def main():
description = "Produce visualisations and find optimal parameters of compression algorithms"
parser = argparse.ArgumentParser(description=description)
parser.add_argument('--verbose', dest='verbose', action='store_true',
help='produce detailed output showing work performed.')
parser.add_argument('--paranoia', dest='paranoia', action='store_true',
help='verify correct operation of compression algorithms by decompressing ' +
'their output and comparing to the original file.')
parser.add_argument('--rerun', dest='rerun', action='store_true',
help='regenerate the data, even if there is a cached result.')
parser.add_argument('--include', dest='include', nargs='+',
help='paths which match the specified regex are included; ' +
'if unspecified, defaults to *.')
parser.add_argument('--exclude', dest='exclude', nargs='+',
help='paths which match the specified regex are excluded.')
parser.add_argument('--num-workers', dest='num_workers', default=config.NUM_WORKERS,
help='number of local processes (default: {0})'.format(config.NUM_WORKERS))
parser.add_argument('--style', dest='style')
parser.add_argument('tests', nargs='*',
help='list of tests to conduct; format is test_name[:parameter1=value1[:...]]')
args = vars(parser.parse_args())
global verbose, paranoia, use_cache
verbose = args['verbose']
paranoia = args['paranoia']
use_cache = not args['rerun']
num_workers = int(args['num_workers'])
if args['style']:
plot.set_style(args['style'])
files = general.include_exclude_files(args['include'], args['exclude'])
if verbose:
print("Operating on: {0}".format(files))
pool = multiprocessing.Pool(num_workers)
if verbose:
print("Splitting work across {0} processes".format(num_workers))
if not args['tests']:
print("WARNING: no tests specified", file=sys.stderr)
for test in args['tests']:
if not test: # empty string
print("ERROR: test name cannot be an empty string", file=sys.stderr)
continue
test_name, *test_args = test.split(":")
test_kwargs = to_kwargs(test_args)
test_id = canonical_name(test_name, test_kwargs)
if test_name in TESTS:
test_runner = TESTS[test_name]
if verbose:
print("Running " + test_id)
test_runner(pool, files, test_id, **test_kwargs)
else:
print("ERROR: unrecognised test '" + test_name + "'")
pool.close()
pool.join()
print("All tests finished.")
def generate_resource_figure(test, settings, data):
file = settings['file']
algos = settings['algos']
resources, format = process_resource_data(settings, data)
resources = resources[file]
# flatten data
flattened = [confidence_interval(resources[algo], config.RESOURCE_ALPHA) for algo in algos]
x = np.arange(len(algos))
y = [z[0] for z in flattened]
yerr = [z[1] for z in flattened]
xticks = list(map(config.ALGO_ABBREVIATIONS.get, algos))
colors = ppl.brewer2mpl.get_map('Set2', 'qualitative', len(algos)).mpl_colors
if settings['style']:
plot.set_style(settings['style'])
plot.new_figure()
fig, ax = plt.subplots()
rects = ppl.bar(x, y, xticklabels=xticks, yerr=yerr, log=True, grid='y', color=colors)
# Annotate
for rect in rects:
bar_x = rect.get_x() + rect.get_width()/2.
bar_y = rect.get_height()
label = format(bar_y)
plt.annotate(label, xy=(bar_x, bar_y), xytext=(0, 10), textcoords='offset points',
horizontalalignment='center', verticalalignment='bottom')
plt.xlabel('Compressor')
if settings['col'] == 'runtime':
plt.ylabel(r'Runtime (\si{\second})')
elif settings['col'] == 'memory':
plt.ylabel(r'Memory (\si{\byte})')
ax.set_yscale('log', basey=2) # units are in powers of two, so scale should be as well
# hack to make labels fit
ymin, ymax = plt.ylim()
plt.ylim((ymin, ymax*2))
plot.save_figure(fig, 'resources', [test])
