Deletion

deleteAllSimulations(; kwargs...)

Delete all simulations. See deleteSimulations for keyword arguments.

deleteMonad(monad_ids; delete_subs, delete_supers)

Delete monads by ID, optionally cascading to their simulations and to empty samplings/trials above them.

deleteSampling(sampling_ids; delete_subs, delete_supers)

Delete samplings by ID, optionally cascading.

deleteSimulation(args...; kwargs...)

Alias for deleteSimulations.

deleteSimulations(simulation_ids; delete_supers, filters)
deleteSimulation(args...; kwargs...)

Delete simulations from the database and from disk.

Also removes each deleted simulation's row (if any) from the post-processing sink (data/outputs/postprocessing.db), keeping it consistent with the central database.

If delete_supers is true (default), also removes any monads/samplings/trials that become empty after the deletion. filters adds extra SQL WHERE conditions.

Examples

deleteSimulations(1:3)
deleteSimulations(4)
deleteSimulations(1:100; filters=Dict("config_id" => 1))

deleteSimulationsByStatus(status_codes_to_delete::Vector{String}=["Failed"]; user_check::Bool=true)
deleteSimulationsByStatus(status_code_to_delete::String; user_check::Bool=true)

Delete simulations filtered by status code.

deleteTrial(trial_ids; delete_subs)

Delete trials by ID, optionally cascading to their samplings.

resetDatabase(; force_reset, force_continue)

Reset the database after user confirmation: delete all output folders, remove the post-processing sink (data/outputs/postprocessing.db), clear all variation files, call clearSimulatorArtifacts on the active simulator, then reinitialize the database.

On a shared filesystem some of what this removes may be staged in data/.trash/ rather than deleted — see rm_hpc_safe. That space is not reclaimed until the files are released and a later session retries them.

resetFolder(location::Symbol, folder::String)

Remove the variations DB and variations folder from folder in location.

rm_hpc_safe(path::String; force, recursive)

Remove path, tolerating shared cluster filesystems that refuse to release files.

Off HPC — anywhere sbatch is absent, unless useHPC says otherwise — this is exactly rm(path; force, recursive), exceptions included.

On HPC that same rm is attempted first, since removing is the only thing that actually frees disk space. A network filesystem routinely refuses to remove a directory that a process on another node still holds open, and rm throws. Whatever survives is then moved into data/.trash/data-YYMMDD/, mirroring its position under data/, which succeeds where removal does not because a rename never releases the file. A path outside data/ is staged under _external/.

Returns:

• :removed — rm succeeded.
• :staged — the residue was moved into data/.trash/. Nothing staged there has had its space reclaimed, and the first time it happens in a project a warning says so and names the directory. initializeModelManager will retry the removal in the background at the start of every later session, so staged paths clear themselves once the jobs holding them exit.
• :unremoved — neither worked. path is left where it is and a warning names it.

In HPC mode a filesystem failure does not throw: every caller deletes the corresponding database rows before calling this, and most call it inside a loop over many simulations, so an exception here would abandon a bulk deletion with the rows already gone. Check the return value where a removal must be guaranteed — resetDatabase does, for the central database file. A missing path with force=false still throws, exactly as rm does.

Examples

rm_hpc_safe(joinpath(dataDir(), "outputs", "simulations", "7"); force=true, recursive=true)

if rm_hpc_safe(path_to_file; force=true) === :staged
    @info "still on disk under $(joinpath(dataDir(), ".trash"))"
end