Trouble deducing type parameter with inner constructor

[Magnus Lie Hetland]

What is the right approach if I want (1) to use an inner constructor, which does some modification/normalization to the arguments, and (2) I want proper type parameter deduction from the arguments? Do I need to write a separate function or something?

I'm not even sure what's going on here (i.e., why it doesn't work with the inner constructor) – or how I could make it work with the outer constructor…

immutable A{N}

    x::NTuple{N, Int}

    y::Int

    A(x, y) = new(x, y + 1)

end

#a = A((1, 2), 3)       # Doesn't work

a = A{2}((1, 2), 3)     # Works (w/cumbersome param)

println(a)

abstract C

immutable B{N} 

    x::NTuple{N, Int}

    y::Int

end

#B(x, y) = B(x, y + 1)  # Stack overflow...

b = B((1, 2), 3)        # Works (w/wrong answer)

println(b)

What am I doing wrong?

[Magnus Lie Hetland]

OK, this works:

immutable A{N}

    x::NTuple{N, Int}

    y::Int

    A(x, y) = new(x, y + 1)

end

A{N}(x::NTuple{N, Int}, y::Int) = A{N}(x, y)

But why is this necessary? (Adding type decl. to the internal constructor doesn't seem to do anything useful.) And why is it not necessary if I don't have an internal constructor?

(Oh, by the way: Please ignore the abstract C in the previous example ;-))

[Tobias Knopp]

I don't really know if this is a bug or if there is some good technical reason for this.
What seems to happen is that the definition

A(x, y) = new(x, y + 1)

removes the default constructor. You can see this by typing "A" on the REPL (or methods(A))

[Magnus Lie Hetland]

Hm, yes. Is this somehow related to the default constructor stuff added in 0.3 (linked to issues #4026 and #7071)? Not that I see exactly how they'd be related, except that they're linked to the same topic :-} Could it be because the default argument uses Any for its argument types, and me adding a constructor with exactly that signature prohibits it from being constructed (reasonable enough)? But even so, not figuring out the type parameter seems like it might be a bug…? (Or maybe it's just a feasible specific case of an infeasible general problem that I haven't noticed?)

[Tobias Knopp]

Magnus,

Please feel free to file a new issue. Even if it is intentional it might have to be better documented.

Cheers,

Tobi

[Magnus Lie Hetland]

Done.