Added functions for join-filtering.

[andrewsmartin]

@nevillelyh @rav RFC please. I often see code to filter an SCollection[(K, V)] by another one of type SCollection[K], where (k, v) pairs are filtered out if k is not present on the right hand side. It usually looks like this:

val s1: SCollection[(K, V)] = getIt()
val s2: SCollection[K] = getThat()

val filtered: SCollection[(K, V)] = 
    s1.join(s2.distinct.map(_, ())).map{ case(k, (v, _)) => (k, v)}

I find this pattern doesn't really speak to the high level intent of the operation, and personally it always takes me a bit of staring to understand what's happening when I see it. So I thought it might be useful to include in the core API. The equivalent of the above would look like this:

val s1: SCollection[(K, V)] = getIt()
val s2: SCollection[K] = getThat()

val filtered: SCollection[(K, V)] = s1.filterJoin(s2)

Do you consider this useful at all? Is there a more obvious way of doing this that I'm not thinking about? If you do think it's useful, any suggestions on naming? filterJoin is the first thing that came to mind since I'm implementing this on top of join but there's probably a better name. Similar functionality is achieved in a plain SCollection[K] via intersect, which internally converts both collections to a dummy SCollection[(K, V)], and inner join is the equivalent for SCollection[(K, V)], SCollection[(K, W)]. Maybe keyIntersect is a better name.

[codecov-io]

Current coverage is 75.26% (diff: 100%)

Merging #202 into master will increase coverage by 0.03%

@@             master       #202   diff @@
==========================================
  Files            61         61          
  Lines          2160       2163     +3   
  Methods        1998       1977    -21   
  Messages          0          0          
  Branches        162        186    +24   
==========================================
+ Hits           1625       1628     +3   
  Misses          535        535          
  Partials          0          0          

Powered by Codecov. Last update 72b4a31...3da9919

[ravwojdyla]

Would not use join in the name (the implementation should most likely use cogroup instead of joins), some variation of intersect (intersectByKey) makes more sense.

[ravwojdyla]

Does it make sense to test on empty left&right hand side?

[ravwojdyla]

+1 pls squash commits