本文介绍如何用matplotlib将The ONE一组MessageStatsReport仿真结果可视化。

1. gnuplot vs matplotlib

最开始我是用gnuplot，直到我做一个简单图例（legend，字符串数组的变量含有空格，titles = "'With XOR' 'Without XOR'"），花去了我两个多小时（详情见如何让gnuplot字符串数组中的变量含有空格）。

在使用gnuplot过程中，了解到Python有个matplotlib库是用来绘图的，加之，那会对Python越来越熟悉，所以决定转向matplotlib。转向matplotlib，不得不说让我节省了很多时间，在此期间，也做了一些笔记。

• Draw with Matplotlib: Table of Contents

给我感觉，matplotlib作出来的图不是很炫，或者说要得到很炫的图需要花很多时间，不过对于学术文章，我想是够了。

2. MessageStatsReport仿真结果可视化

假设我得到了一组仿真结果（详情见The ONE使用笔记：批量仿真），如下（MessageBroadcastStatsReport是我自己定义的Report，跟MessageStatsReport类似）：

4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 01:52 WithXOR-CDSD-TH-3-10_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 02:19 WithXOR-CDSD-TH-3-11_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 513 janv. 17 22:01 WithXOR-CDSD-TH-3-1_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 02:47 WithXOR-CDSD-TH-3-12_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 03:15 WithXOR-CDSD-TH-3-13_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 03:43 WithXOR-CDSD-TH-3-14_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 04:11 WithXOR-CDSD-TH-3-15_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 04:38 WithXOR-CDSD-TH-3-16_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 05:05 WithXOR-CDSD-TH-3-17_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 515 janv. 18 05:32 WithXOR-CDSD-TH-3-18_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 513 janv. 17 22:26 WithXOR-CDSD-TH-3-2_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 17 22:51 WithXOR-CDSD-TH-3-3_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 17 23:16 WithXOR-CDSD-TH-3-4_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 17 23:42 WithXOR-CDSD-TH-3-5_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 18 00:07 WithXOR-CDSD-TH-3-6_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 18 00:34 WithXOR-CDSD-TH-3-7_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 18 01:00 WithXOR-CDSD-TH-3-8_20_MessageBroadcastStatsReport.txt
4 -rw-rw-r-- 1 qiankun qiankun 514 janv. 18 01:26 WithXOR-CDSD-TH-3-9_20_MessageBroadcastStatsReport.txt

2.1 读取文件

每一个MessageBroadcastStatsReport文件大概包含如下内容：

Message stats for scenario WithXOR-CDSD-TH-3-4_20
sim_time: 47400.0000
created: 466683
started: 1945912
emissions: 1928668
nrofXor: 17244
relayed: 1945912
aborted: 0
dropped: 0
removed: 1945912
delivered: 52538
delivery_prob: 0.1126
xor_prob: 0.0089
response_prob: 0.0000
overhead_ratio: 36.0382
overheadBc_ratio: 36.7100
lastReceivedTime: 43901.0000
latency_avg: 15836.0662
latency_med: 14265.0000
hopcount_avg: 7.9397
hopcount_med: 8
buffertime_avg: 2197.7907
buffertime_med: 1539.0000
rtt_avg: NaN
rtt_med: NaN

将所需要的指标（比如：delivered）读到一个数组，源代码如下：

def read_files(filenames, metric):
    """
        Read values with a given metric from a batch of files.
        Return a list of values
    """
    result = list()

    for filename in filenames:
        with open(filename, 'r') as f:
            for line in f:
                if line.startswith(metric):
                    value = float(line.strip().split(':')[1])
                    result.append(value)
                    break

    return result


def main():
    # Step 1: Read files
    # Get filenames
    list_message_intervals = range(1, 14)
    filenames = ['WithXOR-CDSD-TH-3-{interval}_20_MessageBroadcastStatsReport.txt'.format(interval=interval)
                    for interval in list_message_intervals]

    # Read files
    metric = 'delivered'
    list_delivered = read_files(filenames, metric)
    # [47670.0, 49541.0, 50944.0, 52538.0, 53841.0, 55770.0, 57245.0, 58479.0, 58269.0, 58095.0, 58257.0, 61098.0, 62986.0]

2.2 绘图

有了x（list_message_intervals）和y（list_delivered）就可以绘图了，源代码如下：

# Step 2: Plot figures
    x = list_message_intervals
    y = list_delivered

    fig, ax = plt.subplots()
    ax.plot(x, y, 'r-o', label='With NC', linewidth=2,
                    markeredgewidth=1, markerfacecolor='white', markersize=8)

    # Decoration
    plt.ticklabel_format(style='sci', axis='y', scilimits=(0,0))
    plt.grid()

    plt.xlabel(r'The message creation period $\Delta$', fontsize=15)
    plt.ylabel('The number of messages delivered', fontsize=15)

    ax.legend(loc='best')
    plt.xticks(x, x)
    plt.xlim(min(x), max(x))

    out_file = 'delivered.png'
    plt.savefig(out_file)
    plt.show()

运行程序，得到下图：

Fig. 1：The number of delivered messages in different message creation periods.

本文的完整代码已分享到我的Github，在这里的plot_MessageStatsReport.py。