FunTuple silently ignores ThOr in some cases

If I give the following list of functors to the Jpsi branch

PT_refit = F.MAP_INPUT(Functor=F.PT, Relations=DTFRelations),
variables_jpsi = FunctorCollection({
    'THOR_PT': F.PT,
    'LOKI_MAXPT': 'TRACK_MAX_PT',
    'THOR_CHI2' : F.CHI2,
    'DTF_PT': PT_refit 
})

None get added to FunTuple. If I comment out DTF_PT, I get the three others ones added. FunTuple does not issue any warning of a problem adding DTF_PT. Instead the message block

DimuonsTuple                           INFO Conducting checks with LoKi 
DimuonsTuple                           INFO Conducting checks with ThOr 
DimuonsTuple                           INFO Setting the properties of ParticleTupleProp objects!
DimuonsTuple                           INFO Instatiating LoKi functors!
DimuonsTuple                           INFO Instatiating ThOr functors!
FunctorFactory                         INFO Cache miss for functor: ::Functors::Track::TransverseMomentum{}, now trying cling with headers [Event/Particle.h, Functors/TrackLike.h]
FunctorFactory                         INFO Cache miss for functor: ::Functors::Track::Chi2{}, now trying cling with headers [Event/Particle.h, Functors/TrackLike.h]
FunctorFactory                         INFO Cache miss for functor: ::Functors::Composite::Mass{}, now trying cling with headers [Event/Particle.h, Functors/Composite.h]
FunctorFactory                         INFO Cache miss for functor: ::Functors::Track::Momentum{}, now trying cling with headers [Event/Particle.h, Functors/TrackLike.h]
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Track::Momentum{}
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Composite::Mass{}
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Track::TransverseMomentum{}
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Composite::Mass{}
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Track::Momentum{}
FunctorFactory                         INFO Reusing cling compiled factory for functor: ::Functors::Track::TransverseMomentum{}

becomes

DimuonsTuple                           INFO Conducting checks with LoKi 
DimuonsTuple                           INFO Setting the properties of ParticleTupleProp objects!
DimuonsTuple                           INFO Instatiating LoKi functors!

so ThOr is just ignored. (BTW Instantiating is mispelled).

This needs Rec!2504 (merged), Phys!970 (merged) and DaVinci!569 (merged) to test. I do

./run davinci run-mc --inputfiledb Upgrade_Bd2KstarMuMu_ldst Phys/DaVinci/options/DaVinciDB-Example.yaml --user_algorithms DaVinciExamples/python/DaVinciExamples/tupling/example-tupling-DTF-run-mc:main | tee log-DTF

added DPA-WP3 label

assigned to @mferrill and @amathad

changed title from FunTuple silently ignore ThOr in some cases to FunTuple silently ignores ThOr in some cases

Note that the functor itself is untested. I will do so using it as a filter as FunTuple doesn't like it.

This is most certainly related to how the functor is defined. Does DTF_PT have data dependency? Does it take as input the TES to relation tables (@emgabrie flavour tag functors also had data dependency if I remember). From a quick look I could not find where DTF_PT is defined and only found this).

Yes, DTF_PT is defined here Rec!2504 (diffs) . It takes as input a relations table and a functor.

There are two issues here:

1. The fact that FunTuple ignores stuff. That may be my problem, as I don't know yet if this functor works. Or it could be that it doesn't like input, which is a feature that should be supported at some point.
2. The fact that in this case FunTuple does not throw an exception in such a case, which I think is a bug.

I indeed saw this behaviour with the flavour tag functors as well, it was fixed by adding the data dependency (i.e. variables_b = FunctorCollection({'THOR_MISTAG': F.Mistag}) -> variables_b = FunctorCollection({'THOR_MISTAG': F.Mistag('/Event/HLT2/Hlt2BsToJpsiPhiLine/FlavourTagger')})).

But I do agree that it would be nice to catch such bugs so the other functors still work & the user gets a warning regarding the functor with data dependency.

Thanks @emgabrie . But here I do explicitly write PT_refit = F.MAP_INPUT(Functor=F.PT, Relations=DTFRelations) where the DTFRelations are an output of another algorithm.

... but with a spurious comma. Need to try again.

With the comma removed (I now explicitly pass F.MAP_INPUT(Functor=F.PT, Relations=DTFRelations) ) I get an error. That's what I want.

So to come back to the initial issue, the problem is that I was adding a training comma. It seems hard to reproduce. Closing now. I'll re-open if it comes back again.

closed