34344 2010-01-29 13:54:19 -0800 Adding overflow:hidden to TD elements makes hit testing much slower 2013-05-10 05:59:12 -0700 1 1 1 Unclassified WebKit WebCore Misc. 528+ (Nightly build) All All NEW P2 Normal --- 1 jamesr webkit-unassigned hyatt simon.fraser tonikitoo oldest_to_newest 185952 0 47733 jamesr 2010-01-29 13:54:19 -0800 Created attachment 47733 Test page using elementFromPoint Setting overflow:hidden on <td> elements inside a large table increases the amount of time needed to hit test points within the table. This is just a special case of https://bugs.webkit.org/show_bug.cgi?id=33520 - setting overflow:hidden gives each <td> its own RenderLayer, so hit testing within the element causes a lot of very slow coordinate maps. The other issue is that currently when hit testing inside a table we iterate through every child cell instead of taking advantage of rows to restrict the search. This shows up as a 30-50ms delay on an internal app with a large table embedded at a moderate depth inside the DOM (roughly 10 RenderLayers from the document root). 188358 1 jamesr 2010-02-05 16:19:50 -0800 overflow:hidden causes each TD to get its own layer. The patch in https://bugs.webkit.org/show_bug.cgi?id=33520 speeds hit testing up 188359 2 jamesr 2010-02-05 16:20:55 -0800 Pah, bugzilla was a bit overeager. What I meant was: overflow:hidden causes each TD to get its own RenderLayer. The patch in https://bugs.webkit.org/show_bug.cgi?id=33520 speeds this test up by about 33%. 47733 2010-01-29 13:54:19 -0800 2010-01-29 13:54:19 -0800 Test page using elementFromPoint td_overflow.html text/html 1482 jamesr PCFET0NUWVBFIGh0bWw+DQo8c3R5bGUgaWQ9J3MxJz4NCi5jMCB7IG92ZXJmbG93OiB9DQouYzEg eyBvdmVyZmxvdzpoaWRkZW4gfQ0KLmMyIHsgb3ZlcmZsb3c6dmlzaWJsZSB9DQouYzMgeyBvdmVy ZmxvdzpzY3JvbGwgfQ0KLmM0IHsgb3ZlcmZsb3c6YWx3YXlzIH0NCjwvc3R5bGU+DQo8c3R5bGU+ DQp0YWJsZSB7IGJvcmRlci1zcGFjaW5nOiA1MHB4IH0NCjwvc3R5bGU+DQo8c2VsZWN0IG9uY2hh bmdlPSJqYXZhc2NyaXB0OnNlbGVjdENoYW5nZWQodGhpcykiPg0KICA8b3B0aW9uPm92ZXJmbG93 Ojwvb3B0aW9uPg0KICA8b3B0aW9uPm92ZXJmbG93OmhpZGRlbjwvb3B0aW9uPg0KICA8b3B0aW9u Pm92ZXJmbG93OnZpc2libGU8L29wdGlvbj4NCiAgPG9wdGlvbj5vdmVyZmxvdzpzY3JvbGw8L29w dGlvbj4NCiAgPG9wdGlvbj5vdmVyZmxvdzphbHdheXM8L29wdGlvbj4NCjwvc2VsZWN0Pg0KPGRp diBvbmNsaWNrPSJqYXZhc2NyaXB0OmRvSGl0VGVzdCgpIj5jbGljayBoZXJlIHRvIGRvIDEwMCBl bGVtZW50RnJvbVBvaW50KDUwMCw1MDApIGNhbGxzPC9kaXY+DQoNCjx0YWJsZSBpZD0ndHJvb3Qn PjwvdGFibGU+DQo8L2Rpdj4NCjxzY3JpcHQ+DQpmdW5jdGlvbiBidWlsZFRhYmxlKHN6KSB7DQog IHZhciByID0gZG9jdW1lbnQuZ2V0RWxlbWVudEJ5SWQoJ3Ryb290Jyk7DQogIGZvciAodmFyIGk9 MDsgaTxzejsgKytpKSB7DQogICAgdmFyIHRyID0gZG9jdW1lbnQuY3JlYXRlRWxlbWVudCgndHIn KTsNCiAgICBmb3IgKHZhciBqPTA7IGo8c3o7ICsraikgew0KICAgICAgdmFyIHRkID0gZG9jdW1l bnQuY3JlYXRlRWxlbWVudCgndGQnKTsNCiAgICAgIHRkLmlubmVyVGV4dCA9ICcuLi4nOw0KICAg ICAgdHIuYXBwZW5kQ2hpbGQodGQpOw0KICAgIH0NCiAgICByLmFwcGVuZENoaWxkKHRyKTsNCiAg fQ0KfQ0KDQpmdW5jdGlvbiBzZWxlY3RDaGFuZ2VkKHMpIHsNCiAgdmFyIHRkcyA9IGRvY3VtZW50 LmdldEVsZW1lbnRzQnlUYWdOYW1lKCd0ZCcpOw0KICB2YXIgdGRzQyA9IFtdOw0KICBmb3IgKHZh ciBpPTA7IGk8dGRzLmxlbmd0aDsgKytpKSB7DQogICB0ZHNDLnB1c2godGRzLml0ZW0oaSkpOw0K ICB9DQogIGZvciAodmFyIGk9MDsgaTx0ZHNDLmxlbmd0aDsgKytpKSB7DQogICAgdGRzQ1tpXS5j bGFzc05hbWUgPSAiYyIrcy5zZWxlY3RlZEluZGV4Ow0KICB9DQp9DQoNCmZ1bmN0aW9uIGRvSGl0 VGVzdCgpIHsNCiAgZG9jdW1lbnQuYm9keS5vZmZzZXRUb3A7DQogIHZhciBzdGFydCA9IG5ldyBE YXRlKCkuZ2V0VGltZSgpOw0KICBmb3IgKHZhciBpID0gMDsgaTwxMDA7ICsraSkgew0KICAgIGRv Y3VtZW50LmVsZW1lbnRGcm9tUG9pbnQoNTAwLDUwMCk7DQogIH0NCiAgd2luZG93LmFsZXJ0KCJU b29rICIgKyAobmV3IERhdGUoKS5nZXRUaW1lKCkgLSBzdGFydCkgKyAibXMgZm9yIDEwMCBsb29r dXBzIGF0ICg1MDAsNTAwKSIpOw0KfQ0KDQpidWlsZFRhYmxlKDEwMCk7DQo8L3NjcmlwdD4NCg0K