Yeah, this definitely seems like a bug (and still exists in SQL 2017 CU9). Using OPTIMIZE FOR UNKNOWN on the query seems to resolve the issue (and probably makes sense...

I'd generally recommend avoiding using PARSENAME in that fashion, even though it might seem to work. It does a bunch of things you might not expect, for things like handling...

I'm not a fan of triggers in general, but recursive triggers are absolutely the spawn of satan. There's a really good reason they are disabled by default in SQL Server...

I had a stab at implementing your rules, although they seem a bit contradictory:

Select
 MULTI.App_id,
 TableB.App_acc,
 TableB.CID,
 Max(IsNull(MISS.MissingAccount, 0)) As MissingAccount,
 Max(IsNull(MULTI.MultiAccount, 0)) As MultiAccount

Most DDL statements can be rolled back in 2005, though I don't think TRUNCATE can (at least not without jumping through some hoops first). I have an inkling that you...

In terms of which tables to join on, either it's because you need data from that table or because the existence of a row in a table (or existence with...

If you have 3000 explicit values in an IN clause, you probably have a design problem and performance will suffer from SQL Server trying to parse your code. There will...

I'd be inclined to use a calendar table that includes columns to indicate the days that are likely to be excluded (whether that be weekends or a more general non-working...

If it's analysing data produced from a system that has run to completion, I think it can be achieved but you probably need to record additional data about the order...

Am I missing something, or would it be a lot easier as:

select
 datename(month,[date]),
 count(*)
from logs
group by
 year([date]),
 datename(month,[date])
order by

Inline sql basically turns the entire database into a massive number of global variables. If that sounds like a good idea to your developers, you need better developers....