Json aggregation of data with missing values - sql

Question

I have a simple table of supplements
create table consumption (
account bigint not null,
date date not null,
supplement text not null check (supplement in ('multiVitamin', 'calMag', 'omega3', 'potassium', 'salt', 'antiOxidant', 'enzymes')),
quantity integer not null default 0
);
And I want to fetch what people have consumed per day. This would be an exmaple of my desired output
[
{
"date" : "2016-01-01",
"multiVitamin" : 7,
"calMag" : 0,
"omega3" : 3,
"potassium" : 3,
"salt" : 2,
"antiOxidant" : 0,
"enzymes" : 1
},
{
"date" : "2016-01-02",
"multiVitamin" : 2,
"calMag" : 1,
"omega3" : 1,
"potassium" : 2,
"salt" : 2,
"antiOxidant" : 0,
"enzymes" : 1
}
]
I'm confused how to get those values into a json object and coalesce so that I return 0 if there aren't any supplements entered for that day. So everyday should return all supplements. This is what I have so far but its very far from complete - it's at least fetching for the dates selected though.
WITH duration_amount AS (
SELECT date_trunc('day', date)::date AS date_group, json_build_object('quantity', SUM(consumption.quantity) )::jsonb->'quantity' as supplement
FROM consumption
WHERE account = 1667
GROUP BY date_group
)
SELECT DISTINCT date_group, supplement
FROM (
SELECT generate_series(date_trunc('day', '2016-10-20'::date), '2016-10-28'::date, '1 day') AS date_group
) x
LEFT JOIN duration_amount
USING (date_group)
ORDER BY date_group DESC

Answer 1

Example data:
insert into consumption values
(1667, '2016-10-21', 'multiVitamin', 1),
(1667, '2016-10-21', 'calMag', 2),
(1667, '2016-10-22', 'multiVitamin', 3),
(1667, '2016-10-22', 'calMag', 4),
(1667, '2016-10-22', 'omega3', 5);
You should prepare a template table containing rows for all possible values. In the example it will contain 14 rows (cross join 2 days with 7 supplements). Next, left join to it your table using coalesce() for missing values:
select
date_group::date as date,
supplement_group as supplement,
coalesce(quantity, 0) quantity
from generate_series('2016-10-21'::date, '2016-10-22', '1 day') as date_group
cross join (
values
('multiVitamin'), ('calMag'), ('omega3'),
('potassium'), ('salt'), ('antiOxidant'), ('enzymes')
) as supplements(supplement_group)
left join consumption
on date_group = date
and supplement = supplement_group
and account = 1667;
date | supplement | quantity
------------+--------------+----------
2016-10-21 | multiVitamin | 1
2016-10-21 | calMag | 2
2016-10-21 | omega3 | 0
2016-10-21 | potassium | 0
2016-10-21 | salt | 0
2016-10-21 | antiOxidant | 0
2016-10-21 | enzymes | 0
2016-10-22 | multiVitamin | 3
2016-10-22 | calMag | 4
2016-10-22 | omega3 | 5
2016-10-22 | potassium | 0
2016-10-22 | salt | 0
2016-10-22 | antiOxidant | 0
2016-10-22 | enzymes | 0
(14 rows)
The result can be easily aggregated to jsonb, see the full example here.