Skip to content

Function

Snowflake Docs - CREATE FUNCTION

Snowflake Docs - ALTER FUNCTION

Usage

  • A Function is made up of 2 parts
    • YAML Configuration
    • Code with the policy itself (see next sections for supported languages and associated file extensiosn)
  • All masking policies must live inside a [DATABASE]/[SCHEMA]/FUNCTIONS folder and contain both a .yml file and code file with the same name.
  • Object name = name of the yml config file within [DATABASE]/[SCHEMA]/FUNCTIONS folder
  • In Snowflake, multiple functions can share the same name with different signatures. Therefore, the signature with associated data types (no input names, just data types) must be include in the file name to make the file unique. See example sections below.

Supported Languages

Language
 File Extension
SQL .sql
JAVASCRIPT .js
PYTHON .py

Snowflake Attributes

Parameter
 Description
INPUT_ARGS (List[{name: datatype}]) - Optional
IS_SECURE (Bool) - Optional
RETURNS (String) - Optional
LANGUAGE (String) - Optional
  • Valid Values = [SQL,JAVASCRIPT,PYTHON]
NULL_HANDLING (String) - Optional
  • Valid Values = [CALLED ON NULL INPUT, RETURNS NULL ON NULL INPUT, STRICT]
COMMENT (String) - Optional
OWNER (String) - Optional
  • If HANDLE_OWNERSHIP=ERROR, be careful not to set OWNER to a role that the deployer does not have access to as it will no longer have access to manage
TAGS ({KEY:VALUE}) - Optional
GRANTS ({KEY:VALUE}) - Optional

Python Only

| IMPORTS | (List[String]) - Optional | | HANDLER | (String) - Optional | | RUNTIME_VERSION | (String) - Optional | | PACKAGES | (List[String]) - Optional |

Optional Parameter Defaults - if omitted, Snowflake defaults for parameters are used, just like creating an object manually.

Deployment attributes

Parameter
 Description
DEPLOY_ENV (List) - Optional
  • Default = []
  • List of the deployment environments to deploy to
  • Deployment environment set in the environment config yaml using the DEPLOY_ENV parameter. See Setup/Config for docs.
  • When empty or not included, deployed to all environments
DEPLOY_LOCK (Bool) - Optional
  • Default = False
  • Locks the config file to being over written by reverse engineer process.
  • Use if source code should always be source of truth and any changes pulled from database should be ignored

Folder Structure

Configuration: snowflake/data/[database name]/[schema name]/FUNCTIONS/[function name].yml

Code snowflake/data/[database name]/[schema name]/FUNCTIONS/[function name].[extension]

Example Structure

snowflake/data/CONTROL/CODE/FUNCTIONS/GET_FISCAL_YEARS().yml

snowflake/data/CONTROL/CODE/FUNCTIONS/GET_FISCAL_YEARS.sql

snowflake/data/CONTROL/CODE/FUNCTIONS/PY_PI(varchar,int).yml

snowflake/data/CONTROL/CODE/FUNCTIONS/PY_PI(varchar,int).py

This specifies the metadata for a sql function named "GET_FISCAL_YEARS" with an empty signature and a python function "PY_PI" with a signature of (varchar,int). Both within the CONTROL.CODE schema, each with their own yml config file & code based on the function name.

Samples

Config 

INPUT_ARGS:
- TABLENAME: VARCHAR
- ROLE: VARCHAR
RETURNS: TABLE (ID NUMBER, NAME VARCHAR, ROLE VARCHAR)
LANGUAGE: PYTHON
OWNER: INSTANCEADMIN
COMMENT: 
IS_SECURE: false
IMPORTS: 
HANDLER: filter_by_role
RUNTIME_VERSION: 3.8
PACKAGES: 
- snowflake-snowpark-python
TAGS: 
- {{ref('CONTROL__GOVERNANCE__ENV')}}: {{env}}
GRANTS: 
- {{role('SOME_ROLE')}}: USAGE

Code file (.py in this example) 

from snowflake.snowpark.functions import col

def filter_by_role(session, table_name, role):
   df = session.table(table_name)
   return df.filter(col("role") == role)