Unknown random distribution type

[adamburkegh]

The UNKNOWN distribution type which allows reading properties of unsupported distributions. Examples seen in stochastic process mining such as the StochasticPetriNets plugin in ProM include GAUSSIAN_KERNEL, DETERMINISTIC, GAMMA and HISTOGRAM. It is useful to be still be able to read weights, priorities, and so on in pm4py even if full support is not available.

An example snippet produced from running that SPN plugin on the BPIC2013_closed.xes reference log:

         <transition id="n14">
            <graphics>
               <dimension x="25.0" y="20.0"/>
            </graphics>
            <toolspecific tool="StochasticPetriNet" version="0.2">
               <property key="distributionType">GAUSSIAN_KERNEL</property>
               <property key="trainingData">duration;systemLoad;timestamp
0.0;506;1316178833000
8.659166666666666;730;1326434536000
1.7252777777777777;441;1329927075000
</property>
               <property key="weight">1475.0</property>
               <property key="invisible">false</property>
               <property key="distributionParameters">0.0;1.7252777777777777;8.659166666666666</property>
               <property key="priority">0</property>
            </toolspecific>
            <name>
               <text>Completed</text>
            </name>
         </transition>

I don't expect pm4py to drop everything and start supporting all distribution types. I'm not sure if it's even a good idea. But since there doesn't seem to be a central standard for extended stochastic net distributions, it is definitely useful if pm4py doesn't throw away what properties do exist and can be read. In my current use case, I don't need to perform any direct simulation, but I do need to be able to describe the net in question.

This is my first PR for pm4py. I am very open to feedback about alternative ways to do this, or pointers if I've just misunderstood the current capabilities of pm4py.

I ran the unit tests before and after. 4/326 failed both before and after. The failing tests all appear to be related to extensions outside core not affected by this patch. The SPN importer doesn't seem to have specific coverage, and that isn't changed by this patch.

Abbreviated test output (can share full on request)

ERROR: test_bpmn_importing_and_layouting (tests.bpmn_tests.BPMNTests)
ERROR: test_bpmn_layouting (tests.bpmn_tests.BPMNTests)
ERROR: test_emd_1 (tests.other_tests.OtherPartsTests)
ERROR: test_emd_2 (tests.other_tests.OtherPartsTests)

Ran 326 tests in 28.805s

FAILED (errors=4)

Cheers
Adam