Funny coincidence. I have been playing around with its REST API
recently. I was thinking of mirroring the R package (and thus have a
more or less identical interface), which is Apache licensed and is
really nice to use. I did, however, have two concerns that discouraged
me. My main concern is the rapid pace of changes of H2O. I think that
even if you'd had a fully implemented H2O package that it would be quite
some work to keep it up to date. The second, more minor concern is that
I don't see a way of implementing the package without binary
dependencies, since HttpParser.jl introduces one. But this is more of a
personal preference of avoiding binary dependencies
That being said, if you are really interested in doing this via the REST
API, then I am interested in contributing. I do think, though, that it
should be Apache licensed. Don't know if it makes a difference, but just
to be on the safe side.
On 2015-11-17 16:04, Randy Zwitch wrote:
I've been using H2O quite a bit at work, because it does a few things
well (I mostly use it for random forest and GBM) and is easy to use.
I talked with the company at length about creating a Julia package and
the company is supportive of open-source contributions, so I created a
stump of a package. Anyone interested in working on it with me? Right
now, I'm still in between about using PyCall for everything or
attacking the API directly. Anyone interested in helping can help me
derive a development plan...
https://github.com/randyzwitch/H2O.jl
