#************************** PARAMETERS *****************************

fileCollectionSimilarities = "./CollectionSimilarities/dbToDb.gov2.formattedExp"

fileCollectionRanks = "./CollectionRanks/ranks_gov2_crcs_clustfuse_ideal_50_50_0.0_701-850"

fileCollectionRanksUpd = "./CollectionRanks/ranks_gov2_crcs_clustfuse_ideal_50_50_0.0_701-850_upd"

clusteringType = "online"

fileCollectionRanksColumns = 3

clusterSize = 3

collectionRerankCutoff = 10

collectionSearchCutoff = 5

#*******************************************************************

# compute clusters

clustering = Clustering()

clustering._readSimilarities(fileCollectionSimilarities)

if clusteringType == OFFLINE_CLASTERING:

clustering.computeOfflineClusters( fileCollectionSimilarities, clusterSize)

#clustering.printClusters()

# read collection scores

resManager = ResManager()

resManager.clean()

resManager.readResults(fileCollectionRanks, fileCollectionRanksColumns)

# open output stream

otp = open(fileCollectionRanksUpd, 'w')

# compute scores

for query in resManager.getQueries():

if clusteringType == ONLINE_CLASTERING:

initialCollectionSet = resManager.getDocs(query, collectionRerankCutoff)

clustering.computeOnlineClusters(initialCollectionSet, clusterSize)

print query, clustering.clusters.keys()

print query, clustering.clusters

scoreFunction = GeometricScoreFunction()

clusterScores = scoreFunction.scoreClusters( clustering, resManager, query )

# prepare result collection ranking

boost = 100

rerankedCollections = scoreFunction.scoreCollections(clustering, resManager, clusterScores, query, boost )

print query, rerankedCollections

# check if number of re-ranked collections satisfies collection cutoff

# if it does not, then append more collections to the end of the list

if len(rerankedCollections) < collectionSearchCutoff:

initialCollectionIds = resManager.getDocs(query)

initialCollectionScores = resManager.getScores(query)

pos = 0

while pos < len(initialCollectionIds) and len(rerankedCollections) < collectionSearchCutoff :

collectionId = initialCollectionIds[pos]

if not rerankedCollections.has_key(collectionId):

rerankedCollections[collectionId] = initialCollectionScores[pos]

pos += 1

# save results to a file

for collection in sorted( rerankedCollections.iteritems(), key=itemgetter(1), reverse=True ):

otp.write( " ".join( [query, collection[0], str(collection[1])] ) + '\n' )

otp.close()