/document/N/M/ format differs from previous API

[dvogel]

Here is an example of the same document loaded into the old and new versions of superfastmatch, associated with itself using these commands:

echo "Lorem ipsum dolor sit amet, consectetur adipiscing elit." > /tmp/loremipsum.txt
curl --data-urlencode "title=Lorem ipsum" --data-urlencode "text@/tmp/loremipsum.txt" http://localhost:8080/document/1/1/
curl -XPOST http://localhost:8080/association/1/1/
curl http://localhost:8080/document/1/1/ | python -m json.tool

Old format:

{
  "text": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.\n",
  "title": "Lorem ipsum",
  "doctype": 1,
  "docid": 1,
  "characters": 57,
  "documents": {
    "rows": [],
    "metaData": {
      "fields": []
    }
  },
  "success": true
}

New format:

{
  "associations": {
    "Documents": [
      {
        "fragment_count": 1,
        "fragments": [
          [
            0,
            0,
            55,
            239133750739
          ]
        ],
        "valid": true,
        "characters": 57,
        "title": "Lorem ipsum",
        "id": {
          "docid": 1,
          "doctype": 1
        }
      }
    ],
    "Meta": {}
  },
  "valid": true,
  "characters": 57,
  "text": "Lorem ipsum dolor sit amet, consectetur adipiscing elit.\n",
  "title": "Lorem ipsum",
  "id": {
    "docid": 1,
    "doctype": 1
  }
}

A re-implementation of the old API would be best for compatibility.

• Move "doctype" and "docid" out of the "id" object, drop the "id" key
• Either drop or document the new "valid" key. Is this the same as the old "success" key?
• Capitalization changes, e.g. documents vs Documents
• Can we drop the new "Associations" key that wraps the "Documents" key?
• What is the empty "Meta" object for? Is this supposed to be the old "meta_data" object under "documents"?

It seems the old version didn't associate a document with itself while the new old does. Since the full overlap of a document with itself can be assumed, the old behavior makes more sense to me.