Skip to content
/ pipper Public

A serverless Python package manager for private packages that runs on S3

Notifications You must be signed in to change notification settings

sernst/pipper

Repository files navigation

NOTE: This project moved to Pipper @ Gitlab on 11/11/2020. Consider this repo deprecated as it will eventually be archived.

Pipper

Private Python package manager on an S3 bucket

An experimental Python package manager wrapped around pip for lightweight management of non-public packages with an AWS S3 static backend. Requires no server or database resources, only a private S3 bucket that stores the pipper packages. Authentication is handled using standard AWS Identity and Access Management (IAM) users, roles and policies.

Installing pipper

The pipper package can be installed using pip:

$ pip install pipper

Basic Usage

Pipper is primarily used from the command line and consists of multiple sub-command actions. The general format of a pipper command is:

$ pipper <ACTION> <REQUIRED_ARGS> --flag=<VALUE> --other-flag ...

The available actions are:

  • install: add or update new packages
  • download: save remote packages locally
  • info: information on a specific package
  • bundle: bundle a package for publishing
  • publish: release a new or updated package
  • authorize: create a pre-authorized url for download
  • repository: Modify pre-defined pipper repositories

AWS Credentials

Pipper uses AWS credentials for authentication. To maximize flexibility, the AWS credentials can be specified in a myriad of ways. Pipper will try to identify credentials in the following order:

1. Explicit Credentials: You can specify the AWS credentials directly on the command line with the --credentials flag:

  • -c --credentials <AWS_ACCESS_KEY_ID> <AWS_SECRET> <AWS_SESSION_TOKEN>

This can be useful for situations where profiles are not initialized or undesirable. If your credentials do not require a session token, which is usually the case, that argument can be omitted. It's also possible to specify a missing token using a 0 value for the <AWS_SESSION_TOKEN> argument for simplicity in cases where omitting the value is more difficult than including it with an explicit ignore value.

2. Pipper Configuration: Using pipper's repository command action, you can store credentials and remote information in a pipper config file. If you do create a pipper repository configuration, which stores AWS credentials, you can reference that repository configuration by name to provide credentials to the various commands with the --repository command flag:

  • -r --repository <PIPPER_REPOSITORY_NAME>

For more information on how to specify repository configurations for use with this flag, see the repository. This is the recommended way to specify credentials for persistent environments like your local computer.

3. AWS Profiles: Standard AWS profile-based credentials can be used as well. Use the --profile flag to specify the name of the profile you wish to use:

  • -p --profile <AWS_PROFILE_NAME>

4. Pipper Environmental Variables: If none of the previous forms of credentials are provided, pipper will try to use pipper-specific environmental variables:

PIPPER_AWS_ACCESS_KEY_ID

PIPPER_AWS_SECRET_ACCESS_KEY

PIPPER_AWS_SESSION_TOKEN

5. AWS Environmental Variables: If none of the previous forms of credentials are provided, pipper will attempt to use the standard AWS environmental variables:

AWS_ACCESS_KEY_ID

AWS_SECRET_ACCESS_KEY

AWS_SESSION_TOKEN

If neither set of environmental variables exist, pipper will fallback to using the default profile credentials if they exist.

6. Default Pipper Repository Configuration: If none of the other credentials are specified, pipper will try to use the default repository configuration if one exists.

7. System-level credentials: In the end, pipper will try to use the default system-level credentials, which is useful in situations like EC2 instances where the credentials are baked into the instance. However, on remote systems the lack of specified credentials will likely result in authorization exceptions.

Install Action

The pipper command can be used to install packages directly from the command lin in much the same way you install packages with pip. The command is followed by one or more packages to install. Specific package versions can be downloaded by appending the version to package names with a colon separator.

$ pipper install <PACKAGE_NAME[:VERSION]> <PACKAGE_NAME[:VERSION]>

There are a number of flags available to modify how the install command functions:

  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the remote pipper files are stored.

  • -i --input <INPUT_FILE>

    Allows you to load one or more packages from a pipper-formatted JSON file. Use this in place of specifying the packages directly in the command when convenient.

  • -u --upgrade

    When specified currently installed packages will be updated to the latest version. If this flag is not specified the installation process will ignore already installed packages, even if a newer version is available.

When installing pipper packages, pipper dependencies are handled recursively as long as the dependency packages have a properly configured pipper.json file located at the top-level of the repository.

Installation Examples

$ pipper install foo --bucket my_bucket --profile my_profile

Installs the foo package using the default AWS credentials associated with the my_profile AWS profile from the my_bucket S3 bucket.

Download Action

The download action can be used to download pipper packages for later use. This can be helpful if you want to make packages available while offline or when AWS credentials are unavailable.

$ pipper download <PACKAGE_NAME[:VERSION]>
  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the remote pipper files are stored.

  • -d --directory <DIRECTORY_NAME>

    The directory where the pipper bundle file for the package should be saved to when downloaded.

  • -i --input <INPUT_FILE>

    Allows you to download one or more packages from a pipper-formatted JSON file. Use this in place of specifying the packages directly in the command when convenient.

  • -e --extract

    When specified, the downloaded pipper files will be immediately extracted into their consituent wheel and metadata files. Useful if you want to install directly with pip using advanced options such as installing to a specific directly.

Repository Action

The repository action allows you to create and managed named repositories, which can be used to simplify the management of credentials within the command line. The repository command action has a number of sub-actions:

Repository: add

$ pipper repository add <REPOSITORY_NAME>

Adds a new repository configuration with the specified name. Use the -p --profile or -c --credentials flag to specify the AWS credentials to be used by this repository. The add sub-action has other flags:

  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the remote pipper files are stored for this configuration. If the bucket is set in the repository configuration, it will automatically be used by pipper.

  • -d --default

    If this flag is set, this repository configuration will be the default one used when no credentials or other information is specified.

Repository: modify

$ pipper repostory modify <EXISTING_REPOSITORY_NAME>

Modifies an existing repository configuration with new values. This sub-action has the same flags as the add sub-action. Any flags that you set will be used to replace existing values. Any omitted flags will retain their existing values.

Repository: remove

$ pipper repository remove <EXISTING_REPOSITORY_NAME>

Removes an existing repository configuration from the configuration storage.

Repository: list

$ pipper repository list

Use this command to list the currently stored repository configurations. It also lets you know which of the configurations is set to the default value.

Repository: exists

$ pipper repository exists

Displays information on whether or not a repository configuration currently exists.

Authorize Action

There are times when having AWS credentials available isn't practical. To get around those you can create pre-authorized URLs for downloading and installing packages that can be used where credentials are not available.

$ pipper authorize <PACKAGE_NAME[:VERSION]> ...
  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the remote pipper files are stored.

  • -i --input <INPUT_FILE>

    Allows you to load one or more packages from a pipper-formatted JSON file. Use this in place of specifying the packages directly in the command when convenient.

  • -o --ouput <OUTPUT_FILE>

    If specified, a pre-authorized pipper config file will be written that can be used later by download and installation commands.

  • -e --expires <EXPIRES_IN>

    How long the authorized URL is valid before it expires. The format should be <NUMBER><UNIT>, where the number is a positive integer and the unit can be hours, minutes or seconds. Units can be abbreviated, e.g.:

    • 12mins: 12 minutes
    • 130m: 130 minutes
    • 18s: 18 seconds
    • 3hr: 3 hours

Info Action

Prints information on the locally installed and remote versions of the specified package. Also, informs you if there is a newer version of the package available for upgrade.

$ pipper info <PACKAGE_NAME>
  • -l --local

    Only display local package information, which can be useful if you're just looking for what is installed locally and don't want to provide AWS credential information as well.

  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the remote pipper files are stored. This flag is needed unless the local flag is used, which does not communicate with the remote S3 files.

Bundle Action

Creates a pipper package distribution file that can be installed directly or published to a remote S3 bucket for distribution.

$ pipper bundle <PACKAGE_DIRECTORY>
  • -o --output <OUTPUT_DIRECTORY>

    The directory where the pipper bundle should be saved. Defaults to the current working directory.

Publish Action

Deploys a pipper bundle file to a remote S3 bucket for distribution.

$ pipper publish <PIPPER_FILENAME>

If you specify a directory instead of a filename, pipper will search for a pipper file in that directory and upload it. If multiple pipper files are found, the most recently created one will be uploaded.

$ pipper publish <DIRECTORY_CONTAINING_PIPPER_FILE>
  • -b --bucket <BUCKET_NAME>

    Name of the S3 bucket where the package will be published.

  • -f --force

    Unless this flag is specified, publishing a package will be skipped if an identical version of the package has already been published.

Version Locking

Pipper supports version matching/locking in a similar fashion to pip. However, the syntax is a little bit stricter. Values must conform to semantic versions. Consider a library foo. A specific version can be installed using any of the following statements:

  • foo no version will install latest
  • foo:1.2.3 that specific version
  • foo:=1.2.3 that specific version
  • foo:==1.2.3 that specific version
  • foo:1.2.* the latest revision of 1.2.x
  • foo:1.*.* the latest minor version and revision of 1.x.x
  • foo:<1.2.3 any version below the specified one
  • foo:<=1.2.3 any version equal to or below the specified one
  • foo:>1.2.3 any version above the specified one
  • foo:>=1.2.3 any version equal to or above the specified one