Improve treatment of ellipsis in `define-judgment-form` about redex HOT 9 OPEN

I would probably define a judgment-form instead of using ,.

from redex.

I disabled check-clauses in judgement-form.rkt by comment out its body and making it return rest. The error stays the same.

I went even further and redefined raise-syntax-error to a no-op in the scope of judgement-form.rkt. The error stays the same.

I think this rules out the "best case", i.e., a more relaxed well-formedness check isn't going to solve the problem. Do you agree?

In order to explore the "worst case", I believe I should start in compile-clause. There are many things in it that go over my head, so any directions from you are going to help a lot.

Thanks so much for the feedback, @rfindler 😄

from redex.

I think that the call that raises the error is this one:

https://github.com/racket/redex/blob/master/redex-lib/redex/private/judgment-form.rkt#L141

from redex.

Of course, you are right. Disabling the check—i.e., returning cs instead of raising an error with raise-ellipsis-depth-error—changes the error message in the test to "syntax: missing ellipsis with pattern variable in template in: τ", referring to the τ that occurs outside amb in the conclusion of the rule. With that, the easy route has been definitely ruled, I think.

Now would be the time to understand the system a bit better to come up with a solution. I believe I'll need to change how the binding constraint is generated in the case at hand. When you said Redex uses racket/datum, did you mean syntax/datum? I'm studying now how it works.

Side question: When you are working on this, how do you debug? The Debugger in DrRacket fails right away, because the code fails in macro expansion at compile time. The Macro Stepper helps, by allowing me to see the macro expansions, but it is limited in that I can't, for example, set breakpoints.

Thank you!

from redex.

Yes, I meant syntax/datum! Thanks, duh.

When I'm working on this, I use the stacktrace information from syntax errors (that's how I pointed you to the line above) and I use printf debugging, and I combine that with some experience in knowing when things run in this code and just in Racket macros in general. Also I tried to work carefully with small input programs and test individual functions. But generally I don't think using the macro stepper or the debugger is probably what I would do for this particular problem. I think using the macro stepper might be helpful to you to figure out how some of the pieces of Redex fit together, tho. I mostly just try to use my memory for that job -- but the macro stepper is better than my memory and I should be using it more than I do.

from redex.

I was following the tutorial and found this discussion after hitting this issue. Is the following still the best way to answer question 6 for now?

[(types-exercise-6 Γ e t_2) ...
 (side-condition ,(andmap (curry equal? (term t_1)) (term (t_2 ...))))
 -------------------------
 (types-exercise-6 Γ (amb t_1 e ...) t_1)]

from redex.

Ack. I was merely copy-pasting this solution from the other thread but I guess my question was: is a side-condition the best solution for now? Sounds like the answer is yes. Thanks!

from redex.

from redex.

Got it. Thanks.

from redex.