def test_system_get_field_from_file(self):
"""Test returning a distributed field from file."""
field = self.get_field(nrow=5, ncol=7)
if MPI_RANK == 0:
path = self.get_temporary_file_path('data.nc')
else:
path = None
path = MPI_COMM.bcast(path)
with vm.scoped('write test field', [0]):
if MPI_RANK == 0:
field.write(path)
MPI_COMM.Barrier()
rd = RequestDataset(path)
out_field = rd.get()
if MPI_SIZE == 8:
self.assertEqual(vm.size, 8)
if MPI_RANK == 0:
path2 = self.get_temporary_file_path('out_field.nc')
else:
path2 = None
path2 = MPI_COMM.bcast(path2)
with vm.scoped_by_emptyable('out_field write', out_field):
if not vm.is_null:
out_field.write(path2)
MPI_COMM.Barrier()
with vm.scoped('get actual', [0]):
if MPI_RANK == 0:
actual = RequestDataset(path2).get()
actual = actual.data_variables[0].get_value().sum()
else:
actual = None
actual = MPI_COMM.bcast(actual)
desired = field.data_variables[0].get_value().sum()
self.assertAlmostEqual(actual, desired)
def test(self):
gs = self.get_grid_splitter()
desired_dst_grid_sum = gs.dst_grid.parent['data'].get_value().sum()
desired_dst_grid_sum = MPI_COMM.gather(desired_dst_grid_sum)
if MPI_RANK == 0:
desired_sum = np.sum(desired_dst_grid_sum)
desired = [{'y': slice(0, 180, None), 'x': slice(0, 240, None)},
{'y': slice(0, 180, None), 'x': slice(240, 480, None)},
{'y': slice(0, 180, None), 'x': slice(480, 720, None)},
{'y': slice(180, 360, None), 'x': slice(0, 240, None)},
{'y': slice(180, 360, None), 'x': slice(240, 480, None)},
{'y': slice(180, 360, None), 'x': slice(480, 720, None)}]
actual = list(gs.iter_dst_grid_slices())
self.assertEqual(actual, desired)
gs.write_subsets()
if MPI_RANK == 0:
rank_sums = []
for ctr in range(1, gs.nsplits_dst[0] * gs.nsplits_dst[1] + 1):
src_path = gs.create_full_path_from_template('src_template', index=ctr)
dst_path = gs.create_full_path_from_template('dst_template', index=ctr)
src_field = RequestDataset(src_path).get()
dst_field = RequestDataset(dst_path).get()
src_envelope_global = box(*src_field.grid.extent_global)
dst_envelope_global = box(*dst_field.grid.extent_global)
self.assertTrue(does_contain(src_envelope_global, dst_envelope_global))
actual = get_variable_names(src_field.data_variables)
self.assertIn('data', actual)
actual = get_variable_names(dst_field.data_variables)
self.assertIn('data', actual)
actual_data_sum = dst_field['data'].get_value().sum()
actual_data_sum = MPI_COMM.gather(actual_data_sum)
if MPI_RANK == 0:
actual_data_sum = np.sum(actual_data_sum)
rank_sums.append(actual_data_sum)
if MPI_RANK == 0:
self.assertAlmostEqual(desired_sum, np.sum(rank_sums))
index_path = gs.create_full_path_from_template('index_file')
self.assertTrue(os.path.exists(index_path))
MPI_COMM.Barrier()
index_path = gs.create_full_path_from_template('index_file')
index_field = RequestDataset(index_path).get()
self.assertTrue(len(list(index_field.keys())) > 2)
def test_write_parallel(self):
"""Test writing by selective rank."""
if MPI_SIZE != 3 and MPI_SIZE != 1:
raise SkipTest('MPI_SIZE != 1 or 3')
ranks = list(range(MPI_SIZE))
for base_rank in ranks:
for driver in [DriverCSV, DriverVector, DriverNetcdf]:
if MPI_RANK == 0:
path = self.get_temporary_file_path('{}-{}.{}'.format(
driver.key, base_rank, driver.common_extension))
else:
path = None
path = MPI_COMM.bcast(path)
with vm.scoped('field write by rank', [base_rank]):
if not vm.is_null:
geom = GeometryVariable(
value=[Point(1, 2), Point(3, 4)],
name='geom',
dimensions='geom')
data = Variable(name='data',
value=[10, 20],
dimensions='geom')
field = Field(geom=geom)
field.add_variable(data, is_data=True)
self.assertFalse(os.path.isdir(path))
field.write(path, driver=driver)
self.assertFalse(os.path.isdir(path))
rd = RequestDataset(path, driver=driver)
in_field = rd.get()
self.assertEqual(in_field['data'].dimensions[0].size,
2)
MPI_COMM.Barrier()
MPI_COMM.Barrier()
def test_system_parallel_write_ndvariable(self):
"""Test a parallel vector GIS write with a n-dimensional variable."""
ompi = OcgDist()
ompi.create_dimension('time', 3)
ompi.create_dimension('extra', 2)
ompi.create_dimension('x', 4)
ompi.create_dimension('y', 7, dist=True)
ompi.update_dimension_bounds()
if MPI_RANK == 0:
path = self.get_temporary_file_path('foo.shp')
t = TemporalVariable(name='time', value=[1, 2, 3], dtype=float, dimensions='time')
t.set_extrapolated_bounds('the_time_bounds', 'bounds')
extra = Variable(name='extra', value=[7, 8], dimensions='extra')
x = Variable(name='x', value=[9, 10, 11, 12], dimensions='x', dtype=float)
x.set_extrapolated_bounds('x_bounds', 'bounds')
# This will have the distributed dimension.
y = Variable(name='y', value=[13, 14, 15, 16, 17, 18, 19], dimensions='y', dtype=float)
y.set_extrapolated_bounds('y_bounds', 'bounds')
data = Variable(name='data', value=np.random.rand(3, 2, 7, 4), dimensions=['time', 'extra', 'y', 'x'])
dimension_map = {'x': {'variable': 'x', 'bounds': 'x_bounds'},
'y': {'variable': 'y', 'bounds': 'y_bounds'},
'time': {'variable': 'time', 'bounds': 'the_time_bounds'}}
vc = Field(variables=[t, extra, x, y, data], dimension_map=dimension_map, is_data='data')
vc.set_abstraction_geom()
else:
path, vc = [None] * 2
path = MPI_COMM.bcast(path)
vc = variable_collection_scatter(vc, ompi)
with vm.scoped_by_emptyable('write', vc):
if not vm.is_null:
vc.write(path, driver=DriverVector)
MPI_COMM.Barrier()
desired = 168
rd = RequestDataset(path, driver=DriverVector)
sizes = MPI_COMM.gather(rd.get().geom.shape[0])
if MPI_RANK == 0:
self.assertEqual(sum(sizes), desired)
def test_get_intersects_state_boundaries(self):
path_shp = self.path_state_boundaries
geoms = []
with fiona.open(path_shp) as source:
for record in source:
geom = shape(record['geometry'])
geoms.append(geom)
gvar = GeometryVariable(value=geoms, dimensions='ngeom')
gvar_sub = gvar.get_unioned()
if gvar_sub is not None:
subset = gvar_sub.get_value().flatten()[0]
else:
subset = None
subset = MPI_COMM.bcast(subset)
resolution = 1.0
for with_bounds in [False, True]:
grid = self.get_gridxy_global(resolution=resolution,
with_bounds=with_bounds)
vm.create_subcomm_by_emptyable('global grid',
grid,
is_current=True)
if not vm.is_null:
res = grid.get_intersects(subset, return_slice=True)
grid_sub, slc = res
vm.create_subcomm_by_emptyable('grid subset',
grid_sub,
is_current=True)
if not vm.is_null:
mask = Variable('mask_after_subset',
grid_sub.get_mask(),
dimensions=grid_sub.dimensions)
mask = variable_gather(mask)
if vm.rank == 0:
mask_sum = np.invert(mask.get_value()).sum()
mask_shape = mask.shape
else:
mask_sum = None
mask_shape = None
mask_sum = vm.bcast(mask_sum)
mask_shape = vm.bcast(mask_shape)
if with_bounds:
self.assertEqual(mask_shape, (54, 113))
self.assertEqual(
slc, (slice(108, 162, None), slice(1, 114, None)))
self.assertEqual(mask_sum, 1358)
else:
if MPI_SIZE == 2:
grid_bounds_global = [
dim.bounds_global
for dim in grid_sub.dimensions
]
self.assertEqual(grid_bounds_global, [(0, 52),
(0, 105)])
self.assertEqual(mask_shape, (52, 105))
self.assertEqual(
slc, (slice(109, 161, None), slice(8, 113, None)))
self.assertEqual(mask_sum, 1087)
if vm.rank == 0:
path = self.get_temporary_file_path('foo.nc')
else:
path = None
path = vm.bcast(path)
field = Field(grid=grid_sub)
field.write(path)
vm.finalize()
vm.__init__()
MPI_COMM.Barrier()
def test_get_intersects_parallel(self):
if sys.version_info.major == 3 and sys.version_info.minor == 5:
raise SkipTest('undefined behavior with Python 3.5')
grid = self.get_gridxy()
live_ranks = vm.get_live_ranks_from_object(grid)
# Test with an empty subset.
subset_geom = box(1000., 1000., 1100., 1100.)
with vm.scoped('empty subset', live_ranks):
if not vm.is_null:
with self.assertRaises(EmptySubsetError):
grid.get_intersects(subset_geom)
# Test combinations.
subset_geom = box(101.5, 40.5, 102.5, 42.)
keywords = dict(is_vectorized=[True, False],
has_bounds=[False, True],
use_bounds=[False, True],
keep_touches=[True, False])
for ctr, k in enumerate(self.iter_product_keywords(keywords)):
grid = self.get_gridxy()
vm_name, _ = vm.create_subcomm_by_emptyable('grid testing',
grid,
is_current=True)
if vm.is_null:
vm.free_subcomm(name=vm_name)
vm.set_comm()
continue
if k.has_bounds:
grid.set_extrapolated_bounds('xbounds', 'ybounds', 'bounds')
self.assertTrue(grid.has_bounds)
# Cannot use bounds with a point grid abstraction.
if k.use_bounds and grid.abstraction == 'point':
vm.free_subcomm(name=vm_name)
vm.set_comm()
continue
grid_sub, slc = grid.get_intersects(subset_geom,
keep_touches=k.keep_touches,
use_bounds=k.use_bounds,
return_slice=True)
if k.has_bounds:
self.assertTrue(grid.has_bounds)
# Test geometries are filled appropriately after allocation.
if not grid_sub.is_empty:
for t in grid_sub.get_abstraction_geometry().get_value().flat:
self.assertIsInstance(t, BaseGeometry)
self.assertIsInstance(grid_sub, Grid)
if k.keep_touches:
if k.has_bounds and k.use_bounds:
desired = (slice(0, 3, None), slice(0, 3, None))
else:
desired = (slice(1, 3, None), slice(1, 2, None))
else:
if k.has_bounds and k.use_bounds:
desired = (slice(1, 3, None), slice(1, 2, None))
else:
desired = (slice(1, 2, None), slice(1, 2, None))
if not grid.is_empty:
self.assertEqual(grid.has_bounds, k.has_bounds)
self.assertTrue(grid.is_vectorized)
self.assertEqual(slc, desired)
vm.free_subcomm(name=vm_name)
vm.set_comm()
# Test against a file. #########################################################################################
subset_geom = box(101.5, 40.5, 102.5, 42.)
if MPI_RANK == 0:
path_grid = self.get_temporary_file_path('grid.nc')
else:
path_grid = None
path_grid = MPI_COMM.bcast(path_grid)
grid_to_write = self.get_gridxy()
with vm.scoped_by_emptyable('write', grid_to_write):
if not vm.is_null:
field = Field(grid=grid_to_write)
field.write(path_grid, driver=DriverNetcdfCF)
MPI_COMM.Barrier()
rd = RequestDataset(uri=path_grid)
x = SourcedVariable(name='x', request_dataset=rd)
self.assertIsNone(x._value)
y = SourcedVariable(name='y', request_dataset=rd)
self.assertIsNone(x._value)
self.assertIsNone(y._value)
grid = Grid(x, y)
for target in [grid._y_name, grid._x_name]:
self.assertIsNone(grid.parent[target]._value)
self.assertTrue(grid.is_vectorized)
with vm.scoped_by_emptyable('intersects', grid):
if not vm.is_null:
sub, slc = grid.get_intersects(subset_geom, return_slice=True)
self.assertEqual(slc, (slice(1, 3, None), slice(1, 2, None)))
self.assertIsInstance(sub, Grid)
# The file may be deleted before other ranks open.
MPI_COMM.Barrier()