Is there a simple way to use the filesystem using the JCR API?

1
votes

I have an OSGi-based, server side application that uses the filesystem to store scripts and configuration data.

In time, I'd like to move that application to 'the cloud', and that's not going to work well with its current dependency to file system access.

What I'd like to do is insert a JCR layer into this application, so it will still work in the current situation (regular files on the local filesystem), but will pave the way to a cloud situation.

I did find a file connector in modeshape, but I ran into a pretty severe incompatibility with OSGi, which hasn't been fixed. Besides, ModeShape pulls in LOTS of dependencies (about 6 MB, I think), which is a problem for me.

So I don't see any options besides starting to hack my own JCR implementation, which I am reluctant to do.

Any ideas?

2
jackrabbitjcrmodeshape

2 Answers

2
votes

Although you wouldn't be using JCR directly, using the Apache Sling ResourceProvider mechanism should allow you to move easily from filesystem to something else later, and it's OSGi-friendly as Sling is 100% based on OSGi.

You could start now by using Sling's Filesystem resource provider ( http://sling.apache.org/site/accessing-filesystem-resources-extensionsfsresource.html ) and later move to your own custom ResourceProvider, as needed.

The source code of the filesystem provider is at https://svn.apache.org/repos/asf/sling/trunk/bundles/extensions/fsresource - it's quite simple code that can be used as an example for creating your own ResourceProvider.

For your custom system the question would be how many Sling bundles you need to get that working - I don't know off the top of my head but would suggest using the Sling Launchpad to find out, it launches a vanilla Sling system with lots of bundles that you won't need, but you could try reducing it to the minimum that still allows the ResourceProvider mechanism to work.

1
votes

You can also use Apache Commons VFS2, there is for example a JCR connector, or you can use webdav or a JDBC table. I use this in a commercial project on top of an atomic (git like) tree on top of a shared JDBC table.