Snowflake Snowpark Python

Warnings

• breaking Snowpark Python has dropped support for Python 3.8. Version 1.24.0 was the last to support it. Using Snowpark Python with Python 3.8 will trigger deprecation warnings.
  Fix: Upgrade your Python environment to 3.9 or greater. The library now requires Python >=3.9, <3.14.
• gotcha The default 'overwrite' mode for `DataFrameWriter.save_as_table` drops and recreates the target table, leading to potential data loss for non-matching rows and impacting grants. This can be unexpected if partial updates are desired.
  Fix: For targeted delete-insert operations, use the `overwrite_condition` parameter (available since v1.44.0) with `mode='overwrite'`. If you need to preserve grants, specify `copy_grants=True` where applicable.
• gotcha When registering UDFs/SPROCs, specifying an empty list (`[]`) for the `imports` or `packages` argument now explicitly means *no* imports/packages for that specific UDF/SPROC. This behavior changed from older versions where an empty list implicitly meant using session-level imports/packages.
  Fix: To use session-level imports or packages, pass `None` or omit the `imports`/`packages` argument. To explicitly use no imports/packages, pass an empty list `[]`.
• bug A bug existed where `Session.udf.register_from_file` did not properly process the `strict` and `secure` parameters, potentially leading to UDFs not being created with the intended security or null-handling characteristics.
  Fix: Upgrade to `snowflake-snowpark-python` version 1.47.0 or higher.
• gotcha Managing Python packages not available in Snowflake's Anaconda channel for UDFs and stored procedures can be complex. These often require manual zipping and uploading to Snowflake stages, and careful management of `imports` and `packages` parameters.
  Fix: Prioritize packages from Snowflake's Anaconda channel. For custom or unavailable packages, zip them and upload to a Snowflake stage. Reference these staged files using the `imports` parameter when registering UDFs/SPROCs.

Install

• pip install snowflake-snowpark-python Install latest version

Imports

• Session

  from snowflake.snowpark import Session

• functions

  from snowflake.snowpark import functions

  Commonly aliased as 'F' for convenience: `import snowflake.snowpark.functions as F`
• types

  from snowflake.snowpark import types

  Used for specifying UDF/UDTF input/return types.
• DataFrame

  df = session.create_dataframe(...) # DataFrame objects are typically returned by Session methods.

  The `DataFrame` class is generally not imported directly by users; instances are created via `Session` methods like `create_dataframe`, `table`, or `sql`.

Quickstart

This quickstart demonstrates how to establish a Snowpark Session, create a DataFrame from local data, perform a basic transformation, and display the results. Connection parameters are loaded from environment variables for secure and flexible setup. Remember to replace placeholder values with your Snowflake account details.

import os
from snowflake.snowpark import Session
from snowflake.snowpark.functions import col

# Establish a Snowpark Session using environment variables
# Replace with your actual connection parameters, or configure ~/.snowflake/connections.toml
connection_parameters = {
    "account": os.environ.get("SNOWFLAKE_ACCOUNT", "your_account_identifier"),
    "user": os.environ.get("SNOWFLAKE_USER", "your_username"),
    "password": os.environ.get("SNOWFLAKE_PASSWORD", "your_password"),
    "role": os.environ.get("SNOWFLAKE_ROLE", "your_role"),
    "warehouse": os.environ.get("SNOWFLAKE_WAREHOUSE", "your_warehouse"),
    "database": os.environ.get("SNOWFLAKE_DATABASE", "your_database"),
    "schema": os.environ.get("SNOWFLAKE_SCHEMA", "your_schema"),
}

session = Session.builder.configs(connection_parameters).create()
print("Snowpark Session created successfully.")

# Create a simple DataFrame
data = [("Alice", 1), ("Bob", 2), ("Charlie", 3)]
df = session.create_dataframe(data, schema=["name", "id"])

# Perform a simple transformation and show results
df.filter(col("id") > 1).show()

# Close the session
session.close()
print("Snowpark Session closed.")