Re: Is there a reason why Postgres doesn't have Byte or tinyint?

On Sun, Jan 8, 2012 at 21:31, Mike Christensen <mike(at)kitchenpc(dot)com> wrote:
> On Sun, Jan 8, 2012 at 10:12 AM, Francisco Figueiredo Jr.
> <francisco(at)npgsql(dot)org> wrote:
>> On Sun, Jan 8, 2012 at 06:54, Mike Christensen <mike(at)kitchenpc(dot)com> wrote:
>>>>>>>> According to the manuals, Postgres has smallint (2 byte), integer (4
>>>>>>>> bytes) and bigint (8 bytes)..  I use a lot of structures with "bytes"
>>>>>>>> in my code and it's kinda annoying to cast DB output from Int16 to
>>>>>>>> Byte every time, especially since there's no explicit cast in .NET and
>>>>>>>> you have to use System.Convert().
>>>>>>>>
>>>>>>>> Is there a work-around, or do people just cast or use Int16 in their
>>>>>>>> data structures?  Just wondering..  I know on modern computers it
>>>>>>>> probably doesn't make any difference anyway..
>>>>>>>
>>>>>>>
>>>>>>> Is this just about programmer convenience or is it about space efficiency
>>>>>>> in
>>>>>>> the database?  BYTEA might help you.  Or try declaring a DOMAIN over
>>>>>>> SMALLINT that limits allowed values to the range of a byte. -- Darren
>>>>>>> Duncan
>>>>>>
>>>>>>
>>>>>> This is purely programmer convenience.
>>>>>>
>>>>>> Basically, I want Npgsql to marshal the value as a .NET Byte type, if
>>>>>> I can find a way to do that I'm happy.  Perhaps it's more of a Npgsql
>>>>>> question, though I'm curious as to why Postgres doesn't have an
>>>>>> intrinsic tinyint or byte type.
>>>>>
>>>>>
>>>>> Maybe Postgres doesn't need a Byte type predefined because it gives you the
>>>>> means to define the type yourself, such as using DOMAIN.
>>>>>
>>>>> Generally speaking, I believe it is more important for a type system to
>>>>> provide the means for arbitrary user-defined types which can be used in all
>>>>> the places as built-in-defined types, than to have large numbers of
>>>>> built-in-defined types.
>>>>
>>>> Precisely.  postgresql's extensable nature allows you to build your
>>>> own types as well.  If it's popular enough it'll make it into contrib,
>>>> then maybe core.  My guess is that there's some non-trivial cost to
>>>> maintaining each core type, and since a byte type isn't required by
>>>> the SQL spec, it would take some effort to get a standard one included
>>>> in the core.
>>>
>>> That makes sense.
>>>
>>> I guess my question is more of a NpgSql question then.  Is there a way
>>> to create a custom PG type, and have npgsql serialize that type in a
>>> dataset to a .NET Byte type?
>>>
>>> I'd probably be better off posting on the npgsql mailing list, but
>>> perhaps someone here knows as well..
>>>
>>
>>
>> Hi!
>>
>> Npgsql already has support to map .Net byte data type to int2 and vice-versa.
>>
>> You can see all the supported type mappings in this file:
>> http://cvs.pgfoundry.org/cgi-bin/cvsweb.cgi/npgsql/Npgsql2/src/NpgsqlTypes/NpgsqlTypesHelper.cs?rev=1.41&content-type=text/x-cvsweb-markup
>>
>> Check method PrepareDefaultTypesMap().
>>
>> I hope it helps.
>
> Thanks!  I'll have to mess around with this a bit more..  From what
> I've seen so far, functions that have int2 out parameters will return
> Int16 through the DataReader..  Maybe I'm doing something wrong..
>
> Mike

I think you aren't doing anything wrong. int2 postgresql datatypes are
supposed to be mapped to .net int16 types.

Do you have a simple example?

What type were you expecting in the datareader? Maybe there is a
missing mapping in Npgsql.

Thanks in advance.

--
Regards,

Francisco Figueiredo Jr.
Npgsql Lead Developer
http://www.npgsql.org
http://gplus.to/franciscojunior
http://fxjr.blogspot.com
http://twitter.com/franciscojunior