Re: [python-sybase] Sybase module 0.38pre1 released

[Phil P. <pyt...@ph...>]

> Sorry about resending this - I didn't send it from my registered email
address first time.

Hi,

Thanks very much for taking this over.

Like Paul, I posted a patch a while ago that doesn't seem to have made it
into 0.38.  Any chance you could add the following fix as well?

It looks like there are some missing braces in the implementation of
blk_describe.  At the moment it always returns (status, None).  Here is a
patch:

--- blk.c.old   2006-09-09 15:41:49.520844800 +0100
+++ blk.c.new   2006-09-09 15:42:56.657382400 +0100
@@ -86,10 +86,11 @@
         return NULL;
     }

-    if (status != CS_SUCCEED)
+    if (status != CS_SUCCEED) {
        if (self->debug)
            debug_msg(", None\n");
        return Py_BuildValue("iO", status, Py_None);
+    }

     fmt = datafmt_alloc(&datafmt, 0);
     if (fmt == NULL) {

Another point is that blk_describe doesn't seem to behave in the way that
the documentation describes.  The alloc_bufs example given in the docs
doesn't work because blk_describe raises a DatabaseError if the column
number is greater than the number of columns in the table (rather than
returning CS_FAIL).  One solution might be to update the example as
follows...

   def alloc_bufs(self, num):
       bufs = []
       try:
           while 1 :
               status, ofmt = self.blk.blk_describe(len(bufs) + 1)
ofmt.count = num
               bufs.append(DataBuf(ofmt))
       except Sybase.DatabaseError:
           pass
       self.bufs = bufs

Finally, we found that blk_describe reports the wrong maxlength for
numeric types.  We resolved this for our purposes by hacking databuf.c as
follows but I don't think that this is a good solution.  I think it would
be better to do something to blk_describe itself instead.

--- databuf.c   2002-12-24 03:23:39.000000000 +0000
+++ /tmp/databuf.c      2006-09-14 11:43:28.000000000 +0100
@@ -57,14 +57,15 @@
        self->fmt = ((CS_DATAFMTObj*)obj)->fmt;
        if (self->fmt.count == 0)
            self->fmt.count = 1;
-#ifdef HAVE_FREETDS
+//#ifdef HAVE_FREETDS
        /* Seems like FreeTDS reports the wrong maxlength in
         * ct_describe() - fix this when binding to a buffer.
         */
+    /* Seems like Sybase's blk_describe has the same problem - PCP */
        if (self->fmt.datatype == CS_NUMERIC_TYPE
            || self->fmt.datatype == CS_DECIMAL_TYPE)
            self->fmt.maxlength = sizeof(CS_NUMERIC);
-#endif
+//#endif
        if (allocate_buffers(self) == NULL) {
            Py_DECREF(self);
            return NULL;

Hope this is of use.

Regards,

Phil

-- 
Phil Porter



-- 
Phil Porter