Sparx Systems Forum

Enterprise Architect => General Board => Topic started by: mse on September 24, 2024, 11:17:01 pm

Title: How do you put models under configuration control in environments like Git?
Post by: mse on September 24, 2024, 11:17:01 pm

After working for some time with a cloud connection and being frustrated with the very slow performance, I am exploring my options of simply saving the packages to file and putting them in Git configuration control. Is it as simple as simply writing the packages to a file using the export option in EA? Or are there other methods? I see there is a dialog for using other versioning systems, but those are for other environments. I need one specific to Git.

Title: Re: How do you put models under configuration control in environments like Git?
Post by: Geert Bellekens on September 25, 2024, 12:01:51 am

You can't directly use Git because it doesn't support exclusive locks. Only SNV and TFS (Azure Devops), CSV and the generic one.

But you can still control your packages, and use batch xmi export for all your controlled packages.
It's not the same as real version control integration, but it can work (if you are the only user)

Geert

Title: Re: How do you put models under configuration control in environments like Git?
Post by: Takeshi K on September 25, 2024, 09:30:13 am

You should check out the LemonTree, it works perfectly with Git.

Title: Re: How do you put models under configuration control in environments like Git?
Post by: Elpis on October 04, 2024, 09:48:25 pm

Quote from: Geert Bellekens on September 25, 2024, 12:01:51 am

You can't directly use Git because it doesn't support exclusive locks.

Not exactly - see https://github.com/git-lfs/git-lfs/wiki/File-Locking .

However, EA doesn't support the Git. You may try to search for SVN-to-Git adapters, if you really has to stick with Git for that purpose. But IMHO it is not worth of fighting with (expected) problems.

Title: Re: How do you put models under configuration control in environments like Git?
Post by: qwerty on October 05, 2024, 07:28:54 am

Basically a model and code are as similar as fire and water. What works for code does not for models. Changes in code are for small distinct areas. But in a model a single change might affect the whole hing. A commit message "added connector from A to B" is pretty pointless since that change will usually be the end of a along discussion with lots of reasoning.

You might consider snapshots. Though rolling back in most cases is just a PITA. A model is a complex thing at any point in time. It's that one you can deal with and discuss.  A pletora of branches is just not manageable.

Of course YMMV

q.