I am following your discussions on the GraphBLAS email list. Are you interested in the distributed-memory API of GraphBLAS (for clusters, typically over MPI) or can you make use of it if it were available (you may know, CombBLAS already has a distributed-memory implementation)?
Since you are mentioning Redis Graph and I'm not really into academic side of graphs, I want to ask - how do you use it from the user perspective, who wants to do some simple graph traversals using Cypher or some equivalent? For Redis Graph you have some nice higher-level API, is something similar planned for pggraphblas or outside of scope for this project?
Has anyone done a similar thing with Spark and Postgresql for large graphs ? I'm curious to know how you write and structure your algorithms, given that graph partitioning is quite tricky (and not possible for a lot of algorithms).<p>I'm guessing Google uses a variant of this at scale - but not sure if their label propagation algorithms can be implemented on something like Spark and Postgres.
A slightly off topic question regarding the mxm (matrix times matrix), mxv (matrix times vector), and vxm (vector times matrix) functions. AFAIK, Postgres supports function overloading. With the new types of matrix and vector, would it be easier to have a single function for multiplication, than three separate ones?<p>Also, can we overload operators in Postgres?
I found this project on Github a few weeks ago and have been quietly following it.<p>It looks really cool but I didn’t get from the readme how I would best take my relational tables and make a set of matrices from them.<p>I’ll take another look this weekend