 |
|
Writing Object Retrieval
Oracle Database Tips by Donald Burleson
|
Writing a relational interface to a
C++ application
When we look at writing a relational interface
to a C++ application we need to address the method to use for getting objects
from the relational database. Essentially, the retrieval process is very
straightforward. It includes:
1. See if the object is in memory
with a physical address.
2. If the object is not in memory retrieve the object with a logical
pointer.
The logical pointer was created for the purpose
of establishing a primary key for the table. The logical pointer was created
when the object was stored into the relational database, and it will be used to
navigate between orders.
In our example, we have a method called
ordersForCustomers that accepts a "customer" object address and displays the
orders for that customer. The logical pointer is called orderKey and it is the
primary key for our order table. The code is:
customer::ordersForCustomer()
{ // list all orders for a customer
cout << "\n\nOrder
summary for customer " << custName << " \n";
int i;
order * x;
cout << "\n Order = " << (x = orderPtr)->orderNum;
for(i=0;i<orderCount-1;i++) {
x = x->nextOrder;
cout << "\n Order = " << x->orderNum;
};
return 0;
};
Since this is a method of the customer class we
already have "currency" on the customer object and we assume that the "customer"
object is already in memory. We can see from the code example that the first
"order" object is retrieved by checking the orderPtr in the customer object.
From the first order, subsequent orders are retrieved by chasing the nextOrder
pointer within the order objects.
If we were to write this retrieval IN SQL, the
query might look like this:
SELECT orderNum
FROM order
WHERE order.custName = "IBM";
Of course, this retrieval would return many rows
from the order table, and if this query were embedded into a C program, a cursor
would need to be declared to fetch each row, one at a time.
EXEC SQL declare c1
cursor for
SELECT orderNum
FROM order o
WHERE
order.custName = :custname;
EXEC SQL open c1;
EXEC SQL fetch c1 into :orddate;
while (sqlca.sqlcode ==
0){
EXEC SQL fetch c1
into :ordnum;
}
EXEC SQL close c1;
We can now look at how these two retrieval
methods can be combined into a single C++ routine:
customer::ordersForCustomer()
{ // list all orders for a customer
cout << "\n\nOrder
summary for customer " << custName << " \n";
int i;
order * x;
if (orderPtr ==
NULL) getOrder(); // if order not in memory, fetch RDBMS
else
cout << "\n Order = " << (x = orderPtr)->orderNum; // get via
addr
for(i=0;i<orderCount-1;i++)
{
if (nextOrder == NULL) getOrder(); // get object from RDBMS
else { // if
already in-memory,
x = x->nextOrder; // chase physical address
cout << "\n Order = " << x->orderNum;
}
};
return 0;
};
getOrder() {
// fetch the desired object from the RDBMS...
EXEC SQL SELECT * from order into :orderstruct
where orderKey = order.orderKey;
//call the object
constructor...
order o123(IBM,:orderNum);
// Note that the ORDER constructor establishes the in-memory
// pointers to the customer object and related order objects
}
The basic premise here is that the in-memory
pointers are only established after the object has been retrieved from the
RDBMS, and the in-memory pointers have been created by the object constructor.
Note that we first check to see if the object is
in memory with the NULL test. If the in memory pointer is NULL, we then call
the getOrder() function to retrieve the object from the relational database,
calling the object constructor to instantiate the object to the run-time
system. The constructor will make the object exist in memory, and establish all
necessary pointers to owner objects (i.e. those that already exist in memory).
Other pointers to objects that have not yet been retrieved from the relational
database will have NULL pointers.
In summary, the paradigms of object-oriented
programming and relational databases technology are not as far apart as they
appear. If a C++ application is designed with well isolated I/O functions and
linked-list navigation, it can be very simple to back-end the C++ application to
utilize a relational database.
Appendix: original C++
application
#include <iostream.h>
#include <string.h>
#include <math.h>
#include <stdlib.h>
#include <time.h>
class customer;
class order;
class orderline;
class item;
class component;
class customer {
public:
char * custName;
order ** orderList; // pointers to orders for this customer
int orderCount;
customer(char * p) { // constructor for customer
cout << "creating customer " << p << "\n";
int l;
l = strlen(p); // get length of string
custName = new
char[l+1]; // get space for customer name
if (!custName) { cout << "allocation error!\n"; exit(1);}
strcpy(custName,p);
// copy in the customer name
orderCount = 0;
};
ordersForCustomer() {
// list all orders for a customer
cout << "\n\nOrder
summary for customer " << custName << " \n";
int i;
for(i=0;i<orderCount;i++) {
cout << "\n Order = " << orderList[i]->orderNum;
};
return 0;
};
};
class order {
public:
customer * custOwner; // pointer to customer placing this order
orderline ** itemList; // pointers to line items in the order
int orderNum, itemCount;
order(customer &cust,
int ordnum) { // order constructor
cout << "\nCreating order " << ordnum << "\n";
itemCount = 0;
orderNum = ordnum; // set the order number
cust.orderList[cust.orderCount++] = this; // set cust -> order
custOwner = &cust; // set order-> customer pointer
};
itemsInOrder() { // { // list all items for an order
// get the items and count the item names...
cout << "\n\n Item list for order number " << orderNum << "\n";
int i, quant;
float pr, totCost;
for(i=0;i<itemCount;i++) {
// get the orderline object ...
item * temp = itemList[i]->itemOwner;
// get the item object ...
pr = temp->itemPrice;
// save the quantity ordered
quant = itemList[i]->quantity;
totCost = pr * quant;
cout << " Item " << temp->itemName;
cout << " cost " << pr;
cout << " quant " << quant;
cout << " total " << totCost << "\n";
};
return 0;
};
};
class item {
public:
orderline ** orderList; // pointers to orderlines
component ** hasParts; // pointers to has-parts
component ** isAPart; // pointers to is-a-part
char * itemName;
int inventoryLevel, price, orderCount, hasPartCount, isAPartCount;
float itemPrice;
item(char * p, int
quantityReceived, float price ) { // item constructor
cout << "\nCreating item " << p << "\n";
orderCount=0;
itemPrice = price; // set item price;
inventoryLevel = quantityReceived; // set quantity
int l;
l = strlen(p); // get length of string
itemName = new char[l+1]; // get space for name
if (!itemName) { cout << "allocation error!\n"; exit(1);}
strcpy(itemName,p); // copy in item name
};
};
class orderline {
public:
order * orderOwner; // pointer to order
item * itemOwner; // pointer to item
int quantity;
orderline(order &ord, item &it, int qty) { // orderline constructor
cout << " creating orderline\n";
ord.itemList[ord.itemCount++] = this; // set cust -> order
it.orderList[it.orderCount++] = this;
orderOwner = ⩝ // set pointer to order
itemOwner = ⁢ // set pointer to item
quantity = qty; // set quantity ordered
};
};
class component {
public:
item * hasAPart; // pointer to owner item
item * isAPart; // pointer to owner item
int quantity;
component(item &has, item &isA, int quant) { // BOM constructor
has.hasParts[has.hasPartCount++] = this; // set has-part pointer
isA.isAPart[isA.isAPartCount++] = this; // set is-a-part
quantity = quant; // set quantity
}
};
main() {
customer IBM("IBM");
customer ATnT("ATnT");
item pen("pen",300,1.95);
item pencil("pencil",400,1.30);
item pad("pad",100,2.95);
order o123(IBM,123);
orderline i1(o123,pen,3);
orderline i2(o123,pencil,4);
order o456(ATnT,456);
orderline i3(o456,pen,6);
orderline i4(o456,pencil,16);
orderline i5(o456,pad,12);
order o789(ATnT,789);
orderline i6(o789,pencil,50);
orderline i7(o789,pad,10);
orderline i8(o789,pen,2);
IBM.ordersForCustomer();
ATnT.ordersForCustomer();
o123.itemsInOrder();
o456.itemsInOrder();
o789.itemsInOrder();
//ATnT.ordersForCustomer();
//IBM.ordersForCustomer();
return 0;
}