Druid Arbitrary Granularity Extension

The arbitrary granularity extension is used for running Druid queries over custom time intervals. Currently, Druid requires a user to aggregate over fixed time windows (month, day, 36 hours, etc.). While easy to understand, this approach is fairly inflexible. This extension hopes to solve that problem.

Installation

You can either use the prebuilt jar located in dist/ or build from source using the directions below.

Once you have the compiled jar, copy it to your druid installation and follow the including extension druid documentation. You should end up adding a line similar to this to your common.runtime.properties file:

druid.extensions.loadList=["druid-arbitrary-granularity"]

Building from source

Clone the druid repo and this line to pom.xml in the "Community extensions" section:

<module>${druid-arbitrary-granularity-src-root}/arbitrary-granularity</module>

replacing ${druid-arbitrary-granularity-src-root} with your path to this repo.

Then, inside the druid repo, run: mvn package -DskipTests=true -rf :druid-arbitrary-granularity This will build the arbitrary granularity extension and place it in ${druid-arbitrary-granularity-src-root}/arbitrary-granularity/target

Use

The arbitrary granularity takes an array of time intervals which will define the buckets used during querying. It uses the same intervals format as the normal query interval.

{
  ...
  "granularity": {
    "type": "arbitrary",
    "intervals": <Array of time intervals>
  },
  ...
}

Only events that fall in one of the defined intervals will be added to the bucket. If the query interval is longer than any of the granularity intervals, events outside the granularity intervals (but within the query interval) will be dropped. If the query interval partially covers one of the granularity intervals, only the data that is both within the query interval AND within the granularity interval will be included.

Example

Quarters that do not start at the beginning of the calendar year

Often, a business's fiscal year does not align with a calendar year. Grouping by the builtin quarter would not work since it starts with Jan 1, and using the builtin period granularity won't work since each quarter is not the same length. The arbitrary granularity extension can fix this:

Quarter start: October 1, 2015

{
  ...
  "intervals": [
    "2015-10-01/2016-01-01",
    "2016-01-01/2016-04-01",
    "2016-04-01/2016-07-01",
    "2016-07-01/2016-10-01",
  ],
  "granularity": {
    "type": "arbitrary",
    "intervals": [
      "2015-10-01/2016-01-01",
      "2016-01-01/2016-04-01",
      "2016-04-01/2016-07-01",
      "2016-07-01/2016-10-01",
    ]
  },
  ...
}

Weekday vs Weekend

Month: October 2016

{
  ...
  "intervals": [
    "2016-10-01/2016-11-05",
  ],
  "granularity": {
    "type": "arbitrary",
    "intervals": [
      "2016-10-01/2016-10-03",
      "2016-10-03/2016-10-08",
      "2016-10-08/2016-10-10",
      "2016-10-10/2016-10-15",
      "2016-10-15/2016-10-17",
      "2016-10-17/2016-10-22",
      "2016-10-22/2016-10-24",
      "2016-10-24/2016-10-29",
      "2016-10-29/2016-10-31",
      "2016-10-31/2016-11-05",
    ]
  },
  ...
}