Sam Tobin-Hochstadt

icon indicating a website link http://samth.github.io icon indicating an email [email protected]

icon company PLT @ Indiana University icon location Bloomington, IN icon location I'm an associate professor at Indiana University, and a core developer of @racket.

Results 498 comments of Sam Tobin-Hochstadt

[Feature Request] Freeze and microwave treelists and hash

This is basically the "transient" idea, first popularized in Clojure. This paper describes transients for RRB trees (what treelists are) but in C++, and uses some things we don't have...

Fix type-checking list elements

Thanks, this looks like a reasonable approach.

acos unsound in TR

Unfortunately, fixing this exposes problems in a number of other libraries. In particular, some `math` library operations will have different and more limited types (they were of course wrong before).

check-exn should not fail on opaque value

We should just change `typed/rackunit` to use `unsafe-require/typed` in a bunch of places.

Typed Racket pretends mutable sets don't exist, and yet it assumes all hashtables could be mutable

Immutable hash tables would support `hash-set`, and you can't pass a mutable hash table where one is expected. Read-only tables would admit mutable ones, but you could only read them....

Typed Racket pretends mutable sets don't exist, and yet it assumes all hashtables could be mutable

Yes, all of these things would be good ideas, and I think we could make everything compatible.

Typed Racket pretends mutable sets don't exist, and yet it assumes all hashtables could be mutable

`HashTable` would be a supertype of the other types. I think the story for `Set` is less obvious. It's not clear that supporting more without supporting generic sets is useful.

Typed Racket pretends mutable sets don't exist, and yet it assumes all hashtables could be mutable

Sure, if there's need, we should support it. I think we should just split `Set` and `MutableSet` though.