How is the getBBox() SVGRect calculated?

Question

I have a g element that contains one or more path elements. As I mentioned in another question, I scale and translate the g element by computing a transform attribute so that it fits on a grid in another part of the canvas.

The calculation is done using the difference between two rectangles, the getBBox() from the g element and the rectangle around the grid.

Here is the question -- after I do the transform, I update the contents of the g element and call getBBox() again, without removing the transform. The resulting rectangle appears to be calculated without considering the transform. I would have expected it to reflect the change. Is this behavior consistent with the SVG specification? How do I get the bounding box of the transformed rectangle?

This, BTW, is in an HTML 5 document running in Firefox 4, if that makes any difference.

Update: Apparently this behavior seems pretty clearly in violation of the specification. From the text here at w3c:

SVGRect getBBox()

Returns the tight bounding box in current user space (i.e., after application of the ‘transform’ attribute, if any) on the geometry of all contained graphics elements, exclusive of stroking, clipping, masking and filter effects). Note that getBBox must return the actual bounding box at the time the method was called, even in case the element has not yet been rendered.

Am I reading this correctly? If so this seems to be an errata in the SVG implementation Firefox uses; I haven't had a chance to try any other. I would file a bug report if someone could point me to where.

Answer 1

People often get confused by the behavioral difference of getBBox and getBoundingClientRect.

getBBox is a SVG Element's native method as equivalent to find the offset/clientwidth of HTML DOM element. The width and height is never going to change even when the element is rotated. It cannot be used for HTML DOM Elements.

getBoundingClientRect is common to both HTML and SVG elements. The bounded rectangle width and height will change when the element is rotated or when more elements are grouped.

Answer 2

The behaviour you see is correct, and consistent with the spec. The transform gets applied, then the bbox is calculated in "current user units", i.e. the current user space. So if you want to see the result of a transform on the element you'd need to look at the bbox of a parent node or similar. It's a bit confusing, but explained a lot better in the SVG Tiny 1.2 spec for SVGLocatable That contains a number of examples that clarify what it's supposed to do.

Answer 3

there are at least 2 easy but somewhat hacky ways to do what you ask... if there are nicer (less hacky) ways, i haven't found them yet

EASY HACKy #1:
a) set up a rect that matches the "untransformed" bbox that group.getBBox() is returning
b) apply the group's "unapplied transform" to that rect
c) rect.getBBox() should now return the bbox you're looking for

EASY HACKY #2: (only tested in chrome)
a) use element.getBoundingClientRect(), which returns enough info for you to construct the bbox you're looking for

Answer 4

Apparently getBBox() doesn't take the transformations into consideration.

I can point you here, unfortunately I wasn't able to make it working: http://tech.groups.yahoo.com/group/svg-developers/message/22891

Answer 5

The following code takes into account the transformations (matrix or otherwise) from parents, itself, as well as children. So, it will work on a <g> element for example.

You will normally want to pass the parent <svg> as the third argument—toElement—as to return the computed bounding box in the coordinate space of the <svg> (which is generally the coordinate space we care about).

/**
 * @param {SVGElement} element - Element to get the bounding box for
 * @param {boolean} [withoutTransforms=false] - If true, transforms will not be calculated
 * @param {SVGElement} [toElement] - Element to calculate bounding box relative to
 * @returns {SVGRect} Coordinates and dimensions of the real bounding box
 */
function getBBox(element, withoutTransforms, toElement) {

  var svg = element.ownerSVGElement;

  if (!svg) {
    return { x: 0, y: 0, cx: 0, cy: 0, width: 0, height: 0 };
  }

  var r = element.getBBox(); 

  if (withoutTransforms) {
    return {
      x: r.x,
      y: r.y,
      width: r.width,
      height: r.height,        
      cx: r.x + r.width / 2,
      cy: r.y + r.height / 2
    };
  }

  var p = svg.createSVGPoint(); 

  var matrix = (toElement || svg).getScreenCTM().inverse().multiply(element.getScreenCTM()); 

  p.x = r.x;
  p.y = r.y;
  var a = p.matrixTransform(matrix);

  p.x = r.x + r.width;
  p.y = r.y;
  var b = p.matrixTransform(matrix);

  p.x = r.x + r.width;
  p.y = r.y + r.height;
  var c = p.matrixTransform(matrix);

  p.x = r.x;
  p.y = r.y + r.height;
  var d = p.matrixTransform(matrix);

  var minX = Math.min(a.x, b.x, c.x, d.x);
  var maxX = Math.max(a.x, b.x, c.x, d.x);
  var minY = Math.min(a.y, b.y, c.y, d.y);
  var maxY = Math.max(a.y, b.y, c.y, d.y);

  var width = maxX - minX;
  var height = maxY - minY;

  return {
    x: minX,
    y: minY,
    width: width,
    height: height,        
    cx: minX + width / 2,
    cy: minY + height / 2
  };
}

Answer 6

SVG groups have nasty practice - not to accumulate all transformations made. I have my way to cope with this issue. I'm using my own attributes to store current transformation data which I include in any further transformation. Use XML compatible attributes like alttext, value, name....or just x and y for storing accumulated value as atribute.

Example:

<g id="group" x="20" y="100" transform="translate(20, 100)">
<g id="subgroup" alttext="45" transform="rotate(45)">
<line...etc...

Therefore when I'm making transformations I'm taking those handmade attribute values, and when writing it back, I'm writing both transform and same value with attributes I made just for keeping all accumulated values.

Example for rotation:

function symbRot(evt) {

  evt.target.ondblclick = function () {

    stopBlur();
    var ptx=symbG.parentNode.lastChild.getAttribute("cx");
    var pty=symbG.parentNode.lastChild.getAttribute("cy");
    var currRot=symbG.getAttributeNS(null, "alttext");

    var rotAng;
    if (currRot == 0) {
      rotAng = 90
    } else if (currRot == 90) {
      rotAng = 180
    } else if (currRot == 180) {
      rotAng = 270
    } else if (currRot == 270) {
      rotAng = 0
    };
    symbG.setAttributeNS(null, "transform", "rotate(" + rotAng + "," + ptx + ", " + pty + ")");
    symbG.setAttributeNS(null, "alttext", rotAng );
  };
}

Answer 7

I made a helper function, which returns various metrics of svg element (also bbox of transformed element).

The code is here:

SVGElement.prototype.getTransformToElement = 
SVGElement.prototype.getTransformToElement || function(elem) { 
  return elem.getScreenCTM().inverse().multiply(this.getScreenCTM()); 
};

function get_metrics(el) {
    function pointToLineDist(A, B, P) {
        var nL = Math.sqrt((B.x - A.x) * (B.x - A.x) + (B.y - A.y) * (B.y - A.y));
        return Math.abs((P.x - A.x) * (B.y - A.y) - (P.y - A.y) * (B.x - A.x)) / nL;
    }

    function dist(point1, point2) {
        var xs = 0,
            ys = 0;
        xs = point2.x - point1.x;
        xs = xs * xs;
        ys = point2.y - point1.y;
        ys = ys * ys;
        return Math.sqrt(xs + ys);
    }

    var b = el.getBBox(),
        objDOM = el,
        svgDOM = objDOM.ownerSVGElement;
    // Get the local to global matrix
    var matrix = svgDOM.getTransformToElement(objDOM).inverse(),
        oldp = [[b.x, b.y], [b.x + b.width, b.y], [b.x + b.width, b.y + b.height], [b.x, b.y + b.height]],
        pt, newp = [],
        obj = {},
        i, pos = Number.POSITIVE_INFINITY,
        neg = Number.NEGATIVE_INFINITY,
        minX = pos,
        minY = pos,
        maxX = neg,
        maxY = neg;

    for (i = 0; i < 4; i++) {
        pt = svgDOM.createSVGPoint();
        pt.x = oldp[i][0];
        pt.y = oldp[i][1];
        newp[i] = pt.matrixTransform(matrix);
        if (newp[i].x < minX) minX = newp[i].x;
        if (newp[i].y < minY) minY = newp[i].y;
        if (newp[i].x > maxX) maxX = newp[i].x;
        if (newp[i].y > maxY) maxY = newp[i].y;
    }

    // The next refers to the transformed object itself, not bbox
    // newp[0] - newp[3] are the transformed object's corner
    // points in clockwise order starting from top left corner
    obj.newp = newp; // array of corner points
    obj.width = pointToLineDist(newp[1], newp[2], newp[0]) || 0;
    obj.height = pointToLineDist(newp[2], newp[3], newp[0]) || 0;
    obj.toplen = dist(newp[0], newp[1]);
    obj.rightlen = dist(newp[1], newp[2]);
    obj.bottomlen = dist(newp[2], newp[3]);
    obj.leftlen = dist(newp[3], newp[0]);
    // The next refers to the transformed object's bounding box
    obj.BBx = minX;
    obj.BBy = minY;
    obj.BBx2 = maxX;
    obj.BBy2 = maxY;
    obj.BBwidth = maxX - minX;
    obj.BBheight = maxY - minY;
    return obj;
}

and full functional example is here: http://jsbin.com/acowaq/1