Matrix as subclasses of TokenBase and dynamic lua_push* in binding about april-ann HOT 5 CLOSED

I support this change.

When I started working with tokens, I realized they were just containers of matrixes... Not matrixes themselves. I think it would be more intuitive for someone to use tokens as matrixes than not accessing to the matrixes in the tokens.

from april-ann.

That's the point, so, I count one more in favour :-)

from april-ann.

I have some concerns about deriving Matrix from Token. It is basically that Tokens are inspired by dataflow architectures, where data is interchanged between different modules by passing Tokens trough special channel connections. The problem, in my opinion, is that we don't need to be forced to use wrappers of data types. It can be better to allow data types to be serialized over channels in some kind of binary or textual representation. It is a really important thing, we need a better serialization/deserialization scheme in APRIL-ANN.
I will try to implement, in a first a attempt, a simple binding function which dynamically un-wraps tokens into Lua, and tokens will be ignored from the point of view of Lua operations. This will simplify the glue between C/C++ and Lua regarding to Tokens stuff.

from april-ann.

Ok, automatic unwrap of Tokens before push them into Lua is ready. Stills being a question if it would be better to derive Matrix from TokenBase, instead of wrap Matrix into a Token. This discussion will be important when dataflow architecture shall be developed.

from april-ann.

Closing this issue, it can be reopened in the future with the delivering of dataflow stuff into the public repository.

from april-ann.