KSQL Examples

These examples use a pageviews stream and a users table.

Tip

For in-depth tutorials and recommended deployment scenarios, see the KSQL Stream Processing Cookbook.

Creating streams

Prerequisite:
The corresponding Kafka topics must already exist in your Kafka cluster.

Create a stream with three columns on the Kafka topic that is named pageviews. It is important to instruct KSQL the format of the values that are stored in the topic. In this example, the values format is DELIMITED.

CREATE STREAM pageviews \
  (viewtime BIGINT, \
   userid VARCHAR, \
   pageid VARCHAR) \
  WITH (KAFKA_TOPIC='pageviews', \
        VALUE_FORMAT='DELIMITED');

Associating Kafka message keys: The above statement does not make any assumptions about the Kafka message key in the underlying Kafka topic. However, if the value of the message key in Kafka is the same as one of the columns defined in the stream in KSQL, you can provide such information in the WITH clause. For instance, if the Kafka message key has the same value as the pageid column, you can write the CREATE STREAM statement as follows:

CREATE STREAM pageviews \
  (viewtime BIGINT, \
   userid VARCHAR, \
   pageid VARCHAR) \
 WITH (KAFKA_TOPIC='pageviews', \
       VALUE_FORMAT='DELIMITED', \
       KEY='pageid');

Associating Kafka message timestamps: If you want to use the value of one of the columns as the Kafka message timestamp, you can provide such information to KSQL in the WITH clause. The message timestamp is used in window-based operations in KSQL (such as windowed aggregations) and to support event-time based processing in KSQL. For instance, if you want to use the value of the viewtime column as the message timestamp, you can rewrite the above statement as follows:

CREATE STREAM pageviews \
  (viewtime BIGINT, \
   userid VARCHAR, \
   pageid VARCHAR) \
  WITH (KAFKA_TOPIC='pageviews', \
        VALUE_FORMAT='DELIMITED', \
        KEY='pageid', \
        TIMESTAMP='viewtime');

Creating tables

Prerequisite:
The corresponding Kafka topics must already exist in your Kafka cluster.

Create a table with several columns. In this example, the table has columns with primitive data types, a column of array type, and a column of map type:

CREATE TABLE users \
  (registertime BIGINT, \
   gender VARCHAR, \
   regionid VARCHAR, \
   userid VARCHAR, \
   interests array<VARCHAR>, \
   contactinfo map<VARCHAR, VARCHAR>) \
  WITH (KAFKA_TOPIC='users', \
        VALUE_FORMAT='JSON', \
        KEY = 'userid');

Note that specifying KEY is required in table declaration, see Key Requirements.

Working with streams and tables

Now that you have the pageviews stream and users table, take a look at some example queries that you can write in KSQL. The focus is on two types of KSQL statements: CREATE STREAM AS SELECT (a.k.a CSAS) and CREATE TABLE AS SELECT (a.k.a CTAS). For these statements KSQL persists the results of the query in a new stream or table, which is backed by a Kafka topic.

Transforming

For this example, imagine you want to create a new stream by transforming pageviews in the following way:

  • The viewtime column value is used as the Kafka message timestamp in the new stream’s underlying Kafka topic.
  • The new stream’s Kafka topic has 5 partitions.
  • The data in the new stream is in JSON format.
  • Add a new column that shows the message timestamp in human-readable string format.
  • The userid column is the key for the new stream.

The following statement will generate a new stream, pageviews_transformed with the above properties:

CREATE STREAM pageviews_transformed \
  WITH (TIMESTAMP='viewtime', \
        PARTITIONS=5, \
        VALUE_FORMAT='JSON') AS \
  SELECT viewtime, \
         userid, \
         pageid, \
         TIMESTAMPTOSTRING(viewtime, 'yyyy-MM-dd HH:mm:ss.SSS') AS timestring \
  FROM pageviews \
  PARTITION BY userid;

Use a [ WHERE condition ] clause to select a subset of data. If you want to route streams with different criteria to different streams backed by different underlying Kafka topics, e.g. content-based routing, write multiple KSQL statements as follows:

CREATE STREAM pageviews_transformed_priority_1 \
  WITH (TIMESTAMP='viewtime', \
        PARTITIONS=5, \
        VALUE_FORMAT='JSON') AS \
  SELECT viewtime, \
         userid, \
         pageid, \
         TIMESTAMPTOSTRING(viewtime, 'yyyy-MM-dd HH:mm:ss.SSS') AS timestring \
  FROM pageviews \
  WHERE userid='User_1' OR userid='User_2' \
  PARTITION BY userid;
CREATE STREAM pageviews_transformed_priority_2 \
      WITH (TIMESTAMP='viewtime', \
            PARTITIONS=5, \
            VALUE_FORMAT='JSON') AS \
  SELECT viewtime, \
         userid, \
         pageid, \
         TIMESTAMPTOSTRING(viewtime, 'yyyy-MM-dd HH:mm:ss.SSS') AS timestring \
  FROM pageviews \
  WHERE userid<>'User_1' AND userid<>'User_2' \
  PARTITION BY userid;

Joining

When joining objects the number of partitions in each must be the same. You can use KSQL itself to create re-partitioned streams/tables as required. In this example you will join users to the pageviews_transformed topic, which has 5 partitions. First, generate a users topic with a partition count to match that of pageviews_transformed:

CREATE TABLE users_5part \
    WITH (PARTITIONS=5) AS \
    SELECT * FROM USERS;

Now you can use the following query creates a new stream by joining the pageviews_transformed stream with the users_5part table.

CREATE STREAM pageviews_enriched AS \
  SELECT pv.viewtime, \
         pv.userid AS userid, \
         pv.pageid, \
         pv.timestring, \
         u.gender, \
         u.regionid, \
         u.interests, \
         u.contactinfo \
  FROM pageviews_transformed pv \
  LEFT JOIN users_5part u ON pv.userid = u.userid;

Note that by default all the Kafka topics will be read from the current offset (aka the latest available data); however, in a stream-table join, the table topic will be read from the beginning.

Aggregating, windowing, and sessionization

Watch the screencast of Aggregations in KSQL on YouTube.

Now assume that you want to count the number of pageviews per region. Here is the query that would perform this count:

CREATE TABLE pageviews_per_region AS \
  SELECT regionid, \
         count(*) \
  FROM pageviews_enriched \
  GROUP BY regionid;

The above query counts the pageviews from the time you start the query until you terminate the query. Note that we used CREATE TABLE AS SELECT statement here since the result of the query is a KSQL _table_. The results of aggregate queries in KSQL are always a table because it computes the aggregate for each key (and possibly for each window per key) and updates these results as it processes new input data.

KSQL supports aggregation over WINDOW too. Let’s rewrite the above query so that we compute the pageview count per region every 1 minute:

CREATE TABLE pageviews_per_region_per_minute AS \
  SELECT regionid, \
         count(*) \
  FROM pageviews_enriched \
  WINDOW TUMBLING (SIZE 1 MINUTE) \
  GROUP BY regionid;

If you want to count the pageviews for only “Region_6” by female users for every 30 seconds, you can change the above query as the following:

CREATE TABLE pageviews_per_region_per_30secs AS \
  SELECT regionid, \
         count(*) \
  FROM pageviews_enriched \
  WINDOW TUMBLING (SIZE 30 SECONDS) \
  WHERE UCASE(gender)='FEMALE' AND LCASE(regionid)='region_6' \
  GROUP BY regionid;

UCASE and LCASE functions in KSQL are used to convert the values of gender and regionid columns to upper and lower case, so that you can match them correctly. KSQL also supports LIKE operator for prefix, suffix and substring matching.

KSQL supports HOPPING windows and SESSION windows too. The following query is the same query as above that computes the count for hopping window of 30 seconds that advances by 10 seconds:

CREATE TABLE pageviews_per_region_per_30secs10secs AS \
  SELECT regionid, \
         count(*) \
  FROM pageviews_enriched \
  WINDOW HOPPING (SIZE 30 SECONDS, ADVANCE BY 10 SECONDS) \
  WHERE UCASE(gender)='FEMALE' AND LCASE (regionid) LIKE '%_6' \
  GROUP BY regionid;

The next statement counts the number of pageviews per region for session windows with a session inactivity gap of 60 seconds. In other words, you are sessionizing the input data and then perform the counting/aggregation step per region.

CREATE TABLE pageviews_per_region_per_session AS \
  SELECT regionid, \
         count(*) \
  FROM pageviews_enriched \
  WINDOW SESSION (60 SECONDS) \
  GROUP BY regionid;

Working with arrays and maps

The interests column in the users table is an array of strings that represents the interest of each user. The contactinfo column is a string-to-string map that represents the following contact information for each user: phone, city, state, and zipcode.

Tip

If you are using ksql-datagen, you can use quickstart=users_ to generate data that include the interests and contactinfo columns.

The following query will create a new stream from pageviews_enriched that includes the first interest of each user along with the city and zipcode for each user:

CREATE STREAM pageviews_interest_contact AS \
  SELECT interests[0] AS first_interest, \
         contactinfo['zipcode'] AS zipcode, \
         contactinfo['city'] AS city, \
         viewtime, \
         userid, \
         pageid, \
         timestring, \
         gender, \
         regionid \
  FROM pageviews_enriched;