I tried (in ultimate CS nerd fashion) to eliminate the OR's using DeMorgan's laws, and that might lead me to some refactoring of the logic that'll make it work with just sequential joins, but I'm not sure how far it'll take me.
Anyone else have an idea? On Mon, May 9, 2011 at 12:28 PM, Dmitriy Ryaboy <dvrya...@gmail.com> wrote: > My first thought is that you can get there through a combination of > joins, filters, and unions, but there's probably a better way. > > > > On Mon, May 9, 2011 at 7:31 AM, Mark Laczin <mark.lac...@gmail.com> wrote: >> Can the same pattern be used to deal with OR's in a WHERE? >> >> That is, >> SELECT somestuff, somemorestuff >> FROM data1 d1, data2 d2 >> WHERE (d1.d = "somestring" OR d1.d = d2.d) AND (d1.a = d2.a OR d1.b = d2.c) >> >> ? >> >> Quite an odd situation, I know. >> >> -Mark >> >> On Thu, May 5, 2011 at 6:12 PM, Dmitriy Ryaboy <dvrya...@gmail.com> wrote: >>> If you are joining on the same field, you can do that in one statement: >>> >>> x = join a on f1, b on f1, c on f2, d on f3...; >>> >>> If however it's more of a star-like join, you have to do several joins: >>> >>> x = join a on f1, b on f1; >>> xx = join x on b::f2, c on f2; >>> .... >>> >>> D >>> >>> On Thu, May 5, 2011 at 12:06 PM, Mark Laczin <mark.lac...@gmail.com> wrote: >>>> This is more of a "how can I do this" question. >>>> Imagine you have an sql query like the following: >>>> select a.f1, a.f2, a.f3, b.f1, b.f2, c.f1, c.f2, d.f1, d.f2 >>>> from tableA a, tableB b, tableC c, tableD d >>>> where >>>> a.f1 = b.f1 >>>> and b.f2 = d.f1 >>>> and c.f2 = b.f1 >>>> /* etc.... */ >>>> What's the best way to replicate that query in Pig? >>>> It's obvious to me that a simple thing could be done with a simple join. >>>> That is, if you just had >>>> where a.f1 = b.f1 >>>> You can just JOIN and then foreach generate. >>>> But with more than one clause in the where, I'm not sure of a good >>>> approach... >>>> -Mark >>>> >>> >> >
