#!/usr/bin/python

import sys
import os

def main(argv):
    if len(argv) < 2 or not os.access(argv[1], os.R_OK):
        usage()

    # Read in the data
    data = [[conv(x) for x in line.split(",")] for line in open(sys.argv[1])]

    # Get the grade column
    grades = [row[0] for row in data]

    # Now, for each column (other than the first), 
    for i in range(1, max([len(x) for x in data])):
        col = []
        for row in data:
            if len(row) > i:
                col.append(row[i])
            else:
                col.append(None)

        # Process this column against the overall grades
        diff, disc = evalScores(grades, col)

        if type(col[0]) == type(""):
            print col[0],
        else:
            print "Q%03d" % (i,),
        print "Difficulty = %.2f\tDiscrimination = %.2f" % (diff, disc)

def usage():
    msg = """Usage: %s FILE
    FILE is CSV formatted, column 1 is ability estimate (course grade),
    other columns are grades in range [0 .. 1] for each student (row),
    on each question (column)

    Row #1 may be headers to aid in identifying results.""" % sys.argv[0]
    sys.exit(msg)

def conv(x):
    # If we can make a float out of this, do so.
    # If not, and it isn't the empty string, return the string.
    # If it was the empty string, then return None
    try:
        return float(x)
    except:
        if x.strip():
            return x
        return None

def evalDisc(split, data):
    # For everything <= grade, error is difference from 0.
    # For everything else, error is difference from 1.

    error = 0.0
    for grade, score in data:
        if grade <= split:
            error += score
        else:
            error += (1 - score)

    return 1 - error / len(data)

def evalScores(grades, col):
    # We want to find all the split points in the scores in col
    # For each split, find the discrimination
    # Take the max

    # First, build the actual data we will use here:
    data = [(grades[i], col[i]) for i in range(len(grades)) \
            if type(grades[i]) == type(col[i]) == type(0.0)]

    # Now we sort that, which will sort by the first element of the pairs
    data.sort()

    # Now loop through to find the best split
    bestDisc = 0
    bestDiff = 0
    for grade, score in data:
        tempDisc = evalDisc(grade, data)

        if tempDisc > bestDisc:
            bestDisc = tempDisc
            bestDiff = grade

    # Return the best one we found
    return bestDiff, bestDisc

if __name__ == "__main__":
    main(sys.argv)