Ici vous pouvez voir la définition simplifiée et la solution de votre problème (si je vous ai bien compris):http ://sqlfiddle.com/#!9/48a2e1/1
CREATE TABLE foundry
(
lru varchar(50) NOT NULL,
client int NOT NULL,
purchase_date date,
price int NOT NULL
);
INSERT INTO foundry (lru, client, purchase_date, price) VALUES
("article1", 4001, "01-01-16", 100),
("article1", 4001, "01-01-17", 200),
("article1", 4001, "01-02-16", 300),
("article1", 4001, "01-04-16", 400),
("article1", 4001, "01-06-16", 500),
("article1", 4001, "01-08-16", 600),
("article1", 4001, "01-10-16", 700),
("article1", 4001, "01-11-16", 800),
("article1", 4002, "01-01-16", 900),
("article1", 4002, "01-07-16", 1000),
("article1", 4002, "01-12-16", 1100);
Fondamentalement, nous avons un tableau à quatre colonnes :lru (nom de l'article), client, date d'achat et un certain prix.
La solution ressemble à :
SELECT lru, client, avg(price), COUNT(*) as total_items,
MONTHNAME(STR_TO_DATE(L, '%m')) as start_month, MONTHNAME(STR_TO_DATE(R, '%m')) as end_month FROM foundry,
(
SELECT 1 as L, 6 as R
UNION ALL
SELECT 2, 7
UNION ALL
SELECT 3, 8
UNION ALL
SELECT 4, 9
UNION ALL
SELECT 5, 10
UNION ALL
SELECT 6, 11
UNION ALL
SELECT 7, 12
) months
WHERE month(purchase_date) >= L AND month(purchase_date) <= R
GROUP BY lru, client, L, R
L'idée est :
- Générer toutes les combinaisons possibles de mois :1-6, 2-7, ..., 7,12
- Joindre les données source avec la combinaison de mois générée
- Utiliser AVG avec GROUP BY
Et le résultat :
lru client avg(price) total_items start_month end_month
article1 4001 300 5 January June
article1 4001 400 3 February July
article1 4001 500 3 March August
article1 4001 500 3 April September
article1 4001 600 3 May October
article1 4001 650 4 June November
article1 4001 700 3 July December
article1 4002 900 1 January June
article1 4002 1000 1 February July
article1 4002 1000 1 March August
article1 4002 1000 1 April September
article1 4002 1000 1 May October
article1 4002 1000 1 June November
article1 4002 1050 2 July December
