Evaluated Expression Position Not Actual Position

I have a bunch of animation presets that automatically move layers into and out of the comp frame based on the in and out points of the layer. I have created about 50 of these. I took one of them that you can apply to any 2D text layer with any font size and any scale and fly the layer up from the bottom, bounce to a stop, then slide off to the left. I modified the expression so that it would look at a text layer above and compensate for a new layer moving into position. I only spent about 15 minutes on the code, it's not perfect but if you start with an empty text layer as layer 1 and you add additional text layers below and stager their in-point, you'll get a series of text layers that animate in and adjust positions making room for every new layer as they to. 

Here's a screenshot:

This is the expression:

/*
Add an empty text layer above the other text layers and position it at the bottom center of the composition
you can use this expression for the empty text layer to automatically set the layer's position':

	[thisComp.width/2, thisComp.height]
	
*/
sif = effect("slideIn Frames")("Slider");
sof= effect("slideOut Frames")("Slider");
// find in and out pointOfInterest
stime = time - inPoint;
inTime = sif * thisComp.frameDuration;
	if (effect("Match SI SO")("Checkbox") == 0) {
	outTime = sof * thisComp.frameDuration;
	}
	else if (effect("Match SI SO")("Checkbox") == 1) {
	outTime = inTime;
	}
// Find layer size: 
w = thisLayer.sourceRectAtTime().width;
h =thisLayer.sourceRectAtTime().height;w
lyrSize = [w, h];
// Fix  layer size when scaled
sf = scale - [100, 100];
xSize = lyrSize[0] + (lyrSize[0] * sf[0]/100);
ySize = lyrSize[1] + (lyrSize[1] * sf[1]/100);
realSize = [xSize, ySize];

// Get Positions
txtHeight = text.sourceText.style.fontSize * .7;
xRightEdge = 0 - realSize[0] + realSize[0]/2 ;
yTop = height - realSize[1];
rstx = value [0];
rsty = value[1];
epy = thisComp.height + realSize[1] - realSize[1]/2;
startY = yTop + realSize[1] + txtHeight;

// create movement 
freq = effect("bounce Frequency")("Slider");
        amplitude = effect("bounce Amplitude")("Slider");
        decay = effect("bounce Decay")("Slider");
    posCos = Math.abs(Math.cos(freq*time*2*Math.PI));
	nt = time-inPoint;
    y = amplitude*posCos/Math.exp(decay*nt);
yMove = Math.min(linear(stime,0,inTime, startY , y + value[1]));
restPos = [value[0], yMove] 


// check for layer above this one;
topLayer = thisComp.layer(index - 1).position[1];
posOffset = startY - topLayer;


if (posOffset - txtHeight < restPos[1] - ySize)
	restPos
else
	[value[0], topLayer - ySize - txtHeight]

As I said, it's not perfect but it's a starting point. To perfect it I need to factor in the size of the text in the layer above and compensate for the comp center or less being the minimum y value for text layer position.

Here's the animation preset for you to play with: https://www.dropbox.com/s/r8yol0zl28mv1gk/bump%20up%20text%20layers.ffx?dl=0

All you have to do to get started is create a new comp, add an empty text layer positioned at the comp center and bottom, then add an additional text layer with as many lines of text as you want at the default position with no baseline shift. Apply the animation preset, adust the timing and bounce with the sliders, duplicate the layers as many times as you want, then enter your new text and adjust the in points.

I don't know if this helps or not. If I get the time in the next couple of days to fix the bugs in the expression I'll post an update. Dan Ebberts would write some much cleaner code, but I'm not Dan. Maybe he will jump into this thread.

Here's a project file for you: https://www.dropbox.com/s/h862mhg5a42a0g5/bump%20text%20layers%20cc%20%2816.x%29.aep?dl=0

(note: If your browser adds a .txt extension to the AEP file, just delete it)

I appreciate the feedback, but ultimately I think the issue I'm facing is a bug within the expression engine. 

Issues with bounding box calclation were my first thought (hence the various debug layers that are reporting calculated values in my sample project). And I've worked through a lot of this. 

The fundamental variables I've landed on are that with zero changes to any of the expressions, it works exactly as I expect if I make each layer instance of the Message Comp point at a separate copy of the composition.

As it is, with each message layer being an instance of a single Message composition, all the height height and position calculations are correct and work as expected. But the layers don't render in the right place when each is an instance of the same composition. But only when the position is set by expression - if I disable the expression and manually set the position (using the same values that the expression reports) it renders where it should.

That's what I'm trying it illustrate in my screenshots - the expression evaluated position (356.0, 591.9) renders one layer too low, but disabling the expression and entering exactly the same value renders it in a different place - where it should be. 

Maybe this will help. The following expression will put the anchor point at the bottom of any number of lines of text.

h =thisLayer.sourceRectAtTime(t = time, includeExtents = false).height;
pt = text.sourceText.style.fontSize * .7;
[value[0], h - pt]

The multiplier is a good value for most font sizes because font designers typically fill up about 70% of the allotted vertical space with a T.  IOW if the font size in pixels is 100, the height of a "T" will typically be pretty close to 70 pixels. The sourceRectAtTime() will increase if there are descenders so you may have to take that into account in your design. You may have to adjust the multiplier for different fonts, but .7 works for most common fonts.

Starting with that expression for a text layer Anchor Point should give you a way to space layers apart a uniform distance no matter what their position, font size, or the number of lines. 

The math will be different if you create a text area for your text layer, but the idea will still hold. Get the Anchor Point of a layer at the bottom of the bottom line of characters, know the height of the text, and you should be able to easily separate layers a given number of pixels no matter what the font size or the number of lines.

Too much text! Start with sharing the Expressions you are using AND what you are trying to achieve.

Most likely you're messing up things with your many duplicate auxiliary layers and how the expressions are strung together. From past experience this would be a pretty good guess. So the answer likely lies in changing the calculation method of the bounding box to actually force AE to correctly update the position.

Mylenium