throw DataNotAvailable from getCacheDirective?

[markaddleman]

I'm finally getting around to writing support for coordinating multiple executions within a single continuous command.  To recap, the intent is to execute a query once to retrieve data and re-execute only when any of the data sources indicate that new/changed data is available.  When any one data source signals a reexecution, the other executions should be reexecuted but return cached data.

The problem is, using the translator caching API means there's no subsequent execution available from which to throw the DNA.  In fact, the only point of control is the getCacheDirective.  What do you think of providing support for throwing DNA from there?

[shawkins]

> What do you think of providing support for throwing DNA from there?

I think we went over this before, but didn't make a change.  What probably makes the most sense is for the CacheDirective to have an invalidation flag, such that we'll attempt to pull fresh results and replace the old.  Then leave the handling of DNA up to the rest of the normal execution cycle.

[markaddleman]

We may have.  I tried looking up the old thread but couldn't find it.  In any case, sticking an invalidation flag into the CacheDirective makes sense to me.  The closest Jira I can find is [#TEIID-2079] Coordinate closing continuous executions across windows and data sources - JBoss Issue Tracker which doesn't really fit the bill.

Should I open an issue?

[shawkins]

Yes go ahead and open one.

[markaddleman]

It turns out it is there:  https://issues.jboss.org/browse/TEIID-2301