[GRASS-stats] Re: Accessing grass raster data from R via library and not via file

[Rainer M Krug]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 13/01/12 22:19, Hamish wrote:
> Rainer wrote:
>> I am writing to you directly, as you are the author of the 
>> r.out.mat.
> 
> dangerous! my inbox is flooded and I might easily miss it :)
> 

OK - I'll push it to grass-stats - that others can get involved.

>> I have a simulation, which heavily relies on reading and writing 
>> spatial data from GRASS into R and vice versa. Profiling 
>> indicated, that the actual readRAST6() from R takes uip a large 
>> proportion of the simulation. So I looked into alternatives, and 
>> as it is not that difficult to return data from C to R, I thought
>> about that option.
> 
> please post to the statsgrass mailing list. Roger

Done

> may have a good idea about what to do. ISTR there were two
> options, and older & slower but more compatible way and a newer and
> faster way. you could select them via a switch. maybe your system
> is using the old+slow way?

I tried out with the useGDAL option TRUE and FALS, aned the plugin is
not installed on the cluster where I finally want to run the
simulation (and, in addition, I would like to avoid GDAL doe to
continueing hassles wth isnstalaltions etc. on the cluster - locally
GDAL works nicely)

I actualy did some benchmarks between useGDAL TRUE and FALSE and there
was a difference in the 10% range, but still not fast enough.

> 
>> The module r.out.mat actually extracts all the data from GRASS 
>> and saves it (spatial header information and the actual map 
>> data).
> 
> r.out.bin would be more appropriate, as Matlab format saves 
> column-wise not row-wise due to its Fortran origins, so you have
> to read the entire map into memory, which doesn't work for massive
>  datasets.

Thanks - I'll look at that one. The "part of map" is a useful
consideration.

> 
> both are very simple, see doc/raster/r.example/ in the source
> code. no need for libraries or anything.
> 
> 
> header data is available from r.info flags and $MAPSET/cellhd/

The thing is I would like to avoid using files and access GRASS maps
as directly as possible, which, as I understand it at the moment,
would be via C function(s) in GRASS whose return values are the data,
and a C function(s) in R which calls these GRASS C functions to
provide the data to R.

> 
> 
> the alternate idea is to use R's gdal package to read the GRASS 
> raster maps directly.

As mentioned earlier, I would like to avoid GDAL doe to installation
problems - but I should try this.

I understand that the idea of GDAL is to convert spatial data between
different formats, but there is (as far as I now) no way to get GASS
data directly into R, without creating an intermediate format, so
GRASS (disk) --> GDAL (memory) --> new format (disk) --> R

resulting in reading from disk, converting, writing to disk and
reading the converted data again.

No problem if I am doing analysis, but if I am running my simulation
model, data is read and written to GRASS all the time, and this
intermediate step costs lots of time.

I could do some benchmarks with the spearfish dataset next week to
show what I mean.

Cheers and thanks,

Rainer

> 
> 
> double check your region bounds and resolution are correct, or it 
> could waste a lot of time.

I'll do, but I am 99% sure that they are correct and as I need them.


> 
> 
> regards, Hamish


- -- 
Rainer M. Krug, PhD (Conservation Ecology, SUN), MSc (Conservation
Biology, UCT), Dipl. Phys. (Germany)

Centre of Excellence for Invasion Biology
Stellenbosch University
South Africa

Tel :       +33 - (0)9 53 10 27 44
Cell:       +33 - (0)6 85 62 59 98
Fax :       +33 - (0)9 58 10 27 44

Fax (D):    +49 - (0)3 21 21 25 22 44

email:      Rainer at krugs.de

Skype:      RMkrug
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk8RR30ACgkQoYgNqgF2ego03gCeK5DhbC1TQzgVM7OdBMwvqR+u
JoUAniH8s9j0HLb3KafkqoAU6SYNRbcl
=+gkB
-----END PGP SIGNATURE-----