me@1942: #!/usr/bin/python
me@1942: # -*- coding: utf-8 -*-
me@1942: 
me@1942: import xml.etree.ElementTree as ET
me@1942: import os       # for getting files from directory
me@1942: import operator # for sorting data with multiple keys
me@1942: import sys      # for accessing command line arguments
me@1942: 
me@1942: # Command line arguments
me@1942: assert len(sys.argv)<3, "evaluation_stats takes at most 1 command line argument\n"+\
me@1942:                         "Use: python evaluation_stats.py [results_folder]"
me@1942: 
me@1942: # XML results files location
me@1942: if len(sys.argv) == 1:
me@1942:     folder_name = "../saves"    # Looks in 'saves/' folder from 'scripts/' folder
me@1942:     print "Use: python evaluation_stats.py [results_folder]"
me@1942:     print "Using default path: " + folder_name
me@1942: elif len(sys.argv) == 2:
me@1942:     folder_name = sys.argv[1]   # First command line argument is folder
me@1942: 
me@1942: # Turn number of seconds (int) to '[minutes] min [seconds] s' (string)
me@1942: def seconds2timestr(time_in_seconds):
me@1942:     time_in_minutes = int(time_in_seconds/60)
me@1942:     remaining_seconds = int(time_in_seconds%60)
me@1942:     return str(time_in_minutes) + " min " + str(remaining_seconds) + " s"
me@1942: 
me@1942: # stats initialisation
me@1942: number_of_XML_files  = 0
me@1942: number_of_pages      = 0
me@1942: number_of_fragments  = 0
me@1942: total_empty_comments = 0
me@1942: total_not_played     = 0
me@1942: total_not_moved      = 0
me@1942: time_per_page_accum  = 0
me@1942: 
me@1942: # arrays initialisation
me@1942: page_names       = []
me@1942: page_count       = []
me@1942: duration_page    = []      # duration of experiment in function of page content
me@1942: duration_order   = []      # duration of experiment in function of page number
me@1942: fragments_per_page = []    # number of fragments for corresponding page
me@1942: 
me@1942: # get every XML file in folder
me@1942: files_list = os.listdir(folder_name)
me@1942: for file in files_list: # iterate over all files in files_list
me@1942:     if file.endswith(".xml"): # check if XML file
me@1942:         number_of_XML_files += 1
me@1942:         tree = ET.parse(folder_name + '/' + file)
me@1942:         root = tree.getroot()
me@1942:         
me@1942:         print file # print file name (subject name)
me@1942:         
me@1942:         # reset for new subject
me@1942:         total_duration = 0
me@1942:         page_number = 0
me@1942:         
me@1942:         # get list of all page names
me@1942:         for audioholder in root.findall("./audioholder"):   # iterate over pages
me@1942:             page_name = audioholder.get('id')               # get page name
me@1942:             
me@1942:             if page_name is None: # ignore 'empty' audio_holders
me@1942:                 print "WARNING: " + file + " contains empty audio holder. (evaluation_stats.py)"
me@1942:                 break # move on to next
me@1942:             
me@1942:             number_of_comments = 0 # for this page
me@1942:             number_of_missing_comments = 0 # for this page
me@1942:             not_played = 0 # for this page
me@1942:             not_moved = 0 # for this page
me@1942:             
me@1942:             # 'testTime' keeps total duration: subtract time so far for duration of this audioholder
me@1942:             duration = float(audioholder.find("./metric/metricresult[@id='testTime']").text) - total_duration
me@1942:             
me@1942:             # total duration of test
me@1942:             total_duration += duration
me@1942:             
me@1942:             # number of audio elements
me@1942:             audioelements = audioholder.findall("./audioelement") # get audioelements
me@1942:             number_of_fragments += len(audioelements) # add length of this list to total
me@1942:             
me@1942:             # number of comments (interesting if comments not mandatory)
me@1942:             for audioelement in audioelements:
me@1942:                 response = audioelement.find("./comment/response")
me@1942:                 was_played = audioelement.find("./metric/metricresult/[@name='elementFlagListenedTo']")
me@1942:                 was_moved = audioelement.find("./metric/metricresult/[@name='elementFlagMoved']")
me@1942:                 if response.text is not None and len(response.text) > 1: 
me@1942:                     number_of_comments += 1
me@1942:                 else: 
me@1942:                     number_of_missing_comments += 1
me@1942:                 if was_played is not None and was_played.text == 'false': 
me@1942:                     not_played += 1
me@1942:                 if was_moved is not None and was_moved.text == 'false': 
me@1942:                     not_moved += 1
me@1942:             
me@1942:             # update global counters
me@1942:             total_empty_comments += number_of_missing_comments
me@1942:             total_not_played += not_played
me@1942:             total_not_moved += not_moved
me@1942:             
me@1942:             # print audioholder id and duration
me@1942:             print "    " + page_name + ": " + seconds2timestr(duration) + ", "\
me@1942:                   + str(number_of_comments)+"/"\
me@1942:                   +str(number_of_comments+number_of_missing_comments)+" comments"
me@1942:             
me@1942:             # number of audio elements not played
me@1942:             if not_played > 1:
me@1942:                 print 'ATTENTION: '+str(not_played)+' fragments were not listened to!'
me@1942:             if not_played == 1: 
me@1942:                 print 'ATTENTION: one fragment was not listened to!'
me@1942:             
me@1942:             # number of audio element markers not moved
me@1942:             if not_moved > 1:
me@1942:                 print 'ATTENTION: '+str(not_moved)+' markers were not moved!'
me@1942:             if not_moved == 1: 
me@1942:                 print 'ATTENTION: one marker was not moved!'
me@1942:             
me@1942:             # keep track of duration in function of page index
me@1942:             if len(duration_order)>page_number:
me@1942:                 duration_order[page_number].append(duration)
me@1942:             else:
me@1942:                 duration_order.append([duration])
me@1942:                 
me@1942:             # keep list of audioholder ids and count how many times each audioholder id
me@1942:             # was tested, how long it took, and how many fragments there were (if number of 
me@1942:             # fragments is different, store as different audioholder id)
me@1942:             if page_name in page_names: 
me@1942:                 page_index = page_names.index(page_name) # get index
me@1942:                 # check if number of audioelements the same
me@1942:                 if len(audioelements) == fragments_per_page[page_index]: 
me@1942:                     page_count[page_index] += 1
me@1942:                     duration_page[page_index].append(duration)
me@1942:                 else: # make new entry
me@1942:                     alt_page_name = page_name+"("+str(len(audioelements))+")"
me@1942:                     if alt_page_name in page_names: # if already there
me@1942:                         alt_page_index = page_names.index(alt_page_name) # get index
me@1942:                         page_count[alt_page_index] += 1
me@1942:                         duration_page[alt_page_index].append(duration)
me@1942:                     else: 
me@1942:                         page_names.append(alt_page_name)
me@1942:                         page_count.append(1)
me@1942:                         duration_page.append([duration])
me@1942:                         fragments_per_page.append(len(audioelements))
me@1942:             else: 
me@1942:                 page_names.append(page_name)
me@1942:                 page_count.append(1)
me@1942:                 duration_page.append([duration])
me@1942:                 fragments_per_page.append(len(audioelements))
me@1942:                 
me@1942:             # bookkeeping
me@1942:             page_number += 1 # increase page count for this specific test
me@1942:             number_of_pages += 1 # increase total number of pages
me@1942:             time_per_page_accum += duration # total duration (for average time spent per page)
me@1942: 
me@1942:         # print total duration of this test
me@1942:         print "    TOTAL: " + seconds2timestr(total_duration)
me@1942: 
me@1942: 
me@1942: # PRINT EVERYTHING
me@1942: 
me@1942: print "Number of XML files: " + str(number_of_XML_files)
me@1942: print "Number of pages: " + str(number_of_pages)
me@1942: print "Number of fragments: " + str(number_of_fragments)
me@1942: print "Number of empty comments: " + str(total_empty_comments) +\
me@1942:       " (" + str(round(100.0*total_empty_comments/number_of_fragments,2)) + "%)"
me@1942: print "Number of unplayed fragments: " + str(total_not_played) +\
me@1942:       " (" + str(round(100.0*total_not_played/number_of_fragments,2)) + "%)"
me@1942: print "Number of unmoved markers: " + str(total_not_moved) +\
me@1942:       " (" + str(round(100.0*total_not_moved/number_of_fragments,2)) + "%)"
me@1942: print "Average time per page: " + seconds2timestr(time_per_page_accum/number_of_pages)
me@1942: 
me@1942: # Pages and number of times tested
me@1942: page_count_strings = list(str(x) for x in page_count)
me@1942: count_list = page_names + page_count_strings
me@1942: count_list[::2] = page_names
me@1942: count_list[1::2] = page_count_strings
me@1942: print "Pages tested: " + str(count_list)
me@1942: 
me@1942: # Average duration for first, second, ... page
me@1942: print "Average duration per page:"
me@1942: for page_number in range(len(duration_order)): 
me@1942:     print "        page " + str(page_number+1) + ": " +\
me@1942:         seconds2timestr(sum(duration_order[page_number])/len(duration_order[page_number])) +\
me@1942:             " ("+str(len(duration_order[page_number]))+" subjects)"
me@1942: 
me@1942: 
me@1942: # Sort pages by number of audioelements, then by duration
me@1942: 
me@1942: # average duration and number of subjects per page
me@1942: average_duration_page = []
me@1942: number_of_subjects_page = []
me@1942: for line in duration_page:
me@1942:     number_of_subjects_page.append(len(line))
me@1942:     average_duration_page.append(sum(line)/len(line))
me@1942: 
me@1942: # combine and sort in function of number of audioelements and duration
me@1942: combined_list = [page_names, average_duration_page, fragments_per_page, number_of_subjects_page]
me@1942: combined_list = sorted(zip(*combined_list), key=operator.itemgetter(1, 2)) # sort
me@1942: 
me@1942: # Show average duration for all songs
me@1942: print "Average duration per audioholder:"
me@1942: for page_index in range(len(page_names)):
me@1942:     print "        "+combined_list[page_index][0] + ": " \
me@1942:           + seconds2timestr(combined_list[page_index][1]) \
me@1942:           + " (" + str(combined_list[page_index][3]) + " subjects, " \
me@1942:           + str(combined_list[page_index][2]) + " fragments)"
me@1942: 
me@1942: 
me@1942: #TODO
me@1942: # time per page in function of number of fragments (plot)
me@1942: # time per participant in function of number of pages
me@1942: # plot total time for each participant
me@1942: # plot total time
me@1942: # show 'count' per page (in order)
me@1942: 
me@1942: # clear up page_index <> page_count <> page_number confusion
me@1942: 
me@1942: # LaTeX -> PDF print out