Re:Structuring Arrays with Algebraic Shapes [video]
by thesz on Monday July 21, 2025 @06:30AM
(#44633593)
Arithmetic inside the types not necessarily introduces undecidability. One example is telescopes for indices:
To safely access an element in array you have to provide a proof that a telescope can be constructed for given index range.
[ Parent
| on HN ]
Re:Structuring Arrays with Algebraic Shapes [video]
by bgavran on Monday July 21, 2025 @10:24AM
(#44635457)
Indeed, the author seems to have a misunderstanding about both undecidability and complexity as they pertain to dependent types.
Dependent types do not add complexity to our system, they reveal it.
Case in point: here is a fully dependently-typed tensor processing framework written in Idris, which I believe matches most of the desiderata of his talk, capturing even a generalisation of arrays via Naperian functors that is mentioned at one point.