continuity in the development of breedR
Facundo Muñoz
Dear all,
I have been the main (and almost only) developer of breedR for a few years now. As some of you might know, I have switched jobs, from a postdoctoral position at INRA (where breedR was conceived by Leopoldo Sanchez) to a permanent position as a Biostatistician at Cirad.
I think it is fair to keep you informed regarding the prospects of breedR.
Cirad is a different research institution also concerned by
agricultural modeling and by quantitative genetics to some extent.
That means that my domains of interest here are close enough to
justify some investment of time on breedR.
On the other hand, I cannot consider involving myself directly in
the development of new functionalities. _But_ if someone decides
to invest resources in further development this person will have
all my support. This includes guidance on how breedR's code is
organized (there is only little documentation on this), guidance
on the workflows for development, testing and releasing and help
on where to start making changes. I am aware of some future plans
in this regard, but nothing concrete so far.
In summary, I will keep:
- giving support on the use of breedR by reading this mailing list, answering questions, fixing bugs and maintaining the code up to date with respect to PROGSF90, R and other dependencies
- fixing or improving some small usability issues which do not involve too much time
- improving the documentation, guides and use cases
- releasing new versions when necessary and keeping the repository functional
- giving technical support to whoever decides to invest resources
in further development
I hope breedR will continue to be a useful tool for you, and I look forward to hear from you soon.
Best wishes
ƒacu.-
Timothée Flutre
Facundo Muñoz
On 11/02/2017 08:25 AM, Timothée Flutre wrote:
> [...]
> a hindrance wrapping non-free softwares such as PROGSF90 would be on the
> long-term. What do you think about this? Are the developer(s) of
> PROGSF90 aware of breedR? Are they interested in building a community
> around it, as well as with developers/users wrapping it in other
> languages? (Stan is a great example in this respect: http://mc-stan.org/)
using closed-source software such as PROGSF90 comes at a cost.
For instance, that's the main reason breedR is not on CRAN. I have spent
weeks trying to work around it without success. I had to set up and
maintain a separate repository for the binaries for the different
platforms, I had to develop functions for downloading, installing,
checking versions, etc. I had to work around many inconsistencies from
the programs' output that would have been immediate to fix had I had
access to the code.
In the begining of the project I have been to the University of Georgia,
where I met I. Misztal, S. Tsuruta et al. I tried to convince them to
open-source at least some of their programs. But they were not willing
to invest the required resources for that.
On the plus side, PROGSF90 have been used in the animal breeding
community for many years. It's a robust and very well tested suite of
programs. Very efficient and capable of processing large volumes of data.
I must confess though, that I have not performed "stress" tests in order
to compare the performance of PROGSF90 against other possible solutions
from the R ecosystem.
breedR has been conceived from the beginning with PROGSF90 in mind and
some parts are tightly coupled. With time, I made an effort to better
separate the model specification and diagnosis parts from the model
fitting part. But there is more to be done in this respect. In my
dreams, I conceive breedR as almost independent from the "backend",
allowing to specify a model in breedR and fit it under a frequentist or
Bayesian setting with either PROGSF90, or INLA, or STAN, or whatever
comes later. But this requires a lot of work.
>
> * As you said, breedR started while you were working at INRA. Is the
> institution aware of you starting such a free-software project? For
> instance, it could be relevant if Leopoldo Sanchez let the Selgen
> meta-program, and the GA and BAP departments, know about the current
> status of breedR, given that the Inra is getting more and more involved
> in "open-data" aspects, and started to position itself with respect to
> free software (see
"Institutional awareness" is something fuzzy :). But in principle, yes.
Actually it was not me who started the project. I just have been hired
to develop it and when I arrived there was a roadmap in place already
which stated what kind of models were to be developed and that PROGSF90
were to be used as a backend.
I think Leopoldo has talked about breedR more than once within the
SelGen meta-program. Andrés Legarra is also aware of the project. In
fact, one of the trailing ideas was to integrate G3 into breedR.
At the very least, the whole INRA UR AGPF (where Leo works) is very
aware of breedR, which has been considered one of the major outcomes of
the European Project Trees4Future. I've been told there are plans for
future projects in the same line with interest in allocating funding for
improving and extending breedR.
I hope I clarified some of your questions. Don't hesitate to give me or
Leo a call/mail if you want to delve into more details.
ƒacu.-