Alternative query for calculating percentage - sql-server

In SQL Server for calculating percentage I have a function like below:
CREATE FUNCTION [dbo].[fuGetPercentage] ( #part FLOAT, #total FLOAT )
RETURNS FLOAT
AS
BEGIN
DECLARE #Result FLOAT = 0, #Cent FLOAT = 100;
IF (isnull(#total, 0) != 0)
SET #Result = isnull(#part, 0) * #Cent / #total;
RETURN #Result
END
I wonder that is there any better alternative for that, with same checks and a better calculating percentage like below:
SELECT (CASE ISNULL(total, 0)
WHEN 0 THEN 0
ELSE ISNULL(part, 0) * 100 / total
END) as percentage
I want to use it directly after SELECT like above.

There is one issue with using functions such as ISNULL. The query will not use indexes in that case. If the beauty of the code isn't in the first place then you can do something like that:
SELECT
CASE WHEN total * part <> 0 /* will check that both total and part are not null and != 0*/
THEN part * 100 / total
ELSE 0
END AS percentage;

I use #DmitrijKultasev answer but now, I found that it has two problems:
Error on conversion because of the overflow of result of multiply.
Performance problem; because of that math multiply and its conversion.
So I change it to this:
SELECT
CASE
WHEN total <> 0 AND part <> 0 THEN -- This will return 0 for Null values
part * 100 / total
ELSE
0
END AS percentage;

Related

Rounding the value to the nearest 50

I am trying to round down the value to the nearest 50.
1-50 it should round down to below 00 and when its 51-rest then it should round down to 50
ex:
245 (until 1-49) its should round down to 200
258 (from 50-99)then it should round down to 250
I tried this,its wrking good but I need smething other than case statement
#ResultAmount = ROUND(#ResultAmount, -2, 1) +
CASE WHEN RIGHT(CONVERT(INT, FLOOR(#ResultAmount)), 2) IN (00, 50)
THEN RIGHT(CONVERT(INT, FLOOR(#ResultAmount)), 2)
WHEN RIGHT(CONVERT(INT, FLOOR(#ResultAmount)), 2) BETWEEN 1 AND 49
THEN 00
WHEN RIGHT(CONVERT(INT, FLOOR(#ResultAmount)), 2) BETWEEN 51 AND 99
THEN 50
END
Thanks in advance!!!
This is all you need
SELECT FLOOR(#ResultAmount / 50) * 50;
e.g below
declare #ResultAmount decimal(10,2) = 249;
SELECT FLOOR(#ResultAmount / 50) * 50;
SET #ResultAmount = 250;
SELECT FLOOR(#ResultAmount / 50) * 50;
SET #ResultAmount = 200;
SELECT FLOOR(#ResultAmount / 50) * 50;
SET #ResultAmount = 199;
SELECT FLOOR(#ResultAmount / 50) * 50;
It sounds like numbers 0-50 get rounded up to "50", but any number larger than that should just get rounded to the nearest 50. Something like the following should work:
(CASE WHEN f1/50 < 1 THEN 1 ELSE ceiling(f1/50) END) * 50 AS rounded_to_50
You can simply divide the number by 50, round then multiply by 50 again, eg:
select cast(round(#i/50.0,0)*50 as int)
This will return 500 if #i is 524 but 550 if #i is 525.
You can create a function to make this easier:
create function fn_Round_By(#input int,#divider float)
returns int
as
begin
RETURN (cast(round(#input/#divider,0)*#divider as int));
end
Again, select dbo.fn_Round_By(525,50) returns 550 andselect dbo.fn_Round_By(524,50)` returns 500.
If you want values less than 50 to round up to 50, you can use a simple CASE, eg:
create function fn_Round_By(#input int,#divider float)
returns int
as
begin
RETURN (
CASE
WHEN #input <=#divider then #divider
else cast(round(#input/#divider,0)*#divider as int)
END
);
end
Rounding down is performed by the FLOOR function so a function that rounded down to a specific interval would be:
create function fn_Floor_By(#input int,#divider float)
returns int
as
begin
RETURN (cast(FLOOR(#input/#divider)*#divider as int));
end
or, preserving the logic that rounds up anything under 50:
create function fn_Floor_By(#input int,#divider float)
returns int
as
begin
RETURN (
CASE
WHEN #input <=#divider then #divider
else cast(FLOOR(#input/#divider)*#divider as int)
END
);
end
You can calculate modulo 50 and use this to reduce the original value
DECLARE #ResultAmount int = 243
SELECT #ResultAmount - (#ResultAmount%50)

How to protect sql statement from Divide By Zero error

I'm in the process of creating some reports that take a finite total (lets say 2,500) of products (for this example lets say Ice Cream Cones) and counts how many of them were broken before serving.
Now the actual count code of broken cones I've got down.
SELECT COUNT(broken_cones) FROM [ice].[ice_cream_inventory]
WHERE broken_cones = 'Yes'
However, I need a percentage of broken cones from this total as well. I've been playing around with the code but I keep running into a 'Divide By Zero' error with this code below.
SELECT CAST(NULLIF((.01 * 2500)/Count(broken_cones), 0) AS
decimal(7,4)) FROM [ice].[ice_cream_inventory] WHERE broken_cones = 'Yes'
For right now, there aren't any broken cones (and won't be for a while) so the total right now is zero. How can I show the NULL scenario as zero?
I tried to place an ISNULL statement in the mix but I kept getting the 'Divide by Zero' error. Am I even doing this right?
::edit::
Here's what I ended up with.
SELECT
CASE
WHEN COUNT(broken_cones) = 0 then 0
ELSE CAST(NULLIF((.01 * 2500)/Count(broken_cones), 0) AS decimal(7,4))
END
FROM [ice].[ice_cream_inventory] WHERE broken_cones = 'Yes'
Use a case statement.
SELECT
CASE WHEN COUNT(broken_cones) = 0 then 0
ELSE CAST(NULLIF((.01 * 2500)/Count(broken_cones), 0) AS decimal(7,4)) END
FROM [ice].[ice_cream_inventory] WHERE broken_cones = 'Yes'
You already have a solution, but this is why your original solution didn't work.
Your NULLIF needs to be moved in order to be effective. It is doing the division before it gets to the NULLIF call. Dividing by null will return a null value.
SELECT CAST((.01 * 2500)/NULLIF(Count(broken_cones), 0) AS decimal(7,4))
FROM [ice].[ice_cream_inventory]
WHERE broken_cones = 'Yes'`
The NULLIF() function is a great way to prevent divide by zero, since anything divide by NULL returns null. The way to use it is as follows:
<expression> / NULLIF( <expression>, 0 )
Unfortunately you've wrapped your whole divide expression in NULLIF() which is why it isn't working for you. So step one is to get it to return NULL if your COUNT() comes back zero:
SELECT
(0.01 * 2500) / NULLIF( COUNT(broken_cones), 0 )
FROM [ice].[ice_cream_inventory]
WHERE broken_cones = 'Yes'
Now you said you wanted that NULL to come back zero? That is where you use ISNULL():
ISNULL(<expression1>, <expression2>)
If the first expression is NULL then return the second expression, so our SQL now becomes:
SELECT
ISNULL(
(0.01 * 2500) / NULLIF( COUNT(broken_cones), 0 ),
0
)
FROM [ice].[ice_cream_inventory]
WHERE broken_cones = 'Yes'

TSQL - CAST AS PERCENT

In my select list, I want to return a column that shows the percent of appointments missed. When a client misses an appointment, the source column contains "0" for duration. So I would like to sum the instances of "0" and divide by total appointments scheduled.
CAST(SUM(CASE WHEN event_client_duration = 0 THEN 1 ELSE 0 END)/COUNT(event_key) AS FLOAT)
That does not throw an error, but it returns a value of 0 except in the rare case when appointments missed are more than 50% in which case it returns 1.
SUM(CASE WHEN event_client_duration = 0 THEN 1 ELSE 0 END) -- that works
COUNT(event_key) -- that also works, but together they bonk
So, I tried to cast it as a decimal, but this causes an arithmetic overflow.
CAST(SUM(CASE WHEN e_cl_dur = 0 THEN 1 ELSE 0 END)/COUNT(e_key) AS DECIMAL(2,2))
I also tried using percent as a data type. Any ideas?
SUM(CASE event_client_duration WHEN 0 THEN 1.0 ELSE 0.0 END) / COUNT(event_key)

Conversion failed when converting the nvarchar to int

I have a field which is varchar and contains numbers and dates as strings. I want to update all numbers in this field that is greater than 720. I have attempted firstly to do a select but I get this error:
Conversion failed when converting the nvarchar value '16:00' to data type int.
This is my query:
select id, case(isnumeric([other08])) when 1 then [other08] else 0 end
from CER where sourcecode like 'ANE%' --and other08 > 720
It fails when I uncomment the last part.
I am trying to get all numerics greater than 720, but I can't do the comaprison. It also fails when casting and converting.
Thanks all for any help
You also need to perform the checks and conversion in the WHERE clause:
SELECT
id,
CASE WHEN isnumeric([other08]) = 1 THEN CAST([other08] AS INT) ELSE 0 END
FROM CER
WHERE sourcecode LIKE 'ANE%'
AND CASE WHEN isnumeric([other08]) = 1 THEN CAST([other08] AS INT) ELSE 0 END > 720
You need to use IsNumeric in your where clause, to avoid trying to compare strings to the number 720. Eg:
select id, case(isnumeric([other08])) when 1 then [other08] else 0 end
from CER
where sourcecode like 'ANE%' and ISNUMERIC(other08) = 1 and other08 > 720
EDIT
As #Abs pointed out, the above approach won't work. We can use a CTE to compute a reliable field to filter on, however:
WITH Data AS (
select id
, case WHEN isnumeric([other08]) THEN CAST([other08] AS int) else 0 end AS FilteredOther08
, CER.*
from CER
where sourcecode like 'ANE%'
)
SELECT *
FROM Data
WHERE [FilteredOther08] > 720

How do I generate a random number for each row in a T-SQL select?

I need a different random number for each row in my table. The following seemingly obvious code uses the same random value for each row.
SELECT table_name, RAND() magic_number
FROM information_schema.tables
I'd like to get an INT or a FLOAT out of this. The rest of the story is I'm going to use this random number to create a random date offset from a known date, e.g. 1-14 days offset from a start date.
This is for Microsoft SQL Server 2000.
Take a look at SQL Server - Set based random numbers which has a very detailed explanation.
To summarize, the following code generates a random number between 0 and 13 inclusive with a uniform distribution:
ABS(CHECKSUM(NewId())) % 14
To change your range, just change the number at the end of the expression. Be extra careful if you need a range that includes both positive and negative numbers. If you do it wrong, it's possible to double-count the number 0.
A small warning for the math nuts in the room: there is a very slight bias in this code. CHECKSUM() results in numbers that are uniform across the entire range of the sql Int datatype, or at least as near so as my (the editor) testing can show. However, there will be some bias when CHECKSUM() produces a number at the very top end of that range. Any time you get a number between the maximum possible integer and the last exact multiple of the size of your desired range (14 in this case) before that maximum integer, those results are favored over the remaining portion of your range that cannot be produced from that last multiple of 14.
As an example, imagine the entire range of the Int type is only 19. 19 is the largest possible integer you can hold. When CHECKSUM() results in 14-19, these correspond to results 0-5. Those numbers would be heavily favored over 6-13, because CHECKSUM() is twice as likely to generate them. It's easier to demonstrate this visually. Below is the entire possible set of results for our imaginary integer range:
Checksum Integer: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Range Result: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5
You can see here that there are more chances to produce some numbers than others: bias. Thankfully, the actual range of the Int type is much larger... so much so that in most cases the bias is nearly undetectable. However, it is something to be aware of if you ever find yourself doing this for serious security code.
When called multiple times in a single batch, rand() returns the same number.
I'd suggest using convert(varbinary,newid()) as the seed argument:
SELECT table_name, 1.0 + floor(14 * RAND(convert(varbinary, newid()))) magic_number
FROM information_schema.tables
newid() is guaranteed to return a different value each time it's called, even within the same batch, so using it as a seed will prompt rand() to give a different value each time.
Edited to get a random whole number from 1 to 14.
RAND(CHECKSUM(NEWID()))
The above will generate a (pseudo-) random number between 0 and 1, exclusive. If used in a select, because the seed value changes for each row, it will generate a new random number for each row (it is not guaranteed to generate a unique number per row however).
Example when combined with an upper limit of 10 (produces numbers 1 - 10):
CAST(RAND(CHECKSUM(NEWID())) * 10 as INT) + 1
Transact-SQL Documentation:
CAST(): https://learn.microsoft.com/en-us/sql/t-sql/functions/cast-and-convert-transact-sql
RAND(): http://msdn.microsoft.com/en-us/library/ms177610.aspx
CHECKSUM(): http://msdn.microsoft.com/en-us/library/ms189788.aspx
NEWID(): https://learn.microsoft.com/en-us/sql/t-sql/functions/newid-transact-sql
Random number generation between 1000 and 9999 inclusive:
FLOOR(RAND(CHECKSUM(NEWID()))*(9999-1000+1)+1000)
"+1" - to include upper bound values(9999 for previous example)
Answering the old question, but this answer has not been provided previously, and hopefully this will be useful for someone finding this results through a search engine.
With SQL Server 2008, a new function has been introduced, CRYPT_GEN_RANDOM(8), which uses CryptoAPI to produce a cryptographically strong random number, returned as VARBINARY(8000). Here's the documentation page: https://learn.microsoft.com/en-us/sql/t-sql/functions/crypt-gen-random-transact-sql
So to get a random number, you can simply call the function and cast it to the necessary type:
select CAST(CRYPT_GEN_RANDOM(8) AS bigint)
or to get a float between -1 and +1, you could do something like this:
select CAST(CRYPT_GEN_RANDOM(8) AS bigint) % 1000000000 / 1000000000.0
The Rand() function will generate the same random number, if used in a table SELECT query. Same applies if you use a seed to the Rand function. An alternative way to do it, is using this:
SELECT ABS(CAST(CAST(NEWID() AS VARBINARY) AS INT)) AS [RandomNumber]
Got the information from here, which explains the problem very well.
Do you have an integer value in each row that you could pass as a seed to the RAND function?
To get an integer between 1 and 14 I believe this would work:
FLOOR( RAND(<yourseed>) * 14) + 1
If you need to preserve your seed so that it generates the "same" random data every time, you can do the following:
1. Create a view that returns select rand()
if object_id('cr_sample_randView') is not null
begin
drop view cr_sample_randView
end
go
create view cr_sample_randView
as
select rand() as random_number
go
2. Create a UDF that selects the value from the view.
if object_id('cr_sample_fnPerRowRand') is not null
begin
drop function cr_sample_fnPerRowRand
end
go
create function cr_sample_fnPerRowRand()
returns float
as
begin
declare #returnValue float
select #returnValue = random_number from cr_sample_randView
return #returnValue
end
go
3. Before selecting your data, seed the rand() function, and then use the UDF in your select statement.
select rand(200); -- see the rand() function
with cte(id) as
(select row_number() over(order by object_id) from sys.all_objects)
select
id,
dbo.cr_sample_fnPerRowRand()
from cte
where id <= 1000 -- limit the results to 1000 random numbers
select round(rand(checksum(newid()))*(10)+20,2)
Here the random number will come in between 20 and 30.
round will give two decimal place maximum.
If you want negative numbers you can do it with
select round(rand(checksum(newid()))*(10)-60,2)
Then the min value will be -60 and max will be -50.
try using a seed value in the RAND(seedInt). RAND() will only execute once per statement that is why you see the same number each time.
If you don't need it to be an integer, but any random unique identifier, you can use newid()
SELECT table_name, newid() magic_number
FROM information_schema.tables
You would need to call RAND() for each row. Here is a good example
https://web.archive.org/web/20090216200320/http://dotnet.org.za/calmyourself/archive/2007/04/13/sql-rand-trap-same-value-per-row.aspx
The problem I sometimes have with the selected "Answer" is that the distribution isn't always even. If you need a very even distribution of random 1 - 14 among lots of rows, you can do something like this (my database has 511 tables, so this works. If you have less rows than you do random number span, this does not work well):
SELECT table_name, ntile(14) over(order by newId()) randomNumber
FROM information_schema.tables
This kind of does the opposite of normal random solutions in the sense that it keeps the numbers sequenced and randomizes the other column.
Remember, I have 511 tables in my database (which is pertinent only b/c we're selecting from the information_schema). If I take the previous query and put it into a temp table #X, and then run this query on the resulting data:
select randomNumber, count(*) ct from #X
group by randomNumber
I get this result, showing me that my random number is VERY evenly distributed among the many rows:
It's as easy as:
DECLARE #rv FLOAT;
SELECT #rv = rand();
And this will put a random number between 0-99 into a table:
CREATE TABLE R
(
Number int
)
DECLARE #rv FLOAT;
SELECT #rv = rand();
INSERT INTO dbo.R
(Number)
values((#rv * 100));
SELECT * FROM R
select ABS(CAST(CAST(NEWID() AS VARBINARY) AS INT)) as [Randomizer]
has always worked for me
Use newid()
select newid()
or possibly this
select binary_checksum(newid())
If you want to generate a random number between 1 and 14 inclusive.
SELECT CONVERT(int, RAND() * (14 - 1) + 1)
OR
SELECT ABS(CHECKSUM(NewId())) % (14 -1) + 1
DROP VIEW IF EXISTS vwGetNewNumber;
GO
Create View vwGetNewNumber
as
Select CAST(RAND(CHECKSUM(NEWID())) * 62 as INT) + 1 as NextID,
'abcdefghijklmnopqrstuvwxyz0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ'as alpha_num;
---------------CTDE_GENERATE_PUBLIC_KEY -----------------
DROP FUNCTION IF EXISTS CTDE_GENERATE_PUBLIC_KEY;
GO
create function CTDE_GENERATE_PUBLIC_KEY()
RETURNS NVARCHAR(32)
AS
BEGIN
DECLARE #private_key NVARCHAR(32);
set #private_key = dbo.CTDE_GENERATE_32_BIT_KEY();
return #private_key;
END;
go
---------------CTDE_GENERATE_32_BIT_KEY -----------------
DROP FUNCTION IF EXISTS CTDE_GENERATE_32_BIT_KEY;
GO
CREATE function CTDE_GENERATE_32_BIT_KEY()
RETURNS NVARCHAR(32)
AS
BEGIN
DECLARE #public_key NVARCHAR(32);
DECLARE #alpha_num NVARCHAR(62);
DECLARE #start_index INT = 0;
DECLARE #i INT = 0;
select top 1 #alpha_num = alpha_num from vwGetNewNumber;
WHILE #i < 32
BEGIN
select top 1 #start_index = NextID from vwGetNewNumber;
set #public_key = concat (substring(#alpha_num,#start_index,1),#public_key);
set #i = #i + 1;
END;
return #public_key;
END;
select dbo.CTDE_GENERATE_PUBLIC_KEY() public_key;
Update my_table set my_field = CEILING((RAND(CAST(NEWID() AS varbinary)) * 10))
Number between 1 and 10.
Try this:
SELECT RAND(convert(varbinary, newid()))*(b-a)+a magic_number
Where a is the lower number and b is the upper number
If you need a specific number of random number you can use recursive CTE:
;WITH A AS (
SELECT 1 X, RAND() R
UNION ALL
SELECT X + 1, RAND(R*100000) --Change the seed
FROM A
WHERE X < 1000 --How many random numbers you need
)
SELECT
X
, RAND_BETWEEN_1_AND_14 = FLOOR(R * 14 + 1)
FROM A
OPTION (MAXRECURSION 0) --If you need more than 100 numbers

Resources