SQL Server - index on a computed column?

Question

I'm joining to a table dozens of different times, and every time, I join (or filter) based on the results of a SUBSTRING of one of the columns (it's a string, but left-padded with zeros, and I don't care about the last four digits). As a result, even though this column is indexed and my query would use the index, it does a table scan because the SUBSTRING itself isn't indexed, so SQL Server has to calculate it for every row before as it's joining.

I'm looking for any ideas on how to speed up this process. Currently, there's a view on the table (It's a "SELECT * FROM", just to give the table a friendly name), and I'm considering adding a column to the view that's computed, and then indexing that. I'm open to other suggestions, though - any thoughts?

MORE DETAIL: I should have shared this to begin with. The table receives replication from our billing system, so editing the underlying table to add a computed column is not an option. Any computed column would have to be added to the view on the table. Also, the leading zeros aren't always leading zeros - they're sometimes other data that I'm not interested in. I suppose the real question is "How can I join to data in the middle of a VARCHAR column while also making use of an index? Full-text Search?"

Clarifying my example I'm simplifying, but essentially, let's say I'm trying to look up values in a column with the following values:

00000012345MoreStuff
00000012345Whatever
19834212345
Houses12345837443GGD
00000023456MoreStuff

I'm interested in rows where SUBSTRING(7,5)="12345", so I'd want rows 1-4, but not row 5. What I'm proposing is adding a column to my "SELECT *" view that has this substring in it, and then indexing based on that. Does that make more sense?

Answer 1

Change the column to two columns - the data you join on and the extra 4 characters. Using parts of a column slows things down as you hve seen

Answer 2

Can you re-phrase your filter criteria in terms of a LIKE 'something%' statement? (This is applicable to an index)

Answer 3

Assuming you have your fields in this format:

00Data0007
000000Data0011
0000Data0015

, you can do the following:

• Create a computed column: ndata AS RIGHT(REVERSE(data), LEN(data) - 4)

  This will transform your columns into the following:

  ataD00
  ataD000000
  ataD0000
  

• Create an index on that column

• Issue this query to search for the string Data:

  SELECT  *
  FROM    mytable
  WHERE   ndata LIKE N'ataD%'
          AND SUBSTRING(ndata, LEN(N'ataD') + 1, LEN(ndata)) = REPLICATE('0', LEN(ndata) - LEN('ataD'))
  

  The first condition will use an index for coarse filtering.

  The second will make sure that all leading characters (that became the trailing characters in the computed column) are nothing but zeros.

See this entry in my blog for performance detail:

• SQL Server: leading wildcard match using an index

Update

If you just want an index on SUBSTRING without changing your schema, creating a view is an option.

CREATE VIEW v_substring75
WITH SCHEMABINDING
AS
SELECT  s.id, s.data, SUBSTRING(data, 7, 5) AS substring75
FROM    mytable

CREATE UNIQUE CLUSTERED INDEX UX_substring75_substring_id ON (substring75, id)

SELECT  id, data
FROM    v_substring75
WHERE   substring75 = '12345'

Answer 4

Add a calculated column to your table and create an index on this column.

ALTER TABLE MyTable
Add Column CodeHead As LEFT(Code,Len(Code)-4)

Then create an index on this.

CREATE INDEX CodeHeadIdx ON MyTable.CodeHead