proposal: separate Object type into JSON and Buffer types

[kapouer]

There seems to be a confusion in the definition of types: Object represents both JSON or Buffer. They are really different: JSON can be stringified, or in some cases is even a supported database type (postgres 9.3 and of course mongo) while Buffer calls for a binary representation. I suggest orm provides two distinct types, allowing for better usage of existing types in underlying databases. It also improves readability of data in the database itself.

[dresende]

There is a binary type, you can do this:

var Person = db.define("person", {
    photo: Buffer
});

[kapouer]

Okay, Buffer is for binary. But then i don't understand why object is handled the same as binary.

[kapouer]

Also, "Object" cannot always be serialized. "JSON" might be a better name.

[dresende]

It was initially called "object" because you could use Object to define it (instead of "type": "object") and because actually JSON is a simplified javascript object. :)

They are usually handled the same way (when storing them) because databases don't support it, but as you said mongo does. As for postgresql, if you know a way of detecting if the server has native support we could change it. An alternative is using a custom type (see test/integration/property-custom.js for an example).

[kapouer]

At least it should be impossible to put a Buffer instance inside a column accepting objects; so it will eventually be possible to make the distinction between json and binary blobs.

Note that {mycol: JSON} would have worked instead of {mycol: Object}. That's less important ;)