This package provides supplementary metadata generation for registry documents, which is required for registry-api to function correctly, and for common user queries. Execution is idempotent and should be scheduled on a recurring basis.
The repairkit sweeper applies idempotent transformations to targeted subsets of properties, for example ensuring that all properties expected to have array-like values are in fact arrays (as opposed to single-element arrays being flattened to strings during harvest). Documents are processed based on whether their ops:Provenance/ops:registry_sweepers_repairkit_version metadata value is up-to-date relative to the sweeper codebase.
The provenance sweeper generates metadata for linking each version-superseded product with the versioned product which supersedes it. The value of the successor is stored in the ops:Provenance/ops:superseded_by property. This property will not be set for the latest version of any product. All documents are processed, but db writes are optimised based on whether their ops:Provenance/ops:registry_sweepers_provenance_version metadata value is up-to-date relative to the sweeper codebase.
The ancestry sweeper generates membership metadata for each product, i.e. which bundle lidvids and which collection lidvids reference a given product. These values will be stored in properties ops:Provenance/ops:parent_bundle_identifier and ops:Provenance/ops:parent_collection_identifier, respectively. All bundles/collections are processed to populate a lookup table, but db writes are optimised based on whether their ops:Provenance/ops:registry_sweepers_provenance_version metadata value is up-to-date relative to the sweeper codebase, and collection non-aggregate reference pages in registry-refs are skipped entirely if they are marked as up-to-date.
Accepts environment variables to tune performance, primarily trading increased runtime duration for reduced peak memory usage.
- Python >=3.9
MULTITENANCY_NODE_ID= // If running in a multitenant environment, the id of the node, used to distinguish registry/registry-refs index instances
PROV_CREDENTIALS={"admin": "admin"} // OpenSearch username/password, if targeting an OpenSearch host other than AWS AOSS
SWEEPERS_IAM_ROLE_NAME=<value> // AWS IAM role name, if targeting AWS AOSS
PROV_ENDPOINT=https://localhost:9200 // OpenSearch host url and port
LOGLEVEL - an integer log level or anycase string matching a python log level like `INFO` (optional - defaults to `INFO`))
DEV_MODE=1 // disables host verification
// tqdm dependency may cause fatal crashes on some architectures when breakpoints are used in debug mode with Cython speedup extension enabled
PYDEVD_USE_CYTHON=NO // disables Cython speedup extension
With --legacy-sync option, you also need the list of the cross-cluster-search node configured to access all the node's OpensSearch domains:
CCS_CONN=naif-prod-ccs,rms-prod,sbnumd-prod-ccs,geo-prod-ccs,atm-prod-ccs,sbnpsi-prod-ccs,ppi-prod-ccs,img-prod-ccs
Use the connection aliases found in the 'Connections' tab of the Engineering Node OpenSearch Domain on AWS.
After cloning the repository, and setting the repository root as the current working directory install the package with pip install -e .
The wrapper script for the suite of components may be run with python ./docker/sweepers_driver.py
Alternatively, registry-sweepers may be built from its Dockerfile with docker image build --file ./docker/Dockerfile . and run as a container, providing those same environment variables when running the container.
When run against the production OpenSearch instance with ~1.1M products, no cross-cluster remotes, and (only) ~1k multi-version products, from a local development machine, the runtime is ~20min on first run and ~12min subsequently. It appears that OpenSearch optimizes away no-op update calls, resulting in significant speedup despite the fact that registry-sweepers reprocesses metadata from scratch, every run.
The overwhelming bottleneck ops are the O(docs_count) db writes in ancestry.
All users and developers of the NASA-PDS software are expected to abide by our Code of Conduct. Please read this to ensure you understand the expectations of our community.
To develop this project, use your favorite text editor, or an integrated development environment with Python support, such as PyCharm.
For information on how to contribute to NASA-PDS codebases please take a look at our Contributing guidelines.
Install in editable mode and with extra developer dependencies into your virtual environment of choice:
pip install --editable '.[dev]'
Configure the pre-commit hooks:
pre-commit install
pre-commit install -t pre-push
pre-commit install -t prepare-commit-msg
pre-commit install -t commit-msg
These hooks check code formatting and also aborts commits that contain secrets such as passwords or API keys. However, a one time setup is required in your global Git configuration. See the wiki entry on Git Secrets to learn how.
To isolate and be able to re-produce the environment for this package, you should use a Python Virtual Environment. To do so, run:
python -m venv venv
Then exclusively use venv/bin/python, venv/bin/pip, etc.
If you have tox installed and would like it to create your environment and install dependencies for you run:
tox --devenv <name you'd like for env> -e dev
Dependencies for development are specified as the dev extras_require in setup.cfg; they are installed into the virtual environment as follows:
pip install --editable '.[dev]'
All the source code is in a sub-directory under src.
This section describes testing for your package.
A complete "build" including test execution, linting (mypy, black, flake8, etc.), and documentation build is executed via:
tox
Your project should have built-in unit tests, functional, validation, acceptance, etc., tests.
For unit testing, check out the unittest module, built into Python 3.
Tests objects should be in packages test modules or preferably in project 'tests' directory which mirrors the project package structure.
Our unit tests are launched with command:
pytest
If you want your tests to run automatically as you make changes start up pytest in watch mode with:
ptw
pip install wheel
python setup.py sdist bdist_wheel
NASA PDS packages can publish automatically using the Roundup Action, which leverages GitHub Actions to perform automated continuous integration and continuous delivery. A default workflow that includes the Roundup is provided in the .github/workflows/unstable-cicd.yaml file. (Unstable here means an interim release.)
Create the package:
python setup.py bdist_wheel
Publish it as a Github release.
Publish on PyPI (you need a PyPI account and configure $HOME/.pypirc):
pip install twine
twine upload dist/*
Or publish on the Test PyPI (you need a Test PyPI account and configure $HOME/.pypirc):
pip install twine
twine upload --repository testpypi dist/*
The template repository comes with our two "standard" CI/CD workflows, stable-cicd and unstable-cicd. The unstable build runs on any push to main (± ignoring changes to specific files) and the stable build runs on push of a release branch of the form release/<release version>. Both of these make use of our GitHub actions build step, Roundup. The unstable-cicd will generate (and constantly update) a SNAPSHOT release. If you haven't done a formal software release you will end up with a v0.0.0-SNAPSHOT release (see NASA-PDS/roundup-action#56 for specifics).