Fix the res running variable is not consistent with DiffTests

[NeilKleistGao]

Relative Issue: #65

Since every statement will change the res variable, I tried to add the type information into the context so that the correct type could be found. I also used a flag variable to indicate self-define res. I'm not sure expected results and it may need further discussion.

[LPTK]

Cool, thanks! I'd just a minor change to be done, as explained below.

[LPTK]

The outputs look correct. But I'm having trouble figuring out if this implementation is doing the right thing. It looks like it's adding hacks here and there instead of cleanly solving the issue. @chengluyu can you please review and suggest a better way?

[LPTK]

This looks like a huge hack. Making this logic rely on the fresh name encoding of the Scope class is totally brittle (what if we want to change it later?). You should add the correct logic earlier, nearer to the place where this thing is decided.

[LPTK]

Wait so this logic was not needed at all?

[NeilKleistGao]

I found something wrong in shared\src\test\diff\mlscript\BadInherit2.mls, where:

f'
//│ res: 1
//│    = A1 { Foo1: [Function (anonymous)] }

Initially, I thought it might be a mistake because f' should be 1, so I added this check to avoid it. However, I found it wrong in the original file:

f'
//│ res: 1
//│    = 0

So I think maybe it should not print 1 at all, and this check is redundant (or at least should be fixed in another pr).

[LPTK]

Ah I see! This is actually an instance of #65 (comment)

It works without the tick:

def f_ = 1
//│ f_: 1
//│   = 1

f_
//│ res: 1
//│    = 1

We need to replace ticks in JS with some other symbol. This should be doable with a small modification to the scope class. Could you do it in this PR – or in another one if you prefer?

[NeilKleistGao]

I can fix it in another PR later. 😄

[LPTK]

Ok cool, please do!

[LPTK]

LGTM now!

[LPTK]

Thanks @NeilKleistGao