Migration Guide - Removing yield Usage (fivetran-connector-sdk v2.0.0+)

With the release of version 2.0.0 for fivetran-connector-sdk pypi package, you no longer need to use yield with Connector SDK operations that was required in older versions. This update simplifies your code and removes generator-based patterns. You're still free to use generators elsewhere in your code where appropriate, but they're no longer needed for SDK operations.

Background

Prior to version 2.0.0 of the fivetran-connector-sdk PyPI package, data operations such as upsert and checkpoint had to be returned using the yield keyword.

Code example using yield (pre-2.0.0 fivetran-connector-sdk )

from fivetran_connector_sdk import Operations as op

def update(configuration: dict, state: dict):
    # Call the helper function to upsert users
    yield from upsert_users(state)


def upsert_users(state: dict):
    # Sample hardcoded user data
    data = [
        {"id": 1, "name": "Alice"},
        {"id": 2, "name": "Bob"},
        {"id": 3, "name": "Charlie"},
    ]

    # Yield one upsert operation per user
    for row in data:
        yield op.upsert("users", row)

    # Save the progress with a checkpoint
    yield op.checkpoint(state)

In this version, yield was required to perform operations like upsert, delete, checkpoint, etc.

Updated code example without yield (fivetran-connector-sdk v2.0.0+)

from fivetran_connector_sdk import Operations as op

def update(configuration: dict, state: dict):
    # Call the helper function to upsert users
    upsert_users(state)


def upsert_users(state: dict):
    # Sample hardcoded user data
    data = [
        {"id": 1, "name": "Alice"},
        {"id": 2, "name": "Bob"},
        {"id": 3, "name": "Charlie"},
    ]

    # Perform one upsert operation per user
    for row in data:
        op.upsert("users", row)

    # Save the progress with a checkpoint
    op.checkpoint(state)

Now you can simply call data operations directly, without using yield.

Why this matters

• Simpler syntax – easier to read and maintain
• Better performance – no generator overhead
• Cleaner stack traces – easier debugging

What you need to do

• Upgrade to version 2.0.0 or later
• Identify and remove any Connector SDK related usages of yield in your code
• Replace the related yield from statements with direct function calls
• Run fivetran debug to verify everything works as expected
• Run fivetran deploy command to deploy your updated connector code

Related links

• Changelog for version 2.0.0: Fivetran Connector SDK Changelog
• Fivetran Connector SDK code examples