Diffs

Diff communication is the lifeblood of the system. Here's how it works:

• Data is for the most part JSON (there are exceptions)
• JSON objects, including diffs, are zipped together recursively. Colliding names are clobbered from source to destination.
• null is magic; it removes attributes.

For example:

MLB1:
{
    "name": "Joe DiMaggio",
    "average": .325,
    "metadata": {
             "skill": "baseball",
             "girlfriend": "Marilyn Monroe"
         }
}

MLB2:
{
    "name": "A-Rod",
    "metadata": {
             "skill": "football",
             "girlfriend": "Madonna"
        }
}

DIFF:
{
    "average": null,
    "metadata": {
             "skill": null,
             "girlfriend": null
        }
}

If these were diffed against each other conceptually:

MLB = applyDiff(MLB1,MLB2):         //base, then diff

{
    "name": "A-Rod",
    "average": .325,                //which would be in his dreams
    "metadata": {
             "skill": "football",
             "girlfriend": "Madonna"
        }
}

but later:

MLB = applyDiff(MLB,DIFF):          //base, then diff

{
    "name": "A-Rod"
    "metadata": {
        }
}

So, no skill, no average, and no girlfriend.

FIXME: provide a better example with more edge cases.

Arrays are treated as entries, and are not intersected. They clobber each other.

A future incarnation will adopt BSON instead of JSON. Perhaps Qt will make this adoption, or we will have to incorporate it some other way.