# First, we organize the data you've collected into a series of lists
young_data = [cigyoungnone, cigyoungfew, cigyoungmid, cigyoungmany]
teen_data  = [cigteennone,  cigteenfew,  cigteenmid,  cigteenmany ]
adult_data = [cigadultnone, cigadultfew, cigadultmid, cigadultmany]

# We make a list for each kind of category that you have
age_categories = ["1-14","15-17","18+"]
frequency_categories = ["None", "Few", "Mid", "Many"]

# I've shown you "xrange", which is used to build up a list of numbers from a
#   length:
#       > xrange(5)
#       [0,1,2,3,4]
# The NumpPy library (which you have never seen in this class) has a similar
#   thing called "arange", which also builds up a list but with extra abilities:
#       > arange(5) + .5
#       [0.5, 1.5, 2.5, 3.5, 4.5]
#       > arange(5) + .25
#       [0.25, 1.25, 2.25, 3.25, 4.25]
from numpy import arange
indexes = arange(len(frequency_categories))
width = 0.25

# Why does this matter? Because of the way that MatPlotLib places bars.
#   Essentially, it's stupid and needs to be told where you want to place each
#   bar. Remember in Algebra how your graph is a grid, and your x-axis is a
#   line of numbers stretching from 0 to infinity, with each tick being 1?
#       https://www.mathsisfun.com/data/images/x-y-axes.gif
# Well, that's what is happening here. With only one series of bars, you tell
#   Matplotlib to put them at the 0, 1, 2, ... on the x-axis, and each bar
#   defaults to a width of 1 unit.
# When you want to plot multiple, you must horizontally shrink the width of
#   these bars to be less than 1 (0.25 is a good choice when you have three
#   bars, so that there's a bit of empty space on the right) and space them out
#   so that the first series is at [0, 1, 2, ...], the second series is at
#   [0.25, 1.25, 2.25, ...], and so on. In other words, you are offsetting
#   each list of positions by a few decimal places.
# So below, you will notice that each time we call plt.bar, we pass in the
#   "indexes" and add an offset using "width" (which is the reason for using
#   the special "arange" function instead of "xrange". Then we pass in the
#   actual data to plot, the width of the bars, and the color of that bar.
# How did I know how to do all this? I stole it mostly from the documentation!
#   http://matplotlib.org/examples/api/barchart_demo.html
young_plot = plt.bar(indexes,         young_data, width, color='r')
teen_plot  = plt.bar(indexes+width,   teen_data,  width, color='y')
adult_plot = plt.bar(indexes+2*width, adult_data, width, color='b')

# We offset the labels on each x tick so that they are under the center of the
#   of the bars (thse are the frequencies, e.g., "None", "Many"). Remember,
#   these x ticks are the units of the number line!
plt.xticks(indexes+0.5, frequency_categories)

# The labels you know and love
plt.xlabel('Frequency')
plt.ylabel('# of respondents')
plt.title('Cigarettes')

# For clarity, you should add a legend. This looks weirder than it is.
# The first argument are the plots generated by using plt.bar (which actually
#   returns multiple values like your new cigyoung_over_days function,
#   except we only care about the first value from that.
# The second argument is the labels for each of those plots.
# Then we also specify a title for the legend.
plt.legend( (young_plot[0], teen_plot[0], adult_plot[0]),
            age_categories, title='Age at first use' )

# Show and we're done!
plt.show()