The dbt Teradata adapter lets you use dbt with Teradata Vantage.

NOTE: This adapter is maintained by Teradata. We are accelerating our release cadence. Starting October 1st, 2023, we will release dbt-teradata within 4 weeks of a minor release or within 8 weeks of a major release of dbt-core.

pip install dbt-teradata

Starting from dbt-teradata 1.8.0 and above, dbt-core will not be installed as a dependency. Therefore, you need to explicitly install dbt-core. Ensure you install dbt-core 1.8.0 or above. You can do this with the following command:

pip install dbt-core>=1.8.0

Please go through this discussion for more information: dbt-labs/dbt-core#9171

If you are new to dbt on Teradata see dbt with Teradata Vantage tutorial.

NOTE: If the virtual environment in Python is not activating properly on Windows, you can try running the below command in the command-line interface (CLI) before attempting to activate the virtual environment.

Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUse

Here is a working example of a dbt-teradata profile:

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: localhost
      user: dbc
      password: dbc
      schema: dbt_test
      tmode: ANSI

At a minimum, you need to specify host, user, password, schema (database), tmode.

The logon mechanism for Teradata jobs that dbt executes can be configured with the logmech configuration in your Teradata profile. The logmech field can be set to: TD2, LDAP, BROWSER, KRB5, TDNEGO. For more information on authentication options, go to Teradata Vantage authentication documentation.

For the initial BROWSER authentication, the browser opens as expected, asking for the credentials. However, for every subsequent connection, a new browser tab opens, displaying the message 'TERADATA BROWSER AUTHENTICATION COMPLETED,' despite using an existing BROWSER session silently. This is the default behavior of the teradatasql driver, and there is no way to avoid this at the present time.

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      user: <user>
      password: <password>
      schema: dbt_test
      tmode: ANSI
      logmech: LDAP

The logon mechanism for Teradata jobs that dbt executes can be configured with the logdata configuration in your Teradata profile. Addtional data like secure token, distinguished Name, or a domain/realm name can be set in your Teradata profile using logdata. The logdata field can be set to: JWT, LDAP, KRB5, TDNEGO. logdata is not used with the TD2 mechanism.

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      schema: dbt_test
      tmode: ANSI
      logmech: LDAP
      logdata: 'authcid=username password=password'
      port: <port>

For more information on authentication options, go to Teradata Vantage authentication documentation

Stored Password Protection enables an application to provide a connection password in encrypted form to the driver. The plugin supports Stored Password Protection feature through prefix ENCRYPTED_PASSWORD( either in password connection parameter or in logdata connection parameter.

• password

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      user: <user>
      password: ENCRYPTED_PASSWORD(file:PasswordEncryptionKeyFileName,file:EncryptedPasswordFileName)
      schema: dbt_test
      tmode: ANSI
      port: <port>

• logdata

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      schema: dbt_test
      tmode: ANSI
      logmech: LDAP
      logdata: 'authcid=username password=ENCRYPTED_PASSWORD(file:PasswordEncryptionKeyFileName,file:EncryptedPasswordFileName)'
      port: <port>

For full description of Stored Password Protection see https://github.com/Teradata/python-driver#StoredPasswordProtection.

If your Teradata database runs on port different than the default (1025), you can specify a custom port in your dbt profile using port configuration.

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      user: <user>
      password: <password>
      schema: dbt_test
      tmode: ANSI
      port: <port>

Allows an adapter to automatically try again when the attempt to open a new connection on the database has a transient, infrequent error. This option can be set using the retries configuration. Default value is 0. The default wait period between connection attempts is one second. retry_timeout (seconds) option allows us to adjust this waiting period.

If retries is set to 3, the adapter will try to establish a new connection three times if an error occurs.

my-teradata-db-profile:
  target: dev
  outputs:
    dev:
      type: teradata
      host: <host>
      user: <user>
      password: <password>
      schema: dbt_test
      tmode: ANSI
      retries: 3
      retry_timeout: 10

The following fields are required:

The plugin also supports the following optional connection parameters:

Refer to connection parameters for the full description of the connection parameters.

• view
• table
• ephemeral
• incremental

The following incremental materialization strategies are supported:

• append (default)

• delete+insert

• merge

• valid_history

  'valid_history' incremental materialization strategy (early access)

  This strategy is designed to manage historical data efficiently within a Teradata environment, leveraging dbt features to ensure data quality and optimal resource usage. In temporal databases, valid time is crucial for applications like historical reporting, ML training datasets, and forensic analysis.

    {{
        config(
            materialized='incremental',
            unique_key='id',
            on_schema_change='fail',
            incremental_strategy='valid_history',
            valid_from='valid_from_column',
            history_column_in_target='history_period_column'
    )
    }}

  valid_history incremental strategy requires the following parameters:

  • valid_from - Column in the source table of timestamp datatype indicating when each record became valid.
  • history_column_in_target - Column in the target table of period datatype that tracks history.

The valid_history strategy in dbt-teradata involves several critical steps to ensure the integrity and accuracy of historical data management:

• Remove duplicates and conflicting values from the source data:
  • This step ensures that the data is clean and ready for further processing by eliminating any redundant or conflicting records.
  • The process of removing duplicates and conflicting values from the source data involves using a ranking mechanism to ensure that only the highest-priority records are retained. This is accomplished using the SQL RANK() function.
• Identify and adjust overlapping time slices:
  • Overlapping time periods in the data are detected and corrected to maintain a consistent and non-overlapping timeline.
• Manage records needing to be overwritten or split based on the source and target data:
  • This involves handling scenarios where records in the source data overlap with or need to replace records in the target data, ensuring that the historical timeline remains accurate.
• Utilize the TD_NORMALIZE_MEET function to compact history:
  • This function helps to normalize and compact the history by merging adjacent time periods, improving the efficiency and performance of the database.
• Delete existing overlapping records from the target table:
  • Before inserting new or updated records, any existing records in the target table that overlap with the new data are removed to prevent conflicts.
• Insert the processed data into the target table:
  • Finally, the cleaned and adjusted data is inserted into the target table, ensuring that the historical data is up-to-date and accurately reflects the intended timeline.

These steps collectively ensure that the valid_history strategy effectively manages historical data, maintaining its integrity and accuracy while optimizing performance.

  An illustration demonstrating the source sample data and its corresponding target data:  

  -- Source data
      pk |       valid_from          | value_txt1 | value_txt2
      ======================================================================
      1  | 2024-03-01 00:00:00.0000  | A          | x1
      1  | 2024-03-12 00:00:00.0000  | B          | x1
      1  | 2024-03-12 00:00:00.0000  | B          | x2
      1  | 2024-03-25 00:00:00.0000  | A          | x2
      2  | 2024-03-01 00:00:00.0000  | A          | x1
      2  | 2024-03-12 00:00:00.0000  | C          | x1
      2  | 2024-03-12 00:00:00.0000  | D          | x1
      2  | 2024-03-13 00:00:00.0000  | C          | x1
      2  | 2024-03-14 00:00:00.0000  | C          | x1
  
  -- Target data
      pk | valid_period                                                       | value_txt1 | value_txt2
      ===================================================================================================
      1  | PERIOD(TIMESTAMP)[2024-03-01 00:00:00.0, 2024-03-12 00:00:00.0]    | A          | x1
      1  | PERIOD(TIMESTAMP)[2024-03-12 00:00:00.0, 2024-03-25 00:00:00.0]    | B          | x1
      1  | PERIOD(TIMESTAMP)[2024-03-25 00:00:00.0, 9999-12-31 23:59:59.9999] | A          | x2
      2  | PERIOD(TIMESTAMP)[2024-03-01 00:00:00.0, 2024-03-12 00:00:00.0]    | A          | x1
      2  | PERIOD(TIMESTAMP)[2024-03-12 00:00:00.0, 9999-12-31 23:59:59.9999] | C          | x1

Important Note: The target table must already exist before running the model. Ensure that the target table is created and properly structured with the necessary columns, including a column that tracks the history with period datatype, before running a dbt model.

To learn more about dbt incremental strategies please check the dbt incremental strategy documentation.

All dbt commands are supported.

• Enable view column types in docs - Teradata Vantage has a dbscontrol configuration flag called DisableQVCI (QVCI - Queryable View Column Index). This flag instructs the database to build DBC.ColumnsJQV with view column type definitions.

  ℹ️ Existing customers, please see KB0022230 for more information about enabling QVCI.

  To enable this functionality you need to:

  1. Enable QVCI mode in Vantage. Use dbscontrol utility and then restart Teradata. Run these commands as a privileged user on a Teradata node:

     # option 551 is DisableQVCI. Setting it to false enables QVCI.
     dbscontrol << EOF
     M internal 551=false
     W
     EOF
     
     # restart Teradata
     tpareset -y Enable QVCI

  2. Instruct dbt to use QVCI mode. Include the following variable in your dbt_project.yml:

     vars:
       use_qvci: true

     For example configuration, see test/catalog/with_qvci/dbt_project.yml.

The following options apply to table, snapshots and seed materializations.

• table_kind - define the table kind. Legal values are MULTISET (default for ANSI transaction mode required by dbt-teradata) and SET, e.g.:

  • in sql materialization definition file:

    {{
      config(
          materialized="table",
          table_kind="SET"
      )
    }}

  • in seed configuration:

    seeds:
      <project-name>:
        table_kind: "SET"

  For details, see CREATE TABLE documentation.

• table_option - define table options. Legal values are:

  { MAP = map_name [COLOCATE USING colocation_name] |
    [NO] FALLBACK [PROTECTION] |
    WITH JOURNAL TABLE = table_specification |
    [NO] LOG |
    [ NO | DUAL ] [BEFORE] JOURNAL |
    [ NO | DUAL | LOCAL | NOT LOCAL ] AFTER JOURNAL |
    CHECKSUM = { DEFAULT | ON | OFF } |
    FREESPACE = integer [PERCENT] |
    mergeblockratio |
    datablocksize |
    blockcompression |
    isolated_loading
  }

  where:

  • mergeblockratio:

    { DEFAULT MERGEBLOCKRATIO |
      MERGEBLOCKRATIO = integer [PERCENT] |
      NO MERGEBLOCKRATIO
    }

  • datablocksize:

    DATABLOCKSIZE = {
      data_block_size [ BYTES | KBYTES | KILOBYTES ] |
      { MINIMUM | MAXIMUM | DEFAULT } DATABLOCKSIZE
    }

  • blockcompression:

    BLOCKCOMPRESSION = { AUTOTEMP | MANUAL | ALWAYS | NEVER | DEFAULT }
      [, BLOCKCOMPRESSIONALGORITHM = { ZLIB | ELZS_H | DEFAULT } ]
      [, BLOCKCOMPRESSIONLEVEL = { value | DEFAULT } ]

  • isolated_loading:

    WITH [NO] [CONCURRENT] ISOLATED LOADING [ FOR { ALL | INSERT | NONE } ]

  Examples:

  • in sql materialization definition file:

    {{
      config(
          materialized="table",
          table_option="NO FALLBACK"
      )
    }}

    {{
      config(
          materialized="table",
          table_option="NO FALLBACK, NO JOURNAL"
      )
    }}

    {{
      config(
          materialized="table",
          table_option="NO FALLBACK, NO JOURNAL, CHECKSUM = ON,
            NO MERGEBLOCKRATIO,
            WITH CONCURRENT ISOLATED LOADING FOR ALL"
      )
    }}

  • in seed configuration:

    seeds:
      <project-name>:
        table_option:"NO FALLBACK"

    seeds:
      <project-name>:
        table_option:"NO FALLBACK, NO JOURNAL"

    seeds:
      <project-name>:
        table_option: "NO FALLBACK, NO JOURNAL, CHECKSUM = ON,
          NO MERGEBLOCKRATIO,
          WITH CONCURRENT ISOLATED LOADING FOR ALL"

  For details, see CREATE TABLE documentation.

• with_statistics - should statistics be copied from the base table, e.g.:

  {{
    config(
        materialized="table",
        with_statistics="true"
    )
  }}

  This option is not available for seeds as seeds do not use CREATE TABLE ... AS syntax.

  For details, see CREATE TABLE documentation.

• index - defines table indices:

  [UNIQUE] PRIMARY INDEX [index_name] ( index_column_name [,...] ) |
  NO PRIMARY INDEX |
  PRIMARY AMP [INDEX] [index_name] ( index_column_name [,...] ) |
  PARTITION BY { partitioning_level | ( partitioning_level [,...] ) } |
  UNIQUE INDEX [ index_name ] [ ( index_column_name [,...] ) ] [loading] |
  INDEX [index_name] [ALL] ( index_column_name [,...] ) [ordering] [loading]
  [,...]

  where:

  • partitioning_level:

    { partitioning_expression |
      COLUMN [ [NO] AUTO COMPRESS |
      COLUMN [ [NO] AUTO COMPRESS ] [ ALL BUT ] column_partition ]
    } [ ADD constant ]

  • ordering:

    ORDER BY [ VALUES | HASH ] [ ( order_column_name ) ]

  • loading:

    WITH [NO] LOAD IDENTITY

  e.g.:

  • in sql materialization definition file:

    {{
      config(
          materialized="table",
          index="UNIQUE PRIMARY INDEX ( GlobalID )"
      )
    }}

    ℹ️ Note, unlike in table_option, there are no commas between index statements!

    {{
      config(
          materialized="table",
          index="PRIMARY INDEX(id)
          PARTITION BY RANGE_N(create_date
                        BETWEEN DATE '2020-01-01'
                        AND     DATE '2021-01-01'
                        EACH INTERVAL '1' MONTH)"
      )
    }}

    {{
      config(
          materialized="table",
          index="PRIMARY INDEX(id)
          PARTITION BY RANGE_N(create_date
                        BETWEEN DATE '2020-01-01'
                        AND     DATE '2021-01-01'
                        EACH INTERVAL '1' MONTH)
          INDEX index_attrA (attrA) WITH LOAD IDENTITY"
      )
    }}

  • in seed configuration:

    seeds:
      <project-name>:
        index: "UNIQUE PRIMARY INDEX ( GlobalID )"

    ℹ️ Note, unlike in table_option, there are no commas between index statements!

    seeds:
      <project-name>:
        index: "PRIMARY INDEX(id)
          PARTITION BY RANGE_N(create_date
                        BETWEEN DATE '2020-01-01'
                        AND     DATE '2021-01-01'
                        EACH INTERVAL '1' MONTH)"

    seeds:
      <project-name>:
        index: "PRIMARY INDEX(id)
          PARTITION BY RANGE_N(create_date
                        BETWEEN DATE '2020-01-01'
                        AND     DATE '2021-01-01'
                        EACH INTERVAL '1' MONTH)
          INDEX index_attrA (attrA) WITH LOAD IDENTITY"

Seeds, in addition to the above materialization modifiers, have the following options:

• use_fastload - use fastload when handling dbt seed command. The option will likely speed up loading when your seed files have hundreds of thousands of rows. You can set this seed configuration option in your project.yml file, e.g.:

  seeds:
    <project-name>:
      +use_fastload: true

Snapshots uses the HASHROW function of the Teradata database to generate a unique hash value for the 'dbt_scd_id' column. If you want to use your own hash UDF, there is a configuration option in the snapshot model called 'snapshot_hash_udf', which defaults to HASHROW. You can provide a value like <database_name.hash_udf_name>. If only hash_udf_name is provided, it uses the same schema as the model runs.

for e.g. : snapshots/snapshot_example.sql

{% snapshot snapshot_example %}
{{
  config(
    target_schema='snapshots',
    unique_key='id',
    strategy='check',
    check_cols=["c2"],
    snapshot_hash_udf='GLOBAL_FUNCTIONS.hash_md5'
  )
}}
select * from {{ ref('order_payments') }}
{% endsnapshot %}

Grants are supported in dbt-teradata adapter with release version 1.2.0 and above. You can use grants to manage access to the datasets you're producing with dbt. To implement these permissions, define grants as resource configs on each model, seed, or snapshot. Define the default grants that apply to the entire project in your dbt_project.yml, and define model-specific grants within each model's SQL or YAML file.

for e.g. : models/schema.yml

models:
  - name: model_name
    config:
      grants:
        select: ['user_a', 'user_b']

Another e.g. for adding multiple grants:

models:
- name: model_name
  config:
    materialized: table
    grants:
      select: ["user_b"]
      insert: ["user_c"]

ℹ️ copy_grants is not supported in Teradata.

More on Grants can be found at https://docs.getdbt.com/reference/resource-configs/grants

Starting with release 1.3, some macros were migrated from teradata-dbt-utils dbt package to the connector. See the table below for the macros supported from the connector.

For using cross DB macros, teradata-utils as a macro namespace will not be used, as cross DB macros have been migrated from teradata-utils to Dbt-Teradata.

Replace: {{ dbt.replace("string_text_column", "old_chars", "new_chars") }} {{ replace('abcgef', 'g', 'd') }}

Date truncate: {{ dbt.date_trunc("date_part", "date") }} {{ dbt.date_trunc("DD", "'2018-01-05 12:00:00'") }}

datediff macro in teradata supports difference between dates. Differece between timestamps is not supported.

Hash macro needs an md5 function implementation. Teradata doesn't support md5 natively. You need to install a User Defined Function (UDF):

1. Download the md5 UDF implementation from Teradata (registration required): https://downloads.teradata.com/download/extensibility/md5-message-digest-udf.
2. Unzip the package and go to src directory.
3. Start up bteq and connect to your database.
4. Create database GLOBAL_FUNCTIONS that will host the UDF. You can't change the database name as it's hardcoded in the macro:

   CREATE DATABASE GLOBAL_FUNCTIONS AS PERMANENT = 60e6, SPOOL = 120e6;

5. Create the UDF. Replace <CURRENT_USER> with your current database user:

   GRANT CREATE FUNCTION ON GLOBAL_FUNCTIONS TO <CURRENT_USER>;
   DATABASE GLOBAL_FUNCTIONS;
   .run file = hash_md5.btq

6. Grant permissions to run the UDF with grant option.

   GRANT EXECUTE FUNCTION ON GLOBAL_FUNCTIONS TO PUBLIC WITH GRANT OPTION;

last_day in teradata_utils, unlike the corresponding macro in dbt_utils, doesn't support quarter datepart.

• collect statistics - when a table is created or modified significantly, there might be a need to tell Teradata to collect statistics for the optimizer. It can be done using COLLECT STATISTICS command. You can perform this step using dbt's post-hooks, e.g.:

  {{ config(
    post_hook=[
      "COLLECT STATISTICS ON  {{ this }} COLUMN (column_1,  column_2  ...);"
      ]
  )}}

  See Collecting Statistics documentation for more information.

Model contracts are supported with dbt-teradata v1.7.1 and onwards. Constraint support and enforcement in dbt-teradata

To find more on model contracts please follow dbt documentations https://docs.getdbt.com/docs/collaborate/govern/model-contracts

dbt-utils package is supported through teradata/teradata_utils dbt package. The package provides a compatibility layer between dbt_utils and dbt-teradata. See teradata_utils package for install instructions.

Both ANSI and TERA modes are now supported in dbt-teradata. TERA mode's support is introduced with dbt-teradata 1.7.1, it is an initial implementation.

Query Band in dbt-teradata can be set on three levels:

1. Profiles Level: In profiles.yml file, user can provide query_band as below example

   query_band: 'application=dbt;'

2. Project Level: In dbt_project.yml file, user can provide query_band as below

    models:
    Project_name:
       +query_band: "app=dbt;model={model};"

3. Model Level: It can be set on model sql file or model level configuration on yaml files

   {{ config( query_band='sql={model};' ) }}

User can set query_band at any level or on all levels. With profiles level query_band, dbt-teradata will set the query_band for first time for the session and subsequently for model and project level query band will be updated with respective configuration. If a user set some key-value pair with value as '{model}' than internally this '{model}' will be replaced with model name, and it can be useful for telemetry tracking of sql/ dbql logging. Let model that user is running be stg_orders

     models:
     Project_name:
        +query_band: "app=dbt;model={model};"

{model} will be replaced with 'stg_orders' in runtime.

If no query_band is set by user, default query_band will come in play that is : org=teradata-internal-telem;appname=dbt;

The adapter was originally created by Doug Beatty. Teradata took over the adapter in January 2022. We are grateful to Doug for founding the project and accelerating the integration of dbt + Teradata.

The adapter is published using Apache-2.0 License. Please see the license for terms and conditions, such as creating derivative work and the support model.