It seems that your suggestion using ccall from within julia to play around
with the c-api provides some info. Running your ccall command on a couple
of "External Packages", i.e. DSP and HDF5  returns the Null pointer.

pkg_ptr = ccall(:jl_eval_string,Ptr{Void},(Ptr{Cchar},),"DSP.Pkg.hanning")
Ptr{Void} @0x0000000000000000


pkg_ptr = ccall(:jl_eval_string,Ptr{Void},(Ptr{Cchar},),"Base.Pkg.sin")
Ptr{Void} @0x0000000002a431f0

pkg_ptr = ccall(:jl_eval_string,Ptr{Void},(Ptr{Cchar},),"Base.Pkg.cos")
Ptr{Void} @0x00000000032acfa0

pkg_ptr = ccall(:jl_eval_string,Ptr{Void},(Ptr{Cchar},),"HDF5.Pkg.h5open")
Ptr{Void} @0x0000000000000000


So, it seems we are not able to get hold of the external packages from
within C. Any ideas?

Best,

Einar


On Fri, Sep 5, 2014 at 5:46 PM, Jake Bolewski <jakebolew...@gmail.com>
wrote:

> It's jl_eval_string located in jl_api.c in src.
>
> so you would do
> jl_value_t * func2 = jl_eval_string("DSP.hanning")
>
> The best way to play around with Julia's c-api is within julia itself.
> julia> pkg_ptr = ccall(:jl_eval_string, Ptr{Void}, (Ptr{Cchar},),
> "Base.Pkg.clone")
> Ptr{Void} @0x00007fd11c1754a0
>
> julia> unsafe_pointer_to_objref(ans)
> clone (generic function with 2 methods)
>
> julia> typeof(ans)
> Function
>
> That way you can prototype what you want much more easily.
>
> On Friday, September 5, 2014 9:32:43 AM UTC-4, Einar Otnes wrote:
>>
>> Thank you for your help on this. It seems that the 'jl_eval_global_var'
>> function is local as I got the error "undefined reference to
>> `jl_eval_global_var'" when linking. I replaced
>> your suggestion with:
>>
>>     jl_module_t* jl_dsp_module = (jl_module_t*)
>> jl_get_binding(jl_main_module, jl_symbol("DSP"));
>>     jl_function_t* func2 = jl_get_function(jl_dsp_module,"hanning");
>>
>> This compiles OK, but I get a "segfault  (core dumped)" when running the
>> line with 'jl_get_function' .
>>
>> Any ideas or thoughts?
>>
>> Thanks,
>>
>> Einar
>>
>>
>>
>>
>>
>>
>>
>>
>> On Fri, Sep 5, 2014 at 11:15 AM, Ivar Nesje <iva...@gmail.com> wrote:
>>
>>> I would guess that something like
>>>
>>> module = jl_eval_global_var
>>> <https://github.com/JuliaLang/julia/blob/6277015ee3d46f20149136d092525bec95b6e29d/src/julia.h#L917>
>>> (jl_main_module, jl_sym
>>> <https://github.com/JuliaLang/julia/blob/6277015ee3d46f20149136d092525bec95b6e29d/src/julia.h#L666>
>>> ("MyModule"))
>>>
>>> would work, but I don't have the required testing setup to see if it
>>> actually works. (Where is my C REPL?) The embedding API has not gotten much
>>> attention (yet), so it is mostly documented in source, and it is likely
>>> that there will be some adjustments.
>>>
>>>
>>> On Friday, September 5, 2014 10:09:03 AM UTC+2, Einar Otnes wrote:
>>>>
>>>> Sorry, I'm a bit slow. How do I look up a binding for a specific
>>>> module? In other words, how would I explicitly get to call the "fftfreq"
>>>> function in the "DSP" module from C/C++?
>>>>
>>>> Is this this documented anywhere in Julia docs?
>>>>
>>>> Thanks,
>>>> Einar
>>>>
>>>>
>>>>
>>>> On Wednesday, September 3, 2014 3:10:59 PM UTC+2, Isaiah wrote:
>>>>>
>>>>> If you have defined a module already (by eval'ing julia code?) then
>>>>> you can look up the binding and cast that to a jl_module_t.
>>>>> On Sep 3, 2014 8:14 AM, "Einar Otnes" <eot...@gmail.com> wrote:
>>>>>
>>>>>> Dear experts,
>>>>>>
>>>>>> I've looking at the documentation " Embedding Julia" (
>>>>>> http://julia.readthedocs.org/en/latest/manual/embedding/) to figure
>>>>>> out how I can call my own julia functions from within C, and I'm 
>>>>>> struggling
>>>>>> to figure out how I should define the jl_module_t that corresponds the
>>>>>> module I've defined. The examples show that there is an instance of
>>>>>> jl_module_t,  "jl_base_module",  that you need to provide to be able to
>>>>>> call a function defined in the base module. How do I define a 
>>>>>> corresponding
>>>>>> jl_module_t type for the modules that are defined outside of standard
>>>>>> julia, e.g. for the external packages or the modules I have created 
>>>>>> myself?
>>>>>>
>>>>>> Thanks for your help.
>>>>>>
>>>>>> Best regards,
>>>>>>
>>>>>> Einar Otnes
>>>>>>
>>>>>>
>>

Reply via email to