Normalizing a document with an improper id produces empty string

[mattcollier]

This is of concern for systems that use hashes of documents for content-based addressing and also digital signatures. Hashing the result of jsonld.normalize is common practice for these purposes.
Hashing algorithms readily accept '' (an empty string) as a valid input. Since a valid hash is produced, this sort of issue may go unnoticed.

Would it be possible to throw an error instead of returning an empty string? I suggest that an empty string should not be valid output for normalize.

In this case, the empty output is due to this conditional: https://github.com/digitalbazaar/jsonld.js/blob/master/lib/toRdf.js#L108

Sample code:

const jsonld = require('jsonld');
const s1 = {
  "@context": "https://schema.org",
  "id": "urn:07179fd9-4856-4e55-baf8-b2589a037566",
  "name": 'White',
};
const s2 = {
  "@context": "https://schema.org",
  // `id` is not a valid URL
  "id": "07179fd9-4856-4e55-baf8-b2589a037566",
  "name": 'Brown',
};

async function go() {
  const e1 = await jsonld.expand(s1);
  console.log('EXPANDED_1', JSON.stringify(e1, null, 2));
  const e2 = await jsonld.expand(s2);
  console.log('EXPANDED_2', JSON.stringify(e2, null, 2));
  const r1 = await jsonld.normalize(s1);
  console.log('NORMALIZED_1', r1);
  const r2 = await jsonld.normalize(s2);
  console.log('NORMALIZED_2', r2);
  console.log('TEST_EMPTY_STRING', r2 === '');
}

go();

Output:

EXPANDED_1 [
  {
    "@id": "urn:07179fd9-4856-4e55-baf8-b2589a037566",
    "http://schema.org/name": [
      {
        "@value": "White"
      }
    ]
  }
]
EXPANDED_2 [
  {
    "@id": "/07179fd9-4856-4e55-baf8-b2589a037566",
    "http://schema.org/name": [
      {
        "@value": "Brown"
      }
    ]
  }
]
NORMALIZED_1 <urn:07179fd9-4856-4e55-baf8-b2589a037566> <http://schema.org/name> "White" .

NORMALIZED_2 
TEST_EMPTY_STRING true

Related: digitalbazaar/bedrock-ledger-node#12

[dlongley]

Closing as this falls under #199.