Use the row's data AFTER its remove from DefaultTableModel

Question

Yeah I know, this sounds stupid. But before let me illustrate my problem.

I have a GUI with A LOT of JTable. I have various DefaultTableModel, one for each JTable. So, 1:1. And for each DefaultTableModel I have create one a TableModelListener.

For insert and delete elements in my GUI I have create only TWO methods, addVehicle and removeVehicle:

public void addVehicle(final DefaultTableModel model, final String s1, final String s2) {
    Runnable addV = new Runnable() {
        @Override
        public void run() {
            model.addRow(new Object[] { s1, s2 });
        } 
    };
    SwingUtilities.invokeLater(addV);
}


public void removeVehicle(final DefaultTableModel model, final int index) {
    if (row!=null){
        String s1= row.elementAt(0).toString();
        Runnable removeV = new Runnable() {
            @Override
            public void run() {
                model.removeRow(index);
            }
        };
        SwingUtilities.invokeLater(removeV);
    }
}

These methods are perfect, recognize automatically the right tablemodel.

But I have a necessity: since I have many JTable (and corresponding DefaultTableModels) and given that when I add/delete a row from a table a third-party software MUST BE INFORMED, I have created several TableModelListener.

For example listener1:

TableModelListener listener1 = new TableModelListener() {
    public void tableChanged(TableModelEvent e) {
        DefaultTableModel m = (DefaultTableModel) e.getSource();
        int row = e.getLastRow();
        switch (e.getType()) {
            case (TableModelEvent.INSERT): {
                System.out.println("Insert! Key: "+m.getValueAt(row, 0));

                // My agent, the third-party software: 
                shipperAgent.newTruck((String) m.getValueAt(row, 0));
            } break;

            case (TableModelEvent.DELETE): {
                System.out.println("Delete! Key:"+m.getValueAt(row, 0));

                // My agent, the third-party software: 
                shipperAgent.removeTruck((String)m.getValueAt(row, 0));
            } break;

            case (TableModelEvent.UPDATE): {
                // ...
            } break;
        }
    }
};

Another example is listener 2 (I show you just a fragment...):

            case (TableModelEvent.INSERT): {
                System.out.println("Activate! Key: "+m.getValueAt(row, 0));

                // My agent, the third-party software: 
                shipperAgent.activateTruck((String) m.getValueAt(row, 0));
            } break;

            case (TableModelEvent.DELETE): {
                System.out.println("Deactivate! Key:"+m.getValueAt(row, 0));

                // My agent, the third-party software: 
                shipperAgent.deactivateTruck((String)m.getValueAt(row, 0));
            } break;

This is the problem: is all right for the insert, but not for the DELETE. I know why: when I remove a row from the GUI, the row is deleted and in the switch (case TableModelEvent.DELETE) work on another row (that immediately after the one that I deleted). Moreover, if this row is the last (since there are no more lines after) i get a NullPointerException.

HOW I CAN, in the listener, to use a row BEFORE it is actually removed? (or other brilliant ways?)

Answer 1

...when I add/delete a row from a table a third-party software MUST BE INFORMED, I have created several TableModelListener.

If you see TableModelEvent API there's no way to get data from the event itself, all you have is the TableModel which is the source of the event. Having said this, when a table model event is fired the change in the model has been already done and it's nothing you can do about it, so you cannot get the deleted(s) row(s) value(s) just because those won't be there anymore.

However if your concern is to notify this third-party software in order to keep consistency I guess, then you should consider do that before removing rows from the table model.

public void removeVehicle(final DefaultTableModel model, final int index) {
    String value = (String)model.getValueAt(index, 0);
    shipperAgent.deactivateTruck(value);
    // If everything went ok, then remove the row from the table model.
    // Of course, do it in the Event Dispatch Thread.
    model.removeRow(index);
}

This way you can even coordinate that third-party software and your application: if the truck represented by the row cannot be deactivated then don't remove the row in your table model and show some error message instead.

---

Edit

But I wanted to centralize communication with the agent (third party software) and separate it from the rest of the code, and not disseminate it.

Well it makes perfectly sense and I'm sure there are more than one way to accomplish this, but IMHO your design needs another twist in order to fulfil these requirements:

• Have only one point to coordinate between the third-party agent and your table model.
• Don't write lot of methods that do practically the same thing: DRY principle
• Process the actions in the right time.
• All the responsibilities have to be well separated.

I was thinking that something close to Mediator pattern (not exactly this pattern but something close) might help you to encapsulate the coordination logic and decouple the third-party agent and table models involved in the same transaction. For example consider this prototype class:

public abstract class Coordinator {

    /*
     * protected class members so subclasses can access these directly
     */

    protected ShipperAgent shipperAgent;
    protected DefaultTableModel tableModel;

    public Coordinator(ShipperAgent sa, DefaultTableModel tm) {
        shipperAgent = sa;
        tableModel = tm;
    }

    public abstract void notifyAndAddRow(Object[] rowData);

    public abstract void notifyAndDeleteRow(int rowIndex);
}

Since you have managed to identify the right table models by the time you call addVehicle(...) and removeVehicle(...) methods, you will probably can do the same to identy the right coordinator and delegate the task of notify the third-party agent and update the table model. For instance:

ShipperAgent sa = new ShipperAgent(...);
DefaultTableModel model1 = new DefaultTableModel(...);
DefaultTableModel model2 = new DefaultTableModel(...);

Coordinator coordinator1 = new Coordinator(sa, model1) {
    @Override
    public void notifyAndAddRow(Object[] rowData) {
        this.shipperAgent.newTruck((String) rowData[0]);
        this.tableModel.addRow(rowData); // do this in the EDT
    }

    @Override
    public void notifyAndDeleteRow(int rowIndex) {
        String truck = (String)this.tableModel.getValueAt(rowIndex, 0);
        this.shipperAgent.removeTruck(truck);
        this.tableModel.removeRow(rowIndex); // do this in the EDT
    }
};

Coordinator coordinator2 = new Coordinator(sa, model2) {
    @Override
    public void notifyAndAddRow(Object[] rowData) {
        this.shipperAgent.activateTruck((String) rowData[0]);
        this.tableModel.addRow(rowData); // do this in the EDT
    }

    @Override
    public void notifyAndDeleteRow(int rowIndex) {
        String truck = (String)this.tableModel.getValueAt(rowIndex, 0);
        this.shipperAgent.removeTruck(truck);
        this.tableModel.removeRow(rowIndex); // do this in the EDT
    }
};

Then your addVehicle(...) and removeVehicle(...) methods could look like this:

public void addVehicle(Coordinator coordinator, String s1, String s2) {
    coordinator.notifyAndAddRow(new Object[]{s1, s2});
}

public void removeVehicle(Coordinator coordinator, int index) {
    coordinator.notifyAndDeleteRow(index);
}

Of course as I've said you will have to manage to identify the right Coordinator instance by the time you call addVehicle(...) and removeVehicle(...) methods, but I think you have the abstraction required to solve your problem now.