Re: [Discuss-gnuradio] Sub-classing a block defined in an _impl class

[Jeff Long]

The good part is the separation of the API from the implementation. The 
not so good thing is that the implementations themselves often contain 
all the logic, usually because they are fairly simple. So, your choices 
are to subclass or copy/edit the implementation files. [Just saw 
Sylvain's not about phase_detector() being privite - OK].

Also, the implementations are free to change without affecting the API. 
Guaranteeing stability of implementations (even the _impl.h) makes the 
implementations part of the API, which would be bad. If there were a 
family of carrier tracking algorithms, it would make sense to have them 
subclass from a common abstract class.

In more complicated blocks, like some in the filter area, a good portion 
of the logic is separated out and reusable.

The

On 10/10/2014 06:25 AM, Johannes Demel wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hi Daniele,
>
> _impl classes are not meant to be subclassed usually.
> GR distinguishes between interface or public headers, which are
> located in the include directory and private headers which are located
> in the lib directory.
> You can just add new classes to your lib folder and use them within
> this folder. But if you want to export these classes and make them
> accessable from Python, you need a public header for it.
>
> On 10.10.2014 12:03, Daniele Nicolodi wrote:
> > Hello,
> >
> > this question my arise from my ignorance about C++, but I don't
> > understand how, if it is possible, to sub-class a GNURadio block
> > that has moat of his code in an _impl class.
> >
> > To my understanding I would need to subclass both the class
> > defining the block and his _impl class. However, it seems that the
> > _impl class is not part of the exposed API, therefore there is no
> > way to subclass it. There is way to achieve this without copying
> > around significant portions of code?
> The public description of a block is in the corresponding include
> folder. It is pure virtual to make sure it remains as an interface file.
>
> > The case at hand is subclassing gr::analog::pll_carriertracking_cc
> > changing the phase_detector() method to use the regular atan2()
> > instead of the fast version.
> You may just change this in GR directly. You have a personal copy of
> the code so there is nothing that should contradict this approach.
> Unless you want to compare the two but even then you should be able to
> accomplish that.
>
> I hope that gives you an idea of the structure and may help you solve
> your problem.
>
> Cheers
> Johannes
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1
>
> iQIcBAEBAgAGBQJUN7P9AAoJEO7fmkDsqywMpggP/0dQEXmkffTmzeg4cvW5xlmJ
> RWzT+eEA1prFMMR3iBSTrlPnLMyQfVz4nxE63AbBx+pEBsHzu6fUUHWZLtlzVwpy
> Bq89/Sd5G8sM40TWGKrFfUjxd5ND/4HuKar6JC/Wx9tnfZYCa3q3f3O2vi1S/6wu
> Ir6xKGpSGsR9Ro3RZ8iL+d/WoxAv86cH+GxFdo6R496hkqi1gZYdVkBJO1vCAL6S
> 8s2mvNDkSTcXaGcrF3VhhIBnjcwvhn3GYQMCjeSbcbyk++dLhZH+dmitB27lv6qW
> /u/eTNtXVGN2TzzQ0hLW4YI8pLGOj/cENHbBc31zUzRbRwt/ebOWC78bx9dyKALr
> WB0kDVM2XI9veYeyxMwvdtFncO7B7MGh+ocozhg53WaWijnFLGwD55yxZedXNcEr
> iIdq9Px7smsJ87s6ABgB7MfrhGgxT0ZJxop5D68HBLE4/v1OnRxjfi1V0JpKmIxt
> uueDgbKWGQEaDrgQX0BnrQ6bA68Z1HuTHk68j6WOnZ36a2mAZZfnZ8iKVckX/PJv
> CFhJLFtTQpfg3DR7lEN7x1gwbyyQGEk6w68sVK+e/LdPc5BHO9d4AcVCgfLaO/Ov
> i3QYb+Oqjv7P+jl2N2nAohh+AyNjwSRm2DxQGGwnvbhfChii5Cr5mIc+3y/JeoVf
> Vm7RMlaj1ESE9XAifvQC
> =z5WH
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> Discuss-gnuradio mailing list
> address@hidden
> https://lists.gnu.org/mailman/listinfo/discuss-gnuradio