trouble getting rid of duplicate rows - postgresql

Question

I have base data that has multiple lab items (A, B, C) that occur on same date.
id datetime_1 order_datetime item value
-----------------------------------------------------
1 9/1/21 09:57 9/2/21 04:21 A 13
1 9/1/21 09:57 9/2/21 04:21 B 8
1 9/1/21 09:57 9/2/21 04:21 C 11
1 9/1/21 09:57 9/3/21 16:00 A 10
1 9/1/21 09:57 9/3/21 16:00 B 4
1 9/1/21 09:57 9/3/21 16:00 C 7
1 9/2/21 02:30 9/2/21 04:21 A 13
1 9/2/21 02:30 9/2/21 04:21 B 8
1 9/2/21 02:30 9/2/21 04:21 C 11
1 9/2/21 02:30 9/3/21 16:00 A 10
1 9/2/21 02:30 9/3/21 16:00 B 4
1 9/2/21 02:30 9/3/21 16:00 C 7
I need output to show as :
id datetime_1 a_level b_level c_level
------------------------------------------------
1 9/1/21 09:57 13 8 11
1 9/2/21 02:30 13 8 11
My current code is:
with lab_setup as (
select id, datetime_1, row_number() over (partition by id, datetime_1 order by order_datetime) as lab_order)
from data
group by id, datetime_1, order_datetime
)
, lab_first as (
select id, datetime_1,
max(case when item = 'A' then value end) as a_level,
max(case when item = 'B' then value end) as b_level,
max(case when item = 'C' then value end) as c_level
from lab_setup
group by id, datetime_1, item, value
)
select *
from lab_first
group by id, datetime_1, a_level, b_level, c_level
The problem is that I keep getting duplicate rows in response to this code, looking like:
id datetime_1 a_level b_level c_level
------------------------------------------------
1 9/1/21 09:57 13 null null
1 9/1/21 09:57 null null 11
1 9/1/21 09:57 null 8 null
I've tried distinct, group by, max(case when) but so far it still provides multiple rows per datetime_1, which is not what I want. Does anyone have clue how to help merge these multiple rows into one?

Answer 1

You are close, but have a lot of extra, unnecessary work, making the query more complicated then needed. First off the query as posted did not produce the posted results. Seems somewhere along the line you grabbed the wrong query. The posted query is not valid. The lab_first cte uses item and value from lab_setup, but lab_setup does not contain either of them. Further it seems the purpose on of lab first is to derive the column lab_order, but that is not used afterward. Finally the main query selects only what would have been selected from the lab first cte without change. Thus neither cte is needed. Just incorporate the filtering (max functions) into the main. So (see demo)
select id
, datetime_1
, max(case when item = 'A' then value end) as a_level
, max(case when item = 'B' then value end) as b_level
, max(case when item = 'C' then value end) as c_level
from data
group by id, datetime_1
order by id, datetime_1;
Note on Demo: Just repeating the target data values but change the control values. How do I know afterword except for I not picking up the the exact same target values. Further a universe of 1 (id) is not sufficiently discriminating. For demo I changed of the value on a couple rows and added a few. The results look good, but you have to decide.