Thread: [htoolkit-users] HSQL patch

SourceForge Headquarters 225 Broadway Suite 1600 San Diego, CA 92101 +1 (858) 422-6466

Hi,

I just added a new SqlBind instance to HSQL.hsc

instance SqlBind a => SqlBind (Maybe a)

which returns NULL values as Nothing. I needed this because I'm using
an algorithm that uses NULL values and I didn't want to have to call
getFieldValueMB all the time. Actually I'm using my own wrapper module
which uses heterogenous lists so handling NULL values through the type
system is just much more convenient for me:

http://ofb.net/~frederik/futility/src/HSQLExtras.hs

Anyway, it was a bit of work to add this. I had to add a new method to
the SqlBind class

	fromSqlCStringLen :: SqlType -> CString -> Int -> IO (Maybe a)
	fromSqlCStringLen sqlType cstr cstrLen =
            if cstr == nullPtr then return Nothing
            else fromNonNullSqlCStringLen sqlType cstr cstrLen

to be used instead of 

	fromNonNullSqlCStringLen :: SqlType -> CString -> Int -> IO (Maybe a)

This is because apparently NULL values originally get into the system
as nullPtr's, so SqlBind needs to be able to know about those
nullPtr's in order to process NULL values.

Now getFieldValueMB passes fromSqlCStringLen rather than
fromNonNullSqlCStringLen to the driver. The drivers have to be updated
to make use of this feature - currently they intercept NULL values and
return Nothing, but they have to be changed so that they pass these
along to value extractor (which will return Just Nothing in the case
of my new instance). If they aren't changed, the old behavior will
just be preserved so it's *not* necessary to change all of them at
once. I updated MySQL (tested) and PostgreSQL (not tested), so you can
see what has to be done.

Now I can write something like this (it uses another wrapper,
http://ofb.net/~frederik/futility/src/DBTool.hs)

    query0 (sql "insert into $bar values (2,null)")
    Just ((a::Maybe Int) :. (b::Maybe Int) :. Nil) <- query1 (sql "select * from $bar limit 1")
    print (a,b)

==> (Just 2,Nothing)

I hope this patch is suitable. If it isn't accepted then I'll have to
create my own fork of the library, because I depend on this feature in
other stuff I'm distributing.

Cheers,

Frederik