I started working on supporting ndarray inside Postgres before I came
across MADlib. My motivation was the idea that it should be easier to
perform computation local to the data where appropriate.
Currently the only type I've been formally working on is a simple
ndarray, though I've seen at least simple uses of a DataFrame work. I'm
part-way through adding SQL-level versions of basic ndarray functions.
I've got casts between ndarray and relevant Postgres base data types.
The storage is currently rather brain-dead (simply storing the version 0
pickle), but the Postgres ndarray type does have it's own version byte,
so this could be significantly more sophisticated in the future.
Even if this wouldn't be useful to MADlib, I'd love any feedback on what
it could be useful for. I suspect there's a lot that it could be useful
for, but I'm more data architect than data scientist.
--
Jim Nasby, Data Architect, Blue Treble Consulting, Austin TX
Experts in Analytics, Data Architecture and PostgreSQL
Data in Trouble? Get it in Treble! http://BlueTreble.com
