Keep the whole spread-fit ledger out of the data-prep cache key

[eliotmcintire]

Hindcasts and refits redid an hour or two of data preparation per region whenever a new fit was saved anywhere. This stops that, while the later steps that do use the fit still re-run when this region's fit changes.

Affects: No change to results. Data preparation is reused across fits and hindcasts; each region rebuilds once more because the cache key changes once.

Technical details

Defect: fireSense_dataPrepFit.R:244 declared expectsInput("spreadFitPreRun"). SpaDES.core puts every expected input already in the sim into the .inputObjects cache key (simInit, suppliedElsewhere), and fireSense_ELFs puts the whole ledger (rows of all ELFs) there first, so any ELF's new fit missed the cache. .inputObjects and the functions it calls never read spreadFitPreRun (grep: only Init() uses it).

Fix: the expectsInput is gone (createsOutput kept, its description now explains why). The protection it gave stays through .useCacheArgs: dataPrepBuild, prepIgnitionFitData, prepEscapeFitData and prepSpreadFitData each add spreadFitPreRun = sim$spreadFitPreRun to .cacheExtra. SpaDES.core evaluates quoted entries in .runEvent() (R/simulation-spades.R ~1595, eval(envir = environment())), where sim is the live simList; the chaining path (R/cacheChainingJump.R ~240) also evaluates with sim. Init() (line ~411) always assigns sim$spreadFitPreRun (the CacheGeo rows within this study area, or NULL for a held-out fold), and it runs in init, which schedules all later events, so the key is this ELF's rows only, never the whole ledger.

Tests: new test-ledgerNotInInputObjectsKey.R (not an input; .inputObjects does not read it; Init() assigns it unconditionally; each event's .cacheExtra digests the same for the same rows and differently for a new own row or NULL; a ledger differing only in another ELF's rows gives the same readSpreadFitLedger() result for this study area). test-metadata.R updated. Branch: 108 tests, 0 failures, 1 skipped (fixture); base with the new file: 2 of the new tests fail. checkDocFiles/codoc clean; R CMD check (no manual, no tests) has no errors and one WARNING about an import replaced during package conversion, unrelated to this change.

🤖 Generated with Claude Code

https://claude.ai/code/session_019T7wFP7tXPDaRn42BFLzjr