[Brett Connor]

Eclipse 3.1GA
MyEclipse 4.1 GA

I’m finding the sequence diagrammer completely unusable currently; it’s taking me perhaps 10 minutes per message line to find the pixel to start and end on for the tool to accept.
I’ve found a resonable workaround for some situations where the diagrammer refuses to draw the line:
– find two call lifetime boxes that _are_ big enough to be recognised
– draw a message line
– move the line one end at a time to the desired place

However now I’m trying to draw an event line back to an earlier object:

  A                B               C
    ------------>
                     ------------->
    <--------------------------   *2

and after about 15 minutes I’ve given up, I can find no way to draw *2. Is this something that the diagrammer won’t do (lines going R->L)? or is it just a drawing problem again?

Please if there are any tips you can give me to be able to use this diagrammer I would much appreciate it.
Regards
Brett

[Brett Connor]

UPDATE:
I’ve finally got the line to stick to something, but now it refuses to have the two calls coming out of the same execution occurance box. If I drag the message lines up and down they can swap over but there seems nothing I can do to get them both coming out of the same box. Is this a bug or am I missing some trick of the UI?

[Brett Connor]

I now can’t draw a RETURN line, partly perhaps because the diagrammer has split the execution occurance box so if it’s trying to be smart it may think there’s no call to return.
It strikes me that a lot of these problems would go away if you simply let people resize the execution occurance box, and let people draw lines between any two boxes without any kind of checking. It’s not a high level of ambition but it’s not as if you’re trying to generate anything from these diagrams, it would make the diagrammer useful I think.

[Riyad Kalla]

Brett I am sending this thread to the UML team, we appologize for the frustrations you are running into.

[Kapil Kapre]

@brettconnor wrote:

Eclipse 3.1GA
MyEclipse 4.1 GA

I’m finding the sequence diagrammer completely unusable currently; it’s taking me perhaps 10 minutes per message line to find the pixel to start and end on for the tool to accept.
I’ve found a resonable workaround for some situations where the diagrammer refuses to draw the line:
– find two call lifetime boxes that _are_ big enough to be recognised
– draw a message line
– move the line one end at a time to the desired place

You dont need to actually have the start point on the box itself you can use the dashed timeline as well.

However now I’m trying to draw an event line back to an earlier object:

  A                B               C
    ------------>
                     ------------->
    <--------------------------   *2

I believe you should add a B->C and then a B->A return for it to pass the validity check.

[401k]

omg… I’m using 4.1.1 and I feel your pain. I’ve been working on a SINGLE sequence diagram for 2+ hours. Nearly every line has to be attached *at least* 10-20 times before it “sticks’. I finally connected a couple of return lines and then couldn’t go back and place any additional calls between the call and the return. The diagram ended up corrupted somehow after 10 minutes of trying to fix it. Needless to say, I’m searching for another modeling tool while this one is worked on. Right now, it’s simply hideous.

[Kapil Kapre]

rotodaddy,
Could you list the problems enountered with use cases? BTW as I said earlier you dont have to start the line from the box’s edge you can use the actual timeline itself. I think you were probably trying to draw nested calls without any margin between the two messages?

[Brett Connor]

You dont need to actually have the start point on the box itself you can use the dashed timeline as well.

I’ve tried starting on the tiemline, the box, anywhere I could think of. It is at best “unreliable” in my experience.

Re the callback:

I believe you should add a B->C and then a B->A return for it to pass the validity check.

I don’t want to do a return, I want to do a callback (think Visitor pattern for example). I would say this validity check is wrong.

Like rotodaddy I spent almost an entire afternoon trying to do a single diagram – admittedly a little complex, ideally I wanted fragments and conditions as well, but I accept these may not be supported yet.

I am in two minds about whether to persevere with this. I like MyEclipse, I like it’s packaging of functionality, but the UML part of it just isn’t giving me enough steam at the moment. I’m hoping that MyEclipse me in the lead pack as more core modelling behaviour comes with Eclipse, and / or Argo evolves. PLEASE convince me that this is true :-/

[Kapil Kapre]

I don’t want to do a return, I want to do a callback (think Visitor pattern for example). I would say this validity check is wrong.

I think what you need here is an asynchronous call. The returns are synchronous so the current validity check is correct.

I am in two minds about whether to persevere with this. I like MyEclipse, I like it’s packaging of functionality, but the UML part of it just isn’t giving me enough steam at the moment. I’m hoping that MyEclipse me in the lead pack as more core modelling behaviour comes with Eclipse, and / or Argo evolves. PLEASE convince me that this is true :-/

Our goal with MyUML was not to provide a UML tool for hardcore UML users that need to (and can) use all the latest UML features from the latest spec, etc. We tried to create a go-between tool for novice to medium level users who just dont need huge amount of functionality that something like say Rationals tools provide. They typically need a tool to document their ideas or do some reverse engineering, etc. Ofcourse that doesnt mean that we wont dont want to add new features, far from it. It just means while our tool isnt perfect for everyone its just right for others.

[Brett Connor]

I think what you need here is an asynchronous call. The returns are synchronous so the current validity check is correct.

An async call is not what I’m trying to represent – it’s a sync call going on, hopefully clear from a pseudo code:

A::method() {
  B b = ...
  b->someMethod(this);
}

B::someMethod(A caller) {
  caller->areYouSure()
  c->someOtherMethod()
  ...etc

None of these calls are async, A calls B, B messages A amoung other things as part of a single execution occurance. Have I explained this OK?

I understand these tools are not trivial to write – I did modelling tools for several years in a previous life. So far we’re strugging to find any that works, including ones we forked out big wonga for. Can’t blame me for being optimistic, can you?!
I’m hoping that you’ll continue to flesh out functionality on the diagrams you do support. Am I being naive?

[Kapil Kapre]

OK, I dont understand why you need a “return” call from B to A, cant you just use a normal call? I’m trying to understand where our check fails the UML spec.

I understand these tools are not trivial to write – I did modelling tools for several years in a previous life. So far we’re strugging to find any that works, including ones we forked out big wonga for. Can’t blame me for being optimistic, can you?!
I’m hoping that you’ll continue to flesh out functionality on the diagrams you do support. Am I being naive?

No, you *can* expect us to take due note user requests for feature additions / bug fixes, etc. But please keep in mind that most of our development is “user demand” driven so we typically spend time on developing features that will be used by the maximum amount of users (not this case specifically, in general).

[Graham Salmon]

I have also had difficulties with sequence diagrams and messages. I think I have worked out a way to prevent losing my diagram (happened twice now due to weird message behaviour and having to restart).

I am currently unable to attach a call-self message to an object. I managed it twice, but on the second time I was unable to move it anywhere (it needed to be further down) and when I did try to move it, it vanished. I ended up deleting it from the Outline pane, but now I don’t seem to be able to add any more call-self messages, it just does a rubber band box that vanishes on mouse release.

How do I add a call-self message to an object and did I just fluke the first try?

[Graham Salmon]

Okay… I gave up and restarted Eclipse. It turns out that my diagram now has dozens of call-self messages its just it was not showing them when I put them on, so I kept adding them.

[Kapil Kapre]

This looks like a refresh issue. Could you post back all the information from MyEclipse>About>Configuration Summary and errors from your log if any.

[Graham Salmon]

Didn’t notice any error logs, but I have sent the requested config information by PM.

[Author]

Posts