I have a Postgres DB with tables for templates and responses, each template defines how many rows and items it has, the response places the items in rows. I want to be able to query a merged response with the "average" placement across all responses for any template.
template_post
id | data |
---|---|
1 | {"items":[{"id":0},{"id":1},{"id":2},{"id":3}],"rows":[{"id":"0","title":"First row"},{"id":"1","title":"Second row"},{"id":"3","title":"Third row"},{"id":"4","title":"Fourth row"}]} |
2 | ... |
response_post
id | template_id [FK] | data |
---|---|---|
1 | 1 | {"filled_rows":[{"row_id":4,"item_ids":[0,1,3]}]} |
2 | 1 | {"filled_rows":[{"row_id":2,"item_ids":[2]}]} |
3 | 1 | {"filled_rows":[{"row_id":0,"item_ids":[1]},{"row_id":2,"item_ids":[2]},{"row_id":4,"item_ids":[3]}]} |
In this case, the query should return the following merged response:
{
"filled_rows": [
{
"row_id": 0,
"item_ids": [1]
},
{
"row_id": 2,
"item_ids": [2]
},
{
"row_id": 4,
"item_ids": [0,3]
}
]
}
In cases of ambiguity, when an item appears the same number of times in more than one row, it should be placed in the first one of those rows.
You need a few levels of aggregation here.
filled_rows
and item_ids
into separate rows.item_id
and take the min row_id
.row_id
and take a JSON array of item_ids
.SELECT
tp.id,
r.result
FROM template_post tp
CROSS JOIN LATERAL (
SELECT
jsonb_build_object(
'filled_rows', jsonb_agg(byRow.*)
) AS result
FROM (
SELECT
byItem.row_id,
jsonb_agg(byItem.item_id) AS item_ids
FROM (
SELECT
MIN((r.value ->> 'row_id') ::int) AS row_id,
i.value::int AS item_id
FROM response_post rp
CROSS JOIN jsonb_array_elements(rp.data -> 'filled_rows') r
CROSS JOIN jsonb_array_elements_text(r.value -> 'item_ids') i
WHERE rp.template_id = tp.id
GROUP BY
i.value::int
) byItem
GROUP BY
byItem.row_id
) byRow
) r;