Alignment: TypeError: unsupported operand type(s) for +: 'int' and 'NoneType' about pm4py-core

[s-j-v-zelst]

Hi @fmannhardt,

I have looked at this together with @Javert899
Minor side note, alignments can actually be computed on easy sound WF nets, which are non-relaxed ;-)

Anyways, the problem was, that I didn't check in the code if a solution existed for the heuristic.
Clearly, if there is no solution, the heuristic should be set to infinity (or MAX INT).
At the moment, the heuristic was None and then the addition of the heuristic to the cost traveled so far was breaking.

From a theoretical perspective, this implies that the code was only able to handle sound WF nets indeed.
We have released this fix in 1.1.1

[Javert899]

Hi,

The Petri net looks a sound workflow net to my eye, but the initial marking is set as empty, and that is not correct.

[fmannhardt]

I see, there was an issue with my exporting code and now there are both markings in the file.
Anyway, still the same error so it seems something to be wrong with some other part.

This is the correct Petri net:
heur_v2.zip

Note that alignments do not require sound Workflow nets as input, should work for any Petri net. It is just not guaranteed to finish ;)

[Javert899]

Actual implementation in PM4Py, that is based on an Heuristic based on resolving the marking equation, requires at least relaxed soundness ..

Obviously, other implementations could be considered in future ;)

[fmannhardt]

Yes, that is what I meant. Non relaxed sound ones cannot be aligned by definition.
Anyway, there seems to be something else wrong regarding the example I uploaded.

[fmannhardt]

Was this fixed? I tried it again, but it results in the same error.

[Javert899]

Unfortunately there is no much we can do on the current version if the net is not sound

In some time from now, a version that is able to do alignments on accepting Petri nets, although with a much larger number of states evaluated, will be worked

See issue #62

[fmannhardt]

Did you have a look at the last files I uploaded:
https://github.com/pm4py/pm4py-source/files/2884165/heur_v2.zip

This has the correct initial and final marking. it is not a sound workflow net, but definitely a relaxed sound Petri net as required by alignments --- at least in its original definition.
Does the PM4PY implementation only work for sound workflow nets?

[fmannhardt]

Thanks. All those definitions, what was easy sound again? ;)

[s-j-v-zelst]

there just has to exist a path from the initial marking to the final marking (easy sound).

relaxed sound is something along the lines of we are able to enable each transition from the initial marking and after firing such transition we can still reach the final marking.

[fmannhardt]

Oh, then easy sound is what I call relaxed sound for Data Petri nets. :-)