Should random primitives have APL/BQN-like symbols? about bqn

[mlochbaum]

You're right that most APLs use ? for random number generation. I removed it in BQN and instead have defined system values to deal with random numbers, which are currently supported in CBQN. For example the common APL pattern ?4⍴6 would be written 4 •rand.Range 6.

The biggest reason not to make ? a primitive is that pseudorandom generation is stateful, and primitives in BQN are always stateless. Because BQN has the function •MakeRand to create a generator, it gives you much better control over the state, while APL's ⎕RL function doesn't scale well. With •MakeRand different parts of a program can have their own non-interfering generators. Depending on what it does, a tool might always use its own generator (possibly recording the seed), require the user to supply one, or use its own by default but allow the caller to override it. The fact that it's a namespace also allows us to easily add all sorts of distributions and special functions without having to make new primitives.

For areas where random generation is a central part of programming, •ReBQN (draft spec) would allow you to extend BQN with appropriate primitives. Something like the following. But right now only CBQN has •MakeRand and only the online REPL supports primitives⇐, so you can't run this code anywhere yet.

rand ← •MakeRand 1  # Initialize
bqn_rand ← •ReBQN {primitives⇐•primitives∾⟨⟨'?',rand.Range⚇0⊘rand.Deal⟩⟩}
BQN_rand "3?5"

[mlochbaum]

If you get stuck on another primitive, the BQN-Dyalog dictionary has most of them. For ? it does point you to •MakeRand, although •rand would be more accurate now that it exists (added to CBQN three days ago).

[mlochbaum]

Added •rand to the online REPL, so defining ? now works. •MakeRand would be harder because it can't use Math.random(), as that can't be initialized.

[vegabook]

Thanks understood. Makes sense. I'l continue the second few steps into my BQN adventure!

[vegabook]

Yeah giving this some more thought it does seem correct that any "primitive" that relies on a seed is no longer a primitive. Therefore this choice actually makes BQN purer in a mathematical sense.