TeradataML

Warnings

• breaking Starting from teradataml 20.0.0.9, optional dependencies are no longer installed by default due to footprint reduction and modularization. Users must explicitly install them using `pip install teradataml[<feature_name>]` (e.g., `teradataml[automl]`, `teradataml[openml]`, `teradataml[visualization]`, `teradataml[eda-ui]`).
  Fix: Manually install required optional dependencies: `pip install teradataml[feature_name]`.
• gotcha The `fastload()` utility in teradataml does not support table names that start with digits, even though it works with `copy_to_sql()` and SQL directly.
  Fix: Avoid creating or using table names that begin with digits when using `fastload()`. Consider using `copy_to_sql()` or pandas `to_sql()` instead if this is an issue.
• gotcha Connections can fail if the Teradata system has 'Require Confidentiality' enabled in `gtwcontrol` due to data encryption issues with older `teradatasql` versions.
  Fix: Upgrade `teradatasql` (`pip install -U --no-cache-dir teradatasql`) or, for older `teradataml` versions, pass `encryptdata='true'` via a SQLAlchemy engine to `create_context`.
• gotcha When upgrading `teradataml`, `pip install` may use a cached version. To ensure the new version is downloaded, use the `--no-cache-dir` option.
  Fix: Use `pip install --no-cache-dir -U teradataml` for upgrades.
• breaking In version 20.0.0.10, the `set_auth_token` function's return type changed from a boolean to the class object. While functionality remains the same, this might affect code expecting a boolean return.
  Fix: Update code that relies on the return value of `set_auth_token` to expect the class object, or adjust logic if a boolean check was previously performed.
• deprecated In earlier versions (e.g., 16.20.00.01), old analytic functions were deprecated due to namespace changes. While not directly breaking for current versions, users upgrading from very old `teradataml` might encounter this.
  Fix: Refer to the Teradata Python Package User Guide for updated analytic function names and structures.
• gotcha When creating a SQLAlchemy engine for use with `teradataml`'s `create_context`, ensure to specify the `teradatasql` dialect, not a generic `teradata` dialect, to avoid `NoSuchModuleError`.
  Fix: Use `sqlalchemy.create_engine('teradatasql://...')` for creating the engine.

Install

• pip install teradataml Basic Installation
• pip install teradataml[automl] Install with AutoML dependencies
• pip install teradataml[openml] Install with OpenSourceML dependencies
• pip install teradataml[visualization] Install with Visualization dependencies
• pip install teradataml[eda-ui] Install with EDA UI dependencies

Imports

• create_context

  from teradataml import create_context

• DataFrame

  from teradataml import DataFrame

• copy_to_sql

  from teradataml import copy_to_sql

• remove_context

  from teradataml import remove_context

• in_schema
  wrong:

  from teradataml import in_schema

  correct:

  from teradataml.dataframe.dataframe import in_schema

  in_schema is part of the dataframe submodule, not top-level.

Quickstart

This quickstart demonstrates how to establish a connection to Teradata Vantage using `create_context`, create a teradataml DataFrame from an existing table, and display its head. It uses environment variables for sensitive connection details.

import os
from teradataml import create_context, DataFrame, remove_context

host = os.environ.get('TD_HOST', 'your_teradata_host')
username = os.environ.get('TD_USERNAME', 'your_username')
password = os.environ.get('TD_PASSWORD', 'your_password')
temp_database_name = os.environ.get('TD_TEMP_DB', 'your_temp_db')

# Establish connection to Teradata Vantage
try:
    create_context(host=host, username=username, password=password, 
                   logmech='TD2', temp_database_name=temp_database_name)
    print("Successfully connected to Teradata Vantage.")

    # Create a teradataml DataFrame from an existing table
    # Replace 'your_table_name' and 'your_database' with actual values
    # For demo, assuming a table named 'sample_data' in the default database
    td_df = DataFrame(tablename='sample_data')
    print("\nTeradataml DataFrame head:")
    print(td_df.head())

except Exception as e:
    print(f"An error occurred: {e}")
finally:
    # Clean up the connection
    if create_context._active_context:
        remove_context()
        print("Connection context removed.")