Identifying/comparing sets of rows within groups - sql

Question

I have a matter which seemed simple to solve but now I find it troublesome.
In simplification - I need to find a way to identify unique sets of rows within groups defined by another column. In basic example the source table contains only two columns:
routeID nodeID nodeName
1 1 a
1 2 b
2 1 a
2 2 b
3 1 a
3 2 b
4 1 a
4 2 c
5 1 a
5 2 c
6 1 a
6 2 b
6 3 d
7 1 a
7 2 b
7 3 d
So, the routeID column refers to set of nodes which define a route.
What I need to do is to somehow group the routes, so that there will be only one unique sequence of nodes for one routeID.
In my actual case I tried to use window function to add columns which help to identify nodes sequence, but I still have no idea how to get those unique sequences and group routes.
As a final effect I want to get only unique routes - for example routes 1,2 and 3 aggregated to one route.
Do you have any idea how to help me ?
EDIT:
The other table which I would like to join with the one from the example may look like that:
journeyID nodeID nodeName routeID
1 1 a 1
1 2 b 1
2 1 a 1
2 2 b 1
3 1 a 4
3 2 c 4
...........................
...........................

Answer 1

You can try this idea:
DECLARE #DataSource TABLE
(
[routeID] TINYINT
,[nodeID] TINYINT
,[nodeName] CHAR(1)
);
INSERT INTO #DataSource ([routeID], [nodeID], [nodeName])
VALUES ('1', '1', 'a')
,('1', '2', 'b')
,('2', '1', 'a')
,('2', '2', 'b')
,('3', '1', 'a')
,('3', '2', 'b')
,('4', '1', 'a')
,('4', '2', 'c')
,('5', '1', 'a')
,('5', '2', 'c')
,('6', '1', 'a')
,('6', '2', 'b')
,('6', '3', 'd')
,('7', '1', 'a')
,('7', '2', 'b')
,('7', '3', 'd');
SELECT DS.[routeID]
,nodes.[value]
,ROW_NUMBER() OVER (PARTITION BY nodes.[value] ORDER BY [routeID]) AS [rowID]
FROM
(
-- getting unique route ids
SELECT DISTINCT [routeID]
FROM #DataSource DS
) DS ([routeID])
CROSS APPLY
(
-- for each route id creating CSV list with its node ids
SELECT STUFF
(
(
SELECT ',' + [nodeName]
FROM #DataSource DSI
WHERE DSI.[routeID] = DS.[routeID]
ORDER BY [nodeID]
FOR XML PATH(''), TYPE
).value('.', 'VARCHAR(MAX)')
,1
,1
,''
)
) nodes ([value]);
The code will give you this output:
So, you simple need to filter by rowID = 1. Of course, you can change the code as you like in order to satisfy your bussness criteria (for example showing no the first route ID with same nodes, but the last).
Also, ROW_NUMBER function cannot be used directly in the WHERE clause, so you need to wrap the code before filtering:
WITH DataSource AS
(
SELECT DS.[routeID]
,nodes.[value]
,ROW_NUMBER() OVER (PARTITION BY nodes.[value] ORDER BY [routeID]) AS [rowID]
FROM
(
-- getting unique route ids
SELECT DISTINCT [routeID]
FROM #DataSource DS
) DS ([routeID])
CROSS APPLY
(
-- for each route id creating CSV list with its node ids
SELECT STUFF
(
(
SELECT ',' + [nodeName]
FROM #DataSource DSI
WHERE DSI.[routeID] = DS.[routeID]
ORDER BY [nodeID]
FOR XML PATH(''), TYPE
).value('.', 'VARCHAR(MAX)')
,1
,1
,''
)
) nodes ([value])
)
SELECT DS2.*
FROM DataSource DS1
INNER JOIN #DataSource DS2
ON DS1.[routeID] = DS2.[routeID]
WHERE DS1.[rowID] = 1;

Answer 2

ok, let's use some recursion to create a complete node list for each routeID
First of all let's populate source table and journeyes tale
-- your source
declare #r as table (routeID int, nodeID int, nodeName char(1))
-- your other table
declare #j as table (journeyID int, nodeID int, nodeName char(1), routeID int)
-- temp results table
declare #routes as table (routeID int primary key, nodeNames varchar(1000))
;with
s as (
select *
from (
values
(1, 1, 'a'),
(1, 2, 'b'),
(2, 1, 'a'),
(2, 2, 'b'),
(3, 1, 'a'),
(3, 2, 'b'),
(4, 1, 'a'),
(4, 2, 'c'),
(5, 1, 'a'),
(5, 2, 'c'),
(6, 1, 'a'),
(6, 2, 'b'),
(6, 3, 'd'),
(7, 1, 'a'),
(7, 2, 'b'),
(7, 3, 'd')
) s (routeID, nodeID, nodeName)
)
insert into #r
select *
from s
;with
s as (
select *
from (
values
(1, 1, 'a', 1),
(1, 2, 'b', 1),
(2, 1, 'a', 1),
(2, 2, 'b', 1),
(3, 1, 'a', 4),
(3, 2, 'c', 4)
) s (journeyID, routeID, nodeID, nodeName)
)
insert into #j
select *
from s
now let's exctract routes:
;with
d as (
select *, row_number() over (partition by r.routeID order by r.nodeID desc) n2
from #r r
),
r as (
select d.*, cast(nodeName as varchar(1000)) Names, cast(0 as bigint) i2
from d
where nodeId=1
union all
select d.*, cast(r.names + ',' + d.nodeName as varchar(1000)), r.n2
from d
join r on r.routeID = d.routeID and r.nodeId=d.nodeId-1
)
insert into #routes
select routeID, Names
from r
where n2=1
table #routes will be like this:
routeID nodeNames
1 'a,b'
2 'a,b'
3 'a,b'
4 'a,c'
5 'a,c'
6 'a,b,d'
7 'a,b,d'
an now the final output:
-- the unique routes
select MIN(r.routeID) routeID, nodeNames
from #routes r
group by nodeNames
-- the unique journyes
select MIN(journeyID) journeyID, r.nodeNames
from #j j
inner join #routes r on j.routeID = r.routeID
group by nodeNames
output:
routeID nodeNames
1 'a,b'
4 'a,c'
6 'a,b,d'
and
journeyID nodeNames
1 'a,b'
3 'a,c'