[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [O] Rsquared for reproductible research
From: |
Eric Schulte |
Subject: |
Re: [O] Rsquared for reproductible research |
Date: |
Thu, 12 Jan 2012 13:33:08 -0700 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/24.0.92 (gnu/linux) |
brian powell <address@hidden> writes:
>>> Also, I very much agree that a "near exact replica" of the http://
>>> rsquared.stat.uni-muenchen.de/index.rhtml for OrgMode would be great.
>> Yes! Any takers?!?
>>
> ...
> Eric questioned:
> "From looking at the fairly terse web site for R^2 it is not clear to me
> exactly what the system includes (I'm sure I'm missing something
> obvious). It seems to be the addition of a packaging system over-top of
> R source files. What would a potential Org-mode based system provide
> which is not already possible with Org-mode text files, Org-mode
> publishing and a version control repository."
> ...
>
> * I mostly agree with your statements. Good challenges. I did more
> investigation: This link to the paper that "Friedrich Leischa, , Manuel
> Eugsterb, Torsten Hothornb" put together may make things clearer--this
> paper really seems to be the justification/impetus for the R^2 website--it
> has made things clearer and more exciting for me:
>
> http://www.sciencedirect.com/science/article/pii/S1877050911001232
>
Ah, thank you for linking to this paper. It seems I was missing was the
package-management aspect provided by R2 through CRAN. The instillation
of all software dependencies is a huge benefit exactly as installing
software with apt-get or pacman is simpler than running "./configure &&
make" and manually resolving dependencies.
While such a tool makes sense for a single language system like R, I
fear an Org-mode version of such a system would have too wide of a
scope. Given that code blocks may contain arbitrary languages, and that
sh blocks can freely call any command-line executable such a system
would turn into a system-wide package management tool.
Perhaps there already exists a portable package management system
designed for local installs which could handle most of the heavy
lifting.
As another option, distributing Virtual Machine images are one solution
which I think work well and are increasingly realistic. Or similarly
providing the research environment as a cloud server image (e.g., Amazon
EC2).
Certainly an interesting area for further work!
Best,
--
Eric Schulte
http://cs.unm.edu/~eschulte/