new to snowflake - issue with SQL statement with joins and pivoting values - sql

Question

I'm trying to model product by have a product table linking to characteristic table with 3 many to many table
eg
Product (productId, brand, description, model)
ProductCharacteristic (productId, characteristicId)
Characteristic (characteristicId, type, name, value)
From this structure I'm trying pivot back into one result to show product fields with and characteristic.type characteristic.value into a select statement.
For example:
select distinct
p.ProductId, p.name, p.description, p.brand, p.model, body, DoorNum
from
Product p
left join
ProductionCharacteristic pc on p.ProductId = pc.ProductId,
(select CharacteristicId, "'Body'" as body, "'DoorNum'" as DoorNum
from Characteristic c
pivot (max(c.name) for c.type IN ('Body' , 'DoorNum'))) temp
where
pc.CharacteristicId = temp.CharacteristicId
Issue1: is any_Value an Aggregate function? it pivot doesnt like it, i'm using max() instead
Issue2: the above select returns 3 rows. is there a way flatten it down to one record?

Answer 1

Notes
I must insist that you NEVER use mixed join expressions. Meaning that if you are going to use the keywords JOIN and LEFT JOIN then don't use , and vice versa.
You didn't make it clear what you expected your output to look like so I am going with a comma separated list, but will also be showing how you can use an array.
I put this in as a comment, but worth repeating: ANY_VALUE() is an aggregate function, but doesn't work inside the context of PIVOT
https://community.snowflake.com/s/question/0D50Z00008uVHTYSA4/anyvalue-does-not-work-on-pivot
Query Options
Option 1 - Inline Pivot (Not recommended as it's harder to read and the pivot sub-query can't be reused)
select
p.ProductId
, p.name
, p.description
, p.brand
, p.model
, listagg(temp.body,', ') as body
, listagg(temp.DoorNum,', ') as doornum1
, arrayagg(temp.DoorNum) as doornum2
from
Product p
left join
ProductionCharacteristic pc
on p.ProductId = pc.ProductId
join (
select CharacteristicId, body, doornum
from Characteristic c
pivot (max(c.name) for c.type IN ('Body' , 'DoorNum')) as p (CharacteristicId, value, body, doornum)
) temp
on pc.CharacteristicId = temp.CharacteristicId
group by 1,2,3,4,5
;
Option 2 - Pivot in CTE (Recommended)
with temp as (
select CharacteristicId, body, doornum
from Characteristic c
pivot (max(c.name) for c.type IN ('Body' , 'DoorNum')) as p (CharacteristicId, value, body, doornum)
)
select
p.ProductId
, p.name
, p.description
, p.brand
, p.model
, listagg(temp.body,', ') as body
, listagg(temp.DoorNum,', ') as doornum1
, arrayagg(temp.DoorNum) as doornum2
from
Product p
left join
ProductionCharacteristic pc
on p.ProductId = pc.ProductId
join temp
on pc.CharacteristicId = temp.CharacteristicId
group by 1,2,3,4,5
;
Option 3 - Lateral Join (Like a correlated subquery)
select
p.ProductId
, p.name
, p.description
, p.brand
, p.model
, temp.body
, temp.doornum1
, temp.doornum2
from
Product p
join lateral (
select listagg(iff(type = 'Body',name,null),', ') as body
,listagg(iff(type = 'DoorNum',name,null),', ') as doornum1
,arrayagg(iff(type = 'DoorNum',name,null)) as doornum2
from Characteristic c
join ProductionCharacteristic pc
on pc.CharacteristicId = c.CharacteristicId
where p.ProductId = pc.ProductId
) temp
;
Hope this helps.