Hello, just a curiosity: after the latest major graph-tool update, I see the new class WeightedBlockState to be used in place of BlockState when edge covariates are used. What are the motivations for this choice (except for clarity)? Are there specific advantages in WeightedBlockState over the previous BlockState with rec_types?
This choice simplifies the code substantially, since the standard blockmodel makes no references to weights at all. The new class only extends the base when needed.
Besides the greater simplicity of implementation, this has no effect for the user: the new WeightedBlockState is fully equivalent to the previous BlockState.