I think having SQL as a way to generate benchmark has some value, and we
should be open to having that option in e2e harness as well. But I don't
see that as a blocker.
In some cases, I would expect that writing an alternative pig-latin
query to generate benchmark might not be easy, and there is also the
danger that the alternative script also has the same bug which results
buggy benchmark data.
-Thejas
On 7/14/11 3:51 PM, Thejas Nair wrote:
On 7/14/11 2:39 PM, Alan Gates wrote:
I have posted a proposal for changes in Pig's testing that I would
like to make.
https://cwiki.apache.org/confluence/display/PIG/PigTestProposal Please
take a look and provide feedback.
Alan.
+1 for the proposal.
-Thejas