Hi all,
If I can chime in with our 2 cents.
Following the SWG meeting in Trondheim it was great to see the community get behind the need for a standardised data api. Based on what I know of IDataset (and the way it matches numpy) I am convinced that is the way forward. However, I do see there is work to be done to get there.
The issues we see in particular are:
1. There are precompiled jars in the current dataset plugin - this must be resolved so it can be all source.
2. There are dependencies on non-orbit approved 3rd party packages- we started discussing in Trondheim how some of these can be resolve.
3. There are 2 plugins & the api plugin comes with lots of other (non-relevant) apis
4. The code has not been IP-review and established as IP clean.
These are all very surmountable, and in fact we (kichwa coders) would like to go ahead and do something about it.
Jonah is currently working on a fork of IDataset, which will be a subset of the current functionality available in the DAWNSci plug-ins. We need to work with the folks at Diamond about the best way to structure this so it can be reintegrated with DAWNSci & other Diamond projects, but still be re-usable for SWG.
The goal will be to get this into a stage where it can be easily IP-checked and meets this criteria
- available in an Eclipse build infrastructure
- has a reproducible build from source
- IP checked & clean
- generates a p2 site
- can be reusued in other projects in different formats (jars, oomph installers, the whole works)
Where it is made available is not nearly as relevant as that it is made relevant (but FWIW, we would favour a top-level science project).
We'd like to work with all in the community to push this forward so will send a pointer to the fork by the end of the week and welcome comments.