Em Qui, 2006-12-07 às 20:56 -0600, John Goerzen escreveu:
> On Wed, Mar 07, 2007 at 07:26:25PM +0100, Marco Tulio Gontijo e Silva wrote:
> > Hello John!
> >
> > I haven't noticed that instance SqlType (Maybe a), it's a workaround,
> > but I still prefer the solution with my patch. In what do you consider
> > that it could be harmful? The problem for me is just that filter (==
>
> The problem is that you convert a SQL NULL value to a string
> "SqlNull".
>
> That string is indistinguishable from a field that contains the
> literal string "SqlNull". That, in fact, is why the SQL NULL value is
> not a string. It is specifically designed to be not representable as
> a string, and trying to represent it as a string is an error.
I pointed out two solutions, and this one I told that it was worse
because of this problem. I submitted a third solution latter, which was
a problem in the first, better, solution. Here it goes again:
[EMAIL PROTECTED]:~/dados/texto/codigo/haskell/hdbc/Database/HDBC$ diff
-ruN Types.hs Types.2.hs
--- Types.hs 2006-11-22 09:43:30.000000000 -0200
+++ Types.2.hs 2006-11-24 08:05:29.000000000 -0200
@@ -363,6 +363,8 @@
SqlEpochTime a == SqlEpochTime b = a == b
SqlTimeDiff a == SqlTimeDiff b = a == b
SqlNull == SqlNull = True
+ SqlNull == _ = False
+ _ == SqlNull = False
a == b = ((fromSql a)::String) == ((fromSql b)::String)
instance SqlType String where
[EMAIL PROTECTED]:~/dados/texto/codigo/haskell/hdbc/Database/HDBC$
What's the problem of defining that SqlNull compared to other type is
false? I know it's not strictly necessary, cause we can use fromSql,
but i still think it's better to just compare with SqlNull without
fromSql, don't you?
--
malebria
Marco Túlio Gontijo e Silva
Jabber (Google Talk): [EMAIL PROTECTED]
Correio (MSN): [EMAIL PROTECTED]
Skype: marcotmarcot
Telefone: 33346720
Celular: 98116720
Endereço: Rua Paula Cândido, 257/201
Gutierrez 3430-260
Belo Horizonte/MG Brasil
