[DB-SIG] DBAPI-2.0 clarifications

[M.-A. Lemburg]

Federico Di Gregorio wrote:
> 
> Scavenging the mail folder uncovered brian zimmer's letter:
> 
> > I believe they are required to see the changes immediately.  Opening two
> > connections is the only way to isolate the transactions.  Essentially, and
> > please correct me if I'm wrong, a cursor is only a statement from a
> > connection and has no transactional ability on it's own, so the connection
> > must handle them as all together or commit on each one.
> 
> ok, this is clear. so what about putting this stuff in the DBAPI document?
> just to avoid unneccesary quastions on the same argument in the future...

Patches welcome ;-)
 
> > > ok. now, how are multiple SELECTs derived from calling executemany()
> > > on a SELECT statement supposed to work? all the results should be
> > > concatenated into a single set? or we can extend .nextset() to cover
> > > this situation?
> >
> > I have the same question.  Consider this example:
> [example snipped]
> 
> can please the dbapi authors comment on this? thanx...

I think the docs on .nextset() are clear about this. .fetchXXX()
APIs only work on the current result set and a call .nextset()
advances to the next available result set.

Example (untested):

cursor.execute('select * from mytable where id = ?', [(1,), (2,)])
resultsets = []
while 1:
    resultsets.append(cursor.fetchall())
    if not cursor.nextset():
        break

resultsets ... [[(1,'first')], [(2,'second')]]
 
> [snip]
> > > "These two fields can probably be safely set to None" is, imho, not good
> > > *unless* the api specifies that the two fields are optional. this is
> > > why i am asking all that questions and why (in the future) i will probably
> > > propose a little revision of the API to clarify the obscure points.
> >
> > For what it's worth I set them to None as well.
> 
> ok. lets say that all the fields, from 2 onward are optional and should
> be set to None if not implemented. what about putting this in the DBAPI?
> DBAPI editor?

Good idea. Could you send in patches for this ?
 
> > I've included my small testing framework as an example of how I test.  I
> > utilize PyUnit and can test just about any DB API 2.0 compliant
> > implementation.  I essentially read an xml document describing the factory
> > method which opens the connection and then run the connection through any
> > number of tests.  It works in jython as well as cpython with no additional
> > changes.  Try changing the sample 'test.xml' to your factory method and see
> > if it works, I'd be quite curious.  It works perfectly well for mxODBC and
> > zxJDBC at the moment as both are set in the xml document.
> 
> looking at it, thank you very much,
> federico

Perhaps we ought to put this somewhere on the db-sig page or
into the database topic guide... Andrew ?

-- 
Marc-Andre Lemburg
______________________________________________________________________
Company & Consulting:                           http://www.egenix.com/
Python Pages:                           http://www.lemburg.com/python/