Arc Forumnew | comments | leaders | submitlogin
1 point by Darmani 6046 days ago | link | parent

Actually, almkglor's code would (I'm pretty sure) have worked with just a normal with; function definitions are unevaluated at definition, so that (fn (x) (afergergghersgergferg x)) would not give "Error: afergergghersgergferg is not a function" until it's called. If a defined a function called afergergghersgergferg in the meantime, as with a normal with statement, it should have worked.

Anyway, you're not alone in being surprised that x.y.z expands to (x y z) instead of (x (y z)).



2 points by absz 6046 days ago | link

Not quite. It works in the toplevel because doing (def name args ...) is effectively to (= name (fn args ...)); this establishes a global, not a lexical, binding. Using let/with/withs adds a lexical binding, and since functions (or "closures") close over their scopes, they can't get at functions defined at the same time. Thus, you need to use withs. Observe:

  arc> (with (foo (fn (x) (* 2 x)) bar (fn (y) (foo y)))
         (prn (foo 10))
         (prn (bar 10))
         t)
  20
  Error: "reference to undefined identifier: __foo"
  arc> (withs (foo (fn (x) (* 2 x)) bar (fn (y) (foo y)))
         (prn (foo 10))
         (prn (bar 10))
         t)
  20
  20
  t

-----

1 point by bOR_ 6046 days ago | link

That implies that at the top level, the order in which I define my functions (which use each other) doesn't matter..

nice! Then I can group them in a way that is more sensible for humans.

Anyway, the program is now far enough that I've a critter running around the field and filling its memory with associations between the things it encounters. Next stop is to let it learn what is edible and to let it steer itself based on its observations :).

-----