Sketch cache key should include size, isInputThetaSketch.

[gianm]

No description provided.

[himanshug]

@gianm getCacheKey() is a query time concept and not having those two fields in the cache key should be OK? It was intentional to not have those in cache key.
Did you encounter any issues there?

[gianm]

@himanshug wouldn't the size of the sketch affect query results? especially if this is run over a field which at query time is a string, not a sketch metric.

[himanshug]

@gianm yeah , you are right, size would have impact on accuracy, i stand corrected. but isInputThetaSketch shouldn't change anything at query time.

[gianm]

I was thinking that isInputThetaSketch = false on a sketch metric should succeed, and isInputThetaSketch = true on a string column should fail, so if someone changes it from false to true that should fail (rather than pull from cache)

[gianm]

same for the reverse.

[himanshug]

isInputThetaSketch is only useful while ingestion if your input data was sketches... at query time it plays no role.
at query time , users would just leave it be default.

[gianm]

@himanshug ah I see, you use instanceof at query time to determine the type, not isInputThetaSketch. ok, that should be removed from the cache key. will do another pr.

[gianm]

#2899