Closed at15 closed 8 years ago
oops, no, it seems to be impossible to generate the range, unless there is some before after method
yep, the cleanup
method is called when one reducer finished its job, so I can get the max and min key there.
finished in ddcd677c96b6f826273eed983f6d2b02fba3b274
currently, the meta for each partition is generated in the index building phase, however, if using mellowtech's bptree, it seems I need to traverse the data twice to get the min and max key, because iterator is needed for bulk loading, and bptree does not provide max and min key.
change meta mr job source, meta table should still build based on index result because it need the position for the index file