Context: we started implementing the "naive" pagerank algorithm because it was very easy to implement. This algorithm requires inverting a matrix and the way to do this in ndarray is by using the ndarray_linalg crate which piggybacks on Lapack & OpenBlas. These two requires external libraries to be provisioned, for @kim 's joy.
However, this algorithm was flawed: for big matrixes was yielding negative pageranks, and we later transitioned to an iterative (and equally simple) version that doesn't require matrix inversion.
This PR kills the pagerank_naive algorithm and all the library baggage associated with it.
Context: we started implementing the "naive" pagerank algorithm because it was very easy to implement. This algorithm requires inverting a matrix and the way to do this in
ndarray
is by using thendarray_linalg
crate which piggybacks on Lapack & OpenBlas. These two requires external libraries to be provisioned, for @kim 's joy.However, this algorithm was flawed: for big matrixes was yielding negative pageranks, and we later transitioned to an iterative (and equally simple) version that doesn't require matrix inversion.
This PR kills the
pagerank_naive
algorithm and all the library baggage associated with it.