gistfile1.md

Proposed format for specifying synthetic data

This file should be edited freely.

Scenario: Find an event with a value

In this scenario, we want to see if the event "Score" ever happens with the property "{points_earned: 10}" in the properties hash. If we find one, we want to add a new event, "User scored max points".

{
  rule: {
    type: "match",
    pattern: [
      {
        event: "Score",
        points_earned: 10
      }
    ]
  },
  result: {
    clone: "pattern[0]",
    event: "User scored max points",
    properties: null
  }
}

This example looks a little more complicated than it might otherwise be, because it is doing two things:

1. Using a generalizable "pattern" field that takes an array, which will be useful for finding multiple events, and
2. Using a "clone" key, because we would like to create a "complete" event with all the other fields intact, such as session-id, application-id and timestamp.

The above rule would find any event with "event: Score, properties: {value: 10}". Because the Rails application "flattens" the properties column, we can refer to "value: 10" directly.

It then creates a new event which first clones the found event (pattern[0], which refers to the first captured event in the pattern array), then replaces the "event" property with our new event, and removes the "properties" property.

Scenario: Find any combination of properties in a single event

The above pattern searches "event" and "points_earned", but we can search any combination of properties.

{
  rule: {
    type: "match",
    pattern: [
      {
        activity: "Spring and mass",
        action: "exported"
        type: "model"
      }
    ]
  },
  result: {
    clone: "pattern[0]",
    event: "User exported model",
    properties: null
  }
}

Scenario: Find multiple events in a row

Here we want to find n events that happen in sequence. We use the pattern's array to specify multiple events that we want to capture in a row

{
  rule: {
    type: "match",
    pattern: [
      {
        event: "Ended challenge"
      },
      {
        event: "Score",
        points_earned: 10
      },
      {
        event: "Restarted challenge"
      }
    ]
  },
  result: {
    clone: "pattern[2]",
    event: "User scored max points and then restarted",
    properties: null
  }
}

Here when we find the three events in order (for two of them we care only about the event name), we clone the last one (pattern[2]) and create the new event.

Note that, for Geniverse, finding three events in order is a fairly naive search -- events that happen close to each other, like a user finishing a challenge and then earning a score for it, are not guaranteed to arrive in a specified order. For this reason, we might want to use a time-based filter. However, it sounds like this filter would still be useful for IS data.