non unique indexes for paranoid docs

[tomdmaguire]

fixes paranoid issue described in #172 (updated for latest version of code)

[leemhenson]

+1

[digitalplaywright]

Thanks! This is very interesting. I am wondering if it would be better to factor out the index creation into a concern so that people that use the old index format can still upgrade to new versions without recreating indexes. Thought or ideas on better solutions?

[dblock]

Bunp @tomdmaguire, care to answer @digitalplaywright's question above? Thanks!

[johnnyshields]

FYI MongoDB 3.2 will support partial indexes which will make for a better solution here--you can declare unique as long as delete_at does not exist.

http://docs.mongodb.org/manual/release-notes/3.1-dev-series/

[dblock]

@johnnyshields @tomdmaguire After merging #172 (via #227) I read everything there's to read here and there and I don't understand what the problem is with paranoid documents and the slug index. Can someone explain it to me like a 2 yo and tell me what we should do with this PR? Thanks.

[johnnyshields]

• this gem adds a unique DB index on the _slugs field. this means if Doc A has slugs: ["foo"] and I try to create Doc B with slugs: ["foo"], the DB itself will not allow Doc B to be saved.
• paranoid docs are "plain old docs" from the DB perspective. so if I have a paranoid deleted doc with slugs: ["foo"], deleted: true, the DB doesn't care about deleted: true. As long as that deleted paranoid doc is lingering in the DB, I can't save a new doc with slugs: ["foo"]
• one might think: ok, the solution here is to set slugs: nil on the deleted doc. prior to this PR Better handling of unique and sparse index constraints. #227 this wasn't possible because the index was not sparse, meaning that two docs with _slugs: null would cause the unique index validation to fail. Now that PR Better handling of unique and sparse index constraints. #227 is merged, this approach should be possible.

[dblock]

Okay this is a very clear explanation!

So I think this particular PR can be closed, because it's fully addressed in #227 which effectively "fixes" the scenario that was attempted here by introducing a sparse index? And the specs now create database-level indexes so we should be good.

[johnnyshields]

Does paranoid doc remove the value for _slugs? It needs to if it's not doing so already.

[dblock]

Oh yeah, https://github.com/digitalplaywright/mongoid-slug/blob/382b0999212518b222da93e983d31b1a34feac83/lib/mongoid/slug.rb#L108.

Closing ;)