This is the AWS CDK v2 Developer Guide. The older CDK v1 entered maintenance on June 1, 2022 and ended support on June 1, 2023.

Working with the AWS CDK in Python

Python is a fully-supported client language for the AWS Cloud Development Kit (AWS CDK) and is considered stable. Working with the AWS CDK in Python uses familiar tools, including the standard Python implementation (CPython), virtual environments with virtualenv,and the Python package installerpip.The modules comprising the AWS Construct Library are distributed viapypi.org.The Python version of the AWS CDK even uses Python-style identifiers (for example,snake_casemethod names).

You can use any editor or IDE. Many AWS CDK developers useVisual Studio Code(or its open-source equivalentVSCodium), which has good support for Python via anofficial extension.The IDLE editor included with Python will suffice to get started. The Python modules for the AWS CDK do have type hints, which are useful for a linting tool or an IDE that supports type validation.

Get started with Python

To work with the AWS CDK, you must have an AWS account and credentials and have installed Node.js and the AWS CDK Toolkit. SeeGetting started with the AWS CDK.

Python AWS CDK applications require Python 3.6 or later. If you don't already have it installed,download a compatible versionfor your operating system atpython.org.If you run Linux, your system may have come with a compatible version, or you may install it using your distro's package manager (yum,apt,etc.). Mac users may be interested inHomebrew,a Linux-style package manager for macOS.

The Python package installer,pip,and virtual environment manager,virtualenv,are also required. Windows installations of compatible Python versions include these tools. On Linux,pipand virtualenvmay be provided as separate packages in your package manager. Alternatively, you may install them with the following commands:

python -m ensurepip --upgrade
python -m pip install --upgrade pip
python -m pip install --upgrade virtualenv

If you encounter a permission error, run the above commands with the--userflag so that the modules are installed in your user directory, or usesudoto obtain the permissions to install the modules system-wide.

Creating a project

You create a new AWS CDK project by invokingcdk initin an empty directory. Use the --languageoption and specifypython:

mkdir my-project
cd my-project
cdk init app --language python

cdk inituses the name of the project folder to name various elements of the project, including classes, subfolders, and files. Hyphens in the folder name are converted to underscores. However, the name should otherwise follow the form of a Python identifier; for example, it should not start with a number or contain spaces.

To work with the new project, activate its virtual environment. This allows the project's dependencies to be installed locally in the project folder, instead of globally.

source.venv/bin/activate

After activating your virtual environment for the first time, install the app's standard dependencies:

python -m pip install -r requirements.txt

Managing AWS Construct Library modules

Use the Python package installer,pip,to install and update AWS Construct Library modules for use by your apps, as well as other packages you need.pipalso installs the dependencies for those modules automatically. If your system does not recognizepipas a standalone command, invoke pipas a Python module, like this:

python -m pipPIP-COMMAND

Most AWS CDK constructs are inaws-cdk-lib.Experimental modules are in separate modules named like aws-cdk.SERVICE-NAME.alpha.The service name includes an awsprefix. If you're unsure of a module's name,search for it at PyPI.For example, the command below installs the AWS CodeStar library.

python -m pip install aws-cdk.aws-codestar-alpha

Some services' constructs are in more than one namespace. For example, besidesaws-cdk.aws-route53, there are three additional Amazon Route 53 namespaces, namedaws-route53-targets, aws-route53-patterns,andaws-route53resolver.

The names used for importing AWS Construct Library modules into your Python code look like the following.

import aws_cdk.aws_s3 as s3
import aws_cdk.aws_lambda as lambda_

We recommend the following practices when importing AWS CDK classes and AWS Construct Library modules in your applications. Following these guidelines will help make your code consistent with other AWS CDK applications as well as easier to understand.

After installing a module, update your project'srequirements.txtfile, which lists your project's dependencies. It is best to do this manually rather than usingpip freeze.pip freezecaptures the current versions of all modules installed in your Python virtual environment, which can be useful when bundling up a project to be run elsewhere.

Usually, though, yourrequirements.txtshould list only top-level dependencies (modules that your app depends on directly) and not the dependencies of those libraries. This strategy makes updating your dependencies simpler.

You can editrequirements.txtto allow upgrades; simply replace the==preceding a version number with~=to allow upgrades to a higher compatible version, or remove the version requirement entirely to specify the latest available version of the module.

Withrequirements.txtedited appropriately to allow upgrades, issue this command to upgrade your project's installed modules at any time:

pip install --upgrade -r requirements.txt

Managing dependencies in Python

In Python, you specify dependencies by putting them inrequirements.txtfor applications or setup.pyfor construct libraries. Dependencies are then managed with the PIP tool. PIP is invoked in one of the following ways:

pipcommandoptions
python -m pipcommandoptions

Thepython -m pipinvocation works on most systems;piprequires that PIP's executable be on the system path. Ifpipdoesn't work, try replacing it withpython -m pip.

Thecdk init --language pythoncommand creates a virtual environment for your new project. This lets each project have its own versions of dependencies, and also a basicrequirements.txtfile. You must activate this virtual environment by runningsource.venv/bin/activateeach time you begin working with the project. On Windows, run.\venv\Scripts\activateinstead

CDK applications

The following is an examplerequirements.txtfile. Because PIP does not have a dependency-locking feature, we recommend that you use the == operator to specify exact versions for all dependencies, as shown here.

aws-cdk-lib==2.14.0
aws-cdk.aws-appsync-alpha==2.10.0a0

Installing a module withpip installdoes not automatically add it to requirements.txt.You must do that yourself. If you want to upgrade to a later version of a dependency, edit its version number inrequirements.txt.

To install or update your project's dependencies after creating or editingrequirements.txt, run the following:

python -m pip install -r requirements.txt

Third-party construct libraries

In libraries, dependencies are specified insetup.py,so that transitive dependencies are automatically downloaded when the package is consumed by an application. Otherwise, every application that wants to use your package needs to copy your dependencies into theirrequirements.txt.An example setup.pyis shown here.

from setuptools import setup

setup(
name='my-package',
version='0.0.1',
install_requires=[
'aws-cdk-lib==2.14.0',
],
...
)

To work on the package for development, create or activate a virtual environment, then run the following command.

python -m pip install -e.

Although PIP automatically installs transitive dependencies, there can only be one installed copy of any one package. The version that is specified highest in the dependency tree is selected; applications always have the last word in what version of packages get installed.

AWS CDK idioms in Python

Language conflicts

In Python,lambdais a language keyword, so you cannot use it as a name for the AWS Lambda construct library module or Lambda functions. The Python convention for such conflicts is to use a trailing underscore, as in lambda_,in the variable name.

By convention, the second argument to AWS CDK constructs is namedid.When writing your own stacks and constructs, calling a parameterid"shadows" the Python built-in functionid(),which returns an object's unique identifier. This function isn't used very often, but if you should happen to need it in your construct, rename the argument, for exampleconstruct_id.

Arguments and properties

All AWS Construct Library classes are instantiated using three arguments: thescopein which the construct is being defined (its parent in the construct tree), anid,and props,a bundle of key/value pairs that the construct uses to configure the resources it creates. Other classes and methods also use the "bundle of attributes" pattern for arguments.

scopeandidshould always be passed as positional arguments, not keyword arguments, because their names change if the construct accepts a property namedscopeor id.

In Python, props are expressed as keyword arguments. If an argument contains nested data structures, these are expressed using a class which takes its own keyword arguments at instantiation. The same pattern is applied to other method calls that take a structured argument.

For example, in a Amazon S3 bucket'sadd_lifecycle_rulemethod, thetransitionsproperty is a list ofTransitioninstances.

bucket.add_lifecycle_rule(
transitions=[
Transition(
storage_class=StorageClass.GLACIER,
transition_after=Duration.days(10)
)
]
)

When extending a class or overriding a method, you may want to accept additional arguments for your own purposes that are not understood by the parent class. In this case you should accept the arguments you don't care about using the**kwargsidiom, and use keyword-only arguments to accept the arguments you're interested in. When calling the parent's constructor or the overridden method, pass only the arguments it is expecting (often just **kwargs). Passing arguments that the parent class or method doesn't expect results in an error.

class MyConstruct(Construct):
def __init__(self, id, *, MyProperty=42, **kwargs):
super().__init__(self, id, **kwargs)
#...

A future release of the AWS CDK could coincidentally add a new property with a name you used for your own property. This won't cause any technical issues for users of your construct or method (since your property isn't passed "up the chain," the parent class or overridden method will simply use a default value) but it may cause confusion. You can avoid this potential problem by naming your properties so they clearly belong to your construct. If there are many new properties, bundle them into an appropriately-named class and pass it as a single keyword argument.

Missing values

The AWS CDK usesNoneto represent missing or undefined values. When working with **kwargs,use the dictionary'sget()method to provide a default value if a property is not provided. Avoid usingkwargs[...],as this raisesKeyErrorfor missing values.

encrypted = kwargs.get( "encrypted" ) # None if no property "encrypted" exists
encrypted = kwargs.get( "encrypted", False) # specify default of False if property is missing

Some AWS CDK methods (such astryGetContext()to get a runtime context value) may return None,which you will need to check explicitly.

Using interfaces

Python doesn't have an interface feature as some other languages do, though it does haveabstract base classes,which are similar. (If you're not familiar with interfaces, Wikipedia hasa good introduction.) TypeScript, the language in which the AWS CDK is implemented, does provide interfaces, and constructs and other AWS CDK objects often require an object that adheres to a particular interface, rather than inheriting from a particular class. So the AWS CDK provides its own interface feature as part of theJSIIlayer.

To indicate that a class implements a particular interface, you can use the@jsii.implements decorator:

from aws_cdk import IAspect, IConstruct
import jsii

@jsii.implements(IAspect)
class MyAspect():
def visit(self, node: IConstruct) -> None:
print( "Visited", node.node.path)

Type pitfalls

Python uses dynamic typing, where all variables may refer to a value of any type. Parameters and return values may be annotated with types, but these are "hints" and are not enforced. This means that in Python, it is easy to pass the incorrect type of value to a AWS CDK construct. Instead of getting a type error during build, as you would from a statically-typed language, you may instead get a runtime error when the JSII layer (which translates between Python and the AWS CDK's TypeScript core) is unable to deal with the unexpected type.

In our experience, the type errors Python programmers make tend to fall into these categories.

The AWS CDK Python modules do include type annotations, so you can use tools that support them to help with types. If you are not using an IDE that supports these, such asPyCharm,you might want to call theMyPytype validator as a step in your build process. There are also runtime type checkers that can improve error messages for type-related errors.

Synthesizing and deploying

Thestacksdefined in your AWS CDK app can be synthesized and deployed individually or together using the commands below. Generally, you should be in your project's main directory when you issue them.

You can specify the names of multiple stacks to be synthesized or deployed in a single command. If your app defines only one stack, you do not need to specify it.

cdk synth # app defines single stack
cdk deploy Happy Grumpy # app defines two or more stacks; two are deployed

You may also use the wildcards * (any number of characters) and? (any single character) to identify stacks by pattern. When using wildcards, enclose the pattern in quotes. Otherwise, the shell may try to expand it to the names of files in the current directory before they are passed to the AWS CDK Toolkit.

cdk synth "Stack?" # Stack1, StackA, etc.
cdk deploy "*Stack" # PipeStack, LambdaStack, etc.

For full documentation of thecdkcommand, seeAWS CDK Toolkit (cdk command).