uiDrawMatrixScale center

[mischnic]

Are the xCenter and yCenter coordinates relative to the element (as with rotation, right?) or the area?

libui/ui.h

Line 472 in 9cf56e1

_UI_EXTERN void uiDrawMatrixScale(uiDrawMatrix *m, double xCenter, double yCenter, double x, double y);

(I'm guessing it's relative to the area).
I'm trying to scale the rectangle with it's center as the scale center.
Here I drew the rectangle at (50,50) and set the scale center to (50,50). Why does the rectangle move while changing the scaling factor?

[mischnic]

Complete example:

#include <stdio.h>
#include <string.h>
#include "ui.h"

int x = 0;

static void handlerDraw(uiAreaHandler *a, uiArea *area, uiAreaDrawParams *p)
{

	uiDrawBrush brush;
	brush.Type = uiDrawBrushTypeSolid;
	brush.R = 255;
	brush.G = 0;
	brush.B = 0;
	brush.A = 1.0;

	uiDrawMatrix m;
	uiDrawMatrixSetIdentity(&m);
	uiDrawMatrixScale(&m, 50, 50, 0.5+x/300.0, 1);

	uiDrawPath *path = uiDrawNewPath(uiDrawFillModeWinding);
	uiDrawPathAddRectangle(path, 50, 50, 100, 100);
	uiDrawPathEnd(path);
	
	uiDrawTransform(p->Context, &m);

	uiDrawFill(p->Context, path, &brush);
	uiDrawFreePath(path);
}

static void handlerMouseEvent(uiAreaHandler *a, uiArea *area, uiAreaMouseEvent *e) {
	x = e->X;
	uiAreaQueueRedrawAll(area);
}
static void handlerMouseCrossed(uiAreaHandler *ah, uiArea *a, int left) {}
static void handlerDragBroken(uiAreaHandler *ah, uiArea *a) {}
static int handlerKeyEvent(uiAreaHandler *ah, uiArea *a, uiAreaKeyEvent *e) {return 0;}


int main(void)
{
	uiInitOptions o;
	const char *err;

	memset(&o, 0, sizeof (uiInitOptions));
	err = uiInit(&o);
	if (err != NULL) {
		fprintf(stderr, "error initializing ui: %s\n", err);
		uiFreeInitError(err);
		return 1;
	}

	uiWindow *mainwin = uiNewWindow("test", 640, 480, 1);
	uiAreaHandler handler;

	handler.Draw = handlerDraw;
	handler.MouseEvent = handlerMouseEvent;
	handler.MouseCrossed = handlerMouseCrossed;
	handler.DragBroken = handlerDragBroken;
	handler.KeyEvent = handlerKeyEvent;

	uiArea *area = uiNewArea(&handler);

	uiWindowSetChild(mainwin, uiControl(area));

	uiControlShow(uiControl(mainwin));

	uiMain();
	return 0;
}

[andlabs]

I'm not sure what this second example is trying to show but IIRC it was relative to the current (0, 0) in the matrix being applied to...

[mischnic]

I called uiDrawMatrixScale with xCenter = yCenter = 50. (50,50) is also the top left corner of the rectangle. Why is the left side moving? I would expect it to get longer towards the left only.

[mischnic]

it was relative to the current (0, 0) in the matrix being applied to...

So this scale everything and the point (50,50) doesn't move (what I want):

uiDrawMatrixTranslate(&m, 50, 50);
uiDrawMatrixScale(&m, 0, 0, i, i);
uiDrawMatrixTranslate(&m, -50, -50);

This should be equivalent? But it isn't.

uiDrawMatrixScale(&m, 50, 50, i, i);

I think it has something to do with scaleCenter, removing that call makes (xCenter, yCenter) stay on the same position.

diff --git a/darwin/draw.m b/darwin/draw.m
index b4be6e7f..53292521 100644
--- a/darwin/draw.m
+++ b/darwin/draw.m
@@ -329,15 +329,11 @@ void uiDrawMatrixTranslate(uiDrawMatrix *m, double x, double y)
 void uiDrawMatrixScale(uiDrawMatrix *m, double xCenter, double yCenter, double x, double y)
 {
        CGAffineTransform c;
-       double xt, yt;

        m2c(m, &c);
-       xt = x;
-       yt = y;
-       scaleCenter(xCenter, yCenter, &xt, &yt);
-       c = CGAffineTransformTranslate(c, xt, yt);
+       c = CGAffineTransformTranslate(c, xCenter, yCenter);
        c = CGAffineTransformScale(c, x, y);
-       c = CGAffineTransformTranslate(c, -xt, -yt);
+       c = CGAffineTransformTranslate(c, -xCenter, -yCenter);
        c2m(&c, m);
 }

[andlabs]

But does the scaling matrix test in the tester binary still work correctly? Expected output should be https://msdn.microsoft.com/dynimg/IC534042.png or something to that effect (I forget how I set up that test; it is also possible my test code is also wrong in addition to the main code, but then all other platforms would also be wrong)

[mischnic]

It needs to be tested if it's currently wrong on linux as well...
Do you know of how it should look like (are the two rectangles on the left correct)?

Currently/before:

With my change:

[andlabs]

...actually huh. What does it look like on Windows? Can't check right now, but unless they changed that page there's a TODO missing somewhere...

[mischnic]

On Windows: like the second screenshot
On Linux: same situation as on macOS

If I interpret the test code correctly, then the second screenshot is correct.

[andlabs]

So yes, the Linux and macOS implementations were wrong. I wonder why I didn't make a TODO about this... unless I did and forgot.

I now wonder what else is wrong, or if there's a better test suite I can use for this.