| 1 |
"""
|
| 2 |
PerligoDupeGroup
|
| 3 |
|
| 4 |
This project for ashely was to group documents by "near dupe" and report
|
| 5 |
|
| 6 |
"""
|
| 7 |
|
| 8 |
if __name__ =='__main__':
|
| 9 |
matrix = {}
|
| 10 |
matrix2 = {}
|
| 11 |
count = 1
|
| 12 |
contents = open(r"C:\Users\eborges\Box Sync\Client\Honeywell\Perligo\20150109_dedupeProj\NearDupesExport.dat").readlines()
|
| 13 |
for line in contents:
|
| 14 |
line = line.replace("\n","")
|
| 15 |
line = line.replace("\r","")
|
| 16 |
line = line.split("|")
|
| 17 |
test = None
|
| 18 |
for docID in line:
|
| 19 |
if docID in matrix.keys():
|
| 20 |
test = matrix[docID]
|
| 21 |
if test:
|
| 22 |
pass
|
| 23 |
else:
|
| 24 |
test = count
|
| 25 |
count = count + 1
|
| 26 |
for docID in line:
|
| 27 |
matrix[docID] = test
|
| 28 |
print "initial matrix done. %d count"% len(matrix.keys())
|
| 29 |
for begNo in matrix.keys():
|
| 30 |
group = matrix[begNo]
|
| 31 |
if group in matrix2.keys():
|
| 32 |
if begNo in matrix2[group]:
|
| 33 |
pass
|
| 34 |
else:
|
| 35 |
matrix2[group].append(begNo)
|
| 36 |
else:
|
| 37 |
matrix2[group] = [begNo,]
|
| 38 |
outputFile = open(r"C:\Users\eborges\Box Sync\Client\Honeywell\Perligo\20150109_dedupeProj\NearDupesOutput.dat",'w')
|
| 39 |
for i in matrix2.keys():
|
| 40 |
outputFile.write(str(matrix2[i])+"\n")
|
| 41 |
outputFile.close()
|
| 42 |
print "Second matrix done. %d count"% len(matrix2.keys())
|
| 43 |
contents = open(r"C:\Users\eborges\Box Sync\Client\Honeywell\Perligo\20150109_dedupeProj\originals_list.dat").readlines()
|
| 44 |
|
| 45 |
origList = []
|
| 46 |
for line in contents:
|
| 47 |
line = line.replace("\n","")
|
| 48 |
line = line.replace("\r","")
|
| 49 |
origList.append(line)
|
| 50 |
print "orig matrix done. %d count."% len(origList)
|
| 51 |
|
| 52 |
contents = open(r"C:\Users\eborges\Box Sync\Client\Honeywell\Perligo\20150109_dedupeProj\second_orig_list.dat").readlines()
|
| 53 |
origList2 = []
|
| 54 |
for line in contents:
|
| 55 |
line = line.replace("\n","")
|
| 56 |
line = line.replace("\r","")
|
| 57 |
origList2.append(line)
|
| 58 |
print "orig 2 matrix done. %d count."% len(origList2)
|
| 59 |
|
| 60 |
origList.sort(reverse=True)
|
| 61 |
origList2.sort(reverse=True)
|
| 62 |
|
| 63 |
outputFile = open(r"C:\Users\eborges\Box Sync\Client\Honeywell\Perligo\20150109_dedupeProj\finalReport_near.txt",'w')
|
| 64 |
for k in matrix2.keys():
|
| 65 |
group = matrix2[k]
|
| 66 |
group.sort()
|
| 67 |
parent = None
|
| 68 |
for orig in origList:
|
| 69 |
if orig in group:
|
| 70 |
parent = orig
|
| 71 |
for orig2 in origList2:
|
| 72 |
if orig2 in group:
|
| 73 |
parent = orig2
|
| 74 |
if parent:
|
| 75 |
outputFile.write(parent + "|")
|
| 76 |
for i in group:
|
| 77 |
if i == parent:
|
| 78 |
pass
|
| 79 |
else:
|
| 80 |
outputFile.write(i+",")
|
| 81 |
else:
|
| 82 |
outputFile.write(group[0]+"|")
|
| 83 |
for i in group[1:]:
|
| 84 |
outputFile.write(i+",")
|
| 85 |
outputFile.write('\n')
|
| 86 |
|
| 87 |
outputFile.close()
|
| 88 |
|