​
Overview

Data modeling and integration is a core component of Turntable. We use SQLMesh for our data transformation infrastructure, giving you world-class modeling capabilities with simple developer ergonomics.

​
Creating a model

Models are created in the UI by clicking the 3 dots next to a folder or at the root of the model section. Models are plain SQL files with a preamble at the top.

-- models/my_model.sql

MODEL(
  name <schema>.<model_name>,
  tags [...],
  cron '@daily' or '@weekly' or '@monthly' or '@quarterly' or '@yearly' or '* * * * *',
  kind FULL or VIEW
);

select * from my_table

​
Building a project

Clicking ‘build project’ at the top does 3 operations under the covers:

  1. Diffs the models that haven’t changed. (Note: running multiple times is a safe operation)
  2. Runs the transformations for the models that have changed.
  3. Deploys these new models to be available in your project.

​
Creating public models (explores)

To have models appear in your explore, you need to create a bi model. This is useful when you transformations are complicated or involve staging models that you don’t want to surface to the rest of your team to use for analysis.

To publish a model, all you have to do is add the ‘bi’ tags to the model and click ‘build project’.

-- models/my_model.sql

MODEL(
  name db.my_model,
  -- add this line
  tags ['bi'],
  kind FULL
);

select * from my_table

Note: models without the ‘bi’ tag will not be deployed to the explore but can still be used for transformations.

​
Typecasting models

Sometimes, you may need to typecast a column to a different type. This is useful when you want to use a column in a metric or dimension that is a different type than the column is in your data warehouse. This often comes up with datetimes and timestamps from upstream tables.

-- models/my_model.sql
MODEL(
  name db.my_model,
  kind VIEW
);
select
  timestamp_column::datetime as timestamp_column,
  order_amount::float as order_amount,
  customer_id 
from my_table