Re: determine what column(s) form the primary key, in C extention

I'm emailing it to the 'general' list.


(1)

This:


     int i = -1;
     while ((i = bms_next_member(pkattnos , i)) >= 0) {
         /* do stuff with i */
         /* you'll need to use i - FirstLowInvalidHeapAttributeNumber to 
get the pg_attribute.attnum */


         elog(INFO, "bms_next_member i: %d", i);
     }


prints 10 and then 9

Then:

   10 - FirstLowInvalidHeapAttributeNumber (-8) ==> 2

   9 - FirstLowInvalidHeapAttributeNumber (-8) ==> 1


These are indexes of the columns, right?

Do they start from 1, not from 0?



(2)

I'll use this C code as an example to build an extention in Rust. The 
Postgresql bindings for Rust I have don't contain a definition of 
`FirstLowInvalidHeapAttributeNumber` for some reason. I can define it 
since it's simply single digit constant.

However what does in some source files it's defined as (-7) and in some 
as (-8)? Which should I use?




El 28/07/2020 a las 03:20, David Rowley escribió:
> Hi Alex,
>
> On Tue, 28 Jul 2020 at 05:47, alex maslakov <alex@serendipia.email> wrote:
>> I was suggested to use `get_primary_key_attnos` from
>> `src/include/catalog/pg_constraint.h`
>>
>> extern Bitmapset *get_primary_key_attnos(Oid relid, bool deferrableOk)
>>
>>
>> It returns *Bitstamp. And it's got "nwords" and "words[]". But those
>> return just big numbers, not something that look similar to an index of
>> the primary key column.
>>
>>
>> And therefore I haven't had any luck thus far.
>>
>> How to do it?
> You'll need to loop over the return value of that function with
> bms_next_member()
>
> e.g.
> pkattnos = get_primary_key_attnos(oid, false);
> i = -1;
> while ((i = bms_next_member(pkattnos , i)) >= 0)
> {
>    /* do stuff with i */
> /* you'll need to use i - FirstLowInvalidHeapAttributeNumber to get
> the pg_attribute.attnum */
> }
>
> For the future, for questions, you should use the general list.  If
> the question is very source code related then you might have more luck
> in pgsql-hackers.  This is not the right list. Please post any
> followup questions on one of those lists.
>
> Note the description for this list, per
> https://www.postgresql.org/list/ is "Notification of git commits are
> sent to this list. Do not post here!"
>
> David