Utiliser les CTE avec Redshift

Introduction à Redshift

Jason Myers

Principal Architect

Expressions de table communes (CTE)

Ensemble de résultats temporaire
Simplifient les requêtes
Variante des sous-requêtes

Structures : sous-requête et CTE

SELECT division_name,
       revenue_total
       -- Subquery for top 10 divions by revenue
  FROM (SELECT division_id,
               SUM(revenue) AS revenue_total
          FROM orders
         GROUP BY division_id
         ORDER BY revenue_total DESC
         LIMIT 10) top_ten_divisions_by_rev
 WHERE revenue_total > 100000;

-- CTE for top 10 divions by revenue
WITH top_ten_divisions_by_rev AS (
  SELECT division_id,
         SUM(revenue) AS revenue_total
    FROM orders
   GROUP BY division_id
   ORDER BY revenue_total DESC
   LIMIT 10)


-- Main query
SELECT division_id,
       revenue_total
       -- FROM our CTE
  FROM top_ten_divisions_by_rev
 WHERE revenue_total > 100000;

CTE multiples

-- Top 10 divisions by revenue CTE
WITH top_ten_divisions_by_rev AS(
  SELECT division_id,
         SUM(revenue) AS revenue_total
    FROM orders
   GROUP BY division_id
   ORDER BY revenue_total DESC
   LIMIT 10),

-- Division ID and Name CTE
division_names AS(
  SELECT id AS division_id,
         name AS division_name
    FROM divisions)

-- Main query
SELECT division_name,
       revenue_total
       -- FROM Top 10 CTE
  FROM top_ten_divisions_by_rev
  -- Joining so we can use the division name
  JOIN division_names USING (DIVISION_ID)
 WHERE revenue_total > 100000;

CTE multiples, jointures entre CTE

```sql
-- Selecting the division names
WITH division_names AS(
  SELECT id AS division_id,
         name AS division_name
    FROM divisions),
-- Selecting the top ten divisions by revenue
top_ten_divisions_by_rev AS(
  SELECT division_name,
         SUM(revenue) AS revenue_total
    FROM orders
    -- Joining in the division_names CTE
    JOIN divisions USING (division_id)
   GROUP BY division_name
   ORDER BY revenue_total DESC
   LIMIT 10);

Loi de Kernighan

Tout le monde sait que le débogage est deux fois plus difficile que d'écrire un programme au départ. Donc, si vous faites preuve du maximum d'ingéniosité en l'écrivant, comment pourrez-vous jamais le déboguer ?

Performance des CTE

Comme les sous-requêtes
Mieux que les sous-requêtes si réutilisées

Code SQL vs ordre d'exécution

Passons à la pratique !

Introduction à Redshift