#!/usr/bin/env python
'''
Use matplotlib to generate performance charts
Copyright 2011 Joel Martin
Licensed under GPL version 3 (see docs/LICENSE.GPL-3)
'''
# a bar plot with errorbars
import sys, json, pprint
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties
def usage():
print "%s json_file level1 level2 level3 [legend_height]\n\n" % sys.argv[0]
print "Description:\n"
print "level1, level2, and level3 are one each of the following:\n";
print " select=ITEM - select only ITEM at this level";
print " bar - each item on this level becomes a graph bar";
print " group - items on this level become groups of bars";
print "\n";
print "json_file is a file containing json data in the following format:\n"
print ' {';
print ' "conf": {';
print ' "order_l1": [';
print ' "level1_label1",';
print ' "level1_label2",';
print ' ...';
print ' ],';
print ' "order_l2": [';
print ' "level2_label1",';
print ' "level2_label2",';
print ' ...';
print ' ],';
print ' "order_l3": [';
print ' "level3_label1",';
print ' "level3_label2",';
print ' ...';
print ' ]';
print ' },';
print ' "stats": {';
print ' "level1_label1": {';
print ' "level2_label1": {';
print ' "level3_label1": [val1, val2, val3],';
print ' "level3_label2": [val1, val2, val3],';
print ' ...';
print ' },';
print ' "level2_label2": {';
print ' ...';
print ' },';
print ' },';
print ' "level1_label2": {';
print ' ...';
print ' },';
print ' ...';
print ' },';
print ' }';
sys.exit(2)
def error(msg):
print msg
sys.exit(1)
#colors = ['#ff0000', '#0863e9', '#00f200', '#ffa100',
# '#800000', '#805100', '#013075', '#007900']
colors = ['#ff0000', '#00ff00', '#0000ff',
'#dddd00', '#dd00dd', '#00dddd',
'#dd6622', '#dd2266', '#66dd22',
'#8844dd', '#44dd88', '#4488dd']
if len(sys.argv) < 5:
usage()
filename = sys.argv[1]
L1 = sys.argv[2]
L2 = sys.argv[3]
L3 = sys.argv[4]
if len(sys.argv) > 5:
legendHeight = float(sys.argv[5])
else:
legendHeight = 0.75
# Load the JSON data from the file
data = json.loads(file(filename).read())
conf = data['conf']
stats = data['stats']
# Sanity check data hierarchy
if len(conf['order_l1']) != len(stats.keys()):
error("conf.order_l1 does not match stats level 1")
for l1 in stats.keys():
if len(conf['order_l2']) != len(stats[l1].keys()):
error("conf.order_l2 does not match stats level 2 for %s" % l1)
if conf['order_l1'].count(l1) < 1:
error("%s not found in conf.order_l1" % l1)
for l2 in stats[l1].keys():
if len(conf['order_l3']) != len(stats[l1][l2].keys()):
error("conf.order_l3 does not match stats level 3")
if conf['order_l2'].count(l2) < 1:
error("%s not found in conf.order_l2" % l2)
for l3 in stats[l1][l2].keys():
if conf['order_l3'].count(l3) < 1:
error("%s not found in conf.order_l3" % l3)
#
# Generate the data based on the level specifications
#
bar_labels = None
group_labels = None
bar_vals = []
bar_sdvs = []
if L3.startswith("select="):
select_label = l3 = L3.split("=")[1]
bar_labels = conf['order_l1']
group_labels = conf['order_l2']
bar_vals = [[0]*len(group_labels) for i in bar_labels]
bar_sdvs = [[0]*len(group_labels) for i in bar_labels]
for b in range(len(bar_labels)):
l1 = bar_labels[b]
for g in range(len(group_labels)):
l2 = group_labels[g]
bar_vals[b][g] = np.mean(stats[l1][l2][l3])
bar_sdvs[b][g] = np.std(stats[l1][l2][l3])
elif L2.startswith("select="):
select_label = l2 = L2.split("=")[1]
bar_labels = conf['order_l1']
group_labels = conf['order_l3']
bar_vals = [[0]*len(group_labels) for i in bar_labels]
bar_sdvs = [[0]*len(group_labels) for i in bar_labels]
for b in range(len(bar_labels)):
l1 = bar_labels[b]
for g in range(len(group_labels)):
l3 = group_labels[g]
bar_vals[b][g] = np.mean(stats[l1][l2][l3])
bar_sdvs[b][g] = np.std(stats[l1][l2][l3])
elif L1.startswith("select="):
select_label = l1 = L1.split("=")[1]
bar_labels = conf['order_l2']
group_labels = conf['order_l3']
bar_vals = [[0]*len(group_labels) for i in bar_labels]
bar_sdvs = [[0]*len(group_labels) for i in bar_labels]
for b in range(len(bar_labels)):
l2 = bar_labels[b]
for g in range(len(group_labels)):
l3 = group_labels[g]
bar_vals[b][g] = np.mean(stats[l1][l2][l3])
bar_sdvs[b][g] = np.std(stats[l1][l2][l3])
else:
usage()
# If group is before bar then flip (zip) the data
if [L1, L2, L3].index("group") < [L1, L2, L3].index("bar"):
bar_labels, group_labels = group_labels, bar_labels
bar_vals = zip(*bar_vals)
bar_sdvs = zip(*bar_sdvs)
print "bar_vals:", bar_vals
#
# Now render the bar graph
#
ind = np.arange(len(group_labels)) # the x locations for the groups
width = 0.8 * (1.0/len(bar_labels)) # the width of the bars
fig = plt.figure(figsize=(10,6), dpi=80)
plot = fig.add_subplot(1, 1, 1)
rects = []
for i in range(len(bar_vals)):
rects.append(plot.bar(ind+width*i, bar_vals[i], width, color=colors[i],
yerr=bar_sdvs[i], align='center'))
# add some
plot.set_ylabel('Milliseconds (less is better)')
plot.set_title("Javascript array test: %s" % select_label)
plot.set_xticks(ind+width)
plot.set_xticklabels( group_labels )
fontP = FontProperties()
fontP.set_size('small')
plot.legend( [r[0] for r in rects], bar_labels, prop=fontP,
loc = 'center right', bbox_to_anchor = (1.0, legendHeight))
def autolabel(rects):
# attach some text labels
for rect in rects:
height = rect.get_height()
if np.isnan(height):
height = 0.0
plot.text(rect.get_x()+rect.get_width()/2., height+20, '%d'%int(height),
ha='center', va='bottom', size='7')
for rect in rects:
autolabel(rect)
# Adjust axis sizes
axis = list(plot.axis())
axis[0] = -width # Make sure left side has enough for bar
#axis[1] = axis[1] * 1.20 # Add 20% to the right to make sure it fits
axis[2] = 0 # Make y-axis start at 0
axis[3] = axis[3] * 1.10 # Add 10% to the top
plot.axis(axis)
plt.show()