The command line

petroscript (or python -m petroscript) has four commands. Each is one of the Python doors, so a command does what the same lines of Python do:

Command	The Python it runs	Prints
petroscript load DECK FOLDER	ps.load(DECK).to_python(FOLDER)	the script’s path, FOLDER/<NAME>.py
petroscript compile SOURCE FOLDER	ps.compile(model).write(FOLDER)	each file’s sha256 and path, as sha256sum does
petroscript check SOURCE	ps.compile(model).check()	the rows, each once
petroscript run [--no-preflight] SOURCE [FOLDER]	ps.compile(model).run(folder=FOLDER); preflight=False with the flag	the run’s folder, last

A SOURCE is a deck, which ps.load reads, or a Python script that binds model to a ps.Model. The script load writes binds it; a script of your own ends with a line such as model = merle.

run runs deck.check() first: what Flow refuses in its pre-flight (a dry run of the deck and one short step) is refused before the run, in Flow’s own words, and the command exits 1; run --no-preflight skips the pre-flight. Then it runs the simulator ps.simulators() lists first (PETROSCRIPT_FLOW, then OPM Flow on PATH) in FOLDER, or in ./<model name>/ when FOLDER is left out. Read the run in Python with ps.results.open(folder).

Exit codes: 0 done; 1 a refusal, printed with its fix line, or an error row from check; 2 a usage error.

$ cd /work
$ petroscript load CASE.DATA case
case/CASE.py
$ petroscript check case/CASE.py
[preflight.passed] checked by flow 2026.04 at /usr/bin/flow: a dry run of the whole deck (NOSIM) and initialisation with one 0.01-day step, in 2.6 s
$ petroscript run case/CASE.py runs
opm-flow: running CASE.DATA
/work/runs