BrechtDeMan@1071: #!/usr/bin/python
BrechtDeMan@1071: # -*- coding: utf-8 -*-
BrechtDeMan@1071: 
BrechtDeMan@1071: import xml.etree.ElementTree as ET
BrechtDeMan@1071: import os       # for getting files from directory
BrechtDeMan@1071: import operator # for sorting data with multiple keys
BrechtDeMan@1071: import sys      # for accessing command line arguments
BrechtDeMan@1071: import subprocess # for calling pdflatex
BrechtDeMan@1071: import shlex # for calling pdflatex
BrechtDeMan@1071: import matplotlib.pyplot as plt # plots
BrechtDeMan@1071: import numpy as np # numbers
BrechtDeMan@1071: 
BrechtDeMan@1071: # Command line arguments
BrechtDeMan@1071: assert len(sys.argv)<3, "evaluation_stats takes at most 1 command line argument\n"+\
BrechtDeMan@1071:                         "Use: python evaluation_stats.py [results_folder]"
BrechtDeMan@1071: 
BrechtDeMan@1071: # XML results files location
BrechtDeMan@1071: if len(sys.argv) == 1:
BrechtDeMan@1071:     folder_name = "../saves"    # Looks in 'saves/' folder from 'scripts/' folder
BrechtDeMan@1071:     print "Use: python evaluation_stats.py [results_folder]"
BrechtDeMan@1071:     print "Using default path: " + folder_name
BrechtDeMan@1071: elif len(sys.argv) == 2:
BrechtDeMan@1071:     folder_name = sys.argv[1]   # First command line argument is folder
BrechtDeMan@1071: 
BrechtDeMan@1071: # Turn number of seconds (int) to '[minutes] min [seconds] s' (string)
BrechtDeMan@1071: def seconds2timestr(time_in_seconds):
BrechtDeMan@1071:     time_in_minutes = int(time_in_seconds/60)
BrechtDeMan@1071:     remaining_seconds = int(time_in_seconds%60)
BrechtDeMan@1071:     return str(time_in_minutes) + " min " + str(remaining_seconds) + " s"
BrechtDeMan@1071: 
BrechtDeMan@1071: # stats initialisation
BrechtDeMan@1071: number_of_XML_files  = 0
BrechtDeMan@1071: number_of_pages      = 0
BrechtDeMan@1071: number_of_fragments  = 0
BrechtDeMan@1071: total_empty_comments = 0
BrechtDeMan@1071: total_not_played     = 0
BrechtDeMan@1071: total_not_moved      = 0
BrechtDeMan@1071: time_per_page_accum  = 0
BrechtDeMan@1071: 
BrechtDeMan@1071: # arrays initialisation
BrechtDeMan@1071: page_names       = []
BrechtDeMan@1071: page_count       = []
BrechtDeMan@1071: duration_page    = []      # duration of experiment in function of page content
BrechtDeMan@1071: duration_order   = []      # duration of experiment in function of page number
BrechtDeMan@1071: fragments_per_page = []    # number of fragments for corresponding page
BrechtDeMan@1071: 
BrechtDeMan@1071: # get username if available
BrechtDeMan@1071: for name in ('LOGNAME', 'USER', 'LNAME', 'USERNAME'):
BrechtDeMan@1071:     user = os.environ.get(name)
BrechtDeMan@1071:     if user:
BrechtDeMan@1071:         break
BrechtDeMan@1071:     else:
BrechtDeMan@1071:         user = ''
BrechtDeMan@1071: 
BrechtDeMan@1071: 
BrechtDeMan@1071: # begin LaTeX document
BrechtDeMan@1071: header = r'''\documentclass[11pt, oneside]{article} 
BrechtDeMan@1071:           \usepackage{geometry}
BrechtDeMan@1071:           \geometry{letterpaper}
BrechtDeMan@1071:           \usepackage[parfill]{parskip}
BrechtDeMan@1071:           \usepackage{graphicx}
BrechtDeMan@1071:           \title{Report}
BrechtDeMan@1071:           \author{'''+\
BrechtDeMan@1071:           user+\
BrechtDeMan@1071:           r'''}
BrechtDeMan@1071:           \graphicspath{{'''+\
BrechtDeMan@1071:           folder_name+\
BrechtDeMan@1071:           r'''/}}
BrechtDeMan@1071:           \begin{document}
BrechtDeMan@1071:           \maketitle
BrechtDeMan@1071:           \tableofcontents
BrechtDeMan@1071:           '''
BrechtDeMan@1071:           
BrechtDeMan@1071: footer = '\end{document}'
BrechtDeMan@1071: 
BrechtDeMan@1071: body = ''
BrechtDeMan@1071: 
BrechtDeMan@1071: # generate images for later use
BrechtDeMan@1071: subprocess.call("timeline_view_movement.py", shell=True)
BrechtDeMan@1071: 
BrechtDeMan@1071: # get every XML file in folder
BrechtDeMan@1071: files_list = os.listdir(folder_name)
BrechtDeMan@1071: for file in files_list: # iterate over all files in files_list
BrechtDeMan@1071:     if file.endswith(".xml"): # check if XML file
BrechtDeMan@1071:         number_of_XML_files += 1
BrechtDeMan@1071:         tree = ET.parse(folder_name + '/' + file)
BrechtDeMan@1071:         root = tree.getroot()
BrechtDeMan@1071:         
BrechtDeMan@1071:         # PRINT name as section
BrechtDeMan@1071:         body+= '\section{'+file[:-4].capitalize()+'}\n' # make section header from name without extension
BrechtDeMan@1071:         
BrechtDeMan@1071:         # reset for new subject
BrechtDeMan@1071:         total_duration = 0
BrechtDeMan@1071:         page_number = 0
BrechtDeMan@1071:         
BrechtDeMan@1071:         individual_table = '' # table with stats for this individual test file
BrechtDeMan@1071:         
BrechtDeMan@1071:         # get list of all page names
BrechtDeMan@1071:         for audioholder in root.findall("./audioholder"):   # iterate over pages
BrechtDeMan@1071:             page_name = audioholder.get('id')               # get page name
BrechtDeMan@1071:             
BrechtDeMan@1071:             if page_name is None: # ignore 'empty' audio_holders
BrechtDeMan@1071:                 print "WARNING: " + file + " contains empty audio holder. (evaluation_stats.py)"
BrechtDeMan@1071:                 break # move on to next
BrechtDeMan@1071:             
BrechtDeMan@1071:             number_of_comments = 0 # for this page
BrechtDeMan@1071:             number_of_missing_comments = 0 # for this page
BrechtDeMan@1071:             not_played = 0 # for this page
BrechtDeMan@1071:             not_moved = 0 # for this page
BrechtDeMan@1071:             
BrechtDeMan@1071:             # 'testTime' keeps total duration: subtract time so far for duration of this audioholder
BrechtDeMan@1071:             duration = float(audioholder.find("./metric/metricresult[@id='testTime']").text) - total_duration
BrechtDeMan@1071:             
BrechtDeMan@1071:             # total duration of test
BrechtDeMan@1071:             total_duration += duration
BrechtDeMan@1071:             
BrechtDeMan@1071:             # number of audio elements
BrechtDeMan@1071:             audioelements = audioholder.findall("./audioelement") # get audioelements
BrechtDeMan@1071:             number_of_fragments += len(audioelements) # add length of this list to total
BrechtDeMan@1071:             
BrechtDeMan@1071:             # number of comments (interesting if comments not mandatory)
BrechtDeMan@1071:             for audioelement in audioelements:
BrechtDeMan@1071:                 response = audioelement.find("./comment/response")
BrechtDeMan@1071:                 was_played = audioelement.find("./metric/metricresult/[@name='elementFlagListenedTo']")
BrechtDeMan@1071:                 was_moved = audioelement.find("./metric/metricresult/[@name='elementFlagMoved']")
BrechtDeMan@1071:                 if response.text is not None and len(response.text) > 1: 
BrechtDeMan@1071:                     number_of_comments += 1
BrechtDeMan@1071:                 else: 
BrechtDeMan@1071:                     number_of_missing_comments += 1
BrechtDeMan@1071:                 if was_played is not None and was_played.text == 'false': 
BrechtDeMan@1071:                     not_played += 1
BrechtDeMan@1071:                 if was_moved is not None and was_moved.text == 'false': 
BrechtDeMan@1071:                     not_moved += 1
BrechtDeMan@1071:             
BrechtDeMan@1071:             # update global counters
BrechtDeMan@1071:             total_empty_comments += number_of_missing_comments
BrechtDeMan@1071:             total_not_played += not_played
BrechtDeMan@1071:             total_not_moved += not_moved
BrechtDeMan@1071:             
BrechtDeMan@1071:             # PRINT alerts when elements not played or markers not moved
BrechtDeMan@1071:             # number of audio elements not played
BrechtDeMan@1071:             if not_played > 1:
BrechtDeMan@1071:                 body += '\\emph{\\textbf{ATTENTION: '+str(not_played)+' fragments were not listened to in '+page_name+'!}} \\\\ \n'
BrechtDeMan@1071:             if not_played == 1: 
BrechtDeMan@1071:                 body += '\\emph{\\textbf{ATTENTION: one fragment was not listened to in '+page_name+'!}} \\\\ \n '
BrechtDeMan@1071:             
BrechtDeMan@1071:             # number of audio element markers not moved
BrechtDeMan@1071:             if not_moved > 1:
BrechtDeMan@1071:                 body += '\\emph{\\textbf{ATTENTION: '+str(not_moved)+' markers were not moved in '+page_name+'!}} \\\\ \n'
BrechtDeMan@1071:             if not_moved == 1: 
BrechtDeMan@1071:                 body += '\\emph{\\textbf{ATTENTION: one marker was not moved in '+page_name+'!}} \\\\ \n'
BrechtDeMan@1071:                 
BrechtDeMan@1071:             #TODO which one not moved/listened to? 
BrechtDeMan@1071:             
BrechtDeMan@1071:             # PRINT song-specific statistic
BrechtDeMan@1071:             individual_table += page_name+'&'+\
BrechtDeMan@1071:                                 str(number_of_comments) + '/' +\
BrechtDeMan@1071:                                 str(number_of_comments+number_of_missing_comments)+'&'+\
BrechtDeMan@1071:                                 seconds2timestr(duration)+'\\\\'
BrechtDeMan@1071:             
BrechtDeMan@1071:             # get timeline for this audioholder
BrechtDeMan@1071:             img_path = 'timelines_movement/'+file[:-4]+'-'+page_name+'.pdf'
BrechtDeMan@1071:             
BrechtDeMan@1071:             # check if available
BrechtDeMan@1071:             if os.path.isfile(folder_name+'/'+img_path):
BrechtDeMan@1071:                 # SHOW timeline image
BrechtDeMan@1071:                 body += r'''\begin{figure}[htbp]
BrechtDeMan@1071:                          \begin{center}
BrechtDeMan@1071:                          \includegraphics[width=\textwidth]{'''+\
BrechtDeMan@1071:                          folder_name+'/'+img_path+\
BrechtDeMan@1071:                         r'''}
BrechtDeMan@1071:                         \caption{Timeline of '''+\
BrechtDeMan@1071:                          page_name+' by '+ file[:-4].capitalize() +\
BrechtDeMan@1071:                         r'''.}
BrechtDeMan@1071:                          \end{center}
BrechtDeMan@1071:                          \end{figure}
BrechtDeMan@1071:                          '''
BrechtDeMan@1071:             
BrechtDeMan@1071:             # keep track of duration in function of page index
BrechtDeMan@1071:             if len(duration_order)>page_number:
BrechtDeMan@1071:                 duration_order[page_number].append(duration)
BrechtDeMan@1071:             else:
BrechtDeMan@1071:                 duration_order.append([duration])
BrechtDeMan@1071:                 
BrechtDeMan@1071:             # keep list of audioholder ids and count how many times each audioholder id
BrechtDeMan@1071:             # was tested, how long it took, and how many fragments there were (if number of 
BrechtDeMan@1071:             # fragments is different, store as different audioholder id)
BrechtDeMan@1071:             if page_name in page_names: 
BrechtDeMan@1071:                 page_index = page_names.index(page_name) # get index
BrechtDeMan@1071:                 # check if number of audioelements the same
BrechtDeMan@1071:                 if len(audioelements) == fragments_per_page[page_index]: 
BrechtDeMan@1071:                     page_count[page_index] += 1
BrechtDeMan@1071:                     duration_page[page_index].append(duration)
BrechtDeMan@1071:                 else: # make new entry
BrechtDeMan@1071:                     alt_page_name = page_name+"("+str(len(audioelements))+")"
BrechtDeMan@1071:                     if alt_page_name in page_names: # if already there
BrechtDeMan@1071:                         alt_page_index = page_names.index(alt_page_name) # get index
BrechtDeMan@1071:                         page_count[alt_page_index] += 1
BrechtDeMan@1071:                         duration_page[alt_page_index].append(duration)
BrechtDeMan@1071:                     else: 
BrechtDeMan@1071:                         page_names.append(alt_page_name)
BrechtDeMan@1071:                         page_count.append(1)
BrechtDeMan@1071:                         duration_page.append([duration])
BrechtDeMan@1071:                         fragments_per_page.append(len(audioelements))
BrechtDeMan@1071:             else: 
BrechtDeMan@1071:                 page_names.append(page_name)
BrechtDeMan@1071:                 page_count.append(1)
BrechtDeMan@1071:                 duration_page.append([duration])
BrechtDeMan@1071:                 fragments_per_page.append(len(audioelements))
BrechtDeMan@1071:                 
BrechtDeMan@1071:             # bookkeeping
BrechtDeMan@1071:             page_number += 1 # increase page count for this specific test
BrechtDeMan@1071:             number_of_pages += 1 # increase total number of pages
BrechtDeMan@1071:             time_per_page_accum += duration # total duration (for average time spent per page)
BrechtDeMan@1071: 
BrechtDeMan@1071:         # PRINT table with statistics about this test
BrechtDeMan@1071:         body += r'''\begin{tabular}{|p{3.5cm}|c|p{2.5cm}|}
BrechtDeMan@1071:                  \hline
BrechtDeMan@1071:                  \textbf{Song name} & \textbf{Comments} & \textbf{Duration} \\ \hline '''+\
BrechtDeMan@1071:                  individual_table+\
BrechtDeMan@1071:                  r'''\hline
BrechtDeMan@1071:                   \textbf{TOTAL} & & \textbf{'''+\
BrechtDeMan@1071:                   seconds2timestr(total_duration)+\
BrechtDeMan@1071:                  r'''}\\
BrechtDeMan@1071:                   \hline 
BrechtDeMan@1071:                   \end{tabular}'''
BrechtDeMan@1071: 
BrechtDeMan@1071: # join to footer
BrechtDeMan@1071: footer = body + footer
BrechtDeMan@1071: 
BrechtDeMan@1071: # empty body again
BrechtDeMan@1071: body = ''
BrechtDeMan@1071: 
BrechtDeMan@1071: # PRINT summary of everything (at start)
BrechtDeMan@1071: body += '\section{Summary}\n'
BrechtDeMan@1071: 
BrechtDeMan@1071: # PRINT table with statistics
BrechtDeMan@1071: body += '\\begin{tabular}{ll}'
BrechtDeMan@1071: body += r'Number of XML files: &' + str(number_of_XML_files) + r'\\'
BrechtDeMan@1071: body += r'Number of pages: &' + str(number_of_pages) + r'\\'
BrechtDeMan@1071: body += r'Number of fragments: &' + str(number_of_fragments) + r'\\'
BrechtDeMan@1071: body += r'Number of empty comments: &' + str(total_empty_comments) +\
BrechtDeMan@1071:       " (" + str(round(100.0*total_empty_comments/number_of_fragments,2)) + r"\%)\\"
BrechtDeMan@1071: body += r'Number of unplayed fragments: &' + str(total_not_played) +\
BrechtDeMan@1071:       " (" + str(round(100.0*total_not_played/number_of_fragments,2)) + r"\%)\\"
BrechtDeMan@1071: body += r'Number of unmoved markers: &' + str(total_not_moved) +\
BrechtDeMan@1071:       " (" + str(round(100.0*total_not_moved/number_of_fragments,2)) + r"\%)\\"
BrechtDeMan@1071: body += r'Average time per page: &' + seconds2timestr(time_per_page_accum/number_of_pages) + r"\\"
BrechtDeMan@1071: 
BrechtDeMan@1071: 
BrechtDeMan@1071: # Pages and number of times tested
BrechtDeMan@1071: page_count_strings = list(str(x) for x in page_count)
BrechtDeMan@1071: count_list = page_names + page_count_strings
BrechtDeMan@1071: count_list[::2] = page_names
BrechtDeMan@1071: count_list[1::2] = page_count_strings
BrechtDeMan@1071: #body +=  r'Pages tested: &' + str(count_list) + r"\\"
BrechtDeMan@1071: 
BrechtDeMan@1071: body += '\\end{tabular} \\vspace{1.5cm} \\\\ \n'
BrechtDeMan@1071: 
BrechtDeMan@1071: # Average duration for first, second, ... page
BrechtDeMan@1071: body += " \\vspace{.5cm} Average duration per page (see also Figure \\ref{fig:avgtimeperpage}): \\\\ \n"
BrechtDeMan@1071: body += r'''\begin{tabular}{lll}
BrechtDeMan@1071:         \textbf{Page} & \textbf{Duration} & \textbf{\# subjects}\\
BrechtDeMan@1071:         '''
BrechtDeMan@1071: tpp_averages = [] # store average time per page
BrechtDeMan@1071: for page_number in range(len(duration_order)): 
BrechtDeMan@1071:     body += str(page_number+1) + "&" +\
BrechtDeMan@1071:         seconds2timestr(sum(duration_order[page_number])/len(duration_order[page_number])) +\
BrechtDeMan@1071:             "&"+str(len(duration_order[page_number]))+r"\\"
BrechtDeMan@1071:     tpp_averages.append(sum(duration_order[page_number])/len(duration_order[page_number]))
BrechtDeMan@1071:             
BrechtDeMan@1071: body += '\\end{tabular} \\vspace{1.5cm} \\\\ \n'
BrechtDeMan@1071: 
BrechtDeMan@1071: # SHOW bar plot of average time per page
BrechtDeMan@1071: plt.bar(range(1,len(duration_order)+1), tpp_averages)
BrechtDeMan@1071: plt.xlabel('Page')
BrechtDeMan@1071: plt.xlim(.8, len(duration_order)+1)
BrechtDeMan@1071: plt.xticks(np.arange(1,len(duration_order)+1)+.4, range(1,len(duration_order)+1))
BrechtDeMan@1071: plt.ylabel('Time [seconds]')
BrechtDeMan@1071: plt.savefig(folder_name+"/time_per_page.pdf", bbox_inches='tight')
BrechtDeMan@1071: plt.close()
BrechtDeMan@1071: body += r'''\begin{figure}[htbp]
BrechtDeMan@1071:          \begin{center}
BrechtDeMan@1071:          \includegraphics[width=\textwidth]{'''+\
BrechtDeMan@1071:          folder_name+"/time_per_page.pdf"+\
BrechtDeMan@1071:         r'''}
BrechtDeMan@1071:         \caption{Average time spent per audioholder page.}
BrechtDeMan@1071:         \label{fig:avgtimeperpage}
BrechtDeMan@1071:          \end{center}
BrechtDeMan@1071:          \end{figure}
BrechtDeMan@1071:          '''
BrechtDeMan@1071: #TODO add error bars
BrechtDeMan@1071: 
BrechtDeMan@1071: 
BrechtDeMan@1071: # Sort pages by number of audioelements, then by duration
BrechtDeMan@1071: 
BrechtDeMan@1071: # average duration and number of subjects per page
BrechtDeMan@1071: average_duration_page = []
BrechtDeMan@1071: number_of_subjects_page = []
BrechtDeMan@1071: for line in duration_page:
BrechtDeMan@1071:     number_of_subjects_page.append(len(line))
BrechtDeMan@1071:     average_duration_page.append(sum(line)/len(line))
BrechtDeMan@1071: 
BrechtDeMan@1071: # combine and sort in function of number of audioelements and duration
BrechtDeMan@1071: combined_list = [page_names, average_duration_page, fragments_per_page, number_of_subjects_page]
BrechtDeMan@1071: combined_list = sorted(zip(*combined_list), key=operator.itemgetter(1, 2)) # sort
BrechtDeMan@1071: 
BrechtDeMan@1071: # Show average duration for all songs
BrechtDeMan@1071: body += r'''\vspace{.5cm} Average duration per audioholder: \\
BrechtDeMan@1071:         \begin{tabular}{llll}
BrechtDeMan@1071:         \textbf{Audioholder} & \textbf{Duration} & \textbf{\# subjects} & \textbf{\# fragments} \\
BrechtDeMan@1071:         '''
BrechtDeMan@1071: for page_index in range(len(page_names)):
BrechtDeMan@1071:     body +=  combined_list[page_index][0] + "&" +\
BrechtDeMan@1071:              seconds2timestr(combined_list[page_index][1]) + "&" +\
BrechtDeMan@1071:              str(combined_list[page_index][3]) + "&" +\
BrechtDeMan@1071:              str(combined_list[page_index][2]) + r"\\"
BrechtDeMan@1071: body += '\\end{tabular}\n'
BrechtDeMan@1071: 
BrechtDeMan@1071: #TODO
BrechtDeMan@1071: # time per page in function of number of fragments (plot)
BrechtDeMan@1071: # time per participant in function of number of pages
BrechtDeMan@1071: # plot total time for each participant
BrechtDeMan@1071: # plot total time
BrechtDeMan@1071: # show 'count' per page (in order)
BrechtDeMan@1071: 
BrechtDeMan@1071: # clear up page_index <> page_count <> page_number confusion
BrechtDeMan@1071: 
BrechtDeMan@1071: 
BrechtDeMan@1071: texfile = header+body+footer
BrechtDeMan@1071: 
BrechtDeMan@1071: # write TeX file
BrechtDeMan@1071: with open(folder_name + '/' + 'test.tex','w') as f:
BrechtDeMan@1071:     f.write(texfile)
BrechtDeMan@1071: proc=subprocess.Popen(shlex.split('pdflatex -output-directory='+folder_name+' '+ folder_name + '/test.tex'))
BrechtDeMan@1071: proc.communicate()
BrechtDeMan@1071: # run again
BrechtDeMan@1071: proc=subprocess.Popen(shlex.split('pdflatex -output-directory='+folder_name+' '+ folder_name + '/test.tex'))
BrechtDeMan@1071: proc.communicate()
BrechtDeMan@1071: 
BrechtDeMan@1071: #TODO remove auxiliary LaTeX files