3 ms·
You seem to be operating from an overly narrow perspective. Type theories were invented for mathematical logic and only co-opted for programming later. It doesn
by andrewflnr 8y ago
You seem to be operating from an overly narrow perspective. Type theories were invented for mathematical logic and only co-opted for programming later. It doesn't make sense to try to prove things about type systems by referring to a runtime. The rest of your objections show similar misunderstandings of the question. While I said myself a dependent type system is not free, its costs have nothing to do with "indirection", whether you mean type system complication or pointer chasing (the latter is more likely to be reduced). Randomness is about computational purity, which is nearly orthogonal to the expressiveness of your type system. I'm not completely sure what you're trying to say about hardware, not I'm pretty sure it's a red herring, at least from the theoretical perspective relevant to characterizing dependent types.
The Wikipedia article on dependent types is a pretty decent survey: https://en.wikipedia.org/wiki/Dependent_type https://en.wikipedia.org/wiki/Dependent_type
- kazinator 8y ago"Dependent types were created to deepen the connection between programming and logic. {clarification needed}"
- Retra 8y agoI don't know what you're trying to say. The point I am trying to refute is "the whole type-level and value-level is a false dichotomy", and the existence of dependent types do not support that claim, they only provide a narrow domain where the distinction is blurred. > Randomness is about computational purity, which is nearly orthogonal to the expressiveness of your type system. If types and values are to be considered identical, then random types should be just as useful as random values. They are not. That alone should indicate that values and types have a well-motivated distinction. Additionally, dependent types offer no additional capabilities, only convenience. Having a type defined in terms of a value communicates nothing more to a compiler than would a functionally equivalent assertion, though it would certainly be easier to use and work with in some situations. I'm not trying to say anything about dependent types here. I'm trying to say that the goal of producing a language which makes no distinction between type and value is neither useful nor possible.
- andrewflnr 8y agoTo be quite blunt, what I'm saying is that you don't have enough context to have an informed opinion on the relationship between dependent types and the distinction between types and values. It's pretty clear you don't have a clear idea of what a dependent type system actually is. People who study type theory and logic are pretty much on the same page about this.
- Retra 8y agoOn the same page about what, exactly?