Hey! This is a small sample project for a specific use case: a table that contains information for multiple events, with the same column having different meanings.

Project Overview

The project showcases two different ways to handle this case: output as a single table or separate tables.

The project can be separated in three pieces: seeds, models and macros. Long story short:

• seeds are .csv files that are transformed into a table. In this case, we have only one seed, for the events (events_done.csv).
• models are .sql files that are going to be materialized in tables/views. In this case, we have 4 models.
  • single_table/fct_event_done.sql: Model that basically gets all the events in the events_done seed and enrichens them, by adding a surrogate key and a event-specific transformation into a column called event_detail
  • separate_tables/: In this other approach, each event gets it's own table
    • fct_file_uploaded.sql: Event for when a file is uploaded. The value is transformed into file_name, with a file_extension column being added.
    • fct_purchased_done.sql: Event for when a purchase is completed. The value is transformed into revenue, with a revenue_decimals column being added.
    • fct_rate_given.sql: Event for when a file is uploaded. The value is transformed into rate_given, the number of stars a user rated the content with.
• macros are .sql files that are templated in other .sql files. These can receive multiple parameters and template them. In this case, they are only being used in the single_table/fct_event_done.sql, acting as a separator of the logic from the main .sql file. In this specific PoC, the transformations are very simple and could be done on the single_table/fct_event_done.sql file.

Running the project

Running the project requires the setup of a profiles.yml file in ~/.dbt/profiles.yml. Then, the following commands can be used:

• dbt seed
• dbt run

Extra stuff

The models were done using incremental models, which means that they compare against the existing table before running to only process the new events