Created attachment 572648 [details] html_location_in_svg_in_html.xhtml User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/535.7 (KHTML, like Gecko) Chrome/16.0.912.21 Safari/535.7 Steps to reproduce: Embedded XHTML inside a <foreignObject> inside SVG and asked for the getBoundingClientRect() of an HTML element therein. Actual results: Received coordinates of the element within the embedded `<body>` inside the SVG document, not affected by the scaling or placement of the `<foreignObject>` inside the SVG container. Expected results: Received coordinates relative to the viewport of the browser window.
View the test file hosted here: http://phrogz.net/SVG/html_location_in_svg_in_html.xhtml The actual application was the desire to find the location in SVG coordinate space of embedded XHTML contents, in order to connect SVG elements to the embedded HTML contents, as seen here: http://phrogz.net/svg/html_location_in_svg.svg The proposed technique is to translate the embedded HTML element coordinates into screen space (using getBoundingClientRect()) and then transform those coordinates into a local SVG space. This works as intended on Chrome/Safari.
This is the behavior the spec calls for, as far as I can tell. In particular, http://dev.w3.org/csswg/cssom-view/#the-getclientrects-and-getboundingclientrect-methods says: The getClientRects() and getBoundingClientRect() methods provide information about the position of the border box edges of an element relative to the viewport. and http://dev.w3.org/csswg/cssom-view/#viewport says: Viewport and initial containing block are defined by CSS 2.1 unless there is an ancestor foreignObject element in the http://www.w3.org/2000/svg namespace in which case that element acts as viewport and initial containing block.
Or in other words, the return value for getBoundingClientRect in a <foreignObject> is in the coordinate system of that <foreignObject>; you can then use its CTM to do whatever else you want as described in comment 1.
Wow…right you are. I totally missed that definition of viewport. Looks like it's Webkit at fault here. :)
Gavin, are you going to file a bug on WebKit, or should I?
Boris: I am planning on filing a bug. I think I could make a slightly clearer use case/repro case for them first.
Sounds good. Thanks!
FWIW the corresponding Webkit bug is https://bugs.webkit.org/show_bug.cgi?id=71819
Note also http://lists.w3.org/Archives/Public/www-style/2011Nov/0629.html Depending on the outcome of that, the spec (and our behavior) might change... If you have thoughts on the issue, it'd be good to follow up on that mailing list.