[ZPatterns] TransactionAgents: small bug in BoundaryAgents

[Steve Spicklemire]

Hi Steve,

	It would help me a lot if you could describe your application a 
bit, and why it needs TransactionAgents (i.e., elaborate on what 
TransactionAgents allow you to do, or not do, in your application that 
would be impossible/difficult with stock ZPatterns.)

thanks!
-steve

P.S. Did you ever get to test that Python2.1/DynPersist.pyd? It seems to 
import OK for me.

On Thursday, August 2, 2001, at 05:09 PM, Steve Alexander wrote:

> I've just been doing lots of work with large amounts of data, 
> subtransactions, dataskins, and _p_jar.cacheMinimize(3).
>
> What I've found is that DataSkins really need to have their 
> per-transaction _v_ attributes cleared on subtransaction commits as 
> well as when the outer transaction commits. Otherwise, you can get a 
> situation where _v_ attributes you thought would be around until the 
> end of the transaction have actually disappeared.
>
> This isn't an issue for standard-issue ZPatterns, because such 
> attributes are cleared on subtransaction commits, as well as other 
> transaction boundaries.
>
> The patch to TransactionAgents is to change this:
>
>   if not subtransaction:
>       self.runBoundaryAgents() # ignores errors
>
> to this:
>
>   #if not subtransaction:
>   self.runBoundaryAgents() # ignores errors
>
> Find it at around line 70 of __init__.py, in the new_commit method.
>
> I'll release a new TransactionAgents soon with this fix in it.
>
>
> I'm using TransactionAgents with the modified ZPatterns on a fairly 
> complex system. It works, and having IndexingAgents available as well 
> as  RuleAgents has meant that I could get rid of some hacky 
> subtransaction commits in my application code. The difference with 
> having two kinds of agent is that the IndexingAgents for all DataSkins 
> run after the RuleAgents for all DataSkins for a transaction. The 
> equivalent case in standard ZPatterns is that you can only set the 
> order of agents to run for a single Specialist.
>
> Of course, Phillip Eby described this in the ZPatterns wiki months and 
> months ago. I've only properly understood what it was all for now :-)
>
> --
> Steve Alexander
> Software Engineer
> Cat-Box limited
>
>
> _______________________________________________
> ZPatterns mailing list
> ZPatterns@eby-sarna.com
> http://www.eby-sarna.com/mailman/listinfo/zpatterns