Q3/Py implementation should produce p1|p2 output

[szarnyasg]

We select the upper triangle to ensure p1 < p2 in the dense ids. This only maps back to the sparse ids correctly, we need to make sure that they were originally sorted.

Temporary workaround:

sort -g person.csv > person-sorted.csv
mv person-sorted.csv person.csv 
less person.csv 

cc @sandordavid1124

A permanent solution would be to sort the persons.csv file beforehand.

[marci543]

We can keep the current optimization (using of TRIU) without sorting the persons. The query is symmetric and comparison of IDs is only used to filter duplicates and have a consistent output. A simple solution is to transform the output by swapping the IDs if necessary:

p1_id=...
p2_id=...
if(p1_id>p2_id):
  temp=p1_id
  p1_id=p2_id
  p2_id=temp