UNNEST/Flatten in snowflake - snowflake-cloud-data-platform

Question

I have a MySQL query that I want to convert into a snowflake.
MySQL query :
WITH t AS (
select id,
date,
copt,
split(copt, '|') [ 1 ] as "abc",
split(copt, '|') [ 2 ] as "def",
split(copt, '|') [ 3 ] as "xyz",
from tablename
where id in (
123,
456,
789,
)
and date >= dateadd('day', -6, to_date('2021-12-17'))
and date <= date '2021-12-17'
and copt like '%|%|%|%|%|%|%|%|%|%|%'
)
SELECT t.id,
t.date,
catId,
REPLACE(productId, '_', '') as productId,
FROM t
CROSS JOIN UNNEST(
split(t."abc", '_'),
split(t."def", '_'),
split(t."xyz", '_'),
) as x(catId, productId, quantity)
where productid != ''
order by id
I have tried to replace UNNEST() with FLATTEN() but didn't succeed.
Can anyone help me out here to convert this query from MySQL to Snowflake?

Answer 1

It doesn't appear the MySQL has UNNEST support, so I could not read the manual on that.
Also you have not provided any example input, and the output you are expecting, But assuming you abc, def, & xzy are arrays of independent sizes, and you are want all permutations the following SQL should work:
WITH cte_t(id,date, abc, def, xyz, productid) AS (
SELECT * FROM VALUES
(1, '2022-01-23'::date, 'a_b_c_d', 'd_e_f', 'x_y_w', 'not empty'),
(2, '2022-01-23'::date, 'a_d', 'd_f', 'y_w', 'not empty'),
(3, '2022-01-23'::date, 'a_c_d', 'f', 'y1_y2_w', 'not empty')
)
SELECT
t.id,
t.date,
a.value AS catid,
b.value AS productid,
c.value AS quantity
FROM cte_t AS t
,LATERAL SPLIT_TO_TABLE(t.abc, '_') a
,LATERAL SPLIT_TO_TABLE(t.def, '_') b
,LATERAL SPLIT_TO_TABLE(t.xyz, '_') c
WHERE productid != ''
gives:
ID
DATE
CATID
PRODUCTID
QUANTITY
1
2022-01-23
a
d
x
1
2022-01-23
b
d
x
1
2022-01-23
c
d
x
1
2022-01-23
d
d
x
1
2022-01-23
a
e
x
1
2022-01-23
b
e
x
1
2022-01-23
c
e
x
1
2022-01-23
d
e
x
1
2022-01-23
a
f
x
1
2022-01-23
b
f
x
1
2022-01-23
c
f
x
1
2022-01-23
d
f
x
1
2022-01-23
a
d
y
1
2022-01-23
b
d
y
1
2022-01-23
c
d
y
1
2022-01-23
d
d
y
1
2022-01-23
a
e
y
1
2022-01-23
b
e
y
1
2022-01-23
c
e
y
1
2022-01-23
d
e
y
1
2022-01-23
a
f
y
1
2022-01-23
b
f
y
1
2022-01-23
c
f
y
1
2022-01-23
d
f
y
1
2022-01-23
a
d
w
1
2022-01-23
b
d
w
1
2022-01-23
c
d
w
1
2022-01-23
d
d
w
1
2022-01-23
a
e
w
1
2022-01-23
b
e
w
1
2022-01-23
c
e
w
1
2022-01-23
d
e
w
1
2022-01-23
a
f
w
1
2022-01-23
b
f
w
1
2022-01-23
c
f
w
1
2022-01-23
d
f
w
2
2022-01-23
a
d
y
2
2022-01-23
d
d
y
2
2022-01-23
a
f
y
2
2022-01-23
d
f
y
2
2022-01-23
a
d
w
2
2022-01-23
d
d
w
2
2022-01-23
a
f
w
2
2022-01-23
d
f
w
3
2022-01-23
a
f
y1
3
2022-01-23
c
f
y1
3
2022-01-23
d
f
y1
3
2022-01-23
a
f
y2
3
2022-01-23
c
f
y2
3
2022-01-23
d
f
y2
3
2022-01-23
a
f
w
3
2022-01-23
c
f
w
3
2022-01-23
d
f
w
Ok, that was more data the I was expecting, but there you go.
If how ever those three arrays are of the same size then you can use one SPLIT_TO_TALBE and two SPLIT_PART:
WITH cte_t(id,date, abc, def, xyz, productid) AS (
SELECT * FROM VALUES
(1, '2022-01-23'::date, 'a_b_c', 'd_e_f', 'x_y_w', 'not empty')
)
SELECT
t.id,
t.date,
a.value AS catid,
split_part(t.def, '_', a.index) AS productid,
split_part(t.xyz, '_', a.index) AS quantity
FROM cte_t AS t
,LATERAL SPLIT_TO_TABLE(t.abc, '_') a
WHERE productid != ''
giving
ID
DATE
CATID
PRODUCTID
QUANTITY
1
2022-01-23
a
d
x
1
2022-01-23
b
e
y
1
2022-01-23
c
f
w