match_recognize — Regular Expressions Over Rows

Use Cases

The match_recognize clause is useful to implement the following use cases:

• Finding series of consecutive events (see slides)

• Pattern matching: trend reversal, periodic events, …

• Top-N per Group (see slides)

Compatibility

Row pattern matching was introduced by SQL:2016 with three optional features: R010 is the basis: this feature requires the match_recognize clause as explained above with the aggregate functions min, max, sum, count, avg. R020 allows using patterns in the over clause to define frames. R030 supports all aggregate functions for row patterns (e.g. stddev_pop, …).

JDBC and ORA-17041 Missing IN or OUT parameter

The character ? is used as placeholder in JDBC but is also a valid literal in the pattern clause (e.g., for non-greedy matching). Per default, the Oracle JDBC driver treats each ? as bind parameter. To use ? as literal in the pattern clause, the Oracle JDBC driver supports two methods:

Statement.setEscapeProcessing(false)

This is a standard JDBC feature that disables any client side interpretation of the SQL string. It doesn’t work for PreparedStatements and thus disallows the use of bind variables.

{\?\}

The Oracle JDBC driver support a proprietary escape syntax for the question mark: {\?\}. This character sequence can be used in the pattern clause instead of a question mark. This method works for PreparedStatements.