def read(self, indices=None, axis=0, indexing=None, tol=None, keepdims=False):
""" Read values from disk
Parameters
----------
indices : int or list or slice (single-dimensional indices)
or a tuple of those (multi-dimensional)
or `dict` of { axis name : axis values }
axis : None or int or str, optional
if specified and indices is a slice, scalar or an array, assumes
indexing is along this axis.
indexing : {'label', 'position'}, optional
Indexing mode.
- "label": indexing on axis labels (default)
- "position": use numpy-like position index
Default value can be changed in dimarray.rcParams['indexing.by']
tol : float, optional
tolerance when looking for numerical values, e.g. to use nearest
neighbor search, default `None`.
keepdims : bool, optional
keep singleton dimensions (default False)
Returns
-------
DimArray instance or scalar
See Also
--------
DimArray.take, DatasetOnDisk.read, DimArrayOnDisk.write
"""
if type(indices) is slice and indices == slice(None) and self.ndim == 0:
indices = ()
return AbstractDimArray._getitem(self, indices=indices, axis=axis,
indexing=indexing, tol=tol, keepdims=keepdims)
def write(self, indices, values, axis=0, indexing=None, tol=None):
""" Write numpy array or DimArray to netCDF file (The
variable must have been created previously)
Parameters
----------
indices : int or list or slice (single-dimensional indices)
or a tuple of those (multi-dimensional)
or `dict` of { axis name : axis values }
values : np.ndarray or DimArray
axis : None or int or str, optional
if specified and indices is a slice, scalar or an array, assumes
indexing is along this axis.
indexing : {'label', 'position'}, optional
Indexing mode.
- "label": indexing on axis labels (default)
- "position": use numpy-like position index
Default value can be changed in dimarray.rcParams['indexing.by']
tol : float, optional
tolerance when looking for numerical values, e.g. to use nearest
neighbor search, default `None`.
See Also
--------
DimArray.put, DatasetOnDisk.write, DimArrayOnDisk.read
"""
# just create the variable and dimensions
ds = self._ds
dima = values # internal convention: values is a numpy array
values, nctype, cf_attrs = maybe_encode_values(values, format=self._ds.file_format)
assert self._name in ds.variables.keys(), "variable does not exist, should have been created earlier!"
# add attributes
if hasattr(dima,'attrs'):
self.attrs.update(dima.attrs)
self.attrs.update(cf_attrs) # calendar?
# special case: index == slice(None) and self.ndim == 0
# This would fail with numpy, but not with netCDF4
if type(indices) is slice and indices == slice(None) and self.ndim == 0:
indices = ()
indices = self._get_indices(indices,axis=axis, indexing=indexing, tol=tol)
# Perform additional checks on axes if the Data to assign is a DimArray
if isinstance(dima, DimArray):
for i, ax in enumerate(self.axes):
idx = indices[i]
axis = dima.axes[ax.name]
# write unlimited dimensions
if self._ds.dimensions[ax.name].isunlimited():
self.axes[ax.name][idx] = axis
else:
# dimension variable already written, simple check
ondisk = self.axes[ax.name][idx if not (np.isscalar(idx) or np.ndim(idx)==0) else [idx]].values
inmemory = axis.values
# inmemory, _ = maybe_encode_values(axis.values)
if not np.all(ondisk == inmemory):
# assert np.all(ondisk == inmemory)
warnings.warn("axes values differ in the netCDF file, try using a numpy array instead, or re-index\
the dimarray prior to assigning to netCDF", RuntimeWarning)
# do all the indexing and assignment via IndexedArray class
# ==> it will set the values via _setvalues_ortho below
# super(DimArrayOnDisk)._setitem(self, indices, values, **kwargs)
AbstractDimArray._setitem(self, indices, values, indexing='position')