User-defined functions overview | Snowflake Documentation (original) (raw)

You can write user-defined functions (UDFs) to extend the system to perform operations that are not available through thebuilt-in system-defined functions provided by Snowflake. Once you create a UDF, you can reuse it multiple times. A function always returns a value explicitly by specifying an expression, so it’s a good choice for calculating and return a value.

You can use UDFs to extend built-in functions or to encapsulate calculations that are standard for your organization. UDFs you create can be called in a way similar to built-in functions.

You write a UDF’s logic – its handler – in one of the supported languages. Once you have a handler, you can create a UDF using any of several tools included in Snowflake, thenexecute the UDF.

A UDF is like a stored procedure, but the two differ in important ways. For more information, seeChoosing whether to write a stored procedure or a user-defined function.

A UDF is just one way to extend Snowflake. For others, see the following:

• Stored procedures overview
• Writing external functions
• Snowpark API

User-defined function variations¶

You can write a UDF in one of several variations, depending on the input and output requirements your function must meet.

Supported languages and tools¶

You can create and manage UDFs (and other Snowflake entities) by using any of multiple tools, depending on how you prefer to work.

When choosing a language, consider also the following:

• Handler locations supported. Not all languages support referring to the handler on a stage (the handler code must instead be in-line). For more information, see Keeping handler code in-line or on a stage.
• Whether the handler results in a UDF that’s sharable. A sharable UDF can be used with the SnowflakeSecure Data Sharing feature.

Considerations¶

• If a query calls a UDF to access staged files, the operation fails with a user error if the SQL statement also queries a view that calls any UDF or UDTF, regardless of whether the function in the view accesses staged files or not.
• UDTFs can process multiple files in parallel; however, UDFs currently process files serially. As a workaround, group rows in a subquery using the GROUP BY clause. See Process a CSV with a UDTFfor an example.
• Currently, if staged files referenced in a query are modified or deleted while the query is running, the function call fails with an error.
• If you specify the CURRENT_DATABASE or CURRENT_SCHEMA function in the handler code of the UDF, the function returns the database or schema that contains the UDF, not the database or schema in use for the session.

UDF example¶

Code in the following example creates a UDF called addone with a handler written in Python. The handler function isaddone_py. This UDF returns an int.

CREATE OR REPLACE FUNCTION addone(i INT) RETURNS INT LANGUAGE PYTHON RUNTIME_VERSION = '3.9' HANDLER = 'addone_py' AS defaddonepy(i):returni+1def addone_py(i): return i+1defaddonep​y(i):returni+1;

Code in the following example executes the addone UDF.

Guidelines and constraints¶

Snowflake constraints:

You can ensure stability within the Snowflake environment by developing within Snowflake constraints. For more information, see Designing Handlers that Stay Within Snowflake-Imposed Constraints.

Naming:

Be sure to name functions in a way that avoids collisions with other functions. For more information, seeNaming and overloading procedures and UDFs.

Arguments:

Specify the arguments and indicate which arguments are optional. For more information, seeDefining arguments for UDFs and stored procedures.

Data type mappings:

For each handler language, there’s a separate set of mappings between the language’s data types and the SQL types used for arguments and return values. For more about the mappings for each language, see Data Type Mappings Between SQL and Handler Languages.

Handler writing¶

Handler languages:

For language-specific content on writing a handler, see Supported languages and tools.

External network access:

You can access external network locations withexternal network access. You can create secure access to specific network locations external to Snowflake, then use that access from within the handler code.

Logging, tracing, and metrics:

You can record code activity bycapturing log messages, trace events, and metrics data, storing the data in a database you can query later.

Security¶

You can grant privileges on objects needed for them to perform specific SQL actions with a UDF or UDTF. For more information, seeGranting privileges for user-defined functions

Functions share certain security concerns with stored procedures. For more information, see the following:

• You can help a procedure’s handler code execute securely by following the best practices described inSecurity Practices for UDFs and Procedures
• Ensure that sensitive information is concealed from users who should not have access to it. For more information, seeProtecting Sensitive Information with Secure UDFs and Stored Procedures

Handler code deployment¶

When creating a function, you can specify its handler – which implements the function’s logic – as code in-line with the function definition or as code external to the definition, such as code packaged and copied to a stage.

For more information, see Keeping handler code in-line or on a stage.