[GitHub] incubator-spark pull request: [Proposal] Adding sparse data suppor...

[shivaram]

Github user shivaram commented on the pull request:

    https://github.com/apache/incubator-spark/pull/575#issuecomment-35007797
  
    @srowen  Thanks for the summary.
    
    For the external API I wouldn't mind using (i, j, value) -- It results in 
larger files and a `groupBy` to get to a row / column but I agree that this is 
the most compatible format across various tools. On a related note is there an 
external representation that GraphX uses that can be leveraged ? /cc  @jegonzal 
@ankurdave
    
    About the top 3 ideas, I'd also prefer (2) first and then (1). I am not 
sure what the façade is buying us as these are internal library calls that we 
can change going forward.