DB2 to Snowflake Migration - SQL: Deriving a result table using 'values' clause - db2

Question

We are migrating off DB2 to Snowflake, and we have run into an issue with a lack of scope when using the 'values' clause. DB2 scope extends to all tables within the select query; while Snowflake is limited to that within the clause:
Valid in DB2:
select * from
t_1
join
table (values
(t_1.c1, ‘foo’)
,(t_1.c1, ‘bar’)
) as t_2 (c1, c2)
on t_1.c5 = t_2.c2
Not valid in Snowflake (slightly different syntax - remove 'table'):
select * from
t_1
join
(values
(t_1.c1, ‘foo’)
,(t_1.c1, ‘bar’)
) as t_2 (c1, c2)
on t_1.c5 = t_2.c2
This throws an error saying "...t_1 is undefined".
Does anyone know of an equivalent capability in Snowflake?
For some context: This construct sits behind a well embedded process for our business, and to rework it would mean significant investment.
It forms the basis for a dynamic ETL query which is passed into an ADF pipeline. There is a frontend which deploys components into this query - this deployable piece needs to be kept as simple as possible. It is the equivalent to the "(t_1.c1, ‘foo’)" line, so ideally any solution would keep this piece the same.
EDIT:
For some additional context...
In the example above, we have a placeholder which identifies the insert point of newly deployed items from the front end, like this:
select * from
t_1
join
table (values
(t_1.c1, ‘foo’)
,(t_1.c1, ‘bar’)
/*client_X_placeholder*/
) as t_2 (c1, c2)
on t_1.c5 = t_2.c2
union
select * from
t_1
join
table (values
(t_1.c1, ‘foo’)
,(t_1.c1, ‘bar’)
,(t_1.c2, 'barfoo')
/*client_Y_placeholder*/
) as t_2 (c1, c2)
on t_1.c5 = t_2.c2
We are pulling a series of metrics associated with clients, where the label is t_2.c2 and the result is t2.c1. This is still highly simplified relative to the real world. We need a simple way to update the placeholder with a new metric and its calculation - t2.c1 may be wrapped in a pile of functions and/or reside inside case statements, etc.

Answer 1

Lennart deserve this one, his code is so close
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
)
select *
from t_1
join lateral (values
(t_1.c1, 'foo'),
(t_1.c1, 'bar')
) as t_2(c1,c2)
on t_1.c5 = t_2.c2;
but that causes an internal error:
000603 (XX000): SQL execution internal error:
Processing aborted due to error 300002:4143448929; incident 4919401.
But if you put a SELECT * FROM in there:
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
)
select *
from t_1
join lateral (select * from values
(t_1.c1, 'foo'),
(t_1.c1, 'bar')
) as t_2(c1,c2)
on t_1.c5 = t_2.c2;
it works:
C1
C5
C1
C2
a
bar
a
bar
Lenart's try more brackets works also:
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
)
select *
from t_1
join lateral (
select * from (
values
(t_1.c1, 'foo'),
(t_1.c1, 'bar')
)
) as t_2(c1,c2)
on t_1.c5 = t_2.c2;

Answer 2

I have no idea whether this works in snowflake, but SQL requires you to declare t_2 as LATERAL if you want to reference t_1 inside of it. Db2 uses tables as a synonym for lateral (which works as well). So you could try:
select *
from t_1
join LATERAL (
values (t_1.c1, ‘foo’)
, (t_1.c1, ‘bar’)
) as t_2 (c1, c2)
on t_1.c5 = t_2.c2
Lateral was introduced in SQL99 and is supported by many DBMS including Db2.
In Db2, TABLE is typically used when selecting from a function returning a table:
SELECT * FROM TABLE ( myfun() )
which is somewhat similar so I guess that is why TABLE can act as a synonym for LATERAL (just guessing).
I did some googling and LATERAL is described in the docs for snowflake: join-lateral
EDIT:
Apparently, there is a limitation when it comes to joining the values clause in the derived table. The answer by Simeon Pilgrim contains a fix for this

Answer 3

hmm, that string is almost correlated subquery/join with values:
The first thought I have is:
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
)
select t_1.*, t_1.c1 as c1, t_2.c2
from t_1
join values ('foo'),
('bar') as t_2(c2)
on t_1.c5 = t_2.c2
;
but the part you wanting would look like (if it works):
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
)
select t_1.*, t_1.c1 as c1, t_2.c2
from t_1
join values
(t_1.c1, 'foo'),
(t_1.c1, 'bar')
as t_2(c1,c2)
on t_1.c5 = t_2.c2
;
but it doesn't work.... more thinking required.
So IF that block comes in as
(t_1.c1, 'foo'),
(t_1.c1, 'bar')
then you can mash it into block with %%stuff%%
and then rip it apart and hoop jump..
WITH t_1 (c1, c5) as (
SELECT 'a', 'bar'
), magic as (
SELECT
split(t.value, ',') as p
,trim(get(p,0),'"') as p0
,trim(get(p,1),' \'")') as p1
FROM table(split_to_table($$(t_1.c1, 'foo'),
(t_1.c1, 'bar')$$,'(')) as t
WHERE t.value <> ''
), hoop_jumping(c1, c2) as (
SELECT
case p0
WHEN 't_1.c1' then t_1.c1
WHEN 't_1.c5' then t_1.c5
end,
p1
FROM magic, t_1
)
SELECT t.*, h.*
FROM t_1 as t
JOIN hoop_jumping as h
ON t.c5 = h.c2
and as long as you are will to cover all the bases of what was selected:
C1
C5
C1
C2
a
bar
a
bar
but I suspect it makes more sense to inject that block into a snowflake scripting block, and use it as injected code.