Proper ccalls with variadic arguments.

[rschwarz]

We need to call SCIPexprCreate which has variadic arguments. This is supported by Julia's ccall, but only for variadic arguments of the same type.

This was already discussed in #100 and on Discourse.

The suggested solution is to create a thin C wrapper, which exposes function without variadic arguments. Unfortunately, we still need those for SCIP_EXPR_SUM and SCIP_EXPR_PRODUCT.

[rschwarz]

See gist for an incomplete solution sketch.

[rschwarz]

It seems that in CSIP, we were just passing a pointer to an array as the "last" argument to SCIPexprCreate. I did not know that this is valid.

I guess usually, this array needs to be 0-terminated or something, but in this case of n-ary operators, the number of children is passed explicitly, so it might work out.

[bhalonen]

look into CBinding for a solution to this problem: https://github.com/analytech-solutions/CBinding.jl

[rschwarz]

Thanks for the hint. I understand that the issue stems from a limitation of Julia's ccall, so I would be surprised to find a solution in a Julia package, but you never know...

[bhalonen]

julia> func = Cfunction{Cint, Tuple{Cstring, Vararg}}(lib, :printf)    # int (*func)(char *, ...) = dlsym(lib, "printf");
Ptr{Cfunction{Nothing,Tuple{Cstring,Vararg{Any,N} where N}}} @0x000061eefc388930

julia> func("%s i%c %ld great demo of CBinding.jl v%3.1lf%c\n", "this", 's', 1, 0.1, '!')
this is 1 great demo of CBinding.jl v0.1!

[bhalonen]

He actually has it working, and is adding functionality over ccall. I really don't understand how it works, to be totally honest.

Scroll to the bottom of the README.md

[bhalonen]

It is some manual memory mapping onto C types. Very impressive.

[rschwarz]

OK, great, so it does support variadic arguments of mixed types properly 😮