Messages

1

Suggestions and Requests / Viewing more things 'side by side'

« on: September 21, 2009, 09:20:37 pm »

It would be nice to be able to view two diagrams side by side.

And also, to view a diagram alongside the discussion forum.

I guess that means I'm asking for the ability to 'tear off' the tabs at the bottom of the main window & rearrange them. Or have a Window / Tile menu option.

2

Suggestions and Requests / Re: Discussion Forums - Linkage to Model Elements

« on: September 21, 2009, 09:17:15 pm »

We'd also like this.

And why can I only link an element to a post, not to a topic (or even category?).

We're at a review stage, and it seemed sensible to me to create a category for each package, with topic for each class, and posts/threads for each comment against the class. It's a bit simplistic (some comments cross classes & even packages), but combined with an ability to search / display the discussion alongside the model & we'd have a powerful review tracking tool.

3

General Board / EA users, resources; db users

« on: May 12, 2011, 12:06:54 am »

We've been using a shared .EAP for some time, but the team is growing & it seems time to move to a database repository. It seems to make sense to enable EA's security at the same time.

I've read around a lot, including forum searches for 'repository', 'database', 'security'. I have a few questions:

1. is there any relationship between the Enterprise Architect security users & groups, and the database users (perhaps via Windows Active Directory)?

2. If this isn't automatic, is it a good plan to encourage/require users to connect to the database with their own user names? I guess that means, is there any future in trying to use database level security alongside/beneath EA's?

3. Do these EA security users have any relationship to the EA Project Management 'resources', e.g. is there a relationship between assigning a package to a resource & locking it for a user?

4

General Board / Re: PIM to DDL Transformation

« on: July 03, 2009, 11:35:43 pm »

I'm a recent member, looking for an answer to this question - but it looks as if there was no answer four years ago.

The behaviour seems to be the same: when I transform my PIM to DDL, the generalisation associations are treated as generic associations, that is, the (abstract) superclass gets its own table. This is OK, but I would prefer for the inherited attributes to be copied into each subclass's table.

Is there any way I can influence the transformation's behaviour? I've had a brief look at the template, but can't see anything obvious.

5

Bugs and Issues / Re: Model transformation to DDL: multiplicity & NU

« on: May 12, 2011, 01:10:07 am »

I can just about read the XSD one, but the DDL one relies on the TRANSFORM_CURRENT built in function.

Would I need to replicate whatever that does, or can I just tack something to handle attLowerBound on the end, setting "NOT NULL" at the end of the string. Guess I'll have to make time to try it!

6

Bugs and Issues / Model transformation to DDL: multiplicity & NULL

« on: January 26, 2010, 10:46:04 pm »

Hi,
In a UML class, I can set the multiplicity lowe bound to 0, which means to me that the attribute is optional, or 1 (or more), which would mean it is mandatory.

When I transform the package to DDL, the attributes come through stereotyped as columns, but all of them have Not null set to No. The only columns which get not null set to Yes are the primary keys which the transform creates.

Conversely, where a column is introduced to carry the foreign key for an association, it is often given Not Null = Yes, even if the association has a multiplicity lower bound of 0.

It's the "often" that has got me worried - in a given model, some of these optional associations get foreign key columns that can be null, others can't be.

Any thoughts as to why this is?

7

Bugs and Issues / Re: Datatype mapping

« on: May 12, 2011, 01:04:29 am »

Re-raised as http://www.sparxsystems.com/cgi-bin/yabb/YaBB.cgi?num=1305025781

I was referring to the Datatype Map which is used by EA's "Reset DBMS"; the response last year seems to relate to scripting, which could be an option I guess.

8

Bugs and Issues / Datatype mapping

« on: January 26, 2010, 10:25:03 pm »

Searching the forum doesn't return anything for 'datatype map', so I'll start a new topic.

I have created a database product "TC211" containing some of the data types used by ISO TC211. I have created a datatype map to map these to Oracle. The tool for doing that is rather clumsy, but I believe I've got there.

But it doesn't work as I would expect, so CharacterString turns to VARCHAR(36), regardless of the length I specify in the map (and it would be even better to be able to specify it in the logical model); Boolean, which I've specified has a max length of 1 comes out as NUMBER(22).

Any suggestions?

9

Bugs and Issues / datatype mapping

« on: May 10, 2011, 09:09:41 pm »

Hi,
I've created a new "Product" within the Database Datatypes, describing the data types in the TC211 family of standards. This is the default for all our logical class modelling.

I've defined a "Datatype Map" from that to Oracle, which is our physical database technology (wow - that's a painful interface!). For example, GM_Polygon maps to SDO_GEOMETRY. All that is safely stored in our .EAP file.

However, when I choose Code Engineering -> Reset DBMS Options (to open the "Manage DBMS Options" dialogue) and select Convert DBMS Type to convert from TC211 to Oracle, the results are patchy.

At a high level, quite a few things get converted properly - at least, most of the time. For example, CharacterString converts to VARCHAR2. Sadly, all these 'normal' types ignore anything about length, even when they get converted. More sadly, all those which should map to SDO_GEOMETRY come out as BLOB. And some 'normal' things seem to get missed (the odd Integer for example).

In practice, we have to revisit each attribute / column & set it manually. Is there any way to make this work reliably?

And is there anyway to export the mapping & load it into a different .EAP file?
Peter

10

Bugs and Issues / Data type mapping with custom products not working

« on: March 19, 2011, 03:04:26 am »

I have a problem which is very similar to one in the forum from 2009, with the same title.

I've created a 'custom product' (TC211) with the range of data types I want to use in my logical model.

I've struggled through the unhelpful Datatype Map dialogue and created mappings from this product to Oracle. I can get the mappings I want to reappear there (e.g. CharacterString -> VARCHAR2(256), Boolean -> CHAR(1), GM_Point -> SDO_GEOMETRY, GM_Line -> SDO_GEOMETRY,...)

But when I run Model Transformation, it is pretty hit & miss. Some of the data types are changed, but many are not. In the end, it is easier to spend a few days going through class by class & changing them by hand! We're considering exporting to XMI & processing it externally (XSLT) & re-importing. That's not good for a CASE tool, especially as I try to 'sell it' to my developers.

Is there something I'm missing?

I am using EA v8.0, Corporate Edition, build 856 - but I'm pretty sure the behaviour has been the same for several years. The repository is an EAP file.

11

Bugs and Issues / rerunning model transformations

« on: January 26, 2010, 10:34:40 pm »

Hi,
I like the way that I can re-run a model transformation, and EA attempts to only change the bits in the target (DDL, in my case) that I've changed in the source (UML class diagram).

But, sometimes it doesn't work. Particularly, I needed to introduce role names, because the constraint names generated automatically can be too long for Oracle. Redoing the transform creates the new foreign keys, but doesn't delete the old ones. I can't delete the old ones in the DDL package, because they are foreign keys. So I deleted the relevant tables in the DDL package and re-run the model transfomr again.

Now the new table doesn't get joined to the parent, and gets no foriegn key at all. The only solution seems to be to delete the whole target package and re-run the transform from scratch. Shame.