[GRASS-stats] Re: spgrass6 and initGRASS
Roger Bivand
Roger.Bivand at nhh.no
Mon Dec 7 14:58:13 EST 2009
On Mon, 7 Dec 2009, Roger Bivand wrote:
> On Mon, 7 Dec 2009, Patrick Giraudoux wrote:
>
>> Dear Roger,
>>
>> I have moved the thread and the answer below from qgis-user at lists.osgeo.org
>> to this list as advised...
>>>
>>> That is because the GRASS environment variables have not been set at that
>>> point. Do you need to use GRASS in throw-away (initGRASS) mode, or with an
>>> existing location? If the latter, please use the interface in the regular
>>> way, starting R "within" GRASS. I'll try to reproduce the problem, but it
>>> is not easy to do, I'm afraid.
>
> Dear Patrick,
>
> I've tried with Metis with no success. The first problem is addressed in this
> draft release:
>
> http://spatial.nhh.no/R/Devel/spgrass6_0.6-13.zip
>
> but with Metis is simply falls over later. With WinGRASS (last but latest),
> it just works. It even works in Metis once the basic GRASS framework is set
> up using WinGRASS, but behaviour is erratic and unpredictable. To track it
> with Metis (which I just uninstalled to try OSGeo4W) seemed impossible, it
> wasn't clear where the GRASS modules were looking for their environment
> variables and GISRC backstore.
>
> I would probably use file transfer and have QGIS do the desktop mapping and
> WinGRASS the analytical stuff, but I can't say that I have favourites here.
I've now installed QGIS 1.3.0 Mimas and GRASS 6.4.0 SVN from OSGeo4W.
Doing it their way (rather than my way), you then need Carson Farmer's
ManageR plugin and the rpy2 2.0.6 msi for python 2.5 (the one that OSGeo4W
ships). Next probably you need to put your R (mine is 2.10.0) on the PATH
which OSGeo4W and QGIS have manipulated - maybe through the control panel
- needs updating manually.
Start QGIS, look for Carson's repository, install and enable manageR, open
a GRASS mapset in QGIS, enter library(spgrass6) in the manageR console,
and if you're lucky GRASS may be there. I've managed to move spearfish
roads to R and tabulate road length by class using SpatialLinesLength() on
the representation on the R side. Fairly horrible to install for students,
but maybe?
Roger
>
>>
>> I would like to be capable to read into GRASS locations from within R (so
>> not from R started withing GRASS) -what is supposedly done with initGRASS;
>> eg just to pick up a region definition into a "spgrass6" object and compute
>> on that. I though that initGRASS was design for this task.
>>
>
> Right, initGRASS with home=tempdir() (not what I replied earlier - with no
> home= given, it uses whatever Windows says "HOME" is) starts a new location
> and mapset in the R temporary directory, which you move data to for GRASS to
> analyse.
>
>>> Have you tried the WinGRASS stand-alone (I haven't trusted either QGIS
>>> GRASS used this way or OSGeo4W - I follow the bug-tracker for OSGEO4W, so
>>> stay away)?
>>
>> I am trying this essentially on teaching purposes. Students are working
>> under Windows XP and Vista. I have only a limited number of hours, and to
>> board GIS concepts from QGIS then to GRASS is somehow easier for them than
>> straight within GRASS (although it is just the reverse when you know
>> GRASS). The QGIS Metis standalone version was excellent for that. You have
>> seen that the updates to Kore and Mimas are done with no standalone version
>> available on the QGIS official site. There the use of OSGeo4W is encouraged
>> and I am trying to find out what could be best advisable for the students.
>>
>> I did not tried WinGRASS yet, and I am interested in doing it, but have
>> installed the OSGeo4W GRASS 6.4svn version (works well - but definitely
>> the unix/linux environment is more convenient for command lines), as well
>> as OSGeo4W QGIS Kore and Mimas, and try and compare them.
>>
>
> Roger
>
>>> What is your sessionInfo()?
>> sessionInfo()
>> R version 2.10.0 (2009-10-26)
>> i386-pc-mingw32
>>
>> locale:
>> [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 [3]
>> LC_MONETARY=French_France.1252 LC_NUMERIC=C [5]
>> LC_TIME=French_France.1252 attached base packages:
>> [1] grDevices datasets splines graphics stats tcltk utils [8]
>> methods base other attached packages:
>> [1] spgrass6_0.6-12 XML_2.6-0 rgdal_0.6-22 sp_0.9-47 [5]
>> svSocket_0.9-48 TinnR_1.0.3 R2HTML_1.59-1 Hmisc_3.7-0 [9]
>> survival_2.35-7
>>
>> loaded via a namespace (and not attached):
>> [1] cluster_1.12.1 grid_2.10.0 lattice_0.17-26 svMisc_0.9-56 [5]
>> tools_2.10.0
>> Roger Bivand a écrit :
>>> On Mon, 7 Dec 2009, Patrick Giraudoux wrote:
>>>
>>>> Have tried the following:
>>>>
>>>> initGRASS("C:\\OSGeo4W\\bin",home="U:\\Documents and
>>>> Settings\\pgiraudo\\",override=T)
>>>> initGRASS("C:\\Program Files\\Quantum GIS\\grass",home="U:\\Documents and
>>>> Settings\\pgiraudo",override=T)
>>>> initGRASS("C:\\OSGeo4W\\bin",override=T)
>>>>
>>>> with the same results...
>>>>
>>>>> initGRASS("C:\\OSGeo4W\\bin",home="U:\\Documents and
>>>> Settings\\pgiraudo\\",override=T)
>>>> Error in system(paste("g.dirseps.exe -g",
>>>> shQuote(Sys.getenv("WINGISRC"))), :
>>>> g.dirseps.exe not found
>>>>> initGRASS("C:\\Program Files\\Quantum GIS\\grass",home="U:\\Documents
>>>>> and
>>>> Settings\\pgiraudo",override=T)
>>>> Error in Sys.setenv(GISRC = gisrc) : wrong length for argument
>>>>> initGRASS("C:\\OSGeo4W\\bin",override=T)
>>>> Error in Sys.setenv(GISRC = gisrc) : wrong length for argument
>>>>
>>>> Quite puzzling since g.dirseps does not give a message "unknown command"
>>>> when tried within grass (thus it is suppose to be in the GRASS programme)
>>>
>>> That is because the GRASS environment variables have not been set at that
>>> point. Do you need to use GRASS in throw-away (initGRASS) mode, or with an
>>> existing location? If the latter, please use the interface in the regular
>>> way, starting R "within" GRASS. I'll try to reproduce the problem, but it
>>> is not easy to do, I'm afraid.
>>>
>>> What is your sessionInfo()? Have you tried the WinGRASS stand-alone (I
>>> haven't trusted either QGIS GRASS used this way or OSGeo4W - I follow the
>>> bug-tracker for OSGEO4W, so stay away)?
>>>
>>> Would you mind moving to the proper list - this is a GRASS question much
>>> more than an R one:
>>>
>>> http://lists.osgeo.org/mailman/listinfo/grass-stats
>>>
>>> Roger
>>>
>>>>
>>>> Patrick
>>>>
>>>>
>>>>
>>>> Roger Bivand a écrit :
>>>>> On Mon, 7 Dec 2009, Patrick Giraudoux wrote:
>>>>>
>>>>>> Here is the result:
>>>>>>
>>>>>> library(spgrass6)
>>>>>>
>>>>>> debug(initGRASS)
>>>>>>> initGRASS("C:/OSGeo4W/bin",home="U:/Documents and
>>>>>> Settings/pgiraudo/",override=T)
>>>>>>
>>>>>> (...)
>>>>>>
>>>>>> Browse[2]>
>>>>>> debug: gisrc <- system(paste("g.dirseps.exe -g",
>>>>>> shQuote(Sys.getenv("WINGISRC"))),
>>>>>> intern = TRUE)
>>>>>> Browse[2]> Sys.getenv("WINGISRC")
>>>>>> WINGISRC
>>>>>> "U:/Documents and Settings/pgiraudo/\\.grassrc6"
>>>>>> Browse[2]> shQuote(Sys.getenv("WINGISRC"))
>>>>>> [1] "\"U:/Documents and Settings/pgiraudo/\\.grassrc6\""
>>>>>> Browse[2]> gisrc
>>>>>> Error: object 'gisrc' not found
>>>>>
>>>>> Could you try both not setting home= (you get a temporary folder that is
>>>>> deleted when you exit R), or setting
>>>>>
>>>>> home="U:\\Documents and Settings\\pgiraudo"
>>>>>
>>>>> without the trailing forward slash?
>>>>>
>>>>> This a problem connected to protecting spaces on the one hand and system
>>>>> specific directory separators on the other, I think.
>>>>>
>>>>> Roger
>>>>>
>>>>>>
>>>>>> Beforehand, the first time I run
>>>>>> initGRASS("C:/OSGeo4W/bin",home="U:/Documents and
>>>>>> Settings/pgiraudo/",override=T), I got:
>>>>>>
>>>>>> initGRASS("C:/OSGeo4W/bin",home="U:/Documents and
>>>>>> Settings/pgiraudo/",override=T)
>>>>>> Error in system(paste("g.dirseps.exe -g",
>>>>>> shQuote(Sys.getenv("WINGISRC"))), :
>>>>>> g.dirseps.exe not found
>>>>>>
>>>>>> Then, if I try the same commande on QGIS Metis, I got:
>>>>>> initGRASS("C:/Program Files/Quantum GIS/grass",home="U:/Documents and
>>>>>> Settings/pgiraudo",override=T)
>>>>>> Error in Sys.setenv(GISRC = gisrc) : wrong length for argument
>>>>>>
>>>>>>
>>>>>> Best,
>>>>>>
>>>>>> Patrick
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Roger Bivand a écrit :
>>>>>>> On Sun, 6 Dec 2009, Patrick Giraudoux wrote:
>>>>>>>
>>>>>>>> Dear all,
>>>>>>>>
>>>>>>>> I am trying to init a GRASS session from within R, using the function
>>>>>>>> initGRASS from the package spgrass6. GRASS is installed in two ways
>>>>>>>> on my computer (here, the OS is Windows XP): with QGIS metis
>>>>>>>> (precompiled version including grass 6.3 (2008)) and grass 6.4.0svn,
>>>>>>>> just freshly installed with the Osgeo4W installer.
>>>>>>>>
>>>>>>>> In both cases I get the same result:
>>>>>>>>
>>>>>>>>> initGRASS("C:/Program Files/Quantum GIS/grass",home="U:/Documents
>>>>>>>>> and
>>>>>>>> Settings/pgiraudo",override=T)
>>>>>>>> Error in Sys.setenv(GISRC = gisrc) : wrong length for argument
>>>>>>>>
>>>>>>>>> initGRASS("C:/OSGeo4W/bin",home="U:/Documents and
>>>>>>>> Settings/pgiraudo",override=T)
>>>>>>>> Error in Sys.setenv(GISRC = gisrc) : wrong length for argument
>>>>>>>>
>>>>>>>> Any idea about what happens ?
>>>>>>>
>>>>>>> No, without the output of traceback(), it isn't obvious, although here
>>>>>>> traceback() might not help. Run under debug(initGRASS) to the step
>>>>>>> before, that is:
>>>>>>>
>>>>>>> gisrc <- system(paste("g.dirseps.exe -g",
>>>>>>> shQuote(Sys.getenv("WINGISRC"))), intern=TRUE)
>>>>>>>
>>>>>>> then run:
>>>>>>>
>>>>>>> Sys.getenv("WINGISRC")
>>>>>>> shQuote(Sys.getenv("WINGISRC"))
>>>>>>> gisrc
>>>>>>>
>>>>>>> and try to see what is going on. At this stage, Sys.getenv("GISRC")
>>>>>>> should be "junk", assigned a couple of lines earlier.
>>>>>>>
>>>>>>> Maybe the grass-stats list would have a higher proportion of users of
>>>>>>> both GRASS and R.
>>>>>>>
>>>>>>> Hope this helps,
>>>>>>>
>>>>>>> Roger
>>>>>>>
>>>>>>>>
>>>>>>>> Patrick
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>>
>
>
--
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: Roger.Bivand at nhh.no
More information about the grass-stats
mailing list