Re: Query termination clariffication

[Henrik Nordberg <HNordberg@lbl.gov>]

R Jeff Porter x6983 wrote:
> 
> As Arie suggests, if the user-code ~QO (e.g. it goes out
> of scope) "done" is sent to the QE but not to the QM. The
> QM keeps getting more files to any EIs. It will not clear
> its cache until the last file is sent.  I'm not sure
> what happens when the QM tells the QE its finished - after
> the QE has already cleared its memory about the query.

QM never tells QE that it finished. (IMHO)

> This will also not allow the user to later abort the query
> execution. Since niether the QE nor (currently) the QO keeps
> track of the EIs (except perhaps in the QO's initial construction
> of the EI), it becomes the user's responcibility to keep
> the QO active should it need to access any of the QO
> interface functions. I don't see any problems with this
> in our current single process user-code/QO.  This could be
> an issue in the distributed-EI mode.
> 
> I agree that if ~QO occurs on a "paused" query, then
> "abort" should be sent not "done".

Yes.

How about couple the lives of the QO and EI. After all, you can't
have an EI without a QO. Not a very functional anyway. We could kill the
EI(s) from the QO dtor (~QO()).

 - Henrik
___________________________________________________________
Henrik Nordberg <HNordberg@lbl.gov> http://nordberg.lbl.gov
Scientific Data Management Research Group
Lawrence Berkeley Laboratory