On creating a table column of type “float with precision” column gets created as “real” type in SQL server

问题

I created a simple table as below:

create table table1
(
col1 float(1),
col2 float
)

When I check the table definition using sp_help or script table as create, I get the float(1) as real

CREATE TABLE [dbo].[table1](
[col1] [real] NULL,
[col2] [float] NULL
)

As per my understanding from msdn and stackoverflow references, float can have a precision n ranging from 1 to 53.

Why is it getting converted to real by default whenever float has a precision specified?

回答1:

SQL Server doesn't remember what language you used when you create the table - it just consults the current table definition and renders a piece of text that could have been used to create an identical table.

As the documentation points out, SQL Server doesn't support arbitrary precisions - it only supports two precisions - 24 and 53:

SQL Server treats n as one of two possible values. If 1<=n<=24, n is treated as 24. If 25<=n<=53, n is treated as 53.

As it also points out, real is treated as a synonym for float(24):

The ISO synonym for real is float(24).

So, any float column specified with a precision of 24 or lower will in fact be created as a float(24). And a synonym for float(24) is real. The system doesn't remember whether it was originally specified as float(1), float(24) or real.

---

Why is it getting converted to real by default whenever float has a precision specified?

It doesn't. As per the above, if you create a column with type float(25) (or any higher precision), you'll find it gets returned as plain float, rather than real, because it was created as a float(53) and 53 is the default precision when none is supplied.

来源：https://stackoverflow.com/questions/40398692/on-creating-a-table-column-of-type-float-with-precision-column-gets-created-as