Skip to content

hovaesco/dbt-trino

 
 

Repository files navigation

Build Status db-presto-trino Slack

dbt-trino

Introduction

dbt is a data transformation workflow tool that lets teams quickly and collaboratively deploy analytics code, following software engineering best practices like modularity, CI/CD, testing, and documentation. It enables anyone who knows SQL to build production-grade data pipelines.

One frequently asked question in the context of using dbt tool is:

Can I connect my dbt project to two databases?

(see the answered question on the dbt website).

TL;DR dbt stands for transformation as in T within ELT pipelines, it doesn't move data from source to a warehouse.

dbt-trino adapter uses Trino as a underlying query engine to perform query federation across disperse data sources. Trino connects to multiple and diverse data sources (available connectors) via one dbt connection and process SQL queries at scale. Transformations defined in dbt are passed to Trino which handles these SQL transformation queries and translates them to queries specific to the systems it connects to create tables or views and manipulate data.

This repository represents a fork of the dbt-presto with adaptations to make it work with Trino.

Compatibility

This dbt plugin has been tested against Trino version 393 and Starburst Enterprise version 393-e.1.

Installation

This dbt adapter can be installed via pip:

$ pip install dbt-trino

Configuring your profile

A dbt profile can be configured to run against Trino using the following configuration:

Option Description Required? Example
method The Trino authentication method to use Optional (default is none, supported methods are ldap, kerberos, jwt, oauth or certificate) none or kerberos
user Username for authentication Optional (required if method is none, ldap or kerberos) commander
password Password for authentication Optional (required if method is ldap) none or abc123
keytab Path to keytab for kerberos authentication Optional (may be required if method is kerberos) /tmp/trino.keytab
krb5_config Path to config for kerberos authentication Optional (may be required if method is kerberos) /tmp/krb5.conf
principal Principal for kerberos authentication Optional (may be required if method is kerberos) trino@EXAMPLE.COM
service_name Service name for kerberos authentication Optional (default is trino) abc123
jwt_token JWT token for authentication Optional (required if method is jwt) none or abc123
client_certificate Path to client certificate to be used for certificate based authentication Optional (required if method is certificate) /tmp/tls.crt
client_private_key Path to client private key to be used for certificate based authentication Optional (required if method is certificate) /tmp/tls.key
http_headers HTTP Headers to send alongside requests to Trino, specified as a yaml dictionary of (header, value) pairs. Optional X-Trino-Client-Info: dbt-trino
http_scheme The HTTP scheme to use for requests to Trino Optional (default is http, or https for method: kerberos, ldap or jwt) https or http
cert The full path to a certificate file for authentication with trino Optional
session_properties Sets Trino session properties used in the connection Optional query_max_run_time: 5d
database Specify the database to build models into Required analytics
schema Specify the schema to build models into. Note: it is not recommended to use upper or mixed case schema names Required public
host The hostname to connect to Required 127.0.0.1
port The port to connect to the host on Required 8080
threads How many threads dbt should use Optional (default is 1) 8
prepared_statements_enabled Enable usage of Trino prepared statements (used in dbt seed commands) Optional (default is true) true or false
retries Configure how many times a database operation is retried when connection issues arise Optional (default is 3) 10

Example profiles.yml entry:

my-trino-db:
  target: dev
  outputs:
    dev:
      type: trino
      user: commander
      host: 127.0.0.1
      port: 8080
      database: analytics
      schema: public
      threads: 8
      http_scheme: http
      session_properties:
        query_max_run_time: 5d
        exchange_compression: True

Example profiles.yml entry for kerberos authentication:

my-trino-db:
  target: dev
  outputs:
    dev:
      type: trino
      method: kerberos
      user: commander
      keytab: /tmp/trino.keytab
      krb5_config: /tmp/krb5.conf
      principal: trino@EXAMPLE.COM
      host: trino.example.com
      port: 443
      database: analytics
      schema: public

For reference on which session properties can be set on the the dbt profile do execute

SHOW SESSION;

on your Trino instance.

Usage Notes

Supported authentication types

  • none - No authentication
  • ldap - Specify username in user and password in password
  • kerberos - Specify username in user
  • jwt - Specify JWT token in jwt_token
  • certificate - Specify a client certificate in client_certificate and private key in client_private_key
  • oauth - It is recommended to install keyring to cache the OAuth2 token over multiple dbt invocations by running pip install 'trino[external-authentication-token-cache]', keyring is not installed by default.

See also: https://trino.io/docs/current/security/authentication-types.html

Required configuration

dbt fundamentally works by dropping and creating tables and views in databases. As such, the following Trino configs must be set for dbt to work properly on Trino:

hive.metastore-cache-ttl=0s
hive.metastore-refresh-interval = 5s
hive.allow-drop-table=true
hive.allow-rename-table=true

Session properties per model

In some specific cases, there may be needed tuning through the Trino session properties only for a specific dbt model. In such cases, using the dbt hooks may come to the rescue:

{{
  config(
    pre_hook="set session query_max_run_time='10m'"
  )
}}

Materializations

Table

dbt-trino supports two modes in table materialization rename and drop configured using on_table_exists.

  • rename - creates intermediate table, then renames the target to backup one and renames intermediate to target one.
  • drop - drops and recreates a table. It overcomes table rename limitation in AWS Glue.

By default table materialization uses on_table_exists = 'rename', see an examples below how to change it.

In model add:

{{
  config(
    materialized = 'table',
    on_table_exists = 'drop`
  )
}}

or in dbt_project.yaml:

models:
  path:
    materialized: table
    +on_table_exists: drop

Using table materialization and on_table_exists = 'rename' with AWS Glue may result in below error:

TrinoUserError(type=USER_ERROR, name=NOT_SUPPORTED, message="Table rename is not yet supported by Glue service")
View

Adapter supports two security modes in view materialization DEFINER and INVOKER configured using view_security.

See Trino docs for more details about security modes in views.

By default view materialization uses view_security = 'definer', see an examples below how to change it.

In model add:

{{
  config(
    materialized = 'view',
    view_security = 'invoker'
  )
}}

or in dbt_project.yaml:

models:
  path:
    materialized: view
    +view_security: invoker
Incremental

Using an incremental model limits the amount of data that needs to be transformed, vastly reducing the runtime of your transformations. This improves performance and reduces compute costs.

{{
    config(
      materialized = 'incremental', 
      unique_key='<optional>',
      incremental_strategy='<optional>',)
}}
select * from {{ ref('events') }}
{% if is_incremental() %}
  where event_ts > (select max(event_ts) from {{ this }})
{% endif %}
append (default)

The default incremental strategy is append. append only adds the new records based on the condition specified in the is_incremental() conditional block.

{{
    config(
      materialized = 'incremental')
}}
select * from {{ ref('events') }}
{% if is_incremental() %}
  where event_ts > (select max(event_ts) from {{ this }})
{% endif %}
delete+insert

Through the delete+insert incremental strategy, you can instruct dbt to use a two-step incremental approach. It will first delete the records detected through the configured is_incremental() block and re-insert them.

{{
    config(
      materialized = 'incremental',
      unique_key='user_id',
      incremental_strategy='delete+insert',
      )
}}
select * from {{ ref('users') }}
{% if is_incremental() %}
  where updated_ts > (select max(updated_ts) from {{ this }})
{% endif %}
merge

Through the merge incremental strategy, dbt-trino constructs a MERGE statement which INSERTs new and UPDATEs existing records based on the unique key (specified by unique_key).
If unique_key is not unique delete+insert strategy can be used. Note that some connectors in Trino have limited or no support for MERGE.

{{
    config(
      materialized = 'incremental',
      unique_key='user_id',
      incremental_strategy='merge',
      )
}}
select * from {{ ref('users') }}
{% if is_incremental() %}
  where updated_ts > (select max(updated_ts) from {{ this }})
{% endif %}
Incremental overwrite on hive models

In case that the target incremental model is being accessed with hive Trino connector, an insert overwrite functionality can be achieved when using:

<hive-catalog-name>.insert-existing-partitions-behavior=OVERWRITE

setting on the Trino hive connector configuration.

Below is a sample hive profile entry to deal with OVERWRITE functionality for the hive connector called minio:

trino-incremental-hive:
  target: dev
  outputs:
    dev:
      type: trino
      method: none
      user: admin
      password:
      catalog: minio
      schema: tiny
      host: localhost
      port: 8080
      http_scheme: http
      session_properties:
        minio.insert_existing_partitions_behavior: OVERWRITE
      threads: 1

Existing partitions in the target model that match the staged data will be overwritten. The rest of the partitions will be simply appended to the target model.

NOTE that this functionality works on incremental models that use partitioning:

{{
    config(
        materialized = 'incremental',
        properties={
          "format": "'PARQUET'",
          "partitioned_by": "ARRAY['day']",
        }
    )
}}
Snapshots

Commonly, analysts need to "look back in time" at some previous state of data in their mutable tables. While some source data systems are built in a way that makes accessing historical data possible, this is often not the case. dbt provides a mechanism, snapshots, which records changes to a mutable table over time.

Snapshots implement type-2 Slowly Changing Dimensions over mutable source tables. These Slowly Changing Dimensions (or SCDs) identify how a row in a table changes over time. Imagine you have an orders table where the status field can be overwritten as the order is processed. See also the dbt docs about snapshots.

An example is given below.

{% snapshot orders_snapshot %}
{{
    config(
        target_database='analytics',
        target_schema='snapshots',
        unique_key='id',
        strategy='timestamp',
        updated_at='updated_at',
    )
}}
select * from {{ source('jaffle_shop', 'orders') }}
{% endsnapshot %}

Note that the Snapshot feature depends on the current_timestamp macro. In some connectors the standard precision (TIMESTAMP(3) WITH TIME ZONE) is not supported by the connector eg. Iceberg.

If necessary, you can override the standard precision by providing your own version of the trino__current_timestamp() macro as in following example:

{% macro trino__current_timestamp() %}
    current_timestamp(6)
{% endmacro %}

Use table properties to configure connector specifics

Trino connectors use table properties to configure connector specifics.

Check the Trino connector documentation for more information.

{{
  config(
    materialized='table',
    properties={
      "format": "'PARQUET'",
      "partitioning": "ARRAY['bucket(id, 2)']",
    }
  )
}}

Seeds

Seeds are CSV files in your dbt project (typically in your data directory), that dbt can load into your data warehouse using the dbt seed command.

For dbt-trino batch_size is defined in macro trino__get_batch_size() and default value is 1000. In order to override default value define within your project a macro like the following:

{% macro default__get_batch_size() %}
  {{ return(10000) }}
{% endmacro %}

Persist docs

Persist docs optionally persist resource descriptions as column and relation comments in the database. By default, documentation persistence is disabled, but it can be enabled for specific resources or groups of resources as needed.

Column-level comments are not supported in Trino views. Detailed documentation can be found here.

Generating lineage flow in docs

In order to generate lineage flow in docs use ref function in the place of table names in the query. It builts dependencies between models and allows to create DAG with data flow. Refer to examples here.

dbt docs generate          # generate docs
dbt docs serve --port 8081 # starts local server (by default docs server runs on 8080 port, it may cause conflict with Trino in case of local development)

Using Custom schemas

By default, all dbt models are built in the schema specified in your target. But sometimes you wish to build some of the models in a custom schema. In order to do so, use the schema configuration key to specify a custom schema for a model. See here for the documentation. It is important to note that by default, dbt will generate the schema name for a model by concatenating the custom schema to the target schema, as in: <target_schema>_<custom_schema>.

Prepared statements

The dbt seed feature uses Trino's prepared statements.

Python's http client has a hardcoded limit of 65536 bytes for a header line.

When executing a prepared statement with a large number of parameters, you might encounter following error:

requests.exceptions.ConnectionError: ('Connection aborted.', LineTooLong('got more than 65536 bytes when reading header line')).

The prepared statements can be disabled by setting prepared_statements_enabled to true in your dbt profile (reverting back to the legacy behavior using Python string interpolation). This flag may be removed in later releases.

Contributing

Release process

Before doing a release, it is required to bump the dbt-trino version by triggering release workflow version-bump.yml. The major and minor part of the dbt version are used to associate dbt-trino's version with the dbt version.

Next step is to merge the bump PR and making sure that test suite pass.

Finally, to release dbt-trino to PyPi and GitHub trigger release workflow release.yml.

Code of Conduct

Everyone interacting in the dbt project's codebases, issue trackers, chat rooms, and mailing lists is expected to follow the PyPA Code of Conduct.

About

The trino adpter plugin for dbt (https://getdbt.com)

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Python 96.3%
  • Shell 2.2%
  • Other 1.5%