Petri net refactor #21

AdamBanham · 2025-02-16T03:56:39Z

Code related to Petri nets in the repo has been broken down into separated files in a new module petrinets. The refactor introduces the following existing structures across the following inner modules.

Net types:

PetriNets (pn):
- Places, Transitions (without weight now), Arcs
  - To keep backwards compatibility, you can pass ids as ints, or strings. But UserWarnings will appear if you use int ids when exporting.
  - To keep previous behaviour, when hashing or equality check occurs we use the str of the id, i.e. str(n.id) is always used.
- LabelledPetriNet
  - Now precomputes presets and postsets. Tthe methods preset(node) and postset(node) return the set of preset/postset nodes based on the given inpuit.
  - Markings have been removed from structure.
- PetriNetMarking
  - A simple structure to store what places are marked
  - Introduced some shortcuts to modify
  - place in marking, is place marked in the marking
  - marking[place], returns the marking of that place or zero
  - for place in marking, iterates over the marked places in the marking
  - mark + omark, joins the markings over their marked places
  - mark - omark, reduces all marked places in mark by their counterpart in omark, no negatives allowed
  - mark << omark, check whether mark is a subset of omark.
- AcceptingPetriNet
  - A simple structure to store information about initial and final markings
- PetriNetSemantics
  - A structure to enforce the execution semantics of the net, takes the AcceptingPretiNet and a marking.
- ExecutablePetriNet
  - A simple structure to store all information to execute runs over a given PetriNet.
- get_execution_semantics
  - A simple shortcut function to jump from AcceptingPetriNet to executable version.
Data Petri Nets (dpn)
- GuardedTransition
  - Adds a guard to the transition
- and thin wrappers that extend pn counterparts and update hint types:
  - PetriNetWithData, AcceptingDataPetriNet, DataPetriNetSemantics, ExecutableDataPetriNet, get_execution_semantics
Weighted Petri Nets (wpn)
- WeightedTransition
  - adds a weight to transition
- The BuildablePetriNet has been moved into this file and updated to keep original behaviour of making transitions with weights.
- and thin wrappers that extend pn counterparts and update hint types:
  - WeightedPetriNet, WeightedAcceptingPetriNet, WeightedPetriNetSemantics, ExcutableWeightedPetriNet, get_execution_semantics

Import and Exporting pnml files:

reading read:
- parse_pnml_into_lpn, reads in a LabelledPetriNet or AcceptingPetriNet if marking information is present.
- parse_pnml_into_wpn, read in a WeightedPetriNet or WeightedAcceptingPetriNet -- same as above
- parse_pnml_for_dpn, read in a PetriNetWithData or AcceptingDataPetriNet -- same as above
exporting export:
- export_net_to_pnml, exports the given net with as much information as possible, works for all types.

I have moved guard implementation for dpns to be in the same dir and moved all other existing methods in the previous petrinet.py file to into either utils.py or dot.py. There is another file I created for prettier dot files for lpns in the upper directory, but unsure whos should be the goto.

Leaving the possible refactor for reading and importing using a chain of responsibility for another day.

@adamburkegh you may wish to suggest stochastic over weighted, let me know.

While refactoring, I ensured that all previous behaviour expected by the current set of tests is passing. Updated the tests to reflect the file changes and refactoring for types of nets.

the pain, he knew! refactored weights/guards Update test_net_export.py

…alas into petri-net-refactor

adamburkegh

As it's mostly code moves, I think it will be much less painful if we fix the small things I've highlighted and then commit. I really don't want to end up with a snowball patch from hell.

pmkoalas/models/petrinet-redesign/model/petrinet-redesign.unotation

pmkoalas/models/petrinets/dot.py

pmkoalas/models/petrinets/export.py

pmkoalas/models/petrinets/read.py

adamburkegh · 2025-02-18T06:54:39Z

pmkoalas/models/petrinets/wpn.py

+    def weight(self,value:float) -> None:
+        self._weight = value
+
+    # data model methods


technically data model is a bit more of a general term, but I use this term to refer to emulating behaviour https://docs.python.org/3/reference/datamodel.html#emulating-generic-types. But the data model is very broad.

pmkoalas/models/petrinets/wpn.py

adamburkegh · 2025-02-18T06:58:40Z

pmkoalas/models/pnfrag.py

    return PetriNetFragmentParser().create_net(net_title,net_text)

-def parse_net_fragments(net_title:str, *fragments:str) -> LabelledPetriNet:
+def parse_net_fragments(net_title:str, *fragments:str) -> WeightedPetriNet:


I think it may not make sense in the same patch, but the better behaviour here would be to return a different type based on the text passed in. So if you don't included weighted transitions, you don't get a weighted net.

Interesting, what if I mix both? Keeping it simple may be easier for the future. Outside of type instance checks the wpn is functionally equivalent, so the user does not lose anything in returning a wpn.

I guess you are right that WPN is a direct subclass of PN so it is ok at the moment. It would also be good to keep it open to extension (eg a DPN) without breaking the API contract.

In terms of mixing both, I guess you mean weighted and unweighted in the same net, that would not be a valid WPN though.

tests/models/test_petrinet.py

adamburkegh · 2025-02-19T05:49:37Z

pmkoalas/models/petrinets/export.py

+        if isinstance(tran.tid, int):
+            wstring = f"Transition {tran} has an integer id, this is not " \
+                +"recommended, importing into others may break."
+            warn(wstring)


Having something allowed but spitting warnings all over I/O is irritating. It would be better not to warn, but parameterize whether to override with ProM-compatible ids.

would you be okay with being a debug-level message or gone all together?

I would prefer to kill the I/O side effect and the complicating runtime conditional. But it would be good to mention this integer / uuid stuff as part of a package doc comment that describes the design context. Eg how some of the important importers work and how that constrains the way people may want to use the export function. But also the advantages of simple integer ids during development and other uses.

fair enough, maybe we should consider that as a general tennet? Removed warns and adding some design considerations to pn.py.

pmkoalas/models/petrinets/export.py

AdamBanham · 2025-02-19T11:28:10Z

@adamburkegh LGTY enough for a merge onto main and we deal with anything else directly?

adamburkegh

Made some more small comments, another round of review seems excessive though. If they seem reasonable go ahead and push to main when ready

pmkoalas/models/petrinets/export.py

pmkoalas/models/petrinets/read.py

adamburkegh · 2025-02-20T00:41:34Z

pmkoalas/models/petrinets/utils.py

I know that's my comment about utilities, but the first function is weighted net specific, and the second is not specific to pn, so maybe this isn't a win

Unclear instructions.

AdamBanham added 6 commits February 14, 2025 17:37

init discussion

b66ad52

the pain, he knew!

e7ba562

refactored weights/guards

eef9f36

Update test_net_export.py

407bad4

init discussion

46533b9

the pain, he knew! refactored weights/guards Update test_net_export.py

Merge branch 'petri-net-refactor' of https://github.com/AdamBanham/Ko…

c4f0e45

…alas into petri-net-refactor

AdamBanham requested a review from adamburkegh February 16, 2025 03:56

adamburkegh reviewed Feb 18, 2025

View reviewed changes

AdamBanham and others added 3 commits February 18, 2025 20:46

adjustments in pr#1

b2cc24a

oops

754497c

Change expected failure to asserting exception

c972ba2

adamburkegh reviewed Feb 19, 2025

View reviewed changes

adjustments based on pr#2

f27a084

adamburkegh requested changes Feb 20, 2025

View reviewed changes

moving throat clearing

2d3c3f9

AdamBanham merged commit 3333f6b into main Feb 20, 2025
5 checks passed

Petri net refactor #21

Petri net refactor #21

Uh oh!

Conversation

AdamBanham commented Feb 16, 2025

Uh oh!

adamburkegh left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AdamBanham Feb 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

AdamBanham commented Feb 19, 2025

Uh oh!

adamburkegh left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

AdamBanham Feb 18, 2025 •

edited

Loading