How to get an annotation's position ?

Okey so this is first script on this forum that does such thing! Regarding reverted coordinates they were displayed right way, so like written in binary's. I said that I changed their order but true is I had but forgot, and to your post I was sure I did so Now it works as it should. That's not only problem with annotations, the other order of coordinates, I noticed it also in other parts like in path(Item)s. But that's other story.

In current 13th line I changed 28 value to 12. That caused (re)moved notes positions could still be read, but that happened in some not specified cases like only 1 note or some custom. Originally I used 28 value as it seemed range of binary characters can't be match of regular expresion's.

I put resultant coordinates into array. They are alerted the sama way, but with chance to match one of them by N[0/*1*/].

I tested it again. Now it works a little slower (I checked old & new script for the same document). Then it was about 225 ms, now 250). For 1*1*1 document it was 130 ms with one note, but with 100 of them it was not slower for any kind of document, however when I tried it with written annotations, many layer(Sets), channels, paths and big dimmension & resolution it was slower of course, but not much for some reason, just 350 ms ?!

Putting note vertically/horizontally under 0, that is on top/left side of document will give 4294967278, so Math.pow(256, 4)!

function bin(v) {function anno(v) {
          var wh = v.match(/(.{4})(.{4})$/)
          for(pos = '', i = wh.length - 1; i > 0; i--) {
               for(var h = '', b = wh, j = 0; j < b.length; j++) {
                    if (b) h += b.charCodeAt().toString(16)
               }
               pos += eval('0x' + h) + ', '
          }
          return pos.slice(0, -2)
     }
     notes = [], v.open('r'), v.encoding = 'binary'
     var arr = v.read().match(/txtA.{12}/g); v.close()
     for(k = 0; k < arr.length; k++) notes.push(anno(arr))
     return notes
}
function sTT(v) {return stringIDToTypeID(v)} function uN() {executeAction(sTT('undoEvent'))}
function dN() {
     (ref1 = new ActionReference()).putEnumerated(sTT('annotation'),
     sTT('ordinal'), sTT('targetEnum')); (dsc1 = new ActionDescriptor())
     .putReference(sTT('null'), ref1), executeAction(sTT('delete'), dsc1, DialogModes.NO)
}
function pHS(v) {
     (ref1 = new ActionReference()).putEnumerated(sTT('historyState'),
     sTT('ordinal'), sTT('previous')); (dsc1 = new ActionDescriptor()).putReference
     (sTT('null'), ref1), executeAction(sTT('select'), dsc1, DialogModes.NO)
}
function dHS() {
     (ref1 = new ActionReference()).putProperty(sTT('historyState'),
     sTT('currentHistoryState')); (dsc1 = new ActionDescriptor())
     .putReference(sTT('null'), ref1), executeAction(sTT('delete'), dsc1, DialogModes.NO)
}
with(psd = new PhotoshopSaveOptions()) {
     annotations = !(alphaChannels = embedColorProfile = layers = spotColors = false)
}
if ($.level = 0, documents.length) {try{
          (aD = activeDocument).suspendHistory('Notes', '')
          for(dN(), pHS(), m = l = 0; l < (Nn = ['N', 'n']).length; l++) {
               aD.saveAs(fle = File('~/desktop/.psd'), psd, true)
               if (!l || N.length > 1) eval(Nn + ' = bin(fle)')
               !l ? dN() : fle.remove()
          }
          dHS(); while(typeof n != 'undefined' && N == n) m++; alert(N.split(','))
     }
     catch(err) {uN(), alert('No selected annotation!')}
}

A regex I thought to use for over 8 bits depth documents wouldn't be probably different than I used. I found problem is with sole writing/reading 16 and 32 bits deepth documents with .txt file I had to use to study binary record. With annotations that would not be problem to change deepth to 8 bits for time of script work, but if we had to read binary for other goals for not only 8 bits document then I have no idea how we could do it. I wonder why for so many years noone met this problem on Adobe forums? Once Michael L Hale mentioned "out of memory" issue. I tried to do the same with .tif which can store annotations information but some lines covering bitmap data were too long comparing with psd to be read. Any ideas?