One Billion Row challenge in Hs

[atravers]

Hmm:

• @emiruz

  If this problem was to be a motivational case for improvement – speaking as a consumer, as opposed to a language researcher – I’d personally prioritise changing whatever makes possible a fast immutable declarative single threaded solution, over consideration of the parallel aspects.

  Here is 1B rows in 25s in 7 LoC of obvious SQL: to the pragmatist, it speaks for itself.

• @chreekat

  One clear takeaway already is that somebody needs to look into GHC’s handling of n>4 threads. I have heard in the context of GHC’s own CI that anything more than N=4 creates a performance degradation.

…are there any extant proposals to add laziness to the SQL standard? If so, then we could use SQL as a target language (at least it would be portable :-).