MaxCompute:Functions

Object model

PyODPS represents UDFs with two types of objects:

• MaxCompute entry object (o): exposes project-level methods — create_function(), delete_function(), get_function(), list_functions(), and exist_function().

• Function object: exposes instance-level methods — update() and drop() — and holds properties such as class_type and resources.

Understanding which object owns which method helps you write correct PyODPS code: always call project-level operations on o, and call instance-level operations on a function object you retrieved first.

Create a UDF

Call create_function() on the MaxCompute entry object. Pass the function name, the implementing class (class_type), and a list of resources.

The following examples create a UDF that references a resource in the current project and in another project:

# Reference a resource in the current project.
resource = o.get_resource('my_udf.py')
function = o.create_function('test_function', class_type='my_udf.Test', resources=[resource])
# Reference a resource in another project.
resource2 = o.get_resource('my_udf.py', project='another_project')
function2 = o.create_function('test_function2', class_type='my_udf.Test', resources=[resource2])

List UDFs

Call list_functions() on the MaxCompute entry object to get all UDFs in a specific project.

for func in o.list_functions():
    print(func.name)

Check whether a UDF exists

Call exist_function() on the MaxCompute entry object.

o.exist_function('test_function')

Get a UDF

Call get_function() on the MaxCompute entry object to retrieve a function object.

function = o.get_function('test_function')

Update a UDF

Get the function object first, then modify its class_type and resources properties, and call update().

function = o.get_function('test_function')
new_resource = o.get_resource('my_udf2.py')
function.class_type = 'my_udf2.Test'
function.resources = [new_resource, ]
function.update()  # Update the function.

Delete a UDF

Two methods are available:

• delete_function() on the entry object: use this when you only have the function name.

• drop() on the function object: use this when you already hold a function object.

# Delete by name.
o.delete_function('test_function')
# Delete using the function object.
function.drop()  # Call the drop() method if the function exists.