memset works on contigous memory. the zeros you want to add are randomly accessed.

One might question why you do this? If you want a null terminated string, why not use an array of char?


here is what I would have done:

void process( void )
{
typedef char enameType[30];
exec sql type enameType is string(30);
enameType ename[100];

char job[100][20];
exec sql var job is string(20);

int i;

exec sql whenever sqlerror do sqlerror_hard();

exec sql declare c cursor
for
select ename, job
from emp;

exec sql open c;
for( ; sqlca.sqlcode == 0; )
{
exec sql fetch c into :ename, :job;
for( i = 0; i < sqlca.sqlerrd[2]; i++ )
printf( "'%s', '%s'\n", ename[i], job[i] );
}
exec sql close c;
exec sql commit;
}

That shows two different ways.


typedef char enameType[30];
exec sql type enameType is string(30);
enameType ename[100];

is nice if you have lots of homogenous sized variables as it associates the datatype with a string.


char job[100][20];
exec sql var job is string(20);

is nice if you don't. if most variables are "their own size".

The output of that is:


[tkyte@tkyte-pc-isdn t]$ ./t

Connected to ORACLE as user: /

'SMITH', 'CLERK'
'ALLEN', 'SALESMAN'
'WARD', 'SALESMAN'
'JONES', 'MANAGER'
'MARTIN', 'SALESMAN'
'BLAKE', 'MANAGER'
'CLARK', 'MANAGER'
'SCOTT', 'ANALYST'
'KING', 'PRESIDENT'
'TURNER', 'SALESMAN'
'ADAMS', 'CLERK'
'JAMES', 'CLERK'
'FORD', 'ANALYST'
'MILLER', 'CLERK'

1) from the pro*c side? no.

from your code side? perhaps -- it'll change your code and if you made assumptions that "this is varchar" in your code and don't catch that, you'll have a bug of course.

2) you'd have to benchmark. I've never benchmarked it.

3) you tell me, you did it :)

1)
</code> http://docs.oracle.com/docs/cd/B10501_01/appdev.920/a97269/pc_08arr.htm#1988 <code>

2) proc command line.

my car won't start this morning. any clues?

but yes, in general -- the semantics of an UPDATE are not conducive to an array of structs.

A struct of arrays, yes.

use the command line version?

the GUI version is not running "in dos", it is running in the window. the "host cls" command is just firing up a dos window, clearing the screen and exiting.

I never liked that gui version, i always just use the command line version when forced to use windows.

huh? int's are not "null terminated"?

if you fetch a number into a string -- that is just like you used to_char() -- meaning that you do it the same way you did it for "job" -- no different.

if you fetch into a string, you use the same technique.

if you fetch into int's -- you don't null terminate them.

the distinction between sql and plsql blurs - melds together - becomes hard to see - since plsql and sql are tied so tightly together.

in the example above -

declare
    x   varchar2(20);
    eno number;
begin
   x := 'KING';

   select empno into eno from emp where ename = x;
end;
/

x and eno are host variables - but we don't really call them that by convention when speaking about plsql. they are just plsql variables. They are bound into the sql statement transparently, they are fetched into transparently - you don't demark them with a ":" or anything. the line between sql and plsql is "blurry" since they are very tightly coupled.