xoto3

xoto3 (pronounced zoto-three) is a layer of useful micro-utilities for boto3 (the AWS Python library) particularly suitable for serverless development.

These come from years of experience developing a serverless platform at XOi Technologies, and represent real production code.

Features

Some of the features included:

• A more general purpose boto3 client API paginator.

• Higher-level abstractions for DynamoDB, including:

  • pure data transformation safeguards against various sorts of data that DynamoDB won't accept.
  • an multi-item, multi-table composable, retrying, transaction wrapper for TransactWriteItems, allowing arbitrary writes (up to the built in API limitations) to multiple databases to be expressed as pure Python.
  • a transactional single-item update that allows you to express your single-item update transformation in pure Python.
  • transparent BatchGet and BatchWrite utilities that work around the many annoyances of boto3 and DynamoDB itself.
  • composable query interfaces that make writing basic queries against DynamoDB fun.

  see readme for examples

• Cloudwatch Insights and Log Groups Query URL formatters.

• General-purpose AWS Lambda finalization code, to make sure buffered IO gets a chance to flush before your Lambda gets paused.

• Wrapper for SSM parameter puts and gets, including built-in support for parameter values larger than what SSM will accept by automatically splitting your values and reconstructing them on gets.

Various other utilities are included as well - feel free to poke through the source code.

None of these features "rely" on any of the others, so all of the power is left in your hands. This is not a framework; just a set of mostly pure-functional utilities, with a couple of handy wrappers for boto3 functions that perform IO.

Other Utilities

Some fairly general-purpose utilities are also included. See the readme for more details.

Some highlights:

• tree_map - recursively map through a tree of Python builtins
• lazy - general purpose lazy-loading container
• Various serialization utilities (datetimes, decimals, JSON helpers)
• pipe_multiprocessing - a Process Pool for places like AWS Lambda where Python's built-in shared memory-dependent Pool does not work.

Testing

You can run all unit tests with pipenv run pytest tests.

You can additionally include all the DynamoDB integration tests by setting some environment variables:

XOTO3_INTEGRATION_TEST_DYNAMODB_ID_TABLE_NAME: the name of a DynamoDB table with a primary key that is a partition key of id and no range key. XOTO3_INTEGRATION_TEST_NO_RANGE_KEY_INDEX_HASH_KEY: the name of an attribute which is the partition key of a GSI with no range key.

If you don't currently have a table viable for testing, you can use the following script to easily create one: pipenv run python ./scripts/create_integration_test_table.py

Development

Writing tests

Any new changes should be accompanied by unit tests. Integration tests should also be included where they are helpful.

Integration tests should make use of environment variables and pytest's ability to skip a test in the event that an environment variable is not set.

Additionally, pytest is configured to run all tests in parallel, so any integration test you write must be independent of any other test, and must not leave behind test data.

Your integration tests should be written alongside the unit tests. We do not keep them in a separate folder.