How multiple join to the same table with other table? - sql

Question

I am not sure of the question title!, but i have this problem
table1
id | from | to
1 A B
2 C A
3 B A
table2
id | table1_id
1 1
2 1
3 1
4 3
5 3
6 2
7 2
8 2
I need to get data from table1 considering rows with ids (2,3) as one row and joined with table2
fetch the last id between them which is 5
the result
id | from | to | table2_id
3 B | A | 5
2 C | A | 8

Answer 1

As you have not mentioned database, I have written SQL in SQL Server
WITH TAB_FLAT AS
(
SELECT TAB_1.table1_id , COALESCE ( TAB_2.table2_id , TAB_1.table2_id ) max_id , TAB_1.table2_id FROM
( select table1_id , max(id) table2_id from table2 group by table1_id ) TAB_1
LEFT JOIN
( select table1_id , max(id) table2_id from table2 group by table1_id ) TAB_2
ON TAB_1.table2_id = TAB_2.table1_id
)
select TAB_FLAT.table1_id AS ID, COALESCE(C.frm ,COALESCE(B.frm,A.frm) ) AS 'FROM' ,
COALESCE(C.to2 ,COALESCE(B.to2,A.to2) ) AS 'TO' ,
TAB_FLAT.max_id AS table2_id
FROM TAB_FLAT
LEFT JOIN table1 A ON A.id = TAB_FLAT.table1_id
LEFT JOIN table1 B ON B.id = TAB_FLAT.table2_id
LEFT JOIN table1 C ON C.id = TAB_FLAT.max_id
WHERE TAB_FLAT.table1_id IN (1,2)
Demo --> https://rextester.com/DCHUN74655
Explanation:
SELECT TAB_1.table1_id , COALESCE ( TAB_2.table2_id , TAB_1.table2_id ) max_id , TAB_1.table2_id FROM
( select table1_id , max(id) table2_id from table2 group by table1_id ) TAB_1
LEFT JOIN
( select table1_id , max(id) table2_id from table2 group by table1_id ) TAB_2
ON TAB_1.table2_id = TAB_2.table1_id
We are doing a self left join to get the largest table 2 ID for each table 1 ID
AND corresponding intermediate level.
select TAB_FLAT.table1_id AS ID, COALESCE(C.frm ,COALESCE(B.frm,A.frm) ) AS 'FROM' ,
COALESCE(C.to2 ,COALESCE(B.to2,A.to2) ) AS 'TO' ,
TAB_FLAT.max_id AS table2_id
FROM TAB_FLAT
LEFT JOIN table1 A ON A.id = TAB_FLAT.table1_id
LEFT JOIN table1 B ON B.id = TAB_FLAT.table2_id
LEFT JOIN table1 C ON C.id = TAB_FLAT.max_id
We are joining to table 1 based on table1 id , intermediate id and largest id
WHERE TAB_FLAT.table1_id IN (1,2)
Filtering out records with ID = 1 or 2

Answer 2

First, write a query to get each row's max ID in table 2.
select t1.*, max(t2.id) as table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
Then use that as a CTE for two queries. One to get all the rows except 1 and 3. And one to get 1 or 3, whichever has a greater table2_id. union them together.
with max_table2 as (
select t1.*, max(t2.id) as table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
)
select *
from max_table2
where id not in (1,3)
union
(
select *
from max_table2
where id in (1,3)
order by table2_id desc
limit 1
)
If the combined 1/3 row must have the ID of 1, despite 3 having the larger table2_id, this can be done by hard coding the ID in the select query.
with max_table2 as (
select t1.*, max(t2.id) as table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
)
select *
from max_table2
where id not in (1,3)
union
(
select 1 as id, "from", "to", table2_id
from max_table2
where id in (1,3)
order by table2_id desc
limit 1
)
Try it.
I suspect rather than hard coding rows 1 and 3, you're actually counting them as equivalent because they have the same path, just reversed. We can make this query more generic.
First, normalize the from/to so they're in the same order. While we're at it, also get their max table2 id.
select
t1.id,
case when "from" < "to" then "from" else "to" end as "from",
case when "from" < "to" then "to" else "from" end as "to",
max(t2.id) as max_table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
id from to max_table2_id
2 A C 8
3 A B 5
1 A B 3
Then rank the paths with the same from/to.
with normalized_max_table2 as (
select
t1.id,
case when "from" < "to" then "from" else "to" end as "from",
case when "from" < "to" then "to" else "from" end as "to",
max(t2.id) as table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
)
select *,
rank() over (partition by "from", "to" order by table2_id desc) as "rank"
from normalized_max_table2
id from to table2_id rank
3 A B 5 1
1 A B 3 2
2 A C 8 1
And, finally, select only the first ranks.
with normalized_max_table2 as (
select
t1.id,
case when "from" < "to" then "from" else "to" end as "from",
case when "from" < "to" then "to" else "from" end as "to",
max(t2.id) as table2_id
from table2 t2
join table1 t1 on t2.table1_id = t1.id
group by t1.id
),
ranked_max_table2 as (
select *,
rank() over (partition by "from", "to" order by table2_id desc) as "rank"
from normalized_max_table2
)
select id, "from", "to", table2_id
from ranked_max_table2
where "rank" = 1
id from to table2_id
3 A B 5
2 A C 8
That's a long-hand way of doing it. There may be a more compact way.
Try it.