[PEAK] Re: Cross-platform stored-procedure support

[Phillip J. Eby]

At 12:49 PM 12/13/03 -0500, Jeff Kowalczyk wrote:
>Phillip J. Eby wrote:
> > The Python DBAPI is somewhat underspecified (and poorly supported) for
> > stored procedure invocation. (...)
> > I'd like to give PEAK a stored procedure mechanism that is:
> > * Cross-platform
> > * Convenient
> > * Covers all functionality of the platform(s)
>
>I am usually the one to bring this up, and it always plays to crickets,
>but a pythonic rendition(*) of ADO.Net's disconnected-data DataSet and
>supporting classes such as SqlCommand) would be a potential solution to
>these issues.
>
>O'Reilly's ADO.Net Center:
>http://www.ondotnet.com/topics/dotnet/ado.net
>
>Executing Stored Procedures Through a SqlCommand Object:
>http://www.ondotnet.com/lpt/a/1157#58556

I read the latter link; it looks to me like it brings nothing new to the 
table.  That is, it's no different than the Python DBAPI, in that all of 
the interesting parts (such as parameter naming and type conversion) are 
unspecified.

My objective here is to treat stored procedures as Python callables, taking 
Python values and returning Python results.

As for the dataset stuff, that sort of thing will be handled via 
peak.query, for the most part.  I don't plan to have an explicit buffered 
relvar type, but there *will* be a framework for managing any kind of data 
as streams of row-like things, and if you really need an in-memory dataset 
for something you could probably implement one.  But the normal (non-GUI) 
use case for in-memory datasets will be handled via transactions.