I think we can merge this and change the ** associativity to match what most programming languages do as a convention.

Yet, I think the pr and code should be named right associative 🙈 . Left associative is what we have now.

Do you want to do the rename or should I?

[asterite]

should be named right associative 🙈

Oops! 😄

[asterite]

Ready for another review... I can't change the branch name, though :-P

[bcardiff]

👍

FYI, I continued the discussion of unary negation in the forum thread.