Select Data with Order By Respecting Ancestor - sql

Question

I have a table MyStackFiles that has 3 columns:
FileID (The primary key)
FileName
OriginalFileID (This can be either 0 if there is no original file or one of the other file IDs)
My goal is to select the whole data sorted by name. In addition, I need to always have the original files appear before their children. In other words, the desired result will start with the first alphabetical file whose OriginalFileID is 0 followed by all its children (if available) alphabetically. The following SQL script creates the sample data and illustrates exactly what I'm trying to achieve. Notice that the last select command is the desired output.
What is the query that can return the desired result?
The Script:
-------------------------- Creating the Example Schema --------------------------
IF EXISTS (SELECT * FROM INFORMATION_SCHEMA.Tables WHERE Table_Name = 'MyStackFiles')
Drop table MyStackFiles
GO
CREATE TABLE [dbo].[MyStackFiles](
[FileID] [int] IDENTITY(1,1) NOT NULL,
[FileName] [varchar](50) NULL,
[OriginalFileID] [int] NOT NULL DEFAULT (0),
CONSTRAINT [PK_MyStackFiles] PRIMARY KEY CLUSTERED
(
[FileID] 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
-------------------------- Insert Into the Sample Data --------------------------
INSERT INTO MyStackFiles(FileName) values ('S')
INSERT INTO MyStackFiles(FileName) values ('G')
INSERT INTO MyStackFiles(FileName, OriginalFileID) values ('E', 1)
INSERT INTO MyStackFiles(FileName) values ('F')
INSERT INTO MyStackFiles(FileName, OriginalFileID) values ('Q', 2)
INSERT INTO MyStackFiles(FileName, OriginalFileID) values ('N', 3)
INSERT INTO MyStackFiles(FileName) values ('A')
INSERT INTO MyStackFiles(FileName, OriginalFileID) values ('X', 1)
INSERT INTO MyStackFiles(FileName) values ('W')
------------------------------------------------------------------------------
GO
-------------------------- Simple select sorted by FileName --------------------------
SELECT * From MyStackFiles ORDER BY FileName
-------------------------- A representation of the desired result --------------------------
SELECT * FROM MyStackFiles WHERE FileID = 7 UNION ALL -- We insert "A" (respecting the alphabetical order) since its OriginalFileID is 0
SELECT * FROM MyStackFiles WHERE FileID = 4 UNION ALL -- Then we insert F.
SELECT * FROM MyStackFiles WHERE FileID = 2 UNION ALL -- Then we insert G. G has children so we insert them
SELECT * FROM MyStackFiles WHERE FileID = 5 UNION ALL -- Q is the only child of G. We insert it
SELECT * FROM MyStackFiles WHERE FileID = 1 UNION ALL -- Now we insert S. Notice that S has two children (E and X)
SELECT * FROM MyStackFiles WHERE FileID = 3 UNION ALL -- E is before X alphabetically so it gets inserted first
SELECT * FROM MyStackFiles WHERE FileID = 6 UNION ALL -- E happens to have children so we insert them right away (in a depth first fashion)
SELECT * FROM MyStackFiles WHERE FileID = 8 UNION ALL -- Now we insert the other child of S which is X
SELECT * FROM MyStackFiles WHERE FileID = 9 -- Finally we insert W the only file left
--Drop Table MyStackFiles
I'm open to any schema modification if that helps find an efficient query.

Answer 1

I'm using the technique called Recursive CTE to try to solve your problem:
with t (RowID, FileID, FileName, OriginalFileID)
as (
select convert(varchar(max), row_number() over (order by s.FileName)), s.*
from MyStackFiles s
where s.OriginalFileID = 0
union all
select t.RowID + '.' + convert(varchar(max), row_number() over (order by s.FileName)), s.*
from MyStackFiles s
inner join t on t.FileID = s.OriginalFileID
)
select FileID, FileName, OriginalFileID from t
order by RowID
A temporary column RowID is created on-the-fly to chain up the ancestor's RowID to the row's row_number, so that for instance the file "N" will have RowID = '4.1.1', the file "X" will have RowID = '4.2', and this is the column to sort that fits your sorting requirement.