[nvpm-lab]

Adding an optional confluence_selection flag to the vertex post-processor. Off by default, so nothing changes unless it's turned on.

When get_confluence_points finds more than one confluence group, get_vertex currently goes to confluence_dir and then usually ends up in track_length (start of the longest track). With confluence_selection: pointing, each candidate is instead scored by how well the particles point away from it, and the best one is taken. A particle starting within touching_threshold of the candidate counts as 1, otherwise (1 + cos θ)/2, where θ is the angle between its direction and the candidate --> start vector. No beam direction is used, so it should also work for atmospherics.

Usage -->

post:
  vertex:
    confluence_selection: pointing

Motivation

On the DUNE FD-HD LBNF v1 sample, about 38% of CC events have more than one confluence group. These mostly end up in track_length, which accounts for over half of the vertices that are >10 cm off.

Full inference on the same 10k events with/without the flag, CC (numu + nue), overlap > 0.5 (almost 5.3k interactions):

• median: 1.30 --> 1.09 cm
• 68%: 2.4 --> 1.7 cm
• 90%: 11.6 --> 5.8 cm
• fraction > 10 cm: 10.9% --> 7.4%

About 1.6k interactions change vertex, roughly 1.3k of them for the better.

Type of change

• New feature (non-breaking)

Testing

New unit tests for the selection and for the processor passing the flag through. Existing vertex tests still pass. Plus the two inference runs above.

Notes

• Confluence points are defined basically on track–track only.
• Not tried on atmospherics yet.

[francois-drielsma]

Looks good. Added regression coverage for shower-driven selection and unchanged default/fallback paths, plus formatting fixes. All 17 vertex tests and pre-commit checks pass.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.99%. Comparing base (d4de053) to head (a725567).
⚠️ Report is 32 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #182      +/-   ##
==========================================
- Coverage   99.99%   99.99%   -0.01%     
==========================================
  Files         435      440       +5     
  Lines       37678    39316    +1638     
==========================================
+ Hits        37677    39314    +1637     
- Misses          1        2       +1     

Flag	Coverage Δ
unittests	99.99% <100.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/spine/physics/vertex.py	100.00% <100.00%> (ø)
src/spine/post/reco/vertex.py	100.00% <100.00%> (ø)

... and 78 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.