Question about compound metrics and storage?


Background for my question: when raw data is normalized, the normalized version of the data is persisted as another copy (e.g. to Cassandra).

With compound metrics - is any storage utilized ? e.g. are the values for compound metrics computed on the fly with each invocation from metrics or is there anything stored to avoid re-computing compound metrics each time they are accessed?



It is not persisted except if you decide to cache the metric.
However a nice functionality avoids to re-compute the same underlying metric if it is involved several times in several expressions:

For example:

If I do MyType.evalMetrics(expressions: ["2A + 5", 3Aexp(-6A)]) the metric engine will parse and decompose each expression and only evaluate A onceā€¦