Getting Started with dbt-tags package¶

In order to start the Tag-based masking journey in our dbt project, we have 6 essential steps as follows:

Now, let's go 🏃:

1. Configure the namespace (databse, schema)¶

Firstly, we should let dbt know where we would like to store the Tag objects (as well as the Masking Policies' definitions).

We'll put the dbt_tags's variables in dbt_project.yml file as below:

vars:
  # dbt_tags__database: MY_DB # (optional) default to `target.database` if not specified
  dbt_tags__schema: COMMON # (optional) default to `target.schema` if not specified

2. Decide to allow the specific tags only¶

ℹ️ Skip this step if all dbt tags are allowed. Otherwise, see the sample below:

vars:
  dbt_tags__allowed_tags:
    - pii_name
    - pii_email
    - ...

3. Decide to commit masking policies' definition¶

ℹ️ Skip this step if you decide not to use masking policies, but only tags!

Masking Policies' functions vary depending on the dbt project, so it's your responsibility to implement them in advance.

In the dbt root directory, let's create a new one within /macros folder e.g. /macros/mp-ddl.

For each tag name, we need a corresponding macro that holds the masking policy definition.

Given a sample, we have a tag named pii_name, we'll create a macro file as below:

-- File path: /macros/mp-ddl/create_masking_policy__pii_name.sql
{% macro create_masking_policy__pii_name(ns) -%}

  create masking policy if not exists {{ ns }}.pii_name as (val string)
    returns string ->
    case --/ your definition start here /--
      when is_role_in_session('ROLE_HAS_PII_ACCESS') then val
      else sha2(val)
    end;

{%- endmacro %}

{{ ns }} or ns stands for the schema namespace, let's copy the same!

ℹ️ If you want to have multiple masking policies of different data types (they must be different data types) to a single tag, follow these steps:

Given a sample, we have a tag named pii_null, we'll create a macro file as below:

-- File path: /macros/mp-ddl/create_masking_policy__pii_null.sql
{% macro create_masking_policy__pii_null(ns) -%}

  create masking policy if not exists {{ ns }}.pii_null_varchar as (val string)
    returns string ->
    case --/ your definition start here /--
      when is_role_in_session('ROLE_HAS_PII_ACCESS') then val
      else null
    end;

  create masking policy if not exists {{ ns }}.pii_null_number as (val number)
    returns number ->
    case --/ your definition start here /--
      when is_role_in_session('ROLE_HAS_PII_ACCESS') then val
      else null
    end;

{%- endmacro %}

We then must modify the optional var dbt_tags__policy_data_types in the dbt_project.yml file:

vars:
  dbt_tags__policy_data_types:
    - pii_null: ['varchar','number']

These must match the exact same data_type suffix that has been applied to the name of the masking policies in the create macro. This will then assign both of these to the single tag, rather than having to manage multiple of the same tag for the different data types.

Leaving any tags out of the dbt_tags__policy_data_types var definition means that it will expect only a single masking policy which has the exact same name as the tag.

4. Set tags on columns¶

To assign tags to columns, you follow the same process as you would apply dbt tags to columns normally. This is done in the model schema yaml files.

By default, this package assigns the name of the column as the value of the tag. Because of how dbt tags work, there is no out of the box way to assign values for the Snowflake tags, so a separator ("~") has been configured within dbt_tags to facilitate this.

Setting a value for a tag can be useful for Security Governance querying in Snowflake. Or it can be used within a masking policy to allow some dynamic functionality using the Snowflake function system$get_tag_on_current_column('fully.qualified.tag-name').

Looking at a model's schema yaml file:

• If you don't need a tag value

columns:
    - name: first_name
      description: Customer's first name. PII.
      tags:
        - pii_name

• If you do need a tag value

columns:
    - name: membership_number
      description: Customer's membership number. PII.
      tags:
        - pii_mask_last_x_characters~4

The value is then available to use either in the masking policy or in Snowflake.

ℹ️ dbt tags will only deploy tags that have been set on columns. If you have tags or masking policies which aren't assigned to columns, they won't be deployed.

5. Deploy resources (tags, masking policies)¶

❗We don't want to repeat this step on every dbt run(s).

Instead, let's do it as a step in the Production Release process (or manually).

Remove --args '{debug: true}' for a live run

• Deploy the dbt tags to DWH:

dbt run-operation create_tags --args '{debug: true}'

• Deploy the tag-based masking policy functions to DWH:

ℹ️ Skip this step if you decide not to use masking policies, but only tags!

dbt run-operation create_masking_policies --args '{debug: true}'

6. Apply tags to columns¶

ℹ️ Currently, only column tags are supported!

Add a new post-hook to the model level:

models:
  my_project:
    post-hook:
      - > # customize below `if` condition following your need
        {% if flags.WHICH in ('run', 'build') %}
          {{ dbt_tags.apply_column_tags() }}
        {% endif %}

7. Apply masking policies to tags¶

ℹ️ Skip this step if you decide not to use masking policies, but only tags!

Remove --args '{debug: true}' for a live run

dbt run-operation apply_mps_to_tags --args '{debug: true}'

Now, go to Snowflake and check the result! 🚀

Happy Masking 🎉