1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
#!/usr/bin/python3
 
''' SI 335 Spring 2013
    Project 1: OMA Service Selection
 
    This contains the "simple" algorithm to do the service selection,
    provided as example/starter code.
'''
 
from sys import stdin
 
# Read in the number of communities
k = int(stdin.readline())
 
# Read in community names
communities = []
for i in range(k):
    communities.append(stdin.readline().strip())
 
# Read in number of proles
n = int(stdin.readline())
 
# Read in proles' names and rankings
# Rankings will be stored in a list of lists.
# Each list in the list will contain all k rankings for that prole.
proleNames = []
proleRanks = []
 
for i in range(n):
    lineparts = stdin.readline().strip().split()
    proleNames.append(lineparts[0])
    ranks = []
    for j in range(1, k+1):
        ranks.append(int(lineparts[j]))
    proleRanks.append(ranks)
 
# This list will hold the names that are already picked
picked = []
 
# Now do the actual selection.
for i in range(n):
    comm = i % k # which community gets the next pick
    rank = 0 # start with the top-ranked pick
    found = False # set to true once we find the next pick
 
    while not found:
        rank = rank + 1
 
        for j in range(n):
            if proleRanks[j][comm] == rank:
                nextPick = proleNames[j]
 
        # Now nextPick is the name of the prole with rank "rank"
        # by community "comm". But are they already picked?
 
        found = True
        for j in range(len(picked)):
            if picked[j] == nextPick:
                found = False
 
 
    # Now we have the actual pick!
    # Print it and then add to the "picked" list.
    print(nextPick, communities[comm])
    picked.append(nextPick)