On 20 Sep 2015, at 7:15pm, James K. Lowden <jklowden at schemamania.org> wrote:
> Simon Slavin <slavins at bigfraud.org> wrote: > >> Constructions like this >> >>> INSERT INTO _TAG_EXISTS_RESULT_ ( NAME, BOOL ) >>> SELECT 'evil little sister' >> >> should be rewritten so that you are not trying to do a SELECT in the >> middle of your INSERT. > > Why in the world would you say that? That's the SQL assignment > function, the equivalent of > > A = A + B I would argue that that would be UPDATE, not INSERT. But I see your side of the issue. > Anything you'd do to rewrite it would result in something more complex, > doubly so if it included the atomicity guarantee of the original. Suppose the SELECT doesn't return anything. Do you still want to do the INSERT ? Do you now have to look up the values to INSERT elsewhere ? Put the commands in a transaction if that worries you. You're right -- I should probably have written "I preffer" instead if making it an absolute statement. Simon.
