I have a solution for this question, but I'm unsure how to convert it to a train

Try it online!

@JoKing 14 bytes train

...which is one byte shorter than naive dfn

Just beat me to it :) (Sometimes people just read input with ⎕ like this: tio.run/##SyzI0U2pTMzJT///qG@qp/@jtgkGXI862tP@H56ub6h9eLuhto62/…)

@Bubbler Ah thanks, I was just looking at the ∘ operator. Pity I can't use monadic ÷ then

You can with +∘÷

^, which is still the same length

i need to get used to actually using APL practically. I vaguely know most of the operators, but I don't quite know how to organise things

A good place to start is Phase 1 problems of APL student competition.

2019 site still has online checker set up, I think

@Adám Do you have a plan to extend the language bar for Dyalog APL Extended?

@Bubbler Ah, that's an interesting idea.

Now I'm tackling this challenge, and my solution is 61 bytes now in Dyalog APL Extended

just 3 bytes away from beating CJam, but I can't find any more places to golf right now

Only if inverse accepted trains, or APL was lazy like Haskell...

Inverse does accept trains (?)

wat.

⎕←(*×⍟)⍣¯1⊢2

@H.PWiz
1.537201703

Unless it's something with extended

So it happens only in Extended, when a train is assigned to a name before applying inverse

@Bubbler @H.PWiz Some inverses don't work in Extended because ⍣ can't invert the dfn cover functions I inject. You can "un-extend" a primitive by prefixing it with a backtick.