How to implement a cache "loader"?

[jmfaerman]

I would like to use infinispan to cache data aggregated from several systems, for a example in a medical system, cache the patient history (system a), images (system b) and billing (system c), etc...

I would like that, on a cache miss, infinispan loads this data using a "loader" class i would write, to find and build the aggregated document (value) for a given key.

Is there such a feature? How should i implement such a thing?

[galder.zamarreno]

Just look at the implementations of any of the cache loaders in https://docs.jboss.org/author/x/DwY5

Codebase: https://github.com/infinispan/infinispan

Examples: jdbc, file, bdbje...etc.

[jmfaerman]

Will try that, thanks!

[jmfaerman]

I have found the implementation of Cache Loaders to be very dependant on Infinispan APIs (InternalCacheEntry, CacheLoaderConfig, etc).

One should be able to code just the "loading" (retrieving keys and values) without messing with ISPN internals.

Is there any JIRA suggesting something similar so i can vote?

[jmfaerman]

Just to make the point clear, i was looking how other grids do this. Hazelcasts MapLoader looks nice:

http://www.hazelcast.com/docs/2.1/javadoc/com/hazelcast/core/MapLoader.html

Hope our proposal for the new SPI moves this way

[galder.zamarreno]

I see that you've commented on https://community.jboss.org/docs/DOC-17867 - that's the place to suggest changes

[jmfaerman]

I am trying to implement a CacheLoader Adapter do bridge a simple interface, like MapLoader, to our current CacheStore or CacheLoader.

As current implementations are very concerned with cache internals (bucket management, transactions, etc), this is quite a challenging task.

If this would be helpful, i would be glad to contribute, but i would need some assinstance and wouldn't like to polute the forum.

Should i try IRC ?

[galder.zamarreno]

Julio, IRC would be good. You could ask us questions there and get soonish replies.