Attaching JSON schema to KSQL stream records

Question

I've been using KSQL and so far it's been working great. But now I'd like to sink the output to BigQuery via Kafka Connect, and need to attach a JSON schema. I'm having trouble figuring out how to do this. Here's my query:

CREATE STREAM tweets_original (
      CreatedAt BIGINT,
      Id BIGINT,
      Text VARCHAR,
      Source VARCHAR,
      GeoLocation VARCHAR,
      User STRUCT<Id BIGINT, Name VARCHAR, Description VARCHAR, ScreenName VARCHAR, URL VARCHAR, FollowersCount BIGINT, FriendsCount BIGINT>
    )
    WITH (kafka_topic='tweets', value_format='JSON');

    CREATE STREAM tweets_new
    WITH (kafka_topic='tweets-new') AS
    SELECT
      CreatedAt as created_at,
      Id as tweet_id,
      Text as tweet_text,
      Source as source,
      GeoLocation as geo_location,
      User->Id as user_id,
      User->Name as user_name,
      User->Description as user_description,
      User->ScreenName as user_screenname
    FROM tweets_original ;

Here's an example of a record that was written to the output topic ( tweets-new ).

{
  "CREATED_AT": 1535036410000,
  "TWEET_ID": 1032643668614819800,
  "TWEET_TEXT": "Sample text",
  "SOURCE": "<a href=\"http://twitter.com\" rel=\"nofollow\">Twitter Web Client</a>",
  "GEO_LOCATION": null,
  "USER_ID": 123,
  "USER_NAME": "John Smith",
  "USER_DESCRIPTION": "Developer in Chief",
  "USER_SCREENNAME": "newphonewhodis"
}

However, in order for Kafka Connect to sink these records to BigQuery, I need to attach a schema, like so:

{
  "schema": {
    "type": "struct",
    "fields": [
      {
        "type": "int64",
        "optional": false,
        "field": "CREATED_AT"
      },
      {
        "type": "int64",
        "optional": false,
        "field": "TWEET_ID"
      },
      {
        "type": "string",
        "optional": false,
        "field": "TWEET_TEXT"
      }
      ...
    ],
    "optional": false,
    "name": "foobar"
  },
  "payload": {...}
}

Anyways, I'm not seeing anything thing in the docs that shows how I might approach this (maybe I'm looking in the wrong place). Any help would be greatly appreciated!

Answer 1

This is a simple solution for KSQL, just update your 2nd stream to AVRO.

CREATE STREAM tweets_new
    WITH (kafka_topic='tweets-new', value_format='AVRO') AS
    SELECT
      CreatedAt as created_at,
      Id as tweet_id,
      Text as tweet_text,
      Source as source,
      GeoLocation as geo_location,
      User->Id as user_id,
      User->Name as user_name,
      User->Description as user_description,
      User->ScreenName as user_screenname
    FROM tweets_original ;

Then in your Kafka Connect configuration, you can use the AvroConvertor and allow for schema evolution/management in Google Big Query.