How to animate tennis balls dropping out of a tennis can.

Revealing the tennis balls is all layering. Depending on your design you may need more than one copy of the can. Parenting can help and an expression that simulates gravity and a bounce can help, but you should not have any need for masks unless the layer that contains the can does not have a transparent background. 

By far the easiest solution is to spend some mone on Newton, and amazing 2D simulation package from aeScripts.com

If you don't have the budget you can modify and expression shared by Dan Ebberts on motionscript.com and set up your comp like this:

The top layer is a null parented to the can and the opening layer that is below the balls. I just created the can with a rectangle and an ellipse, then used another lighter ellipse below the rectangle to give the balls a hole to come out of. 

Here's my modification of Dan Ebbert's expression:

elev = degreesToRadians(86);
v = 2500;
e = .6;
f = .5;
g = 5000;
nMax = 9;
tLaunch = thisLayer.inPoint - .5;
 
vy = v*Math.sin(elev);
vx = v*Math.cos(elev);
if (time >= tLaunch){
  t = time - tLaunch;
  tCur = 0;
  segDur = 2*vy/g;
  tNext = segDur;
  d = 0; // x distance traveled
  nb = 0; // number of bounces
  while (tNext < t && nb <= nMax){
    d += vx*segDur;
    vy *= e;
    vx *= f;
    segDur *= e;
    tCur = tNext;
    tNext += segDur;
    nb++
  }
  if(nb <= nMax){
    delta = t - tCur;
    x = d + delta*vx;
    y = delta*(vy - g*delta/2);
  }else{
    x = d;
    y = 0;
  }
  value + [x,-y]
}else
  value

Right up at the top are the variables that I changed. here is my reasoning:

elev = degreesToRadians(86); // Reset to a very steep angle to match the speed of the moving can

v = 2500; // This is the velocity I needed to get enough elevation in the bounce

e = .6; // I needed a little less bounce

f = .5; // The friction is OK

g = 5000; // The gravity is OK

nMax = 9; The number of bounces is OK

tLaunch = thisLayer.inPoint - .5; // I adjusted the start time to be a half second before layer start so the ball is on it's way down

All that is necessary is to adjust the Y position to match the position of the null at the start of the layer and stagger the copies of the ball layer. 

You could also go to Dan's website and add in an expression to get the balls to roll and come to a stop.

That's the easiest and most realistic solution I can think of without spending any money. If you have a lot of work to do that requires simulations Newton is a very good solution to your problem.