A reverse solver implementation for project Thoth.
Thoth's solver can resolve requirements of Python packages that are subsequently synced into Thoth's knowledge base. Note this solver operates only one way - from a package to its dependencies. This causes out-of-date issues when a new release of a dependency is released which would satisfy version specification of an already solved package. The reverse solver deals with this issue.
The reverse solver checks what packages depend on the given package (dependents) and re-computes dependency information so that the dependency graph stored in Thoth's knowledge base is up to date - considering releases.
This is tightly bound to over-pinning and under-pinning issues often seen in Python packages.
See this intro video for more info.
This is suitable only for development purposes.
The implementation talks to OpenShift to retrieve solvers installed in a deployment. You need to be logged into OpenShift cluster and have at least view permissions to a Thoth deployment (thoth-test-core in this example):
THOTH_INFRA_NAMESPACE=thoth-test-core KUBERNETES_VERIFY_TLS=0 PYTHONPATH=. pipenv run python3 ./app.py --package-name tensorflow --package-version 2.0.0
The output of the above command is a JSON document stating what packages depend on TensorFlow in version 2.0.0. Note the Python ecosystem does not provide any way how to depend on a particular build available on different package indexes.
The reverse solver is run in the cluster as part of solver workflow when a new package release is detected.