A stable content hash (rlang::hash, xxHash) of the model with volatile
bookkeeping stripped: every @misc list (paths, on-disk markers,
timestamps) and the inMemory flag are removed, recursively, before
hashing. Two models built identically hash identically; renaming a folder,
saving, or loading does not change the hash — changing any modeled content
does.
Note: a model whose data slots were thinned to disk (e.g. inside an
on-disk scenario) hashes differently from its in-memory original, because
the data is genuinely absent from the object. Hash models before/without
thinning — save_model() does.
The generalization of model_hash() to any S4 object of the package
(technology, storage, trade, scenario, calendar, horizon, weather, ...):
rlang::hash of the object with volatile bookkeeping stripped, so two
identically-built objects hash identically regardless of @misc contents,
paths, or the inMemory flag. Models and repositories delegate to their
dedicated hashes. Used as the object-identity half of the report and
levcost cache keys.
The thinning caveat of model_hash() applies: an object whose data slots
live on disk hashes differently from its in-memory original.
Arguments
- object
a model object.
- x
an S4 object (model and repository delegate to
model_hash()/repository_hash()).