Add non-interactive and real-time visualization generation support to…

[aiqubits]

… throughput benchmark script in Walrus.

The usage is as follows:
WALRUS_FSYNC=no-fsync WALRUS_DURATION=2m cargo test --release --test multithreaded_benchmark_writes python scripts/visualize_throughput_noninteractive.py --file benchmark_throughput.csv python scripts/visualize_throughput_noninteractive.py --file benchmark_throughput.csv --single-shot --output test_non_interactive.png

[nubskr]

Thanks for the PR! Headless support is useful for CI/remote servers, but I'd prefer not to maintain two nearly-identical scripts.

Could you modify the existing visualize_throughput.py and visualize_batch_benchmark.py instead? Something like:

parser.add_argument('--headless', action='store_true', help='Save to file instead of displaying (for environments without a display)')
parser.add_argument('--output', '-o', help='Output file path (implies --headless)')

args = parser.parse_args()

if args.headless or args.output:
    import matplotlib
    matplotlib.use('Agg')

import matplotlib.pyplot as plt
# ... existing code ...

# At the end:
if args.output:
    plt.savefig(args.output, dpi=150, bbox_inches='tight')
else:
    plt.show()

This way we get headless support without duplicating all the plotting logic. The --single-shot and --max-frames options could be added to the existing realtime mode as well.

[aiqubits]

Thanks for the PR! Headless support is useful for CI/remote servers, but I'd prefer not to maintain two nearly-identical scripts.

Could you modify the existing visualize_throughput.py and visualize_batch_benchmark.py instead? Something like:

parser.add_argument('--headless', action='store_true', help='Save to file instead of displaying (for environments without a display)')
parser.add_argument('--output', '-o', help='Output file path (implies --headless)')

args = parser.parse_args()

if args.headless or args.output:
    import matplotlib
    matplotlib.use('Agg')

import matplotlib.pyplot as plt
# ... existing code ...

# At the end:
if args.output:
    plt.savefig(args.output, dpi=150, bbox_inches='tight')
else:
    plt.show()

This way we get headless support without duplicating all the plotting logic. The --single-shot and --max-frames options could be added to the existing realtime mode as well.

After receiving feedback, I have newly submitted a PR to modify and merge the headless visualization into the original file.

[aiqubits]

Deprecated headless visualization script has been removed.