Warn on useless comparisons

[Papipo]

Hi there,

I had a typo in some code and I was comparing a variable with itself, like group_id != group_id. One of those was supposed to be group.id, and I think it's a common typo.

The idea is that the compiler is able to detect such comparisons and warn about them, since they will always evaluate to the same and are completely useless.

Thanks 🙇🏽

[Ramkarthik]

@lpil and @Papipo If we are doing this, would something like the below work?

Variable:

warning: Same operand on either side of the equality operator
  ┌─ /gleam/test/equality/src/equality.gleam:8:8
  │
8 │   case greeting != greeting {
  │        ^^^^^^^^^^^^^^^^^^^^ This will always be false

Field access:

warning: Same operand on either side of the equality operator
  ┌─ /gleam/test/equality/src/equality.gleam:8:8
  │
8 │   case wibble.age == wibble.age {
  │        ^^^^^^^^^^^^^^^^^^^^^^^^ This will always be true

Would we want to implement this for anything else apart from variables and field access?

[giacomocavalieri]

Just field access, module selects, literal values and variables I'd say. @Ramkarthik are you planning to work on this?

[Ramkarthik]

@giacomocavalieri I have implemented this for field access and variables to emit the warning mentioned in my previous comment. I can look into implementing it for module selects and literal values as well. I'll give it a try today.

I'm assuming we would also want this for gt, gte, 'lt, and lte` for Int and Float.

[giacomocavalieri]

Great thank you!! Not sure we also want it for those operators, I'd just start with equality checks 👍

[lpil]

Hello! Good idea @Papipo, thank you.

@Ramkarthik @giacomocavalieri The goal is to warn for useless comparisons generally, not for equality checks with the same value specifically.

These should emit a warning saying they always evaluate to true:

1 < 2
3 == 3
1 != 2
1.5 >. 1.1
xyz == xyz
wibble.wobble == wibble.wobble

These should emit a warning saying they always evaluate to false:

1 > 2
3 != 3
1 == 2
1.5 <. 1.1

[Ramkarthik]

@lpil Got it, that makes sense. I can work on this on the same PR (or a new one) if you'd like, or leave it open for someone/future. Would like to know your opinion on what the warning title should be. The hint can be "This is always [true | false]" depending on the condition.