While reading 'From Anchor Model to Relational Database' it occurred to me that converting a relational database into an anchor modeled database would not be a direct inverse. Poorly-un-non-denormalized data, anomalies, etc may present challenges. I haven't put much thought into it, but intuitively, it seems an additional set of transforms, refactorings if you will, would be required as well as some human decision making where semantic interpretation was required.
Challenges aside I do see value in reverse engineering existing relational models. Is there any tooling or work being done in the area of transforming from relational to anchor model?
Yes, there is a tool that reverse engineer existing relational models to anchor models. It's very much in its infancy though, and not developed by us, but I have seen it in action. Let me check what the current status is of that project and get back to you.
Since you are asking, are you interested in helping out with the development of such a tool?
Thank you. I need to reevaluate when and what I can contribute before committing however, yes, I am interested. A colleague and I briefly discussed creating such a tool a few days ago. If there's an effort underway already, and it fits what we're doing, then that works too.
One project, also in infancy I've begun, is essentially an IoT scale geotemporal infrastructure, for lack of a better term. I recently discovered anchor modeling, and it seems to fit, and also seems to solve many problems in that domain. I need to take the time to consider the architecture with anchor modelling factored in, because so far, it seems to be a game changer in terms of what will be implemented and development schedule, scope and possibly direction. It hasn't quite taken form yet, but seems it will soon.
I think the tool went defunct. Follow the link below for a script that you can use that will give you a model to start with. The generated model will most likely still need some manual labor in order to become ready for production environments.