me@1942: #!/usr/bin/python
me@1942: 
me@1942: import sys
me@1942: import os
me@1942: import csv
me@1942: import matplotlib.pyplot as plt
me@1942: import numpy as np
me@1942: import scipy as sp
me@1942: import scipy.stats
me@1942: 
me@1942: # COMMAND LINE ARGUMENTS
me@1942: 
me@1942: #TODO: Merge, implement this functionality
me@1942: #TODO: Control by CLI arguments (plot types, save and/or show, ...) 
me@1942: 
me@1942: assert len(sys.argv)<4, "score_plot takes at most 2 command line arguments\n"+\
me@1942:                         "Use: python score_plot.py [ratings_folder_location]."+\
me@1942:                         "Type 'python score_plot.py -h' for more options"
me@1942: 
me@1942: # initialise plot types (false by default) and options
me@1942: enable_boxplot    = False     # show box plot
me@1942: enable_confidence = False     # show confidence interval
me@1942: confidence        = 0.90      # confidence value (for confidence interval plot)
me@1942: enable_individual = False     # show all individual ratings
me@1942: show_individual   = []        # show specific individuals (empty: show all individuals found)
me@1942: show_legend       = False     # show names of individuals
me@1942: 
me@1942: # DEFAULT: Looks in 'saves/ratings/' folder from 'scripts/' folder
me@1942: rating_folder = "../saves/ratings/" 
me@1942: 
me@1942: # XML results files location
me@1942: if len(sys.argv) == 1: # no extra arguments
me@1942:     enable_boxplot    = True # show box plot
me@1942:     print "Use: python score_plot.py [rating folder] [plot_type] [-l/-legend]"
me@1942:     print "Type 'python score_plot.py -h' for help."
me@1942:     print "Using default path: " + rating_folder + " with boxplot."
me@1942: else:
me@1942:     for arg in sys.argv: # go over all arguments
me@1942:         if arg == '-h':
me@1942:             # show help
me@1942:             #TODO: replace with contents of helpfile score_plot.info (or similar)
me@1942:             print "Use: python score_plot.py [rating_folder] [plot_type] [-l] [confidence]"
me@1942:             print "   rating_folder:"
me@1942:             print "            folder where output of 'score_parser' can be found, and"
me@1942:             print "            where plots will be stored."
me@1942:             print "            By default, '../saves/ratings/' is used."
me@1942:             print ""
me@1942:             print "PLOT TYPES"
me@1942:             print " Can be used in combination."
me@1942:             print "    box | boxplot | -b"
me@1942:             print "            Enables the boxplot" 
me@1942:             print "    conf | confidence | -c"
me@1942:             print "            Enables the confidence interval plot" 
me@1942:             print "    ind | individual | -i"
me@1942:             print "            Enables plot of individual ratings" 
me@1942:             print ""
me@1942:             print "PLOT OPTIONS"
me@1942:             print "    leg | legend | -l"
me@1942:             print "            For individual plot: show legend with individual file names"
me@1942:             print "    numeric value between 0 and 1, e.g. 0.95"
me@1942:             print "            For confidence interval plot: confidence value"
me@1942:             assert False, ""# stop immediately after showing help #TODO cleaner way
me@1942:             
me@1942:         # PLOT TYPES
me@1942:         elif arg == 'box' or arg == 'boxplot' or arg == '-b':
me@1942:             enable_boxplot    = True     # show box plot
me@1942:         elif arg == 'conf' or arg == 'confidence' or arg == '-c':
me@1942:             enable_confidence = True     # show confidence interval
me@1942:             #TODO add confidence value input
me@1942:         elif arg == 'ind' or arg == 'individual' or arg == '-i':
me@1942:             enable_individual = True     # show all individual ratings
me@1942:             
me@1942:         # PLOT OPTIONS
me@1942:         elif arg == 'leg' or arg == 'legend' or arg == '-l':
me@1942:             if not enable_individual: 
me@1942:                 print "WARNING: The 'legend' option is only relevant to plots of "+\
me@1942:                       "individual ratings"
me@1942:             show_legend = True     # show all individual ratings
me@1942:         elif arg.isdigit():
me@1942:             if not enable_confidence: 
me@1942:                 print "WARNING: The numeric confidence value is only relevant when "+\
me@1942:                       "confidence plot is enabled"
me@1942:             if float(arg)>0 and float(arg)<1:
me@1942:                 confidence = float(arg)
me@1942:             else: 
me@1942:                 print "WARNING: The confidence value needs to be between 0 and 1"
me@1942:         
me@1942:         # FOLDER NAME
me@1942:         else: 
me@1942:              # assume it's the folder name
me@1942:              rating_folder = arg
me@1942: 
me@1942: # at least one plot type should be selected: box plot by default
me@1942: if not enable_boxplot and not enable_confidence and not enable_individual:
me@1942:     enable_boxplot = True
me@1942: 
me@1942: # check if folder_name exists
me@1942: if not os.path.exists(rating_folder):
me@1942:     #the file is not there
me@1942:     print "Folder '"+rating_folder+"' does not exist."
me@1942:     sys.exit() # terminate script execution
me@1942: elif not os.access(os.path.dirname(rating_folder), os.W_OK):
me@1942:     #the file does exist but write rating_folder are not given
me@1942:     print "No write privileges in folder '"+rating_folder+"'."
me@1942: 
me@1942: 
me@1942: # CONFIGURATION
me@1942: 
me@1942: # Font settings
me@1942: font = {'weight' : 'bold',
me@1942:         'size'   : 10}
me@1942: plt.rc('font', **font)
me@1942: 
me@1942: 
me@1942: # CODE
me@1942: 
me@1942: # get every csv file in folder
me@1942: for file in os.listdir(rating_folder): # You have to put this in folder where rating csv files are.
me@1942:     if file.endswith(".csv"):
me@1942:         page_name = file[:-4] # file name (without extension) is page ID
me@1942: 
me@1942:         # get header
me@1942:         with open(rating_folder+file, 'rb') as readfile: # read this csv file
me@1942:             filereader = csv.reader(readfile, delimiter=',')
me@1942:             headerrow = filereader.next() # use headerrow as X-axis
me@1942:             headerrow = headerrow[1:]
me@1942: 
me@1942:         # read ratings into matrix
me@1942: #         ratings = np.loadtxt(open(rating_folder+file,"rb"),
me@1942: #                             delimiter=",",
me@1942: #                             skiprows=1,
me@1942: #                             usecols=range(1,len(headerrow)+1)
me@1942: #                             )
me@1942:             ratings = np.genfromtxt(readfile,
me@1942:                                    delimiter=",",
me@1942:                                    #skip_header = 1,
me@1942:                                    converters = {3: lambda s: float(s or 'Nan')},
me@1942:                                    usecols=range(1,len(headerrow)+1)
me@1942:                                    )
me@1942:         
me@1942:         # assert at least 2 subjects (move on to next file if violated)
me@1942:         if ratings.shape[0]<2:
me@1942:             print "WARNING: Just one subject for " + page_name + ". Moving on to next file."
me@1942:             break
me@1942: 
me@1942:         # BOXPLOT
me@1942:         if enable_boxplot:
me@1942:             plt.boxplot(ratings)
me@1942:             
me@1942:         # CONFIDENCE INTERVAL
me@1942:         if enable_confidence:
me@1942:             iterator = 0
me@1942:             for column in ratings.T: # iterate over transposed matrix
me@1942:                 # remove all 'Nan's from column
me@1942:                 column = column[~np.isnan(column)]
me@1942:             
me@1942:                 # get number of non-Nan ratings (= #subjects)
me@1942:                 n = column.size
me@1942:         
me@1942:                 # get mean
me@1942:                 mean_rating = np.mean(column)
me@1942:         
me@1942:                 # get errors
me@1942:                 err = scipy.stats.sem(column)* sp.stats.t._ppf((1+confidence)/2., n-1)
me@1942:         
me@1942:                 # draw plot
me@1942:                 plt.errorbar(iterator+1, 
me@1942:                             mean_rating, 
me@1942:                             yerr=err,
me@1942:                             marker="x",
me@1942:                             color ="k",
me@1942:                             markersize=12,
me@1942:                             linestyle='None')
me@1942:                             
me@1942:                 iterator += 1 # increase counter
me@1942:     
me@1942:     
me@1942:         # INDIVIDUAL PLOT
me@1942:         if enable_individual or show_individual:
me@1942:             # marker list and color map to cycle through
me@1942:             markerlist = ["x", ".", "o", "*", "+", "v", ">", "<", "8", "s", "p"]
me@1942:             colormap = ['b', 'r', 'g', 'c', 'm', 'y', 'k'] 
me@1942:             increment = 0
me@1942:             linehandles = []
me@1942:             legendnames = []
me@1942:             with open(rating_folder+file, 'rb') as readfile: # read this csv file
me@1942:                 filereader = csv.reader(readfile, delimiter=',')
me@1942:                 headerrow = filereader.next() # use headerrow as X-axis
me@1942:                 headerrow = headerrow[1:]
me@1942:                 for row in filereader:
me@1942:                     subject_id = row[0][:-4] # read from beginning of line
me@1942:                     # assume plotting all individuals if no individual(s) specified
me@1942:                     if not show_individual or subject_id in show_individual:
me@1942:                         plothandle, = plt.plot(range(1,len(row)), # x-values
me@1942:                                  ratings[increment,:],#row[1:], # y-values: csv values except subject name
me@1942:                                  color=colormap[increment%len(colormap)],
me@1942:                                  marker=markerlist[increment%len(markerlist)],
me@1942:                                  markersize=10,
me@1942:                                  linestyle='None',
me@1942:                                  label=subject_id
me@1942:                                 )
me@1942:                         linehandles.append(plothandle)
me@1942:                         legendnames.append(subject_id)
me@1942:                         if show_legend:
me@1942:                             plt.legend(linehandles, legendnames,
me@1942:                                        loc='upper right',
me@1942:                                        bbox_to_anchor=(1.1, 1),
me@1942:                                        borderaxespad=0.,
me@1942:                                        numpoints=1 # remove extra marker
me@1942:                                        )
me@1942:                     increment += 1 # increase counter
me@1942: 
me@1942:         # TITLE, AXIS LABELS AND LIMITS
me@1942:         plt.title(page_name)
me@1942:         plt.xlabel('Fragment')
me@1942:         plt.xlim(0, len(headerrow)+1) # only show relevant region, leave space left & right)
me@1942:         plt.xticks(range(1, len(headerrow)+1), headerrow, rotation=90) # show fragment names
me@1942:         plt.ylabel('Rating')
me@1942:         plt.ylim(0,1)
me@1942:         
me@1942:         
me@1942: 
me@1942:         # SHOW PLOT
me@1942:         #plt.show()
me@1942:         #exit()
me@1942:     
me@1942:         # SAVE PLOT
me@1942:         # automatically 
me@1942:         plot_type = ("-box" if enable_boxplot else "") + \
me@1942:                     ("-conf" if enable_confidence else "") + \
me@1942:                     ("-ind" if enable_individual else "")
me@1942:         plt.savefig(rating_folder+page_name+plot_type+".pdf", bbox_inches='tight')
me@1942:         plt.close()