Postgres sum values from json only if pair of values exist - sql

Question

My json field looks something like this
{
"Ui": [
{
"element": "TG1",
"mention": "in",
"time": 123
},
{
"element": "TG1",
"mention": "out",
"time": 125
},
{ "element": "TG2",
"mention": "in",
"time": 251
},
{
"element": "TG2",
"mention": "out",
"time": 259
},
{ "element": "TG2",
"mention": "in",
"time": 251
}
]
}
I am trying to get the sum of difference of time per element which is as below
| element | Timespent |
| TG1 | 2 |
| TG2 | 8 |
The problem is ideally for every "in" element there should be an "out" element which is clearly not the case in the above example. I want to only calculate the difference of this pair of values and ignore any value that doesn't have a corresponding out to a in. How can I do that?
Below is what I am using to get the time difference
select element, sum(time) as time_spent
from my_table
cross join lateral (
select
value->>'element' as element,
case value->>'mention' when 'in' then -(value->>'time')::numeric else (value->>'time')::numeric end as time
from json_array_elements(json_column->'Ui')) as elements
group by 1
order by 1

Answer 1

I was not sure about you json_column attribute - you need to group by it in order to not mix values between rows, so I included it into window aggregations in CTE part. But you don't have it in your results, so I skipped it in final qry as well. In short - you can check if order number is even and equal to max order number and then just skip it:
select json_column
, e->>'element' element
, case when
mod(lag(i) over (partition by json_column::text ,e->>'element' order by i),2) = 0
and
max(i) over (partition by json_column::text ,e->>'element') = i
then true else false end "skip"
, case when e->>'mention' = 'in' then -(e->>'time')::int else (e->>'time')::int end times
from my_table, json_array_elements(json_column->'Ui') with ordinality o(e,i)
)
select element, sum (times)
from logic
where not skip
group by element
;
element | sum
---------+-----
TG1 | 2
TG2 | 8
(2 rows)