Documentation for developers

[brainstorm]

Hello NIDASH TF!

We are sitting here with Tristan Glatard and Joris Slob and spent some time trying to figure out how NIDM could be integrated with CBRAIN during the Hackathon in Leiden:

http://wiki.incf.org/mediawiki/index.php/Hackathons/Leiden-2014

Apparently the developer docs/intro link is broken:

https://github.com/incf-nidash/nidm/wiki/Getting-Started-with-NI-DM

And while some of the notebooks provide some information we couldn't yet figure out what are the steps required to export information from CBRAIN database to the NI-DM prov model.

We would like to export the information related to a set of jobs ran in CBRAIN:

1. Executed jobs and software.
2. Input data.
3. Derived data.

We would be happy to have some feedback on this.

@glatard, @jorisslob

[satra]

you might want to first look at this:

https://github.com/nipy/nipype/blob/master/nipype/utils/provenance.py

that was an initial pass at provenance in nipype - there is a current PR where a few tweaks have been made and then all of this will need to be updated to the current prov 1.x library.

[brainstorm]

Thanks @satra! Yes, we saw that one and later on we went through the following:

https://github.com/incf-nidash/nidm-results_fsl/blob/master/NIDMStat.py

Plus some other NIDM people walked into the room and briefed us on the efforts to have a simplified developer example(s). Right now there are many notebooks and really good specs but few central, concise examples for developers.

[dbkeator]

@satra Couple things we discussed in Leiden (more obviously on our next NIDM call on Monday), we decided we'd work with the e-phys TF on workflow representations using NIDM. I mentioned nipype had a draft workflow description. Also, there is some difficulty finding nidm.nidash.org with google unless one types "Neuroimaging Data Model" as the search term. I submitted nidm.nidash.org to google but I think we need to add some keywords/metadata and possibly schema.org markups to make the website more accessible. Lastly we discussed creating a "Getting Started with NIDM" specification (probably the primer document) which essentially explains how to get started in modeling your data with PROV and NIDM.

[nicholsn]

@dbkeator, I think the primer is more of a conceptual overview that shouldn't go into enough depth to start modeling. How about something like the best practices on provenance: http://www.w3.org/International/its/wiki/Provenance_Best_Practice

[satra]

@brainstorm - the nidm stat model is very specific to the results representation while the nipype provenance is more about how to encode workflow provenance using PROV-DM. you should definitely read the pointer that @nicholsn provided.

@dbkeator - thanks for the updates. i agree with nolan that we don't want to get into general data modeling. there are plenty of very useful resources (that we should point to). i think our primer should highlight nidm specific concepts and decisions we have taken.