How to pivot columns from EAV by group (SQL Server) - sql-server

Question

I have an EAV table in SQL Server and would like to pivot some of the table into columns.
EXAMPLE DATA
| id | key | value |
|-----|---------------|-----------------|
| 001 | Name | Hand Protection |
| 001 | Family | Gloves |
| 001 | Brand | Memphis |
| 001 | Style | 6030 |
| 001 | Material | Nitrile |
| 002 | Name | Hand Protection |
| 002 | Family | Gloves |
| 002 | Brand | Gladiator |
| 002 | Color | Black |
| 002 | Size | Large |
| 003 | Name | Head Protection |
| 003 | Family | Hats |
| 003 | Brand | Gladiator |
| 003 | Color | Black |
| 003 | Size | Large |
For each name I would like to aggregate
DESIRED OUTPUT
| name | list |
|-----------------|---------------------|
| Hand Protection | [{"attribute": "Family", "values": ["Gloves"], "count": 2},{"attribute": "Brand", "values": ["Memphis", "Gladiator"], "count": 2},{"attribute": "Style", "values": [6030], "count": 1},{"attribute": "Material", "values": ["Nitrile"], "count": 1},{"attribute": "Color", "values": ["Black"], "count": 1},{"attribute": "Size", "values": ["Large"], "count": 1} |
| Head Protection | [{"attribute": "Family", "values": ["Hats"], "count": 1},{"attribute": "Brand", "values": ["Gladiator"], "count": 1},{"attribute": "Color", "values": ["Black"], "count": 1},{"attribute": "Size", "values": ["Large"], "count": 1}
SQL
-- Get reference item for attributes
select distinct
[id]
into #tmp
from data..item_attributes
where [key] = 'name'
and [value] = #category
-- Get attributes for a given name
select distinct
'attribute' = [key],
'values' = (
select distinct
'value' = [value]
from data..item_attributes
where [key] = #attribute
for json path
),
'count' = count(*)
from data..item_attributes
where id in (select * from #tmp)
-- Make new table
update table set [list] = (
**???***
)
where [name] = [key]
How can I pivot out associated keys and append them into a new column with a count?

Answer 1

db<>fiddle
SELECT MAX(CASE YT.[key] WHEN 'Name' THEN YT.[Value] END) AS [name]
, ( SELECT [key] AS attribute
, JSON_QUERY(REPLACE(REPLACE(
( SELECT value
FROM dbo.YourTable AT
WHERE AT.id = YT.id AND AT.[key] = ZT.[key]
FOR JSON PATH), '{"value":', ''), '"}', '"')) AS [values]
, COUNT(*) AS [count]
FROM dbo.YourTable ZT
WHERE ZT.id = YT.id AND Zt.[key] <> 'Name'
GROUP BY id, [key]
FOR JSON PATH) AS list
FROM dbo.YourTable YT
GROUP BY YT.id;
Returns something like this
name list
HandProtection [{"attribute":"Brand","values":["Memphis"],"count":1},{"attribute":"Family","values":["Gloves"],"count":1},{"attribute":"Material","values":["Nitrile"],"count":1},{"attribute":"Style","values":["6030"],"count":1}]
HandProtection [{"attribute":"Brand","values":["Gladiator"],"count":1},{"attribute":"Color","values":["Black","White"],"count":2},{"attribute":"Family","values":["Gloves"],"count":1},{"attribute":"Size","values":["Large"],"count":1}]
HeadProtection [{"attribute":"Brand","values":["Gladiator"],"count":1},{"attribute":"Color","values":["Black"],"count":1},{"attribute":"Family","values":["Hats"],"count":1},{"attribute":"Size","values":["Large"],"count":1}]
This one turned out harder than it should have been. For instance had to use the JSON_QUERY(REPLACE(REPLACE kludge to produce a JSON array of strings (from the JSON array of objects FOR JSON PATH insists on doing).