Le mardi 07 juillet 2015 à 02:29 -0700, Simon Byrne a écrit :
> Thanks everyone.
>
> A bit more context: I'm trying to implement callbacks in RCall.jl.
> There are multiple ways R objects could be converted to Julia ones
> (as singletons, vectors, Nullable singletons/vectors, dictionaries,
> etc.), so what I was thinking was:
> 1) pass the R objects (known as SEXPRECs) straight to the Julia
> method
> 2) the first time a Julia function is converted to an R object (via
> the sexp function), I wanted to define an initial generic method
> which would take SEXPRECs and perform a default conversion.
>
> function sexp(f::Function)
> if method_defined(f,(VarArgs{SEXPREC},))
> global f
> f(x::SEXPREC...)
> y = rcopy(...) # do default conversions
> f(y...)
> end
> end
> return callback(f) # construct R object for callback
> end
>
> In that way if a user wanted a different conversion, they could
> define their own method(s) to do this. But I don't want to overwrite
> this if it already exists.
>
> Any ideas?
Interesting design issue!
I think another solution could work, but I'm not sure it's really
better. Instead of checking whether
method_defined(f,(Vararg{SEXPREC},))
you could always define a fallback wrapper function like this:
function f(x::AbstractSEXPREC...)
y = rcopy(x...) # do default conversions
f(y...)
end
AbstractSEXPREC would just be an abstract type from which SEXPREC would
inherit, so that the fallback would not be called when a more specific
f(x::SEXPREC...) method exists.
Then, you would call f() on the VarArgs{SEXPREC}, and Julia would take
care of calling either the user-defined function or your fallback,
based on the standard dispatching rules.
As I said, your solution might be equally good.
My two cents
> -Simon
>
>
> On Monday, 6 July 2015 19:09:43 UTC+1, Mauro wrote:
> > If I understand correctly what you want, then have a look at
> > Traits.jl
> > where I check method signatures of trait-definitions against
> > methods of
> > a generic function. See the loop at:
> > https://github.com/mauro3/Traits.jl/blob/master/src/Traits.jl#L158
> > and
> > in particular isfitting.
> >
> > Turns out that this is a relatively hard problem (unless I made a
> > mess
> > of it), in particular once parametric methods are involved.
> > However,
> > your problem might be a bit easier than what Traits does as it test
> > for
> > equality. Let me know if you got questions.
> >
> > On Mon, 2015-07-06 at 19:25, Simon Byrne <[email protected]>
> > wrote:
> > > If I have a generic method foo, is there a way I can tell if a
> > particular
> > > signature has been defined?
> > >
> > > Note that I don't want method_exists (which simply determines if
> > something
> > > can be dispatched), I want to determine if a particular
> > definition has been
> > > made, e.g. if
> > >
> > > foo(x) = x
> > >
> > > then I want
> > >
> > > method_defined(foo,(Int,)) == false
> > > method_defined(foo,(Any,)) == true
> >
