def test_reconciliation(self):
"Tests that reconciling multiple rasters works correctly for multiple views."
lims = {}
ns = {}
hfs = {}
views = []
for order in ['xy','yx']:
for first in ['+','-']:
for second in ['+','-']:
view = order[0]+first+order[1]+second
views.append(view)
lim_lo = np.random.uniform(-1,-.5,size=2)
lim_hi = np.random.uniform(.5,1,size=2)
lim = [lim_lo[0], lim_hi[0], lim_lo[1], lim_hi[1]]
n = np.random.randint(200,500,size=2)
lims[view]=np.array(lim)*180./np.pi
ns[view]=n
x = np.linspace(lim[0],lim[1],n[0])
y = np.linspace(lim[2],lim[3],n[1])
print view
fname, hf = write_datafile(x,y,view=view)
hfs[view] = hf
lon, lat, layers = reconcile_multiple_rasters([hfs[v].root for v in views], thin=2)
# Check that the layers all have the right shape
assert(np.all([(l.shape == (len(lon),len(lat))) for l in layers]))
# Check that the limits are inside the joint intersection
assert(np.all([(lon.min() >= l[0]) * (lon.max() <= l[1]) * (lat.min() >= l[2]) * (lat.max() <= l[3]) for l in lims.itervalues()]))
# Check that the coarseness is maximal
assert([(lims[v][1]-lims[v][0])/float(ns[v][0]-1) <= lon[1]-lon[0] for v in views])
assert([(lims[v][2]-lims[v][3])/float(ns[v][1]-1) <= lat[1]-lat[0] for v in views])
lonmin = np.min([l[0] for l in lims.itervalues()])
lonmax = np.max([l[1] for l in lims.itervalues()])
latmin = np.min([l[2] for l in lims.itervalues()])
latmax = np.max([l[3] for l in lims.itervalues()])
import pylab as pl
for i in xrange(len(views)):
pl.figure(figsize=(10,6))
pl.subplot(1,2,2)
pl.imshow(grid_convert(layers[i],'x+y+','y+x+'),extent=[lon.min(),lon.max(),lat.min(),lat.max()])
pl.title(views[i] + ': ' + str(hfs[views[i]].root.data.shape))
pl.axis([lonmin, lonmax, latmin, latmax])
pl.subplot(1,2,1)
pl.imshow(grid_convert(hfs[views[i]].root.data[:], views[i], 'y+x+'), extent = lims[views[i]])
pl.plot([lon.min(), lon.max(), lon.max(), lon.min()],[lat.min(), lat.min(), lat.max(), lat.max()],'r-')
pl.axis([lonmin, lonmax, latmin, latmax])
def make_covering_raster(thin=1, env_variables=(), **kwds):
a='MODIS-hdf5/raw-data.land-water.geographic.world.version-4'
names = [a]+env_variables
names.sort()
cache_fname = 'rasters_%i_%s_.hdf5'%(thin, hashlib.sha1('~*~lala~*~oolala'.join(names)).hexdigest())
cache_found = False
if 'anopheles-caches' in os.listdir('.'):
if cache_fname in os.listdir('anopheles-caches'):
hf = tables.openFile(os.path.join('anopheles-caches',cache_fname))
lon = hf.root.lon[:]
lat = hf.root.lat[:]
layers = []
for n in [a] + env_variables:
layers.append(getattr(hf.root,os.path.split(n)[1])[:])
hf.close()
cache_found = True
if not cache_found:
lon,lat,layers = reconcile_multiple_rasters([get_datafile(n) for n in [a] + env_variables], thin=thin)
hf = tables.openFile(os.path.join('anopheles-caches',cache_fname),'w')
hf.createArray('/','lon',lon)
hf.createArray('/','lat',lat)
names = [a] + env_variables
for i in xrange(len(layers)):
hf_arr=hf.createCArray('/',os.path.split(names[i])[1],shape=layers[i].shape,chunkshape=layers[i].shape,atom=tables.FloatAtom(),filters=tables.Filters(complevel=1))
hf_arr[:] = layers[i]
hf.close()
mask = np.round(layers[0]).astype(bool)
img_extent = [lon.min(), lat.min(), lon.max(), lat.max()]
lat_grid, lon_grid = np.meshgrid(lat*np.pi/180.,lon*np.pi/180.)
x=np.dstack([lon_grid,lat_grid]+[l for l in layers[1:]])
if x.shape[:-1] != mask.shape:
raise RuntimeError, 'Shape mismatch in mask and other layers.'
return mask, x, img_extent