Open sebastian opened 4 years ago
This would be useful to combine with an EXPLAIN
feature!
I.e. we could in fact always embed an EXPLAIN
query plan with the results we return. For queries that execute it could then include such information as data loss etc at different stages as well as potentially some rough execution time etc.
At Telefonica the admins have the concern that they don't really know, nor how to explain, why an analysts query produces a result that's different from what the analyst would have expected. They are requesting some sort of ability to get more insights into what happens at the different query processing stages. The things they said that would be the most useful to them would be along the lines of:
These don't need to be exact numbers (which we likely couldn't give them without privacy risks), but could still be useful in answering questions such as why is there no results after anonymization, because they could help pinpoint that the data vanished at "subquery x" and furthermore pinpoint whether a problem might have been a result of Oracle not delivering data (i.e. an Oracle problem) or Aircloak dropping the ball.
Some creativity is required here.
Ideas welcome!