yall/fullset-f2-repro/README.md

0001 # FullSet F2 reproduction
0002
0003 This example reproduces Fredi's SPS H2 `FullSetF_2` calibration chain from raw
0004 TB2026 data using Yall and HTCondor.
0005
0006 It is a larger validation workflow, not the recommended first Yall/LFHCal test.
0007 For a fresh BNL setup, complete the checks in [`../SETUP.md`](../SETUP.md)
0008 first.
0009
0010 For the numerical comparison with the published calibration, see
0011 [`CALIBRATION_COMPARISON.md`](CALIBRATION_COMPARISON.md).
0012
0013 ## Inputs
0014
0015 The workflow uses pedestal run `471` and merges muon runs
0016 `472`, `475`, `476`, `479`, `480`, and `483` as `Muon_FullSetF_2`.
0017
0018 The raw inputs must be readable as:
0019
0020 ```text
0021 $LFHCAL_DATA/Run471.h2g
0022 $LFHCAL_DATA/Run472.h2g
0023 $LFHCAL_DATA/Run475.h2g
0024 $LFHCAL_DATA/Run476.h2g
0025 $LFHCAL_DATA/Run479.h2g
0026 $LFHCAL_DATA/Run480.h2g
0027 $LFHCAL_DATA/Run483.h2g
0028 ```
0029
0030 The Yallfile also uses the TB2026 mapping and run database, the FullSetA-F bad
0031 channel map, and the published FullSetF ToA offsets. The ToA offsets are not
0032 rederived by this workflow.
0033
0034 ## Run table and dependencies
0035
0036 The Yallfile declares the seven runs once in a typed
0037 `@table runs type run:` table. One `convert-{type}-{run}` family converts every
0038 row. `merge-muon` binds `type=muon`, restricting both its patterned parent
0039 fan-in and `@input.parts` to runs `472`, `475`, `476`, `479`, `480`, and `483`;
0040 pedestal run `471` cannot leak into the merge.
0041
0042 The `{type}-{run}` pedestal family uses `@each type pedestal`. That explicitly
0043 binds only `type`; the compatible `convert-{type}-{run}` parent supplies the
0044 remaining `run=471`, producing `pedestal-471`. `transfer-f2` depends on the
0045 patterned pedestal family, so changing the pedestal row makes the conversion,
0046 pedestal task, and transfer dependency follow it automatically.
0047
0048 This partial-binding syntax requires yall-run PR #33, merged as commit
0049 `1081e9dd39418262588248272618130ce0503b8a`.
0050
0051 ## Run at BNL
0052
0053 Use the normal BNL `tcsh` login/submit session. Yall and Condor run on the host;
0054 the scientific payloads are sent through the configured EIC shell wrapper.
0055
0056 ```tcsh
0057 cd "$LFHCAL_REPO/examples/yall/fullset-f2-repro"
0058 source env.tcsh
0059 ```
0060
0061 Check the raw inputs before submitting:
0062
0063 ```tcsh
0064 foreach r (471 472 475 476 479 480 483)
0065     ls -lh "$LFHCAL_DATA/Run${r}.h2g"
0066 end
0067 ```
0068
0069 Validate and inspect the graph:
0070
0071 ```tcsh
0072 yall-run validate
0073 yall-run plan
0074 ```
0075
0076 Create and submit a campaign:
0077
0078 ```tcsh
0079 set C = `yall-run create --campaigns-dir "$LFHCAL_WORK/campaigns"`
0080 echo "$C"
0081 yall-run start "$C"
0082 yall-run status "$C"
0083 ```
0084
0085 Repeat `yall-run status "$C"` to inspect progress.
0086
0087 Do not start another campaign against an existing output tree. Use a fresh
0088 `LFHCAL_WORK` for a clean rerun.
0089
0090 ## Outputs
0091
0092 Results are written under:
0093
0094 ```text
0095 $LFHCAL_WORK/fullset-f2-repro/
0096 ```
0097
0098 The workflow converts the seven raw runs, constructs the pedestal calibration,
0099 merges and calibrates the muon sample, selects MIP-triggered events, and performs
0100 five reduced/no-event-tree refinement passes through `Imp5R`.
0101
0102 The final products are:
0103
0104 ```text
0105 $LFHCAL_WORK/fullset-f2-repro/final/calib_Final_Muon_FullSetF_2.root
0106 $LFHCAL_WORK/fullset-f2-repro/final/calib_Final_Muon_FullSetF_2_calib.txt
0107 ```
0108
0109 `final-f2` does not perform another fit; it copies the `Imp5R` products into the
0110 final output directory.
0111
0112 Campaign records and scheduler logs are kept under `$LFHCAL_WORK/campaigns` when
0113 the commands above are used.