Hope you guys can help me little bit with this one.
I have a table in HBase with following structure (simplified)
key: epoch in seconds rounded to the day.
cq: 0..23 (hourly counters)
I have no problem to create horizontal VIEW so I can do query like
SELECT k1, k2, ts, 01,02,03...23 FROM t1 GROUP BY
but I would like an option to have a VIEW in phoenix where I will build ts by doing ts+hour*3600 (and may be even convert it to DATE)
so I can have
SELECT k1, k2, ts, count() FROM t1 GROUP BY ... what would resulted in 24 records per day with ts being DDMMYYY hh:00 - (basically turn horizontal table to vertical)
SELECT k1, k2, ts, hour, count() FROM t1 GROUP BY ...
Normally in my old days in SQL I could use some UNPIVOT or CROSS APPLY (depends on RDBMS) or simply do some union like
SELECT k1, k2, ts+1*3600 AS EPOCH, h01 as COUNT_PER_HOUR FROM t1
SELECT k1, k2, ts+2*3600 AS EPOCH, h02 as COUNT_PER_HOUR FROM t1
SELECT k1, k2, ts+23*3600 AS EPOCH, h23 as COUNT_PER_HOUR FROM t1
but it does not look like I can do it in phoenix
#1 it does not look like I can create VIEW of the VIEW
#2 there is no UNION (yes I know work is happening on UNION ALL now)
I hope someone has an idea how I can do it in phoenix with what we have now and if it even possible.
An if not what will it take to make that happens? Even if lets say we got UNION ALL would it be enough end how well those kind of queries will live in phoenix?
Would it be insane to even dream about lets say UNPIVOT when we would be able to point it to cf:regexstring (for selecting list of columns)
and turn it into row ? Even just UNPIVOT (value for cell in (h:01,h:02...h:23) unpiv; What kind of amount of work we are looking for to implement something like this ?