Support for sparse edge covariates?

[krivit]

Currently, edgecov() and dyadcov() expect n*n matrices, but it should be straightforward to also implement support for sorted edge lists and perhaps even RLE-encoded dyad matrices (RLEBDM for binary and perhaps a new type for non-binary).

Would that be useful?

[mbojan]

Would a sparse matrix object already go most of the way to solve the problem?

[krivit]

Some of the way; the term C code would need to be able to interpret them as well.