Avancerad användning av Jinja med dbt

Fallstudie: Bygg e-handelsdatamodeller med dbt

Susan Sun

Freelance Data Scientist

Jinja-sats: loop

Jinja satser omges av {% %}:

Loopar genom element

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

Exempel:

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

Jinja-sats: loop

Upprepad SQL:

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-sats: loop

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

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

Jinja-sats: loop

dbt compile-utdata:

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-sats: makron

Ett Jinja-makro:

Är en funktion skriven i Jinja
Är återanvändbar
Lagras i mappen /macros

COALESCE():

returnerar det första icke-null-värdet
syntax: COALESCE(value1, value2, ...)
t.ex. COALESCE(nickname, first_name)

Ett exempel på en makrofunktion:

En skärmbild av IDE:n med filen `coalesce_and_round.sql` öppen i editorn, som visar makrologiken.

Jinja-sats: makron

Jinja satser omges av {% %}:

En återanvändbar makrofunktion:

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

Exempel:

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

Jinja-sats: makron

Upprepad 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)

Skapa ett makro kallat coalesce_and_round.sql
Lagra i /macros

Anropa:

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

Hantera blanksteg i Jinja

Som skrivet:

{% 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,
...

Hantera blanksteg i Jinja

Kompilerat:




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

Hantera blanksteg i Jinja

{%- ... %} tar bort blanksteget före.
{% ... -%} tar bort blanksteget efter.
{%- ... -%} tar bort båda.
Ett exempel:

{%- 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,
...

Nu kör vi en övning!

Fallstudie: Bygg e-handelsdatamodeller med dbt