[Discussion] Should we clear global sets of headers?

[vepadulano]

The set of headers (and shared libraries) is kept as a global variable after the execution of the graph. Is this intended?

[vepadulano]

In my opinion, this is a problem only in a very specific environment: multiple analysis are running in the same python session but the spark backend is reinstantiated from scratch multiple times. This can happen while running our suite of tests for example. Possibly could also happen if the spark backend runs out of scope but the python session with PyRDF stays the same, e.g. the user writes a standalone function with the analysis