Re-surfacing the questions. This sounds like a common use case, butI am still stuck -:) Thanks!
On Thu, Mar 1, 2012 at 1:03 AM, prash987 prash987 <mysub...@gmail.com>wrote: > Thank You Norbert for your reply. > > > > I am still not sure how/if I can do this through Pig script, though? > Given the below adjacency list, > How would I find ancestors of R (P,C,A) in Pig? > > > > cat testgraph.txt > > A B:C > > B D > > C P > > P Q:R > > > > From the pig script I can get immediate parent, i.e. C > > But wouldn’t I need multiple iterations to get parent of C? > > > > -----------------script------------------------ > > A=LOAD 'testgraph.txt' USING PigStorage() AS (subject:chararray, > link:chararray); > > B = FOREACH A GENERATE subject, FLATTEN(STRSPLIT(link,':',2)) AS L; > > C = FILTER B BY $1 == 'P' > > ----------------------------------------------- > > > On Wed, Feb 29, 2012 at 10:16 PM, Norbert Burger <norbert.bur...@gmail.com > > wrote: > >> Prash, you can just model this tree as a simple graph adjacency list: >> >> A1,A2 >> A2,A3 >> A3,A4 >> A4,Am >> ... >> >> For nodes with more than one child, you simply extend each row >> horizontally. Child/parent/descendant/ancestor are straightforward >> applications of a traversal on this graph (BFS would be a good choice). >> >> Norbert >> >> On Wed, Feb 29, 2012 at 9:02 AM, prash987 prash987 <mysub...@gmail.com >> >wrote: >> >> > Hi All, >> > How do I represent hierarchical information in flat file and process it >> in >> > Pig? >> > >> > Let’s say I have objects of type A. >> > I want to have a Tree representation with their parent-child >> > relationships. >> > >> > In scenario 1: >> > A1 points to A2; A2 points to A3; A3 points to A4; A4 points to Am and >> > so on till An. >> > Given above definition; I want to be able to answer following : >> > >> > Child(A1) = A2 >> > Parent(A4) = A3 >> > Descendant(A1) = A2,A3,A4, Am… An >> > Ancestor(A4) = A3,A2,A1 >> > Ancestor (An) = Am,…A4,A3,A2,A1 >> > >> > Can this be represented in text file and queried in Pig. >> > >> > Appreciate any pointers/suggestions. >> > Thanks! >> > >> > >
