MySQL, last in a group (again) - mysql

Question

I found tons of examples in Stackoverflow about how to select first and last row in a group but I cannot adapt them to my need. Alas my limited knowledge of MySQL doesn't help.
Some data (date_time, val1 and val2) are saved in random moments. I need to group them for 15 minutes and then calculate, for each 15 min interval:
number of row in the interval (done)
val1 min (done)
val1 avg (done)
val1 max (done)
first val2 in the group (done, easy with COALESCE)
last val2 in the group (here I need your help)
Here's my data, desired result, and best effort to date...
# what I would like to obtain:
#_timeslice _count _min_val1 _avg_val1 _max_val1 _first_val2 _last_val2
# 2017-11-01T00:00:00Z 4 90 100 110 200 210
# 2017-11-01T00:15:00Z 3 100 110 120 240 230
# 2017-11-01T00:30:00Z 2 110 120 130 270 265
# 2017-11-01T00:45:00Z 4 80 112.5 150 290 320
CREATE TABLE `test-table`
(`date_time` datetime, `val1` float, `val2` float);
INSERT INTO `test-table`
(`date_time`, `val1`, `val2`)
VALUES
('2017-11-01 00:00:00', 100, 200), # first 15 min
('2017-11-01 00:01:00', 110, 190),
('2017-11-01 00:02:05', 90, 220),
('2017-11-01 00:14:00', 100, 210),
('2017-11-01 00:15:00', 100, 240), # second 15 min
('2017-11-01 00:16:00', 110, 250),
('2017-11-01 00:28:00', 120, 230),
('2017-11-01 00:30:00', 110, 270), # third 15 min
('2017-11-01 00:44:59', 130, 265),
('2017-11-01 00:50:00', 120, 290), # fourth 15 min
('2017-11-01 00:55:00', 150, 300),
('2017-11-01 00:57:00', 100, 280),
('2017-11-01 00:58:00', 80, 320)
;
SELECT FROM_UNIXTIME(FLOOR(UNIX_TIMESTAMP(`date_time`)/900)*900) AS _timeslice,
COUNT(*) AS _count,
min(`val1`) as _min_val1,
avg(`val1`) as _avg_val1,
max(`val1`) as _max_val1,
coalesce(`val2`) as _first_val2 # returns the first val2 in the group
# ----> how to add here the last val2 for the group?
FROM `test-table`
GROUP BY _timeslice;
...and here's a fiddle of same SQLFiddle_grouping_15_min
Can somebody help me with the syntax to use?

Answer 1

SELECT a.*
FROM `test-table` a
JOIN
( SELECT FLOOR(UNIX_TIMESTAMP(date_time)/900) timeslice
, MIN(date_time) min_date_time
, MAX(date_time) max_date_time
FROM `test-table`
GROUP
BY timeslice
) b
ON b.timeslice = FLOOR(UNIX_TIMESTAMP(a.date_time)/900)
AND a.date_time IN(b.min_date_time,b.max_date_time);
+---------------------+------+------+
| date_time | val1 | val2 |
+---------------------+------+------+
| 2017-11-01 00:00:00 | 100 | 200 |
| 2017-11-01 00:14:00 | 100 | 210 |
| 2017-11-01 00:15:00 | 100 | 240 |
| 2017-11-01 00:28:00 | 120 | 230 |
| 2017-11-01 00:30:00 | 110 | 270 |
| 2017-11-01 00:44:59 | 130 | 265 |
| 2017-11-01 00:50:00 | 120 | 290 |
| 2017-11-01 00:58:00 | 80 | 320 |
+---------------------+------+------+

Answer 2

I found a simple solution with a straightforward syntax. It is possibly the easiest for a newbie like me:
SELECT FROM_UNIXTIME(FLOOR(UNIX_TIMESTAMP(`date_time`)/900)*900) AS _timeslice,
COUNT(*) AS _count,
min(`val1`) as _min_val1,
avg(`val1`) as _avg_val1,
max(`val1`) as _max_val1,
SUBSTRING_INDEX(GROUP_CONCAT(CAST(`val2` AS CHAR) ORDER BY date_time), ',', 1) AS _first_val2,
SUBSTRING_INDEX(GROUP_CONCAT(CAST(`val2` AS CHAR) ORDER BY date_time DESC), ',', 1) AS _last_val2
FROM `test-table`
GROUP BY _timeslice;
Fiddle at SQLFiddle_solved