I don't really understand all the code;I 'm just guessing. The code runs reliably for point with integers coords, so afterall we are not so far away ? I thought the problem could come from the IEEE754 function ( which is a totaly brand new universe for me ). Can you confirm the problem doesn't come from this function ? By @Panchromatic For example, when using Points units in a ruler, the coordinates of pixels are still fractional, despite being integer in pixels. This function (IEEE75) is for writing numbers in float mode (32 bits) as they are stored in memory in machine code. The coordinates of path points are stored (I think so) in double mode (64 bits). Perhaps due to the difference in accuracy, the operation does not work accurately for many points and selection does not occur. I chose this operation because it contains the minimum amount of data. I calculated the click coordinates empirically and the fact that they are stored in float mode. There still remains some unknown data, the meaning of which I have not yet discovered and how it affects execution. Even if you fully understand how everything works there, it will not work reliably. Example. I only have Photoshop CS6 for now. You can check it on your version as well. Create an Action where you record a click on your point, which is not highlighted by the script or any other point on the path. Unhighlight the path and run the Action. Very often the action will not work, i.e. will not select the point/path. It also depends on the zoom of the document. And since the Action works unreliably, it is stupid to try to reproduce it in a script, even knowing all the nuances of the data structure. If you record a small square selection around a point in the Action using the PathSelection tool instead of point clicking, it will work. But at the same time, the amount of data will be gigantic. It's impossible to figure it out on your own. Moreover, execution will be very slow. So forget about it. )
... View more