Preparing a local folder with configured Snowflake Native App artifacts¶

Create a local folder with configured artifacts¶

The snow app bundle command creates a local directory in your project, populates it with the file structure you specified in the project definition file, and generates CREATE FUNCTION or CREATE PROCEDURE declarations in Snowflake Native App setup scripts from Snowpark Python code that includes decorators (such as @sproc or @udaf). For more information, see the Snowpark Python documentation corresponding to your chosen function decorator, such as snowflake.snowpark.functions.udaf.

The snow app deploy and snow app run commands already use this functionality. However, now with an explicit snow app bundle command at your disposal, you can explore this directory before it gets uploaded to the stage, to verify the artifacts were created as expected.

To create a local folder with the configured artifacts, do the following:

1. Create or verify your Snowflake snowflake.yml project definition file, such as:

   definition_version: 2
   entities:
     codegen_nativeapp_pkg:
       type: application package
       manifest: root_files/_manifest.yml
       artifacts:
         - src: root_files/README.md
           dest: README.md
         - src: root_files/_manifest.yml
           dest: manifest.yml
         - src: root_files/setup_scripts/*
           dest: setup_scripts/
         - src: python/user_gen/echo.py
           dest: user_gen/echo.py
         - src: python/cli_gen/*
           dest: cli_gen/
           processors:
             - snowpark
     codegen_nativeapp_pkg:
       type: application
       from:
         target: codegen_nativeapp_pkg
   

   Copy

2. From your project directory, run the snow app bundle command to create the temporary output/deploy directory that contains your configured artifacts.

   snow app bundle
   

   Copy

3. Verify the contents of the output or deploy directory match the rules you specified in the snowflake.yml. file. If you invoked Snowpark annotation processing in your Python files, you can see the generated code in the amended setup script in the directory.

For more information, see the snow app bundle command.

Generate SQL code using Snowpark annotation processing¶

As a Snowflake Native App developer with a limited SQL background, you might find it cumbersome to write and maintain setup scripts, which can get quite large and complicated over time. Setup scripts contain all the application logic that a customer can use with their data, and hence are a required part of developing a Snowflake Native App. One of the core components of setup scripts is your ability to use Snowpark Python extension functions for functions and stored procedures. In addition to writing Snowpark code in Python, Java, or other Snowpark supported languages, you need to write the corresponding portions of those functions and procedures using SQL in the setup script.

For example, you could create a basic function and stored procedure using Snowpark Python, as shown:

# Example python file "echo.py" that a developer writes

def echo_fn(data):
    return 'echo_fn: ' + data

def echo_proc(session, data):
    return 'echo_proc: ' + data

You would then need to upload the file to a stage and refer to it from the setup script SQL code, similar to the following:

-- Sample setup_script.sql SQL file for a Snowflake Native App

CREATE APPLICATION ROLE IF NOT EXISTS app_instance_role;

CREATE OR ALTER VERSIONED SCHEMA ext_code_schema;
GRANT USAGE ON SCHEMA ext_code_schema TO APPLICATION ROLE app_instance_role;

CREATE OR REPLACE PROCEDURE ext_code_schema.py_echo_proc(DATA string)
  RETURNS STRING
  LANGUAGE PYTHON
  RUNTIME_VERSION = 3.9
  PACKAGES=('snowflake-snowpark-python')
  HANDLER='echo.echo_proc'
  IMPORTS=('/echo.py');

    GRANT USAGE ON PROCEDURE ext_code_schema.py_echo_proc(string)
      TO APPLICATION ROLE app_instance_role;

-- Wraps a function from a python file
CREATE OR REPLACE FUNCTION ext_code_schema.py_echo_fn(string)
RETURNS STRING
LANGUAGE PYTHON
RUNTIME_VERSION = 3.9
PACKAGES=('snowflake-snowpark-python')
HANDLER='echo.echo_fn'
IMPORTS=('/echo.py');

GRANT USAGE ON FUNCTION ext_code_schema.py_echo_fn(DATA string)
  TO APPLICATION ROLE app_instance_role;

# some python file echo.py
@udf(name="echo_fn")
def echo_fn(data) -> str:
  return 'echo_fn: ' + str

-- Sample setup_script.sql SQL file for a Snowflake Native App

-- User-written code
CREATE OR REPLACE APPLICATION ROLE app_instance_role;

CREATE OR ALTER VERSIONED SCHEMA ext_code_schema;
GRANT USAGE ON SCHEMA ext_code_schema TO APPLICATION ROLE app_instance_role;

-- Snowflake CLI generated code
CREATE OR REPLACE FUNCTION ext_code_schema.py_echo_fn(DATA string)
  RETURNS STRING
  LANGUAGE PYTHON
  RUNTIME_VERSION = 3.9
  PACKAGES=('snowflake-snowpark-python')
  HANDLER='echo.echo_fn'
  IMPORTS=('/echo.py');

  GRANT USAGE ON FUNCTION ext_code_schema.py_echo_fn(string)
    TO APPLICATION ROLE app_instance_role;

Using the Snowpark Python decorators¶

While the Snowpark decorators in Snowflake CLI work the same as regular Snowpark Python decorators, you should be aware of the following differences when writing Python code files specifically for a Snowflake Native App:

• You can’t use any Session objects in these files, as Snowflake CLI executes these Python files in a sandbox environment with no connection to Snowflake. As a result, any reference of a Snowpark Session results in an error.

• You can use only the @udf, @sproc, @udaf and @udtf Snowpark Python decorators.

• You can’t use these decorators as regular functions to register your code as a Snowflake object. Only code explicitly annotated with the supported decorators is recognized. Therefore, the Python function must be a named function. Lambda functions are not supported.

• Snowflake CLI always generates CREATE OR REPLACE statements, as recommended by Snowflake, for creating functions and procedures in your setup scripts.

# output/deploy/dest_dir1/dest_dir2/echo.py
#: @sproc(
#:    return_type=IntegerType(),
#:    input_types=[IntegerType(), IntegerType()],
#:    packages=["snowflake-snowpark-python==1.15.0"],
#:    native_app_params={
#:        "schema": "ext_code_schema",
#:        "application_roles": ["app_instance_role"],
#:    },
#: )
def add_sp(session_, x, y):
    return x + y