dbt-prql

Original docs at https://github.com/prql/dbt-prql

  1. As of Feb 2022, we're again considering how to best integrate with
  2. dbt more closely. Ideally a file with a `.prql` extension will just work™.
  3. If you're interested in this, subscribe or 👍 to
  4. <https://github.com/dbt-labs/dbt-core/pull/5982>.
  5. The original plugin is hosted here, but only works with a minority of
  6. dialects, and isn't a focus of development at the moment.

dbt-prql allows writing PRQL in dbt models. This combines the benefits of PRQL’s power & simplicity within queries, with dbt’s version control, lineage & testing across queries.

Once dbt-prql in installed, dbt commands compile PRQL between {% prql %} & {% endprql %} Jinja tags to SQL as part of dbt’s compilation. No additional config is required.

Examples

Simple example

  1. {% prql %}
  2. from employees
  3. filter (age | in 20..30)
  4. {% endprql %}

…would appear to dbt as:

  1. SELECT
  2. employees.*
  3. FROM
  4. employees
  5. WHERE
  6. age BETWEEN 20
  7. AND 30

Less simple example

  1. {% prql %}
  2. from in_process = {{ source('salesforce', 'in_process') }}
  3. derive expected_sales = probability * value
  4. join {{ ref('team', 'team_sales') }} [name]
  5. group name (
  6. aggregate (sum expected_sales)
  7. )
  8. {% endprql %}

…would appear to dbt as:

  1. SELECT
  2. name,
  3. sum(in_process.probability * in_process.value) AS expected_sales
  4. FROM
  5. {{ source('salesforce', 'in_process') }} AS in_process
  6. JOIN {{ ref('team', 'team_sales') }} USING(name)
  7. GROUP BY
  8. name

…and then dbt will compile the source and refs to a full SQL query.

Replacing macros

dbt’s use of macros has saved many of us many lines of code, and even saved some people some time. But imperatively programming text generation with code like if not loop.last is not our highest calling. It’s the “necessary” part rather than beautiful part of dbt.

Here’s the canonical example of macros in the dbt documentation:

  1. {%- set payment_methods = ["bank_transfer", "credit_card", "gift_card"] -%}
  2. select
  3. order_id,
  4. {%- for payment_method in payment_methods %}
  5. sum(case when payment_method = '{{payment_method}}' then amount end) as {{payment_method}}_amount
  6. {%- if not loop.last %},{% endif -%}
  7. {% endfor %}
  8. from {{ ref('raw_payments') }}
  9. group by 1

Here’s that model using PRQL1, including the prql Jinja tags.

  1. {% prql %}
  2. func filter_amount method -> s"sum(case when payment_method = '{method}' then amount end) as {method}_amount"
  3. from {{ ref('raw_payments') }}
  4. group order_id (
  5. aggregate [
  6. filter_amount bank_transfer,
  7. filter_amount credit_card,
  8. filter_amount gift_card,
  9. ]
  10. )
  11. {% endprql %}

As well the query being simpler in its final form, writing in PRQL also gives us live feedback around any errors, on every keystroke. Though there’s much more to come, check out the current version on PRQL Playground.

1

  1. Note that when <https://github.com/prql/prql/issues/82> is implemented, we
  2. can dispense with the s-string, and optionally dispense with the function.
  3. ```elm
  4. from {{ ref('raw_payments') }}
  5. group order_id (
  6. aggregate [
  7. bank_transfer_amount = amount | filter payment_method == 'bank' | sum,
  8. credit_card_amount = amount | filter payment_method == 'credit_card' | sum,
  9. gift_amount = amount | filter payment_method == 'gift_card' | sum,
  10. ]
  11. )
  12. ```
  13. or
  14. ```elm
  15. func filter_amount method -> amount | filter payment_method == method | sum
  16. from {{ ref('raw_payments') }}
  17. group order_id (
  18. aggregate [
  19. bank_transfer_amount = filter_amount 'bank'
  20. credit_card_amount = filter_amount 'credit_card'
  21. gift_amount = filter_amount 'gift_card'
  22. ]
  23. )
  24. ```

What it does

When dbt compiles models to SQL queries:

  • Any text in a dbt model between {% prql %} and {% endprql %} tags is compiled from PRQL to SQL before being passed to dbt.
  • The PRQL compiler passes text that’s containing {{ & }} through to dbt without modification, which allows us to embed Jinja expressions in PRQL. (This was added to PRQL specifically for this use-case.)
  • dbt will then compile the resulting model into its final form of raw SQL, and dispatch it to the database, as per usual.

There’s no config needed in the dbt project; this works automatically on any dbt command (e.g. dbt run) assuming dbt-prql is installed.

Installation

  1. pip install dbt-prql

Current state

Currently this is new, but fairly feature-complete. It’s enthusiastically supported — if there are any problems, please open an issue.

How does it work?

It’s some dark magic, unfortunately.

dbt doesn’t allow adding behavior beyond the database adapters (e.g. dbt-bigquery) or Jinja-only plugins (e.g. dbt-utils). So this library hacks the Python import system to monkeypatch dbt’s Jinja environment with an additional Jinja extension on Python’s startup2.

2

  1. Thanks to
  2. [mtkennerly/poetry-dynamic-versioning](https://github.com/mtkennerly/poetry-dynamic-versioning)
  3. for the technique.

This approach was discussed with the dbt team here and here.

This isn’t stable between dbt versions, since it relies on internal dbt APIs. The technique is also normatively bad — it runs a few lines of code every time the Python interpreter starts — whose errors could lead to very confusing bugs beyond the domain of the problem (though in the case of this library, it’s small and well-constructed™).

If there’s ever any concern that the library might be causing a problem, just set an environment variable DBT_PRQL_DISABLE=1, and this library won’t monkeypatch anything. It’s also fully uninstallable with pip uninstall dbt-prql.

Roadmap

Open to ideas; at the moment it’s fairly feature-complete. If we were unconstrained in dbt functionality:

  • If dbt allowed for external plugins, we’d enthusiastically move to that.
  • We’d love to have this work on .prql files without the {% prql %} tags; but with the current approach that would require quite invasive monkeypatching.
  • If we could add the dialect in automatically (i.e. prql dialect:snowflake), that would save a line per model.
  • If we could upstream this into dbt-core, that would be awesome. It may be on PRQL to demonstrate its staying power before that, though.

We may move this library to the https://github.com/prql/PyPrql or https://github.com/prql/prql repos. We’d prefer to keep it as its own package given the hackery above, but there’s no need for it to be its own repo.