BinaryDocumentDb 1.0.0

BinaryDocumentDb

Created to enable storing any information no matter how small, as a uniquely identifiable and retrievable byte array. This is really more like an on-disk version of a Dictionary<uint, byte[ ]>

When this project was started, there was thought given to how SSDs have a limited capacity to overwrite the same area of the drive again and again. Therefore, it was important to find a way of allowing for changes to be stored whilst making as few changes to the physical NAND Flash memory as possible. To this end, there is no index stored.

We conpensate for this by scanning the database file and creating an in-memory index. In reality a Dictionary<uint,uint> where the key is a unique value generated when you create a record in the store, and the Value part of the dictionary is a physical Offset into the underlying FileStream.

Entries in the physical file look like this:

Empty Entry

• (byte) 0 : meaning an empty space entry.
• (uint) 123 : meaning 123 available bytes including the header byte, and this 'length' of 4 bytes.

Blob Entry

• (byte) 1 : meaning a blob entry.
• (uint) 12 : meaning a length of 12 bytes including header byte, and this 'length' and 4 bytes.
• (uint) 34 : the KEY that uniquely identified the following data
• (byte[ ]) : the raw data that was originally asked to be stored.

I/O Operations

• Deleting or Updating an entry with more or less data that originally was stored may cause the existing entry to me marked as empty space (changing header byte only).
• If smaller (by 5 or more bytes) will re-use the same entry, but create an empty entry of the remaining space.
• If larger, even by one byte, will try to find an existing empty space to re-use, otherwise the existing entry will be marked as free space, and the new data is added to the end of the file.

All operations result in returning an ExecResponse object. These objects encapsulate success or failure, an error code, exception messages, and a Result property (type depends on operation).

You should ensure the success of the operation with each interaction.

In any operation that may create empty spaces, the free spaces are defragmented, so where more than one empty space exist on disk next to each other are merged into a single entry.

Events!

BinaryDocumentDb is meant as the basis of other simple stores, or repositories. We find it useful know when IO operations that could affect an in-memory cache occurs. To that end, the QueuedEvent type from the InFlux library has been used, because if its weakly refernced events that are all queued for processing, givening a more predictable sequence to event subscription calls.

Usage

var instance = BinaryDocumentDbFactory.Make(new BdDbConfig() 
    { FilePathAndName = "databaseFile.db" }); 

• Create a record in the database

   var createResponse = instance.Create(new byte[]{1,2,3});
   var KEY = createResponse.Result; // such as 123

• Read a record from the database

   var readResult = instance.Read(123);
   byte[] data = readResult.Result.ToArray();

• Update an existing record in the database

   var response = instance.Update(123, new byte[] { 4, 5, 6 });

• Delete an existing record in the database

   var response = instance.Delete(123);