[GRASS-dev] Working with TGIS without starting GRASS explicitly

[Laurent C.]

Hi Sören,

I followed you suggestion and now each simulation run in its own process
using the multiprocessing library. It works well, thanks.
However, there was indeed something strange going on:
The new mapset, location and accessible mapsets were set properly, as seen
by the output of g.mapset and g.mapsets
However, the output of ciface.available_mapsets() and therefore
core.get_available_temporal_mapsets() were not updated after the first
simulation, so tgis.init() were failing, not seeing the new current mapset
in its own list of available mapsets.

Below was the output of the software:

g.mapset -p
swashes_1d_1000m
g.mapsets -p
Accessible mapsets:
swashes_1d_1000m PERMANENT
ciface.available_mapsets()
['swashes_1d_1000m', 'PERMANENT']
get_available_temporal_mapsets()
{'swashes_1d_1000m': ('sqlite',
'/home/laurent/grassdata/flood_test/swashes_1d_1000m/tgis/sqlite.db')}
Starting simulation for configuration file macdo1000.ini...
[...]
g.mapset -p
hull
g.mapsets -p
Accessible mapsets:
hull PERMANENT hull_rain
ciface.available_mapsets()
['swashes_1d_1000m', 'PERMANENT']
get_available_temporal_mapsets()
{'swashes_1d_1000m': ('sqlite',
'/home/laurent/grassdata/flood_test/swashes_1d_1000m/tgis/sqlite.db')}
ERROR: Unable to execute sql statement. There is no temporal database
connection defined for mapset <hull>

It seems that tgis.init() is not doing what it is supposed to, according to
the documentation:

"Re-run this function in case the following GRASS variables change while
the process runs:
- MAPSET
- LOCATION_NAME
- GISDBASE
- TGIS_DISABLE_MAPSET_CHECK
- TGIS_DISABLE_TIMESTAMP_WRITE
"

Thanks again.

Regards,
Laurent

2016-09-22 13:52 GMT-05:00 Sören Gebbert <soerengebbert at googlemail.com>:
> Hi Laurent,
> the temporal framework was neither tested nor designed to work in a script
> that changes the location while running. I would suggest to run the
temporal
> framework related code in a subprocess that is created each time the
> location or mapset changes. Unfortunately i have no other suggestion.
>
> However, have you tested your code with grass 7.2 or 7.3? Many
improvements
> have been implement in the recent version that would be beneficial to use.
>
> Best regards
> Soeren
>
> 2016-09-22 2:47 GMT+02:00 Laurent C. <lrntct at gmail.com>:
>>
>> Hello all,
>>
>> I've ran into another related issue.
>> One of the goal of running the software outside of GRASS shell is to
>> batch process simulations in various Locations/mapsets.
>>
>> I set the GRASS session for each case. The simulation works well for
>> the first case.
>> However when starting the second case, tgis.init() fails with the
>> following error:
>>
>> ERROR: Unable to execute sql statement. There is no temporal database
>> connection defined for mapset <hull>
>>
>> The mapset and location are properly set.
>> If this case is run first, it works well and it's the second one that
>> fail.
>> Running t.connect -c between two cases does not solve the problem.
>> Actually, t.connect -p shows the correct connection parameters, but
>> tgis.init() still fails.
>>
>> Regards,
>> Laurent
>>
>>
>> 2016-09-21 18:27 GMT-05:00 Laurent C. <lrntct at gmail.com>:
>> > Hi Sören,
>> >
>> > Setting LD_LIBRARY_PATH is working, thanks. But it cannot be set at
>> > run-time, which is not very user-friendly.
>> > I guess this issue is related to the ticket 2424 [1].
>> >
>> > I managed to get it work at run-time by restarting the program with
>> > sys.execv() after setting the path [2], but I find it a bit ugly and
>> > quite verbose to be multi-platform.
>> >
>> > It would be great if an easier option was possible.
>> >
>> > Regards,
>> > Laurent
>> >
>> > [1] https://trac.osgeo.org/grass/ticket/2424
>> > [2]
>> >
http://stackoverflow.com/questions/6543847/setting-ld-library-path-from-inside-python
>> >
>> >
>> > 2016-09-21 15:40 GMT-05:00 Sören Gebbert <soerengebbert at googlemail.com
>:
>> >> Hi,
>> >> i think you have to put the GRASS libraries into your LD_LIBRARY_PATH
>> >> so
>> >> that the python wrapper can access them.
>> >>
>> >> Best regards
>> >> Soeren
>> >>
>> >> 2016-09-21 20:37 GMT+02:00 Laurent C. <lrntct at gmail.com>:
>> >>>
>> >>> Hello all,
>> >>>
>> >>> I'm trying to adapt a python module in order for it to be run outside
>> >>> of
>> >>> GRASS.
>> >>> I followed the instructions from the wiki and
>> >>> lib/python/script/setup.py
>> >>>
>> >>> gsetup.init() run without error and I can list maps like in the
>> >>> example.
>> >>>
>> >>> However, when I try to import the temporal module, I receive this
>> >>> error:
>> >>>
>> >>> import grass.temporal as tgis
>> >>> File "/usr/lib/grass70/etc/python/grass/temporal/__init__.py", line
>> >>> 1, in <module>
>> >>> from core import *
>> >>> File "/usr/lib/grass70/etc/python/grass/temporal/core.py", line 38,
>> >>> in <module>
>> >>> from c_libraries_interface import *
>> >>> File
>> >>>
"/usr/lib/grass70/etc/python/grass/temporal/c_libraries_interface.py",
>> >>> line 19, in <module>
>> >>> import grass.lib.gis as libgis
>> >>> File "/usr/lib/grass70/etc/python/grass/lib/gis.py", line 23, in
>> >>> <module>
>> >>> _libs["grass_gis.7.0.4"] = load_library("grass_gis.7.0.4")
>> >>> File "/usr/lib/grass70/etc/python/grass/lib/ctypes_loader.py", line
>> >>> 55, in load_library
>> >>> return self.load(path)
>> >>> File "/usr/lib/grass70/etc/python/grass/lib/ctypes_loader.py", line
>> >>> 71, in load
>> >>> raise ImportError,e
>> >>> ImportError: libgrass_datetime.7.0.4.so: cannot open shared object
>> >>> file: No such file or directory
>> >>>
>> >>>
>> >>> Did I forgot to set-up something or this is a bug?
>> >>>
>> >>> Cheers,
>> >>> Laurent
>> >>> _______________________________________________
>> >>> grass-dev mailing list
>> >>> grass-dev at lists.osgeo.org
>> >>> http://lists.osgeo.org/mailman/listinfo/grass-dev
>> >>
>> >>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-dev/attachments/20160923/5bb1b0a9/attachment-0001.html>