[Ur] XML records from higher-order type classes

Adam Chlipala adamc at csail.mit.edu
Fri Apr 10 08:37:49 EDT 2015


On 04/09/2015 09:28 PM, Benjamin Barenblat wrote:
> On Thursday, April  9, 2015, at  8:41 pm EDT, Adam Chlipala wrote:
>> Probably putting a full type annotation on the formal parameter
>> [witness] will solve this problem.
> I’ll be darned, Adam, that did the trick.  How did you know that was
> going to fix it?  I never would have guessed that.

Adding more type annotations on polymorphically typed formal parameters 
is the general way to help the Ur inference engine cope with 
undecidability.  It should more or less always be the case that fully 
annotated function definitions go through, if they really are 
type-correct; and there are no guarantees otherwise, but many other 
cases do "get lucky" and work.



More information about the Ur mailing list