[Qgis-developer] SEXTANTE: R output?

[Matteo Mattiuzzi]

Hi Victor,
Having a separated libloc causes also a maintenance problem, as the libs in
there can only be updated (with ease) by SEXTANTE, that does not (?) update
them. I also think that it is better to have everything on system level but
there are also environments where a user has no writing privileges outside
its user space and a user libloc is required (this does R by default in
such a case, but it asks the user to create that dir if it doesn't already
exits, and R re-connects to that user libloc every time it starts with the
same user).

So except for the problem (esily to solve) of permissions and user libloc I
would stay within R defaults.

I think that SEXTANTE offers a good place to share scripts, _especially_
for non R experts (like the example scripts). I.e. I create an R-script
that my colleagues can use without knowing R. I think this is one of the
great things of SEXTANTE!

I think it is (quite) easily possible to remove the need of R knowledge for
a lot of situations, one of these it the installation and update of
packages. It is much easier to share a Rscript (rsx) to another system if
not installed packages are added automatically.

It could be interesting to have also a online repository of RSX scripts,
isn't every R-script is a bit like a plugin!?

Cheers, Matteo










2013/7/12 Victor Olaya <volayaf at gmail.com>

> Matteo
>
> Thanks for your help
>
> After some thinking and discussion, I think that the best and easiest
> solutions is just to have a separate SEXTANTE environment, with its
> own libPath. This is the current configuration, but it was not
> correctly set, so libs were installed tere, but not used from there
> later. With Donovan's solution, this wil be easy to fix.
>
> The R connector in SXTANTE is supposed to be used by people that can
> use R and write R scripts, so I guess we can let them add the
> necessary install commands for their scripts. SEXTANTE will just take
> care of rgdal, since that is used very often.
>
> What I will try to do is:
> - document this in detail in the manual
> - improve the error handling, so if the R script cannot be run due to
> a missing dependency, SEXTANTE will inform the user. For this, your
> scripts might be useful.
>
> I want to treat the R scripts much in the same way as the Python
> scripts. If you need a lib (other that the ones installed with QGIS),
> you have to install it yourself, modify the PYTHONPATH add the import
> sentences, etc...and I assume that you know Python well enough to take
> care of that. The same goes with R scripts. By having a separate R
> environment, I guess this is easier and also it will not interfere in
> the "normal" usage of R outside SEXTANTE. And we also get rid of the
> problem of not having priviledges to install new packages, since all
> goes to the home folder.
>
> What do you think of this?
>
> Thanks again!
> Victor
>
>
>
> 2013/7/12 Matteo Mattiuzzi <matteo at mattiuzzi.com>:
> > Hi guys I hope I'm not to late for the lunch,
> >
> > I have attached a (mature draft) Rscript that check if all needed
> packages
> > are installed (needed because in the given rscript.rsx there are some
> > "require(..." or "library(...")
> > One my Linux it works fine.
> > Note that it currently doesn't check for SEXTANTE R default packages as
> > raster/rgdal... but it would be no problem to add them by default.
> > Currently it installs to the default lib location (so you need to have
> write
> > access (in the R promt if the folder does not exist R asks you if you
> want
> > to create a libloc where you have write access, this could be problematic
> > and this is probably the reason for the sextante /R lib loc?, if this is
> a
> > problem, I can use the sextante R lib loc).
> >
> > Depending on where you save these files the command is:
> > R CMD BATCH "--args ~/exampleScriptNotOk.rsx" ~/checkPac.R
> > R CMD BATCH "--args ~/exampleScriptOk.rsx" ~/checkPac.R
> >
> > This command should be started on every Rscripr.rsx execution before the
> > real execution as it tries to make sure thet the packages requirement are
> > all ok. If no installation is required it is very fast, here 1/3 of a
> sec.
> > see ceckPac.Rout for the full execution log
> >
> > If this method is an option some testing from your side would be great!
> This
> > is only a suggestion, ant it can be easily addapted/changed/enhanced.
> >
> > Matteo
> >
> >
> > 2013/7/12 Alister Hood <Alister.Hood at synergine.com>
> >>
> >> Hi guys,
> >> Speaking of R, would it be worth adding some maintenance scripts? e.g.
> >>
> >> update.packages(checkBuilt=TRUE)
> >>
> >> and even:
> >>
> >> if(!require(installr)) {
> >> install.packages("installr"); require(installr)} #load / install+load
> >> installr
> >> updateR()
> >>
> >> I'm not much of an R guy - maybe there are other things too.
> >>
> >> Regards,
> >> Alister
> >> _______________________________________________
> >> Qgis-developer mailing list
> >> Qgis-developer at lists.osgeo.org
> >> http://lists.osgeo.org/mailman/listinfo/qgis-developer
> >
> >
> >
> > _______________________________________________
> > Qgis-developer mailing list
> > Qgis-developer at lists.osgeo.org
> > http://lists.osgeo.org/mailman/listinfo/qgis-developer
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/qgis-developer/attachments/20130712/ef941d05/attachment.html>