Skip to main content

About flags variable

The flags variable contains values of flags provided on the command line.

Example usage:

flags.sql
{% if flags.FULL_REFRESH %}
drop table ...
{% else %}
-- no-op
{% endif %}

The list of available flags is defined in the flags module within dbt-core.

Recommended use cases include:

  • different materialization logic based on "run modes," such as flags.FULL_REFRESH and flags.STORE_FAILURES
  • running hooks conditionally based on the current command / task type, via flags.WHICH

Note: It is not recommended to use flags as an input to parse-time configurations, properties, or dependencies (ref + source). Flags are likely to change in every invocation of dbt, and their parsed values will become stale (and yield incorrect results) in subsequent invocations that have partial parsing enabled. For more details, see the docs on parsing.

invocation_args_dict​

For the full set of information passed from the CLI—subcommand, flags, arguments—you can use invocation_args_dict. This is equivalent to the args dictionary in run_results.json.

models/my_model.sql
-- invocation_args_dict:
-- {{ invocation_args_dict }}

-- dbt_metadata_envs:
-- {{ dbt_metadata_envs }}

select 1 as id

The invocation_command key within invocation_args_dict includes the entire subcommand when it compiles:

$ DBT_ENV_CUSTOM_ENV_MYVAR=myvalue dbt compile -s my_model

12:10:22  Running with dbt=1.6.0-b8
12:10:22  Registered adapter: postgres=1.6.0-b8
12:10:22  Found 1 seed, 1 model, 349 macros
12:10:22
12:10:22  Concurrency: 5 threads (target='dev')
12:10:22
12:10:22  Compiled node 'my_model' is:
-- invocation_args_dict:
-- {'log_format_file': 'debug', 'log_level': 'info', 'exclude': (), 'send_anonymous_usage_stats': True, 'which': 'compile', 'defer': False, 'output': 'text', 'log_format': 'default', 'macro_debugging': False, 'populate_cache': True, 'static_parser': True, 'vars': {}, 'warn_error_options': WarnErrorOptions(include=[], exclude=[]), 'quiet': False, 'select': ('my_model',), 'indirect_selection': 'eager', 'strict_mode': False, 'version_check': False, 'enable_legacy_logger': False, 'log_path': '/Users/jerco/dev/scratch/testy/logs', 'profiles_dir': '/Users/jerco/.dbt', 'invocation_command': 'dbt compile -s my_model', 'log_level_file': 'debug', 'project_dir': '/Users/jerco/dev/scratch/testy', 'favor_state': False, 'use_colors_file': True, 'write_json': True, 'partial_parse': True, 'printer_width': 80, 'print': True, 'cache_selected_only': False, 'use_colors': True, 'introspect': True}

-- dbt_metadata_envs:
-- {'MYVAR': 'myvalue'}

select 1 as id

flags.WHICH​

flags.WHICH is a global variable that gets set when you run a dbt command. If used in a macro, it allows you to conditionally change behavior depending on the command currently being executed. For example, conditionally modifying SQL:

{% macro conditional_filter(table_name) %}
    {# Add a WHERE clause only during `dbt run`, not during `dbt test` or `dbt compile` #}

    select *
    from {{ table_name }}
    {% if flags.WHICH == "run" %}
        where is_active = true
    {% elif flags.WHICH == "test" %}
        -- In test runs, restrict rows to keep tests fast
        limit 10
    {% elif flags.WHICH == "compile" %}
        -- During compile, just add a harmless comment
        -- compile mode detected
    {% endif %}
{% endmacro %}

The following commands are supported:

flags.WHICH valueDescription
"build"Build and test all selected resources.
"clean"Remove artifacts like target directory and packages.
"clone"Clone models and other resources.
"compile"Compile SQL, but do not execute.
"debug"Test connections and validate configs.
"deps"Download package dependencies.
"docs"Generate and serve documentation.
"environment"Workspace environment commands (cloud CLI).
"help"Show help for commands and subcommands.
"init"Bootstrap a new project.
"invocation"For interacting with or inspecting current invocation (cloud CLI).
"list"List resources.
"parse"Parse project and report errors, but don’t build/test.
"retry"Retry the last invocation from the point of failure.
"run"Execute models.
"run-operation"Invoke arbitrary macros or SQL ops.
"seed"Load CSV(s) into the database.
"show"Inspect resource definitions or materializations.
"snapshot"Execute snapshots.
"source"Validate freshness and inspect source definitions.
"test"Schema and data tests.
"version"Display dbt version.

Was this page helpful?

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

0
Edit this page