Skip to content



PRs welcome!

  • Consider starting a discussion to see if there's interest in what you want to do.
  • Submit PRs from feature branches on forks to the develop branch.
  • Ensure PRs pass all CI checks.
  • Maintain test coverage at 100%.




This project uses Hatch for dependency management and packaging.



Hatch can be installed with Homebrew or pipx.

Install project with all dependencies: hatch env create.

Key commands

# Basic usage:
hatch env create  # create virtual environment and install dependencies
hatch env find  # show path to virtual environment
hatch env show  # show info about available virtual environments
hatch run COMMAND  # run a command within the virtual environment
hatch shell  # activate the virtual environment, like source venv/bin/activate
hatch version  # list or update version of this package
export HATCH_ENV_TYPE_VIRTUAL_PATH=.venv  # install virtualenvs into .venv

Testing with pytest

General info

Integration testing instructions

Integration tests will be skipped if cloud credentials are not present. Running integration tests locally will take some additional setup.

Make buckets on each supported cloud platform
Upload objects to each bucket

Upload an object to each bucket named .env.testing.

The file should have this content:

# .env
# comment
INLINE_COMMENT=no_comment  # inline comment
JSON_EXAMPLE='{"array": [1, 2, 3], "exponent": 2.99e8, "number": 123}'
QUOTES_AND_WHITESPACE='text and spaces'
OBJECT_STORAGE_VARIABLE='DUDE!!! This variable came from object storage!'
Generate credentials for each supported cloud platform

There are three sets of credentials needed:

  1. AWS temporary credentials
  2. AWS static credentials
  3. Backblaze static credentials

The object storage docs have general info on generating the static credentials.

For AWS static credentials, create a non-admin user. The user will need an IAM policy like the following. This project doesn't do any listing or deleting at this time, so those parts can be omitted if you're going for least privilege.

    "Version": "2012-10-17",
    "Statement": [
            "Effect": "Allow",
            "Action": ["s3:ListBucket"],
            "Resource": ["arn:aws:s3:::<AWS_S3_BUCKET_NAME>"]
            "Effect": "Allow",
            "Action": ["s3:PutObject", "s3:GetObject", "s3:DeleteObject"],
            "Resource": ["arn:aws:s3:::<AWS_S3_BUCKET_NAME>/*"]

After attaching the IAM policy to the non-admin user, generate an access key for the non-admin user, set up an AWS CLI profile named fastenv, and configure it with the access key for the non-admin user. AWS static credentials are now ready.

AWS temporary credentials work a little differently. Create an IAM role, with a resource-based policy called a "role trust policy." The role trust policy would look like this (<AWS_IAM_USERNAME> is the IAM user that owns the static credentials, not your admin username):

    "Version": "2012-10-17",
    "Statement": [
            "Effect": "Allow",
            "Principal": {
                "AWS": "arn:aws:iam::<AWS_ACCOUNT_ID>:user/<AWS_IAM_USERNAME>"
            "Action": ["sts:AssumeRole", "sts:TagSession"]

Attach the identity-based policy you created for the IAM user to the role as well.

The end result is that the IAM user can assume the IAM role and obtain temporary credentials. The temporary credentials have the same IAM policy as the regular access key, so tests can be parametrized accordingly.

Run all the tests

Once you're finally done with all that, maybe go out for a walk or something.

Then come back, and run these magic words:

# set the required input variables
# leading space to avoid storing secret key in shell history
# set `HISTCONTROL=ignoreboth` for Bash or `setopt histignorespace` for Zsh

# get AWS account ID from STS (replace fx with jq or other JSON parser as needed)
AWS_ACCOUNT_ID=$(aws sts get-caller-identity | fx .Account)

# assume the IAM role to get temporary credentials
  aws sts assume-role \
  --role-arn arn:aws:iam::$AWS_ACCOUNT_ID:role/$AWS_IAM_ROLE_NAME \
  --role-session-name fastenv-testing-local-aws-cli \
  --profile fastenv

# run all tests by providing the necessary input variables
AWS_IAM_ACCESS_KEY_FASTENV=$(aws configure get fastenv.aws_access_key_id) \
  AWS_IAM_ACCESS_KEY_SESSION=$(echo $ASSUMED_ROLE | fx .Credentials.AccessKeyId) \
  AWS_IAM_SECRET_KEY_SESSION=$(echo $ASSUMED_ROLE | fx .Credentials.SecretAccessKey) \
  AWS_IAM_SECRET_KEY_FASTENV=$(aws configure get fastenv.aws_secret_access_key) \
  AWS_IAM_SESSION_TOKEN=$(echo $ASSUMED_ROLE | fx .Credentials.SessionToken) \
  pytest --cov-report=html --durations=0 --durations-min=0.5

Code quality

Code style

  • Python code is formatted with Black. Configuration for Black is stored in pyproject.toml.
  • Python imports are organized automatically with isort.
    • The isort package organizes imports in three sections:
      1. Standard library
      2. Dependencies
      3. Project
    • Within each of those groups, import statements occur first, then from statements, in alphabetical order.
    • You can run isort from the command line with hatch run isort ..
    • Configuration for isort is stored in pyproject.toml.
  • Other web code (JSON, Markdown, YAML) is formatted with Prettier.

Static type checking

  • To learn type annotation basics, see the Python typing module docs, Python type annotations how-to, the Real Python type checking tutorial, and this gist.
  • Type annotations are not used at runtime. The standard library typing module includes a TYPE_CHECKING constant that is False at runtime, but True when conducting static type checking prior to runtime. Type imports are included under if TYPE_CHECKING: conditions so that they are not imported at runtime. These conditions are ignored when calculating test coverage.
  • Type annotations can be provided inline or in separate stub files. Much of the Python standard library is annotated with stubs. For example, the Python standard library logging.config module uses type stubs. The typeshed types for the logging.config module are used solely for type-checking usage of the logging.config module itself. They cannot be imported and used to type annotate other modules.
  • The standard library typing module includes a NoReturn type. This would seem useful for unreachable code, including functions that do not return a value, such as test functions. Unfortunately mypy reports an error when using NoReturn, "Implicit return in function which does not return (misc)." To avoid headaches from the opaque "misc" category of mypy errors, these functions are annotated as returning None.
  • Mypy is used for type-checking. Mypy configuration is included in pyproject.toml.
  • Mypy strict mode is enabled. Strict includes --no-explicit-reexport (implicit_reexport = false), which means that objects imported into a module will not be re-exported for import into other modules. Imports can be made into explicit exports with the syntax from module import x as x (i.e., changing from import logging to import logging as logging), or by including imports in __all__. This explicit import syntax can be confusing. Another option is to apply mypy overrides to any modules that need to leverage implicit exports.


Pre-commit runs Git hooks. Configuration is stored in .pre-commit-config.yaml. It can run locally before each commit (hence "pre-commit"), or on different Git events like pre-push. Pre-commit is installed in the Python virtual environment. To use:

❯ cd path/to/fastenv

❯ hatch env create

❯ hatch shell

# install hooks that run before each commit
.venv  pre-commit install

# and/or install hooks that run before each push
.venv  pre-commit install --hook-type pre-push

Spell check

Spell check is performed with CSpell.

In GitHub Actions, CSpell runs using cspell-action.

To run spell check locally, consider installing their VSCode extension or running from the command line.

CSpell can be run with pnpm if pnpm is installed:

pnpm -s dlx cspell --dot --gitignore "**/*.md"

or with npx if npm is installed:

npx -s -y cspell --dot --gitignore "**/*.md"

CSpell also offers a pre-commit hook through their cspell-cli repo. A .pre-commit-config.yaml configuration could look like this:

    - repo:
      rev: v6.16.0
          - id: cspell
            files: "^.*.md$"
            args: ["--dot", "--gitignore", "**/*.md"]

CSpell is not currently used with pre-commit in this project because behavior of the pre-commit hook is inconsistent.

GitHub Actions workflows

GitHub Actions is a continuous integration/continuous deployment (CI/CD) service that runs on GitHub repos. It replaces other services like Travis CI. Actions are grouped into workflows and stored in .github/workflows. See Getting the Gist of GitHub Actions for more info.

GitHub Actions and AWS

Static credentials

As explained in the section on generating credentials for local testing, a non-admin IAM user must be created in order to allow GitHub Actions to access AWS when using static credentials. The IAM user for this repo was created following IAM best practices. In AWS, there is a GitHubActions IAM group, with a fastenv IAM user (one user per repo). The fastenv user has an IAM policy attached specifying its permissions.

On GitHub, the fastenv user access key is stored in GitHub Secrets.

The bucket host is stored in GitHub Secrets in the "virtual-hosted-style" format (<bucketname>.s3.<region>

Temporary credentials

In addition to the static access key, GitHub Actions also retrieves temporary security credentials from AWS using OpenID Connect (OIDC). See the GitHub docs for further info.

The OIDC infrastructure is provisioned with Terraform, using a similar approach to the example in br3ndonland/terraform-examples.

GitHub Actions and Backblaze B2

A B2 application key is stored in GitHub Secrets, along with the corresponding bucket host in "virtual-hosted-style" format (<bucket-name>.s3.<region-name>

See the Backblaze B2 S3-compatible API docs for further info.


  • The default branch is develop.
  • PRs should be merged into develop. Head branches are deleted automatically after PRs are merged.
  • The only merges to main should be fast-forward merges from develop.
  • Branch protection is enabled on develop and main.
    • develop:
      • Require signed commits
      • Include administrators
      • Allow force pushes
    • main:
      • Require signed commits
      • Include administrators
      • Do not allow force pushes
      • Require status checks to pass before merging (commits must have previously been pushed to develop and passed all checks)
  • To create a release:
    • Bump the version number in fastenv.__version__ with hatch version and commit the changes to develop.
      • Follow SemVer guidelines when choosing a version number. Note that PEP 440 Python version specifiers and SemVer version specifiers differ, particularly with regard to specifying prereleases. Use syntax compatible with both.
      • The PEP 440 default (like 1.0.0a0) is different from SemVer. Hatch and PyPI will use this syntax by default.
      • An alternative form of the Python prerelease syntax permitted in PEP 440 (like 1.0.0-alpha.0) is compatible with SemVer, and this form should be used when tagging releases. As Hatch uses PEP 440 syntax by default, prerelease versions need to be written directly into fastenv.__version__.
      • Examples of acceptable tag names: 1.0.0, 1.0.0-alpha.0, 1.0.0-beta.1
    • Push to develop and verify all CI checks pass.
    • Fast-forward merge to main, push, and verify all CI checks pass.
    • Create an annotated and signed Git tag.
      • List PRs and commits in the tag message:
        git log --pretty=format:"- %s (%h)" \
          "$(git describe --abbrev=0 --tags)"..HEAD
      • Omit the leading v (use 1.0.0 instead of v1.0.0)
      • Example: git tag -a -s 1.0.0
    • Push the tag. GitHub Actions will build and publish the Python package and update the changelog.
    • Squash and merge the changelog PR, removing any Co-authored-by trailers before merging.