Seal, mark, runs, storage and lineage of a saved object

What state a saved object is in: whether it is sealed, whether it is queued for deletion, what runs it holds, what it costs on disk, and what it shares with other scenarios. Reads the manifests, so the answer does not depend on the registry being up to date.

Usage

get_status(x = NULL, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class '`NULL`'
get_status(x = NULL, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class 'scenario'
get_status(x, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class 'model'
get_status(x, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class 'repository'
get_status(x, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class 'character'
get_status(x, type = NULL, sizes = NULL, lineage = TRUE, ...)

# Default S3 method
get_status(x, type = NULL, sizes = NULL, lineage = TRUE, ...)

# S3 method for class 'en_status'
print(x, ...)

Arguments

x: a saved object, a name, or NULL for every saved object.
type: "model", "repository", "dataset" or "scenario", when x is a name that needs disambiguating. A dataset is always named this way: load_dataset() returns an unclassed payload, so there is nothing to dispatch on.
sizes: measure folder sizes with dir_size(). Defaults to TRUE for one object and FALSE for a whole project, where walking a large store would dominate the call.
lineage: resolve used_by and check for broken references. Needs the store walk; with FALSE, used_by is NA and nothing is checked.
...: unused.

Value

A tibble of class en_status, one row per object.

Details

With an object, a name, or nothing:

get_status(scen), get_status(mod), get_status(repo) — that object.
get_status("REF_sub"), get_status("merra_cf", type = "dataset") — the named entry, resolved in its store. A bare name present in more than one store is refused rather than guessed; name the type.
get_status() — every saved model, repository, dataset and scenario.

A run carries no seal and no deletion mark — those verbs act on the four entry kinds, not on a run — so runs appear as the runs and solved counts of the scenario that holds them.

Lineage

A derived scenario records its origin twice, and the two can disagree. base is the name from base: in the manifest: provenance, written by interpolate_model(base = ). shares counts the tables it actually borrows from another scenario: the coupling, and the only one of the two that blocks unseal_scenario() or deletion.

used_by is the other direction, for every kind of entry and not only scenarios – the scenarios built on a model, the models holding a repository, the scenarios borrowing a scenario's tables. It counts the same edges list_dependents() walks, so a non-zero used_by is what a refusal will cite.

base set with shares == 0 is therefore the ordinary safe case — the base was not sealed when the derivation was built, or every table fell under getOption("energyRt.param_ref_min_rows") and was copied. lineage_warn is for the cases that are not safe: a borrowed table whose source scenario is gone, or a recorded base that no longer exists.

Examples

if (FALSE) { # \dontrun{
get_status()                              # the whole project
get_status(scen)
get_status("merra_cf", type = "dataset")

st <- get_status()
st[st$marked, c("type", "name", "importance")]   # what delete_marked() sees
st[nzchar(st$lineage_warn), ]                    # broken references
} # }