只要你用 UTF-8 的編碼來儲存描述關係圖的 dot 檔，Graphviz 也可以產生包含中文的關係圖。但 Windows 版的 Graphviz 似乎沒有設定好 fontconfig，因此會找不到中文字型，只要設定好 fontconfig 即可解決。依照預設安裝路徑，設定檔在 C:\\Program Files\\Graphviz 2.21\\etc\\fonts\\fonts.conf，找到以下這行：

    <dir>#FONTDIR#</dir>
改成 Windows 字型檔的路徑：

    <dir>C:\\WINDOWS\\Fonts</dir>
Graphviz 就可以使用中文了！但注意目前似乎只能輸入英文的字型名稱，範例如下：

digraph g {
    node[fontname = "PMingLiu"];
    "中文" -> "英文";
}

View Code

(4) 可以运行gvedit.exe写入代码，f5直接预览效果

(5) python库，其实用处不大

https://pypi.python.org/pypi/graphviz

3.参考文档

官网： http://graphviz.org/Documentation.php

使用DOT语言和Graphviz绘图(翻译)

一次性批量定义属性

    node [color=Red,fontname=Courier,shape=box] //All nodes will this shape and colour
    edge [color=Blue, style=dashed] //All the lines look like this

标题: 【合集】dot language Graphviz流程图的问题

　　（连线连在角上）

　　portPos headport tailport 这几个属性

画图神器--graphviz

digraph login3times{

    edge[fontname="Microsoft YaHei"]
    node[shape=box, fontname="Microsoft YaHei" size="20,20"]
    fontname="Microsoft YaHei"
    label="三次登录流程图"

4.语法注意事项

(1)框图为node，默认形状 ellipse 椭圆形， shape = box 长方形，diamond 菱形，circle 圆形

(2)连线为edge，箭头方向 [dir=forward, back, both, or none]

(3)在有必要的时候加双引号，如：“do sth”, "func()"

(4)注释为 //，或 /* */

(5)不方便之处：

A -> B -> C ->D[label=to] ，则会在三条连线上都添加 to 字样

对比：

A -> B

B -> C[label=to]

C ->D

5.综合应用

//G:\\pydata\\pycode>dot eg.dot -Tpng -o eg.png
digraph graphname
{
//rankdir=LR
node[fontname = "simsun"]
fontname="Microsoft YaHei"
label="代理IP获取流程图"

START[label="入口get_mimvp()" shape=box]
END[label=出口ip_port_type_tuple_list, shape=box]
JUDGE[label="feature_vectors == []" color=Blue, fontsize=24, shape=diamond]

START -> JUDGE
                {rank=same; "extract_features()", "load_mimvp()"}
                JUDGE -> "extract_features()"[label=True tailport=w]// headport=n]
                        "extract_features()"[style=filled, fillcolor=orange]
                JUDGE -> "load_mimvp()"[label=False tailport=s headport=n]
                    
"extract_features()" -> "load_mimvp()"[label=return, fontcolor=red, dir=forward]   //dir=forward, back, both, or none        


port_src_list[style=filled, fillcolor=yellow]
port_list[style=filled, fillcolor=green]

"load_mimvp()" -> {ip_list, type_list, port_src_list}
                                        port_src_list -> port_list
{rank=same; ip_list, type_list, port_list} -> "merge_result" -> END


    subgraph cluster1
    {
    label=子图1_特征提取
    
    "invoke extract_features()"[style=filled, fillcolor=orange]
    judge[label="filepath is not None" color=Blue, shape=diamond]
    feature_vectors[style=filled, fillcolor=red]
    
    "invoke extract_features()" -> judge
        judge -> "load_images_from_filepath()"[label=True tailport=w]
        judge -> "load_images_from_src_list()"[label=False tailport=s]
        {rank=same; "load_images_from_filepath()", "load_images_from_src_list()"}
            "load_images_from_filepath()" -> "split_image()"[label=img_list]
            "load_images_from_src_list()" -> "split_image()"[label=img_list]
                "split_image()" -> "build_vector()"[label="split_imgs"]
                    "build_vector()" -> feature_vectors[label="item = {input: vector}"]   // -> "load_mimvp()"
    } 

 
    subgraph cluster2
    {
    label=子图2_余弦相似度比较
    port_src[style=filled, fillcolor=yellow]
    port[style=filled, fillcolor=green]

    port_src -> "get_port()" -> "load_image_from_src()" -> "invoke split_image()" -> "invoke build_vector()"
        "invoke build_vector()" -> "cos_similarity()"[label="vector\\nfeature_vectors", fontcolor=red]
            "cos_similarity()" -> port
    }  
}

View Code

代码另存为eg.dot, 编码utf-8

在命令行中运行：G:\\pydata\\pycode>dot eg.dot -Tpng -o eg.png

以上是关于流程图 Graphviz - Graph Visualization Software的主要内容，如果未能解决你的问题，请参考以下文章

一款适合程序员的流程图/思维导图利器

doxygen+graphviz轻松绘制函数调用图（call graph）

如何在 scikit-learn 的随机森林的 graphviz-graph 中找到一个类？

看开源代码利器—用Graphviz + CodeViz生成C/C++函数调用图(call graph) - 转

Graphviz中文乱码问题