Command-Line Arguments

Command-Line Arguments#

Flexible, Reusable, and Automatable Simulations Optional

OpenSees (both Tcl and Python) supports passing command-line arguments, allowing you to control your simulation behavior at runtime without editing the script itself. This is essential for parameter studies, calibration runs, job arrays, and high-throughput automation.

Why Use Command-Line Arguments?#

  • Avoid duplicating scripts for each input variation

  • Customize simulations dynamically at runtime

  • Enable parameter sweeps, job arrays, and automation

  • Keep job inputs visible in logs and reproducible

  • Simplify integration with Tapis, Slurm, or bash scripts

Use Cases#

Use Case

Benefit

Parameter studies

Vary geometry, loads, or materials across runs

Job arrays on Slurm or Tapis

Pass different arguments to identical jobs

Model calibration or UQ

Use command-line control for thousands of simulation runs

Organizing output

Pass output folder names from the command line

How to Pass Arguments#

You append arguments after the script name:

Tcl:

OpenSees model.tcl 2 output_folder/

Python (OpenSeesPy):

python model.py 2 output_folder/

Parallel Execution:

mpiexec -np N OpenSeesMP modelMP.tcl 2 output_folder/
mpiexec -np N python model.py 2 output_folder/

How to Read Arguments Inside the Script#

Tcl (OpenSees, OpenSeesMP, OpenSeesSP)

Tcl automatically provides:

  • argv — List of command-line arguments

  • argc — Number of arguments

Example – model.tcl

set argc [llength $argv]
set modelType [lindex $argv 0]
set outputDir [lindex $argv 1]

puts "Running model type: $modelType"
puts "Saving results to: $outputDir"
file mkdir $outputDir
Python (OpenSeesPy)

Use the sys module to access:

  • sys.argv[0] → Script name

  • sys.argv[1:] → Arguments

Example – model.py

import sys
import os
import openseespy.opensees as ops

model_type = int(sys.argv[1])
output_dir = sys.argv[2]

print(f"Running model type: {model_type}")
print(f"Saving results to: {output_dir}")
os.makedirs(output_dir, exist_ok=True)

if model_type == 1:
    # Define Model 1
    pass
elif model_type == 2:
    # Define Model 2
    pass
Using argparse in Python (Recommended)

For more robust input handling, use the argparse module:

Run:

python model.py --model 2 --output output_folder/

Script:

import argparse
import os
import openseespy.opensees as ops

parser = argparse.ArgumentParser()
parser.add_argument("--model", type=int, required=True)
parser.add_argument("--output", type=str, default="output")
args = parser.parse_args()

print(f"Model: {args.model}, Output Dir: {args.output}")
os.makedirs(args.output, exist_ok=True)

if args.model == 1:
    # Define Model 1
    pass
elif args.model == 2:
    # Define Model 2
    pass

Automating Parameter Sweeps#

You can control OpenSees (Tcl or python) runs from another Python script (or Jupyter Notebook cell):

Python controller script:

import os

model_types = [1, 2, 3]
for m in model_types:
    out_dir = f"results/model_{m}"
    os.system(f"OpenSees model.tcl {m} {out_dir}")
    os.system(f"python model.py --model {m} --output {out_dir}")

Tip: Use This in Jupyter!

You can run the controller above inside a Jupyter notebook cell to launch multiple simulations interactively.

Great for:

  • Parameter studies

  • Monte Carlo simulations

  • Generating data for plots

Summary Table#

Language

Access Pattern

Example

Tcl

$argv, [lindex $argv N]

[lindex $argv 0]

Python

**sys.argv[N]` or argparse

sys.argv[1], args.output
Contents