Cool article. The first time I've seen linear algebra done in RDBMS was with MadLib, a PostgreSQL extension which implements a lot of these operations https://madlib.apache.org/docs/latest/group__grp__matrix.htm...

So all the operations described in the article, and more are implemented here https://github.com/apache/madlib/tree/master/src/ports/postg...

I'm not sure why Madlib is not mentioned in that CACM article

Madlib is mentioned in the Related Work section of the PDF article.

Lot of interesting ideas to move computation close to where the data reside can be found in this paper: http://www.cs.ox.ac.uk/dan.olteanu/papers/soc-sigmod16.pdf this is an example of how linear regression can be implemented which is intimately related to the article.

Check out RIOT[0][1] for R if you think this is cool.

[0]: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.157...

[1]:https://github.com/riot-xx/riotdb

PDF version of the paper: http://cmj4.web.rice.edu/mat_vec.pdf

I'm convinced this is ultimately going to happen.

I liken it to how we used to test on how to sort and merge data, with the answer being that it is always a trade off. Which is exactly what an execution plan is. More, index projections is half of the magic developers relearn every year.

To that end, getting linear algebra primitives easily accessed in a query language feels like it will be a game changer.