OHDSI Home | Forums | Wiki | Github

Replacing ff with RSQLite?

(Vojtech Huser) #21

Is there a work group around OHDSI R package development

Currently I don’t think there is any such group.

The idea of creating an R user work group is a great idea (in my view). It would not be about any specific package, but about the universe of them. E.g., OhdsiRTools could be further developed. In my view, having more R coders involved with development might be good…

(Martijn Schuemie) #22

There is no formal R developer group (as @Adam_Black asked), and there is also no R user group (as @Vojtech_Huser suggested). In our current organization of open source software development there is a lead for each R package. Mechanisms for others to contribute are the usual: creating issues, creating pull requests, etc., and of course participating in forum discussions such as these.

I prefer to keep it ‘simple’, with a single backend. If we have something that works as ff was intended to work, it would work for everyone. If we have to write our code considering all the different implementations we would be supporting, we just made our life a lot more difficult.

(Mark Seal) #23

I like this idea. I have over a decade of enterprise experience in Java/C++/C#/Python, but I know just enough about R to shoot myself in the foot. I would love to lurk in said user group to learn the design principles of the tools.

(Adam Black) #24

This makes sense. Sounds like RSQLite might be the way to go then.

I would like to contribute but would need to start with something small. I still have a ways to go to understand the OHDSI R ecosystem. I also just have some basic questions like the one I posted here. Is it possible to use odbc instead of jdbc with the OHDSI tools?

(Martijn Schuemie) #25

I just tried to release a new version of DatabaseConnector to CRAN, and it was refused because it depends on an orphaned package (bit).

So I guess we’ll have to speed this up.

(Seng Chan You) #26

Since FeatureExtraction package is the one of the most fundamental packages in OHDSI, if we need to revise, we should do ASAP, even though we’re distracted for the COVID-19 now.

(Martijn Schuemie) #27

Well, we can work without CRAN, it is just annoying.

(Seng Chan You) #28

@schuemie As you know, revision of FeatureExtraction is highly related with reproducibility. And, I think we need to escape from ff. If so, we should do it now, until more packages are developed based on ff in OHDSI, such as clinical characterization package for the covid-19.

(Jreps) #29

One thing I just thought while reading this thread - when we change from ffdf we need to make sure we still have a way to load old data that was saved in ffdf files otherwise we won’t be able to go back easily to view old study data.

(Seng Chan You) #30

@jreps Indeed. I don’t think we can maintain the reproducibility before and after the revision of FeatureExtraction package, and that’s why we need to start to revise it now…

(Martijn Schuemie) #31

We always need to make sure to have the exact same versions of the packages to reproduce a result. This is no different from all other changes we’ve made to our software over the years.

I agree we need to work on this sooner rather than later, but it is not a must-have for any evidence generation we do in the short term.