NoData Utilities

The utils.nodata module provides two utility functions for working with NoData values.

Default NoData

By default, Raster objects try to ensure they have a NoData value. So if you create a Raster object from a dataset that lacks a NoData value (for example, from a raw numpy array), then the object will assign a default NoData value based on the raster’s dtype. Default values are assigned as follows:

Dtype	Default NoData
Float	nan
Signed integer	Lowest represetable value
Unsigned integer	Highest representable value
bool	False
Anything else	None

You can use the nodata.default function to query the default value for specific dtypes:

>>> from pfdf.utils import nodata

>>> nodata.default('float32')
nan
>>> nodata.default('int64')
-9223372036854775808
>>> nodata.default('int16')
-32768
>>> nodata.default('uint16')
65535
>>> nodata.default(bool)
False

Data Mask

Tip

It is usually preferable to use the Raster.data_mask or Raster.nodata_mask properties, rather than this function.

The mask function returns a NoData mask or data mask, given an array and a NoData value. In a NoData mask, NoData elements are marked as True:

>>> from pfdf.utils import nodata
>>> nodata.mask([1,2,3], nodata=2)
array([False,  True, False])

A data mask is the opposite - data elements are marked as True. Use the “invert” option to return a data mask instead:

>>> nodata.mask([1,2,3], invert=True)
array([ True, False,  True])