Optimize jQuery.expr.filters.hidden

[dmethvin]

​http://forum.jquery.com/topic/optimize-jquery-expr-filters-hidden

Wouldn't it make sense to alter the order of the expressions in jQuery.expr.filters.hidden to only do the expensive work (elem.offsetWidth) after evaluating the parts that don't trigger a reflow? In the case the element is determined to be hidden by the style, it does not trigger a reflow anymore, what would be a nice behavior for a hidden element.

So instead of

return ( elem.offsetWidth === 0 && elem.offsetHeight === 0 ) || (!jQuery.support.reliableHiddenOffsets && ((elem.style && elem.style.display) || curCSS( elem, "display" )) === "none");

do

return (!jQuery.support.reliableHiddenOffsets && ((elem.style && elem.style.display) || curCSS( elem, "display" )) === "none") || ( elem.offsetWidth === 0 && elem.offsetHeight === 0 );

Roel

[Timmy Willison]

I'm fine with it, but isn't display calculation a lot more expensive in IE than .offsetWidth/.offsetHeight?

[dmethvin]

I think display just looks at the property for that element so no recalc should be required. The offset properties definitely require a recalc if one hasn't already been done, I'm always seeing big delays on offsetWidth/Height in profile runs for IE.

[r.spilker@…]

Created pull request: ​https://github.com/jquery/jquery/pull/1027

[Timmy Willison]

@dmethvin: I'm wondering, is offsetWidth/Height faster even when the codepath hits defaultDisplay? But even if it is, this is probably a better way to go.

[Timmy Willison]

Will create a perf

[dmethvin]

Based on the perf we decided not to land this.