def create_question(question):
if not question:
return False
question_query = Question.all(keys_only=True)
question_query.filter("question_id =", question.question_id)
if question_query.get():
return False
logging.info("Create a new question")
question.terms = extract_terms(question.title)
question.terms += generate_local_terms(question.terms, question.place_ids)
if question.terms:
term_dict = dict(zip(question.terms, [1]*len(question.terms)))
update_termstats(term_dict)
db.put(question)
return True
def find_relevant_questions(query, place_ids=[], max_num=10):
query = query.strip()
query_terms = extract_terms(query)
query_terms += generate_local_terms(query_terms, place_ids)
# NOTE:
# the following code uses list-properties and merge-join to implement keyword search
# but it leads to the problem of exploding index if len(query_terms) >= 2
#
# select_str = "SELECT * FROM Question WHERE"
# where_str = " AND ".join([("terms = '%s'" % term) for term in query_terms])
# order_str = "ORDER BY create_time DESC" # useful as sorted() is guaranteed to be stable
# limit_str = "LIMIT 200"
# questions = db.GqlQuery(select_str+" "+where_str+" "+order_str+" "+limit_str)
#
questions = []
if query_terms:
termstats = TermStat.get_by_key_name(query_terms)
term_dict = dict([(termstat.key().name(),termstat.docfreq) for termstat in termstats if termstat])
terms = sorted(term_dict.keys(),
key=lambda term: term_dict[term])
best_terms = []
k = 0
for term in terms:
k += 1
if len(best_terms) < 1:
best_terms.append(term)
else:
if (k <= 5) and (term_dict[term] <= 20):
best_terms.append(term)
if best_terms:
question_query = Question.all()
question_query.filter("terms IN", best_terms)
question_query.order("-create_time")
questions = question_query.fetch(max_num*10) # the number of questions to be ranked
if questions:
questions.sort(key=lambda question: question_score(question,term_dict),
reverse=True)
return questions[:max_num]
