create window group based on value of preceding row - sql

Question

I have a table like so:
#standardSQL
WITH k AS (
SELECT 1 id, 1 subgrp, 'stuff1' content UNION ALL
SELECT 2, 2, 'stuff2' UNION ALL
SELECT 3, 3, 'stuff3' UNION ALL
SELECT 4, 4, 'stuff4' UNION ALL
SELECT 5, 1, 'ostuff1' UNION ALL
SELECT 6, 2, 'ostuff2' UNION ALL
SELECT 7, 3, 'ostuff3' UNION ALL
SELECT 8, 4, 'ostuff4'
)
and like to group based on the subgrp value to re-create the missing grp: if subgrp value is smaller than previous row, belongs to same group.
Intermediate result would be:
| id | grp | subgrp | content |
| 1 | 1 | 1 | stuff1 |
| 2 | 1 | 2 | stuff2 |
| 3 | 1 | 3 | stuff3 |
| 4 | 1 | 4 | stuff4 |
| 5 | 2 | 1 | ostuff1 |
| 6 | 2 | 2 | ostuff2 |
| 7 | 2 | 3 | ostuff3 |
| 8 | 2 | 4 | ostuff4 |
on which I can then apply
SELECT id, grp, ARRAY_AGG(STRUCT(subgrp, content)) rcd
FROM k ORDER BY id, grp
to have I nice nested structure.
Notes:
with 'id' ordered, subgrp is always in sequence so no 3 before 2
groups are not always 4 subgrp's - this is just to illustrate so cannot hardcode
Problem: how can I (re)create the grp column here ? I played with several Window functions to no avail.
EDIT
Although Gordon's answer work, it took 3min over 104M records to run and I had to remove an ORDER BY on the final resultset because of Resources exceeded during execution: The query could not be executed in the allotted memory. ORDER BY operator used too much memory.
Anyone having an alternative solution for large dataset ?

Answer 1

A simple way to assign the group is to do a cumulative count of the subgrp = 1 values:
select k.*,
sum(case when subgrp = 1 then 1 else 0 end) over (order by id) as grp
from k;
You can also do it your way, using lag() and a cumulative sum. That requires a subquery:
select k.*,
sum(case when prev_subgrp = subgrp then 0 else 1 end) over (order by id) as grp
from (select k.*,
lag(subgrp) over (order by id) as prev_subgrp
from k
) k

Answer 2

Below can potentially perform better - but has limitation - I assume there is no gaps in numbering within subgroups and respective ids
#standardSQL
WITH k AS (
SELECT 1 id, 1 subgrp, 'stuff1' content UNION ALL
SELECT 2, 2, 'stuff2' UNION ALL
SELECT 3, 3, 'stuff3' UNION ALL
SELECT 4, 4, 'stuff4' UNION ALL
SELECT 5, 1, 'ostuff1' UNION ALL
SELECT 6, 2, 'ostuff2' UNION ALL
SELECT 7, 3, 'ostuff3' UNION ALL
SELECT 8, 4, 'ostuff4'
)
SELECT
ROW_NUMBER() OVER(ORDER BY id) grp,
rcd
FROM (
SELECT
MIN(id) id,
ARRAY_AGG(STRUCT(subgrp, content)) rcd
FROM k
GROUP BY id - subgrp
)
result is
Row grp rcd.subgrp rcd.content
1 1 1 stuff1
2 stuff2
3 stuff3
4 stuff4
2 2 1 ostuff1
2 ostuff2
3 ostuff3
4 ostuff4