HI Mark, of course I can't have an opinion on such low-level details of the rule engine. But, in general terms, this proposal seems related to a topic we slightly touched in the past, i.e. if it would not be a good idea to create a sort of "experimental" "new feature" or whatever branch for such new things, to be used also for other new stuff. To better reinforce the concept, I strongly believe that we must keep creating new features and experimenting new approaches, but at the same time we have to balance the impact on the principal branch, that should be kept "stable". Reason for that is that currently there is a lot of "abandoned" code in our repositories, sort of POC or attempt that IINW no-one actually used and no-one actually maintains (except if some test gets broken for some unrelated modifications). Again, I can't tell how the code you are writing fits in the general picture, but since you opened a "PROPOSAL", it seems that it is not just a simple reorganization or clean-up. With that "new-feature" branch, curious or interested users could start actually testing such a new code and, later on, we could have a better opinion if the feature is mature enough (and used enough) to be merged (and maintained) in the main branch.
I hope to have been clear. WDYT ? Il giorno mer 4 dic 2024 alle ore 14:51 Mark Proctor <[email protected]> ha scritto: > I have opened a ticket for the Sequential Logic work I'm doing. > https://github.com/apache/incubator-kie-drools/issues/6182 > > I had a slight pause, but I have now resumed this work. > > The sequential engine is a dedicated subsystem that integrates with Rete's > alpha network but uses a state machine rather than Rete's beta network to > drive the sequence matching. The current temporal operators (before, after, > etc.) all require beta joins, which are not as efficient as a state machine > and are not as flexible for complex sequences with loops. > > When a sequence is matched, it enters the beta network for additional > relational joins or rule firing. Future versions will optimise the network > to avoid entering the beta network and fire immediately. > > Under the hood is a dedicated state machine that moves through states when > signals are received from the SignalProcessors; the state machine is > optimised for integrations for the sources it interacts with. Those signals > are provided by logic gates once they become satisfied. The logic gates > matched data received from the event stream, which itself is from the alpha > network. The "feel" of the system is similar to how regular expression > matching works. > > Because it only wants to attempt to match data for the current active > sequence or subsequence, it uses a flat list of current patterns to match > against. Bitmasks are used to get efficient routing from the input of data, > to the pattern matching to the appropriate logic gates and their parent > sequence. > > The branch can be found here: > https://github.com/mdproctor/drools/tree/sequentiallogic > > It's early days, however a lot of the unit tests are in place to show how > it works, but it's currently still very low level. For those interested, a > lot of the unit tests can be found here: > > https://github.com/mdproctor/drools/tree/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing > > > A sequence has 1...n steps. Currently, it supports LogicCircuits, Actions, > Aggregators and Subsequences. > > > https://github.com/mdproctor/drools/tree/sequentiallogic/drools-base/src/main/java/org/drools/base/reteoo/sequencing/steps > > > This test shows how to construct the logic gates, which are used within a > LogicCircuit: > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-base/src/main/java/org/drools/base/reteoo/sequencing/signalprocessors/Gates.java > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerLogicGateTest.java > > > A LogicGate is a type of SignalProcessor: > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-base/src/main/java/org/drools/base/reteoo/sequencing/signalprocessors/LogicGate.java > > > Loops can occur on individual SignalProcesses or on Subsequences. Where > it's on a LogicGate, it tests the number of times the individual LogicGate > is matched. > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerSignalProcessorCounterTest.java > > > Here is the conditional loop for SubSequence: > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerSequenceLoopTest.java > > > Timers are supported as well on signal processors and subsequences. The > timer can be used to add delays or do timeouts. > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerSignalProcessorTimerTest.java > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerSequenceTimerTest.java > > > Lastly, you can do aggregations over subsequences: > > > https://github.com/mdproctor/drools/blob/sequentiallogic/drools-test-coverage/test-compiler-integration/src/test/java/org/drools/mvel/integrationtests/phreak/sequencing/PhreakSequencerAggregatorTest.java > > > The next steps are to improve the unit tests and start to try and expose > this in the DRL language itself so end users can start to use it. > > > Regards > > > Mark > > > > > > > > > > Mark >
