@Alec Mihailovs I just have in mind not to write down all these combinations (ignoring special stuff like 'struct') and the WRAPPER seems to do it - however for each input from Maple, no? Perhaps there is some voodoo in C++, how one could avoid to write down the signatures explicitely and 'derive' it from input donw in Maple (but it is above my head).

@Alec Mihailovs Thx, I did not know it has a name.

Can one use Maple to generate the signatures using define_external(...,WRAPPER,...) ?

Well ... I have given up ... and reduced my activity

Yes, one needs the signature of the function to be called by that address.

It is some years ago, that I tried to learn and play with that. However it was all
'manually', http://axelvogt.de/axalom/maple/maple-external/index.html,
see especially for cplx_GSL.

To have something better one could try to parse the input of 'define_external'
to read off the desired signature needed for a C program.

But that is beyond my abilities, my coding experience is quite limited.

This is puzzling :-( ... and I keep to the initial Digits=10.

EE := 1/(sqrt(convert(0.8750001, rational,exact) - 1/invee)) - 3162;
``;
eval(EE, [x=21/10,y=24/10]); # 1/invee = x/y ---> 7/8
evalf(%);

                                 1/2
                          1000 10    - 3162

                               0.277660

Now sqrt(10) ~ 3 and to have 10 decimal places I would need those 4
leading ones + 10 trailing ones ( + 1) to avoid the cancellation in
the substraction (i.e. it is the fractional part of 1/sqrt(...) ):

1000*sqrt(10) - 3162; evalf[4+10+1](%);

                            0.27766016838

I would work with rationals, if being aware of the cancellation in *advance*.

But without that ... please continue that story!

@Alec Mihailovs 

I see ... I think I had similar if trying to use complex versions from GSL.

Does the following shed some light on the problem?

Id:=define_external('gsl_permutation_calloc',
 'MAPLE', # wrong, of course
     'n'::integer[4],
     'RETURN'::integer[4],
     'LIB'="gsl.dll"):

Id(3);
                    System error, , "PopErrorProc"

I am not sure which version I have and usually have something like

  BesselK1_GSL := define_external(
   'gsl_sf_bessel_K1',
   'C',
   't'::float[8],
   'RETURN'::float[8],
   LIB="gsl.DLL"):

  BesselK1_GSL(1.23);

                         0.414744832850106770

For 'gsl_permutation_calloc' the documentation says, that
it returns a structure - but then return = integer is not right
(and it crashes for me as well).

If I remember correctly one can use the DLL from GSL without wrapping, if input and output are just 1 variable (but have no Maple installed here right now).

For PARI: that was a by-product: Karim Belabas (the maintainer of PARI) once helped me to compile PARI with MS (or gcc) in a native way on Windows. From that one go a step further and provide a 'raw' interfacing to Maple.

may be he means http://en.wikipedia.org/wiki/Modern_portfolio_theory#The_efficient_frontier_with_no_risk-free_asset

may be he means http://en.wikipedia.org/wiki/Modern_portfolio_theory#The_efficient_frontier_with_no_risk-free_asset

That was quite quickly!

But is there really any need to display it in italics? Otherwise it would be much better to read ...

oh ... that would be worth

1. a new thread
2. a beer for me (it means, that some of my code would not work as such)

PS: klicking on "reply to this" (=Alejandro's comment) and being asked to log in does not give a reply to this ... great.

oh ... that would be worth

1. a new thread
2. a beer for me (it means, that some of my code would not work as such)

PS: klicking on "reply to this" (=Alejandro's comment) and being asked to log in does not give a reply to this ... great.

Is that by using 'Standard' sheets and Maple 14 at least?

PS - the XML issue: our admin solved it for MS OpenWebAcess, it can be configured for such (I can ask him, it would help you and roughly - IRIC - it is some issue with MS Share Point [not sure, if it would help you with your ISP, but may be you have a chance at your workplace])

Yes, a hovering would be helpfull.

Though it would be much better to bury those concepts and their ambiguities ...
(=ceterum censeo ...)