b@2264: #!/usr/bin/python
b@2264: 
b@2264: import sys
b@2264: import os
b@2264: import csv
b@2264: import matplotlib.pyplot as plt
b@2264: import numpy as np
b@2264: import scipy as sp
b@2264: import scipy.stats
b@2264: 
b@2264: # COMMAND LINE ARGUMENTS
b@2264: 
b@2264: #TODO: Merge, implement this functionality
b@2264: #TODO: Control by CLI arguments (plot types, save and/or show, ...) 
b@2264: 
b@2264: assert len(sys.argv)<4, "score_plot takes at most 2 command line arguments\n"+\
b@2264:                         "Use: python score_plot.py [ratings_folder_location]."+\
b@2264:                         "Type 'python score_plot.py -h' for more options"
b@2264: 
b@2264: # initialise plot types (false by default) and options
b@2264: enable_boxplot    = False     # show box plot
b@2264: enable_confidence = False     # show confidence interval
nicholas@3107: enable_combined   = False     # show combined plots
b@2264: confidence        = 0.90      # confidence value (for confidence interval plot)
b@2264: enable_individual = False     # show all individual ratings
b@2264: show_individual   = []        # show specific individuals (empty: show all individuals found)
b@2264: show_legend       = False     # show names of individuals
b@2264: 
b@2264: # DEFAULT: Looks in 'saves/ratings/' folder from 'scripts/' folder
b@2264: rating_folder = "../saves/ratings/" 
b@2264: 
b@2264: # XML results files location
b@2264: if len(sys.argv) == 1: # no extra arguments
b@2264:     enable_boxplot    = True # show box plot
b@2272:     print("Use: python score_plot.py [rating folder] [plot_type] [-l/-legend]")
b@2272:     print("Type 'python score_plot.py -h' for help.")
b@2272:     print("Using default path: " + rating_folder + " with boxplot.")
b@2264: else:
b@2264:     for arg in sys.argv: # go over all arguments
b@2264:         if arg == '-h':
b@2264:             # show help
b@2264:             #TODO: replace with contents of helpfile score_plot.info (or similar)
b@2272:             print("Use: python score_plot.py [rating_folder] [plot_type] [-l] [confidence]")
b@2272:             print("   rating_folder:")
b@2272:             print("            folder where output of 'score_parser' can be found, and")
b@2272:             print("            where plots will be stored.")
b@2272:             print("            By default, '../saves/ratings/' is used.")
b@2272:             print("")
b@2272:             print("PLOT TYPES")
b@2272:             print(" Can be used in combination.")
b@2272:             print("    box | boxplot | -b")
b@2272:             print("            Enables the boxplot" )
b@2272:             print("    conf | confidence | -c")
b@2272:             print("            Enables the confidence interval plot" )
b@2272:             print("    ind | individual | -i")
b@2272:             print("            Enables plot of individual ratings" )
b@2272:             print("")
b@2272:             print("PLOT OPTIONS")
b@2272:             print("    leg | legend | -l")
b@2272:             print("            For individual plot: show legend with individual file names")
b@2272:             print("    numeric value between 0 and 1, e.g. 0.95")
b@2272:             print("            For confidence interval plot: confidence value")
b@2264:             assert False, ""# stop immediately after showing help #TODO cleaner way
b@2264:             
b@2264:         # PLOT TYPES
b@2264:         elif arg == 'box' or arg == 'boxplot' or arg == '-b':
b@2264:             enable_boxplot    = True     # show box plot
b@2264:         elif arg == 'conf' or arg == 'confidence' or arg == '-c':
b@2264:             enable_confidence = True     # show confidence interval
b@2264:             #TODO add confidence value input
b@2264:         elif arg == 'ind' or arg == 'individual' or arg == '-i':
b@2264:             enable_individual = True     # show all individual ratings
nicholas@3107:         elif arg == 'comb' or arg == 'combined' or arg == '-m':
nicholas@3107:             enable_combined   = True     # show combined plot with error bars  
b@2264:         # PLOT OPTIONS
b@2264:         elif arg == 'leg' or arg == 'legend' or arg == '-l':
b@2264:             if not enable_individual: 
b@2272:                 print("WARNING: The 'legend' option is only relevant to plots of "+\
b@2272:                       "individual ratings")
b@2264:             show_legend = True     # show all individual ratings
b@2264:         elif arg.isdigit():
b@2264:             if not enable_confidence: 
b@2272:                 print("WARNING: The numeric confidence value is only relevant when "+\
b@2272:                       "confidence plot is enabled")
b@2264:             if float(arg)>0 and float(arg)<1:
b@2264:                 confidence = float(arg)
b@2264:             else: 
b@2272:                 print("WARNING: The confidence value needs to be between 0 and 1")
b@2264:         
b@2264:         # FOLDER NAME
b@2264:         else: 
b@2264:              # assume it's the folder name
b@2264:              rating_folder = arg
b@2264: 
b@2264: # at least one plot type should be selected: box plot by default
nicholas@3107: if not enable_boxplot and not enable_confidence and not enable_individual and not enable_combined:
nicholas@3107:     print("Default to enable boxplot")
b@2264:     enable_boxplot = True
b@2264: 
b@2264: # check if folder_name exists
b@2264: if not os.path.exists(rating_folder):
b@2264:     #the file is not there
b@2272:     print("Folder '"+rating_folder+"' does not exist.")
b@2264:     sys.exit() # terminate script execution
b@2264: elif not os.access(os.path.dirname(rating_folder), os.W_OK):
b@2264:     #the file does exist but write rating_folder are not given
b@2272:     print("No write privileges in folder '"+rating_folder+"'.")
b@2264: 
b@2264: 
b@2264: # CONFIGURATION
b@2264: 
b@2264: # Font settings
b@2264: font = {'weight' : 'bold',
b@2264:         'size'   : 10}
b@2264: plt.rc('font', **font)
b@2264: 
b@2264: 
b@2264: # CODE
nicholas@3107: combined = {}
b@2264: 
b@2264: # get every csv file in folder
b@2264: for file in os.listdir(rating_folder):
b@2264:     if file.endswith(".csv"):
b@2264:         page_name = file[:-4] # file name (without extension) is page ID
b@2264: 
b@2272:         # get header (as text)
b@2272:         with open(rating_folder+file, 'rt') as readfile: # read this csv file
b@2272:             filereader = csv.reader(readfile, delimiter=',')
b@2272:             headerrow = next(filereader) # use headerrow as X-axis
b@2272:             headerrow = headerrow[1:]
b@2272: 
b@2272:         # read ratings into matrix (as bytes)
b@2264:         with open(rating_folder+file, 'rb') as readfile: # read this csv file
b@2264:             filereader = csv.reader(readfile, delimiter=',')
b@2264:             ratings = np.genfromtxt(readfile,
b@2264:                                    delimiter=",",
b@2272:                                    skip_header = 1,
b@2264:                                    converters = {3: lambda s: float(s or 'Nan')},
b@2272:                                    usecols=list(range(1,len(headerrow)+1))
b@2264:                                    )
b@2264:         
b@2264:         # assert at least 2 subjects (move on to next file if violated)
b@2264:         if ratings.shape[0]<2:
b@2272:             print("WARNING: Just one subject for " + page_name + ". Moving on to next file.")
b@2264:             break
b@2264: 
b@2489:         if len(ratings.shape) <= 1: # if only single subject
b@2489:             ratings = [[r] for r in ratings] # turn into array of arrays
b@2489: 
b@2264:         # BOXPLOT
b@2264:         if enable_boxplot:
b@2264:             plt.boxplot(ratings)
b@2264:             
b@2264:         # CONFIDENCE INTERVAL
b@2264:         if enable_confidence:
b@2264:             iterator = 0
b@2264:             for column in ratings.T: # iterate over transposed matrix
b@2264:                 # remove all 'Nan's from column
b@2264:                 column = column[~np.isnan(column)]
b@2264:             
b@2264:                 # get number of non-Nan ratings (= #subjects)
b@2264:                 n = column.size
b@2264:         
b@2264:                 # get mean
b@2264:                 mean_rating = np.mean(column)
b@2264:         
b@2264:                 # get errors
b@2264:                 err = scipy.stats.sem(column)* sp.stats.t._ppf((1+confidence)/2., n-1)
b@2264:         
b@2264:                 # draw plot
b@2264:                 plt.errorbar(iterator+1, 
b@2264:                             mean_rating, 
b@2264:                             yerr=err,
b@2264:                             marker="x",
b@2264:                             color ="k",
b@2264:                             markersize=12,
b@2264:                             linestyle='None')
b@2264:                             
b@2264:                 iterator += 1 # increase counter
b@2264:     
b@2264:     
b@2264:         # INDIVIDUAL PLOT
b@2264:         if enable_individual or show_individual:
b@2264:             # marker list and color map to cycle through
b@2264:             markerlist = ["x", ".", "o", "*", "+", "v", ">", "<", "8", "s", "p"]
b@2264:             colormap = ['b', 'r', 'g', 'c', 'm', 'y', 'k'] 
b@2264:             increment = 0
b@2264:             linehandles = []
b@2264:             legendnames = []
b@2272:             with open(rating_folder+file, 'r') as readfile: # read this csv file
b@2264:                 filereader = csv.reader(readfile, delimiter=',')
b@2272:                 headerrow = next(filereader) # use headerrow as X-axis
b@2264:                 headerrow = headerrow[1:]
b@2264:                 for row in filereader:
b@2264:                     subject_id = row[0][:-4] # read from beginning of line
b@2264:                     # assume plotting all individuals if no individual(s) specified
b@2264:                     if not show_individual or subject_id in show_individual:
b@2264:                         plothandle, = plt.plot(range(1,len(row)), # x-values
b@2264:                                  ratings[increment,:],#row[1:], # y-values: csv values except subject name
b@2264:                                  color=colormap[increment%len(colormap)],
b@2264:                                  marker=markerlist[increment%len(markerlist)],
b@2264:                                  markersize=10,
b@2264:                                  linestyle='None',
b@2264:                                  label=subject_id
b@2264:                                 )
b@2264:                         linehandles.append(plothandle)
b@2264:                         legendnames.append(subject_id)
b@2264:                         if show_legend:
b@2264:                             plt.legend(linehandles, legendnames,
b@2264:                                        loc='upper right',
b@2264:                                        bbox_to_anchor=(1.1, 1),
b@2264:                                        borderaxespad=0.,
b@2264:                                        numpoints=1 # remove extra marker
b@2264:                                        )
b@2264:                     increment += 1 # increase counter
nicholas@3107:         if enable_combined:
nicholas@3107:             print(page_name)
nicholas@3107:             combined[page_name] = {"labels": headerrow, "r": ratings}
b@2264: 
nicholas@3107:         if enable_boxplot or enable_confidence or enable_individual:
nicholas@3107:             # TITLE, AXIS LABELS AND LIMITS
nicholas@3107:             plt.title(page_name)
nicholas@3107:             plt.xlabel('Fragment')
nicholas@3107:             plt.xlim(0, len(headerrow)+1) # only show relevant region, leave space left & right)
nicholas@3107:             plt.xticks(range(1, len(headerrow)+1), headerrow, rotation=90) # show fragment names
nicholas@3107:             plt.ylabel('Rating')
nicholas@3107:             plt.ylim(0,1)
nicholas@3107:             # SHOW PLOT
nicholas@3107:             #plt.show()
nicholas@3107:             #exit()
b@2264:     
nicholas@3107:             # SAVE PLOT
nicholas@3107:             # automatically 
nicholas@3107:             plot_type = ("-box" if enable_boxplot else "") + \
b@2264:                     ("-conf" if enable_confidence else "") + \
b@2264:                     ("-ind" if enable_individual else "")
nicholas@3107:             plt.savefig(rating_folder+page_name+plot_type+".pdf", bbox_inches='tight')
nicholas@3107:             plt.close()
nicholas@3107: 
nicholas@3107: if enable_combined:
nicholas@3107:     plt.figure()
nicholas@3107:     pages = combined.keys()
nicholas@3107:     numcombined = len(pages)
nicholas@3107:     spacing = 1.0/float(numcombined+2)
nicholas@3107:     for i in range(0,numcombined):
nicholas@3107:         page_name = pages[i]
nicholas@3107:         N = len(combined[page_name]['labels'])
nicholas@3107:         mean = np.percentile(combined[page_name]['r'], 50, 0)
nicholas@3107:         p25 = np.percentile(combined[page_name]['r'], 25, 0)
nicholas@3107:         p75 = np.percentile(combined[page_name]['r'], 75, 0)
nicholas@3107:         yerr = [mean-p25, p75-mean]
nicholas@3107:         print(yerr)
nicholas@3107:         plt.errorbar(np.arange(0,N)+(spacing*(i+1)), combined[page_name]['r'].mean(0), yerr=yerr, fmt='x', elinewidth=0.5)
nicholas@3107:     ax = plt.gca()
nicholas@3107:     ax.grid(which='major', axis='x', linewidth=2, color='k')
nicholas@3107:     plt.show()