# coding=UTF-8
#from django.utils.encoding import smart_str, smart_unicode
from xml.parsers import expat
import sys

class Parser:

	def __init__(self):
		self._parser = expat.ParserCreate()
		self._parser.StartElementHandler = self.start
		self._parser.EndElementHandler = self.end
		self._parser.CharacterDataHandler = self.data

	def feedFile(self, inputFileName, createSQLFileName, max_inserts_num, insert_header_sql):
		# init basic flag(s)
		self.inPage = 0
		self.page_counter = 0
		self.inTag = {}
		self.max_inserts_count = 0
		# init stuff from params
		self.max_inserts = max_inserts_num
		self.insert_header = insert_header_sql
		self._f = open (inputFileName+'.sql', 'wb')
		# add create sql
		ftmp = open (createSQLFileName, 'rb')
		self._f.write(ftmp.read())
		# run parser
		self._parser.ParseFile(open(inputFileName))

	def close(self):
		self._parser.Parse("", 1) # end of data
		del self._parser # get rid of circular references

	# Any tag start found. Note! page must not contain other page tag!
	def start(self, tag, attrs):
		if tag == 'page':
			self.inPage = 1
			self.page_counter += 1
			self.rev_row_clear()
		elif self.inPage:
			self.inTag[tag] = self.inTag.get(tag,0) + 1
			if tag == 'revision':
				self.rev_row_partial_clear()

	# Any tag end found
	def end(self, tag):
		if tag == 'page':
			self.inPage = 0
		elif self.inPage:
			self.inTag[tag] -= 1
			if tag == 'revision':
				self.rev_row_insert()
	
	# Any data found
	def data(self, data):
		if self.inPage:
			#data = unicode(data, "utf-8")
			if self.inTag.get('contributor', 0):
				if self.inTag.get('id', 0):
					self.rev_row['user'] = data
			else:
				if self.inTag.get('title', 0):
					print('@page: ' + data)
					if data.find("MediaWiki:")==0:
						self.rev_row['is_mw_ns'] = 1
						
				elif self.inTag.get('revision', 0):
					if self.inTag.get('id', 0):
						self.rev_row['id'] = data
						
					elif self.inTag.get('timestamp', 0):
						self.rev_row['log_timestamp'] = str(data).translate(None, "-T:Z")

	# clear rev_row and init it's data if needed
	def rev_row_clear(self):
		self.rev_row = {}
		self.rev_row['is_mw_ns'] = 0
	def rev_row_partial_clear(self):
		tmp = self.rev_row['is_mw_ns']
		self.rev_row = {}
		self.rev_row['is_mw_ns'] = tmp
	
	# insert rev_row (output SQL)
	def rev_row_insert(self):
		#sys.stderr.write('Adding revision with id: %s\n' % self.rev_row['id'])	#self.page_counter)
		#print('Adding revision with id: %s' % self.rev_row['id'])	#self.page_counter)
		if (self.max_inserts_count >= self.max_inserts):
			self.max_inserts_count = 0
			self._f.write(";\n")
		if (self.max_inserts_count==0):
			self._f.write(self.insert_header)

		if (self.max_inserts_count!=0):
			self._f.write(",")
		self._f.write("(")
		#self._f.write(self.rev_row['id'])
		#self._f.write(",'")
		self._f.write("'")
		self._f.write(self.rev_row.get('user', '0').encode('utf-8'))
		self._f.write("','")
		self._f.write(str(self.rev_row['is_mw_ns']))
		self._f.write("','")
		self._f.write(self.rev_row['log_timestamp'].encode('utf-8'))
		self._f.write("')\n")
		
		self.max_inserts_count+=1
#
# Run
#
p = Parser()
#p.feedFile('plwiki-latest-stub-meta-history.xml', 'revision_create.sql', 5000, "INSERT INTO revision (rev_user, rev_is_mw_ns, rev_timestamp) VALUES\n")
p.feedFile('revision-test.xml', 'revision_create.sql', 5000, "INSERT INTO revision (rev_user, rev_is_mw_ns, rev_timestamp) VALUES\n")
p.close()