Goal
Find a perfect, flexible schema for storing many different types of objects with a wide variety of links between them in a relational database.
Problem
EAV is a workaround to the normal confinements of a RDBMS.
If you were to normalize an EAV schema, it would be ugly.
Idea
If EAV was normalized, it would be ugly.
Does the fact that we traditionally maintain these schema by hand limit their complexity and power?
But if it was maintained and queried programmatically, what would it matter?
Graphs
If you have n different entities in n different tables, why not let your code generate n(n+1)/2 link tables and the queries between them? Would this not result in a true graph in a normalized schema?
In a highly interlinked database, there will always be exponentially more edges than vertices. Why not focus on creating proper, normalized verticles (n entity tables) and let our code maintain the edges (n^x link tables)?
Conclusion
Can a system normalize EAV and maintain the resulting complex schema?
Can complex graphs be stored in (and remain true to) relational databases?
I’m sure this has been done before, but I’ve never seen it. What am I missing?
Example problem
Storing printed works and their bibliographic data
- Many properties which might be not just strings but whole objects.
- In the library world, there is no simple (and relational) schema which can store data “losslessly” without extremely complex schemas.
- Many different types of associations and associated objects
- And their relevant properties (which can vary wildly).
- And their many relationships, of different types, amongst themselves.
Questions
“What problem are you trying to solve?“
-Piet
I’m looking for a normalized solution to EAV, graphs, and polymorphic relationships in a relational database system.
“I would hate to be the guy who has to understand or maintain it after it’s gone into production.“
-Andrew
This “traditional maintenance” is the exact thing I’m saying we should be automating. Isn’t it largely grunt work?
Since you are editing the question, it must be active.
Yes, there are much better ways of designing this, for the purpose and use you describe.
The first issue is EAV, which is usually very badly implemented. More precisely, the EAV crowd, and therefore the literature is not of high quality, and standards are not maintained, therefore the basic integrity and quality of a Relational Database is lost. Which leads to the many well-documented problems.
You should consider the proper academically derived alternative. This retaiins full Relational integrity and capability. It is called Sixth Normal Form. EAV is in fact a subset of 6NF, without the full understanding; the more commonly known rendition of 6NF.
6NF implemented correctly is particularly fast, in that it stores columns, not rows. Therefore you can map your data (graph series, data points) in such a way, as to gain a flat high speed regardless of the vectors that you use to access the graphs. (You can eliminate duplication to a higher order than 5NF, but that is advanced use.)
“Highly-interlinked” is not a problem at all. That is the nature of a Relational Database. The caveat here is, it must be truly Normalised, not a inlerlinked bunch of flat files.
The automation or code generation is not a problem. Of course, you need to extend the SQL catalogue, and ensure it is table-driven, if you want quality and maintainability.
My answers to these questions provide a full treatment of the subject. The last one is particularly long due the the context and arguments raised.
EAV-6NF Answer One
EAV-6NF Answer Two
EAV-6NF Answer Three
And this one is worthwhile as well:
Schema-Related Problem