FreeTDS Development Group

[Brian Bruns <camber AT ais.org>]

On Fri, 30 Aug 2002, ZIGLIO Frediano wrote:

> This structure has many field...
> I think we can join column_textvalue/column_offset and
> column_textsize/cur_row_size (why cur_row_size and not cur_column_size ??)
> 
> column_textvalue/column_offset
> Use a dataptr to point to the buffer. If this is a blob point to an
> allocated buffer. If this is a constant length buffer point to buffer
> allocated per row. Is quite easy to test is this pointer point to row buffer
> or not.

We used to do this, but tell me who will free this?  rows can be buffered 
(under dblib) and you may end up with 100, 200, 1000 or more pointers to 
text values.  It gets to be a mess.  Right now we simply don't buffer text 
values.
 
> column_textsize/cur_row_size
> just use one of them (best to use column_cur_size).

cur_row_size was added after the fact, not sure by whom (mark?), to 
support dbdatlen().  The name refers to the fact that it is the size of 
the column in the current row.  I agree that they could be combined.  I 
generally like to avoid overloading the meaning of column, but I think 
this is clear enough.

> freddy77
> 

Brian