def get_between(self, dim, lower, upper):
pos = self._axis_map[dim]
ref = getattr(self, dim)
_, indices = ref.get_between(lower, upper, return_indices=True)
slc = get_reduced_slice(indices)
slc_field = [slice(None)] * 5
slc_field[pos] = slc
ret = self[slc_field]
return ret
def get_between(self,dim,lower,upper):
pos = self._axis_map[dim]
ref = getattr(self,dim)
## TODO: minor redundancy in slicing and returning dimension
new_dim,indices = ref.get_between(lower,upper,return_indices=True)
slc = get_reduced_slice(indices)
slc_field = [slice(None)]*5
slc_field[pos] = slc
ret = self[slc_field]
return(ret)
def get_subset_bbox(self,min_col,min_row,max_col,max_row,return_indices=False,closed=True):
assert(min_row <= max_row)
assert(min_col <= max_col)
if self.row is None:
r_row = self.value[0,:,:]
real_idx_row = np.arange(0,r_row.shape[0])
r_col = self.value[1,:,:]
real_idx_col = np.arange(0,r_col.shape[1])
if closed:
lower_row = r_row > min_row
upper_row = r_row < max_row
lower_col = r_col > min_col
upper_col = r_col < max_col
else:
lower_row = r_row >= min_row
upper_row = r_row <= max_row
lower_col = r_col >= min_col
upper_col = r_col <= max_col
idx_row = np.logical_and(lower_row,upper_row)
idx_col = np.logical_and(lower_col,upper_col)
keep_row = np.any(idx_row,axis=1)
keep_col = np.any(idx_col,axis=0)
row_slc = get_reduced_slice(real_idx_row[keep_row])
col_slc = get_reduced_slice(real_idx_col[keep_col])
new_mask = np.invert(np.logical_or(idx_row,idx_col)[row_slc,col_slc])
# import ipdb;ipdb.set_trace()
# idx_row_start = None
# idx_row_stop = None
# for ii in range(idx_row.shape[0]):
# if not idx_row[ii,:].any():
# continue
# else:
# if idx_row_start is None:
# idx_row_start = ii
# import ipdb;ipdb.set_trace()
#
# ## if all data is returned the indices are equivalent to the size
# ## of the array.
# if not r_value_mask.any():
# row_slc = slice(0,r_value.shape[1])
# col_slc = slice(0,r_value.shape[2])
# else:
# import ipdb;ipdb.set_trace()
else:
new_row,row_indices = self.row.get_between(min_row,max_row,return_indices=True,closed=closed)
new_col,col_indices = self.col.get_between(min_col,max_col,return_indices=True,closed=closed)
row_slc = get_reduced_slice(row_indices)
col_slc = get_reduced_slice(col_indices)
# new_uid = self.uid[row_slc,col_slc]
#
# if self._value is None:
# new_value = None
# else:
# new_value = self._value[:,row_slc,col_slc]
#
# ret = copy(self)
# ret._value = new_value
# ret.row = new_row
# ret.col = new_col
# ret.uid = new_uid
ret = self[row_slc,col_slc]
try:
ret._value.mask = new_mask
except UnboundLocalError:
if self.row is not None:
pass
else:
raise
if return_indices:
ret = (ret,(row_slc,col_slc))
return(ret)
def get_subset_bbox(self,min_col,min_row,max_col,max_row,return_indices=False,closed=True,
use_bounds=True):
assert(min_row <= max_row)
assert(min_col <= max_col)
if self.row is None:
r_row = self.value[0,:,:]
real_idx_row = np.arange(0,r_row.shape[0])
r_col = self.value[1,:,:]
real_idx_col = np.arange(0,r_col.shape[1])
if closed:
lower_row = r_row > min_row
upper_row = r_row < max_row
lower_col = r_col > min_col
upper_col = r_col < max_col
else:
lower_row = r_row >= min_row
upper_row = r_row <= max_row
lower_col = r_col >= min_col
upper_col = r_col <= max_col
idx_row = np.logical_and(lower_row,upper_row)
idx_col = np.logical_and(lower_col,upper_col)
keep_row = np.any(idx_row,axis=1)
keep_col = np.any(idx_col,axis=0)
## slice reduction may fail due to empty bounding box returns. catch
## these value errors and repurpose as subset errors.
try:
row_slc = get_reduced_slice(real_idx_row[keep_row])
except ValueError:
if real_idx_row[keep_row].shape[0] == 0:
raise(EmptySubsetError(origin='Y'))
else:
raise
try:
col_slc = get_reduced_slice(real_idx_col[keep_col])
except ValueError:
if real_idx_col[keep_col].shape[0] == 0:
raise(EmptySubsetError(origin='X'))
else:
raise
new_mask = np.invert(np.logical_or(idx_row,idx_col)[row_slc,col_slc])
else:
new_row,row_indices = self.row.get_between(min_row,max_row,return_indices=True,closed=closed,use_bounds=use_bounds)
new_col,col_indices = self.col.get_between(min_col,max_col,return_indices=True,closed=closed,use_bounds=use_bounds)
row_slc = get_reduced_slice(row_indices)
col_slc = get_reduced_slice(col_indices)
ret = self[row_slc,col_slc]
try:
grid_mask = np.zeros((2,new_mask.shape[0],new_mask.shape[1]),dtype=bool)
grid_mask[:,:,:] = new_mask
ret._value = np.ma.array(ret._value,mask=grid_mask)
ret.uid = np.ma.array(ret.uid,mask=new_mask)
except UnboundLocalError:
if self.row is not None:
pass
else:
raise
if return_indices:
ret = (ret,(row_slc,col_slc))
return(ret)
