Should Encoder default to vector behaviour?

[Johan Rosenkilde]

Right now, it is rather tedious to write an Encoder: one needs to define encode, unencode, message_space, and generator_matrix.

We could take the stand that most encoders will be vector-based. If we did so, there could be standard implementations of encode, unencode, and message_space, and subclasses have only to implement generator_matrix.

This doesn't reduce flexibility, since if you have a different message space, e.g. F[x], then you just override all these default implementaitons.

[Johan Rosenkilde]

Also, generator_matrix makes no sense for encoders whise message space is not F^k. So in that design sense, we already defaulted somewhat to vector behaviour.

[David Lucas]

Partially resolved issue #66 : encode() and message_space() now have default impl. in encoder.py

→ <<cset bceaa3447483>>

[David Lucas]

Problem with auto-resolve system in Bitbucket : partially resolved is understood as resolved... So, partially resolved in bceaa34