SQL server query, sort on multiple columns - sql

Question

We have a nested structure of tasks in which every task can contain other tasks. Order of tasks in a task is important and is defined by the Sequence field starting at zero.
Here is my table structure:
USE [MyDB]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE TABLE [dbo].[Relation](
[PK_ID] [int] IDENTITY(1,1) NOT NULL,
[SourceEntityId] [uniqueidentifier] NOT NULL,
[TargetEntityId] [uniqueidentifier] NOT NULL,
CONSTRAINT [PK_Relation] PRIMARY KEY CLUSTERED
(
[PK_ID] ASC
)WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, IGNORE_DUP_KEY = OFF, ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON) ON [PRIMARY]
) ON [PRIMARY]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE TABLE [dbo].[TaskTable1](
[Id] [uniqueidentifier] NOT NULL,
[Title] [nvarchar](max) NULL,
[SequenceId] [int] NULL
) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]
GO
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
CREATE TABLE [dbo].[TaskTable2](
[Id] [uniqueidentifier] NOT NULL,
[Title] [nvarchar](max) NULL,
[SequenceId] [int] NULL
) ON [PRIMARY] TEXTIMAGE_ON [PRIMARY]
GO
SET IDENTITY_INSERT [dbo].[Relation] ON
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (1, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'5b266fd1-cbc8-c16a-91c4-5675a35c9ecf')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (2, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'e499ca68-8103-b8ec-06ba-110fa3f6eb5b')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (4, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'645ad2eb-df10-0d5b-0526-408aad45a145')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (5, N'785227d1-393c-ae18-02e5-03ab08d577af', N'5655aeb7-b8b5-dca9-38af-37687c668c14')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (6, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'030cdefc-0e45-01e6-e2a5-a69e303bda4b')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (7, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'0375c7a1-8cc5-a4c8-151c-966e4af83f73')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (8, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'785227d1-393c-ae18-02e5-03ab08d577af')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (9, N'030cdefc-0e45-01e6-e2a5-a69e303bda4b', N'8324bba9-252f-bef8-c018-8b86491e2361')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (10, N'030cdefc-0e45-01e6-e2a5-a69e303bda4b', N'f1cbe8a3-3285-4cf0-096d-aad0327bdb0b')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (11, N'dab00c89-961c-84dd-bb43-cffd18e63594', N'0189f0af-5045-a498-2d70-99187bf3f0ae')
GO
INSERT [dbo].[Relation] ([PK_ID], [SourceEntityId], [TargetEntityId]) VALUES (12, N'785227d1-393c-ae18-02e5-03ab08d577af', N'ffecd091-c17b-ee5f-a64d-54ea9ff65aa9')
GO
SET IDENTITY_INSERT [dbo].[Relation] OFF
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'5b266fd1-cbc8-c16a-91c4-5675a35c9ecf', N'First', 0)
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'e499ca68-8103-b8ec-06ba-110fa3f6eb5b', N'Second', 1)
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'0189f0af-5045-a498-2d70-99187bf3f0ae', N'Fourth', 3)
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'0375c7a1-8cc5-a4c8-151c-966e4af83f73', N'Sixth', 5)
GO
INSERT [dbo].[TaskTable2] ([Id], [Title], [SequenceId]) VALUES (N'030cdefc-0e45-01e6-e2a5-a69e303bda4b', N'Fifth', 4)
GO
INSERT [dbo].[TaskTable2] ([Id], [Title], [SequenceId]) VALUES (N'785227d1-393c-ae18-02e5-03ab08d577af', N'Seventh', 6)
GO
INSERT [dbo].[TaskTable2] ([Id], [Title], [SequenceId]) VALUES (N'645ad2eb-df10-0d5b-0526-408aad45a145', N'Third', 2)
GO
INSERT [dbo].[TaskTable2] ([Id], [Title], [SequenceId]) VALUES (N'8324bba9-252f-bef8-c018-8b86491e2361', N'sub1', 0)
GO
INSERT [dbo].[TaskTable2] ([Id], [Title], [SequenceId]) VALUES (N'f1cbe8a3-3285-4cf0-096d-aad0327bdb0b', N'sub2', 1)
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'ffecd091-c17b-ee5f-a64d-54ea9ff65aa9', N'sub 1', 0)
GO
INSERT [dbo].[TaskTable1] ([Id], [Title], [SequenceId]) VALUES (N'5655aeb7-b8b5-dca9-38af-37687c668c14', N'sub 2', 1)
GO
To get the tasks in order with their child tasks right beneath their parents, I tried the following query to no avail:
;With TaskCTE
AS
(
select R.SourceEntityId AS ParentTask_Id, R.TargetEntityId AS Task_Id , cast(null as uniqueidentifier) AS ParentTask, 0 AS Level
, ROW_NUMBER() OVER (ORDER BY (SELECT 100)) / power(10.0,0) as x
from Relation R
where (R.SourceEntityId = 'DAB00C89-961C-84DD-BB43-CFFD18E63594')
UNION ALL
select R1.SourceEntityId , R1.TargetEntityId, TaskCTE.Task_Id , Level + 1
, x + ROW_NUMBER() OVER (ORDER BY (SELECT 100)) / power(10.0,level+1)
from Relation R1
INNER JOIN TaskCTE
ON R1.SourceEntityId = TaskCTE.Task_Id
)
select ParentTask_Id, Task_Id, ParentTask, Level
, COALESCE(TT1.Title, TT2.Title) AS Title
, COALESCE(TT1.SequenceId, TT2.SequenceId) AS SequenceId
, x
from TaskCTE
LEFT OUTER JOIN TaskTable1 TT1
ON TaskCTE.Task_Id = TT1.Id
LEFT OUTER JOIN TaskTable2 TT2
ON TaskCTE.Task_Id = TT2.Id
order by level , SequenceId
If you follow the structure of required output (shown in below image), the sequence ** column along with the **Level column must determine the sort order.
Thanks in advance
Edit: My query output which is wrong:

Answer 1

If your problem is that sequence field in other table rather than relation table, then why do not you join them before running recursion? But it likely will be slower than your initial query. Here's a sample
with cte as (
select
r.SourceEntityId, r.TargetEntityId, t.SequenceId, 0 k
from
Relation r
join (
select * from TaskTable1
union all
select * from TaskTable2
) t on r.TargetEntityId = t.id
---------------------------------------
union all select * from cte where k = 1
---------------------------------------
)
, rcte as (
select
SourceEntityId, TargetEntityId, ParentTask = cast(null as uniqueidentifier)
, SequenceId, rn = cast(row_number() over (order by SequenceId) as varchar(8000)), 1 step
from
cte
where
SourceEntityId = 'DAB00C89-961C-84DD-BB43-CFFD18E63594'
union all
select
a.TargetEntityId, b.TargetEntityId, a.SourceEntityId, b.SequenceId
, cast(concat(a.rn, '.', row_number() over (partition by b.SourceEntityId order by b.SequenceId)) as varchar(8000))
, step + 1
from
rcte a
join cte b on a.TargetEntityId = b.SourceEntityId
)
select
*
from
rcte
order by rn
I have not included your X column, I can not get what are trying to calculate. Also, in your expected output values of ParentTask and ParentTask_Id are same. Should be so?

Answer 2

I am using same query as #Uzi with minor correction.I am having same doubts as him.#Yasser should clearly show what output is desire in proper output and remove unnecessary columns.
if row_number only purpose is to order record then why convert it to varchar(8000).Also you can avoid expensive Row_number all together.
Take advantage of PK_ID instead of expensive row_number,even if PK_ID is not in sequence in this case.
If performance is big issue then user should mention number of rows in 3 TABLE AND WHAT OTHER FILTER be applied IN WHERE CONDITION ?
Why data type is uniqueidentifier ?Will it solve the purpose if it is INT?
Read this
Check this query,
WITH cte
AS (
SELECT r.PK_ID
,r.SourceEntityId
,r.TargetEntityId
,t.SequenceId,0 k
FROM #Relation r
JOIN (
SELECT id
,SequenceId
FROM #TaskTable1
UNION ALL
SELECT id
,SequenceId
FROM #TaskTable2
) t ON r.TargetEntityId = t.id
---------------------------------------
--union all select * from cte where k = 1
---------------------------------------
)
,rcte
AS (
SELECT SourceEntityId
,TargetEntityId
,ParentTask = cast(NULL AS UNIQUEIDENTIFIER)
,SequenceId
, rn = cast(row_number() over (order by SequenceId) as decimal(3,1))
--, rn = cast( SequenceId+1 as decimal(3,1))--**
,1 step
FROM cte
WHERE SourceEntityId = 'DAB00C89-961C-84DD-BB43-CFFD18E63594'
UNION ALL
SELECT a.TargetEntityId
,b.TargetEntityId
,a.SourceEntityId
,b.SequenceId
,cast((a.rn+(b.SequenceId/10.0)) as decimal(3,1))
,step + 1
FROM rcte a
JOIN cte b ON a.TargetEntityId = b.SourceEntityId
)
SELECT *
FROM rcte
ORDER BY rn
--**
--SELECT *
--FROM rcte
--ORDER BY rn,st
-- 2nd Edit,
I understand that there is no way of changing database.
In that case it is very logical to create index view where task table id is Clustered index.
select id, SequenceId from #TaskTable1
union all
select id, SequenceId from #TaskTable2
Create nonclustered index NCI_Relation_SourceID on Relation([SourceEntityId])
Create nonclustered index NCI_Relation_TargetEntityId on Relation([TargetEntityId])
you can once try this combination,
Remove PK_ID as clustered index and make TargetEntityId as clustered index.
you can once try creating view on this query,
SELECT r.PK_ID
,r.SourceEntityId
,r.TargetEntityId
,t.SequenceId
FROM #Relation r
JOIN (
SELECT id
,SequenceId
FROM #TaskTable1
UNION ALL
SELECT id
,SequenceId
FROM #TaskTable2
) t ON r.TargetEntityId = t.id

Answer 3

by adding a new column named Hierarchy in CTE expression and sorting outcome according to this value could solve your requirement
Here is the modified CTE query
;With TaskCTE AS
(
select
R.SourceEntityId AS ParentTask_Id,
R.TargetEntityId AS Task_Id , cast(null as uniqueidentifier) AS ParentTask, 0 AS Level
, ROW_NUMBER() OVER (ORDER BY (SELECT 100)) / power(10.0,0) as x
,CAST( ROW_NUMBER() OVER (ORDER BY R.SourceEntityId) as varchar(max)) Hierarchy
from Relation R
where (R.SourceEntityId = 'DAB00C89-961C-84DD-BB43-CFFD18E63594')
UNION ALL
select R1.SourceEntityId , R1.TargetEntityId, TaskCTE.Task_Id , Level + 1
, x + ROW_NUMBER() OVER (ORDER BY (SELECT 100)) / power(10.0,level+1)
,CAST(Hierarchy + ':' + CAST(ROW_NUMBER() OVER (ORDER BY R1.SourceEntityId) as varchar(max)) as varchar(max)) as Hierarchy
from Relation R1
INNER JOIN TaskCTE
ON R1.SourceEntityId = TaskCTE.Task_Id
)
select ParentTask_Id, Task_Id, ParentTask, Level
, COALESCE(TT1.Title, TT2.Title) AS Title
, COALESCE(TT1.SequenceId, TT2.SequenceId) AS SequenceId
, x
,Hierarchy
from TaskCTE
LEFT OUTER JOIN TaskTable1 TT1
ON TaskCTE.Task_Id = TT1.Id
LEFT OUTER JOIN TaskTable2 TT2
ON TaskCTE.Task_Id = TT2.Id
order by Hierarchy
Please note that I have added Hierarchy column and its value is calculated using a ROW_NUMBER() function which creates a unique integer value for each task
You can find implemantation of this hierarchy query with SQL CTE at refereced tutorial
I hope it helps
I am also adding the output as screenshot here to show how data is sorted according to Hierarchy
Although childs are listed after parents, it does not one-to-one match with your desired outcome as I could see