Proposal: readonly local variables + Explicit 'mutable' specifier to help with code analysis + Related compiler settings

[Lehonti]

Currently, fields can be assigned and re-assigned by default, unless an explicit readonly modifier is used. readonly fields exist to help the programmer ensure certain invariants during the lifetime of an object.

My first proposal is to provide a similar functionality for local variables. They would have to be assigned a value during declaration and this would ensure certain invariants during a function call. Something like this:

void DoSomething (ExecutionContext context) {
    readonly int number = _numberFactory.Create(context); // OK
    number = 4; // Error: 'number' can't be assigned a value after declaration since it is read-only.
}

My second proposal is to create a new mutable modifier for fields and variables, which would allow us to explicitly specify the semantics that fields now have by default (when the readonly modifier is not present). This would help code analysis. Someone could configure their compiler such that fields and variables without any specifier are treated as readonly by default, or at least give info/warnings/errors when a non-mutable field or variable is re-assigned to, just like we can currently configure the compiler to allow for non-nullable reference types to exist.

Edit: It would also be a good idea to have readonly parameters, as proposed here. Thanks, @svick

[svick]

My first proposal is to provide a similar functionality for local variables.

That sounds like a duplicate of #188.

My second proposal is to create a new mutable modifier for fields and variables

At least for fields, that sounds like something that could be achieved using an attribute.

[Lehonti]

Thanks @svick for noting that my proposal includes something that was already suggested. Still, I think it makes the most sense in combination with the mutable part. Also, if attributes could be applied to local variables, I could probably buy the solution you are proposing. Sadly, this is not the case as far as I know.

[svick]

if attributes could be applied to local variables

There's a proposal for that too: #2478. Though it hasn't seen any movement in quite some time.

[Richiban]

Wow. #188. That issue was championed 3 ½ years ago!

I know about the prioritisation process (roughly) for features but, damn, I thought this one would have been chosen for development by now.