merge_csv.py

   1 #!/usr/bin/env python
   2 # -*- coding: utf-8 -*-
   3
   4 import os, sys
   5 import codecs
   6
   7 sys.stdout = codecs.getwriter('utf_8')(sys.stdout)
   8 sys.stderr = codecs.getwriter('cp932')(sys.stderr)
   9 input_codec = "cp932"
  10 output_codec = "cp932"
  11
  12 try:
  13         ga_data_path = sys.argv[1]
  14         otp_data_path = sys.argv[2]
  15         output_path = sys.argv[3]
  16 except IndexError:
  17         sys.exit(sys.argv[0] + " <ga file> <otp file> <output_file>")
  18
  19 ga_file = codecs.open(ga_data_path, "r", input_codec)
  20 output_file = codecs.open(output_path, "w", output_codec)
  21
  22 ga_dict = {}
  23 ga_titles = []
  24 for row in ga_file:
  25 #Page Title,Pageviews,Unique Pageviews,Avg. Time on Page,Bounce Rate,% Exit,$ Index
  26
  27         items = row.strip().split( "," )
  28         title = items.pop(0)
  29         ga_dict[title] = items
  30         ga_titles.append(title)
  31 #       print title
  32 ga_file.close()
  33
  34 otp_file = codecs.open(otp_data_path, "r", input_codec)
  35 updated_dict = {}
  36 ga_notfounds = []
  37 for row in otp_file:
  38 #0   1     2      3   4        5    6
  39 #url,title,editor,PVs,comments,date,tags
  40         items = row.strip().split( "," )
  41
  42         for ga_title in ga_titles:
  43                 if ga_title.find( items[1] ) != -1:
  44                         ga_info = ga_dict.pop(ga_title)
  45                         ga_info.append( items[5] )
  46                         ga_info.append( items[6] )
  47                         updated_dict[ga_title] = ga_info
  48                         break
  49         else:
  50                 sys.stderr.write( "! %s - %s\n" % (items[1],items[5]) )
  51
  52 otp_file.close()
  53 for title in updated_dict:
  54         print >> output_file, title, ",",  ",".join( updated_dict[title] )
  55
  56 for title in ga_dict:
  57         print >> output_file, title, ",",  ",".join( ga_dict[title] )
  58
  59 output_file.close()