How can I pass pure sql into a macro and then do the iteration in a dbt jinja2 macro - jinja2

Question

I am building a data pipeline that ingests data from one single table with different groupby objects. The following pure sql example demonstrates how it looks like:
select
a1,b1,
r,s,t
from table
group by a1, b1
union all
select
a2,b2,
r,s,t
from table
group by a2, b2
union all
select
a3,b3,
r,s,t
from table
group by a3, b3
union all
...
The problem I am trying to solve is that this sql has to repeat 40 times of union all with only groupby object changed for each time. I want to make the SQL configurable and dynamic to load its group by objects, so I only need to code the SQL query once along with a list of 40 elements of the changed groupby objects.
Theoretically, using macro I can do
(1) iteratively create a single table 40 times with different groupby objects.
(2) combine 40 tables together.
For (1) one approach I can think of is to embed the sql into a macro something like this pseudo code
{% macro create_table(groupby = [[a1, b1], [a2, b2], [a3, b3]]) %}
{% for a, b in groupby %}
table =
select
a,b,
r,s,t
from table_a
group by a1, b1
{% end_macro %}
However, one constraint is that I don't want to code the sql within a macro file. It would be better to place the sql logic in a sql file for some reason. Then how to code this out and pass the sql into a macro? I saw a sql can call macro but never saw the reverse way.
Also, for (2), I can't find any straightforward approach .
I would appreciate it if anyone could help to have a real macro code implementation since I don't have too much advance experience with dbt and jinja macro and am having a hard time researching on it
Edit: the column names are not strictly a1, a2, ... a40. I used a1 for simplicity only. You can imagine the columns are real English words like State, Country, Year, and things like these. And it doesn't necessarily have to iterative the whole list of columns. That's why macro is necessarily required.

Answer 1

I'd not use a macro for this. As you mentioned, you can define the SQL on a model and then just iterate through the different field names with a range.
One possible approach would be the following:
{% set max_range = 40 %}
{% for range_ in range(max_range) %}
select
a{{ range_ + 1 }},
b{{ range_ + 1 }},
r,
s,
t
from {{ source('your_schema', 'your_table') }} -- This could also be a `ref` if selecting from an existing model
group by 1, 2
{% if not loop.last -%} union all {% endif -%}
{% endfor %}
Update: I based my answer on the fact that fields would follow an increasing integer value on the field name. That's not the case according to the OP, so I propose another solution using an array of arrays containing pairs of field names:
{% set field_pairs = [['country', 'city'], ['first_name', 'last_name'], ['address', 'postal_code']] %}
{% for row in field_pairs %}
select
{% for value in row -%}
{{ value }},
{% endfor %}
r,
s,
t
from {{ source('your_schema', 'your_table') }} -- This could also be a `ref` if selecting from an existing model
group by 1, 2
{% if not loop.last -%} union all {% endif -%}
{% endfor %}