Postgresql - Reducing rows to one if a condition is met - postgresql

Question

I have a table that returns a series of objects. This is a small section of an example result (in JSON):
Query:
SELECT id, starttime, endtime, duration, type FROM things
Result:
{
"id": 3,
"starttime": "2016-09-15T03:27:09",
"endtime": "2016-09-15T03:31:43",
"duration": 274,
"type": "bad"
},
{
"id": 2,
"starttime": "2016-09-15T03:26:48",
"endtime": "2016-09-15T03:27:09",
"duration": 20,
"status": "good"
},
{
"id": 1,
"starttime": "2016-09-15T03:19:46",
"endtime": "2016-09-15T03:26:48",
"duration": 422,
"status": "bad"
},
I am trying to exclude anything less than 30 seconds - this is simple enough. However I also need to combine the top two together - their durations combined, the starttime with id 1's starttime and endttime as id 3's endtime. So this:
{
"starttime": "2016-09-15T03:19:46",
"endtime": "2016-09-15T03:31:43",
"duration": 696,
"status": "bad"
},
Is this possible in Postgres/SQL? I think I could figure something out in Java/C# but would rather do it in the query.

Answer 1

You can use ROW_NUMBER() :
SELECT MAX(CASE WHEN s.rnk = 1 THEN s.starttime END) as starttime,
MAX(CASE WHEN s.rnk = 2 THEN s.endtime END) as endtime,
SUM(s.duration) as dur,
--You didn't say which status you want
FROM (
SELECT t.*,
ROW_NUMBER() OVER(ORDER BY t.id) as rnk
FROM things t
WHERE t.duration >= 30) s
WHERE s.rnk < 3