Veit Heller

I just tried coming up with one, but failed. The best I can offer is this: ```clojure (load "Test.carp") (use Test) (load "[email protected]:carpentry-org/[email protected]") (def uri "scheme://usr:[email protected]:123/path?query=1#fragment") (deftest test (assert-equal test...

There’s also a problem using `System.get-arg`: ```clojure (defn main [] (for [i 0 (System.get-args-len)] (let [x (System.get-arg i)] ; this is inferred as dead (if (ends-with? x "Untitled") (println* x)...

Not at concretization time. Then, the job is to figure out which version of the function to use. If we were going for extensibility later, we would have to re-concretize...

This is an awesome idea! I think adding more malleability like this is always a good idea; I also don’t think that moving away from our hardocded solution would be...

But wouldn’t making the lookup order dynamic just mean taking all of the parts of the monadic pipeline and making their order dynamic, in pseudocode `reduce >>= lookup pipeline`? More...

I think I was still not quite clear about this: what I want to make scriptable is _the way in which we perform the lookup_, i.e. where to lookup when....

I think the default would likely not change—the code would, but it would be completely unimportant to the users, since its behaviors would hopefully stay the same. The tiny fraction...

A good example for the compiler being mysteriously slow (at least on my machine) is `tests/map.carp`. That test takes almost 20 seconds to compile.

I think originally the idea was to separate expansion and evaluation more, so that the top level of the evaluation knows whether the returned thing is the result of expansions...

We have a flag for that in `eval`, which tells the evaluator whether to prefer dynamic or static code. The REPL will prefer dynamic code; whether that’s the right behavior...