803124 - Implement isPointInStroke

Status: RESOLVED FIXED

(Core :: Graphics: Canvas2D, enhancement)

Description

[Tobias Schneider [:tobytailor]]

Attachment: Implement mozIsPointInStroke

In addition to isPointInPath, the Canvas API should also provide isPointInStroke for hit testing. Right now there is no efficient way to test if a point lies inside the filled AND stroked region of a path. (except of checking for nontransparend pixels using getImageData, which is awful slow if 2D rendering is hw accelerated).

Comment 1

[Tobias Schneider [:tobytailor]]

Attachment: Patch v2

Comment 2

[Tobias Schneider [:tobytailor]]

Attachment: Patch v3

Comment 3

[Tobias Schneider [:tobytailor]]

Attachment: Patch v4

Comment 4

[Tobias Schneider [:tobytailor]]

Attachment: Patch v5

Comment 5

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 673128 [details] [diff] [review]
Patch v5

Review of attachment 673128 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/canvas/test/test_mozIsPointInStroke.html
@@ +33,5 @@
> +
> +ok(ctx.mozIsPointInStroke(20, 20) === false);
> +ok(ctx.mozIsPointInStroke(80, 20) === false);
> +ok(ctx.mozIsPointInStroke(80, 30) === false);
> +ok(ctx.mozIsPointInStroke(20, 30) === false);

This test should include curves and transformed paths/transformed userspace. It might also be worth testing different cap modes.

Comment 6

[Tobias Schneider [:tobytailor]]

Attachment: Patch v6

Comment 7

[Jeff Muizelaar [:jrmuizel]]

Comment on attachment 675015 [details] [diff] [review]
Patch v6

Review of attachment 675015 [details] [diff] [review]:
-----------------------------------------------------------------

::: content/canvas/test/test_mozIsPointInStroke.html
@@ +191,5 @@
> +ok(ctx.mozIsPointInStroke(14, 15) === true, 'ctx.mozIsPointInStroke(14, 15) === true');
> +ok(ctx.mozIsPointInStroke(81, 15) === true, 'ctx.mozIsPointInStroke(81, 15) === true');
> +ok(ctx.mozIsPointInStroke(79, 41) === true, 'ctx.mozIsPointInStroke(79, 41) === true');
> +ok(ctx.mozIsPointInStroke(73, 21) === true, 'ctx.mozIsPointInStroke(73, 21) === true');
> +

How about a final test that does a curve scaled by 3x?

Also you should test changing the scale during path construction.

::: gfx/2d/PathSkia.cpp
@@ +184,5 @@
> +                    int32_t(SkFloatToScalar(transformed.y + 1)));
> +
> +  SkRegion pathRegion;
> +
> +  return pathRegion.setPath(strokePath, pointRect);

This should use SkPathContainsPoint if possible.

Comment 8

[Tobias Schneider [:tobytailor]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
> 
> How about a final test that does a curve scaled by 3x?
>
> Also you should test changing the scale during path construction.

Sounds reasonable, will add those tests.

> > +  return pathRegion.setPath(strokePath, pointRect);
> 
> This should use SkPathContainsPoint if possible.

Don't think SkPathContainsPoint is a standard Skia API, is it? Seems to be an internal helper defined and used in Chromium (http://code.google.com/p/chromium/source/search?q=SkPathContainsPoint&origq=SkPathContainsPoint&btnG=Search+Trunk) and itself uses setPath().

Comment 9

[Jeff Muizelaar [:jrmuizel]]

(In reply to Tobias Schneider from comment #8)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
>
> Don't think SkPathContainsPoint is a standard Skia API, is it? Seems to be
> an internal helper defined and used in Chromium
> (http://code.google.com/p/chromium/source/
> search?q=SkPathContainsPoint&origq=SkPathContainsPoint&btnG=Search+Trunk)
> and itself uses setPath().

True. I thought they had fixed this in Skia.

Comment 10

[Jeff Muizelaar [:jrmuizel]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #9)
> (In reply to Tobias Schneider from comment #8)
> > (In reply to Jeff Muizelaar [:jrmuizel] from comment #7)
> >
> > Don't think SkPathContainsPoint is a standard Skia API, is it? Seems to be
> > an internal helper defined and used in Chromium
> > (http://code.google.com/p/chromium/source/
> > search?q=SkPathContainsPoint&origq=SkPathContainsPoint&btnG=Search+Trunk)
> > and itself uses setPath().
> 
> True. I thought they had fixed this in Skia.

Actually, the function is SkPath::contains(x,y)

Comment 11

[Tobias Schneider [:tobytailor]]

Attachment: Patch v7

Comment 12

[Ryan VanderMeulen [:RyanVM]]

Pushed to Try since I don't see any results posted here.
https://tbpl.mozilla.org/?tree=Try&rev=079dc1ebac54

Comment 13

[Ryan VanderMeulen [:RyanVM]]

Unfortunately, this causes hangs in android armv6 builds.
https://tbpl.mozilla.org/?tree=Try&rev=7f236ae2ee94

Comment 14

[Jeff Muizelaar [:jrmuizel]]

I wonder if it still hangs on armv6 without the test changes. I find it pretty unlikely that this would cause all the mochitests to fail...

Comment 15

[Tobias Schneider [:tobytailor]]

Didn't happen in https://tbpl.mozilla.org/?tree=Try&rev=8cab7ad6c0bf.

Comment 16

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/69c7ee7683e0

Comment 17

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/69c7ee7683e0

Comment 18

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

Was this ever even suggested on a standards list?

Comment 19

[Jeff Muizelaar [:jrmuizel]]

(In reply to :Ms2ger from comment #18)
> Was this ever even suggested on a standards list?

I don't believe it was. Would you be willing to do so?

Comment 20

[:Ms2ger (he/him; ⌚ UTC+1/+2)]

I have no idea what this API is good for; it's probably better to let someone who does suggest it.

Comment 21

[Jeff Muizelaar [:jrmuizel]]

http://lists.whatwg.org/pipermail/whatwg-whatwg.org/2012-November/037911.html

Comment 22

[Hixie (not reading bugmail)]

There's no need for this to be prefixed, by the way. It'll either be added, in which case I'll ensure it's compatible and so it won't conflict, or it won't, in which case nothing with this name will be added and it still won't conflict.

The idea seems pretty sound, the main thing that would stop me from adding it to the spec right now is lack of interest from other vendors. I'll ask around, but if you can convince another vendor to indicate an interest that would certainly help.

Comment 23

[Henri Sivonen (:hsivonen)]

Can this be unprefixed, please? As Hixie says, the prefix is useless and unprefixing later will cause breakage later.

Comment 24

[Tobias Schneider [:tobytailor]]

Attachment: Patch to remove prefix

Comment 25

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/eed5cbd0b622

Comment 26

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/eed5cbd0b622

Comment 27

[Masatoshi Kimura [:emk]]

Comment on attachment 684393 [details] [diff] [review]
Patch to remove prefix

[Approval Request Comment]
Bug caused by (feature/regressing bug #): N/A
User impact if declined: proliferating more -moz prefixes
Testing completed (on m-c, etc.): on m-c
Risk to taking this patch (and alternatives if risky): maybe low, simple renaming
String or UUID changes made by this patch: none

Comment 28

[bhavana bajaj [:bajaj]]

Comment on attachment 684393 [details] [diff] [review]
Patch to remove prefix

Approving based on the m-c testing.

Comment 29

[Masatoshi Kimura [:emk]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/a55bc8cb8a8e

Comment 30

[Florian Scholz (Open Web Docs)]

Documented:
https://developer.mozilla.org/en-US/docs/DOM/CanvasRenderingContext2D#isPointInStroke%28%29
https://developer.mozilla.org/en-US/docs/Firefox_19_for_developers#DOM