[PEAK] Adding psycopg support to PEAK

[John Landahl]

On Wed, 12 Nov 2003 09:18:42 -0500, Phillip J. Eby <pje at telecommunity.com> 
wrote:

> John, if you've tested this, I'll go ahead and add it to 
> peak.storage.SQL.

I used it for a data converter I was working on and was able to transfer 
~15,000 records from one database to another with no hitches.  It could 
still use more real world testing, but I'd say it's pretty functional 
already.

> Maybe instead of an 'AddingPsycopgSupport' page, this could become a 
> document on how to add support for an unsupported DBAPI database?

That's what I want to do next, after finishing up some really urgent 
deliverables I'm working on.  I was going to e-mail the info, then thought 
it would be both more readable and more accessible if I put it on the 
wiki.  [I actually only took on the task because I had finally had enough 
of pgdb's "internal errors" that kept popping up while writing the 
aforementioned data converters.  There were no such problems with psycopg, 
and as Victor pointed out it certainly does seem faster.  (Coming soon, 
some actual numbers).]

> I.e., take my previous notes on how to identify what types a driver uses,
> and whether it uses .values, and explain how to add support.

I must have missed those notes; could you supply an href?  I figured out 
the supported types through trial and error at first, then by reading 
psycopg's source code.

> Actually, it might be worthwhile to split it into two how-tos: adding an 
> address scheme, and creating a DB driver.

Good idea.  This was my first foray into adding an address scheme, but I 
could experiment some more and come up with a few more examples and turn 
it into its own howto.