Quicklinks not working in our v16.1 version of our MDG

[Mats Gejnevall]

Hi
We are just moving from v15.1 to v16.1 and after generating our v16.1 MDG the qucklinks have stopped working. Any ideas on what could be the problem. We have been using a list of quicklinks in the profile.

Thanks for any help
Mats

[Geert Bellekens]

are you using the quicklink CSV, or the new metamodel format?

Metamodel works fine for me. I've abandoned the CVS a while ago

Geert

[Mats Gejnevall]

We are using the CSV format. Is the new metamodel format when you model the same stuff as in the CSV file? Tried doing in v15, that but there were some functions that was not avaiable in that format. So we were stuck with CSV.
Cheers
Mats

[Geert Bellekens]

Yes, indeed. You can mostly do the same thing in metamodels as with the CSV, but then 100 times easier and less error prone.

I do remember some stuff that you could do in CSV but not in metamodels, but I never needed any of those.

Geert

[Paolo F Cantoni]

Yes, indeed. You can mostly do the same thing in metamodels as with the CSV, but then 100 times easier and less error-prone.

I do remember some stuff that you could do in CSV but not in metamodels, but I never needed any of those.

Geert

Wot 'e sed!

We also haven't missed the CSV format!  I'd just caution that there are still parts of the new mechanism that are "less than optimal", but it is still way better than the CSV approach.  Search the forum to find where the issues still are.

Paolo

[wikitect]

Funny that. I thought I'd done something wrong. In my MDG I've got c 1,500 lines of a Quicklinker document that has worked for some years i.e. largely unchanged by me. Recently I found that nothing was being suggested. Bizarrely I did a quick test in older versions of EA and those too were no longer working whereas they did previously work.

I've been working on adding a metaconstraint + stereotyped-relationship definition version of the metamodel and this does work. It isn't, however, as good a way of implementing something because yoo have no control over whether you allow a new target element stereotype to be created. In the old CSV you can allow a relationship to be created with an existing target element from another diagram but disallow creating a new target on the 'wrong' diagram. The new meta- based method can't control this - it is based on a combination of the meta-definition and the stereotypes present on the visible toolbox. It would be relatively easy to fix this by adding an attribute / flag on the custom toolbox palette definition to allow / disable new element creation.

Controlling on what diagrams elements can be created is important to providing an understanding of where to go look [particularly where you haven't the tool and a Traceability pane] e.g declaring on a UML Class Diagram before using the Classes elsewhere etc.

Simply stating that the older CSV method shouldn't be used isn't helpful - it should still work.

Luckily for me as the CSV method seems to be broken I haven't yet encountered whatever happens when both methods are present.

[Eve]

Luckily for me as the CSV method seems to be broken I haven't yet encountered whatever happens when both methods are present.

EA assumes that the CSV is there to use with old versions of EA and only uses the metaconstraints. There are exceptions, the notable one being any rows with a diagram filter.

[wikitect]

This just isn't good enough. An unsolicited turning-off of a 1,500 row Quicklinker definition amounts to poor treatement of developers and needlessly then aggravates users of an affected MDG.

Where in the revision history is this 'feature' advertised - https://sparxsystems.com/products/ea/16.0/history.html ?

In addition the meta- contraint / relationship method doesn't provide the fineness of control that the CSV method did (I've just had Sparx Support tell me this). For example in the CSV definition it is possible to allow the creation of a link to an existing target stereotype but prevent creation of a new target stereotype. This provides a much-needed mechanism to ensure that users 'declare' (create) stereotypes in the expected diagram types and thereby help consistency of presentation.

Since the meta- method depends on a set of meta-definitons which define what is possible and the visible toolbox palette stereotypes (which define what is allowable) there is currently no means to prevent creation of a new target element stereotype. Removing it from the toolbox palette then also prevents creation of the triple.

It is perfectly feasible to add an attribute (switch) to the toolbox element definition allowing / disabling new element creation which could then reinstate the control that was removed when the meta method was introduced. [Feature Request]

In addition, if the meta- method is the preferred approach:-
1) where is this stated?
2) what happens where both methods are present? Is the CSV method completely disabled? If not which takes precedence?
3) At which Sparx EA version was the CSV Quicklink mechanism disabled?

[Eve]

You won't find it in the release notes because it has been the behavior since the metaconstraints feature was introduced. (Build 1400)

Yes the metamodel constraints are used in priority to the quicklinker table. It's not formally stated anywhere, but I'll say it here. If you care about having your technology behave consistently within itself or to the rest of the tool you should replace the quicklinker table with the metamodel constraints.

On the other hand, we don't force anyone to upgrade to the better method. If you feel like you can create a better technology by using the quicklinker table, feel free.

[wikitect]

So we're back to the bug observed by Mats Gejnevall that the CSV-defined Quicklink now doesn't work

[Eve]

So we're back to the bug observed by Mats Gejnevall that the CSV-defined Quicklink now doesn't work

No. We're back to a report by one user that a CSV-defined Quicklink now doesn't work.

We don't know what the cause is, which means we don't know that the cause is a bug. There is not enough information here to determine what the problem is and calling it a bug is pure speculation.

I don't know if Mats Gejnevall reported the issue and provided their technology to support, but if they haven't that is the next step.

[wikitect]

No - as I said it's affected to my TRAK MDG to what was a (large) set of unchanged CSV quicklink definitions.

Looks as though it'll be me feeding the machine and submitting a bug report then....

[Eve]

No - as I said it's affected to my TRAK MDG to what was a (large) set of unchanged CSV quicklink definitions.

I'm confused. I thought the answer to your issue was that your technology had been changed to incorporate the metaconstraints, which you didn't realize would disable the corresponding portions of your quicklink CSV.