Talk:LlDetectedTouchUV

Discussion

This topic was recently discussed on Jira at SVC-8090, here is an excerpt.

Moon Metty added a comment - 06/Aug/12 3:14 PM Yes Spootie, UV is very handy for that kind of menu. :) Strife, maybe this is a start: llDetectedTouchUV() returns the texture coordinates where an object was clicked. The result is the same as llDetectedTouchST() transformed by UV-repeats, -offset and -rotation. Where ST coordinates are always in the interval [0, 1] , UV coordinates are not because of the transformation, Example: When we wrap a UV coordinate in [0, 1] , the result represents a location in the texture. This is the fractional part of llDetectedTouchUV(). The integer part (rounded down for positive and negative numbers) represents which repetition of the texture was clicked. <lsl> touch_start(integer num_detected) { vector UV = llDetectedTouchUV(0); // store the UV coordinates as a vector variable integer repetition_x = llFloor(UV.x); // round the x coordinate down integer repetition_y = llFloor(UV.y); // round the y coordinate down float texture_x = UV.x - repetition_x; // calculate the fractional part of x float texture_y = UV.y - repetition_y; // calculate the fractional part of y } </lsl> Planar mapping has never worked well (see MAINT-667, VWR-4860 or search the jira for "planar") However, no matter how mangled the texture looks, llDetectedTouchUV still appears to return the correct result, which is nice.

Spootie added a comment - 06/Aug/12 11:45 PM Moon's approach that considers DetectedTouchUV as a transformation of DetectedTouchST is an interesting, if sometimes confusing approach, at least for someone who came to DetectedTouchUV looking for a way to determine simply where on a texture one clicks. I came to think of a texture as a map, irregardless of what part of the map was shown on the face. Strife schooled me here on how detecting which of multiple copies might be useful, and now Moon is schooling me on a different, equally valid paradigm. You can see an expression of my different approach on the discussion page, signed by Windy Schor. I do like Moon's LSL example better than my own. Strife's point about future mapping issues makes me lean away from mixing ST into the UV explanation. Another source of confusion is that a positive fractional offset puts us in a >1 repeat. Let's take any further discussion to the page itself, since this clearly isn't a bug. Thank you Moon and Strife for your input.

Offsets and repeats need to be clarified

If someone would please look at the following with an editor's eye, I'd like to add this clarification to the page. Should there also be an example of how offsets can make the results be offset?

llDetectedTouchUV returns the U,V corresponding to the part of the texture touched, plus any U and V index of tiled repeats. For example, if you set two repeats vertically and one horizontally, with no texture-offset and no texture-rotation, clicking in the middle of the upper half yields <1.5,0.5,0.0> and clicking in the exactly same spot on the lower copy of the texture, i.e. the middle of the bottom half, yields <0.5,0.5,0.0>. Thanks to this feature, one can determine not only where on a texture the touch was detected, but on which repeat of the texture, in case that matters.

The position touched on the texture, ignoring repeats and/or determining which repeat, is determined thusly: <lsl> vector DTUV=llDetectedTouchUV(index); vector position_on_texture = <DTUV.x - (float)llFloor(DTUV.x), DTUV.y - (float)llFloor(DTUV.y), 0.0>; list which_repeat = [ llFloor(DTUV.x), llFloor(DTUV.y) ]; </lsl> Windy Schor 18:08, 6 August 2012 (PDT)