mirror of
https://github.com/QuantumPackage/qp2.git
synced 2024-11-03 12:43:48 +01:00
73 lines
1.8 KiB
Python
Executable File
73 lines
1.8 KiB
Python
Executable File
#!/usr/bin/env python2
|
|
# -*- coding: utf-8 -*-
|
|
|
|
import urllib
|
|
import sys
|
|
from bs4 import BeautifulSoup
|
|
|
|
address="http://gernot-katzers-spice-pages.com/character_tables/%s.html?fmt=simple"
|
|
|
|
def clean_up(text):
|
|
soup = BeautifulSoup(text, "lxml")
|
|
pre = soup.pre
|
|
group = pre.b.get_text()
|
|
sop = {}
|
|
irred = {}
|
|
irred_count = 0
|
|
sop_count = 0
|
|
for span in pre.find_all('span'):
|
|
cls = span.get('class')
|
|
if cls == ['sop']:
|
|
a = span.decode_contents()
|
|
if a not in sop:
|
|
sop[a] = sop_count
|
|
sop_count += 1
|
|
elif cls == ['irred']:
|
|
a = span.decode_contents()
|
|
if a not in irred:
|
|
irred[a] = irred_count
|
|
irred_count += 1
|
|
table = [ [] for j in sop ]
|
|
data = pre.get_text().splitlines()
|
|
def f(x):
|
|
y = x.split()
|
|
if len(y) == 0:
|
|
return False
|
|
else:
|
|
return y[0] in irred
|
|
data = filter(f,data)[:len(irred)]
|
|
for line in data:
|
|
s = line.replace('*','').split()
|
|
l = irred[s[0]]
|
|
data[l] = map(float,s[1:len(irred)+1])
|
|
|
|
d = {}
|
|
e = {}
|
|
for k in irred:
|
|
d[irred[k]] = k
|
|
for k in sop:
|
|
e[sop[k]] = k
|
|
n = len(irred)
|
|
print "Group\t", group, "\nn\t", n
|
|
print "\n \tIrred \tOperation"
|
|
for i in range(n):
|
|
print "%4d \t %s \t %s"%(i+1, d[i].ljust(10), e[i].ljust(10))
|
|
|
|
print "\nTable\n ",
|
|
for j in range(n):
|
|
print "%8s "%(str(j+1).center(8)),
|
|
for i in range(n):
|
|
print "\n%4d "%(i+1),
|
|
for j in range(n):
|
|
print "%8.5f "%(data[i][j]),
|
|
print "\n"
|
|
|
|
def main():
|
|
for group in sys.argv[1:]:
|
|
f = urllib.urlopen(address%(group))
|
|
html = f.read().split('\n',1)[1]
|
|
text = clean_up(html)
|
|
|
|
if __name__ == "__main__":
|
|
main()
|