import codecs
from collections import defaultdict
import sys
import math

truth = defaultdict(set)

if __name__ == "__main__":
    num = int(sys.argv[1])
else:
    num = 1000

with codecs.open('data/saldo/saldo_baseforms.txt', encoding='utf-8') as f:
    for l in f:
        (w,p) = l[:-1].split('\t')
        truth[w].add(p)

trainp = defaultdict(set)
        
with codecs.open('data/saldo/p_names.txt', encoding='utf-8') as f:
    for l in f:
        (w,p) = l[:-1].split('\t')
        trainp[w].add(p)

paradigms = defaultdict(set)
with codecs.open('paradigms/se_nouns_train.para', encoding='utf-8') as f:
    for l in f:
        (_,insts) = l[:-1].split('\t')
        xs = [v.split('=')[1] for i in insts.split('#') for v in i.split(',') if v[0] == '0']
        for x in xs:
            paradigms['p_%s' % xs[0]].update(trainp[x])

paradigms_in_selection = set()

for s in paradigms.values():
    paradigms_in_selection.update(s)
    
data = []

with codecs.open('output/sv_c_nouns.txt', encoding='utf-8') as f:
    for l in f:
        (d,r,_) = l[:-1].split('\t')
        (p,w) = r.split(' ')
        data.append((float(d),p,w))

(correct,unknown,wrong) = (0,0,0)

count = 0

bins = {1:{'correct':0,'wrong':0}}
bin_number = 1

for (d,p,w) in sorted(set(data),reverse=True):
    if count == num:
        break
    if w in trainp:
        pass
    else:
        count += 1
        if count > bin_number*(num/10):
            bin_number += 1
            bins[bin_number] = {'correct':bins[bin_number-1]['correct'],'wrong':bins[bin_number-1]['wrong']}
        if len(paradigms[p].intersection(truth[w])) > 0:
            s = '+\t%d\t%s %s\t%f\t%s' % (count,p,w,d," ".join(paradigms[p].intersection(truth[w])))
            bins[bin_number]['correct'] += 1
        elif len(paradigms_in_selection.intersection(truth[w])) > 0:
            s = '-\t%d\t%s %s\t%f\t%s' % (count,p,w,d, list(paradigms[p])[0] + " => " + " ".join(paradigms_in_selection.intersection(truth[w])))
            bins[bin_number]['wrong'] += 1
        elif len(set([x.split('_')[1] for x in paradigms[p]]).intersection(set([x.split('_')[1] for x in truth[w]]))) > 0:
            s = '+?\t%d\t%s %s\t%f\t%s' % (count,p,w,d, list(paradigms[p])[0] + " =? " + " ".join(truth[w]))
            bins[bin_number]['correct'] += 1
        else:
            s = '-?\t%d\t%s %s\t%f\t%s' % (count,p,w,d, list(paradigms[p])[0] + " =>? " + " ".join(truth[w]))
            bins[bin_number]['wrong'] += 1
        print s.encode('utf-8')

print
for (n,sc) in bins.iteritems():
    (c,w) = (bins[n]['correct'],bins[n]['wrong'])
    total = float(c + w)
    (cp,wp) = (c/total*100, w/total*100)
    print '%s\tC: %d (%s), W: %d (%s)' % (str(n*10)+'%', bins[n]['correct'], '%.1f' % cp+'%' ,bins[n]['wrong'],'%.1f' % wp+'%')