Вот 2 примера использования MS SQL SERVER в SQL.
Пример установки:
create table test (id int identity primary key, val varchar(20) )
--id is a pk for the cursor so it can update using "where current of"
-- name a is not duplicated
-- name b is duplicated 3 times
-- name c is duplicated 2 times
insert test values('name a')
insert test values('name b')
insert test values('name c')
insert test values('name b')
insert test values('name b')
insert test values('name c')
Sql 2005 \ 2008: (Вычисление вычисленной таблицы)
begin tran; -- Computed table expressions require the statement prior to end with ;
with cte(val,row) as (
select val, row_number() over (partition by val order by val) row
--partiton is important. it resets the row_number on a new val
from test
where val in ( -- only return values that are duplicated
select val
from test
group by val
having count(val)>1
)
)
update cte set val = val + ltrim(str(row))
--ltrim(str(row)) = converting the int to a string and removing the padding from the str command.
select * from test
rollback
Sql 2000: (пример курсора)
begin tran
declare @row int, @last varchar(20), @current varchar(20)
set @last = ''
declare dupes cursor
for
select val
from test
where val in ( -- only return values that are duplicated
select val
from test
group by val
having count(val)>1
)
order by val
for update of val
open dupes
fetch next from dupes into @current
while @@fetch_status = 0
begin
--new set of dupes, like the partition by in the 2005 example
if @last != @current
set @row = 1
update test
--@last is being set during the update statement
set val = val + ltrim(str(@row)), @last = val
where current of dupes
set @row = @row + 1
fetch next from dupes into @current
end
close dupes
deallocate dupes
select * from test
rollback
Я откатил каждое из обновлений, потому что мой файл сценария содержит оба примера. Это позволило мне проверить функциональность без сброса строк в таблице.