def test_write_variable_collection_object_arrays(self):
"""Test writing variable length arrays in parallel."""
with vm.scoped('write', [0]):
if not vm.is_null:
path_actual = self.get_temporary_file_path('in.nc')
path_desired = self.get_temporary_file_path('out.nc')
value = [[1, 3, 5],
[7, 9],
[11]]
v = Variable(name='objects', value=value, fill_value=4, dtype=ObjectType(int), dimensions='values')
v.write(path_desired)
else:
v, path_actual, path_desired = [None] * 3
path_actual = MPI_COMM.bcast(path_actual)
path_desired = MPI_COMM.bcast(path_desired)
dest_mpi = OcgDist()
dest_mpi.create_dimension('values', 3, dist=True)
dest_mpi.update_dimension_bounds()
scattered = variable_scatter(v, dest_mpi)
outvc = VariableCollection(variables=[scattered])
with vm.scoped_by_emptyable('write', outvc):
if not vm.is_null:
outvc.write(path_actual)
if MPI_RANK == 0:
self.assertNcEqual(path_actual, path_desired)
def test_system_changing_field_name(self):
path1 = self.get_temporary_file_path('foo1.nc')
path2 = self.get_temporary_file_path('foo2.nc')
vc1 = VariableCollection(name='vc1')
var1 = Variable('var1',
value=[1, 2, 3],
dimensions='three',
parent=vc1)
vc2 = VariableCollection(name='vc2')
vc1.add_child(vc2)
var2 = Variable('var2',
value=[4, 5, 6, 7],
dimensions='four',
parent=vc2)
vc1.write(path1)
rd = RequestDataset(path1)
# rd.inspect()
nvc = rd.get_variable_collection()
nvc2 = nvc.children['vc2']
self.assertIsNone(nvc2['var2']._value)
self.assertEqual(nvc2.name, 'vc2')
nvc2.set_name('extraordinary')
self.assertIsNotNone(nvc2['var2'].get_value())
self.assertEqual(nvc2['var2'].get_value().tolist(), [4, 5, 6, 7])
nvc.write(path2)
rd2 = RequestDataset(path2)
# rd2.inspect()
n2vc = rd2.get_variable_collection()
self.assertEqual(n2vc.children[nvc2.name].name, nvc2.name)
def test_system_changing_field_name(self):
path1 = self.get_temporary_file_path('foo1.nc')
path2 = self.get_temporary_file_path('foo2.nc')
vc1 = VariableCollection(name='vc1')
var1 = Variable('var1', value=[1, 2, 3], dimensions='three', parent=vc1)
vc2 = VariableCollection(name='vc2')
vc1.add_child(vc2)
var2 = Variable('var2', value=[4, 5, 6, 7], dimensions='four', parent=vc2)
vc1.write(path1)
rd = RequestDataset(path1)
# rd.inspect()
nvc = rd.create_raw_field()
nvc2 = nvc.children['vc2']
self.assertIsNone(nvc2['var2']._value)
self.assertEqual(nvc2.name, 'vc2')
nvc2.set_name('extraordinary')
self.assertIsNotNone(nvc2['var2'].get_value())
self.assertEqual(nvc2['var2'].get_value().tolist(), [4, 5, 6, 7])
nvc.write(path2)
rd2 = RequestDataset(path2)
# rd2.inspect()
n2vc = rd2.create_raw_field()
self.assertEqual(n2vc.children[nvc2.name].name, nvc2.name)
def test_write_variable_collection_object_arrays(self):
"""Test writing variable length arrays in parallel."""
with vm.scoped('write', [0]):
if not vm.is_null:
path_actual = self.get_temporary_file_path('in.nc')
path_desired = self.get_temporary_file_path('out.nc')
value = [[1, 3, 5], [7, 9], [11]]
v = Variable(name='objects',
value=value,
fill_value=4,
dtype=ObjectType(int),
dimensions='values')
v.write(path_desired)
else:
v, path_actual, path_desired = [None] * 3
path_actual = MPI_COMM.bcast(path_actual)
path_desired = MPI_COMM.bcast(path_desired)
dest_mpi = OcgDist()
dest_mpi.create_dimension('values', 3, dist=True)
dest_mpi.update_dimension_bounds()
scattered = variable_scatter(v, dest_mpi)
outvc = VariableCollection(variables=[scattered])
with vm.scoped_by_emptyable('write', outvc):
if not vm.is_null:
outvc.write(path_actual)
if MPI_RANK == 0:
self.assertNcEqual(path_actual, path_desired)
def test_get_dist_default_distribution(self):
"""Test using default distributions defined by drivers."""
with vm.scoped('write', [0]):
if not vm.is_null:
path = self.get_temporary_file_path('foo.nc')
varx = Variable('x',
np.arange(5),
dimensions='five',
attrs={'axis': 'X'})
vary = Variable('y',
np.arange(7) + 10,
dimensions='seven',
attrs={'axis': 'Y'})
vc = VariableCollection(variables=[varx, vary])
vc.write(path)
else:
path = None
path = MPI_COMM.bcast(path)
rd = RequestDataset(path)
dist = rd.driver.dist
distributed_dimension = dist.get_dimension('seven')
self.assertTrue(distributed_dimension.dist)
def test_system_parallel_write_ndvariable(self):
"""Test a parallel CSV 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.csv')
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'])
vc = VariableCollection(variables=[t, extra, x, y, data])
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,
iter_kwargs={
'variable': 'data',
'followers': ['time', 'extra', 'y', 'x']
},
driver=DriverCSV)
if MPI_RANK == 0:
desired = 169
with open(path, 'r') as f:
lines = f.readlines()
self.assertEqual(len(lines), desired)
def test_system_parallel_write_ndvariable(self):
"""Test a parallel CSV 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.csv')
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'])
vc = VariableCollection(variables=[t, extra, x, y, data])
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, iter_kwargs={'variable': 'data', 'followers': ['time', 'extra', 'y', 'x']},
driver=DriverCSV)
if MPI_RANK == 0:
desired = 169
with open(path, 'r') as f:
lines = f.readlines()
self.assertEqual(len(lines), desired)
def test_get_dist_default_distribution(self):
"""Test using default distributions defined by drivers."""
with vm.scoped('write', [0]):
if not vm.is_null:
path = self.get_temporary_file_path('foo.nc')
varx = Variable('x', np.arange(5), dimensions='five', attrs={'axis': 'X'})
vary = Variable('y', np.arange(7) + 10, dimensions='seven', attrs={'axis': 'Y'})
vc = VariableCollection(variables=[varx, vary])
vc.write(path)
else:
path = None
path = MPI_COMM.bcast(path)
rd = RequestDataset(path)
dist = rd.driver.dist
distributed_dimension = dist.get_dimension('seven')
self.assertTrue(distributed_dimension.dist)
def test_system_renaming_dimensions_on_variables(self):
var1 = Variable(name='var1', value=[1, 2], dimensions='dim')
var2 = Variable(name='var2', value=[3, 4], dimensions='dim')
vc = VariableCollection(variables=[var1, var2])
path = self.get_temporary_file_path('out.nc')
vc.write(path)
rd = RequestDataset(path)
meta = rd.metadata
meta['dimensions']['new_dim'] = {'size': 2}
meta['variables']['var2']['dimensions'] = ('new_dim',)
field = rd.get()
self.assertEqual(field['var2'].dimensions[0].name, 'new_dim')
self.assertEqual(field['var2'].get_value().tolist(), [3, 4])
path2 = self.get_temporary_file_path('out2.nc')
field.write(path2)
rd2 = RequestDataset(path2)
field2 = rd2.get()
self.assertEqual(field2['var2'].dimensions[0].name, 'new_dim')
self.assertEqual(field2['var2'].get_value().tolist(), [3, 4])
def write_subsets(self, src_template, dst_template, wgt_template, index_path):
"""
Write grid subsets to netCDF files using the provided filename templates. The template must contain the full
file path with a single curly-bracer pair to insert the combination counter. ``wgt_template`` should not be a
full path. This name is used when generating weight files.
>>> template_example = '/path/to/data_{}.nc'
:param str src_template: The template for the source subset file.
:param str dst_template: The template for the destination subset file.
:param str wgt_template: The template for the weight filename.
>>> wgt_template = 'esmf_weights_{}.nc'
:param index_path: Path to the output indexing netCDF.
"""
src_filenames = []
dst_filenames = []
wgt_filenames = []
dst_slices = []
# nzeros = len(str(reduce(lambda x, y: x * y, self.nsplits_dst)))
for ctr, (sub_src, sub_dst, dst_slc) in enumerate(self.iter_src_grid_subsets(yield_dst=True), start=1):
# padded = create_zero_padded_integer(ctr, nzeros)
src_path = src_template.format(ctr)
dst_path = dst_template.format(ctr)
wgt_filename = wgt_template.format(ctr)
src_filenames.append(os.path.split(src_path)[1])
dst_filenames.append(os.path.split(dst_path)[1])
wgt_filenames.append(wgt_filename)
dst_slices.append(dst_slc)
for target, path in zip([sub_src, sub_dst], [src_path, dst_path]):
if target.is_empty:
is_empty = True
target = None
else:
is_empty = False
field = Field(grid=target, is_empty=is_empty)
ocgis_lh(msg='writing: {}'.format(path), level=logging.DEBUG)
with vm.scoped_by_emptyable('field.write', field):
if not vm.is_null:
field.write(path)
ocgis_lh(msg='finished writing: {}'.format(path), level=logging.DEBUG)
with vm.scoped('index write', [0]):
if not vm.is_null:
dim = Dimension('nfiles', len(src_filenames))
vname = ['source_filename', 'destination_filename', 'weights_filename']
values = [src_filenames, dst_filenames, wgt_filenames]
grid_splitter_destination = GridSplitterConstants.IndexFile.NAME_DESTINATION_VARIABLE
attrs = [{'esmf_role': 'grid_splitter_source'},
{'esmf_role': grid_splitter_destination},
{'esmf_role': 'grid_splitter_weights'}]
vc = VariableCollection()
grid_splitter_index = GridSplitterConstants.IndexFile.NAME_INDEX_VARIABLE
vidx = Variable(name=grid_splitter_index)
vidx.attrs['esmf_role'] = grid_splitter_index
vidx.attrs['grid_splitter_source'] = 'source_filename'
vidx.attrs[GridSplitterConstants.IndexFile.NAME_DESTINATION_VARIABLE] = 'destination_filename'
vidx.attrs['grid_splitter_weights'] = 'weights_filename'
x_bounds = GridSplitterConstants.IndexFile.NAME_X_BOUNDS_VARIABLE
vidx.attrs[x_bounds] = x_bounds
y_bounds = GridSplitterConstants.IndexFile.NAME_Y_BOUNDS_VARIABLE
vidx.attrs[y_bounds] = y_bounds
vc.add_variable(vidx)
for idx in range(len(vname)):
v = Variable(name=vname[idx], dimensions=dim, dtype=str, value=values[idx], attrs=attrs[idx])
vc.add_variable(v)
bounds_dimension = Dimension(name='bounds', size=2)
xb = Variable(name=x_bounds, dimensions=[dim, bounds_dimension], attrs={'esmf_role': 'x_split_bounds'},
dtype=int)
yb = Variable(name=y_bounds, dimensions=[dim, bounds_dimension], attrs={'esmf_role': 'y_split_bounds'},
dtype=int)
x_name = self.dst_grid.x.dimensions[0].name
y_name = self.dst_grid.y.dimensions[0].name
for idx, slc in enumerate(dst_slices):
xb.get_value()[idx, :] = slc[x_name].start, slc[x_name].stop
yb.get_value()[idx, :] = slc[y_name].start, slc[y_name].stop
vc.add_variable(xb)
vc.add_variable(yb)
vc.write(index_path)
vm.barrier()
def write_chunks(self):
"""
Write grid subsets to netCDF files using the provided filename templates. This will also generate ESMF
regridding weights for each subset if requested.
"""
src_filenames = []
dst_filenames = []
wgt_filenames = []
dst_slices = []
src_slices = []
index_path = self.create_full_path_from_template('index_file')
# nzeros = len(str(reduce(lambda x, y: x * y, self.nchunks_dst)))
ctr = 1
ocgis_lh(logger=_LOCAL_LOGGER,
msg='starting self.iter_src_grid_subsets',
level=logging.DEBUG)
for sub_src, src_slc, sub_dst, dst_slc in self.iter_src_grid_subsets(
yield_dst=True):
ocgis_lh(
logger=_LOCAL_LOGGER,
msg='finished iteration {} for self.iter_src_grid_subsets'.
format(ctr),
level=logging.DEBUG)
src_path = self.create_full_path_from_template('src_template',
index=ctr)
dst_path = self.create_full_path_from_template('dst_template',
index=ctr)
wgt_path = self.create_full_path_from_template('wgt_template',
index=ctr)
src_filenames.append(os.path.split(src_path)[1])
dst_filenames.append(os.path.split(dst_path)[1])
wgt_filenames.append(wgt_path)
dst_slices.append(dst_slc)
src_slices.append(src_slc)
# Only write destinations if an iterator is not provided.
if self.iter_dst is None:
zip_args = [[sub_src, sub_dst], [src_path, dst_path]]
else:
zip_args = [[sub_src], [src_path]]
cc = 1
for target, path in zip(*zip_args):
with vm.scoped_by_emptyable('field.write' + str(cc), target):
if not vm.is_null:
ocgis_lh(logger=_LOCAL_LOGGER,
msg='write_chunks:writing: {}'.format(path),
level=logging.DEBUG)
field = Field(grid=target)
field.write(path)
ocgis_lh(
logger=_LOCAL_LOGGER,
msg='write_chunks:finished writing: {}'.format(
path),
level=logging.DEBUG)
cc += 1
# Increment the counter outside of the loop to avoid counting empty subsets.
ctr += 1
# Generate an ESMF weights file if requested and at least one rank has data on it.
if self.genweights and len(
vm.get_live_ranks_from_object(sub_src)) > 0:
vm.barrier()
ocgis_lh(logger=_LOCAL_LOGGER,
msg='write_chunks:writing esmf weights: {}'.format(
wgt_path),
level=logging.DEBUG)
self.write_esmf_weights(src_path,
dst_path,
wgt_path,
src_grid=sub_src,
dst_grid=sub_dst)
vm.barrier()
# Global shapes require a VM global scope to collect.
src_global_shape = global_grid_shape(self.src_grid)
dst_global_shape = global_grid_shape(self.dst_grid)
# Gather and collapse source slices as some may be empty and we write on rank 0.
gathered_src_grid_slice = vm.gather(src_slices)
if vm.rank == 0:
len_src_slices = len(src_slices)
new_src_grid_slice = [None] * len_src_slices
for idx in range(len_src_slices):
for rank_src_grid_slice in gathered_src_grid_slice:
if rank_src_grid_slice[idx] is not None:
new_src_grid_slice[idx] = rank_src_grid_slice[idx]
break
src_slices = new_src_grid_slice
with vm.scoped('index write', [0]):
if not vm.is_null:
dim = Dimension('nfiles', len(src_filenames))
vname = [
'source_filename', 'destination_filename',
'weights_filename'
]
values = [src_filenames, dst_filenames, wgt_filenames]
grid_chunker_destination = GridChunkerConstants.IndexFile.NAME_DESTINATION_VARIABLE
attrs = [{
'esmf_role': 'grid_chunker_source'
}, {
'esmf_role': grid_chunker_destination
}, {
'esmf_role': 'grid_chunker_weights'
}]
vc = VariableCollection()
grid_chunker_index = GridChunkerConstants.IndexFile.NAME_INDEX_VARIABLE
vidx = Variable(name=grid_chunker_index)
vidx.attrs['esmf_role'] = grid_chunker_index
vidx.attrs['grid_chunker_source'] = 'source_filename'
vidx.attrs[GridChunkerConstants.IndexFile.
NAME_DESTINATION_VARIABLE] = 'destination_filename'
vidx.attrs['grid_chunker_weights'] = 'weights_filename'
vidx.attrs[GridChunkerConstants.IndexFile.
NAME_SRC_GRID_SHAPE] = src_global_shape
vidx.attrs[GridChunkerConstants.IndexFile.
NAME_DST_GRID_SHAPE] = dst_global_shape
vc.add_variable(vidx)
for idx in range(len(vname)):
v = Variable(name=vname[idx],
dimensions=dim,
dtype=str,
value=values[idx],
attrs=attrs[idx])
vc.add_variable(v)
bounds_dimension = Dimension(name='bounds', size=2)
# TODO: This needs to work with four dimensions.
# Source -----------------------------------------------------------------------------------------------
self.src_grid._gc_create_index_bounds_(RegriddingRole.SOURCE,
vidx, vc, src_slices,
dim, bounds_dimension)
# Destination ------------------------------------------------------------------------------------------
self.dst_grid._gc_create_index_bounds_(
RegriddingRole.DESTINATION, vidx, vc, dst_slices, dim,
bounds_dimension)
vc.write(index_path)
vm.barrier()
def create_merged_weight_file(self, merged_weight_filename, strict=False):
"""
Merge weight file chunks to a single, global weight file.
:param str merged_weight_filename: Path to the merged weight file.
:param bool strict: If ``False``, allow "missing" files where the iterator index cannot create a found file.
It is best to leave these ``False`` as not all source and destinations are mapped. If ``True``, raise an
"""
if vm.size > 1:
raise ValueError(
"'create_merged_weight_file' does not work in parallel")
index_filename = self.create_full_path_from_template('index_file')
ifile = RequestDataset(uri=index_filename).get()
ifile.load()
ifc = GridChunkerConstants.IndexFile
gidx = ifile[ifc.NAME_INDEX_VARIABLE].attrs
src_global_shape = gidx[ifc.NAME_SRC_GRID_SHAPE]
dst_global_shape = gidx[ifc.NAME_DST_GRID_SHAPE]
# Get the global weight dimension size.
n_s_size = 0
weight_filename = ifile[gidx[ifc.NAME_WEIGHTS_VARIABLE]]
wv = weight_filename.join_string_value()
split_weight_file_directory = self.paths['wd']
for wfn in map(
lambda x: os.path.join(split_weight_file_directory,
os.path.split(x)[1]), wv):
ocgis_lh(msg="current merge weight file target: {}".format(wfn),
level=logging.DEBUG,
logger=_LOCAL_LOGGER)
if not os.path.exists(wfn):
if strict:
raise IOError(wfn)
else:
continue
curr_dimsize = RequestDataset(wfn).get().dimensions['n_s'].size
# ESMF writes the weight file, but it may be empty if there are no generated weights.
if curr_dimsize is not None:
n_s_size += curr_dimsize
# Create output weight file.
wf_varnames = ['row', 'col', 'S']
wf_dtypes = [np.int32, np.int32, np.float64]
vc = VariableCollection()
dim = Dimension('n_s', n_s_size)
for w, wd in zip(wf_varnames, wf_dtypes):
var = Variable(name=w, dimensions=dim, dtype=wd)
vc.add_variable(var)
vc.write(merged_weight_filename)
# Transfer weights to the merged file.
sidx = 0
src_indices = self.src_grid._gc_create_global_indices_(
src_global_shape)
dst_indices = self.dst_grid._gc_create_global_indices_(
dst_global_shape)
out_wds = nc.Dataset(merged_weight_filename, 'a')
for ii, wfn in enumerate(
map(lambda x: os.path.join(split_weight_file_directory, x),
wv)):
if not os.path.exists(wfn):
if strict:
raise IOError(wfn)
else:
continue
wdata = RequestDataset(wfn).get()
for wvn in wf_varnames:
odata = wdata[wvn].get_value()
try:
split_grids_directory = self.paths['wd']
odata = self._gc_remap_weight_variable_(
ii,
wvn,
odata,
src_indices,
dst_indices,
ifile,
gidx,
split_grids_directory=split_grids_directory)
except IndexError as e:
msg = "Weight filename: '{}'; Weight Variable Name: '{}'. {}".format(
wfn, wvn, str(e))
raise IndexError(msg)
out_wds[wvn][sidx:sidx + odata.size] = odata
out_wds.sync()
sidx += odata.size
out_wds.close()
def write_chunks(self):
"""
Write grid subsets to netCDF files using the provided filename templates. This will also generate ESMF
regridding weights for each subset if requested.
"""
src_filenames = []
dst_filenames = []
wgt_filenames = []
dst_slices = []
src_slices = []
index_path = self.create_full_path_from_template('index_file')
# nzeros = len(str(reduce(lambda x, y: x * y, self.nchunks_dst)))
ctr = 1
ocgis_lh(logger='grid_chunker', msg='starting self.iter_src_grid_subsets', level=logging.DEBUG)
for sub_src, src_slc, sub_dst, dst_slc in self.iter_src_grid_subsets(yield_dst=True):
ocgis_lh(logger='grid_chunker', msg='finished iteration {} for self.iter_src_grid_subsets'.format(ctr),
level=logging.DEBUG)
src_path = self.create_full_path_from_template('src_template', index=ctr)
dst_path = self.create_full_path_from_template('dst_template', index=ctr)
wgt_path = self.create_full_path_from_template('wgt_template', index=ctr)
src_filenames.append(os.path.split(src_path)[1])
dst_filenames.append(os.path.split(dst_path)[1])
wgt_filenames.append(wgt_path)
dst_slices.append(dst_slc)
src_slices.append(src_slc)
# Only write destinations if an iterator is not provided.
if self.iter_dst is None:
zip_args = [[sub_src, sub_dst], [src_path, dst_path]]
else:
zip_args = [[sub_src], [src_path]]
cc = 1
for target, path in zip(*zip_args):
with vm.scoped_by_emptyable('field.write' + str(cc), target):
if not vm.is_null:
ocgis_lh(logger='grid_chunker', msg='write_chunks:writing: {}'.format(path),
level=logging.DEBUG)
field = Field(grid=target)
field.write(path)
ocgis_lh(logger='grid_chunker', msg='write_chunks:finished writing: {}'.format(path),
level=logging.DEBUG)
cc += 1
# Increment the counter outside of the loop to avoid counting empty subsets.
ctr += 1
# Generate an ESMF weights file if requested and at least one rank has data on it.
if self.genweights and len(vm.get_live_ranks_from_object(sub_src)) > 0:
vm.barrier()
self.write_esmf_weights(src_path, dst_path, wgt_path, src_grid=sub_src, dst_grid=sub_dst)
vm.barrier()
# Global shapes require a VM global scope to collect.
src_global_shape = global_grid_shape(self.src_grid)
dst_global_shape = global_grid_shape(self.dst_grid)
# Gather and collapse source slices as some may be empty and we write on rank 0.
gathered_src_grid_slice = vm.gather(src_slices)
if vm.rank == 0:
len_src_slices = len(src_slices)
new_src_grid_slice = [None] * len_src_slices
for idx in range(len_src_slices):
for rank_src_grid_slice in gathered_src_grid_slice:
if rank_src_grid_slice[idx] is not None:
new_src_grid_slice[idx] = rank_src_grid_slice[idx]
break
src_slices = new_src_grid_slice
with vm.scoped('index write', [0]):
if not vm.is_null:
dim = Dimension('nfiles', len(src_filenames))
vname = ['source_filename', 'destination_filename', 'weights_filename']
values = [src_filenames, dst_filenames, wgt_filenames]
grid_chunker_destination = GridChunkerConstants.IndexFile.NAME_DESTINATION_VARIABLE
attrs = [{'esmf_role': 'grid_chunker_source'},
{'esmf_role': grid_chunker_destination},
{'esmf_role': 'grid_chunker_weights'}]
vc = VariableCollection()
grid_chunker_index = GridChunkerConstants.IndexFile.NAME_INDEX_VARIABLE
vidx = Variable(name=grid_chunker_index)
vidx.attrs['esmf_role'] = grid_chunker_index
vidx.attrs['grid_chunker_source'] = 'source_filename'
vidx.attrs[GridChunkerConstants.IndexFile.NAME_DESTINATION_VARIABLE] = 'destination_filename'
vidx.attrs['grid_chunker_weights'] = 'weights_filename'
vidx.attrs[GridChunkerConstants.IndexFile.NAME_SRC_GRID_SHAPE] = src_global_shape
vidx.attrs[GridChunkerConstants.IndexFile.NAME_DST_GRID_SHAPE] = dst_global_shape
vc.add_variable(vidx)
for idx in range(len(vname)):
v = Variable(name=vname[idx], dimensions=dim, dtype=str, value=values[idx], attrs=attrs[idx])
vc.add_variable(v)
bounds_dimension = Dimension(name='bounds', size=2)
# TODO: This needs to work with four dimensions.
# Source -----------------------------------------------------------------------------------------------
self.src_grid._gc_create_index_bounds_(RegriddingRole.SOURCE, vidx, vc, src_slices, dim,
bounds_dimension)
# Destination ------------------------------------------------------------------------------------------
self.dst_grid._gc_create_index_bounds_(RegriddingRole.DESTINATION, vidx, vc, dst_slices, dim,
bounds_dimension)
vc.write(index_path)
vm.barrier()
def create_merged_weight_file(self, merged_weight_filename, strict=False):
"""
Merge weight file chunks to a single, global weight file.
:param str merged_weight_filename: Path to the merged weight file.
:param bool strict: If ``False``, allow "missing" files where the iterator index cannot create a found file.
It is best to leave these ``False`` as not all source and destinations are mapped. If ``True``, raise an
"""
if vm.size > 1:
raise ValueError("'create_merged_weight_file' does not work in parallel")
index_filename = self.create_full_path_from_template('index_file')
ifile = RequestDataset(uri=index_filename).get()
ifile.load()
ifc = GridChunkerConstants.IndexFile
gidx = ifile[ifc.NAME_INDEX_VARIABLE].attrs
src_global_shape = gidx[ifc.NAME_SRC_GRID_SHAPE]
dst_global_shape = gidx[ifc.NAME_DST_GRID_SHAPE]
# Get the global weight dimension size.
n_s_size = 0
weight_filename = ifile[gidx[ifc.NAME_WEIGHTS_VARIABLE]]
wv = weight_filename.join_string_value()
split_weight_file_directory = self.paths['wd']
for wfn in map(lambda x: os.path.join(split_weight_file_directory, os.path.split(x)[1]), wv):
if not os.path.exists(wfn):
if strict:
raise IOError(wfn)
else:
continue
n_s_size += RequestDataset(wfn).get().dimensions['n_s'].size
# Create output weight file.
wf_varnames = ['row', 'col', 'S']
wf_dtypes = [np.int32, np.int32, np.float64]
vc = VariableCollection()
dim = Dimension('n_s', n_s_size)
for w, wd in zip(wf_varnames, wf_dtypes):
var = Variable(name=w, dimensions=dim, dtype=wd)
vc.add_variable(var)
vc.write(merged_weight_filename)
# Transfer weights to the merged file.
sidx = 0
src_indices = self.src_grid._gc_create_global_indices_(src_global_shape)
dst_indices = self.dst_grid._gc_create_global_indices_(dst_global_shape)
out_wds = nc.Dataset(merged_weight_filename, 'a')
for ii, wfn in enumerate(map(lambda x: os.path.join(split_weight_file_directory, x), wv)):
if not os.path.exists(wfn):
if strict:
raise IOError(wfn)
else:
continue
wdata = RequestDataset(wfn).get()
for wvn in wf_varnames:
odata = wdata[wvn].get_value()
try:
split_grids_directory = self.paths['wd']
odata = self._gc_remap_weight_variable_(ii, wvn, odata, src_indices, dst_indices, ifile, gidx,
split_grids_directory=split_grids_directory)
except IndexError as e:
msg = "Weight filename: '{}'; Weight Variable Name: '{}'. {}".format(wfn, wvn, str(e))
raise IndexError(msg)
out_wds[wvn][sidx:sidx + odata.size] = odata
out_wds.sync()
sidx += odata.size
out_wds.close()