SKNanoAnalyzer

---

THIS IS DEVELOPMENT VERSION!!

Introduction

• Start with the documentation index.
• For installation and analysis submission, see Getting Started.
• Before contributing code, read the Development Guide.

Analyzer development and ownership

SKNano keeps the shared event-processing backend separate from runnable physics analyses. Before adding or moving an analyzer, read the full Analyzer development and ownership guide.

• AnalyzerFramework owns AnalyzerCore, loading, branch management, standard NanoAOD views, and output services. Do not add concrete physics analyzers to this directory.
• CommonAnalyzers is only for analyzers that are useful to all SKNano users and have no analysis-group-specific policy. Changes require a normal backend review.
• Personal and analysis-group analyzers belong in an external module repository. The module owns its CMake target, ROOT dictionary, tests, runtime data, documentation, and custom input schema. Start from examples/AnalysisModule.
• Official NanoAOD schemas and broadly reusable persistent objects belong in the backend. A custom schema stays with its consuming module until multiple independent modules need it; only then should it be promoted to shared input infrastructure.
• Backend branches pin reviewed external modules as Git submodules. Analyzer names and the usual SKNano.py -a AnalyzerName ... submission interface do not change; the installed analyzer manifest validates availability before a job is created.

The old flat Analyzers/ layout and flat includes such as #include "AnalyzerCore.h" are unsupported. Use namespaced includes such as #include <AnalyzerFramework/AnalyzerCore.h>.

Changelog

[2.0.0] - 2026-07-22

SKNano 2.0.0 is the NanoAODv15 and RNTuple release. This section summarizes the user-visible differences from the main branch rather than listing every internal commit.

Breaking changes

• RNTuple is the only supported input format.
• Analyzer datasets are written as typed RNTuples through Output().Book(), typed fields, and stable analyzer-owned buffers. Histograms remain ordinary ROOT objects.
• NanoAOD collections now use event-scoped *ViewCollection APIs such as GetAllMuonViews() and GetAllJetViews(). The former owning input-object classes and materialization APIs have been removed. Views and selected ranges must not be retained after advancing to the next event.
• The supported Linux environment is the repository's ROOT 6.40.02 toolchain. Create the shared Nano environment from docs/Nano-linux-64.lock, or use docs/Nano.yml as the solver-based fallback. ROOT, correctionlib, Abseil, ONNX Runtime, and the compiler toolchain must be kept ABI-compatible.

Analysis and I/O

• Added schema-driven lazy branch binding for NanoAODv15, including generated scalar/vector accessors and event-scoped physics-object views.
• Added native multi-file RNTuple input and schema-preserving RNTuple skimming.
• Added atomic output publication and scripts/sknano_merge.py, which validates RNTuple and histogram schemas and verifies the merged output before removing input shards.
• Added typed RNTuple output handles, histogram groups, sparse output profiles, and collision/schema checks.
• Added a shared analyzer task API. Flag-selectable studies can register validation, booking, and event callbacks while reusing one baseline event selection.
• Added configurable event failure handling through --failure-policy and --max-event-errors.
• Added deterministic counter-based random-number handling, batched correction and ONNX helpers, execution planning, cache tuning, and optional performance telemetry.

Submission and user tools

• SKNano.py now accepts data-period filtering with -p, sample exclusion with --exclude, and explicit failure policies.
• Every submission records a source snapshot, compressed source archive, sample-metadata snapshot, and run_manifest.json in the run directory.
• User flags are included in skim and analysis output paths, preventing outputs from different task selections from colliding.
• Telegram submission/completion reporting is optional. setup.sh reports only whether it is configured and no longer prints credentials.
• Merge jobs use scheduler defaults unless resources are explicitly requested, and batch setup activates the package environment before enabling strict shell checks.
• Added the schema-driven sknano_plotter workflow with inspect, validate, local plot, and Condor submission modes.
• scripts/build.sh supports incremental builds by default plus --clean, Ninja, build-type, and AddressSanitizer options.

Physics content

• Added 2024 NanoAODv15 sample metadata, trigger paths, golden JSON, muon corrections, tagging inputs, and modelling/reweighting resources.
• Updated jet corrections and JES/JER-to-MET propagation, lepton IDs, trigger object access, systematic handling, and Vcb calibration workflows.

Migration quick start

source setup.sh
./scripts/build.sh --clean

# RNTuple input
SKNano.py -a AnalyzerName -i 'Sample*' -e 2024 -n 10

See docs/README.md for the maintained API and workflow documentation.

To do

Assigned

• FatJet and Tau: Youngwan
• SKNano.py skimming test Yeonjoon
• Auto-Validation: Check the consistency of the output for a specific NanoAOD file: Jin

For DY, check the before/after of correction

• Leptons Jihoon
• Jets(B-tagging) Taehee

Make Tutorial

• From make Training tree, Export ONNX, Use this model in the Analyzer : Yeonjoon(Review: Youngwan)

• Simple DY tutorial : Eunsu(Review: Jin)

• tt reconstruction using Kinematic Fit : Minseok(Review: Byunghun)

• tt reconstruction using KinFit Library : Yeonjoon

• NLO vs LO comparison in MadGraph Tutorial: Jin

• Tau+Fatjet Tutorial : Youngwan

• At least try to install until next meeting : All

Useful links

• MiniAOD
• NanoAOD
• NanoAOD-Tools (for postproc)
• SNU_SKFlatProduction
• NanoAOD wiki@gitlab
• Run2 Summer20 UL
• Run3 Common Bkg Sample Preparation
• SKFlatMaker
• SKFlatAnalyzer
• SKFlat Manual
• NanoAOD Branches
• PdmV Run3 DATA full table
• CMS 13.6TeV SM cross sections
• LHC ttbar NNLO+NNLL cross sections
• LHC Single Top NNLO+NNLL cross sections
• LHC Higgs Cross Section
• Rules for Run3 2024 dataset names
• Rules for Run3 dataset names