dbt seeds

Intermediate dbt

Mike Metzger

Data Engineer

What are dbt seeds?

CSV files to be loaded into data warehouse
Typically rarely changing sets of data
- List of countries
- List of postal codes
NOT meant for raw data

Seeds

¹ Photo by Maddi Bazzocco on Unsplash

Why?

Easy to manage
Easy to use in various scenarios
Source controllable

How are seeds defined?

Add CSV file to the seeds directory
Make sure the header is the first row
Import using the dbt seed command

zipcode,place,state
99553,Akutan,Alaska
99571,Cold Bay,Alaska
99583,False Pass,Alaska

bash> dbt seed

¹ Postal codes provided via https://github.com/zauberware/postal-codes-json-xml-csv

Further configuration

Several options
- Which schema?
- What database?
- Column quoting
- Column data types
Can be applied to the whole project or to individual seeds
Can be added to dbt_project.yml or seeds/properties.yml

Defining datatypes

Available data types depend on data warehouse
Typical ones are available
- Integer
- Varchar
- etc
If type is not defined, it is inferred based on the data

version: 2

seeds:
  - name: zipcodes
    config:
      column_types:
        zipcode: varchar(5)

Test

Support tests
Just like models and sources

version: 2

seeds:
  - name: zipcodes
    config:
      column_types:
        zipcode: varchar(5)
    columns:
      - name: zipcode
        tests:
          - unique

Accessing seeds

Available via the {{ ref() }} command
Behaves as a model after initial import

select * from
  {{ ref('zipcodes') }}

Let's practice!

Intermediate dbt