Consider changing subgraph serialization format

[michael]

The current serialization format (JSON) for subgraphs (aka documents) is an array of nodes, where the last node in the array is the entry point (= root node).

However, as I'm integrating Svedit into an app, I realized that possibly it would be advantageous to have a homogenous format for documents, no matter if serialized or in memory.

// the nav doc - stored as a self-contained document in SQLITE
const nav_doc = {
  document_id: 'nav_1',
  nodes: {
    "nav_item_1": {
      id: 'nav_item_1',
      type: 'nav_item',
      url: '/',
      label: 'Home'
    },
    "nav_1": {
      id: 'nav_1',
      type: 'nav',
      nav_items: ['nav_item_1']
    },
  }
}

// the page doc - stored as a self-contained document in SQLITE
// however, nav_1 is not included, but only referenced
const home_page_doc = {
  document_id: 'page_1',
  nodes: {
    "text_1": {
      id: 'text_1',
      type: 'text',
      layout: 1,
      content: { text: 'This is the home page.', annotations: [] }
    },
    "page_1": {
      id: 'page_1',
      type: 'page',
      body: ['text_1'],
      nav: 'nav_1'
    }
  }
}

Now when we fetch the home_page_doc from the database, the home_page_doc and the subdocs (nav_doc) will be joined like so:

{
  document_id: 'page_1',
  nodes: {...nav_doc.nodes, ...home_page_doc.nodes}
}

This is how Svedit (the client) sees the data (as just one self-contained document, always)