In a graph, should a reentry transition show only trigger specific entry methods?

[leeoades]

Consider a state machine with states A and B.

• Trigger X transitions from A to B
• State B has 2 entry functions
  • OnEntry for any trigger called OnEveryEntry
  • OnEntryFrom(Trigger.X) called OnEntryFromX
• B allows Trigger X to reentry into B.

So from State A, on Trigger.X, we expect OnEveryEntry and OnEntryFromX to fire.
In State B, on Trigger X, we expect the same.

The Digraph generated is as such:

digraph {
compound=true;
node [shape=Mrecord]
rankdir="LR"
"A" [label="A"];
"B" [label="B|entry / OnEveryEntry"];

"A" -> "B" [style="solid", label="X / OnEntryFromX"];
"B" -> "B" [style="solid", label="X / OnEveryEntry"];
 init [label="", shape=point];
 init -> "A"[style = "solid"]
}

I would argue that the reentry transition currently labelled

X / OnEveryEntry

would be more usefully labelled as

X / OnEntryFromX

The State B already displays entry / OnEveryEntry to denote all entries.
The label on the X reentry repeats this information, and neglects to mention OnEntryFromX.

So I would expect this:

Thoughts?

[leeoades]

A possibly cleaner representation would be to declare a conditional entry as such.

digraph {
compound=true;
node [shape=Mrecord]
rankdir="LR"
"A" [label="A"];
"B" [label="B|entry / OnEveryEntry\nentry[X] / OnEntryFromX"];

"A" -> "B" [style="solid", label="X"];
"B" -> "B" [style="solid", label="X"];
 init [label="", shape=point];
 init -> "A"[style = "solid"]
}

This cleans up the labels on the transition arrows quite nicely, whilst also grouping the entry functions together.

[mclift]

Hi @leeoades,

Well spotted, this is definitely not what we should expect. I can see some spurious logic in GraphStyleBase.FormatAllTransitions, where it selects the entry action: https://github.com/dotnet-state-machine/stateless/blob/dev/src/Stateless/Graph/GraphStyleBase.cs#L83. Could this also be the cause of issue #509?

Additionally, the re-entry action name gets dropped entirely if you remove the entry action for any trigger:

digraph {
compound=true;
node [shape=Mrecord]
rankdir="LR"
"A" [label="A"];
"B" [label="B"];

"A" -> "B" [style="solid", label="X / OnEntryFromX"];
"B" -> "B" [style="solid", label="X"];
 init [label="", shape=point];
 init -> "A"[style = "solid"]
}

Regarding the alternative representation you've proposed, this looks clearer to me too.

Do you have a fix in mind?

[leeoades]

@mclift - Wait? You want me to actually code it up too? 😂
Happy to give it a bash. Wanted to run the ideas past the collective first though as I'm a noob here.

[mclift]

@mclift - Wait? You want me to actually code it up too? 😂

😁 only if you'd like to!

I'll try to take a look myself soon otherwise.

[leeoades]

@mclift Regarding the fix to #509, I believe it is more correct to store the reference to the trigger so that it can be interrogated later rather than simply adding a check for it inside the delegate.
f755ddf#diff-0de579f20c3805bba5083d369cbe249e3df5328c0c35c8c029f7239c92db7d00
Anyway, that's been merged in now so that's nice.