Quick fixes for record positional record literals

[CoderDake]

Tasks

• Implement new diagnostic RECORD_LITERAL_ONE_POSITIONAL_NO_TRAILING_COMMA b0078ad
• Add a quick fix for RECORD_LITERAL_ONE_POSITIONAL_NO_TRAILING_COMMA 43fbb1d

---

Context

While testing records I found that the following gives an error:

(int,) intRecord1 = (1);

The error is:

A value of type 'int' can't be assigned to a variable of type '(int)'.
Try changing the type of the variable, or casting the right-hand type to '(int)'.

After doing some reserearch into the spec I found out that records that look like this must have a trailing comma to show that they are a record. This was not immediately apparent when reading the errors or looking at the quick fixes though.

So the proper way to write this would have been:

(int,) intRecord1 = (1,);

It would be helpful if the error mentioned that records must end in a comma, or perhaps there was a quick fix that offered to add the trailing comma.

[CoderDake]

@bwilkerson I heard you might be interested in this

[bwilkerson]

Yes, I'm definitely interested. @keertip @pq might also be interested.

I think the right solution here is to define a new diagnostic for this case (because that's how we can have a better message here). We might even be able to re-use ParserErrorCode.RECORD_LITERAL_ONE_POSITIONAL_NO_TRAILING_COMMA, though we don't normally produce parse errors outside the parser, so that might be too odd.

Then we can create a fix that will add the comma. (That fix could be applied to a couple of other diagnostics as well.)

[bwilkerson]

I gave this a P2, but I'd really like to see it ship with the feature because I think it will help keep users on the rails.

[pq]

I agree. This feels like a significant foot-gun. I'l take a look.

[pq]

Completed w/ 43fbb1d.

Thanks again for the report @CoderDake!