Since we use different RoIs when running the precision SA muon finding than we do for the fast SA muon finding, we need to run the decoding again in the muon precision muon RoIs. Otherwise we may miss hits that are included in the updated RoIs.

Switching the msonly muon triggers to use exactly the same SA muon trigger sequence as other chains. This is to avoid having two different hypo algs attach the same feature twice in the navigation (which can lead to retrieving the incorrect features from the navigation later on in the monitoring).

Add the MuonIso RoiDescriptors o the EDM so that we can retrieve the collections
created in the iso view

Savanna Shaw authored 4 years ago and Vakhtang Tsulaia committed 4 years ago

The precision muon trigger hypo loops first on the previous decisions, then gets the event view for each decision, and finally retrievs the muons from the view. In the case where we have an input maker that merges based on feature, we run into the problem of having one decision per unique muon, which means we have N decisions for a given view, and so end up creating N new decisions per muon. To avoid this, I've introduced a check that the muon corresponds to the muon from the previous decision. Since this is only a problem for combined muons, we first check if the extrapolated muon spectrometer track for the combined muons matches the extrapolated track from the previous step's SA muons. In case there is no extrapolated track (which is the case for muons reconstructed by the InsideOut reconstruction), we check the dR between the combined muons and the previous step's muons. This makes sure we don't create new decisions multiple times for the same muon (which is reflected in the reference update), and avoids warnings of the form:

`WARNING Found 'feature' by multiple paths, but the links are coming from different places in the navigation graph`

as discussed in ATR-21548.

Muon algorithms where imported from `TrigMuonEFConfig` instead of
`TrigMuonEFConf`. This only worked because the former did a `import *`
of the latter.

Updating the invariant mass hypo used by the muon trigger to be a combo hypo tool:
- Removing the invariant mass hypo alg, since it's not longer needed
- Changing the invariant mass hypo tool to be a ComboHypoTool, and updating the logic a bit compared to what was in the hypo alg to make sure we a) avoid looking at the same muon twice, b) look only at muons that passed the combined muon hypothesis
- Remove the specific menu sequence for the unvariant mass hypo, and added instead a ComboHypoTool to the combined muon menu sequence for invariant mass chains.
- Updated the references, since this removes one step from the chain, but also because the counting for this last step wasn't quite correct before (it is now).

Changing the RoIs used to the RoIs that are not using the extrapolation to the IP (which can reduce the efficiency for SA muons not originating from the IP). This brings the Run 3 muon triggers a small step closer to the Run 2 triggers.

Due to the way the input maker was configured for the EF SA muon triggers, we were duplicating processing of RoIs when running ms-only and combined muon chains in the same steps. In order to avoid this:
- Have changed the input maker for the EF SA step to use the roi tool that fetches the RoIs from previous views (this allows us to use the same RoIs in both types of chains, instead of the using the L1 RoIs for ms-only chains and the L2 RoIs in the combined chains).
- Change the input maker for the EF SA step to merge based on the initialRoI, which removes the duplication of views
- Adjust the ViewCreatorFetchFromViewROITool to be able to fetch the RoIs from a specific view from earlier in the chain (default is to fetch from the view from the previous step)

Since we can find multiple SA muons in a single view, changing the input maker for the subsequent steps to merge-by-feature instead of merge-by-roi as discussed in ATR-21548.
Reference updated since this affects the way the features are counted.

Vladimir Lyubushkin authored 4 years ago and Frank Winklmeier committed 4 years ago

TrigBphysHypo: TrigMultiTrkHypo algorithm is modified to be used as ComboHypo at muComb and muEFCB steps of BLS trigger chains (ATR-20836, ATR-21092, ATR-21112)

Removing a few instances where we were defining a track particle container name (and also an algorithm name) globally within different muon sequences, which can cause problems depending on which order chains are configured in the menu (since the names were being overwritten). All the sequences that need the track particle collection name use the same default variable name that we define only once. All sequences that were changing the collection name did not have any later dependency on the new name in other sequences.

Savanna Shaw authored 5 years ago and Walter Lampl committed 5 years ago

Since the TrigMuonEFMSonlyHypo alg/tool was basically a copy of the TrigMuonEFCombinerHypo alg/tool, have combined them into a single hypo alg/tool (TrigMuonEFHypo). The only difference was what type of muons were being considered (MSonly vs Combined), so this is now a configurable property of the tool. The corresponding configuration has been updated to configure the different instances of the hypo alg/tool and the monitoring configuration has similarly been condensed.

A few updates related to the event view creation for L2 combined muons:
- Use the ViewCreatorFetchFromViewROITool when creating teh event views to spawn views from the RoIs created by the L2 SA algorithm. This allows us to easily use the varying RoI sizes defined in the L2 SA algorithm, depending on where in the detector the muon is.
- Record the output of the ViewCreatorFetchFromViewROITool to the EDM, and remove the recording of the 'forID' RoI collection that was created inside the L2 SA views. Now the RoI collection is recorded at the event level to the EDM and can be used in the offline monitoring (as needed in ATR-20850)
- Fix a bug in the creation of the output RoI descriptor in ViewCreatorFetchFromViewROITool (need to create the new roi descriptor from the input RoI itself, and not the pointer)

Disabling the error tuning for the muon triggers (it's not needed, and uses quite a lot of CPU)

As pointed out in ATR-20903, the RoIs created in muon views should have 'MU' in the name instead of 'EM' to avoid confusion.

Merging the offline and trigger configuration for the IndetCandidate and MuonCombined algorithms, and updating the trigger use case to follow the Run 2 configuration (ATR-20389).

Make sure we always configure output collections for the EventViewCreatorWithMuons (or disable initializing write handles for cases they aren't needed), and switched t using the RoIs written out by the EV creator.

Savanna Shaw authored 5 years ago and Frank Winklmeier committed 5 years ago

Fixing a problem with the previous implementation of the late muon triggers that leads to a crash due to trying to record multiple containers in a single view (ATR-20407).
- Split the out of time RoI making into it's own sequence that runs before the reco sequence.
- Add a simple hypo alg/tool to select the out of time RoIs (no real selection added at the moment, but it could be added in the future)
- Update the event view creator algorithm to be able to add the out of time RoIs to the views