Move to updated PID model, add track dt/dt fit info

[michaelmackenzie]

This follows PRs Mu2e/MLTrain#9 and Mu2e/ArtAnalysis#8 to update the input PID model and to also add an optional branch with the track dt/dt fit results. I added rooutil information as well and a test script plotting the results.

[AndrewEdmonds11]

Looks great! Thanks, Michael. Could you also add the new branch to the ntuplehelper and recreate branches.md? Thanks

[michaelmackenzie]

Looks great! Thanks, Michael. Could you also add the new branch to the ntuplehelper and recreate branches.md? Thanks

I just added this, thanks for reminding me!

[oksuzian]

Issues in PR #386:

1. Branch misalignment risk (src/EventNtupleMaker_module.cc) — when the DtDt handle is invalid or shorter than the KalSeedPtr collection, nothing is pushed, so trkdtdt ends up shorter/misaligned with trk. RooUtil assumes 1:1. Fix: default-fill a TrkDtDtInfo for missing entries.

2. handle->at(i_kseedptr) can throw — if the DtDt collection is valid but smaller than the seed collection.

3. branches.md regression — regenerated from an older base; drops previously-merged Calorimeter/CRV/Trigger docs (field tables, calohitsmc alignment warning, trigger array explanation).

4. Empty CRV explanation columns — crvpulses, crvdigis, crvpulsesmc rows now have blank explanations in branches.md.

5. Wrong leaf link — crvpulsesmc now points to CrvPulseInfoReco.hh (reco) instead of an MC struct.

6. Missing trkDtDtTag on De fit config (fcl/prolog.fcl) — with fillTrkDtDt set globally true, fit branches lacking a trkDtDtTag get an empty tag → invalid handle → misaligned/empty branch.

[michaelmackenzie]

Responding to the AI comments:

1. I fixed this
2. I now actively throw on this
3. I merged in main and regenerated the documentation
4. I don't know why this is missing
   @AndrewEdmonds11 I used ntuplehelper --list-all-branches --export-to-md >| doc/branches.md but it doesn't list these descriptions? Looking at these files it seems like they're missing this comment line, maybe it was done by-hand before?
5. I added the missing tag

I think this should be merge ready, where another PR should fix the CRV object files, but I can also do this if preferred

[oksuzian]

I'll review PR #386 to find issues. Let me gather the relevant data.Here's my review of PR #386 – Move to updated PID model, add track dt/dt fit info.

Summary

This PR updates the PID model and adds a new optional trkdtdt branch storing the track dt_hit/dt_trk fit results (slope, offset, uncertainty, chi², dof). It's an approved, low-risk PR (12 files, +225/−104) that follows upstream PRs in MLTrain and ArtAnalysis.

Reviews: 1 approval from @AndrewEdmonds11 with a request: "add the new branch to ntuplehelper and recreate branches.md." Both appear addressed in the current diff (helper/ntuplehelper.py and doc/branches.md are updated).

Issue found: silently mismatched dt/dt info when the collection is missing

This is the one substantive bug worth flagging before merge. In EventNtupleMaker_module.cc:

if(_conf.trk().fillTrkDtDt()) {
  auto handle = _allKSDtDtHs.at(i_trk_fit_branch);
  if(handle.isValid()) {
    if(i_kseedptr >= handle->size()) throw cet::exception("EventNtuple")
                                       << "Index " << i_kseedptr
                                       << " out of range of DtDt list size = "
                                       << handle->size();
    const auto& dtdt = handle->at(i_kseedptr);
    _infoStructHelper.fillTrkDtDtInfo(dtdt, _allTDtDtIs.at(i_trk_fit_branch));
  } else {
    const KalSeedDtDt dtdt; // empty dt/dt object
    _infoStructHelper.fillTrkDtDtInfo(dtdt, _allTDtDtIs.at(i_trk_fit_branch));
  }
}

The concern is the handle.isValid() == false branch. When fillTrkDtDt is true but the configured trkDtDtTag collection is absent (or mis-tagged), the code silently fills a default-constructed KalSeedDtDt (all zeros) for every track, rather than throwing. Contrast this with the valid-handle path, which does throw on any inconsistency (i_kseedptr >= handle->size()).

The result: a misconfigured tag produces an ntuple full of zero-valued trkdtdt entries that look like legitimate fit results (slope=0, offset=0, chisq=0, dof=0) with no indication anything went wrong. Since the user explicitly opted in via fillTrkDtDt: true, a missing collection is more likely a configuration error than an intended empty state.

Suggestions (pick one):

• Throw (like the out-of-range case) when the handle is invalid but fillTrkDtDt is true, so misconfiguration fails loudly; or
• Keep the graceful fallback but make the "no data" state distinguishable — e.g. initialize dof = -1 in TrkDtDtInfo so consumers can tell an unfilled entry from a real fit with dof=0. The rooutil example only null-checks track.trkdtdt (the pointer), not the fit validity, so a zeroed struct would be plotted as a real slope=0 entry.

Other minor observations

• branches.md collateral edits: This PR reverts several previously-detailed sections (Calorimeter, CRV, Trigger docs) back to terser versions — e.g. crvpulses/crvpulsesmc descriptions become blank, and the detailed calohitsmc alignment warning is removed. If those richer docs landed in a separate recent PR, this may be an unintended merge/rebase regression rather than an intended doc change. Worth double-checking the base is current.
• Whitespace noise: rooutil/inc/Track.hh is mostly trailing-whitespace cleanup (17/32 lines), which is fine but inflates the diff.