I’m just learning about ORM’s, and am wondering… if you have an object (a JSON let’s say… and assume it’s dense, contains nested objects, has a total number of properties that could vary from object to object, etc), what would be the argument against simply turning it into a serialized array, and storing it in a table?
The object I’m describing likely won’t normalize anyway; it would have an indeterminate amount of properties meaning an indeterminate columns… map this kind of thing to a table would require an insane amount of rows and keys all over the place.
If you’re ultimately planning on querying the db for an object to then process via code, what is the downside to simply serializing it or using some kind of OODB? What would I gain by using an ORM?
My experience is that because the Storage layer for the ORM expects the storage data to be dynamic and therefore holds no preconceptions about its format, it can be better equipped to deal with the exceptions to the norm. (Not actual error exceptions but cases where your object doesn’t match your Database Schema)
When you’re dealing with Dynamic objects, the rigidity enforced by classic storage generally forces you to either handle the exceptions to norm or create a database schema so loose, that using it defeats the optimisations generally granted by the Database engine; think computed stats and various indices.
However, ultimately I think you’ve hit the nail on the head in your second paragraph: If it won’t normalise then you will have trouble representing it into a schema that’s good enough for your database to work with efficiently.
Sure you could serialise the entire object to an array and store that, but you lose the potency of good indexing, full text search and being able to cross-reference objects without having to do multiple reads.
An example for the above, say your DB is modelling ecommerce orders and you want to find subsequent purchase orders to the initial one. The database would need to know how to read each serialised item to find it’s parentId property and then rescan the table for matches.
Long story short, ORMs are an answer to a problem that’s been extant since Object Orientated programming was dreamed about – Don’t worry about it and use them, unless you’re sure that your data structures/schema are rigid and sensible to SQL.