|
The unique index on start and end can not ensure B IDs are not overlapped, eg 1, 10 and 2,, 20. This results lines crossed when being joined even though after sorted. This can be clearly seen given a traditional merge join, eg, t1 has rows 1,3,5 and t2 has rows 1,5,9 (two joined lines never cross between t1 and t2 and therefore inner loop needs not start from beginning again every time with outer loop going on in comparison with nested loop join). |
|