Decoders.connected_encoder not done anywhere

[Johan Rosenkilde]

No description provided.

[David Lucas]

First proposal in commit : 54cb3e0

For now, it looks for the specified key (hardcoded : "EvaluationVector" for GRS codes and "GeneratorMatrix" for linear codes) of the _cached_encoders dictionnary, and returns the associated encoder.

Of course, if the user creates the decoder before the encoder, and tries to call connected_encoder, it returns a KeyError. I think it makes sense to consider that you will always try to encode before decoding, especially when it comes to connected encoders. But I can change the code in a way that, if I attempt to access the connected_encoder before it was created, I create it, cache it and return it.

By the way, I used a direct call to code._cached_encoders for now, but I'd prefer use a helper function to access _cached_encoders. Because cached_encoders() does not sound as a nice name (to me), can you imagine something better?

[Johan Rosenkilde]

Why don't you just call self.code.encoder(name=self._connected_encoder_name)? Then the check and possible caching will be done for you.

It is important that the connected encoder can be returned without the user having created it, since he might have encoded using a different encoder, but decodes using this. For instance, this is relevant for the Gao GRS decoder, since decode_to_code will itself call decode_to_message and use the connected encoder to retrieve the corresponding codeword. This is because Gao naturally retrieves the message polynomial f.