Thread: may be a mismatch between the construct_array and construct_md_array comments

While working on the "POC, WIP: OR-clause support for indexes" project [0], it was suggested to use the construct_array function to form a one-dimensional array.

I noticed that there is a comment that values ​​with NULL are not processed there, but in fact this function calls the construct_md_array function, which
contains a comment that it can handle NULL values.

/*
 * construct_array    --- simple method for constructing an array object
 *
 * elems: array of Datum items to become the array contents
 *          (NULL element values are not supported).

*/

/*
 * construct_md_array    --- simple method for constructing an array object
 *                            with arbitrary dimensions and possible NULLs

*/

In the places where the construct_md_array function is called, I did not see a check for NULL and a limitation on the use of the function, if any.

The tests during the check did not show that there is a problem with this [1].

Is this comment correct or we should update it?

[0] https://commitfest.postgresql.org/49/4450/

[1] https://www.postgresql.org/message-id/CACJufxHCJvC3X8nUK-jRvRru-ZEXp16EBPADOwTGaqmOYM1Raw%40mail.gmail.com

Alena Rybakina <a.rybakina@postgrespro.ru> writes:
> I noticed that there is a comment that values ​​with NULL are not 
> processed there, but in fact this function calls the construct_md_array 
> function, which
> contains a comment that it can handle NULL values.

Right.  construct_md_array has a "bool *nulls" argument, but
construct_array doesn't --- it passes NULL for that to
construct_md_array, which will therefore assume there are no null
array elements.

            regards, tom lane

On 12.09.2024 20:44, Tom Lane wrote:
> Alena Rybakina <a.rybakina@postgrespro.ru> writes:
>> I noticed that there is a comment that values ​​with NULL are not
>> processed there, but in fact this function calls the construct_md_array
>> function, which
>> contains a comment that it can handle NULL values.
> Right.  construct_md_array has a "bool *nulls" argument, but
> construct_array doesn't --- it passes NULL for that to
> construct_md_array, which will therefore assume there are no null
> array elements.
>
Understood.

At first I thought this comment was related to the value of a NULL 
element that might be in the Array, but now I realized that this is not 
the case.

Thanks for the explanation, it helped a lot!