405300 - <canvas> isPointInPath doesn't transform its coordinates

Status: VERIFIED FIXED

(Core :: Graphics: Canvas2D, defect)

Description

[Ilmari Heikkinen]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686 (x86_64); fi; rv:1.9b1) Gecko/2007110903 Firefox/3.0b1
Build Identifier: Mozilla/5.0 (X11; U; Linux i686 (x86_64); fi; rv:1.9b1) Gecko/2007110903 Firefox/3.0b1

When using ctx.isPointInPath(x,y), the x,y-coords should be multiplied by the current transformation matrix to have isPointInPath work for transformed paths. 

The WhatWG spec[1] wording is a bit confusing on this matter: 
"The isPointInPath(x, y) method must return true if the point given by the x and y coordinates passed to the method, when treated as coordinates in the canvas' coordinate space unaffected by the current transformation..."

The idea being that the x,y-coordinates are in canvas space and hence must be transformed to the current space to do the isPointInPath-test.

[1] http://www.whatwg.org/specs/web-apps/current-work/multipage/section-the-canvas.html#ispointinpath

Reproducible: Always

Steps to Reproduce:
1. Change canvas transform (translate/rotate/scale.)
2. Use isPointInPath on a point that would have had its return value changed by the transform.

Actual Results:  
The return value didn't change.

Expected Results:  
The return value should've had changed.

Comment 1

[Ilmari Heikkinen]

Attachment: Transform coords to device-space before cairo_in_fill-check

Comment 2

[Anne (:annevk)]

This bug is INVALID as far as I can tell.

Comment 3

[Ilmari Heikkinen]

Opera handles isPointInPath coordinates as untransformed coordinates. See test in URL with Firefox vs. Opera. With setTransform, Firefox's behaviour can be worked around.

Comment 4

[Ilmari Heikkinen]

My original explanation of "how to fix" is wrong. The patch works, but the math is: multiply mouse coords by the inverse of the current transformation matrix.  Which the patch does implicitly (changing the matrix changes the path's matrix as well, so transforming the path to the point's matrix equals to transforming the point to the path's matrix.)

It can be worked around on a web page by using setTransform or by manually keeping track of the current transformation matrix and doing multiply(invert(matrix), [x,y,1]). Latter way + bbox / circle test can then be used to implement a crude version of isPointInPath for browsers that lack it.

Comment 5

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

WONTFIX'ing this as per bug 380338 -- paths don't get transformed, points do.  Changing the transformation matrix has no effect on the path that's currently built up.

Comment 6

[Ilmari Heikkinen]

Attachment: JS workaround to normalize isPointInPath behaviour across browsers

Note that you need to keep track of the current transformation matrix on Firefox 2. Implementation is left to the reader. Cheers!

Comment 7

[Jordan Osete]

Attachment: isPointInPath with transformations testcase

I don't understand why this still is a WONTFIX, but here is a simple testcase of isPointInPath with transformations, and a (very simple) workaround for Mozilla included.

Comment 8

[Jordan Osete]

(Sorry, the circles should highlight when you move your mouse over them in the testcase.)

Comment 9

[Charles Schmidt]

Comment #5 is incorrect. Paths are transformed "IF" there transform is set before the path is drawn. The transform should not be applied retroactively as was being requested in the referenced bug.  As per the specification the isPointInPath method should test the query point "AS IS" again the "transformed" Path.

Please reopen this issue!

Comment 10

[Tim]

Please reopen this, all other modern browsers are implementing correctly and this functionality is critical to making canvas interactive (e.g. is my mouse inside this path after transforms have been applied). 

example

http://philip.html5.org/tests/canvas/suite/tests/index.2d.path.isPointInPath.html

tested on IE9, Chrome, Opera, Safari (iphone and ipad)

Comment 11

[Jordan Osete]

For those struggling with this, a "simple" workaround is to detect the bug, then replace isPointInPath() with a correct one (please excuse the poor formatting, just a copy/paste of old code):


var c2dp = CanvasRenderingContext2D.prototype;

//  special isPointInPath method to workaround Mozilla bug 405300
//      [https://bugzilla.mozilla.org/show_bug.cgi?id=405300]
function isPointInPath_mozilla( x, y )
{
    this.save();
    this.setTransform( 1, 0, 0, 1, 0, 0 );
    var ret = this.isPointInPath_old( x, y );
    this.restore();
    return ret;
};

//  test for the presence of the bug, and set the workaround function only if needed
var ctx = document.createElement( "canvas" ).getContext( "2d" );
ctx.translate( 50, 0 );
ctx.moveTo( 125, 50 );
ctx.arc( 100, 50, 25, 0, 360, false );
if( !ctx.isPointInPath( 150, 50 ) )
{
    c2dp.isPointInPath_old = c2dp.isPointInPath;
    c2dp.isPointInPath = isPointInPath_mozilla;
}

Comment 12

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

We should get either the spec or our implementation changed.

Comment 14

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Relevant tests:

http://www.w3c-test.org/html/tests/approved/canvas/2d.path.isPointInPath.transform.1.html
http://www.w3c-test.org/html/tests/approved/canvas/2d.path.isPointInPath.transform.2.html
http://www.w3c-test.org/html/tests/approved/canvas/2d.path.isPointInPath.transform.3.html
http://samples.msdn.microsoft.com/ietestcenter/html5/canvas_harness.htm?url=canvas-complexShapes-isPointInPath-001

Comment 15

[:Benjamin Peterson]

Attachment: updated patch

This is an updated patch.

Comment 16

[Joe Drew (not getting mail)]

Comment on attachment 533682 [details] [diff] [review]
updated patch

Review of attachment 533682 [details] [diff] [review]:
-----------------------------------------------------------------

What should we do about the spec issue? The spec needs to be changed if everyone else does something different.

Comment 17

[:Benjamin Peterson]

Everyone else is doing something different, but they're all doing the same thing: following the spec.

Comment 18

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 533682 [details] [diff] [review]
updated patch

Review of attachment 533682 [details] [diff] [review]:
-----------------------------------------------------------------

Benjamin's right. This change makes us follow the spec.

Comment 19

[Dão Gottwald [:dao]]

http://hg.mozilla.org/mozilla-central/rev/8a133638f5b6

Comment 20

[Eric Shepherd [:sheppy]]

Documentation updated:

https://developer.mozilla.org/en/DOM/CanvasRenderingContext2D#isPointInPath()

And mentioned on Firefox 7 for developers.

Comment 21

[Mihaela Velimiroviciu (:mihaelav)]

Build identifier: Mozilla/5.0 (X11; Linux x86_64; rv:7.0) Gecko/20100101 Firefox/7.0

All tests mentioned in comment 7 , comment 10 and comment 14 passed. Marking bug as VERIFIED-FIXED.