netgen/tests/pytest/compare_results.py

104 lines
2.9 KiB
Python
Raw Normal View History

2019-10-24 21:22:25 +05:00
import json
import sys
import subprocess
import statistics
def readData(a, files):
amin=[]
amax=[]
amin1=[]
amax1=[]
bad=[]
ne1d=[]
ne2d=[]
ne3d=[]
file=[]
2019-10-24 21:22:25 +05:00
for f in files:
for t in a[f]:
file.append(f)
2019-10-24 21:22:25 +05:00
if t['ne1d']>0:
ne1d.append(t['ne1d'])
if t['ne2d']>0:
ne2d.append(t['ne2d'])
if t['ne3d']>0:
ne3d.append(t['ne3d'])
if t['total_badness']>0.0:
bad.append(t['total_badness'])
if 'angles_tet' in t:
amin.append(t['angles_tet'][0])
amax.append(t['angles_tet'][1])
if 'angles_trig' in t:
amin1.append(t['angles_trig'][0])
amax1.append(t['angles_trig'][1])
return {
"min tet angle":amin,
"max tet angle" : amax,
"min trig angle":amin1,
"max trig angle" : amax1,
"badness" : bad,
"#edges" : ne1d,
"#trigs" : ne2d,
"#tets" : ne3d,
"file" : file,
2019-10-24 21:22:25 +05:00
}
import matplotlib.pyplot as plt
ref = 'master'
if len(sys.argv)>1:
ref = sys.argv[1]
res = subprocess.run(['git','show','{}:./results.json'.format(ref)], capture_output=True)
s = json.loads(res.stdout.decode())
if len(sys.argv) > 2:
ref2 = sys.argv[2]
res = subprocess.run(['git','show','{}:./results.json'.format(ref2)], capture_output=True)
s2 = res.stdout.decode()
else:
ref2 = 'current'
s2 = open('results.json','r').read()
s2 = json.loads(s2)
filenames = [f for f in s if f in s2]
data = readData(s, filenames)
data2 = readData(s2, filenames)
assert(len(data) == len(data2))
for bad1,bad2, f1, f2 in zip(data['badness'], data2['badness'], data['file'], data2['file']):
assert f1==f2
if bad2>0 and bad2>1.1*bad1:
print(f"file {f1} got worse: {bad1} -> {bad2}")
if bad2>0 and bad2<0.9*bad1:
print(f"file {f1} got better: {bad1} -> {bad2}")
2019-10-24 21:22:25 +05:00
n = len(data)+1
fig,ax = plt.subplots(figsize=(10,7))
for i,d in enumerate(['min trig angle','min tet angle','max trig angle','max tet angle']):
ax = plt.subplot(2,5,i+1)
plt.title(d)
ax.set_xticks([1,2])
if len(data[d])==0 or len(data2[d])==0:
continue
plt.violinplot([data[d],data2[d]], showmedians=True)
med = statistics.median(data[d])
plt.hlines(med, 1,2, linestyle='dotted')
if d=='badness':
ax.set_yscale('log')
ax.set_xticklabels([ref, ref2])
for i,d in enumerate(['badness','#edges','#trigs','#tets']):
ax = plt.subplot(2,5,6+i)
plt.title('difference '+d+' (in %)')
# plt.violinplot([(y-x)/x for x,y in zip(data[d],data2[d])], showmedians=True)
plt.boxplot([100.0*(y-x)/x for x,y in zip(data[d],data2[d])])
2019-10-24 21:22:25 +05:00
plt.hlines(0.0, 0.5,1.5, linestyle='dotted')
# plt.savefig('comparison.png', dpi=100)
plt.show()