Re: [HACKERS] [BUG?] strange PQgetisnull

"Park, Chul-Su" <pcs@mhlx01.kek.jp> writes:
>     I got strange result with PQgetisnull, say
> for empty table(I think that it's empty so null)
> I guess that
> after fetch some selection
> PQgetisnull(result, 0, 0) should give me "1 or true"
> But
> it gives me
> ERROR! field 0(of 0) of row 0(of 0) is not available... Segmentation
> Fault

PQgetisnull is buggy in 6.3.2: it range-checks the tuple and field
numbers, and complains if they are out of range ... but then falls
through and tries to reference the tuple info anyway.  Thus, segfault.

It should return a default value (probably 1 to pretend the field is
NULL) when the indexes are out of range.  This is already fixed in the
current development sources, but if you want to stick with a 6.3.2
server then you will have to modify fe-exec.c yourself.

> so I tried
> if (PQntuples(result) == 0)
> ...
> But  PQntules gives me "1" not "0" !

The PQgetisnull error message you quoted above indicates (after looking
at the 6.3.2 sources) that nfields was 1 and ntuples was 0.  I think you
are testing the results of a different query here.

> do I have to check result with
> char* value = PQgetvalue(result, 0, 0);
> and test value is "" or not?  any idea???

No, you should be checking PQntuples and perhaps also PQnfields to be
sure that the indexes you are going to use are OK.

            regards, tom lane