Code to replicate a bug in grass.pipe_command

replicate_error.py

import os
import tempfile
import shutil

import grass.script as grass
import grass.script.setup as gsetup

# first do:
# mkdir /tmp/grasstest
# wget http://grass.osgeo.org/sampledata/north_carolina/nc_spm_latest.zip
# unzip *.zip

GIS_BASE = '/usr/lib/grass64/'
GRASS_DB = '/tmp/grasstest/'

loc_name = 'nc_spm_08'
rast_name = 'elev_ned_30m'


def export(loc, name, out_fn='/tmp/exported.tif'):
    gsetup.init(GIS_BASE, GRASS_DB, loc, 'PERMANENT')
    cmd = 'r.out.gdal'
    kwargs = {
        'quiet': True,
        'flags': 'c',
        'input': name,
        'output': out_fn
    }
    process = grass.pipe_command(cmd, **kwargs)
    stdout, stderr = process.communicate()


def run():
    i = 0
    while True:
        i += 1
        print 'Iteration %s' % i
        work_dir = tempfile.mkdtemp()
        out_fn = os.path.join(work_dir, 'rast.tif')
        export(loc_name, rast_name, out_fn)
        shutil.rmtree(work_dir)
        
run()

Since this can pop out in some search, it might be useful to note that the problem was not in pipe_command but in subprocess call hitting operating system limit for environment + parameters. The environment (env vars + their values) was full of paths accumulated from gsetup.init calls.

The solution is not to call gsetup.init repetitively (it is enough just to all it once and then switch Mapsets if needed). See also https://trac.osgeo.org/grass/ticket/2468 and https://gist.github.com/max-nova/487a82de00651a33f2c2.