[FFmpeg-devel] [PATCH] Added the interface for the Turing codec

[Alexander Strasser]

On 2016-11-19 14:29 +0100, Moritz Barsnick wrote:
> On Fri, Nov 18, 2016 at 20:15:30 +0100, wm4 wrote:
> > This lib has a really weird API...
> 
> I can't judge the algorithms or the development direction and stuff -
> I'm sure it's all state of the art. But the API sure gives me
> goosebumps, and I don't mean the sexy feelgood type.
> 
> Sure, it has all the flexibility, but also almost all the disadvantages
> of calling an external program. This makes it look so much like a proof
> of concept.

  I looked at the source for fun yesterday to find out what
that was about.

  IMHO it is that the real encoder class takes C++ arguments 
(boost variables map) and they have a CLI program which also
needs to create that variables map for the encoder. Thus it
seems the quickest way to make a C interface was to use that
argument parsing code for initializing the encoder.

  I hope it copes well with multiple definitions of the same
option, like the last overwriting the previous ones.


[...]


  Alexander