Complex usage of jinja with dbt

Casestudy: e-commercedatamodellen bouwen met dbt

Susan Sun

Freelance Data Scientist

Jinja-statement: lus

Jinja-statements staan tussen {% %}:

Lussen door items

Sjabloon:
```
{% for ... %} ... {% endfor %}
```

Voorbeeld:

{% for order_status in order_statuses %}
  SUM(
      CASE WHEN status = '{{ order_status }}' 
        THEN order_id 
      END)
{% endfor %}

Jinja-statement: lus

SQL herhalen:

SELECT 
    user_id,
    SUM(CASE WHEN status = 'Shipped' THEN 1 ELSE 0 END) 
        AS num_orders_Shipped,
    SUM(CASE WHEN status = 'Complete' THEN 1 ELSE 0 END) 
        AS num_orders_Complete,
    SUM(CASE WHEN status = 'Processing' THEN 1 ELSE 0 END) 
        AS num_orders_Processing,        
FROM {{ ref('stg_looker__orders') }}
GROUP BY 1

Jinja-statement: lus

{% set order_statuses = ['Shipped', 'Complete', 'Processing'] %}

SELECT 
    user_id,
    -- Jinja-lus
    {% for order_status in order_statuses %}
        SUM(CASE WHEN status = '{{ order_status }}' THEN 1 ELSE 0 END) 
        -- Geparametriseerde kolomnaam
        AS num_orders_{{ order_status }}
    {% endfor %}
FROM {{ ref('stg_looker__orders') }}
GROUP BY 1

Jinja-statement: lus

dbt compile-output:

SELECT 
    user_id,
    SUM(CASE WHEN status = 'Shipped' THEN 1 ELSE 0 END) 
        AS num_orders_Shipped,
    SUM(CASE WHEN status = 'Complete' THEN 1 ELSE 0 END) 
        AS num_orders_Complete,
    SUM(CASE WHEN status = 'Processing' THEN 1 ELSE 0 END) 
        AS num_orders_Processing,        
FROM "dbt"."main"."stg_looker__orders"
GROUP BY 1

Jinja-statement: macro's

Een Jinja-macro:

Is een functie, geschreven in Jinja
Is herbruikbaar
Staat in de map /macros

COALESCE():

geeft de eerste niet-null waarde terug
syntax: COALESCE(value1, value2, ...)
bijv. COALESCE(nickname, first_name)

Voorbeeld van een macrofunctie:

Een screenshot van de IDE, met het bestand `coalesce_and_round.sql` geopend in de editor met de macrologica.

Jinja-statement: macro's

Jinja-statements staan tussen {% %}:

Een herbruikbare macro-functie:

Sjabloon:
```
{% macro ... %} ... {% endmacro %}
```

Voorbeeld:

{% macro coalesce_and_round(column_name, decimal_places = 2) %}
      ROUND(COALESCE({{ column_name }}, 0), {{ decimal_places }})
{% endmacro %}

Jinja-statement: macro's

Herhaalde SQL

ROUND(COALESCE(sales_amount, 0), 2) AS sales_amount,
ROUND(COALESCE(cost_of_goods_sold, 0), 2) AS cost_of_goods_sold,

DRY (Don't Repeat Yourself)

Maak een macro coalesce_and_round.sql
Sla op in /macros

Aanroepen:

{{ coalesce_and_round('sales_amount', 2) }} AS sales_amount,
{{ coalesce_and_round('cost_of_goods_sold', 2) }} AS cost_of_goods_sold,

Witruimte in Jinja beheren

Zoals geschreven:

{% set traffic_source_values = ['Adwords', 'Email', 'Facebook'] %}
{% set browser_values = ['Chrome', 'Firefox', 'Safari', 'IE', 'Other'] %}

SELECT 
    user_id,
    COUNT(DISTINCT session_id) AS num_web_sessions,
...

Witruimte in Jinja beheren

Gecompileerd:




SELECT 
    user_id,
    COUNT(DISTINCT session_id) AS num_web_sessions,
...

Witruimte in Jinja beheren

{%- ... %} stript de witruimte ervoor.
{% ... -%} stript de witruimte erna.
{%- ... -%} stript beide.
Voorbeeld:

{%- set traffic_source_values = ['Adwords', 'Email', 'Facebook'] -%}
{%- set browser_values = ['Chrome', 'Firefox', 'Safari', 'IE', 'Other'] -%}

SELECT 
    user_id,
    COUNT(DISTINCT session_id) AS num_web_sessions,
...

Laten we oefenen!

Casestudy: e-commercedatamodellen bouwen met dbt