import codecs
import time

from sklearn.feature_extraction.text import TfidfVectorizer

class NipTester:

    def load_reference_nips(self):
        self.nips = []
        for line in codecs.open("nips.txt", "r", "utf8"):
            line = line.strip()
            if len(line) > 0:
                for n in range(0,7):
                    self.nips.append(line)

    def init(self):
        self.load_reference_nips()
        self.vect = TfidfVectorizer(analyzer='char', ngram_range=(1, 3), min_df=0.0, max_df=1.0)
        self.tfidf = self.vect.fit_transform(self.nips)

    def find_similar(self, nip):
        nip_tfidf = self.vect.transform([nip])
        mat = (self.tfidf * nip_tfidf.T).A

        similar = []
        for i in range(0, len(self.nips)):
            sim = float(mat[i][0])
            if sim > 0.6:
                similar.append((self.nips[i], sim))
        return similar


test_nips = []
test_nips.append("986o239432")
test_nips.append("986o239.32")


nipTester = NipTester()

start = time.time()
nipTester.init()
print("Loading time: %4.2f s" % (time.time() - start))
print("Size of the reference NIP database: %d" % len(nipTester.nips))

loops = 5
searches = 0

start = time.time()
for n in range(0, loops):
    for nip in test_nips:
        matches = nipTester.find_similar(nip)
        searches += 1
        for match in matches:
            print("'%s' is similar to '%s' with score of %8.5f" % (nip, match[0], match[1]))

print("Searching time: %4.2f s" % (time.time() - start))
print("Number of searches: %d" % searches)