Generate Random Test Data with ORDER BY NEWID() , include duplicate rows - sql

Question

I need to select random rows from a table for test data. There may be times I need more rows of test data than there are records in the table. Duplicates are okay. How do I structure my select so that I can get duplicate rows?
CREATE TABLE [Northwind].[dbo].[Persons]
(PersonID int, LastName varchar(255))
INSERT INTO [Northwind].[dbo].[Persons]
VALUES
(1, 'Smith'),
(2, 'Jones'),
(3, 'Washington')
SELECT TOP 5 *
FROM [Northwind].[dbo].[Persons]
ORDER BY NEWID()
How do I get the Select statement to give me five records in random order, with repeats? Currently, it only returns three in random order.
I'd like to be able to extend this to get 100 rows or 1000 rows or however many I need.

Answer 1

Use a recursive CTE to union enough rows so that they are larger than what you desire. Then select from that as you have done before.
declare
#desired int = 5,
#actual int = (select count(*) from persons);
with
persons as (
select personId,
lastName,
batch = 0
from Persons
union all
select personId,
lastName,
batch = batch + 1
from persons
where (batch + 1) * #actual < #desired
)
select
top (#desired) personId, lastName
from persons
order by newid()

Answer 2

As mentioned. You could instead us a tally table and then get the random rows;
WITH N AS(
SELECT N
FROM (VALUES(NULL),(NULL),(NULL),(NULL),(NULL),(NULL),(NULL),(NULL),(NULL),(NULL)))N(N)),
Tally AS(
SELECT ROW_NUMBER() OVER (ORDER BY (SELECT NULL)) AS I
FROM N N1, N N2, N N3, N N4) --Repeat for more
SELECT TOP 500 YT.*
FROM Tally T
CROSS JOIN YourTable YT
ORDER BY NEWID();

Answer 3

I was thinking of how you would solve this without ordering all the records, especially multiple times.
One method is to generate random numbers and to use those for looking up values in your data:
with n as (
select rand(checksum(newid())) as r, 1 as n
union all
select rand(checksum(newid())) as r, n + 1
from n
where n < 10
),
tt as (
select t.*, lag(tile_end, 1, 0) over (order by tile_end) as tile_start
from (select t.*, row_number() over (order by newid()) * 1.0 / count(*) over () as tile_end
from t
) t
)
select tt.*, n.r, (select count(*) from n)
from n left join
tt
on n.r >= tt.tile_start and n.r < tt.tile_end;
Here is a db<>fiddle. The row_number() does not need to use order by newid(). It can order by a key that has an index -- which makes that component much more efficient.
For more than 100 rows, you will need OPTION (MAXRECURSION 0).

Answer 4

I added a temp results table and looped through the query and pushed the results into the temp table.
declare #results table(
SSN varchar(10),
Cusip varchar(10),
...
EndBillingDate varchar(10))
DECLARE #cnt INT = 0;
WHILE #cnt < #trades
BEGIN
INSERT INTO #results
Select ...
set #cnt = #cnt + 10
END
select * from #results