dbt-python-imports 0.1.1

dbt plugin to add `modules.import` method to Jinja context to import arbitrary Python modules

dbt-python-imports

This is a dbt plugin to allow importing arbitrary Python modules from Jinja templates.

Usage

1. Install dbt-python-imports via PyPI into the same Python environment as dbt itself.

2. Call modules.import("<module_name>") to import the given module from a dbt Jinja string.

   Examples:

   • modules.import("os.path").dirname
   • modules.import("requests").get (already a dependency of dbt-core)

3. Profit!

[!warning]

This allows arbitrary code execution. (That's kindof the point.)

Jinja's sandboxing is not foolproof (e.g. some adapters allow arbitrary reads/writes to disk), so you already shouldn't be running untrusted dbt code/packages. However the limited Jinja context available by default does currently make it harder to run arbitrary code.

Make sure you're not installing any packages (dbt or Python) whose source you haven't inspected.

Available dbt Jinja Contexts

Context	Available?
model & hook SQL	✅
model properties.yml	✅
dbt_project.yml	⚠️ (partial: vars work, configs seem to not)
profiles.yml	❌

See unit tests for more details on context availability.

Example

# download the artifacts from your latest Databricks dbt job run

$ echo '
{#- call via `dbt build --state="$(dbt run-operation -q fetch_remote_state)" --defer ... #}
{%- macro fetch_remote_state(base_path=invocation_args_dict.project_dir ~ "/" ~ var("remote_state_base_path", "target/remote-state"), job_name="dbt build") %}
  {#- https://databricks-sdk-py.readthedocs.io/en/latest/workspace/jobs/jobs.html #}
  {%- set jobs_api = adapter.config.credentials.authenticate().api_client.jobs %}
  {%- set job = jobs_api.list(name=job_name, limit=1) | first %}
  {%- set job_run = jobs_api.list_runs(job_id=job["job_id"], limit=1, completed_only=true, expand_tasks=true) | first %}
  {%- set task_run = job_run["tasks"] | first %}
  {%- set task_run_output = jobs_api.get_run_output(run_id=task_run["run_id"]) %}
  {%- set download_url = task_run_output.dbt_output.artifacts_link %}

  {%- set state_path = base_path ~ "/" ~ task_run["run_id"] %}
  {%- set tar_path = state_path ~ "/dbt-artifacts.tar.gz" %}

  {%- set system = modules.import("dbt_common.clients.system") %}
  {%- do system.make_directory(state_path) %}
  {%- do system.download(download_url, tar_path) %}
  {%- do system.untar_package(tar_path, state_path) %}

  {#- print file path to stdout #}
  {%- do print(state_path ~ "/target") %}
{%- endmacro %}
' > macros/fetch_remote_state.sql

$ dbt run-operation -q fetch_remote_state
.../target/remote-state/55665502170299/target