DEVELOPMENT_GUIDE.md

Development Guide

Welcome hacker!

This document will make your life easier by helping you setup a development environment, IDEs, tests, coding practices, or anything that will help you be more productive. If you found something is missing or inaccurate, update this guide and send a Pull Request.

Note: pyenv currently only supports macOS and Linux. If you are a Windows users, consider using pipenv.

1-click ready-to-hack IDE

For setting up a local development environment, we recommend using Gitpod - a service that allows you to spin up an in-browser Visual Studio Code-compatible editor, with everything set up and ready to go for development on this project. Just click the button below to create your private workspace:

This will start a new Gitpod workspace, and immediately kick off a build of the code. Once it's done, you can start working.

Gitpod is free for 50 hours per month - make sure to stop your workspace when you're done (you can always resume it later, and it won't need to run the build again).

Environment setup

1. Install Python versions

Our officially supported Python versions are 3.8, 3.9 and 3.10. Our CI/CD pipeline is setup to run unit tests against Python 3 versions. Make sure you test it before sending a Pull Request. See Unit testing with multiple Python versions.

pyenv is a great tool to easily setup multiple Python versions. For

Note: For Windows, type export PATH="/c/Users/<user>/.pyenv/libexec:$PATH" to add pyenv to your path.

1. Install PyEnv - curl -L https://github.com/pyenv/pyenv-installer/raw/master/bin/pyenv-installer | bash
2. Restart shell so the path changes take effect - exec $SHELL
3. pyenv install 3.8.16
4. pyenv install 3.9.16
5. pyenv install 3.10.9
6. Make Python versions available in the project: pyenv local 3.8.16 3.9.16 3.10.9

Note: also make sure the following lines were written into your .bashrc (or .zshrc, depending on which shell you are using):

export PATH="$HOME/.pyenv/bin:$PATH"
eval "$(pyenv init -)"
eval "$(pyenv virtualenv-init -)"

2. Install additional tooling

Black

We format our code using Black and verify the source code is black compliant during PR checks. Black will be installed automatically with make init.

After installing, you can run our formatting through our Makefile by make format or integrating Black directly in your favorite IDE (instructions can be found here)

(Workaround) Integrating Black directly in your favorite IDE

Since black is installed in virtualenv, when you follow this instruction, which black might give you this

(sam38) $ where black
/Users/<username>/.pyenv/shims/black

However, IDEs such PyChaim (using FileWatcher) will have a hard time invoking /Users/<username>/.pyenv/shims/black and this will happen:

pyenv: black: command not found

The `black' command exists in these Python versions:
  3.8.16/envs/sam38
  sam38

A simple workaround is to use /Users/<username>/.pyenv/versions/sam38/bin/black instead of /Users/<username>/.pyenv/shims/black.

Pre-commit

If you don't wish to manually run black on each pr or install black manually, we have integrated black into git hooks through pre-commit. After installing pre-commit, run pre-commit install in the root of the project. This will install black for you and run the black formatting on commit.

3. Activate virtualenv

Virtualenv allows you to install required libraries outside of the Python installation. A good practice is to setup a different virtualenv for each project. pyenv comes with a handy plugin that can create virtualenv.

Depending on the python version, the following commands would change to be the appropriate python version.

1. Create Virtualenv sam38 for Python3.8: pyenv virtualenv 3.8.16 sam38
2. Activate Virtualenv: pyenv activate sam38

4. Install dev version of SAM transform

We will install a development version of SAM transform from source into the virtualenv.

1. Activate Virtualenv: pyenv activate sam38
2. Install dev version of SAM transform: make init

Running tests

Unit testing with one Python version

If you're trying to do a quick run, it's ok to use the current python version. Run make test or make test-fast. Once all tests pass make sure to run make pr before sending out your PR.

Unit testing with multiple Python versions

Currently, our officially supported Python versions are 3.8, 3.9 and 3.10. For the most part, code that works in Python3.8 will work in Pythons 3.9 and 3.10. You only run into problems if you are trying to use features released in a higher version (for example features introduced into Python3.10 will not work in Python3.9). If you want to test in many versions, you can create a virtualenv for each version and flip between them (sourcing the activate script). Typically, we run all tests in one python version locally and then have our ci (appveyor) run all supported versions.

Transform tests

Successful transforms

Transform tests ensure a SAM template transforms into the expected CloudFormation template.

We provide a script to help generate the transform test input and output files in the correct directory given a SAM template. For example:

python3 bin/add_transform_test.py --template-file template.yaml

Warning

Always check the generated output is as expected. This tool does not guarantee correct output.

Transform failures

To test a SAM template results in a specific transform error, add the SAM template under tests/translator/input, and a JSON file with the expected errorMessage as a top-level value under tests/translator/output. Both files must have the same basename and be prefixed with error_ (e.g. error_my_cool_template.yaml for input, and error_my_cool_template.json for the expected error).

See #2993 for an example.

Integration tests

Integration tests are covered in detail in the INTEGRATION_TESTS.md file of this repository.

Development guidelines

1. Do not resolve intrinsic functions. Adding AWS::LanguageExtensions before the AWS::Serverless-2016-10-31 transform resolves most of them (see #2533).

2. Do not break backward compatibility. A specific SAM template should always transform into the same CloudFormation template. A template that has previously deployed successfully should continue to do so. Do not change logical IDs. Add opt-in properties for breaking changes. There are some exceptions, such as changes that do not impact resources (e.g. Metadata) or abstractions that can by design change over time.

3. Stick as close as possible to the underlying CloudFormation properties. This includes both property names and values. This ensures we can pass values to CloudFormation and let it handle any intrinsic functions. In some cases, it also allows us to pass all properties as-is to a resource, which means customers can always use the newest properties, and we don’t spend effort maintaining a duplicate set of properties.

4. Only validate what’s necessary. Do not validate properties if they’re passed directly to the underlying CloudFormation resource.

5. Add type hints to new code. Strict typing was enabled in #2558 by sprinkling # type: ignore across the existing code. Don't do that for new code. Avoid # type: ignores at all cost. Instead, add types to new functions, and ideally add types to existing code it uses as well.

6. Do not use PropertyType for new Resource properties.

   For new properties of SAM resources, use Property or PassThroughProperty instead of PropertyType. This avoids sneaky bugs and ensures valid templates do not cause transform failures.

   For new properties of CloudFormation resources, use GeneratedProperty. It performs no runtime validation, reducing the risk of valid values causing transform failures.

7. Write all new code under samtranslator/internal if possible. This ensures we don't increase our "public" library interface and cause unnecessary breakages to consumers. While Python doesn't have private access modifiers, we assume anything by convention private or under internal to be internal code, not bound by typical expectations of backward compatibility.

Code conventions

Please follow these code conventions when making your changes. This will align your code to the same conventions used in rest of the package and make it easier for others to read/understand your code. Some of these conventions are best practices that we have learnt over time.

• Don't write any code in __init__.py file unless there is a really strong reason.
• Module-level logger variable must be named as LOG
• If your method wants to report a failure, it must raise a custom exception. Built-in Python exceptions like TypeError, KeyError are raised by Python interpreter and usually signify a bug in your code. Your method must not explicitly raise these exceptions because the caller has no way of knowing whether it came from a bug or not. Custom exceptions convey are must better at conveying the intent and can be handled appropriately by the caller. In HTTP lingo, custom exceptions are equivalent to 4xx (user's fault) and built-in exceptions are equivalent to 5xx (Service Fault)
• Don't use *args or **kwargs unless there is a really strong reason to do so. You must explain the reason in great detail in docstrings if you were to use them.
• Do not catch the broader Exception, unless you have a really strong reason to do. You must explain the reason in great detail in comments.

Making schema changes

The AWS SAM specification includes a JSON schema (see #2645). All test templates must validate against it.

To add new properties, do the following:

1. Add the property to the relevant resource schema under samtranslator/internal/schema_source (e.g. samtranslator/internal/schema_source/aws_serverless_function.py for AWS::Serverless::Function).
2. You can leave out the assignement part; it adds documentation to the schema properties. The team will take care of documentation updates once code changes are merged. Typically we update documentation by running make update-schema-data.
3. Run make schema.

Profiling

Install snakeviz: pip install snakeviz

python -m cProfile -o sam_profile_results bin/sam-translate.py translate --template-file=tests/translator/input/alexa_skill.yaml --output-template=cfn-template.json
snakeviz sam_profile_results

Verifying transforms

If you make changes to the transformer and want to verify the resulting CloudFormation template works as expected, you can transform your SAM template into a CloudFormation template using the following process:

# Optional: You only need to run the package command in certain cases; e.g. when your CodeUri specifies a local path
# Replace MY_TEMPLATE_PATH with the path to your template and MY_S3_BUCKET with an existing S3 bucket
aws cloudformation package --template-file MY_TEMPLATE_PATH/template.yaml --output-template-file output-template.yaml --s3-bucket MY_S3_BUCKET

# Transform your SAM template into a CloudFormation template
# Replace "output-template.yaml" if you didn't run the package command above or specified a different path for --output-template-file
bin/sam-translate.py --template-file=output-template.yaml

# Deploy your transformed CloudFormation template
# Replace MY_STACK_NAME with a unique name each time you deploy
aws cloudformation deploy --template-file cfn-template.json --capabilities CAPABILITY_NAMED_IAM --stack-name MY_STACK_NAME