Partitioning data by date

[Ghislain Fourny]

Hi Andrew,

If you are interested, with the tumbling windows Michael is mentioning (again, assuming your metrics are ordered by date) it could look like this:

<metrics>{
  for tumbling window $w in doc("metricsCollection.xml")//metric
  start at $s when fn:true()
  end $m at $e next $n when
    xs:date(xs:dateTime($m/timestamp)) ne xs:date(xs:dateTime($n/timestamp))
  return
  <metric>
    <date>{xs:date(xs:dateTime($m/timestamp))}</date>
    <num_actions>{$e - $s + 1}</num_actions>
  </metric>
}</metrics>

Tumbling windows cannot overlap, so you can start one whenever it is possible, and end a window as soon as the date in an item $m is not the same as in the following item $n. $s and $e indicate the start and end positions of the current window.

You can test this query in Zorba (http://try.zorba-xquery.com).

Kind regards,
Ghislain


> One would also expect a solution using XQuery 1.1 "tumbling windows" to be O(n)