This is a follow up to last week's !27749 (merged)
I realised (a bit too late for that MR) that the new interface still had a couple of areas in which it could be improved.
- It is better to consume a
const SG::ReadHandle<CONTAINER>&
than aconst SG::ReadHandleKey<CONTAINER>&
as with the latter, one also needs access to theEventContext
to construct the actual handle. - The function can be templated on the
CONTAINER
type rather thanDataVector<OBJECT>
, we don't need to directly assign theOBJECT
. We can get away with a double-dereference instead (objectROI = accessor( **it )
)
@jmasik - I have been working more on https://twiki.cern.ch/twiki/bin/view/Atlas/TrigDecisionToolMTFeatureAccess and I will document the interface as in this MR and then send out a wider mail.
Another small change added is a try-catch in the TrigEDMChecker.cxx
as otherwise this gets tripped up by a photon chain which has an interface container but no aux. Will open a bug report for the underlying cause.
Edited by Tim Martin