Skip to content

Latest commit

 

History

History
100 lines (72 loc) · 3.39 KB

Specification.md

File metadata and controls

100 lines (72 loc) · 3.39 KB
Raw

Storcord

Storcord is a document-oriented NoSQL manager that uses Discord as its "main" storage.

Setting up a database

Creating databases are like creating Discord guilds. Every guild represents a database and every Discord channel acts as a "collection"(from Mongo).

Adding documents

When adding documents, an entry is generated for that document. A document has to be in JSON(no BSON bullshit from mongo).

A document entry is the document object with metadata about the document. The ObjectID concept from Mongo is ported to Storcord through a tuple of the channel ID of the entry and the message ID of the entry.

ObjectID isn't provided in the document entry itself, the client needs to index it manually thgouth its own means.

The basic metadata a document entry can provide is:

  • _type[int]: The type of this entry
    • 0 for an actual full document(not sharded).
    • 1 for a document shard.

Document Examples

{
    "_type": 0,
    "raw": "{\"a\": 2}"
}

Document Sharding

For documents that their total JSON(plus metadata) gets bigger than 2000 characters, the document becomes sharded across document shards, each shard has chunks of the actual JSON encoded document but they have metadata on how to find the other shards so you can rebuild the entire document witn only one of the shards.

Document shards have extra fields:

  • _shard[int]: The Shard ID of this entry.
  • _range[tuple[int, int]]: The character location range that this shard provides.

Document shard example

{
    "_type": 1,
    "_shard": 0,
    "_range": [0, 100],
    "raw": "\"a\": 2, 3: 4,",
}

Querying documents

Queries have two types:

  • Simple query: where you just want 1 document returned, the query finishes itself.
  • Multiple query: where you want all the documents matching it, way more expensive.

Query cases:

  • You want to Simple query a document and you have its ObjectID.

    • Get the channel and message IDs from the ObjectID and get the document.
    • If it is a shard, get the other shards.
    • Multiple query case: giving just a channel ID returns all documents that are in it.

  • Second case: You want to Simple query a document by its raw content.

    • O(n*C) operation, with C being the number of channels you are separating your documents with
    • You have to go through all the channels and find the 1st document containing your query.
    • For Multiple query, you'll have to go through all the documents.

  • Third case: You want to Simple query a document by its JSON data.

    • Follow the Second Case strategy, but load the document's JSON and search through it.
    • Same follows for Mutiple query.

Updating documents

If the document to be updated is a single unsharded document, updating is easy: edit the message the document is contained in with a new version of it.

  • If the edited version of the document goes over 2000 characters we will have to delete the old document and write a new sharded document.

If the document is a sharded document, you can:

  • Easy: Delete all document shards and recreate them.
  • Hard: Edit existing shards and create new shards as required by the document's size.

Deleting documents

Deleting documents is straightforward.

  • If the document is not sharded, deleting the message the document is contained is enough.
  • If the document is sharded, all document shards have to be deleted from their respective channels.