Was curious about the change that came about in the thread:
Re: [Dbdpg-general] benchmarking old Pg and DBD::Pg
... which (I gather) hinged on an efficiency issue, of type-mismatch between
passed parameters and indexed columns.
I'm asking because a fair amount of code broke in our app,
when "?" appeared in a context where PG couldn't determine a type from context,
but VARCHAR was appropriate.
In general, Perl, and perl programmers, assume it's okay for any scalar to
reduce to a string. One even claims that the DBI spec actually says that default
binding is VARCHAR; I have yet to track down the reference.
I was under the impression that type-mismatch of constants vs indexed-column
types was something that was fixed in Postgres8; should the above "fix" then not
be a non-issue? To un-break code here, I'm having to reverse the patch on
"Dreams come true, not free." -- S.Sondheim.
pgsql-interfaces by date
|Next:||From: Van Tress,Michael J - PGPW||Date: 2005-09-09 20:56:37|
|Subject: ecpg -C INFORMIX|
|Previous:||From: Norman Yamada||Date: 2005-09-08 19:51:44|
|Subject: Re: getting julian number from date in libpq|