com.linkedin.feathr.common

Interface FeatureExtractor<SOURCE_DATA>

All Known Subinterfaces:

ParameterizedFeatureExtractor<SOURCE_DATA,PARAM>

@Experimental
public interface FeatureExtractor<SOURCE_DATA>

A feature extractor that extracts FeatureValue from source data. The class must have 0-argument constructor so Feathr engine can instantiate the class with 0-argument constructor. We expect this to replace Extractor eventually. WARNING: This interface is in beta testing and is not approved for public use yet.

Method Summary

Modifier and Type	Method and Description
java.util.Map<java.lang.String,FeatureValue>	batchExtract(java.util.Set<java.lang.String> featuresToExtract, SOURCE_DATA sourceData) Extract the set of features specified by featuresToExtract from the sourceData.
default FeatureValue	extract(java.lang.String featureName, SOURCE_DATA sourceData) Extract a single feature specified by featureName from sourceData.

Method Detail

extract

default FeatureValue extract(java.lang.String featureName,
                             SOURCE_DATA sourceData)

Extract a single feature specified by featureName from sourceData. If you only extract a single feature a time and it can be more performant than batchExtract(java.util.Set<java.lang.String>, SOURCE_DATA), you should implement this.

Parameters:

featureName - Feature name in the Feathr config.

sourceData - The source data from your Feathr Source.

Returns:

The FeatureValue associated with the requested feature name. When the requested feature cannot be extracted, null may be returned.

batchExtract

java.util.Map<java.lang.String,FeatureValue> batchExtract(java.util.Set<java.lang.String> featuresToExtract,
                                                          SOURCE_DATA sourceData)

Extract the set of features specified by featuresToExtract from the sourceData.

Parameters:

featuresToExtract - A set of features that this extractor will extract.

sourceData - The source data from your Feathr Source.

Returns:

A map of feature name to FeatureValue. For requested features that cannot be extracted, the expectation is that the result map will not contain entries of those features.