DRP Data Types
Custom data types can be used as Requirements and Products by Recipes. New data types can be derived as follows.
Create a new DataType
New Data Types must derive from numina.core.DataType or one of its subclasses. In the constructor, we must declare the base type of the objects ot this Data Product.
For example, a MasterBias Data Product is an image, so its base type is a DataFrame. A table of 2D coordinates will have a numpy.ndarray base type.
In general, we are interested in defining new DataTypes for objects that will contain information that will be used as inputs in different recipes. In this case, we must derive from numina.core.DataProductType.
As an example, we create a DataType that will store information about the trace of a spectrum. The information will be stored in Python dict.
class TraceMap(DataProductType):
def __init__(self, default=None):
super(TraceMap, self).__init__(dict, default)
Construction of objects
The input of a recipe is created by inspecting the Recipe Requirements. The Recipe Loader is in charge of finding an appropriated value for each requirement. The value is passed to Requirement.convert, that in turn calls DataType.convert. The default implementation just returns in input object unchanged.
Loading and Storage with the command line Recipe Loader
Each Recipe Loader can implement its own mechanism to store and load Data Products. The Command Line Recipe Loader uses text files in YAML format.
To define how a particular DataProduct is stored under the default Recipe Loader, two functions must be defined, a store function and a load function. Then thse two functions must be registered with the global functions numina.store.dump and numina.store.load.
from numina.store import dump, load
from .products import TraceMap
@dump.register(TraceMap)
def dump_tracemap(tag, obj, where):
filename = where.destination + '.yaml'
with open(filename, 'w') as fd:
yaml.dump(obj, fd)
return filename
@load.register(TraceMap)
def load_tracemap(tag, obj):
with open(obj, 'r') as fd:
traces = yaml.load(fd)
return traces
In this example, tag is an argument of type TraceMap and obj is of type dict.