Alternative Analysis Approaches

The main episodes read EDM4eic data with uproot through an MCP server. Here is the same Λ⁰ → p π⁻ analysis written four ways — each reproduces the peak near 1.1157 GeV. Every example reads a dataset root:// file directly, selects protons (PDG 2212) and π⁻ (PDG -211), builds proton–π⁻ pairs, and histograms the invariant mass. Full scripts: extras/. Replace root://epicxrd1.sdcc.bnl.gov:1095//... with the file you located.

Callout

All four agree

Same masses (\(M_p = 0.9382720813\), \(M_\pi = 0.13957061\) GeV), same 200 bins over 1.05–1.25 GeV, same peak. Choose for convenience.

1. Stand-alone uproot (no MCP)

---

Plain Python — no server, no ROOT install. The engine the MCP server runs internally. Environment: a pip venv with uproot awkward numpy. Script: extras/standalone_uproot/lambda_uproot.py

BASH

$ python3 extras/standalone_uproot/lambda_uproot.py root://epicxrd1.sdcc.bnl.gov:1095//...

PYTHON

protons = four_vectors(px[pdg == 2212], ..., M_PROTON)
pions   = four_vectors(px[pdg == -211], ..., M_PION)
m = pair_mass(protons, pions)          # invariant mass of every p x pi- pair

2. ROOT RDataFrame (PyROOT)

---

Columnar and declarative; a JIT-compiled C++ helper does the per-event pairing, RDataFrame histograms it in one pass. Scales to many files and cores. Environment: ROOT (e.g. eic-shell). Script: extras/rdataframe/lambda_rdf.py

BASH

$ eic-shell -- python3 extras/rdataframe/lambda_rdf.py root://epicxrd1.sdcc.bnl.gov:1095//...

PYTHON

df = ROOT.RDataFrame("events", path)
df = df.Define("m_lambda", "lambda_mass(ReconstructedChargedParticles.PDG, ...)")
h  = df.Histo1D(("h", ";m(p,pi) [GeV];pairs", 200, 1.05, 1.25), "m_lambda")

3. ROOT TTreeReader (C++ macro)

---

The classic ROOT event loop, the form most ePIC analysis code uses. Reconstructed momentum branches are float, so the TTreeReaderArray must be templated on <float>. Environment: ROOT (e.g. eic-shell). Script: extras/ttreereader/lambda_ttreereader.C

BASH

$ eic-shell -- root -l -b -q 'extras/ttreereader/lambda_ttreereader.C("root://epicxrd1.sdcc.bnl.gov:1095//...")'

4. Native PODIO Frame API

---

PODIO deserializes each event into a Frame; you iterate over typed objects (p.getPDG(), p.getMomentum()). Environment: Key4hep / eic-shell with the podio and edm4eic dictionaries — the heaviest to set up. Script: extras/podio_frame/lambda_podio.py

BASH

$ eic-shell -- python3 extras/podio_frame/lambda_podio.py root://epicxrd1.sdcc.bnl.gov:1095//...

Callout

Same physics, lower barrier

PODIO needs the full Key4hep stack; the uproot path needs only what is already in eic-shell. Both find the same Λ⁰ — pairing an AI assistant with an uproot MCP server is a cheap on-ramp.