Descriptor

The Descriptors interface lets users define the properties and relationships between the features being used in their models or other data-intensive workflows.

Descriptor supertypes define some general properties that can be overloaded for specific subtypes as required, such as the tables to load or the quantitative variable to use.

Users may define their own Descriptor` type system for their own purposes but must extend the following methods:

• sources: The sources for the data needed to build the Desriptor feature.
• quantity_key: The quantitative variable for the Descriptor.
• categorical_keys: The categorical variables for the Descriptor.

categorical_keys(::Descriptor) -> Vector

Returns the categorical variables for the Descriptor, e.g. [:colour, :shape]. Returns an empty vector if no categorical variables exist.

label(D::Descriptor) -> Symbol

Returns a symbol representation of the Descriptor.

parents(D::Descriptor) -> Vector

Returns the parent Descriptor for each table names returned from sources. Similar to Base.parent, we return input D when there is no parent for a source.

quantity_key(::Descriptor) -> Symbol

Returns the quantitative variable for the Descriptor, e.g. :temperature.

sources(::Descriptor) -> Vector{String}

Returns the data sources for the Descriptor, e.g. the names or paths to the tables containing the required data: ["temperature.csv", "locations.csv"].

FeatureDescriptors.TestUtils

Provides a fake Descriptor for testing purposes and a test_interface function to check that any Descriptor subtype implements the expected API.

DerivedFakeDescriptor <: Descriptor

A fake derived Descriptor for testing purposes only.

FakeDescriptor <: Descriptor

A fake Descriptor for testing purposes only.

OtherFakeDescriptor <: Descriptor

Another fake Descriptor for testing purposes only.

TestUtils.test_interface(D::Type{<:Descriptor})

Test that a subtype of Descriptor implements the expected API.