def broadcast_apply(cls, axis, apply_func, left, right):
client = _get_global_client()
right_parts = np.squeeze(right)
if len(right_parts.shape) == 0:
right_parts = np.array([right_parts.item()])
assert (
len(right_parts.shape) == 1
), "Invalid broadcast partitions shape {}\n{}".format(
right_parts.shape, [[i.get() for i in j] for j in right_parts]
)
return np.array(
[
[
PandasOnDaskFramePartition(
client.submit(
deploy_func,
part.future,
right_parts[col_idx].future
if axis
else right_parts[row_idx].future,
apply_func,
part.call_queue,
right_parts[col_idx].call_queue
if axis
else right_parts[row_idx].call_queue,
pure=False,
)
)
for col_idx, part in enumerate(left[row_idx])
]
for row_idx in range(len(left))
]
)
def deploy(cls, func, num_returns, kwargs):
client = _get_global_client()
remote_task_future = client.submit(func, **kwargs)
return [
client.submit(lambda l, i: l[i], remote_task_future, i)
for i in range(num_returns)
]
def get_indices(cls, axis, partitions, index_func):
"""This gets the internal indices stored in the partitions.
Note: These are the global indices of the object. This is mostly useful
when you have deleted rows/columns internally, but do not know
which ones were deleted.
Args:
axis: This axis to extract the labels. (0 - index, 1 - columns).
index_func: The function to be used to extract the function.
old_blocks: An optional previous object that this object was
created from. This is used to compute the correct offsets.
Returns:
A Pandas Index object.
"""
client = _get_global_client()
ErrorMessage.catch_bugs_and_request_email(not callable(index_func))
func = cls.preprocess_func(index_func)
if axis == 0:
# We grab the first column of blocks and extract the indices
new_idx = (
[idx.apply(func).future for idx in partitions.T[0]]
if len(partitions.T)
else []
)
else:
new_idx = (
[idx.apply(func).future for idx in partitions[0]]
if len(partitions)
else []
)
new_idx = client.gather(new_idx)
return new_idx[0].append(new_idx[1:]) if len(new_idx) else new_idx
def data_array(self):
""" Lazy load model/analysis data into memory.
"""
client = _get_global_client()
xr_data = xr.open_mfdataset(self.path_to_files,
chunks=self.chunks,
parallel=True)
if not all(x in list(xr_data.coords) for x in self.DIMS):
xr_data = xr_data.rename({
'latitude': 'lat',
'longitude': 'lon',
})
if isinstance(xr_data.time.values[0], cftime._cftime.DatetimeNoLeap):
datetime_index = xr_data.indexes['time'].to_datetimeindex()
xr_data['time'] = datetime_index
if self.subset_dict is not None:
xr_data = self.cut(xr_data)
print('Cut data')
if self.season is not None:
xr_data = xr_data.where(xr_data.time.dt.season == self.season,
drop=True)
if self.rescale_longitude is True:
xr_data = xr_data.assign_coords(lon=(((xr_data.lon + 180) % 360) -
180)).sortby('lon')
return xr_data[self.DIMS + [self.temp_var]].squeeze()
def __init__(self, future, length=None, width=None, call_queue=None):
self.future = future
if call_queue is None:
call_queue = []
self.call_queue = call_queue
self._length_cache = length
self._width_cache = width
self.client = _get_global_client()
def _get_client(client):
if client is None:
return _get_global_client()
elif isinstance(client, Client):
return client
else:
# e.g., connection string
return Client(client)
def _determine_worker(self):
try:
get_worker()
self.worker = True
self.fs = filesystem(self.protocol, **self.storage_options)
except ValueError:
self.worker = False
self.client = _get_global_client()
self.rfs = dask.delayed(self)
def put(cls, obj):
"""A factory classmethod to format a given object.
Args:
obj: An object.
Returns:
A `RemotePartitions` object.
"""
client = _get_global_client()
return cls(client.scatter(obj))
def _column_widths(self):
"""Compute the column widths if they are not cached.
Returns:
A list of column widths.
"""
client = _get_global_client()
if self._column_widths_cache is None:
self._column_widths_cache = client.gather([
obj.apply(lambda df: len(df.columns)).future
for obj in self._partitions[0]
])
return self._column_widths_cache
def _row_lengths(self):
"""Compute the row lengths if they are not cached.
Returns:
A list of row lengths.
"""
client = _get_global_client()
if self._row_lengths_cache is None:
self._row_lengths_cache = client.gather([
obj.apply(lambda df: len(df)).future
for obj in self._partitions.T[0]
])
return self._row_lengths_cache
def deploy_axis_func(cls, axis, func, num_splits, kwargs,
maintain_partitioning, *partitions):
client = _get_global_client()
axis_result = client.submit(PandasFrameAxisPartition.deploy_axis_func,
axis, func, num_splits, kwargs,
maintain_partitioning, *partitions)
if num_splits == 1:
return axis_result
# We have to do this to split it back up. It is already split, but we need to
# get futures for each.
return [
client.submit(lambda l: l[i], axis_result)
for i in range(num_splits)
]
def deploy_func_between_two_axis_partitions(
cls, axis, func, num_splits, len_of_left, kwargs, *partitions
):
client = _get_global_client()
axis_result = client.submit(
PandasFrameAxisPartition.deploy_func_between_two_axis_partitions,
axis,
func,
num_splits,
len_of_left,
kwargs,
*partitions
)
# We have to do this to split it back up. It is already split, but we need to
# get futures for each.
return [client.submit(lambda l: l[i], axis_result) for i in range(num_splits)]
def broadcast_apply(cls, axis, apply_func, left, right, other_name="r"):
def mapper(df, others):
other = pandas.concat(others, axis=axis ^ 1)
return apply_func(df, **{other_name: other})
client = _get_global_client()
return np.array([[
PandasOnDaskFramePartition(
client.submit(
deploy_func,
part.future,
mapper,
part.call_queue,
[obj[col_idx].call_queue for obj in right]
if axis else [obj.call_queue for obj in right[row_idx]],
*([obj[col_idx].future for obj in right]
if axis else [obj.future for obj in right[row_idx]]),
pure=False,
)) for col_idx, part in enumerate(left[row_idx])
] for row_idx in range(len(left))])
def get_indices(cls, axis, partitions, index_func):
"""
This gets the internal indices stored in the partitions.
Parameters
----------
axis : 0 or 1
This axis to extract the labels (0 - index, 1 - columns).
partitions : NumPy array
The array of partitions from which need to extract the labels.
index_func : callable
The function to be used to extract the function.
Returns
-------
Index
A Pandas Index object.
Notes
-----
These are the global indices of the object. This is mostly useful
when you have deleted rows/columns internally, but do not know
which ones were deleted.
"""
client = _get_global_client()
ErrorMessage.catch_bugs_and_request_email(not callable(index_func))
func = cls.preprocess_func(index_func)
if axis == 0:
# We grab the first column of blocks and extract the indices
new_idx = ([idx.apply(func).future for idx in partitions.T[0]]
if len(partitions.T) else [])
else:
new_idx = ([idx.apply(func).future
for idx in partitions[0]] if len(partitions) else [])
new_idx = client.gather(new_idx)
return new_idx[0].append(new_idx[1:]) if len(new_idx) else new_idx
def dataset(self):
client = _get_global_client()
if client is None:
print(f'WARNING! No Dask client available in environment!')
_full_dataset = xr.open_mfdataset(self.path_to_files,
chunks=self.chunks,
concat_dim='time',
preprocess=self.preprocess_mf)
self.year_range = np.unique(_full_dataset.time.dt.year)[[0, -1]]
if self.season == 'DJF':
try:
_full_dataset['time'] = _full_dataset.indexes[
'time'].normalize()
except AttributeError:
_full_dataset['time'] = _full_dataset.indexes[
'time'].to_datetimeindex().normalize()
_dataset = self.sel_winters(_full_dataset, *self.year_range)
return _dataset
elif season == 'all':
return _full_dataset
else:
raise NotImplementedError
if threading.current_thread().name == "MainThread":
ray.init(
redirect_output=True,
include_webui=False,
redirect_worker_output=True,
ignore_reinit_error=True,
)
num_cpus = ray.global_state.cluster_resources()["CPU"]
except AssertionError:
pass
elif execution_engine == "Dask":
from distributed.client import _get_global_client
if threading.current_thread().name == "MainThread":
# initialize the dask client
client = _get_global_client()
if client is None:
from distributed import Client
client = Client()
num_cpus = sum(client.ncores().values())
elif execution_engine != "Python":
raise ImportError(
"Unrecognized execution engine: {}.".format(execution_engine))
DEFAULT_NPARTITIONS = max(4, int(num_cpus))
__all__ = [
"DataFrame",
"Series",
"read_csv",
def materialize(cls, future):
client = _get_global_client()
return client.gather(future)
