Improved, explicit model checker

[danionita]

The checker:

• checks for errors such as those listed in Sect 4.2 (page 64) of the user manual. Specifically, pages 64, 65 for now.
• checks if there are unconnected ports of any kind (also starts/ends, gate ports, etc)
• checks formulas for correct syntax and resolvable references. This is easily achieved by attempting to create an evaluatedModel. However, error reporting for this operation needs to be beefed up for this to be useful.
• checks that, for each AND-gate, all incoming Connection Elements have the same occurrence rate.
• checks that all connected ports have the same value object.

It should run

• Manually (button in the Model menuItem). This shows a complete list of what is wrong.
• Automatically before every fraud generation/profitability analysis. This shows a warning, with three options: "Show errors", "Ignore" and "Cancel".

[bobismijnnaam]

Can you send me the pdf for the manual? I can't find it in my gmail :(

[bobismijnnaam]

I'm thinking about the error "Cannot find end-stimulus when calculating path” (starting at start stimulus [ID])". If this is the case, you have either a flow collision ("cannot connect end stimulus as down/up") or an unconnected signal port (an error we check for). Does the error "cannot find end stimulus" still make sense (do we need it) as a separate error or do the two earlier mentioned errors describe all the cases in which an end stimulus might not be found?

[danionita]

I think the more granular errors make more sense. The only other situation in which an end-stimulus is missing - except if it is not connected or if a start stimulus is connected instead - is if there is a loop. Do we or can we check for this?

On Sun, Oct 16, 2016 at 8:20 PM, Bob Rubbens notifications@github.com wrote:

I'm thinking about the error "Cannot find end-stimulus when calculating path” (starting at start stimulus [ID])". If this is the case, you have either a flow collision ("cannot connect end stimulus as down/up") or an unconnected signal port (an error we check for). Does the error "cannot find end stimulus" still make sense (do we need it) as a separate error or do the two earlier mentioned errors describe all the cases in which an end stimulus might not be found?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/danionita/e3tools/issues/67#issuecomment-254064007, or mute the thread https://github.com/notifications/unsubscribe-auth/AKjEpAflKDQEoZoyZmFjRaThHdmS88rJks5q0mr4gaJpZM4KK6Op .

[bobismijnnaam]

I was planning to write something that detects a loop. So then the error "end stimulus not found" can never appear I guess.

On 18 October 2016 at 11:11, danionita notifications@github.com wrote:

I think the more granular errors make more sense. The only other situation in which an end-stimulus is missing - except if it is not connected or if a start stimulus is connected instead - is if there is a loop. Do we or can we check for this?

On Sun, Oct 16, 2016 at 8:20 PM, Bob Rubbens notifications@github.com wrote:

I'm thinking about the error "Cannot find end-stimulus when calculating path” (starting at start stimulus [ID])". If this is the case, you have either a flow collision ("cannot connect end stimulus as down/up") or an unconnected signal port (an error we check for). Does the error "cannot find end stimulus" still make sense (do we need it) as a separate error or do the two earlier mentioned errors describe all the cases in which an end stimulus might not be found?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/danionita/e3tools/issues/67#issuecomment-254064007, or mute the thread https://github.com/notifications/unsubscribe-auth/ AKjEpAflKDQEoZoyZmFjRaThHdmS88rJks5q0mr4gaJpZM4KK6Op .

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/danionita/e3tools/issues/67#issuecomment-254451214, or mute the thread https://github.com/notifications/unsubscribe-auth/AC8IaMzzvHWhBWlkOoG3C4JEZkN1UYZdks5q1I1bgaJpZM4KK6Op .

[danionita]

What is the status of this? I see the button in the menu but when I press it something is outputted on System.out but nothing happens in the GUI.

[bobismijnnaam]

Implemented as of de2047e0801f8c58afff8ffcdb9f77ca78e4393e.

[danionita]

Issues so far:

• in this model, highlighting does not work.
• the model checker should run before any of the analyses (profitability, fraud generation, NCF

[bobismijnnaam]

@danionita Fixed your earlier two points. Is this enough for now?

[danionita]

Looks like it. If further testing reveals other bugs, we can re-open the issue.