datasqrl/sqrl/implement-udf

Implement User Defined Function

Implement New UDF

Only implement a UDF if the functionality cannot be implemented with existing Flink SQL or DataSQRL functions. Look up existing functions first.

Follow these steps:

• From the project root directory, initialize function with /opt/agent/cmd.sh add-func <function-name> to create a scalar UDF. Append --aggregate to create an aggregate function.
• New function implementation is added in the functions/ folder with the name <function-name>.java
• The generated file declares //DEPS org.apache.flink:flink-table-common:<version> with the compiler's Flink version. Keep this line: the UDF does not compile without it.
• Import into SQRL script with IMPORT functions.<function-name> or IMPORT functions.* for all functions. functions/ sits at the project root, so from a script in a subfolder use the reserved root prefix — IMPORT root.functions.<function-name> — since SQRL relative dot-paths cannot climb upward. (Inside an included project, keep the import relative so it travels with the copied files.)

Edit UDF

Follow these steps:

• Edit the .java file in the functions/ folder
• Try to use standard java libraries
• Use JBANG //DEPS to include dependencies, next to the generated flink-table-common one
• compile, test and run build the UDFs with JBang. A build failure stops the command with [FATAL] Failed to build JBang UDF(s) [...] followed by the javac output: fix the reported file and line before anything else
• Implement tests for the function in the main method
• Execute the main method after each edit to ensure function compiles and works correctly