Join Event Streams with ksqlDB for Confluent Platform

You can use ksqlDB to merge streams of events in real time by using the JOIN statement, which has a SQL join syntax. A ksqlDB join and a relational database join are similar in that they both combine data from two or more sources based on common values. The result of a ksqlDB join is a new stream or table that’s populated with the column values that you specify in a SELECT statement.

With ksqlDB, you don’t need to write the low-level logic around joining streams and tables, so you can focus on the business logic for combining your streaming data.

You can join streams and tables in these ways:

• Join multiple streams to create a new stream.
• Join multiple tables to create a new table.
• Join multiple streams and tables to create a new stream.

JOIN Clause

The ksqlDB JOIN clause has the familiar syntax of a SQL JOIN clause. The following example creates a pageviews_enriched stream, which is a combination of a pageviews stream and a users table:

CREATE STREAM pageviews_enriched AS
  SELECT
     users.userid AS userid,
     pageid,
     regionid,
     gender
  FROM pageviews
    LEFT JOIN users ON pageviews.userid = users.userid
  EMIT CHANGES;

Note

Currently, the join expression must be a single column equal comparison. Non-equi joins, or joins on multiple columns are not supported.

When you join two streams, you must specify a WITHIN clause for matching records that both occur within a specified time interval. For valid time units, see Time Units.

Here’s an example stream-stream-stream join that combines orders, payments and shipments streams. The resulting shipped_orders stream contains all orders paid within 1 hour of when the order was placed, and shipped within 2 hours of the payment being received.

CREATE STREAM shipped_orders AS
  SELECT
     o.id as orderId,
     o.itemid as itemId,
     s.id as shipmentId,
     p.id as paymentId
  FROM orders o
     INNER JOIN payments p WITHIN 1 HOURS ON p.id = o.id
     INNER JOIN shipments s WITHIN 2 HOURS ON s.id = o.id;

Joins and Windows

ksqlDB enables grouping records that have the same key for stateful operations, like joins, into windows. A window has a grace period, which means that a record can arrive out-of-order for a specified period of time, and it’s still joined correctly based on its timestamp.

If a record arrives after the window’s grace period has passed, the record is discarded and isn’t processed.

For more information on grace periods, see Out-of-order events.

Note

Only stream-stream joins are windowed.

Windows are tracked per record key. In join operations, ksqlDB uses a windowing state store to store all of the records received so far within the defined window boundary. Old records in the state store are purged after the window grace period passed.

For more information on windows, see Windows in ksqlDB Queries.

Join Requirements

Your ksqlDB applications must meet specific requirements for joins to be successful.

Co-partitioned data

For most joins, input data must be co-partitioned when joining. This ensures that input records with the same key, from both sides of the join, are delivered to the same stream task during processing. For some cases, ksqlDB may repartition the data automatically, if required. But there are some cases for which it’s your responsibility to ensure data co-partitioning when joining. The only join that does not require co-partitioning is a foreign-key table-table join. For more information, see Partition Data to Enable Joins.

Join Capabilities

ksqlDB supports a large set of join operations for streams and tables, including INNER, LEFT OUTER, and FULL OUTER. Frequently, LEFT OUTER is shortened to LEFT JOIN, and FULL OUTER is shortened to OUTER JOIN.

The following table shows the supported combinations.

Type	Windowed?	INNER	LEFT OUTER	RIGHT OUTER	FULL OUTER
Stream-Stream	Windowed	Supported	Supported	Supported	Supported
Table-Table	Non-windowed	Supported	Supported	Supported	Supported
Stream-Table	Non-windowed	Supported	Supported	Not Supported	Not Supported

Stream-Stream Joins

ksqlDB supports INNER, LEFT OUTER, RIGHT OUTER, and FULL OUTER joins between streams.

All of these operations support out-of-order records.

To join two streams, you must specify a windowing scheme by using the WITHIN clause. A new input record on one side produces a join output for each matching record on the other side, and there can be multiple such matching records within a join window.

Joins cause data re-partitioning of a stream only if the stream was marked for re-partitioning. If both streams are marked, both are re-partitioned.

Important

Kafka guarantees the relative order of any two messages from one source partition only if they are also both in the same partition after the repartition. Otherwise, Kafka is likely to interleave messages. The use case will determine if these ordering guarantees are acceptable.

LEFT OUTER joins will contain leftRecord-NULL records in the result stream, which means that the join contains NULL values for fields selected from the right-hand stream where no match is made.

RIGHT OUTER joins will contain rightRecord-NULL records in the result stream, which means that the join contains NULL values for fields selected from the left-hand stream where no match is made.

FULL OUTER joins will contain leftRecord-NULL or NULL-rightRecord records in the result stream, which means that the join contains NULL values for fields coming from a stream where no match is made.

Semantics of Stream-Stream Joins

The semantics of the various stream-stream join variants are shown in the following table. In the table, each row represents a new incoming record. The following assumptions apply:

• All records have the same key.
• All records are processed in timestamp order.
• The join window is 15 seconds, and the grace period is 5 seconds.

When new input is received, the join is triggered under the conditions listed in the table. Input records with a NULL key or a NULL value are ignored and don’t trigger the join.

Important

If you don’t specify a grace period, left/outer join results are emitted eagerly, and the observed result might differ from the result shown below.

Timestamp	Left stream	Right stream	INNER JOIN	LEFT JOIN	RIGHT JOIN	OUTER JOIN
1	null
2		null
3	A
4		a	[A, a]	[A, a]	[A, a]	[A, a]
5	B		[B, a]	[B, a]	[B, a]	[B, a]
6		b	[A, b], [B, b]	[A, b], [B, b]	[A, b], [B, b]	[A, b], [B, b]
7	null
8		null
9	C		[C, a], [C, b]	[C, a], [C, b]	[C, a], [C, b]	[C, a], [C, b]
10		c	[A, c], [B, c], [C, c]	[A, c], [B, c], [C, c]	[A, c], [B, c], [C, c]	[A, c], [B, c], [C, c]
11		null
12	null
13		null
14		d	[A, d], [B, d], [C, d]	[A, d], [B, d], [C, d]	[A, d], [B, d], [C, d]	[A, d], [B, d], [C, d]
15	D		[D, a], [D, b], [D, c], [D, d]	[D, a], [D, b], [D, c], [D, d]	[D, a], [D, b], [D, c], [D, d]	[D, a], [D, b], [D, c], [D, d]
…
40	E
…
60	F			[E,null]		[E,null]
…
80		f		[F,null]	[F,null]	[F,null]
…
100	G				[null,f]	[null,f]

Stream-Table Joins

ksqlDB supports INNER and LEFT joins between a stream and a table. An OUTER join is not available because it can’t be defined with sound semantics.

Stream-table joins are always non-windowed joins. You can perform table lookups against a table when a new record arrives on the stream. Only events arriving on the stream side trigger downstream updates and produce join output. Updates on the table side don’t produce updated join output.

Stream-table joins cause data re-partitioning of the stream only if the stream was marked for re-partitioning.

Important

ksqlDB provides best-effort on time synchronization, but there are no guarantees, which can cause missing results or leftRecord-NULL results.

Semantics of Stream-Table Joins

The semantics of the various stream-table join variants are shown in the following table. In the table, each row represents a new incoming record. The following assumptions apply:

• All records have the same key.
• All records are processed in timestamp order.

Only input records for the left-side stream trigger the join. Input records for the right-side table update only the internal right-side join state.

Input records for the table with a NULL value are interpreted as tombstones for the corresponding key, which indicate the deletion of the key from the table. Tombstones don’t trigger the join.

Timestamp	Left Stream	Right Table	INNER JOIN	LEFT JOIN
1	null
2		null (tombstone)
3	A			[A, null]
4		a
5	B		[B, a]	[B, a]
6		b
7	null
8		null (tombstone)
9	C			[C, null]
10		c
11		null
12	null
13		null
14		d
15	D		[D, d]	[D, d]

Notice that the INNER JOIN doesn’t result in any output if the table-side does not already contain a value for the key, even if the table-side is later populated. For the LEFT JOIN the same scenario results in an output of leftRecord-NULL. It is therefore important that the table data is loaded before the stream event is received.

ksqlDB attempts to process both sides of a join in event-time order, but it can’t offer strong guarantees, especially in the presence of out-of-order rows.

To maximise join predictability, ensure historic table data is available in the source topic, the query is running, and ksqlDB has had enough time to process the table data before starting to produce to your stream.

Table-Table Joins

ksqlDB supports primary-key (1:1) as well as foreign-key (1:N) joins between tables. Many-to-many (N:M) joins are not supported currently. For a foreign-key join, you can use any left table column in the join condition to join it with the primary-key of the right table.

For primary-key joins INNER, LEFT OUTER, and FULL OUTER joins are supported. For foreign-key joins INNER and LEFT OUTER joins are supported.

Table-table joins are always non-windowed joins.

Table-table joins are eventually consistent.

Important

ksqlDB provides best-effort on time synchronization, but there are no guarantees, which can cause missing results or leftRecord-NULL results.

Semantics of Primary-Key Table-Table Joins

The semantics of the various 1:1 table-table join variants are shown in the following table. In the table, each row represents a new incoming record. The following assumptions apply:

• All records have the same key.
• All records are processed in timestamp order.

Input records with a null value are interpreted as tombstones for the corresponding key, which indicate the deletion of the key from the table. When an input tombstone is received, an output tombstone may be forwarded to the join result table, if the corresponding join result exists in the result table.

Timestamp	Left Table	Right Table	INNER JOIN	LEFT JOIN	RIGHT JOIN	OUTER JOIN
1	null (tombstone)
2		null (tombstone)
3	A			[A, null]		[A, null]
4		a	[A, a]	[A, a]	[A, a]	[A, a]
5	B		[B, a]	[B, a]	[B, a]	[B, a]
6		b	[B, b]	[B, b]	[B, b]	[B, b]
7	null (tombstone)		null (tombstone)	null (tombstone)	[null, b]	[null, b]
8		null (tombstone)			null (tombstone)	null (tombstone)
9	C			[C, null]		[C, null]
10		c	[C, c]	[C, c]	[C, c]	[C, c]
11		null (tombstone)	null (tombstone)	[C, null]	null (tombstone)	[C, null]
12	null (tombstone)			null (tombstone)		null (tombstone)
13		null (tombstone)
14		d			[null, d]	[null, d]
15	D		[D, d]	[D, d]	[D, d]	[D, d]

Semantics of Foreign-Key Table-Table Joins

The semantics of the various 1:N table-table join variants are shown in the following table. In the table, each row represents a new incoming record. The following assumptions apply:

• The primary-key of the left and output table is denoted in upper-case letters.
• The foreign-key from the left table and the primary-key of the right table is denoted in lower-case letters.
• All records are processed in timestamp order.

Input records with a null value are interpreted as tombstones for the corresponding key, which indicate the deletion of the key from the table. When an input tombstone is received, one or multiple output tombstones may be forwarded to the join result table, if a corresponding join result exists in the result table.

Timestamp	Left Table	Right Table	INNER JOIN	LEFT JOIN
1	[A] null (tombstone)
2		[a] null (tombstone)
3	[A,a,1]			[A,a,1,NULL,NULL]
4		[a,10]	[A,a,1,a,10]	[A,a,1,a,10]
5		[a,11]	[A,a,1,a,11]	[A,a,1,a,11]
6	[A,a,2]		[A,a,2,a,11]	[A,a,2,a,11]
7	[B,a,3]		[B,a,3,a,11]	[B,a,3,a,11]
8		[a,12]	[A,a,2,a,12], [B,a,3,a,12]	[A,a,2,a,12], [B,a,3,a,12]
9		[c,13]
10	[C,c,4]		[C,c,4,c,13]	[C,c,4,c,13]
11	[C] null (tombstone)		[C] null (tombstone)	[C] null (tombstone)
12		[c] null (tombstone)
13		[a] null (tombstone)	[A] null (tombstone), [B] null (tombstone)	[A,a,2,NULL,NULL], [B,a,3,NULL,NULL]
14	[B] null (tombstone)			[B] null (tombstone)

N-Way Joins

ksqlDB supports joining more than two sources in a single statement. These joins are semantically equivalent to joining N sources consecutively, and the order of the joins is controlled by the order in which the joins are written.

Consider the following query as an example, where A is a stream of events and B and C are both tables:

CREATE STREAM joined AS
  SELECT *
  FROM A
    JOIN B ON A.id = B.product_id
    JOIN C ON A.id = C.purchased_id;

The output of this query is a stream, and the intermediate join result would is the stream A ⋈ B. If C were a stream instead of a table, you would rewrite the join accordingly, by adding a WITHIN clause because joining A ⋈ B with C is a stream-stream join:

CREATE STREAM joined AS
  SELECT *
  FROM A
    JOIN B ON A.id = B.product_id
    JOIN C WITHIN 10 SECONDS ON A.id = C.purchased_id;

Limitations of N-Way Joins

• The limitations and restrictions described in the previous sections to each intermediate step in N-way joins. For example, FULL OUTER joins between streams and tables are not supported. This means that if any stage in the N-way join resolves to a FULL OUTER join between a stream and a table the entire query fails:

  --- This JOIN fails with the following exception:
  --- Join between invalid operands requested: left type: KTABLE, right type: KSTREAM
  CREATE STREAM joined AS
    SELECT *
    FROM A
      JOIN B WITHIN 10 SECONDS ON A.id = B.product_id
      FULL OUTER JOIN C ON A.id = C.purchased_id;
  

• Although ksqlDB supports N-way joins in general, it’s not possible to have a FULL OUTER N-way table join in a single query. You must break the query into multiple queries, with each doing a single FULL OUTER table-table join.