"""
Test filtering.
arguments:
    topics file
    annotated topics file
    relevance judgements
    path of ids and content per query for realtime filtering (test set)
    training set dir
    results dir
    "external" for using external information, otherwise "internal"
    parameters: classifier (R, NC), classifier parameter, number of negative samples,
        minimum link probability, feature extraction function names (divided by .)
        for twitter status, for generic feature extraction (used on queries), for entity linking expansion.
        e.g. R-0.2-100-....-terms.bigrams-terms-surfaceForms
    [query numbers divided by :]
"""


import sys, errno, os

from CipCipPy.utils.io import readQueries, readQrels
from CipCipPy.filtering import SupervisedFilterer
from CipCipPy.classification import *
from CipCipPy.classification.feature import *

#FIXME use argparse

queries = readQueries(sys.argv[1])

with open(sys.argv[2]) as ann_qfile:
    queriesAnnotated = json.load(ann_qfile)

if len(sys.argv) > 8:
    queries = [q for q in queries if q[0] in set(sys.argv[8].split(':'))]

qrels = readQrels(sys.argv[3], set(q[0] for q in queries))
dataset_path = sys.argv[4]
resultsPath = sys.argv[5]

external = False
if sys.argv[6] == 'external':
    external = True

param = sys.argv[7].split('-')

classifier, classifierParam, neg, minLinkProb, expansion_limit, statusFeatures, genericFeatures, \
    entityFeatures = param

if classifier == 'NC':
    classifier = NCClassifier(shrink=float(classifierParam) if classifierParam != 'None' else None)
elif classifier == 'R':
    classifier = RClassifier(alpha=float(classifierParam))
elif classifier == 'ADA':
    classifier = ADAClassifier(estimators=int(classifierParam))
elif classifier == 'L':
    classifier = LClassifier(C=float(classifierParam))
elif classifier == 'DT':
    classifier = DTClassifier()
elif classifier == 'KNN':
    classifier = KNNClassifier()
elif classifier == 'RF':
    classifier = RFClassifier()
elif classifier == 'RO':
    classifier = RocchioClassifier(threshold=float(classifierParam))

f = SupervisedFilterer(classifier)

f.setFeatureExtractor([eval(feat) for feat in statusFeatures.split('.')],
                      [eval(feat) for feat in genericFeatures.split('.')],
                      [eval(feat) for feat in entityFeatures.split('.')],
                      float(minLinkProb),
                      expansion_limit=float(expansion_limit))

runName = 'run' + '-'.join(param) + ('_external' if external else '_internal')

dumpsPath = os.path.join(resultsPath, 'dumps_' + runName)
if not os.path.exists(dumpsPath):
    try:
        os.makedirs(dumpsPath)
    except OSError as e:
        if e.errno == errno.EEXIST:
            print 'Warning: ', e
            print 'continuing...'
        else:
            raise

results, printOut = f.get(queries, queriesAnnotated, int(neg), dataset_path,
            qrels, external, dumpsPath=dumpsPath)

#print printOut
#printEval(sys.argv[1], sys.argv[3], results)