Re: BUG #14268: NULL parameter conversion

So, you're saying SQL is too sane a language for you and you'd rather have
30+ non-convertible types of null?

2016-07-28 16:56 GMT+03:00 David G. Johnston <david.g.johnston@gmail.com>:

> On Thu, Jul 28, 2016 at 9:30 AM, <coladict@gmail.com> wrote:
>
>> The following bug has been logged on the website:
>>
>> Bug reference:      14268
>> Logged by:          Jordan Gigov
>> Email address:      coladict@gmail.com
>> PostgreSQL version: 9.3.13
>> Operating system:   Ubuntu 14.04
>> Description:
>
> =E2=80=8B[...]=E2=80=8B
>
>>
>>
> When the value is NULL, the type
>> shouldn't matter.
>>
>
> =E2=80=8BAccording to what authority?
>
> =E2=80=8B[...]
>>
> INSERT INTO my_array_table(somedata) VALUES (NULL::bytea);
>>
>
> =E2=80=8BIf you are saying the above should work I'd say that is arguable=
 at
> best.=E2=80=8B  All values, even NULL, are typed in PostgreSQL.
>
>
>> I recognize that the JDBC driver explicitly sends the parameter type in =
a
>> prepared statement, but I think it should be ignored when the value is
>> NULL.
>>
>
> =E2=80=8BAs your first example shows if the NULL remains unknown it will =
be
> auto-cast according to the context in which it is used.  Its not
> PostgreSQL's place to discard type information.
>
> =E2=80=8BThe JDBC API defines "setNull(int parameterIndex, int=E2=80=8B s=
qlType); which
> makes me inclined to say your complaint should be directed at JPA and not
> either PostgreSQL itself or JDBC.
>
> Now, as I am lacking knowledge about the specific problem, JPA, and how
> other databases function, I am unable to meaningfully comment further.  B=
ut
> I can say this isn't a bug.  PostgreSQL is operating as expected given ho=
w
> it handles NULL.
>
> David J.
>
>
>