Re: [Openflow-dev] OpenFlow searching.

[Daniele <d.t...@ic...>]

On Monday 09 December 2002 18:49, Tommaso Cucinotta wrote:
> Hi all,
>
> I understand that OpenFlow uses the ZODB storage facility
> in order to store permanent workflow data.

Right.

> As I don't know how the ZODB works, I'm just wondering
> about what could happen in case of a great number of
> process instances going on in the system. Did you ever
> test OpenFlow with some thousand of instances ?

Nope. We trust in ZCatalog: it should be able to efficiently index such 
number of objects. Afterall none of the store objects have full-text indexes.

> As far as I can understand from a very fast look at the
> code, the Catalog object inside openflow is sometimes
> used to find objects. Is this a way of dealing with
> the just cited issue ? If yes, what are the index-ed
> items inside the catalog (process definition ids,
> activity ids, ... ??).

Yes. 

The indexed objects are all of the openflow objects: processes, activities, 
transitions, instances and workitems. If you want just one kind of object you 
should discriminate using the 'meta_type' index.

> Do you know what kind of transactionality model is enforced
> within the ZODB (when used in the Zope default mode, I've read
> you can put ZODB above an external DBMS) ?

Probably someone can help you better than me. Still, if with 
"transactionality model" you mean you are worried about the 'rollback' on the 
ZODB transaction then rejoice: ZODB is fully transactional and any operation 
being aborted (intentionally or for some mishap including zope errors) 
cancels any ZODB changes of the current transaction.


Daniele =)