When it's 'too late' is when they eventually need to bring in someone (read consultant) to look over why their 'well-thought' system is slower than a dog and absolutely useless...

A...

This behaviour is a documented property of the count() function - it's not quite what Jesper asked for.

/Kenneth

For the most parts, we 'get answers to what we ask'... So, check your data. Does it look like you expect? Easiest way to validate an update is to do...

I don't think you can avoid looping in one way or other in this case, and as noted, REPLACE doesn't deal well with wildcards... but PATINDEX does. So, you can...

If you want to avoid the warning, you can set ansi_warnings off

/Kenneth

Also, the ansi settings is a session setting, so imagine the chaos if col = col would behave differently depending on each connections specific settings..

Your problem seems similar to the problem how to calculate someones age. You want to count years, but within days precision depending on if the date of the birthday has...

I don't think that the ansi setting applies in this case. Imagine what would happen if it would..

Take your everyday inner join:

select foo from bar a join foobar b on...

The only 'good' thing about people as clueless as this, is that they do provide for a neverending source of work-opportunities for those not-so-clueless...

/Kenneth

I don't think that it's designed that way...

For example 1, you have a comparison between datavalue = datavalue, which by design should always evaluate to null, if either of the...

It should work just as fine with ascii 160 as any other.. I believe that it's a codepage problem rather than with replace itself.

ASCII 160 is in the high ASCII...

Hmm... Maybe better check your original db..? I can't see how any objects can disappear when doing a detach-attach. Make sure that the PKs you miss really is there in...

What datatype is your edeb column? If it's a float, then you're going to get some rounding errors and no real control on how many decimals the SUM() will produce.

The...

Well, have you tried it..?

While it looks like it would work, it's imo unnecessary resourcedemanding to write the join with a derived table...

I think that there may be some misunderstanding on how you can 'nest' transactions. Thing is, that transactions cannot be nested in any other way than syntax.

I've posted a short...