Gérer les null

Création de rapports en SQL

Tyler Pernes

Learning & Development Consultant

Que signifie vraiment null ?

+--------+-----------------+
| order  | price_per_unit  |
|--------|-----------------|
| 1      | 4.50            |
| 2      | 2.25            |
| 3      | null            |
+--------+-----------------+

À traiter plus tard ?
Gratuit ?
Prix fixe ?

Problèmes liés aux null

soccer_games
+---------+-------+-------+
| game_id | home  | away  |
|---------|-------|-------|
| 123     | 3     | 2     |
| 124     | 2     | null  |
| 125     | null  | 1     |
+---------+-------+-------+

SELECT *, home + away AS total_goals
FROM soccer_games;

Problèmes liés aux null

+---------+-------+-------+--------------+
| game_id | home  | away  | total_goals  |
|---------|-------|-------|--------------|
| 123     | 3     | 2     | 5            |
| 124     | 2     | null  | null         |
| 125     | null  | 1     | null         |
+---------+-------+-------+--------------+

Problèmes liés aux null

SELECT 
    region, 
    COUNT(DISTINCT athlete_id) AS athletes
FROM summer_games AS s
JOIN countries AS c
ON s.country_id = c.id
GROUP BY region;

+----------+-----------+
| region   | athletes  |
|----------|-----------|
| BALTICS  | 42        |
| OCEANIA  | 62        |
| null     | 10        |
+----------+-----------+

Peu clair : que représente null ?

Correctif 1 : filtrer les null

original_table
+--------+-----------------+
| order  | price_per_unit  |
|--------|-----------------|
| 1      | 4.50            |
| 2      | 2.25            |
| 3      | null            |
+--------+-----------------+

SELECT *
FROM original_table
WHERE price_per_unit IS NOT NULL;

Correctif 1 : filtrer les null

+--------+-----------------+
| order  | price_per_unit  |
|--------|-----------------|
| 1      | 4.50            |
| 2      | 2.25            |
+--------+-----------------+

Correctif 2 : COALESCE()

Syntaxe : COALESCE(field, null_replacement)

SELECT 
    COALESCE(region,'Independent Athletes') AS region,
    COUNT(DISTINCT athlete_id) AS athletes
FROM summer_games AS s
JOIN countries AS c
ON s.country_id = c.id;

+-----------------------+-----------+
| region                | athletes  |
|-----------------------|-----------|
| BALTICS               | 42        |
| OCEANIA               | 62        |
| Independent Athletes  | 10        |
+-----------------------+-----------+

Correctif 2 : COALESCE()

soccer_games
+---------+-------+-------+
| game_id | home  | away  |
|---------|-------|-------|
| 123     | 3     | 2     |
| 124     | 2     | null  |
| 125     | null  | 1     |
+---------+-------+-------+

SELECT *, COALESCE(home,0) + COALESCE(away,0) AS total_goals
FROM soccer_games;

Correctif 2 : COALESCE()

+---------+-------+-------+--------------+
| game_id | home  | away  | total_goals  |
|---------|-------|-------|--------------|
| 123     | 3     | 2     | 5            |
| 124     | 2     | null  | 2            |
| 125     | null  | 1     | 1            |
+---------+-------+-------+--------------+

Null générés par une requête

Causes :

LEFT JOIN ne fait pas correspondre toutes les lignes
Aucune condition de l'instruction CASE n'est satisfaite
Et d'autres !

Mesurer l'impact des null

Proportion de lignes où la valeur est null

SELECT SUM(CASE when country IS NULL then 1 else 0 end) / SUM(1.00)
FROM orders;

+-------+
| .12   |
+-------+

Proportion des revenus où la valeur est null

SELECT SUM(CASE when country IS NULL then revenue else 0 end) / SUM(revenue)
FROM orders;

+-------+
| .25   |
+-------+

Passons à la pratique !

Création de rapports en SQL