One of the most prominent features in HTML 5, and one of the key to allowing developers to provide offline functionality in websites, is the local storage API. This allows developers to store simple strings under values in a JavaScript API, which could be all kinds of application data.
Google have recently been talking about an open source project called LevelDB. This project was inspired by some of the underlying technology to their BigTable infrastructure, and is a variation of what is to be used in upcoming versions of Google Chrome to provide the aforementioned HTML 5 local storage API.
There are a number of interesting features of the database. First of all, the API for it is extremely simple and very nice to use. It also doesn't have many dependencies, which should make it nice and lightweight and portable.
You can only have one connection to the database, which is backed by a file on the filesystem, open at any one time - while the developers say that it may be possible to use multiple threads to access the same database, multiple processes is out of the question and even if you go down the multi thread route you will have to implement your own locking; this is very much designed for single process caching or other lightweight semi-persistant storage uses.
LevelDB has been designed to be extremely performant, and uses the Snappy project for ludicrously fast compression of values. It also has some atomic update functionality available, in a similar way to the Redis project - you queue a set of commands which are then guaranteed to be executed in sequence.
More on the Google Code blog.
Comments