Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Plain View
Drill >> mail # dev >> Example aggregate queries

Julian Hyde 2013-05-18, 21:06
Copy link to this message
Re: Example aggregate queries
I'll start with the answers.  It seems like you've made a bunch of
progress past the rest already.

> 1. is it valid for "carryovers" to be omitted or empty (I get NPE if I omit it)?

It should be.  If it doesn't work, we need to fix it.

> 2. what's the syntax for aggregation expression (I tried "COUNT()", "COUNT(*)", "COUNT")

The spec doesn't do a great job of covering expressions.  In this
case, COUNT(expr) where expr is another expression (anything but
another tree containing an aggregation function).

> 3. the spec seems to imply that I could write something as rich as "5 + SUM(salary) / COUNT()" for aggregation expression. Is that true?

it should be.  The rule is basically any expression except ones where
an aggregation is inside an aggregation.
> 4. "within" is a segment. Does that mean a holder field, like the "toppings" field in donuts.json?

It resets aggregations at each segment boundary.

> 5. is it valid for aggregations to be omitted or empty?

It should be.  If it doesn't work, it should be a jira.

> 6. in the plan spec [ https://docs.google.com/document/d/1QTL8warUYS2KjldQrGUse7zp8eA72VKtLOHwfXy6c7I/edit ] does "*" mean that a
field is optional and dagger mean system-generated? It does't say explicitly.

Star means optional.  Dagger means required outside a sequence, not
allowed in a sequence.  I added notes to the doc.

> 7. if I want to implement "SELECT deptno, COUNT(*) FROM emp GROUP BY deptno", I presume I will generate a plan that consists of scan-segment-collapsingaggregate?

Yes.  Although the count(*) would need to be remapped to
count(constant) or count(_MAP) since I don't believe we currently
support asterisk at the logical plan level.
> If the example queries answer those questions, no need to answer them explicitly.
> Julian