Topics

1

Bugs and Issues / Publishing patterns does not consider package hierarchy

« on: June 06, 2017, 06:31:46 pm »

Sometimes good quick ideas turn into medium high effort ones if your enthusiasm is doing the front runner dance with reality 

The idea:
Create a pattern for virtual model documents to have the same documentation structure at hand over and over again.

The challenge:
To have a common header with history, etc. these have to be stored in a document artifact inside a package to be colelcted by the document generator.

What has been accomplished:
Setting up the structure of model documents (empty) to be filled with packages later on (after expanding the pattern).
Setting up a package with the document artifact containing the history.
Creation of a diagram with all elements on it including the package with the artifact plus the artifact itself separately on the diagram (it will otherwise not be saved with the pattern).

So the structure is:

model document 1 (referencing the header package below)
model document 2 (empty)
model document 3 (empty)
Header package
|
 ---Artifact with history

On the diagram all 5 objects are placed flat and the diagram is saved as a pattern.xml

After importing the pattern and applying it to a diagram the following is observable:

• The model documents are in place including tagged values and ready to be filled for documentation. This is nice. Check.
  
• The header package is present. Check.
  
• The artifact is no longer inside the package. Not good.
• The package is no longer referenced in the model document. Not good.

To conclude: Creating patterns does not consider package hierarchies as they are saved from diagrams. A package in turn does not automatically include its contents and putting the content flat in the diagram leads to having it reflected flat in the pattern.

I am not tempted to state this is a bug. Currently I felt this to be a great idea to handle a certain use case of mine to solve a specific problem and I have found my expectations are not matched.

Therefore I will not file a bug report but raise this as a note to everyone expecting recursive inclusion of items in patterns. It does not work.
I am pretty much convinced there are more important problems Sparx should take care of and I even know a few of these.

Oliver

2

Automation Interface, Add-Ins and Tools / Document reporting the challenging way

« on: May 30, 2017, 07:41:04 pm »

For some years now I am riding the document generation challenge through several companies and versions of EA and I can proudly say that I have mastered various cliffs to gather lots of expertise in that field. Still I have to admit the EA still surprises me with good old bugs I have encountered before and new issues I haven´t seen before.
Lots of things have changed in the last years to the positive, the editor has new features improving the template creation experience, we now have fragments (great thing) and stylesheets (also great thing).

On the other hand we still do not have  a reliable way of creating documents which look the way they should. First of all  I can still fool the editor (or the editor fools me depending on the point of view) by moving around items up and down the sections list which suddenly leads the editor to extend the SSBookmark style to sections it does not belong to thus marking these as protected text. The other way round I can remove line breaks under certain conditions in the template which also confuses the editors handling of the SSBookmark style. OK, I can get around that, this one is an old one. I got used to.
Another old topic is the section numbering between MS Word and EA which often enough messes up and requires careful fine tuning between the styles on both sides, especially (at least it seems so) when non-English installations are involved.

But a very new friend of mine appeared just today: I am able to create the (virtual) document using the same style, templates and stylesheet as RTF, Word and PDF and they look completely different.
First of all the .docx has lost all styles in the process. No styles (Heading 1,...) can be seen in the style bar if I am using our custom stylesheet. No, the sheet is ok. It has all those styles in it and I can see it in EA in the preview. The styles in turn are correct when rendering the same document as RTF. In RTF on the other hand there are no longer tabs and spaces in the style after the numbering which in turn the docx has. Fun fact: The chapter numbering in both documents starts with 1, but not in their ToC where Heading 1 starts with 3 which in turn is nowhere set in any list level and list override.
The docx in turn has no level numbering below level 1.
And the pdf document in general starts numbering with 3 while looking almost perfect in any other aspect (if we ignore the fact that the image caption now uses level 1 numbering which is also never set in any style in any template).

Don´t get me wrong, EA is still a superior product and the document generation is rather helpful. On the other hand it is highly disappointing to see the old problems still being alive and that I still have to find out which issue is caused by me and which one is EA specific by time-consuming investigations. I wished I could solely focus on improving templates and not chase issues between myself, EA and MS Word 2010.


Oliver

3

Suggestions and Requests / Disarm Restore to Baseline option

« on: September 21, 2009, 10:37:17 pm »

OK, that one has been on my focus for a while and now I find it is time to kick a feature request.
The button "Restore to Baseline" has to be disarmed. I consider this one being one of the most dangerous items in the whole EA.
This element is placed so close below the "Show differences" button that somebody with the appropriate rights quickly pushes it by accident. And there is no warning- the baseline is just imported and overwrites the previous content.

I call for disarmament of this thing- asap - before bad things happen

Oliver

4

Suggestions and Requests / EA protocol handler - again

« on: September 17, 2009, 08:11:42 pm »

Some time ago Canonic announced a protocol handler for EA.
As Canonic does not seem to be able to deliver this a feature request has been started.
What we would like to achieve: Open Enterprise Architect from an external application and show the corresponding element or diagram.
This can be done the same way DOORS does it (doors://doors_ID -> ea://element_uid).

This solution will provide the necessary traceability from DOORS and other external applications.
Additional information: Yes, we are already using the EA MDG link for DOORS, however importing requirements is not enough as we need the backlink DOORS->EA while we already have EA->DOORS.

I am also grateful for hints how to create such a capability via scripting or addin or by other means.

Oliver

5

Suggestions and Requests / Transformation language needs a loop statement

« on: September 10, 2009, 07:34:50 pm »

Having worked a lot with the transformation templates I found a high need for iterative language constructs to write loop statements.
Actually there is none.
Under linear conditions this is not obvious, it becomes important when using the EXEC_ADD_IN macro which might deliver multiple values eg. as a comma separated list to process. Classifiers eg. might have several tagged values which values are used to process attributes or parent elements.
Actually such a parsing can be done recursively with custom templates I did this yesterday but it is tedious.

So a loop statement would be a rather good idea.

Olive

6

Suggestions and Requests / Feature request fil: Generate report from EA forum

« on: September 07, 2009, 05:37:58 pm »

Just filed this feature request at Sparx:

====================
The ability to create reports from the EA forum would be a valid and helpful feature.
We use the forum for feature discussions between all parties involved and if we could create a report from this discussion to archive the result for QM issues that would be rather helpful.

Procedure as usual: Select a topic/discussion, press F8, select a tenmplate and the generator runs through the topics reporting the text and attached elements. Reporting elements is important here as they carry agreed information.
================

Oliver

7

Suggestions and Requests / EA Lite and automation interface

« on: April 02, 2009, 10:23:07 pm »

Having finally found the link to EA Lite which is promoted as a "read-only" version of EA I found out that it indeed is not completely read-only. In fact the automation interface seems to be completely functional and lets me modify elements at will.
Though I am not sure whether this is in the nature of the original idea of EA Lite I consider this a feature as it lets us deploy a deamon process which not only does HTML exports on a regular basis but also manipulates elements frequently (like creating baselines, etc.)
This daemon runs on a computer which is not operated by architects and on the very positive side does not require to spend a license for a not existing resource (due to the free character of EA Lite).

My interest is that this is a supported scenario and that therefore this situation will not change. I can imagine a scenario in which this functionality slipped through the fingers of Sparx and which will lead into disabling write access for the automation interface in the mid or long run.
A feedback from Sparx regarding this would be appreciated.

EA Lite is a great addon for us and highly useful in several aspects as it allows a better integrated communication with our stakeholders.

Oliver

8

Suggestions and Requests / Undocumented web template features

« on: December 16, 2008, 03:30:51 am »

While trying to elaborate how to write corporate web templates I stumbled across some undocumented features of the HTML reporting.
In fact the whole web templating mechanism is mostly undocumented, as I found out.

First thing to note: There is no way to copy an existing corporate web style template to a new one. If you create something new from scratch, it creates the EA default template.
So I did what smart guys would do: Export reference data => import reference data. Failure: Simply overrides the existing one.
Again smart guys procedure: Go to the XML file and look what is inside.
Result: <Column name="TRX" value="xxx"/> Voila, here is the name, change it and it will work -> Went through, checked.

Now what is the remaining of the XML file good for ?

Code: [Select]

<Column name="Description" value="HTMLSTYLE"/>
                  <Column name="NumericWeight" value="1.000000"/>
                  <Column name="TRX" value="xxx"/>
                  <Column name="TRX_ID" value="9"/>
                  <Column name="Style" value="2"/>

The most interesting part is hidden behind the "Style"  section. If you change the value to something, it will use a totally different style and structure. Changing it to a different value and importing it, then generating a report with that style will lead to tons of messages about unmatched field values- propably because some fields used in the template have no expression in EA today.
But the whole layout looks much different.

So two conclusions:

• The web template handling is not satisfying as it is not well documented and not manageable (see deriving of templates above)
• There seem to be more features and styles available which are simply not visible to the user and I suspect more. It also leads to the conclusion that a compelte rearrangement of the layout is currently not possible without changing that style attribute- EA seems to expect a certain layout with frames and tables- which seems to be hardcoded.

I'd like to see Sparx comment on that. It would be worth improving this area as I believe it is one of the few (and I mean few) disappointing areas of EA. More documentation here would make it more comfortable.

Just a few cts of mine.

Oliver

9

Suggestions and Requests / Basline management in automation interface

« on: September 24, 2008, 11:35:55 pm »

For everyones information, I just filed the following feature request:

In http://www.sparxsystems.com/cgi-bin/yabb/YaBB.cgi?num=1216872479/4#4 Roy proposes to create backup baselines for EA repository models.

Unfortunately creating and removing baselines is supported in the automation interface which prevents from writing a daemon for regularly creating that baseline.

Request: Introduce two methods createBaseline and deleteBaseline in the automation interface so we can access it from outside.

Regards,

Oliver

10

General Board / Got a bit quiet there in down under

« on: June 30, 2009, 09:08:25 pm »

After sending several bugs and support requests to Sparx and some forum posts I haven't heard anything since. Is this just me or is there winter holiday season in down under?
Anybody else experienced that sort of delay or quietness?

Oliver

11

General Board / WANOptimizer result thread

« on: March 20, 2009, 09:53:23 pm »

As promised in this thread http://www.sparxsystems.com/cgi-bin/yabb/YaBB.cgi?num=1237197542 I made some tests with the WAN Optimizer (furthermore referred to as WO).
But why not have a generic thread for results so, folks, feel free to post your experiences below.

WANOptimizer beta2 requires at least EA build 842 (7.5beta3). A test with EA 7.5 beta2 resulted in dozens of "database session not authenticated"  messages.

All tests have been applied on our corporate system model with 900 MB in size, 2000 packages and 13000 elements. Database is a MSSQL server.

All scenarios are run from the same Dell laptop by just using different connection strings.

1) Inside the corporate network with direct connection to the SQLServer, 100 MBit/s network lines.
Opening the model takes without WO takes approximately 12-15 seconds. The same constellation with WO active use s exactly 7 seconds.
Opening element properties or accessing the context menu brings no advantages.
A large and (way too) complex diagram is opened in 16 seconds without WO, with WO it takes 12 seconds. Opening it the second time used 4 (!) seconds (WO only).

It seems that WO also optimises database access in standard scenarios.

2) VPN connection via DSL, 2MBit downstream/256kBit  upstream.
VPN is routed through several corporate locations  and servers countrywide so some latency can be expected.
Opening the model take 160 seconds without WO. With WO, it opens in 45 seconds.
Accessing an element contect menu takes up to 15 seconds without WO, under WO control it is 5-6 seconds on average.
The properties dialog of an element appears after 5-7 seconds unoptimised, optimised it showed up almost right after double clicking an element.
The diagram tested above showed up after 30 seconds in standard scenario, with WO it took about 15 seconds.
10 seconds the second time it was called.


So far I can say WO substatially increases performance. It can not resolve latency issues nor can it do miracles on slow upstream links.
I have also not modified elements or done complex searches or report generation, import/export, etc.

But it definitely improves the workflow. The question is will it be enough for a constant and comfortable workflow via VPN.

More hopefully to come.

Oliver

12

General Board / Sharing search criteria in repository

« on: June 02, 2008, 06:19:55 pm »

Maybe it´s just me but how (or another EAUI) but how can a team share search criteria ?
Let´s say a team of 20 members working on the same model would like to see a common view on actors, components etc. defined by search criteria.
However if I define a search criteria on my laptop (eg. all actors of a certain stereotype) and attach it to a model view nobody else on the same model is able to get elements in this view (though the view is there) because the search criteria seems to be defined locally.

What am I missing ? It is not convenient to do a file export/import.
It would be acceptable to add the search to the profile or MDG Technology.

Oliver

13

General Board / Re: Are you an Enterprise Integration Architects ?

« on: April 12, 2008, 02:37:49 am »

Someone from the admins please remove that recruiting spam and remove the account of that individual.

If I would like to read about "fantastic opportunities" I know places to go to.
This spam has almost nothing to do with the EA product- that looks like a blind shot into an air which seems to smell known but it is not.

Oliver

14

Bugs and Issues / Datatype mapping inconsistent

« on: March 19, 2009, 06:39:13 pm »

Filed this bug some days ago:

=========================

Data type mapping with custom products not working

Details:  

EA offers the possibility of creating database products with custom
datatypes.
However the mapping dialog shows an inconsistent behavior and the
mapping is not applied when the database is changing from and to a
custom database product.
The reference data export DBMap does not include any custom product
mapping.
In the mapping dialog one can apply a mapping, which is not refreshed
when changing the datatype in the left list box, it always shows the one
from the previous selection. Plus the "Save" button gets grayed out when a size of a datatype is being entered.

Steps to Reproduce:  

- Create a custom database product A and add some types.
- Add some datatypes x,y,z to the product with ranges, size, etc.
- Open the datatypes mapping dialog, create a mapping of product A type
x to eg. Oracle VARCHAR. Press "Save".
- Stay in the mapping dialog and select datatype y. The dialog still
shows a mapping to Oracle VARCHAR. Change this to eg. DB2 REAL and click
in the "Size" field -> the "Save" button gets grayed out unless "None"
or "Precision" is chosen. Save the mapping
- select datatype x again -> the dialog shows a mapping to DB2 REAL for
x, though it should be Oracle VARCHAR.
The correct mapping is shown only if the target product is chosen again.
- Export reference data DBmap and open the file -> Only internal product
mappings are included.
- Create a table of product A with attributes of type x and y, then
change its database product to Oracle -> mapping is not applied, it
still shows a type x.

===============

Sparx has notified me that this bug has been reproduced and that improvements in that area will be included in a future release.

Oliver

15

Bugs and Issues / DDL transformation: Attributes do not transform

« on: February 13, 2010, 01:26:17 am »

Hi all.
Despite that it is a bit quiet around me due to work load, I filed a new bug:
=================
DDL transformation: Attributes do not transform correctly

Several issues came up when transforming class elements to table elements via DDL transformation templates.

- %attallowDuplicates% always returns an empty string even if in the attributes properties in the collection section the duplicates checkbox is ticked
- when applying %TRANSFORM_CURRENT()% without excluding the "containment" section a string "containment=" is created with no value. The "containment" keyword is unknown to the parser, even if the value is explicitely set to something like "1".
- It is unclear which keyword  in the DDL template will set the "Not Null" option of a column. There does not seem to be a allowDuplicates or nullable keyword available.

Reproduction:

Create a class with at least two attributes, of which one has set the Allow Duplicates option from the attributes properties (collection section in the Details tab)
Create a DDL transformation template which shall transform the class into a table.
The "Attribute" template contains a %TRANSFORM_CURRENT()% with no further parameter and a section in which %attallowDuplicates" is read and visualised.
Run the class through the DDL transformation using the template adapted.

Result: The parser stops after encountering the empty "containment=" entry.

Add a section with containment="1" to the template, rerun the transformation.
Result: The parser stops after encountering the containment="1" entry.

Add  "containment" in the exclusion list of the %TRANSFORM_CURRENT% macro
Rerun the transformation.

Result: Transformation runs through but %attallowDuplicates% returns an empty string for both attributes.
============

These seem to fall in the annoying not-thought-till-the-end section

Oliver