DataMapper

Comments and changes to this ticket

• 

  Dan Kubb June 5th, 2008 @ 01:35 AM

  jeff, I'm curious how you think this would work? I've not used hivedb before, but how do you see it fitting in with DataMapper?

• 

  Sam Smoot June 5th, 2008 @ 11:03 AM

  Yehuda and I have been talking about this a bit. I think we can do something in Repository to handle this. So instead of Repository -> Adapter being 1:1, it'd become 1:N(readers)+N(writers).

  Obviously the idea needs more fleshing out, but whatever the solution, I think we can extend the capabilities of the existing system rather than adding a bunch of new classes, and we have to be ultimately concerned that the Adapters should not be aware of the sharding at all.

  This way, eventually, you might have in-memory storage for readers for example, and database storage for a writer, and the readers are updated through a background synchronization process.

  Just to throw out something wild. ;-) But it's vitally important the Adapters remain ignorant as much as possible. dm-core should do all the heavy lifting as a general philosophy.

• 

  Adam French June 12th, 2008 @ 01:00 PM

  • → Milestone changed from “” to “1.x”
• 

  Diogo Almeida August 21st, 2008 @ 05:31 AM

  • → Tag changed from “” to “feature suggestion”

  For reference:

  ActiveRecord sharding with DataFabric, by FiveRuns: http://blog.fiveruns.com/2008/7/...

  Read slaves with ActsAsReadOnlyable plugin (not sharding but a lot of sharding setups use read slave clusters): http://revolutiononrails.blogspo...