[Derby] Allow select/delete/update conditions with comparison to NULL using '='

SQL has three valued logic, where boolean expressions can be "true", "false" or "NULL" where NULL means "I don't know." It's unfortunate that they chose NULL to denote this, since every where else in computerland it means "no value."

When you say "title = NULL", you're saying "is title equal to some value I don't know?" Of course, if you don't know what you're comparing to, you can't know whether the comparison is true or false. So the result of the expression is not "true" or "false" but NULL. If you want to test whether a value is NULL or not, you use "title is NULL," which always returns true or false.

http://en.wikipedia.org/wiki/Null_(SQL)

Unless I'm misunderstanding the issue...

I believe you misunderstand the issue

The construct

Entry.find(:all, :conditions => ['title = ?', my_expression])

is a common ActiveRecord usage that tries to find records in the "entries" table where the title field matches the given expression (my_expression). This construct works flawlessly on PostgreSQL (and maybe other DBMSs), but DerbyDB does not allow it. It demands the "IS NULL" construct if my_expression evaluates to nil. This is correct SQL behaviour (I expect). Nevertheless, it makes programming much harder since you need to make a special case for "nil" values for "my_expression". Like this:

if my_expression
  Entry.find(:all, :conditions => ['title = ?', my_expression])
else
  Entry.find(:all, :conditions => ['title IS NULL'])
end

My patch allows the first construct, avoiding treating nil as a special case. The result is less coding and a more DMBS independent ActiveRecord-JDBC usage.

This isn't a JRuby issue, so I'm removing the fix version. It'll get into AR-JDBC soon hopefully.

Uwe, There are a couple of things that it would be nice if you could fix and resubmit on this patch:

• = NULL probably needs to be =\s*NULL
• !=\s*NULL should become IS NOT NULL
• INSERT GUARD needs to be added similiar to UPDATE check (and override_execute similiar to what is done in jdbc_adapter.rb)

It is a bummer that we are post-processing what AR generates, but since the programmer themselves do this in condition string there is not much we can do. A possibly better method would be to monkey-patch AR method which processes conditions. I think that is rife with peril however. So for now your strategy seems the most manageable right now.

Sorry I was slow on this. I had iteration delivery last week. Thanks for fixing it!