Jason Merrill

I understand how `every` works now--the `notify` wrapper just drops anything that comes in while the reactor is paused: https://github.com/elm-lang/elm-reactor/blob/60b8dca16d5ba4a95a6a2aafe7d89a5a207639d6/frontend/debugger-implementation.js#L785

@vilterp does your work on the reactor affect the plan I outlined above? Is there somewhere I can find your work? It isn't merged in yet, is it?

This looks like nice work. Was `zig fmt` able to do a lot of this? One problem: it looks to me like `make test` isn't actually running any tests. On...

Looks like this along with #45 are very similar to the existing #35. The approaches a little different, though, so I guess you can pick which one suits your tastes.

I ended up deciding to switch to [dogpile.cache](https://dogpilecache.readthedocs.org/en/latest/). Its API has a lot of the same advantages of python-cache, but it also takes care of the "dogpile problem" i.e. avoiding...

Benchmark results at various stages of this work: cf7c6c74c04479d14036fe57daf46b18302524d4 ``` test/benchmark/equality.lox min: 6458 mean: 6530 test/benchmark/binary_trees.lox min: 3290 mean: 3374 test/benchmark/properties.lox min: 704 mean: 766 test/benchmark/invocation.lox min: 484 mean: 564...

This is a big performance improvement across all benchmarks, but for some reason, clox from the crafting interpreters book is still faster on most benchmarks.

For reference, there's some interesting discussion so far on the julia-users mailing list: https://groups.google.com/forum/#!topic/julia-users/LbFCCeMv76M

Since it's possible (and probable) for arithmetic on unums to produce ubounds, and ubounds are the objects that are closed under arithmetic, I think it might make sense to make...

I did read that part of the book. I might have misunderstood something, but I was under the impression that unify and smartunify always return either a single unum, or...