Rambling Intro, Nostalgia, and Crankiness
This week I got a request troubleshoot a legacy Microsoft Access application that has been floating around our company for ages, but still gets used daily because dang it, it does the job and always has. Seems like most companies that are standardized on MS Office have a few of these lurking out on the network.
Earlier in my career I did a ton of work in MS Access and have garnered a reputation within my company for being an expert in this oft maligned platform so I got the call to look into the problem. It had been quite a while since I’d done any real work on MS Access and I’d forgotten about how quirky it could be. Also, I am more than a little disappointed at how Microsoft has mangled the UI of my old friend Access in the 2007 version. It is almost painful to work with it as a power-user in the current incarnation.
So anyway, the issue turned out to be that someone increased the length of a field in the underlying SQL Server table linked into the Access application. They increased it past the magical border (255 characters) between what Access considers a text and a memo field, which imposed new limits on how it could be used. In particular, Access doesn’t allow either end of a join in a query to be a memo field.
The solution is painfully simple. So much so that I have to wonder why Access doesn’t just do it behind the scenes. Perhaps it is just trying to discourage you from building databases that link on big text fields for your own good (see “The Caveat” below).
The trick is to move the join into the WHERE clause of the query like so:
SELECT Table1.*, Table2.* FROM Table1, Table2 WHERE (Table1.MemoField=table2.MemoField);
Here is the same query in the query builder for those of you who prefer it to the SQL view:
Access will raise nary a complaint if you run this query which is logically equivalent to the one it abhorred. That’s all there is to it.
A final note. It is a definite database smell for an application to be joining tables on long text fields and will likely be the source of some performance issues in a database of non-trivial size. However, as was the case for the application I was tweaking, joining on long text fields is sometimes necessary in queries used for data clean-up, validation, or replication. Still, use this type of join with caution avoiding it whenever possible.