dbt_coves 1.7.3

CLI tool for dbt users adopting analytics engineering best practices.

dbt-coves

Sponsor

Hosted VS Code, dbt-core, SqlFluff, and Airflow, find out more at Datacoves.com.

What is dbt-coves?

dbt-coves is a CLI tool that automates certain tasks for dbt, making life simpler for the dbt user.

dbt-coves generates dbt sources, staging models and property(yml) files by analyzing information from the data warehouse and creating the necessary files (sql and yml).

Finally, dbt-coves includes functionality to bootstrap a dbt project and to extract and load configurations from Airbyte.

Supported dbt versions

Version	Status
< 1.0	❌ Not supported
>= 1.0	✅ Tested

From dbt-coves 1.4.0 onwards, our major and minor versions match those of dbt-core. This means we release a new major/minor version once it's dbt-core equivalent is tested. Patch suffix (1.4.X) is exclusive to our continuous development and does not reflect a version match with dbt

Supported adapters

Feature	Snowflake	Redshift	BigQuery
dbt project setup	✅ Tested	✅ Tested	✅ Tested
source model (sql) generation	✅ Tested	✅ Tested	✅ Tested
model properties (yml) generation	✅ Tested	✅ Tested	✅ Tested

NOTE: Other database adapters may work, although we have not tested them. Feel free to try them and let us know so we can update the table above.

Here's the tool in action

Installation

pip install dbt-coves

We recommend using python virtualenvs and create one separate environment per project.

Command Reference

For a complete list of options, please run:

dbt-coves -h
dbt-coves <command> -h

Environment setup

You can configure different components:

Set up git repository of dbt-coves project

dbt-coves setup git

Set up dbt within the project (delegates to dbt init)

dbt-coves setup dbt

Set up SSH Keys for dbt project. Supports the argument --open_ssl_public_key which generates an extra Public Key in Open SSL format, useful for configuring certain providers (i.e. Snowflake authentication)

dbt-coves setup ssh

Set up pre-commit for your dbt project. In this, you can configure different tools that we consider essential for proper dbt usage: sqlfluff, yaml-lint, and dbt-checkpoint

dbt-coves setup precommit

Models generation

dbt-coves generate <resource>

Where <resource> could be sources, properties, metadata, docs or airflow-dags.

dbt-coves generate sources

This command will generate the dbt source configuration as well as the initial dbt staging model(s). It will look in the database defined in your profiles.yml file or you can pass the --database argument or set up default configuration options (see below)

dbt-coves generate sources --database raw

Supports Jinja templates to adjust how the resources are generated. See below for examples.

Every dbt-coves generate <resource> supports --no-prompt flag, which will silently generate all sources/models/properties/metadata without asking anything to the user.

Source Generation Arguments

dbt-coves can be used to create the initial staging models. It will do the following:

1. Create / Update the source yml file
2. Create the initial staging model(sql) file and offer to flatten VARIANT(JSON) fields
3. Create the staging model's property(yml) file.

dbt-coves generate sources supports the following args:

See full list in help

dbt-coves generate sources -h

--database
# Database to inspect

--schemas
# Schema(s) to inspect. Accepts wildcards (must be enclosed in quotes if used)

--select-relations
# List of relations where raw data resides. The parameter must be enclosed in quotes. Accepts wildcards.

--exclude-relations
# Filter relation(s) to exclude from source file(s) generation. The parameter must be enclosed in quotes. Accepts wildcards.

--sources-destination
# Where sources yml files will be generated, default: 'models/staging/{{schema}}/sources.yml'

--models-destination
# Where models sql files will be generated, default: 'models/staging/{{schema}}/{{relation}}.sql'

--model-props-destination
# Where models yml files will be generated, default: 'models/staging/{{schema}}/{{relation}}.yml'

--update-strategy
# Action to perform when a property file already exists: 'update', 'recreate', 'fail', 'ask' (per file)

--templates-folder
# Folder with jinja templates that override default sources generation templates, i.e. 'templates'

--metadata
# Path to csv file containing metadata, i.e. 'metadata.csv'

--flatten-json-fields
# Action to perform when JSON fields exist: 'yes', 'no', 'ask' (per file)

--overwrite-staging-models
# Flag: overwrite existent staging (SQL) files

--skip-model-props
# Flag: don't create model's property (yml) files

--no-prompt
# Silently generate source dbt models

Properties Generation Arguments

You can use dbt-coves to generate and update the properties(yml) file for a given dbt model(sql) file.

dbt-coves generate properties supports the following args:

--destination
# Where models yml files will be generated, default: '{{model_folder_path}}/{{model_file_name}}.yml'

--update-strategy
# Action to perform when a property file already exists: 'update', 'recreate', 'fail', 'ask' (per file)

-s --select
# Filter model(s) to generate property file(s)

--exclude
# Filter model(s) to exclude from property file(s) generation

--selector
# Specify dbt selector for more complex model filtering

--templates-folder
# Folder with jinja templates that override default properties generation templates, i.e. 'templates'

--metadata
# Path to csv file containing metadata, i.e. 'metadata.csv'

--no-prompt
# Silently generate dbt models property files

Note: --select (or -s), --exclude and --selector work exactly as dbt ls selectors do. For usage details, visit dbt list docs

Metadata Generation Arguments

You can use dbt-coves to generate the metadata file(s) containing the basic structure of the csv that can be used in the above dbt-coves generate sources/properties commands. Usage of these metadata files can be found in metadata below.

dbt-coves generate metadata supports the following args:

--database
# Database to inspect

--schemas
# Schema(s) to inspect. Accepts wildcards (must be enclosed in quotes if used)

--select-relations
# List of relations where raw data resides. The parameter must be enclosed in quotes. Accepts wildcards.

--exclude-relations
# Filter relation(s) to exclude from source file(s) generation. The parameter must be enclosed in quotes. Accepts wildcards.

--destination
# Where csv file(s) will be generated, default: 'metadata.csv'
# Supports using the Jinja tags `{{relation}}` and `{{schema}}`
# if creating one csv per relation/table in schema, i.e: "metadata/{{relation}}.csv"

--no-prompt
# Silently generate metadata

Metadata

dbt-coves supports the argument --metadata which allows users to specify a csv file containing field types and descriptions to be used when creating the staging models and property files.

dbt-coves generate sources --metadata metadata.csv

Metadata format: You can download a sample csv file as reference

database	schema	relation	column	key	type	description
raw	raw	_airbyte_raw_country_populations	_airbyte_data	Year	integer	Year of country population measurement
raw	raw	_airbyte_raw_country_populations	_airbyte_data		variant	Airbyte data columns (VARIANT) in Snowflake
raw	raw	_airbyte_raw_country_populations	_airbyte_ab_id		varchar	Airbyte unique identifier used during data load

Docs generation arguments

You can use dbt-coves to improve the standard dbt docs generation process. It generates your dbt docs, updates external links so they always open in a new tab. It also has the option to merge production catalog.json into the local environment when running in deferred mode, so you can run dbt-checkpoint hooks even when the model has not been run locally.

dbt-coves generate docs supports the following args:

--merge-deferred
# Merge a deferred catalog.json into your generated one.
# Flag: no value required.

--state
# Directory where your production catalog.json is located
# Mandatory when using --merge-deferred

Generate airflow-dags

dbt-coves generate airflow-dags

Translate YML files into their Airflow Python code equivalent. With this, DAGs can be easily written with some key:value pairs.

The basic structure of these YMLs must consist of:

• Global configurations (description, schedule_interval, tags, catchup, etc.)
• default_args
• nodes: where tasks and task groups are defined
  • each Node is a nested object, with it's name as key and it's configuration as values.
    • this configuration must cover:
      • type: 'task' or 'task_group'
      • operator: Airflow operator that will run the tasks (full module.class naming)
      • dependencies: whether the task is dependent on another one(s)
      • any key:value pair of Operator arguments

Airflow DAG Generators

When a YML Dag node is of type task_group, Generators can be used instead of Operators.

They are custom classes that receive YML key:value pairs and return one or more tasks for the respective task group. Any pair specified other than type: task_group will be passed to the specified generator, and it has the responsibility of returning N amount of task_name = Operator(params).

We provide some prebuilt Generators:

• AirbyteGenerator creates AirbyteTriggerSyncOperator tasks (one per Airbyte connection)
  • It must receive Airbyte's host and port, airbyte_conn_id (Airbyte's connection name on Airflow) and a connection_ids list of Airbyte Connections to Sync
• FivetranGenerator: creates FivetranOperator tasks (one per Fivetran connection)
  • It must receive Fivetran's api_key, api_secret and a connection_ids list of Fivetran Connectors to Sync.
• AirbyteDbtGenerator and FivetranDbtGenerator: instead of passing them Airbyte or Fivetran connections, they use dbt to discover those IDs. Apart from their parent Generators mandatory fields, they can receive:
  • dbt_project_path: dbt/project/folder
  • virtualenv_path: path to a virtualenv in case dbt has to be ran with another Python executable
  • run_dbt_compile: true/false
  • run_dbt_deps: true/false

Basic YML DAG example:

description: "dbt-coves DAG"
schedule_interval: "@hourly"
tags:
  - version_01
default_args:
  start_date: 2023-01-01
catchup: false
nodes:
  airbyte_dbt:
    type: task_group
    tooltip: "Sync dbt-related Airbyte connections"
    generator: AirbyteDbtGenerator
    host: http://localhost
    port: 8000
    dbt_project_path: /path/to/dbt_project
    virtualenv_path: /virtualenvs/dbt_160
    run_dbt_compile: true
    run_dbt_deps: false
    airbyte_conn_id: airbyte_connection
  task_1:
    operator: airflow.operators.bash.BashOperator
    bash_command: "echo 'This runs after airbyte tasks'"
    dependencies: ["airbyte_dbt"]

Create your custom Generator

You can create your own DAG Generator. Any key:value specified in the YML DAG will be passed to it's constructor.

This Generator needs:

• a imports attribute: a list of module.class Operator of the tasks it outputs
• a generate_tasks method that returns the set of "task_name = Operator()" strings to write as the task group tasks.

class PostgresGenerator():
    def __init__(self) -> None:
        """ Any key:value pair in the YML Dag will get here """
        self.imports = ["airflow.providers.postgres.operators.postgres.PostgresOperator"]

    def generate_tasks(self):
        """ Use your custom logic and return N `name = PostgresOperator()` strings """
        raise NotImplementedError

airflow-dags generation arguments

dbt-coves generate airflow-dags supports the following args:

--yml-path --yaml-path
# Path to the folder containing YML files to translate into Python DAGs

--dag-path
# Path to the folder where Python DAGs will be generated.

--validate-operators
# Ensure Airflow operators are installed by trying to import them before writing to Python.
# Flag: no value required

--generators-folder
# Path to your Python module with custom Generators

--generators-params
# Object with default values for the desired Generator(s)
# For example: {"AirbyteGenerator": {"host": "http://localhost", "port": "8000"}}

--secrets-path
# Secret files location for DAG configuration, i.e. 'yml_path/secrets/'
# Secret content must match the YML dag spec of `nodes -> node_name -> config`

Extract configuration from Airbyte

dbt-coves extract airbyte

Extracts the configuration from your Airbyte sources, connections and destinations (excluding credentials) and stores it in the specified folder. The main goal of this feature is to keep track of the configuration changes in your git repo, and rollback to a specific version when needed.

Full usage example:

dbt-coves extract airbyte --host http://airbyte-server --port 8001 --path /config/workspace/load/airbyte

Load configuration to Airbyte

dbt-coves load airbyte

Loads the Airbyte configuration generated with dbt-coves extract airbyte on an Airbyte server. Secrets folder needs to be specified separately. You can use git-secret to encrypt secrets and make them part of your git repo.

Loading secrets

Secret credentials can be approached in two different ways: locally or remotely (through a provider/manager).

In order to load encrypted fields locally:

dbt-coves load airbyte --secrets-path /path/to/secret/directory

# This directory must have 'sources', 'destinations' and 'connections' folders nested inside, and inside them the respective JSON files with unencrypted fields.
# Naming convention: JSON unencrypted secret files must be named exactly as the extracted ones.

To load encrypted fields through a manager (in this case we are connecting to Datacoves' Service Credentials):

--secrets-manager datacoves

--secrets-url https://api.datacoves.localhost/service-credentials/airbyte

--secrets-token <secret token>

Full usage example:

dbt-coves load airbyte --host http://airbyte-server --port 8001 --path /config/workspace/load/airbyte --secrets-path /config/workspace/secrets

Extract configuration from Fivetran

dbt-coves extract fivetran

Extracts the configuration from your Fivetran destinations and connectors (excluding credentials) and stores it in the specified folder. The main goal of this feature is to keep track of the configuration changes in your git repo, and rollback to a specific version when needed.

Full usage example:

dbt-coves extract fivetran --credentials /config/workspace/secrets/fivetran/credentials.yml --path /config/workspace/load/fivetran

Load configuration to Fivetran

dbt-coves load fivetran

Loads the Fivetran configuration generated with dbt-coves extract fivetran on a Fivetran instance. Secrets folder needs to be specified separately. You can use git-secret to encrypt secrets and make them part of your git repo.

Credentials

In order for extract/load fivetran to work properly, you need to provide an api key-secret pair (you can generate them here).

These credentials can be used with --api-key [key] --api-secret [secret], or specyfing a YML file with --credentials /path/to/credentials.yml. The required structure of this file is the following:

account_name: # Any name, used by dbt-coves to ask which to use when more than one is found
  api_key: [key]
  api_secret: [secret]
account_name_2:
  api_key: [key]
  api_secret: [secret]

This YML file approach allows you to both work with multiple Fivetran accounts, and treat this credentials file as a secret.

:warning: Warning: --api-key/secret and --credentials flags are mutually exclusive, don't use them together.

Loading secrets

Secret credentials can be approached via --secrets-path flag

dbt-coves load fivetran --secrets-path /path/to/secret/directory

Field naming convention

Although secret files can have any name, unencrypted JSON files must follow a simple structure:

• Keys should match their corresponding Fivetran destination ID: two words automatically generated by Fivetran, which can be found in previously extracted data.
• Inside those keys, a nested dictionary of which fields should be overwritten

For example:

{
  "extract_muscle": {
    // Internal ID that Fivetran gave to a Snowflake warehouse Destination
    "password": "[PASSWORD]" // Field:Value pair
  },
  "centre_straighten": {
    "password": "[PASSWORD]"
  }
}

Run dbt commands

dbt-coves dbt <arguments> -- <command>

Run dbt commands on special environments such as Airflow, or CI workers, with the possibility of changing dbt project location and activating a specific virtual environment in which running commands.

Arguments

dbt-coves dbt supports the following arguments

--project-dir
# Path of the dbt project where command will be executed, i.e.: /opt/user/dbt_project

--virtualenv
# Virtual environment path. i.e.: /opt/user/virtualenvs/airflow

Sample usage

dbt-coves dbt --project-dir /opt/user/dbt_project --virtualenv /opt/user/virtualenvs/airflow -- run -s model --vars \"{key: value}\"
# Make sure to escape special characters such as quotation marks
# Double dash (--) between <arguments> and <command> are mandatory

Settings

dbt-coves will read settings from .dbt_coves/config.yml. A standard settings files could look like this:

generate:
  sources:
    database: "RAW" # Database where to look for source tables
    schemas: # List of schema names where to look for source tables
      - RAW
    select_relations: # list of relations where raw data resides
      - TABLE_1
      - TABLE_2
    exclude_relations: # Filter relation(s) to exclude from source file(s) generation
      - TABLE_1
      - TABLE_2
    sources_destination: "models/staging/{{schema}}/{{schema}}.yml" # Where sources yml files will be generated
    models_destination: "models/staging/{{schema}}/{{relation}}.sql" # Where models sql files will be generated
    model_props_destination: "models/staging/{{schema}}/{{relation}}.yml" # Where models yml files will be generated
    update_strategy: ask # Action to perform when a property file already exists. Options: update, recreate, fail, ask (per file)
    templates_folder: ".dbt_coves/templates" # Folder where source generation jinja templates are located. Override default templates creating  source_props.yml, source_model_props.yml, and source_model.sql under this folder
    metadata: "metadata.csv" # Path to csv file containing metadata
    flatten_json_fields: ask

  properties:
    destination: "{{model_folder_path}}/{{model_file_name}}.yml" # Where models yml files will be generated
    # You can specify a different path by declaring it explicitly, i.e.: "models/staging/{{model_file_name}}.yml"
    update-strategy: ask # Action to perform when a property file already exists. Options: update, recreate, fail, ask (per file)
    select: "models/staging/bays" # Filter model(s) to generate property file(s)
    exclude: "models/staging/bays/test_bay" # Filter model(s) to generate property file(s)
    selector: "selectors/bay_selector.yml" # Specify dbt selector for more complex model filtering
    templates_folder: ".dbt_coves/templates" # Folder where source generation jinja templates are located. Override default template creating model_props.yml under this folder
    metadata: "metadata.csv" # Path to csv file containing metadata

  metadata:
    database: RAW # Database where to look for source tables
    schemas: # List of schema names where to look for source tables
      - RAW
    select_relations: # list of relations where raw data resides
      - TABLE_1
      - TABLE_2
    exclude_relations: # Filter relation(s) to exclude from source file(s) generation
      - TABLE_1
      - TABLE_2
    destination: # Where metadata file will be generated, default: 'metadata.csv'

  docs:
    merge_deferred: true
    state: logs/
    dbt_args: "--no-compile --select foo --exclude bar"

  airflow_dags:
    yml_path:
    dags_path:
    generators_params:
      AirbyteDbtGenerator:
        host: "{{ env_var('AIRBYTE_HOST_NAME') }}"
        port: "{{ env_var('AIRBYTE_PORT') }}"
        airbyte_conn_id: airbyte_connection

        dbt_project_path: "{{ env_var('DBT_HOME') }}"
        run_dbt_compile: true
        run_dbt_deps: false

extract:
  airbyte:
    path: /config/workspace/load/airbyte # Where json files will be generated
    host: http://airbyte-server # Airbyte's API hostname
    port: 8001 # Airbyte's API port
  fivetran:
    path: /config/workspace/load/fivetran # Where Fivetran export will be generated
    api_key: [KEY] # Fivetran API Key
    api_secret: [SECRET] # Fivetran API Secret
    credentials: /opt/fivetran_credentials.yml # Fivetran set of key:secret pairs
    # 'api_key' + 'api_secret' are mutually exclusive with 'credentials', use one or the other

load:
  airbyte:
    path: /config/workspace/load
    host: http://airbyte-server
    port: 8001
    secrets_manager: datacoves # (optional) Secret credentials provider (secrets_path OR secrets_manager should be used, can't load secrets locally and remotely at the same time)
    secrets_path: /config/workspace/secrets # (optional) Secret files location if secrets_manager was not specified
    secrets_url: https://api.datacoves.localhost/service-credentials/airbyte # Secrets url if secrets_manager is datacoves
    secrets_token: <TOKEN> # Secrets auth token if secrets_manager is datacoves
  fivetran:
    path: /config/workspace/load/fivetran # Where previous Fivetran export resides, subject of import
    api_key: [KEY] # Fivetran API Key
    api_secret: [SECRET] # Fivetran API Secret
    secrets_path: /config/workspace/secrets/fivetran # Fivetran secret fields
    credentials: /opt/fivetran_credentials.yml # Fivetran set of key:secret pairs
    # 'api_key' + 'api_secret' are mutually exclusive with 'credentials', use one or the other

env_var

From dbt-coves 1.6.28 onwards, you can consume environment variables in you config file using "{{env_var('VAR_NAME', 'DEFAULT VALUE')}}". For example:

generate:
  sources:
    database: "{{env_var('MAIN_DATABASE', 'dev_database')}}"
    schemas:
      - "{{env_var('DEV_SCHEMA', 'John')}}"
      - "{{env_var('STAGING_SCHEMA', 'Staging')}}"

Telemetry

dbt-coves has telemetry built in to help the maintainers from Datacoves understand which commands are being used and which are not to prioritize future development of dbt-coves. We do not track credentials nor details of your dbt execution such as model names. The one detail we do use related to dbt is the anonymous user_id to help us identify distinct users.

By default this is turned on – you can opt out of event tracking at any time by adding the following to your dbt-coves config.yaml file:

disable-tracking: true

Override generation templates

Customizing generated models and model properties requires placing template files under the .dbt-coves/templates folder.

There are different variables available in the templates:

• adapter_name refers to the Adapter's class name being used by the target, e.g. SnowflakeAdapter when using Snowflake.
• columns contains the list of relation columns that don't contain nested (JSON) data, it's type is List[Item].
• nested contains a dict of nested columns, grouped by column name, it's type is Dict[column_name, Dict[nested_key, Item]].

Item is a dict with the keys id, name, type, and description, where id contains an slugified id generated from name.

dbt-coves generate sources

Source property file (.yml) template

This file is used to create the sources yml file

source_props.yml

Staging model file (.sql) template

This file is used to create the staging model (sql) files.

staging_model.sql

Staging model property file (.yml) template

This file is used to create the model properties (yml) file

staging_model_props.yml

dbt-coves generate properties

This file is used to create the properties (yml) files for models

model_props.yml

Thanks

The project main structure was inspired by dbt-sugar. Special thanks to Bastien Boutonnet for the great work done.

Authors

• Sebastian Sassi @sebasuy -- Datacoves
• Noel Gomez @noel_g -- Datacoves
• Bruno Antonellini -- Datacoves

About

Learn more about Datacoves.

⚠️ dbt-coves is still in development, make sure to test it for your dbt project version and DW before using in production and please submit any issues you find. We also welcome any contributions from the community

Metrics