User defined analytics

Declarative stateful math expressions evaluated in real-time

Objective

The analytics processor provides a rich set of features that enable developers to define and execute analytics. This feature is idea for executing an event based analytic expression using attributes present within the event, linked reference data and passed variables.

Currently only Javascript and Python are supported

Processor architecture

The key analytic processor objective is to provide analytic deployment flexibility for the developer. The ambition is to enable a quick test, learn and refine cycle that reduces time to production deployment.

Processor overview

Execution models

This section covers methods for executing expressions in stream processing, focusing on three key approaches:

1. Expression only Execute independent event-based calculations using simple mathematical formulas by defining core attributes.

2. Expression and state Incorporate constants and manage state to initialise calculations and update them with new data.

3. Expression and script Leverage pre-existing scripts alongside expressions for more complex calculations.

See User Defined Scripts documentation for script based execution models.

Expression only

When you only want to execute an expression that can be defined as a mathematical formula use this method to execute independent event based calculations provide the expression along with the required core attributes.

analytic:
  expression: (bid - ask) / scalar
  assign to: new_event_variable

Expression and state

This example demonstrates how you would use provided constants for a calculation. This same method can be used to prime a calculation with a starting value for the initial calculation and then being it replaced with updated values by using the same assign to variable.

See Stateful variables for further documentation.

analytic:
  expression: "(ask + (bid - ask) / 2.0) / scaled_price"
  assign to: scaled_price
  variables:
      scaled_price : 120.21
  stateful:
      assign to: scaled_price
      memory capacity: 1

Expression and script

Use this option when you have pre-existing scripts which you want to leverage within a stream processing context.

analytic:
  script: ./scripts/js/preExistingFunctions.js
  expression: bid - squareRoot(`${ask}` / `${bid}`)
  assign to: new_event_variable

Javascript function

The example will apply the bid and ask event attributes to the expression.

Note the current implementation requires the script to be provided using the js extension due to the way the expression is defined.

Javascript

export function squareRoot(num) {
    if (num < 0) return NaN;
    return Math.sqrt(num);
}

Python

import math

def square_root(num):
    if num < 0:
        return float('nan')  # Return NaN for negative input
    return math.sqrt(num)  # Return the square root for non-negative input

Variables

Variables are provided as a map of key and numerical values.

Variables and the event are passed in to the execution context per event, including stateful variables. This example demonstrates how to apply a constant as a scaling factor for a calculation.

analytic:  
  expression: (highest_price - lowest_price) / scaling_factor
  assign to: scaled_price
  variables:
      scaling_factor : 120.21 

Stateful variables

Stateful variables provide a key function whereby the previous value(s) can be used within the current calculation context.

The example primes the first calculation with a starting value for the initial calculation and thereafter replaced with the computed values from the stateful memory. This is achieved using the same assign to variable set within the variables section.

analytic: 
  expression: (ask + (bid - ask) / 2.0) / Math.avg(scaled_price)
  assign to: scaled_price
  variables:
      scaled_price : 100.00
  stateful:
      assign to: scaled_price
      memory capacity: 25  

Attributes schema

The follow attributes are used for expression and scripting.

Attribute	Description	Data Type	Required
language	Language runtime to use to execute required execution definitions. Currently Javascript (js) and Python (python) are supported	String Default: js
assign to	Assignment variable for the evaluated expression result	String Default: result
variables	Map of constants and seed values for the expression to use. This is an optional variable.	Map<String, Number>
stateful	Store previous computed values for next computation cycle	See stateful section
assignment datatype	Data type to cast too for the assignment variable. See scripting supported data types section	DataType Default: Double

Execution attributes schema

These attributes define what is need to execute analytical functions over a stream events.

Attribute	Description	Data Type
expression	Math expression without an assignment variable. Required if method has not been provided.	String
script	Path of the script to loaded within the Joule processing context	String

Stateful attributes schema

Store previous computed values for next computation cycle. Honours the groupby definition. The previous computed value can be used on the next calculation.

Attribute	Description	Data Type	Required
assign to	Assignment variable for the result	String
memory capacity	Number of rolling elements to store within a FIFO array	Integer Default: 10
attach memory	Attach a copy of the array contents to the computed event	Boolean Default: false