Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - darren.sampson

Pages: [1] 2 3
1
Suggestions and Requests / BUG: Project Integrity Check Completion
« on: October 16, 2006, 11:03:55 pm »
The Project Integrity Check does not indicate when it is complete.

The buttons do not grey while the check is in progress and the Checks Completed: progress bar never completely fills.

2
Suggestions and Requests / Re: IDEA: Stereotyped Template Diagrams
« on: January 09, 2007, 08:54:49 pm »
While extended diagram definitions may be helpful for some situations (such as the Database diagram example), it seems to me that more flexibility in the existing template mechanism may be sufficient for many uses.

I mentioned one such use in my other post ([size=13]Idea: Show Diagram Stereotype in Project Browser[/size])

As we define classes of diagrams, it would be extremely useful to have a template for each class of (say) component diagrams used in the Architecure.

From an outside point of view, it would seem that adding a stereotype combo to the New Diagram dialog would be a simple matter (given stereotype is a property of all diagrams).  From that point, it should be simple to select the particular template to use when creating a new diagram by matching the stereotype.

If this is as small an effort as it seems from an outsider's point of view, it would be really handy while we wait for a more sophisticated solution.

Comments?

ps. I've noticed that diagram notes and text elements don't seem to be copied from the template. Getting a consistent base diagram including things like Diagram Notes would seem to be a typical use for a template.  The properties seem to come over OK though.

3
Suggestions and Requests / Re: Suggestion: Merge on XMI import
« on: November 14, 2006, 04:18:06 pm »
I'd vote for this.

If you have two people working on the same controlled package (one potentially off site or remote), it is currently extremely difficult to merge both their work back together again.

A selective merge that prompts on conflicts would be extremely useful.  At the moment, we would have to use an external diff tool and merge the XMI.


4
Suggestions and Requests / Re: IDEA: Control of connector display
« on: January 10, 2007, 09:23:32 pm »
The current default behaviour (as of 6.5 802) is getting up my nose very fast.

I have a set of class diagrams that have been reviewed, and I'm adding some extra diagrams to show traceability between the classes and the components.  This involves adding links between classes and sub-classes nested in the classes.

These new links show in the reviewed diagrams which I have not touched :o

These reviews are in preparation for a formal EN50128 software audit.  The changes in the diagrams are unlikely to go down well with the auditor!

Please, pretty please fix this annoying feature!!

As has been said:

Quote
Also, we agree that the diagram is not the model.  But it is a view into the model.   Again, from my first post, I'm saying that if I decide what I want to see in any given diagram (that is, the view this diagram is presenting), I don't want EA to muck it up for me by using the current default behaviour.


Stop mucking up my diagrams.  Please!!

Darren


5
Diagram stereotypes should be shown in the project browser.

This is a consistency issue (everything else in the project browser seems to show stereotypes if they exist), and so I'm not really sure if I'd call it an idea or a bug, but I'll stick with idea for now.

We are increasingly using diagram stereotypes to identify common classes of diagrams and hope to use the diagram stereotype for automatic creation and maintenance of the diagrams in the future.  

They will also be more useful if Paolo's idea is accepted - IDEA: Stereotyped Template Diagrams

Due to this, this inconsistency is getting increasingly irritating.

Votes?


6
Quote
Hi Darren,
have you already looked here?


I take it you meant to post that here as Simon suggests. I'd agree with the other thread too - why not kill two birds with one stone?

7
Hi Everyone,

I've just been playing with diagram hyperlinks and it occurred to me that it would be helpful if the hyperlink glyph actually indicated the type of diagram.

This should be simple - the glyph used in the project browser would be perfectly adequate.

Comments? Votes?

Darren.

8
Suggestions and Requests / Word Delimiters in Spell Checking
« on: November 13, 2006, 08:23:15 pm »
We've been setting up a user dictionary which contains all the "names" from the model. We then plan to use this to enforce correct use of the names in notes.

We've noticed that '_',  ':'  characters are delimiters, and the words on each side of the delimiter are checked as expected.

We've also noticed that the '.' character is not a delimiter. This is an issue.

We have two options that we can think of:

1. Insert all combinations of name1.name2 in the dictionary
2. Request delimiters be configurable.

Does anyone know any other options?

While I'm on the topic, the spelling options are only available when you have found the first spelling error - this is a little irritating.

Sparxians: Are you using the Wintertree spelling engine?




9
Hi Simon,

Quote
The solution is to import it again after the information is available.


That's not really a solution - it's a workaround.  

What you're actually saying is that you should always load everything twice to ensure there is no data loss.  Manually checking every diagram is not a realistic option.

It also means if someone else makes changes to a number of controlled packages in their own private EAP file, the only way you can be certain you have everything is to re-load all the changed packages in your own EAP twice.

Please fix this!

Darren

10
I can confirm it is not fixed in 799 - I tried it today while I was looking at the "dirty flag" for controlled packages (in another thread).

It depends on the order you load the packages - some orders work, others don't.

11
I'm very interested in keeping track of this bug fix...

Sparxians:  How is this fix progressing?  When can we expect to see the fix in a build?

Thanks!

12
Thinking more about it, it seems that the following is true:

There is no way to reliably re-create an EAP with multiple project roots with references between elements within the project roots...
at least without knowing a dependency graph of the whole EAP file and analysing it to find an import order.  

If there are cyclic dependencies, then it is impossible.

This is true because you cannot export a single XMI for the whole EAP file.

This means the EAP is (in reality) irreplaceable and not reproducible from XMI.

13
I'm in agreement with Mahaveer - this is a bug.

We use the XMI as our version-controlled "original" for each component.

If we have a package A in an XMI, and a package B in a second XMI and B references an element in A...

If I import A then B into an EAP file, then I have a complete model.

If I import just B into a new EAP (which deletes the reference to the element in A), update an unrelated diagram (as an example) then re-export the XMI, the XMI now no longer has the reference to the element in A. At a later date repeat the process in the paragraph above, I have lost information - the link from B to A.

What Simon is saying is that I have to import both XMI to make any change. In this simple example, that would be OK.  When I've got hundreds of components with complex relationships, this is not realistic.  Even trying to work out what order I should import all the XMIs is near impossible.  If I want to automate this, it's completely impractical.

In fact, cyclic references make it impossible to successfully import anything!

Please fix this - you should never throw away data even if you're not sure what to do with it.

14
General Board / Re: Losing Changes in SVN Controlled Model
« on: August 20, 2008, 11:29:00 am »
Hi guys,

Another thing to note is that XMI import can lose data, so storing a model in more than 1 XMI file (including using EA's version control) is a bad idea unless:

1) You have no connectors between objects in the different XMI packages
2) You are extremely careful!

If you decide to be extremely careful, some guidelines:

EXPORT:
If you export a package with connectors to objects in another package, you must also export the other package.  There is no indication in EA of these connectors, so it is a bit hard to spot.  At Ripple we developed a tool to detect this.

IMPORT:
When importing a set of XMI files with connectors across the packages into a "clean" model, you must import all XMI files TWICE!!  This is a flaw in EA's XMI Import which throws away any connector when the object in the other package does not yet exist.  Importing twice works around this issue.

Hope that helps!
Darren

15
General Board / Re: Subversion - Issue with EA
« on: January 16, 2007, 09:10:20 pm »
Hi!

Have you put a request into Sparx Support here?  If you're a registered user, there is a priority support page as well.

The Sparxians do monitor this board, but it's best to ask them more directly.

d.

Pages: [1] 2 3