Oracle SQL (Pro*C): knowing number of rows of a select beforehand

In a Pro*C file, I have a fetch pattern like this one:

EXEC SQL BEGIN DECLARE SECTION;
    typedef struct my_row_t
    {
        unsigned field1;
        unsigned field2;
    } my_row;
EXEC SQL END DECLARE SECTION;

void do_query()
{
    EXEC SQL BEGIN DECLARE SECTION;
        struct my_row* buffer;
        unsigned buffer_size;
    EXEC SQL END DECLARE SECTION;

    buffer_size = /* some tied upper bound on the number of rows to fetch */;
    buffer = (my_row*)malloc(buffer_size * sizeof(my_row));
  
    // Create my_cursor based on some select query
    // ...
    // and now:

    EXEQ SQL FOR :buffer_size FETCH my_cursor INTO :buffer;
    unsigned number_of_fetched_rows = sqlca.sqlerrd[2];

    // At this point, with just one fetch I have all the rows
}

The problem is that, sometimes, you can’t estimate a proper upper bound. I’m talking about queries that won’t return more than a few thousand rows, but I don’t want to allocate storage for, for example, 5000 rows, when a particular query for a particular input will only give you 500.

Is there a way to know in an efficient way, before doing your first fetch (for example, just after opening the cursor), the number of rows to fetch, without executing any count(*) query (for performance reasons)?

My goal is to “fetch all” in a single pass in the most performant way as possible.

I think it can be done with PL/SQL tables/records, but I don’t know how to “save” or extract the content of a PL/SQL table in a buffer of my own using Pro*C syntax.

Michael Caine

Michael Caine