It could be an idea, yeah. This should also come with the potential of
reducing the number of methods exposed via swig.
On Sun, 30 Jun 2019 at 14:41, Ahmed Essam <theartful...@gmail.com> wrote:
> The main use case I had in mind was making Distribution a base class.
> Methods like:
> float64_t CHMM::best_path(int32_t dimension);
> SGVector<float64_t> CGMM::cluster(SGVector<float64_t> point);
> can be translated easily, and still fall within the meaning of what a
> "get" method does.
> But on a second thought, I think I might have been a little rash/lazy ":D
>
> On Sun, Jun 30, 2019 at 1:33 PM Heiko Strathmann <
> heiko.strathm...@gmail.com> wrote:
>
>> gmm.get_real_vector("cluster", point)
>>
>> Things like these should be done via a base interface, here simply :apply
>>
>> On Sun, 30 Jun 2019 at 13:31, Heiko Strathmann <
>> heiko.strathm...@gmail.com> wrote:
>>
>>> We could template the arguments. But I am unsure this really is a good
>>> idea with the functions. Where would be the use case where the getter
>>> doesn’t work?
>>>
>>> On Sat, 29 Jun 2019 at 20:26, Ahmed Essam <theartful...@gmail.com>
>>> wrote:
>>>
>>>> Hello all,
>>>>
>>>> Do you think it's worth considering extending the behaviour of
>>>> get(name, index) to try to run a function that accepts an integer
>>>> parameter.
>>>> So for example we can run hmm.get("best_path", i) to call best_path
>>>> with i as a parameter.
>>>>
>>>> I believe we can also extend this to accept any number of parameters of
>>>> the same type by having something like get(std::string name, std::vector<T>
>>>> params). So in python we might do something like
>>>> get_sometype("some_lazy_parameter", [1, 2, 3, 4]). This can be done using
>>>> C++ function binding magic. So in GMM for example we might do
>>>> gmm.get_real_vector("cluster", point). And this can be made as fancy as we
>>>> like (to specify parameter names and having default values for example).
>>>>
>>>> Thanks,
>>>> Ahmed Essam.
>>>>
>>> --
>>> Sent from my phone
>>>
>> --
>> Sent from my phone
>>
>
