Simple, scalable, optionally replicated web service which can shorten and expand a URL, in a tinyurl / bitly fashion.
This project is primarily a thought exercise and a way of trying out a few different ideas. It was built using TDD/BDD so should have good test coverage. Particular attention has been paid to efficient use of RAM, and fast shortening/expansion of tokens.
The URL shortener is a JSON web service, and does not have a pretty GUI for ordinary web users. In a future iteration I may add a GUI.
There are three 'databases' available:
- An optimised RAM based map, which can persist to disk. Limited by available RAM.
- A MapDB backed disk store which can support significantly larger data sets. Limited by disk space.
- A MapDB+JGroups distributed store which can replicate across multiple servers. Limited by disk space.
There is a config.properties file which configures the system.
BASE_URL=http://b.io/
DISK_BACKUP_FILEPATH=/var/lib/urlshortener/backup.mapdb
DATABASE_FACTORY_CLASSPATH=com.perry.urlshortener.persistence.BigOrderedReplicatedMapDBSetFactory
REPLICATION_CLUSTER_NAME=urlshortenermap
REPLICATION_TIMEOUT_MS=5000
If you have multiple instances running on the same file system, you will need to give each instance a different backup file path. This can be done by starting the JVM with -DDISK_BACKUP_FILEPATH=/var/lib/urlshortener/backup_2.mapdb (for example).
Assuming the system is running on http://b.io/,
To shorten a URL: http://b.io/?shorten=http://some.url/etc/
This will return a shortened URL e.g. http://b.io/jhakdj Hitting that URL will take you to the originally shortened URL.
To view the expanded URL without redirecting to it, simply add a + http://b.io/xhgjgh+
A JSONP service also exists to allow manipulatin of URLs with JavaScript. Simply add a ?callback= parameter to the request, e.g.
http://b.io/xhghg?callback=myfunc
This will return e.g.
myfunc({"url": "http://fred.1"});
Equally shortening a URL can return JSONP responses;
http://b.io/?shorten=http://bob&callback=myfunc
myfunc({"url": "http://b.io/ashdk"});
- The service could be improved upon with support for whitelists.
- Cloudbase could be used as a database - this would likely be a more robust storage solution.