This function will run on each line which takes more than than writing it in the select. Secondly, David's version will run only once, therefore much faster than yours...

I should send the link to sushila... proves I'm not the only one who HTHs guys to add to the posts count .

.

BTW if you nuke Quebec yours truly will not be here anymore to save your arse when you screw up a query

No. Sql server picks up the best sequence to apply the where condition. You can't really have much control on that part (unless you use a case statement).

Takes the underscore.

Uglier is the right word here .

That's more like it .

Shall we dare to ask him why he's doing that?

Have you tested each case (view/function)?? The optimizer should come up with the same plan.

Since you have a left join am indexed view is out of the question.

I think...

You know better... I never use that datatype.

You're almost right .

declare @str varchar(32)

select @str = '0010A0'

select (ltrim(replace(@str,'0',' ')))

--1 A

select REPLACE(ltrim(replace(@str,'0',' ')), ' ', '0')

--10A0

They all work fine.. some are just a tad faster... and easier to undertsand .

You seem to be the expert on paterns. Do you have a tutorial or a list of exemples of real life pattern searches (bol doesn't offer much in real life...

You'll have to use dynamic sql in this case. Make sure that the users cannot access those values in any way (other than picking the db from a combo).

Make one proc per server and call the good server with an if/else.

can you paste the plans... maybe we can spot something you missed?