Diagrams as Code: The Complete How-to-Use Guide

We're seeing more and more tools that enable you to create software architecture and other Diagrams as Code. The main benefit of using this concept is that majority of the Diagrams as Code tools can be scripted and integrated into a built pipeline for generating automatic documentation. The other benefit responsible for the growing use of Diagrams as code to create software architecture is that it enables the use of text-based tooling, which most software developers already use. Furthermore, text is easily version-controllable and diff’able.

Table of Contents

• What is Diagram as Code?
• How to install Diagrams
• How to use Diagrams
• Conclusion

What Is Diagrams as Code?

In early 2020, Korean developer MinJae Kwon decided to create Diagrams. Diagrams allows you to draw cloud system architecture in Python code, allowing you to track your Diagram in any SCM. It supports major providers such as AWS, Azure, GCP, Kubernetes, OpenStack, Oracle Cloud, etc. but supports drawing on-premise infrastructure as well.

How to Install Diagrams

Requirement: Python 3.6 or higher

Depending on your environment, run one of those commands below:

# using Homebrew for macOS users$  
brew install graphviz 
# using pip (pip3)  
$ pip install diagrams  
# using pipenv  
$ pipenv install diagrams  
# using poetry  
$ poetry add diagrams

How to Use Diagrams

You must import the necessary modules you want to add to your diagrams. They are called Nodes. They represent a node or system component. Nodes are composed of three parts: Provider, Resource type, Name.

You can import OnPrem, AWS, Azure, GCP, Kubernetes nodes, and more.  If you aren't able to find something, you could always use this custom module.

  from diagrams.gcp.network import LoadBalancing

In the above example, the LoadBalancing is a node of the Network resource type provided by the GCP provider.

Custom Graphviz dot attributes options are supported. graph_attr note_attr and edge_attr can be used.

Reference link: https://www.graphviz.org/doc/info/attrs.html

  graph_attr = {    
      "layout": "dot",   
      "concentrate": "true",   
      "compound": "true",   
      "splines": "spline", 
}

Diagram represents a global diagram context.

  with Diagram("Client to Application Flow", show=False, graph_attr=graph_attr):

show=False Will disable the automatic file opening when you generate the diagram.

Data Flow and how to connect nodes together:

>> Connect nodes in the left to the right direction.

<< Connect nodes in right to left direction.

- Connect nodes in no direction. Undirected.

You can change the data flow direction with the direction parameter. Default is LR. (Left to Right)

from diagrams import Diagram
from diagrams.aws.compute import EC2
from diagrams.aws.database import RDS
from diagrams.aws.network import ELB
from diagrams.aws.storage import S3 


with Diagram("Web Services", show=False):   
    ELB("lb") >> EC2("web") >> RDS("userdb") >> S3("store")   
    ELB("lb") >> EC2("web") >> RDS("userdb") << EC2("stat")    
    (ELB("lb") >> EC2("web")) - EC2("web") >> RDS("userdb")

from diagrams import Diagram
from diagrams.aws.compute import EC2
from diagrams.aws.database import RDS
from diagrams.aws.network import ELB 


with Diagram("Workers", show=False, direction="TB"):    
    lb = ELB("lb")   
    db = RDS("events")   
    lb >> EC2("worker1") >> db    
    lb >> EC2("worker2") >> db    
    lb >> EC2("worker3") >> db   
    lb >> EC2("worker4") >> db   
    lb >> EC2("worker5") >> d

Clusters allow you to group the nodes in an isolated group. You can create a cluster context with the Cluster class. And you can also connect the nodes in a cluster to other nodes outside a cluster. There is no depth limit of nesting so that you can imagine the possibilities.

from diagrams import Cluster, Diagram, Node, Edge
from diagrams.k8s.compute import Pod
from diagrams.k8s.network import Ing
from diagrams.gcp.network import LoadBalancing
from diagrams.onprem.network import Nginx  

with Diagram("Cluster nesting", show=False):   

    gcp_lb = LoadBalancing("GCP LB")   

    with Cluster("Kubernetes"):      
  
        with Cluster("Nginx"):          
            nginx= Nginx("")        

        with Cluster("MyApp"):          
            myapp_ing = Ing("")          
            with Cluster("Pods"):             
                myapp_pods = Pod("myapp")      


       with Cluster("MySQL"):           
           myapp_db = Pod("myapp-db")    

   gcp_lb >> Edge(headport="c", tailport="c", minlen="1", lhead='cluster_Kubernetes') >> nginx    
   nginx >> Edge(headport="c", tailport="c", minlen="1", lhead='cluster_MyApp') >> myapp_ing >> Edge(headport="c", tailport="c", minlen="1", lhead='cluster_MyApp pods') >> myapp_pods >> myapp_db

Edges represent a link between Nodes. It contains three attributes: label, color, and style.

https://diagrams.mingrammer.com/docs/guides/edge

Custom Node

We usually import icons externally hosted so they can be accessed when we generate the diagrams.

from diagrams import Cluster, Diagram, Node, Edge
from diagrams.custom import Custom
from urllib.request import urlretrieve 


with Diagram("Import logo CertManager", show=False):     


    with Cluster("This is a logo"):      
        certmanager_url = "https://github.com/jetstack/cert-manager/raw/master/logo/logo.png"      
        certmanager_icon = "logo.png"       
        urlretrieve(certmanager_url, certmanager_icon)       
        certmanager = Custom("Cert Manager", certmanager_icon)

If you want to import local icons saved in your repository, please review the example here.

Generate the Diagram

Once you're happy with the code, generate the diagram by running the command below, which will create a .png version of your diagram.

  python my_diagram.py

Conclusion

Diagrams by Mindgrammer isn't the only option out there with Cloudgram, PlantUML, or even Draw.io having similar functionality of saving the XML generated in Git. We clearly do not lack options with open-source tools to create diagrams. Similarly, Python is so overly present that learning how to use Diagrams wasn't a steep learning curve.