Overview: I'm trying to create the situation in the subject line:
from the outside, the database handle thinks it's just one database,
but inside it's really two databases containing tables with the same
schema but not necessarily the same data.
Specifics: Where I work we have a web site with content determined
jointly by user interaction and by data in various tables in a very
large database. I would like to test what would happen if I add other
data to the database. Typically what I have to do is copy the
relevant parts of the very large database to my box, add in the new
data, and test locally. (The very large database is far too large to
copy between machines.)
There are a number of things that make this setup difficult:
1) Determining what is "relevant" is not always an easy task and is
often fraught with errors.
2) The very large database can be modified by many, many other people
at will. So, each time I perform a test, I have to go back to the
very large database and copy the relevant parts to my box.
3) Handles to the database are created through an intermediary module.
So, the call XXX::DBH->new('YYY') (where XXX::DBH is a subclass of
DBI::db) returns a $dbh to the correct database (names have been
changed to protect the innocent). This module relies on a
configuration file to turn 'YYY' into the correct parameters for
DBI->connect(); that means I'm always fiddling with the configuration
file, depending if I want to use the original very large database or
my local one.
4) I'd like to fiddle with the original code as little as possible,
including the intermediary module.
My thoughts have turned to something like DBD::Proxy, because then I
could just change the environment variable DBI_PROXY when needed to
switch between servers. But DBI::ProxyServer doesn't seem very easily
adaptable to this situation; it seems restricted to the rather
specific task of acting as a proxy for another driver.
I have also thought about writing a driver from scratch, using
SQL::Statement to help out, but that breaks the first rule of driver
writing ("Don't!"), and since this is for my job, the second rule
("Don't -- get someone else to do it for you!") isn't really an
option.
I'm not afraid of putting in the time to do this right, and I'm not
trying to get something for free. But I'd like some opinions about
what might be the correct approaches to this problem, especially if
there's something obvious that I've overlooked. (So OK, I'm trying to
get some advice for free :)
Thanks.
