Flat to multi-level nested data - sql

Question

In BQ, I have used ARRAY_AGG(STRUCT(... to restructure some flat data but wanted to go a level further: create another array of records within an array of records.
Although STRUCT does not exist in PostgreSQL, I am interested how one would tackle that there too.
Considering the flat data:
WITH a AS (
SELECT 'ABC' company, 'adress1' address, 'name1' name, 'email1' email, 'work' ph_type, '+123' ph_nr
UNION ALL
SELECT 'ABC' company, 'adress1' address, 'name1' name, 'email1' email, 'cell' ph_type, '+987'
UNION ALL
SELECT 'DEF' company, 'adress2' address, 'name2' name, 'email2' email, 'work' ph_type, '+127'
UNION ALL
SELECT 'DEF' company, 'adress2' address, 'name2' name, 'email2' email, 'cell' ph_type, '+988'
UNION ALL
SELECT 'XYZ' company, 'adress3' address, 'name3' name, 'email3' email, 'work' ph_type, '+456'
)
I can nest contact like so
SELECT company, address, ARRAY_AGG(STRUCT(name, email, ph_type, ph_nr)) contact
FROM a
GROUP BY company, address
ORDER BY 1
but how can I nest, in the same select statement, phones as well (array of records within contact) ?
The JSON representation would look like - for the first contact:
[
{
"company": "ABC",
"address": "adress1",
"contact": [
{
"name": "name1",
"email": "email1",
"phone": [
{
"ph_type": "work",
"ph_nr": "+123"
},
{
"ph_type": "cell",
"ph_nr": "+987"
}
},
...
This can probably be done with a WITH clause or subselect to process the aggregations sequentially but not sure this would perform well (data read twice ?).
I have 600M records to parse daily so wondering about the most efficient way.
EDIT: corrected name definition

Answer 1

The answer to your question is two levels of aggregation.
However, the question itself confuses me, because the query uses name but that is not defined in the data.
Here is an example of what to do:
SELECT company, address, ARRAY_AGG(STRUCT(email, phones)) as contact
FROM (SELECT company, name, address, email, ARRAY_AGG(STRUCT(ph_type, ph_nr)) as phones
FROM a
GROUP BY company, name, address, email
) a
GROUP BY company, address
ORDER BY 1