How to map multiple nested parent and child ids in BigQuery - google-bigquery

Question

I have the data produced by the following query:
with example_data as (
select 'A' as category, 1 as child_id, 0 as parent_id, 'a' as label
union all select 'A' as category, 2 as child_id, 1 as parent_id, 'b' as label
union all select 'A' as category, 3 as child_id, 1 as parent_id, 'c' as label
union all select 'A' as category, 4 as child_id, 1 as parent_id, 'd' as label
union all select 'A' as category, 5 as child_id, 2 as parent_id, 'e' as label
union all select 'A' as category, 6 as child_id, 2 as parent_id, 'f' as label
union all select 'A' as category, 7 as child_id, 2 as parent_id, 'g' as label
union all select 'A' as category, 8 as child_id, 3 as parent_id, 'h' as label
union all select 'A' as category, 9 as child_id, 3 as parent_id, 'i' as label
union all select 'A' as category, 10 as child_id, 3 as parent_id, 'j' as label
union all select 'B' as category, 1 as child_id, 0 as parent_id, 'k' as label
union all select 'B' as category, 2 as child_id, 1 as parent_id, 'l' as label
union all select 'B' as category, 3 as child_id, 1 as parent_id, 'm' as label
union all select 'B' as category, 4 as child_id, 1 as parent_id, 'n' as label
union all select 'B' as category, 5 as child_id, 2 as parent_id, 'o' as label
union all select 'B' as category, 6 as child_id, 2 as parent_id, 'p' as label
union all select 'B' as category, 7 as child_id, 2 as parent_id, 'q' as label
union all select 'B' as category, 8 as child_id, 3 as parent_id, 'r' as label
union all select 'B' as category, 9 as child_id, 3 as parent_id, 's' as label
union all select 'B' as category, 10 as child_id, 3 as parent_id, 't' as label
)
select *
from example_data
In this table, we have some labels where each one has a parent_id. What I want to do is to get the following result. The first row can be explained as follows: label e has the label b as a parent, and label b has the label a as a parent, based on the parent ids.
category label_1 label_2 label_3
A a b e
A a b f
A a b g
A a c h
A a c i
A a c j
A a d null
B k l o
B k l p
B k l q
B k m r
B k m s
B k m t
B k n null

Answer 1

Consider below approach
with recursive iterations as (
select category, child_id, label as labels
from example_data where parent_id = 0
union all
select e.category, e.child_id, concat(labels, '||', label)
from iterations i join example_data e
on e.category = i.category and e.parent_id = i.child_id
)
select * except(labels) from (
select * from (
select category, labels from iterations
qualify not ifnull(starts_with(lead(labels) over(partition by category order by labels), labels || '||'), false)
), unnest(split(labels, '||')) label with offset
)
pivot (any_value(label) as label for offset + 1 in (1, 2, 3))
order by category, labels
if applied to sample data in your question - output is

Answer 2

I believe there must be a better way than my initial try. Hope this is helpful until somebody find it.
WITH RECURSIVE tree AS (
SELECT category, 1 AS level, parent_id, child_id, label, [label] labels
FROM example_data WHERE parent_id = 0
UNION ALL
SELECT e.category, level + 1 AS level, e.parent_id, e.child_id, e.label, ARRAY_CONCAT(labels,[e.label])
FROM tree t JOIN example_data e ON e.category = t.category AND e.parent_id = t.child_id
),
filtered AS (
SELECT * EXCEPT(labels, max_level, i), i + 1 AS level FROM (
SELECT * REPLACE(IF(ARRAY_LENGTH(labels) <> max_level, ARRAY_CONCAT(labels, ['null']), labels) AS labels) FROM (
SELECT category, labels, MAX(ARRAY_LENGTH(labels)) OVER () max_level,
FROM tree
WHERE child_id NOT IN (SELECT parent_id FROM example_data)
)
), UNNEST(labels) label WITH OFFSET i
),
pivotted AS (
SELECT *
FROM filtered
PIVOT (ARRAY_AGG(label IGNORE NULLS) AS level FOR level IN (1, 2, 3))
)
-- Unnest a pivotted result to make final output.
SELECT category, lv1 AS level_1, lv2 AS level_2, lv3 AS level_3
FROM pivotted, UNNEST(level_1) lv1 WITH OFFSET
JOIN UNNEST(level_2) lv2 WITH OFFSET USING(offset)
JOIN UNNEST(level_3) lv3 WITH OFFSET USING(offset)
ORDER BY 1, 2, 3, 4
;