更改具有相关索引的表

Question

As part of a system upgrade, we need to be able to update a database structure to conform to the newest version.

I'm writing a tool to that end, but am stuck on the correct procedure for updating a column which is used in an index.

If a column which is in an index must be changed, which approach is likely to be the least problematic:

1) Disable the index, alter the column, re-enable the index

2) Drop the index, alter the column, re-create the index

There are a number of instances where this change must be applied, and I would like to reduce the overall time as much as possible, hence my preference for not recreating the index if it can be avoided.

Answer 1

I did some tests,it seems you cannot alter index columns .

test data:
create table idd
(
id int identity(1,1),
name char(33),
name2 varchar(40) null
)

create unique clustered index nci_id on idd(id)
create index nci_test1 on idd(name2)

--disable index
alter index nci_test1 on idd disable

--alter column
alter table idd
alter column test1 varchar(100) not null

below is the error:
Msg 5074, Level 16, State 1, Line 36
The index 'nci_test1' is dependent on column 'test1'.
Msg 4922, Level 16, State 9, Line 36
ALTER TABLE ALTER COLUMN test1 failed because one or more objects access this column.

This is obvious since I have clustered key.so what happens if I drop clustered key and then do an alter operation on non clustered index key column,Result is same.We can alter index columns only after dropping them

drop index  [nci_id] on idd

--alter column
alter table idd
alter column test1 varchar(100) not null 

I think you got some idea on what is the impact

1.We have to drop clustered key first ..heavy tlog writes,since non clustered key also have to change there pointers
2.Again we have to rebuild indexes

You can only drop them.Further I would suggest you go ahead with this approach(since either way you have to drop clustered index) of

1.Drop index
2.Alter column datatype
3.recreate index

Further try changing database recovery model to simple,so as to minimize TLOG writes prior to this operation and also add nocheck option .Below questions has some interesting answers which may help you

https://dba.stackexchange.com/questions/48872/quickly-change-null-column-to-not-null

How do you add a NOT NULL Column to a large table in SQL Server?