Previously, we didn't differentiate between evaluation and target metrics in terms of storing and processing them as separate entities. This meant that we could not use the global target feature to shrink evaluation metrics. The global target feature means that a unique evaluation metric is a combination of unique flag + variation values, where as before the target would have been used.
This change separates out the metric types as follows:
The original single cache is used. I have not created a separate cache.
The metric types are differentiated using a subclassing approach, where we store each type in the same cache.
When we come to process the metrics, we check the type accordingly.
What
Previously, we didn't differentiate between
evaluationandtargetmetrics in terms of storing and processing them as separate entities. This meant that we could not use theglobal targetfeature to shrinkevaluationmetrics. Theglobal targetfeature means that a unique evaluation metric is a combination of uniqueflag+variationvalues, where as before thetargetwould have been used.This change separates out the metric types as follows:
Also enables global target feature.
Testing
Without
global targetfeatureWith
global targetfeature