Why is casting from float to varchar being rounded in SQL Server?

Question

The following SQL,

declare @a as float, @b as float

select @a=1.353954 , @b=1.353956
select 
CAST(@a as VARCHAR(40)) AS a_float_to_varchar ,
CAST(@b as VARCHAR(40)) AS b_float_to_varchar

results in

a_float_to_varchar                       b_float_to_varchar
---------------------------------------- ----------------------------------------
1.35395                                  1.35396

based on 'float' and 'real' (Transact-SQL).

Float has a precision of 15 digits, so I am not sure why the number is being rounded when converted to varchar.

Answer 1

You can specify style to include more digits.

declare @gg float
set @gg = 124.323125453
SELECT @gg,Convert(varchar, @gg,128)

For newer versions of SQL Server, use SELECT @gg,Convert(varchar, @gg,3)

returns

124.323125453    124.323125453

Reference: CAST and CONVERT (Transact-SQL)

Or with STR():

declare @gg float
set @gg = 124.323124354234524
SELECT @gg,str(@gg,16,15)

It should give you all the possible digits. 16 is the total possible length (includes period) while 15 places after the decimal is possible (actually 0.2323... the 0 count toward length, so the length needs to be 17 if all numbers are less that 1). STR(), however, pads the results with leading spaces and trailing 0.

Answer 2

Cast to decimal before casting to varchar:

declare @a as float, @b as float

select @a=1.353954 , @b=1.353956
select
CAST(CAST(@a AS DECIMAL(38,18)) as VARCHAR(40)) AS a_float_to_varchar ,
CAST(CAST(@b AS DECIMAL(38,18)) as VARCHAR(40)) AS b_float_to_varchar

Answer 3

Also from your link (it's actually the first line):

Approximate-number data types...

If you want exact precision, don't use float.

That being said, there is a function STR() specifically for converting float to a character data type.