be used to retrieve multiple columns at once.  The
      structure enables handling multiple column values in a single
      host variable.
     </para>

     <para>
      The following example retrieves OIDs, names, and sizes of the
      available databases from the <literal>pg_database</literal>
      system table and using
      the <function>pg_database_size()</function> function.  In this
      example, a structure variable <varname>dbinfo_t</varname> with
      members whose names match each column in
      the <literal>SELECT</literal> result is used to retrieve one
      result row without putting multiple host variables in
      the <literal>FETCH</literal> statement.
<programlisting>
EXEC SQL BEGIN DECLARE SECTION;
    typedef struct
    {
       int oid;
       char datname[65];
       long long int size;
    } dbinfo_t;

    dbinfo_t dbval;
EXEC SQL END DECLARE SECTION;

    memset(&amp;dbval, 0, sizeof(dbinfo_t));

    EXEC SQL DECLARE cur1 CURSOR FOR SELECT oid, datname, pg_database_size(oid) AS size FROM pg_database;
    EXEC SQL OPEN cur1;

    /* when end of result set reached, break out of while loop */
    EXEC SQL WHENEVER NOT FOUND DO BREAK;

    while (1)
    {
        /* Fetch multiple columns into one structure. */
        EXEC SQL FETCH FROM cur1 INTO :dbval;

        /* Print members of the structure. */
        printf("oid=%d, datname=%s, size=%lld\n", dbval.oid, dbval.datname, dbval.size);
    }

    EXEC SQL CLOSE cur1;
</programlisting>
     </para>

     <para>
      This example shows following result. (The exact values depend on
      local circumstances.)
<screen>
oid=1, datname=template1, size=4324580
oid=11510, datname=template0, size=4243460
oid=11511, datname=postgres, size=4324580
oid=313780, datname=testdb, size=8183012
</screen>
     </para>

     <para>
      Structure host variables <quote>absorb</quote> as many columns
      as the structure as fields.  Additional columns can be assigned
      to other host variables. For example, the above program could
      also be restructured like this, with the <varname>size</varname>
      variable outside the structure:
<programlisting>
EXEC SQL BEGIN DECLARE SECTION;
    typedef struct
    {
       int oid;
       char datname[65];
    } dbinfo_t;

    dbinfo_t dbval;
    long long int size;
EXEC SQL END DECLARE SECTION;

    memset(&amp;dbval, 0, sizeof(dbinfo_t));

    EXEC SQL DECLARE cur1 CURSOR FOR SELECT oid, datname, pg_database_size(oid) AS size FROM pg_database;
    EXEC SQL OPEN cur1;

    /* when end of result set reached, break out of while loop */
    EXEC SQL WHENEVER NOT FOUND DO BREAK;

    while (1)
    {
        /* Fetch multiple columns into one structure. */
        EXEC SQL FETCH FROM cur1 INTO :dbval, :size;

        /* Print members of the structure. */
        printf("oid=%d, datname=%s, size=%lld\n", dbval.oid, dbval.datname, size);
    }

    EXEC SQL CLOSE cur1;
</programlisting>
     </para>
    </sect4>

    <sect4 id="ecpg-variables-nonprimitive-c-typedefs">
     <title>Typedefs</title>
     <indexterm>
      <primary>typedef</primary>
      <secondary>in ECPG</secondary>
     </indexterm>

     <para>
      Use the <literal>typedef</literal> keyword to map new types to already
      existing types.
<programlisting>
EXEC SQL BEGIN DECLARE SECTION;
    typedef char mychartype[40];
    typedef long serial_t;
EXEC SQL END DECLARE SECTION;
</programlisting>
      Note that you could also use:
<programlisting>
EXEC SQL TYPE serial_t IS long;
</programlisting>
      This declaration does not need to be part of a declare section;
      that is, you can also write typedefs as normal C statements.
     </para>

     <para>
      Any word you declare as a <literal>typedef</literal> cannot be used as
      an SQL keyword in <literal>EXEC SQL</literal> commands later in the same
      program. For example, this won't work:
<programlisting>
EXEC SQL BEGIN DECLARE SECTION;
    typedef int start;
EXEC SQL END DECLARE SECTION;