D'abord quelques données de test :
create table seasons (seasonId int primary key
, "date" datetime not null unique
, tariffId int not null)
insert into seasons values (1, '2009-01-01', 1)
insert into seasons values (2, '2009-01-02', 1)
insert into seasons values (3, '2009-01-03', 2)
insert into seasons values (4, '2009-01-04', 3)
insert into seasons values (5, '2009-01-05', 3)
insert into seasons values (6, '2009-01-06', 1)
insert into seasons values (7, '2009-01-07', 1)
insert into seasons values (8, '2009-01-08', 3)
-- add a tarrif with a datespan larger than 2
insert into seasons values (9, '2009-01-09', 4)
insert into seasons values (10, '2009-01-10', 4)
insert into seasons values (11, '2009-01-11', 4)
En s'appuyant sur la réponse de Dave Barker, dans les vues en ligne, ajoutez row_number() afin que nous sachions quelles sont les premières valeurs minimales, lesquelles sont les secondes, etc. par tarifId. (En fait, puisqu'une date ne peut pas avoir plus d'un tarifId, nous n'avons pas besoin de partitionner par tarifId.)
SELECT MinValues.Seasonid, MinValues.Date, MaxValues.Date, MaxValues.tariffid
FROM (
SELECT *, row_number() over (partition by tariffId order by "date") as RN
FROM [dbo].[Seasons] tbl1
WHERE NOT EXISTS (SELECT *
FROM [dbo].[Seasons] tbl2
WHERE tbl1.seasonid - tbl2.seasonid = 1
AND tbl1.tariffId = tbl2.tariffId)) as minValues
JOIN (
SELECT *, row_number() over (partition by tariffId order by "date") as RN
FROM [dbo].[Seasons] tbl1
WHERE NOT EXISTS (SELECT *
FROM [dbo].[Seasons] tbl2
WHERE tbl2.seasonid - tbl1.seasonid = 1
AND tbl1.tariffId = tbl2.tariffId)) as maxValues
ON MinValues.TariffId = MaxValues.tariffId
and MinValues.RN = MaxValues.RN
order by MinValues.Date
Résultats :
1 2009-01-01 00:00:00.000 2009-01-02 00:00:00.000 1
3 2009-01-03 00:00:00.000 2009-01-03 00:00:00.000 2
4 2009-01-04 00:00:00.000 2009-01-05 00:00:00.000 3
6 2009-01-06 00:00:00.000 2009-01-07 00:00:00.000 1
8 2009-01-08 00:00:00.000 2009-01-08 00:00:00.000 3
9 2009-01-09 00:00:00.000 2009-01-11 00:00:00.000 4
