At some point, I assume that anyone developing for couchdb (or other nosql option) must still identify what must be stored in each document. And it would seem that, although very rarely, every once in a while we would need to have some relationships between these documents.
So, are people still using ER-type software (like Workbench for instance) to create a visual representation of the data stored in the different documents? Is anyone working on developing nosql specific methods to identify the data that must be capture?
Right now we are using a bullet list in wiki pages, any suggestion for something better?
I don't anticipate tools like this to show up for a while, especially considering that CouchDB just released 1.0 a few months ago and there is still a lot that can change in the project. In addition, tools like this need to hold off until the real best practices come to light, but that can only come after years of real-world experience.
Although there are other reasons for this, relational databases simply didn't have graphical tools for years, if not decades. I can't imagine new database systems like these would progress much faster.
I would use a class diagram:
- replace every association with a composition or aggregation,
- see which classes must be read (as a whole) and modified,
- add simple attributes (strings or ints).
With this diagram, you have design hints:
- for "documents" granularity,
- URLs patterns,
- JSON "schemas".
来源：https://stackoverflow.com/questions/5350202/what-data-modeling-tools-are-nosql-couchdb-developers-using