Current implementation of a dep solving supports only strong (@Requires:@) simple (@package_name@) dependencies.
In order to support strong (@Requires:@) rich (aka "boolean":http://rpm.org/user_doc/boolean_dependencies.html) dependencies an external library should be used.
The suggestion is to drop the current implementation of dep solving and for any kind of a dep solving start using:
- either @libdnf@ via its Python interface (python2-dnf)
- or @libsolv@ via its Python bindings (python2-solv)
@libdnf@ uses @libsolv@ and exposes a higher level API to a user.
Both @libdnf@ and @libsolv@ requires @repodata@ to perform dep solving.
Some temporary working directory should be provided to a solver to store its cache there.
Functionality of a solver we are looking for: Use cases to keep in mind:
* list dependencies for a given rpm package (can be a part of a synchronous task)
* list fulfilled dependencies (from a given list of available packages) for a given rpm package
* validate that all the dependencies (from a given list of available packages) for a given rpm package are fulfilled.
* perform dep solving against multiple repositories
The actions mentioned above will be a part of an asynchronous task (see a copy example below).
An example of the dep solving use case in Pulp is a @recursive@ flag during a copy: https://docs.pulpproject.org/plugins/pulp_rpm/user-guide/recipes.html?highlight=recursive#copy-errata-from-one-repository-to-another
Dep solving related functionality to support: https://pulp.plan.io/issues/2478
Example usage of the @libsolv@ Python bindings: https://github.com/fedora-modularity/depchase
Ping ttereshc for more details about examples for @libdnf@ case.
Figure out which library supports all required functionality. Keep in mind that performance is important as well. There could be repositories with 100K+ packages against which dep solving can be requested.