def _cast(val, schema):
val = _unmake_node(val)
if val is not None:
if 'datatype' in schema:
# Handle lazy array
if isinstance(val, ndarray.NDArrayType):
val = val._make_array()
dtype = ndarray.asdf_datatype_to_numpy_dtype(schema['datatype'])
val = util.gentle_asarray(val, dtype)
if dtype.fields is not None:
val = _as_fitsrec(val)
if 'ndim' in schema and len(val.shape) != schema['ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected {0}, got {1}".format(
schema['ndim'], len(val.shape)))
if 'max_ndim' in schema and len(val.shape) > schema['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected <= {0}, got {1}".format(
schema['max_ndim'], len(val.shape)))
if isinstance(val, np.generic) and np.isscalar(val):
val = np.asscalar(val)
return val
def _cast(val, schema):
val = _unmake_node(val)
if val is not None:
if 'datatype' in schema:
# Handle lazy array
if isinstance(val, ndarray.NDArrayType):
val = val._make_array()
dtype = ndarray.asdf_datatype_to_numpy_dtype(schema['datatype'])
val = util.gentle_asarray(val, dtype)
if dtype.fields is not None:
val = _as_fitsrec(val)
if 'ndim' in schema and len(val.shape) != schema['ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected {0}, got {1}".
format(schema['ndim'], len(val.shape)))
if 'max_ndim' in schema and len(val.shape) > schema['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected <= {0}, got {1}"
.format(schema['max_ndim'], len(val.shape)))
if isinstance(val, np.generic) and np.isscalar(val):
val = np.asscalar(val)
return val
def _get_hdu_type(hdu_name, schema=None, value=None):
hdu_type = None
if hdu_name in (0, 'PRIMARY'):
hdu_type = fits.PrimaryHDU
elif schema is not None:
dtype = ndarray.asdf_datatype_to_numpy_dtype(schema['datatype'])
if dtype.fields is not None:
hdu_type = fits.BinTableHDU
elif value is not None:
if hasattr(value, 'dtype') and value.dtype.names is not None:
hdu_type = fits.BinTableHDU
return hdu_type
def _get_hdu_type(hdu_name, schema=None, value=None):
hdu_type = None
if hdu_name in (0, 'PRIMARY'):
hdu_type = fits.PrimaryHDU
elif schema is not None:
dtype = ndarray.asdf_datatype_to_numpy_dtype(schema['datatype'])
if dtype.fields is not None:
hdu_type = fits.BinTableHDU
elif value is not None:
if hasattr(value, 'dtype') and value.dtype.names is not None:
hdu_type = fits.BinTableHDU
return hdu_type
def fill_hdu(subschema, path, combiner, shape, recurse):
if 'fits_hdu' not in subschema:
return
dtype = subschema.get('datatype')
if dtype is None:
return
ndim = subschema.get('ndim')
if ndim and ndim != len(shape):
return
dtype = ndarray.asdf_datatype_to_numpy_dtype(dtype)
keyword = '.'.join(path)
default = subschema.get('default', 0.0)
im[keyword] = np.full(shape, default, dtype=dtype)
def fill_hdu(subschema, path, combiner, shape, recurse):
if 'fits_hdu' not in subschema:
return
dtype = subschema.get('datatype')
if dtype is None:
return
ndim = subschema.get('ndim')
if ndim and ndim != len(shape):
return
dtype = ndarray.asdf_datatype_to_numpy_dtype(dtype)
keyword = '.'.join(path)
default = subschema.get('default', 0.0)
im[keyword] = np.full(shape, default, dtype=dtype)
def _cast(val, schema):
val = _unmake_node(val)
if val is not None:
if 'datatype' in schema:
val = util.gentle_asarray(
val, ndarray.asdf_datatype_to_numpy_dtype(schema['datatype']))
if 'ndim' in schema and len(val.shape) != schema['ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected {0}, got {1}".
format(schema['ndim'], len(val.shape)))
if 'max_ndim' in schema and len(val.shape) > schema['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected <= {0}, got {1}"
.format(schema['max_ndim'], len(val.shape)))
tag = schema.get('tag')
if tag is not None:
val = tagged.tag_object(tag, val)
return val
def _cast(val, schema):
val = _unmake_node(val)
if val is not None:
if 'datatype' in schema:
val = util.gentle_asarray(
val, ndarray.asdf_datatype_to_numpy_dtype(schema['datatype']))
if 'ndim' in schema and len(val.shape) != schema['ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected {0}, got {1}".format(
schema['ndim'], len(val.shape)))
if 'max_ndim' in schema and len(val.shape) > schema['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected <= {0}, got {1}".format(
schema['max_ndim'], len(val.shape)))
tag = schema.get('tag')
if tag is not None:
val = tagged.tag_object(tag, val)
return val
def _make_default_array(attr, schema, ctx):
dtype = schema.get('datatype')
if dtype is not None:
dtype = ndarray.asdf_datatype_to_numpy_dtype(dtype)
ndim = schema.get('ndim', schema.get('max_ndim'))
default = schema.get('default', None)
primary_array_name = ctx.get_primary_array_name()
if attr == primary_array_name:
if ctx.shape is not None:
shape = ctx.shape
elif ndim is not None:
shape = tuple([0] * ndim)
else:
shape = (0,)
else:
if dtype.names is not None:
if ndim is None:
shape = (0,)
else:
shape = tuple([0] * ndim)
default = None
else:
has_primary_array_shape = False
if primary_array_name is not None:
primary_array = getattr(ctx, primary_array_name, None)
has_primary_array_shape = primary_array is not None
if has_primary_array_shape:
if ndim is None:
shape = primary_array.shape
else:
shape = primary_array.shape[-ndim:]
elif ndim is None:
shape = (0,)
else:
shape = tuple([0] * ndim)
array = np.empty(shape, dtype=dtype)
if default is not None:
array[...] = default
return array
def _make_default_array(attr, schema, ctx):
dtype = schema.get('datatype')
if dtype is not None:
dtype = ndarray.asdf_datatype_to_numpy_dtype(dtype)
ndim = schema.get('ndim', schema.get('max_ndim'))
default = schema.get('default', None)
primary_array_name = ctx.get_primary_array_name()
if attr == primary_array_name:
if ctx.shape is not None:
shape = ctx.shape
elif ndim is not None:
shape = tuple([0] * ndim)
else:
shape = (0,)
else:
if dtype.names is not None:
if ndim is None:
shape = (0,)
else:
shape = tuple([0] * ndim)
default = None
else:
has_primary_array_shape = False
if primary_array_name is not None:
primary_array = getattr(ctx, primary_array_name, None)
has_primary_array_shape = primary_array is not None
if has_primary_array_shape:
if ndim is None:
shape = primary_array.shape
else:
shape = primary_array.shape[-ndim:]
elif ndim is None:
shape = (0,)
else:
shape = tuple([0] * ndim)
array = np.empty(shape, dtype=dtype)
if default is not None:
array[...] = default
return array
def _cast(val, schema):
val = _unmake_node(val)
if val is None:
return None
if 'datatype' in schema:
# Handle lazy array
if isinstance(val, ndarray.NDArrayType):
val = val._make_array()
if (_is_struct_array_schema(schema) and len(val) and
(_is_struct_array_precursor(val) or _is_struct_array(val))):
# we are dealing with a structured array. Because we may
# modify schema (to add shape), we make a deep copy of the
# schema here:
schema = copy.deepcopy(schema)
for t, v in zip(schema['datatype'], val[0]):
if not isinstance(t, Mapping):
continue
aval = np.asanyarray(v)
shape = aval.shape
val_ndim = len(shape)
# make sure that if 'ndim' is specified for a field,
# it matches the dimensionality of val's field:
if 'ndim' in t and val_ndim != t['ndim']:
raise ValueError(
"Array has wrong number of dimensions. "
"Expected {}, got {}".format(t['ndim'], val_ndim)
)
if 'max_ndim' in t and val_ndim > t['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. "
"Expected <= {}, got {}".format(t['max_ndim'], val_ndim)
)
# if shape of a field's value is not specified in the schema,
# add it to the schema based on the shape of the actual data:
if 'shape' not in t:
t['shape'] = shape
dtype = ndarray.asdf_datatype_to_numpy_dtype(schema['datatype'])
val = util.gentle_asarray(val, dtype)
if dtype.fields is not None:
val = _as_fitsrec(val)
if 'ndim' in schema and len(val.shape) != schema['ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected {}, got {}"
.format(schema['ndim'], len(val.shape)))
if 'max_ndim' in schema and len(val.shape) > schema['max_ndim']:
raise ValueError(
"Array has wrong number of dimensions. Expected <= {}, got {}"
.format(schema['max_ndim'], len(val.shape)))
if isinstance(val, np.generic) and np.isscalar(val):
val = val.item()
return val
