On this page

Show all

Help Center/ MapReduce Service/ User Guide (Paris Region)/ Overview/ Components/ Flink/ Flink Enhanced Open Source Features/ Flink CEP in SQL

Flink CEP in SQL

Updated on 2024-10-11 GMT+08:00

View PDF

Flink CEP in SQL

Flink allows users to represent complex event processing (CEP) query results in SQL for pattern matching and evaluate event streams on Flink engines.

SQL Query Syntax

CEP SQL is implemented through the MATCH_RECOGNIZE SQL syntax. The MATCH_RECOGNIZE clause is supported by Oracle SQL since Oracle Database 12c and is used to indicate event pattern matching in SQL. Apache Calcite also supports the MATCH_RECOGNIZE clause.

Flink uses Calcite to analyze SQL query results. Therefore, this operation complies with the Apache Calcite syntax.

MATCH_RECOGNIZE (
      [ PARTITION BY expression [, expression ]* ]
      [ ORDER BY orderItem [, orderItem ]* ]
      [ MEASURES measureColumn [, measureColumn ]* ]
      [ ONE ROW PER MATCH | ALL ROWS PER MATCH ]
      [ AFTER MATCH
            ( SKIP TO NEXT ROW
            | SKIP PAST LAST ROW
            | SKIP TO FIRST variable
            | SKIP TO LAST variable
            | SKIP TO variable )
      ]
      PATTERN ( pattern )
      [ WITHIN intervalLiteral ]
      [ SUBSET subsetItem [, subsetItem ]* ]
      DEFINE variable AS condition [, variable AS condition ]*
      )

The syntax elements of the MATCH_RECOGNIZE clause are defined as follows:

(Optional) -PARTITION BY: defines partition columns. This clause is optional. If this parameter is not defined, the parallelism 1 is used.

(Optional) -ORDER BY: defines the sequence of events in a data flow. The ORDER BY clause is optional. If it is ignored, non-deterministic sorting is used. Since the order of events is important in pattern matching, this clause should be specified in most cases.

(Optional) -MEASURES: specifies the attribute value of the successfully matched event.

(Optional) -ONE ROW PER MATCH | ALL ROWS PER MATCH: defines how to output the result. ONE ROW PER MATCH indicates that only one row is output for each matching. ALL ROWS PER MATCH indicates that one row is output for each matching event.

(Optional) -AFTER MATCH: specifies the start position for processing after the next pattern is successfully matched.

-PATTERN: defines the matching pattern as a regular expression. The following operators can be used in the PATTERN clause: join operators, quantifier operators (*, +, ?, {n}, {n,}, {n,m}, and {,m}), branch operators (vertical bar |), and differential operators ('{- -}').

(Optional) -WITHIN: outputs a pattern clause match only when the match occurs within the specified time.

(Optional) -SUBSET: combines one or more associated variables defined in the DEFINE clause.

-DEFINE: specifies the Boolean condition, which defines the variables used in the PATTERN clause.

In addition, the MATCH_RECOGNIZE clause supports the following functions:

-MATCH_NUMBER(): Used in the MEASURES clause to allocate the same number to each row that is successfully matched.

-CLASSIFIER(): Used in the MEASURES clause to indicate the mapping between matched rows and variables.

-FIRST() and LAST(): Used in the MEASURES clause to return the value of the expression evaluated in the first or last row of the row set mapped to the schema variable.

-NEXT() and PREV(): Used in the DEFINE clause to evaluate an expression using the previous or next row in a partition.

-RUNNING and FINAL keywords: Used to determine the semantics required for aggregation. RUNNING can be used in the MEASURES and DEFINE clauses, whereas FINAL can be used only in the MEASURES clause.

- Aggregate functions (COUNT, SUM, AVG, MAX, MIN): Used in the MEASURES and DEFINE clauses.

Query Example

The following query finds the V-shaped pattern in the stock price data flow.

SELECT *
    FROM MyTable
    MATCH_RECOGNIZE (
      ORDER BY rowtime  
      MEASURES
        STRT.name as s_name,
        LAST(DOWN.name) as down_name,
        LAST(UP.name) as up_name
      ONE ROW PER MATCH
      PATTERN (STRT DOWN+ UP+)
      DEFINE
        DOWN AS DOWN.v < PREV(DOWN.v),
        UP AS UP.v > PREV(UP.v) 
    )

In the following query, the aggregate function AVG is used in the MEASURES clause of SUBSET E consisting of variables related to A and C.

SELECT * 
    FROM Ticker 
    MATCH_RECOGNIZE ( 
      MEASURES
        AVG(E.price) AS avgPrice 
      ONE ROW PER MATCH 
      AFTER MATCH SKIP PAST LAST ROW 
      PATTERN (A B+ C) 
      SUBSET E = (A,C) 
      DEFINE 
        A AS A.price < 30, 
        B AS B.price < 20,
        C AS C.price < 30 
    )

Parent topic: Flink Enhanced Open Source Features

Previous topic: Stream SQL Join

Next topic: Flume

Feedback

Was this page helpful?

Helpful Not helpful

Provide feedback

Thank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.

The system is busy. Please try again later.

Which of the following issues have you encountered?

Content is inconsistent with the product UI

Unclear descriptions

Lack of examples or code

Incorrect steps

Can't find what I need

Lack of best practices

Feedback (optional)

0/500

Select at least one type of issue, and enter your comments or suggestions.

Enter a maximum of 500 characters.

Submit Cancel

For any further questions, feel free to contact us through the chatbot.

Chatbot