hindi_search.py

import math
import argparse
import linecache
from collections import Counter
from hindi_indexer import *


'''
This class traverses various index files and return the required data
'''
class FileTraverser():

	def __init__(self):

		pass

	def binary_search_token_info(self, high, filename, inp_token):
		
		low = 0
		while low < high:
			
			mid = (low + high)//2
			line = linecache.getline(filename, mid)
			token = line.split('-')[0]

			if inp_token == token:
				token_info = line.split('-')[1:-1]
				return token_info
			
			elif inp_token > token:
				low = mid + 1
			
			else:
				high = mid

		return None

	def title_search(self, page_id):

		title = linecache.getline('../output_data/hindi_wiki_index/id_title_map.txt', int(page_id)+1).strip()
		title = title.split('-', 1)[1]

		return title


	def search_field_file(self, field, file_num, line_num):
		
		if line_num != '':
			line = linecache.getline(f'../output_data/hindi_wiki_index/{field}_data_{str(file_num)}.txt', int(line_num)).strip()
			postings = line.split('-')[1]

			return postings

		return ''

'''
This class implements the ranking functionality on the returned postings.
'''
class Ranker():

	def __init__(self, num_pages):

		self.num_pages = num_pages

	def do_ranking(self, page_freq, page_postings):
		
		result = defaultdict(float)
		weightage_dict = {'title':1.0, 'body':0.6, 'category':0.4, 'infobox':0.75, 'link':0.20, 'reference':0.25}
		
		for token, field_post_dict in page_postings.items():
			
			for field, postings in field_post_dict.items():
				
				weightage = weightage_dict[field]
				
				if len(postings)>0:
					for post in postings.split(';'):
						
						id, post = post.split(':')
						result[id] += weightage*(1+math.log(int(post)))*math.log((self.num_pages-int(page_freq[token]))/int(page_freq[token]))

		return result


'''
This class takes query as input and returns the corresponding postings along with theis fields
'''
class QueryResults():

	def __init__(self, file_traverser):

		self.file_traverser = file_traverser

	def simple_query(self, preprocessed_query, num_tokens, tokens_info_pointer):

		page_freq, page_postings = {}, defaultdict(dict)
    	
		char_list = [chr(i) for i in range(97,123)]
		for token in preprocessed_query:
			
			token_info = self.file_traverser.binary_search_token_info(num_tokens, tokens_info_pointer, token)

			if token_info:
				file_num, freq, title_line, body_line, category_line, infobox_line, link_line, reference_line = token_info
				line_map = {
						'title' : title_line, 'body' : body_line, 'category' : category_line, 'infobox' : infobox_line, 'link' : link_line, 'reference' : reference_line
					}

				for field_name, line_num in line_map.items():
					
					if line_num!='':
						posting = self.file_traverser.search_field_file(field_name, file_num, line_num)

						page_freq[token] = len(posting.split(';'))
						page_postings[token][field_name] = posting

		return page_freq, page_postings


	def field_query(self, preprocessed_query_final, num_tokens, tokens_info_pointer):

		page_freq, page_postings = {}, defaultdict(dict)
		char_list = [chr(i) for i in range(97,123)]

		for field, token in preprocessed_query_final:

			token_info = self.file_traverser.binary_search_token_info(num_tokens, tokens_info_pointer, token)

			if token_info:
				
				file_num, freq, title_line, body_line, category_line, infobox_line, link_line, reference_line = token_info
				line_map = {
					'title':title_line, 'body':body_line, 'category':category_line, 'infobox':infobox_line, 'link':link_line, 'reference':reference_line
				}
				field_map = {
					't':'title', 'b':'body', 'c':'category', 'i':'infobox', 'l':'link', 'r':'reference'
				}

				field_name = field_map[field]
				line_num = line_map[field_name]

				posting = self.file_traverser.search_field_file(field_name, file_num, line_num)

				page_freq[token] = len(posting)
				page_postings[token][field_name] = posting

		return page_freq, page_postings


'''
This class runs the above functions to implement search and ranking and returns the required results.
'''	
class RunQuery():

	def __init__(self, text_pre_processor, file_traverser, ranker, query_results):

		self.file_traverser = file_traverser
		self.text_pre_processor = text_pre_processor
		self.ranker = ranker
		self.query_results = query_results

	def identify_query_type(self, query):
		
		field_replace_map = {
				' t:':';t:',
				' b:':';b:',
				' c:':';c:',
				' i:':';i:',
				' l:':';l:',
				' r:':';r:',
			}

		if ('t:' in query or 'b:' in query or 'c:' in query or 'i:' in query or 'l:' in query or 'r:' in query) and query[0:2] not in ['t:', 'b:', 'i:', 'c:', 'r:', 'l:']:

			for k, v in field_replace_map.items():
				if k in query:
					query = query.replace(k, v)

			query = query.lstrip(';')

			return query.split(';')[0], query.split(';')[1:]

		elif 't:' in query or 'b:' in query or 'c:' in query or 'i:' in query or 'l:' in query or 'r:' in query:

			for k, v in field_replace_map.items():
				if k in query:
					query = query.replace(k, v)

			query = query.lstrip(';')

			return query.split(';'), None

		else:
			return query, None

	def return_query_results(self, query, query_type, num_tokens, tokens_info_pointer):

		if query_type=='field':
			preprocessed_query = [[qry.split(':')[0], self.text_pre_processor.preprocess_text(qry.split(':')[1])] for qry in query]
		else:
			preprocessed_query = self.text_pre_processor.preprocess_text(query)

		if query_type == 'field':

			preprocessed_query_final = []
			for field, words in preprocessed_query:
				for word in words:
					preprocessed_query_final.append([field, word])

			page_freq, page_postings = self.query_results.field_query(preprocessed_query_final, num_tokens, tokens_info_pointer)
		
		else:
			
			page_freq, page_postings = self.query_results.simple_query(preprocessed_query, num_tokens, tokens_info_pointer)

		ranked_results = self.ranker.do_ranking(page_freq, page_postings)

		return ranked_results

	def take_input_from_file(self, file_name, num_results, num_tokens, tokens_info_pointer):
		results_file = file_name.split('.txt')[0]

		with open(file_name, 'r') as f:
			fp = open(results_file+'_op.txt', 'w')
			for i, query in enumerate(f):
				s = time.time()

				query = query.strip()
				query1, query2 = self.identify_query_type(query)

				if query2:
					ranked_results1 = self.return_query_results(query1, 'simple', num_tokens, tokens_info_pointer)

					ranked_results2 = self.return_query_results(query2, 'field', num_tokens, tokens_info_pointer)

					ranked_results = Counter(ranked_results1) + Counter(ranked_results2)
					results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
					results = results[:num_results]

					if results:
						for id, _ in results:
							title= self.file_traverser.title_search(id)
							fp.write(id + ', ' + title)
							fp.write('\n')
					else:
						fp.write('No matching Doc found')
						fp.write('\n')

				elif type(query1)==type([]):

					ranked_results = self.return_query_results(query1, 'field', num_tokens, tokens_info_pointer)

					results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
					results = results[:num_results]

					if results:
						for id, _ in results:
							title= self.file_traverser.title_search(id)
							fp.write(id + ', ' + title)
							fp.write('\n')
					else:
						fp.write('No matching Doc found')
						fp.write('\n')

				else:
					ranked_results = self.return_query_results(query1, 'simple', num_tokens, tokens_info_pointer)

					results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
					results = results[:num_results]

					if results:
						for id, _ in results:
							title= self.file_traverser.title_search(id)
							fp.write(id + ', ' + title)
							fp.write('\n')
					else:
						fp.write('No matching Doc found')
						fp.write('\n')
				e = time.time()
				fp.write('Finished in ' + str(e-s) + ' seconds')
				fp.write('\n\n')

				print('Done query', i+1)

			fp.close()

		print('Done writing results')

	def take_input_from_user(self, num_results, num_tokens, tokens_info_pointer):

		start = time.time()

		while True:
			query = input('Enter Query:- ')

			s = time.time()

			query = query.strip()
			query1, query2 = self.identify_query_type(query)

			if query2:
				ranked_results1 = self.return_query_results(query1, 'simple', num_tokens, tokens_info_pointer)

				ranked_results2 = self.return_query_results(query2, 'field', num_tokens, tokens_info_pointer)

				ranked_results = Counter(ranked_results1) + Counter(ranked_results2)
				results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
				results = results[:num_results]

				for id, _ in results:
					title= self.file_traverser.title_search(id)
					print(id+',', title)

			elif type(query1)==type([]):

				ranked_results = self.return_query_results(query1, 'field', num_tokens, tokens_info_pointer)

				results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
				results = results[:num_results]

				for id, _ in results:
					title= self.file_traverser.title_search(id)
					print(id+',', title)

			else:
				ranked_results = self.return_query_results(query1, 'simple', num_tokens, tokens_info_pointer)

				results = sorted(ranked_results.items(), key = lambda item : item[1], reverse=True)
				results = results[:num_results]

				for id, _ in results:
					title= self.file_traverser.title_search(id)
					print(id+',', title)

			e = time.time()
			print('Finished in', e-s, 'seconds')
			print()


'''
This is the main function which does entire searching task.
'''
if __name__=='__main__':
	
	start = time.time()

	arg_parser = argparse.ArgumentParser()
	arg_parser.add_argument('--filename', action='store', type=str)
	arg_parser.add_argument('--num_results', action='store', default=10, type=int)

	args = arg_parser.parse_args()

	file_name = args.filename
	num_results = args.num_results

	print('Loading search engine ')
	html_tags = re.compile('&amp;|&apos;|&gt;|&lt;|&nbsp;|&quot;')
	with open('hindi_stopwords.txt', 'r') as f:
		stop_words = [word.strip() for word in f]

	with open('hindi_stem_words.txt', 'r') as f:
		stem_words = [word.strip() for word in f]

	with open('../output_data/hindi_wiki_index/num_pages.txt', 'r') as f:
		num_pages = float(f.readline().strip())

	with open('../output_data/hindi_wiki_index/num_tokens.txt', 'r') as f:
		num_tokens = int(f.readline().strip())

	tokens_info_pointer = '../output_data/hindi_wiki_index/tokens_info.txt'
	text_pre_processor = TextPreProcessor(html_tags, stem_words, stop_words)
	file_traverser = FileTraverser()
	ranker = Ranker(num_pages)
	query_results = QueryResults(file_traverser)
	run_query = RunQuery(text_pre_processor, file_traverser, ranker, query_results)

	temp = linecache.getline('../output_data/hindi_wiki_index/id_title_map.txt', 0)

	print('Loaded in', time.time() - start, 'seconds')

	print('Starting Querying')

	start = time.time()
	
	if file_name is not None:

		run_query.take_input_from_file(file_name, num_results, num_tokens, tokens_info_pointer)
	
	else:

		run_query.take_input_from_user(num_results, num_tokens, tokens_info_pointer)

	
	print('Done querying in', time.time() - start, 'seconds')