On Tue, 29 Oct 2002, Tim Bunce wrote:
> That seems like a rather poor design.
>
I take it you have worked with PostgreSQL much :)
> Beware that your semantics for SQL_BINARY must match those of ODBC.
> I think ODBC requires SQL_BINARY values to be formated as pairs of
> hex chars.
So you are saying that when execute is called on a column that is bound as
SQL_BINARY it should expect a string of hex digits? something like,
execute("fffcfdfa010c")? Then all execute need do is slap a X''
around the string before submitting to the db (or in PGs case converting
everything to octal escapes).
Do you have a reference for the ODBC spec that you could point me to?
If it be the case that the values submitted to execute() are required to
be in the form of 'deadbeef', then the current behavour of execute() is
broken and has been broken since DBD::Pg .98(?). How would you suggest
handling this, a) Call it a bug, deprecate the usage of SQL_BINARY with
binary data in the next version, and then fix it in the version after?
b) leave it as is c) just fix it and say that the next version is not
compatible with earlier versions of DBD::Pg wrt binding as SQL_BINARY d)
something else.
-r
