Package org.apache.beam.sdk.extensions.sql.impl.transform.agg

Class VarianceFn<T extends Number>

java.lang.Object
org.apache.beam.sdk.transforms.Combine.CombineFn<T,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T>
org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceFn<T>
All Implemented Interfaces:
Serializable, CombineFnBase.GlobalCombineFn<T,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T>, HasDisplayData
@Internal public class VarianceFn<T extends Number> extends Combine.CombineFn<T,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T>
Combine.CombineFn for Variance on Number types.

Calculates Population Variance and Sample Variance using incremental formulas described, for example, by Chan, Golub, and LeVeque in "Algorithms for computing the sample variance: analysis and recommendations", The American Statistician, 37 (1983) pp. 242--247.

If variance is defined like this:

  • Input elements: (x[1], ... , x[n])
  • Sum of elements: {sum(x) = x[1] + ... + x[n]}
  • Average of all elements in the input: mean(x) = sum(x) / n
  • Deviation of ith element from the current mean: deviation(x, i) = x[i] - mean(n)
  • Variance: variance(x) = deviation(x, 1)^2 + ... + deviation(x, n)^2

Then variance of combined input of 2 samples (x[1], ... , x[n]) and (y[1], ... , y[m]) is calculated using this formula:

  • variance(concat(x,y)) = variance(x) + variance(y) + increment, where:
  • increment = m/(n(m+n)) * (n/m * sum(x) - sum(y))^2

This is also applicable for a single element increment, assuming that variance of a single element input is zero

To implement the above formula we keep track of the current variation, sum, and count of elements, and then use the formula whenever new element comes or we need to merge variances for 2 samples.

See Also:

  • Method Details

    • newPopulation

      public static VarianceFn newPopulation(Schema.TypeName typeName)

    • newPopulation

      public static <V extends Number> VarianceFn newPopulation(SerializableFunction<BigDecimal,V> decimalConverter)

    • newSample

      public static VarianceFn newSample(Schema.TypeName typeName)

    • newSample

      public static <V extends Number> VarianceFn newSample(SerializableFunction<BigDecimal,V> decimalConverter)

    • createAccumulator

      public org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator createAccumulator()
      Description copied from class: Combine.CombineFn
      Returns a new, mutable accumulator value, representing the accumulation of zero input values.
      Specified by:
      createAccumulator in class Combine.CombineFn<T extends Number,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T extends Number>

    • addInput

      public org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator addInput(org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator currentVariance, T rawInput)
      Description copied from class: Combine.CombineFn
      Adds the given input value to the given accumulator, returning the new accumulator value.
      Specified by:
      addInput in class Combine.CombineFn<T extends Number,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T extends Number>
      Parameters:
      currentVariance - may be modified and returned for efficiency
      rawInput - should not be mutated

    • mergeAccumulators

      public org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator mergeAccumulators(Iterable<org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator> variances)
      Description copied from class: Combine.CombineFn
      Returns an accumulator representing the accumulation of all the input values accumulated in the merging accumulators.
      Specified by:
      mergeAccumulators in class Combine.CombineFn<T extends Number,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T extends Number>
      Parameters:
      variances - only the first accumulator may be modified and returned for efficiency; the other accumulators should not be mutated, because they may be shared with other code and mutating them could lead to incorrect results or data corruption.

    • getAccumulatorCoder

      public Coder<org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator> getAccumulatorCoder(CoderRegistry registry, Coder<T> inputCoder)
      Description copied from interface: CombineFnBase.GlobalCombineFn
      Returns the Coder to use for accumulator AccumT values, or null if it is not able to be inferred.

      By default, uses the knowledge of the Coder being used for InputT values and the enclosing Pipeline's CoderRegistry to try to infer the Coder for AccumT values.

      This is the Coder used to send data through a communication-intensive shuffle step, so a compact and efficient representation may have significant performance benefits.

      Specified by:
      getAccumulatorCoder in interface CombineFnBase.GlobalCombineFn<T extends Number,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T extends Number>

    • extractOutput

      public T extractOutput(org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator accumulator)
      Description copied from class: Combine.CombineFn
      Returns the output value that is the result of combining all the input values represented by the given accumulator.
      Specified by:
      extractOutput in class Combine.CombineFn<T extends Number,org.apache.beam.sdk.extensions.sql.impl.transform.agg.VarianceAccumulator,T extends Number>
      Parameters:
      accumulator - can be modified for efficiency

    • getDefaultOutputCoder

      public Coder<T> getDefaultOutputCoder(CoderRegistry registry, Coder<T> inputCoder) throws CannotProvideCoderException
      Description copied from interface: CombineFnBase.GlobalCombineFn
      Returns the Coder to use by default for output OutputT values, or null if it is not able to be inferred.

      By default, uses the knowledge of the Coder being used for input InputT values and the enclosing Pipeline's CoderRegistry to try to infer the Coder for OutputT values.

      Specified by:
      getDefaultOutputCoder in interface CombineFnBase.GlobalCombineFn<InputT,AccumT,OutputT>
      Throws:
      CannotProvideCoderException

    • getIncompatibleGlobalWindowErrorMessage

      public String getIncompatibleGlobalWindowErrorMessage()
      Description copied from interface: CombineFnBase.GlobalCombineFn
      Returns the error message for not supported default values in Combine.globally().
      Specified by:
      getIncompatibleGlobalWindowErrorMessage in interface CombineFnBase.GlobalCombineFn<InputT,AccumT,OutputT>

    • getInputTVariable

      public TypeVariable<?> getInputTVariable()
      Returns the TypeVariable of InputT.

    • getAccumTVariable

      public TypeVariable<?> getAccumTVariable()
      Returns the TypeVariable of AccumT.

    • getOutputTVariable

      public TypeVariable<?> getOutputTVariable()
      Returns the TypeVariable of OutputT.

    • populateDisplayData

      public void populateDisplayData(DisplayData.Builder builder)
      Register display data for the given transform or component.

      populateDisplayData(DisplayData.Builder) is invoked by Pipeline runners to collect display data via DisplayData.from(HasDisplayData). Implementations may call super.populateDisplayData(builder) in order to register display data in the current namespace, but should otherwise use subcomponent.populateDisplayData(builder) to use the namespace of the subcomponent.

      By default, does not register any display data. Implementors may override this method to provide their own display data.

      Specified by:
      populateDisplayData in interface HasDisplayData
      Parameters:
      builder - The builder to populate with display data.
      See Also: