Skip to main content
Version: Java (Groovy)

EmMin

EmMin creates an EM Min (exponential moving minimum) for an updateBy table operation. The formula for an EM Min is:

a=e−dtτa = e^{\frac{-dt}{\tau}}

ϵn=min((a∗ϵn−1),x)\epsilon_{n} = min((a * \epsilon_{n-1}), x)

Where:

  • τ\tau is the window size, an input parameter to the method.
  • ϵ\epsilon is the EM Min.
  • xx is the current value.
  • nn denotes the step. The current step is nn, and the previous step is n−1n - 1.

Syntax​

EmMin(tickDecay, pairs...)
EmMin(control, tickDecay, pairs...)
EmMin(control, timestampColumn, timeDecay, pairs...)
EmMin(control, timestampColumn, durationDecay, pairs...)
EmMin(timestampColumn, timeDecay, pairs...)
EmMin(timestampColumn, durationDecay, pairs...)

Parameters​

ParameterTypeDescription
tickDecaydouble

The decay rate in ticks (rows).

pairsString...

The input/output column name pairs.

controlOperationControl

Defines how special cases should behave. If not given, default OperationControl settings are used.

timestampColumnString

The column in the source table to use for timestamps.

timeDecaylong

The decay rate in nanoseconds.

durationDecayDuration

The decay rate in a Duration object.

Returns​

An UpdateByOperation to be used in an updateBy table operation.

Examples​

One column, no groups​

The following example calculates the tick-based and time-based EM Max of the X column, renaming the resultant column to EmMin_X. The tick decay rate is set to 5 rows, and the time decay rate is set to 5 seconds. No grouping columns are specified, so the EM Max is calculated for all rows.

baseTime = parseInstant("2023-01-01T00:00:00 ET")

source = emptyTable(10).update("Timestamp = baseTime + i *  SECOND", "Letter = (i % 2 == 0) ? `A` : `B`", "X = randomInt(0,25)")

result = source.updateBy([EmMin(5, "EmMin_Tick_X = X"), EmMin("Timestamp", 5 * SECOND, "EmMin_Time_X = X")])

One EM Max column, one grouping column​

The following example builds on the previous by specifying Letter as the key column. Thus, the EM Max is calculated on a per-letter basis.

baseTime = parseInstant("2023-01-01T00:00:00 ET")

source = emptyTable(10).update("Timestamp = baseTime + i *  SECOND", "Letter = (i % 2 == 0) ? `A` : `B`", "X = randomInt(0,25)")

result = source.updateBy([EmMin(5, "EmMin_Tick_X = X"), EmMin("Timestamp", 5 * SECOND, "EmMin_Time_X = X")], "Letter")

Multiple EM Max columns, multiple grouping columns​

The following example builds on the previous by calculating the EM Max of multiple columns with each UpdateByOperation. Also, the groups are defined by unique combinations of letter and boolean in the Letter and Truth columns, respectively.

baseTime = parseInstant("2023-01-01T00:00:00 ET")

source = emptyTable(20).update("Timestamp = baseTime + i *  SECOND", "Letter = (i % 2 == 0) ? `A` : `B`", "Truth = randomBool()", "X = randomInt(0, 25)", "Y = randomInt(0, 25)")

result = source.updateBy([EmMin(2, "EmMin_Tick_X = X", "EmMin_Tick_Y = Y"), EmMin("Timestamp", 3 * SECOND, "EmMin_Time_X = X", "EmMin_Time_Y = Y")], "Letter", "Truth")

Multiple UpdateByOperations, multiple grouping columns​

The following example builds on the previous by calculating the tick- and time-based EM Max of the X and Y columns using different EM Max UpdateByOperations. This allows each EM Max to have its own decay rate. The decay rates are reflected in the renamed resultant columns.

baseTime = parseInstant("2023-01-01T00:00:00 ET")

source = emptyTable(20).update("Timestamp = baseTime + i *  SECOND", "Letter = (i % 2 == 0) ? `A` : `B`", "Truth = randomBool()", "X = randomInt(0, 25)", "Y = randomInt(0, 25)")

emminTickX = EmMin(1, "EmMin_Tick_X_1row = X")
emminTickY = EmMin(5, "EmMin_Tick_Y_5rows = Y")
emminTimeX = EmMin("Timestamp", 2 * SECOND, "EmMin_Time_X_2sec = X")
emminTimeY = EmMin("Timestamp", 4 * SECOND, "EmMin_Time_Y_4sec = Y")

result = source.updateBy([emminTickX, emminTickY, emminTimeX, emminTimeY], "Letter", "Truth")