I greatly appreciate all the responses so far. They will be immensely helpful. Specially thanks Jerson for sharing the trick. I'll surely fit it somewhere in my code.

And yes I was guessing too that a layer above the keypad routines will be required with low level routines unchanged. So I didn't change the low level ones.

I have 2 functions for scanning keypad at present:

unsigned char get_key_status(void)
{
    unsigned char code keypad[4][3]= {  '1','2','3',
                                        '4','5','6',
                                        '7','8','9',
                                        '*','0','#'  };
...
return keypad[x][y];
...

 

That is, a function which when called returns the instantaneous key being pressed during it's execution. Otherwise returns 0 and exits. This is useful when continuous scanning is required e.g in while loops etc.
The other one is bring_key() which when called stays in loop until a key is pressed, returns that key & exits. This is useful when some menu comes on the interface & user has to choose options from 1,2,3 etc. and there's nothing really special to do until user chooses an option.

To start with, I decided to make a routine that prints 'a' if 2 is pressed. If 2 is pressed again within 1 sec, it replaces 'a' with 'b'.
Now the very primitive pseudocode that I have thought of for just getting the '2' key to work for inputting a,b is

multitap_mode()    //called from main()
{
    WHILE FOREVER
    { 
        IF any_key_pressed        //key = '2' here
         THEN 
        {    print_base_char_for_key    //eg 'a' for 2
             
            FOR(i=0;i<1000,i++)    //scan 1000 times in 1 sec    
            {    IF same_key_pressed_again        //no support for pressing other keys yet
                THEN
                {   DECREMENT cursor_position
                    print_next_char_for_key    //should print 'b' now
                //will be 1 more loop here
                }    
                DELAY (1ms)    //wait for 1 second total in for loop
            }
        }
    }
}
 

There should be 1 more embedded for loop for 'c' which I haven't added yet. No exit paths to main() too. There is much to do here.

I have to use for loop since unfortunately all my timers are used for other purposes. So using timers for delay length measurement is not possible.

Any improvements/suggestions?

-Regards,
Munish