Support integers as well as strings

[nolanlawson]

This library is amazing, but it would be neat if I could also do:

docuri.stringify({
  type: 'mytype',
  id: 'myid',
  subtype: 'mysubtype',
  index: 'myindex',
  version: 1
});

If you require the pouchdb-collate project, you could accomplish this:

var collate = require('pouchdb-collate');
collate.toIndexableString(2) < collate.toIndexableString(10); // true

In fact you could even stringify booleans or arrays or whatever you want. Ordering will be different for strings in PouchDB (ASCII vs ICU), but for numbers and booleans it should be the same across both. Then you could support semver:

docuri.stringify({
  type: 'mytype',
  id: 'myid',
  subtype: 'mysubtype',
  index: 'myindex',
  version: [1, 10, 2]
});

[jo]

Hi @nolanlawson, thanks for sharing this idea! I have nonetheless concerns about adding a dependency to docuri, because it would make docuri less easy to use in CouchDB functions like views.

I also dislike that stringifying with PouchDB collate makes ids ugly and unreadable:

> collate.toIndexableString(1)
'323241\u0000'
> collate.toIndexableString([1,2,3])
'5323241\u0000323242\u0000323243\u0000\u0000'

Speaking ids were one of the goals creating docuri.
Anyway I see your point that preserving sort order is an issue. Another way to keep sorting would be to simply add zero padding, eg 001. As this depends highly on the specific requirements of the user I would leave that up to him.
So what about using PouchDB collate outside from docuri? Maybe we can support this via sort of a simple transform api?

[nolanlawson]

Yep, I understand your concerns. The indexable strings are definitely not human readable.

I was thinking of making a plugin for PouchDB where basically you could just use any object as a doc ID, and it would transform them into strings using pouchCollate. So I can move this idea over to a new project.