Re: Performance of records with variant parts

[John Perry <john.perry@usm.edu>]

On Thursday, April 1, 2021 at 8:38:49 AM UTC-5, Niklas Holsti wrote:
> I assume by "X is variant" you mean that X has a variant_part (RM 3.8.1(2)). 
> 
> If the discriminant of Thing_Type has a default value, Intersection_Type 
> can have a Thing_Type component without supplying it with a discriminant 
> value, and Intersection_Type need not have discriminants nor a variant part.

You're right. Originally Thing_Type didn't have a default value. Later I added one, but I didn't think to change Intersection_Type.

Either way, the discriminant on Intersection_Type was the problem. Once I modify Intersection_Type so that it is no longer variant, Intersections can return an Intersection_Type that contains a Thing_Type without the performance hit it had when Intersection_Type was variant.