I want to create a donut chart using an SVG circle element by setting stroke-dasharray
and varying stroke-dashoffset
. The SVG element needs to be rotat
Indeed, set the rotate transformation to something like 90.1deg solves the issue...
I have test many things, reported here: https://codepen.io/KevinNTH/pen/ZBgKdG
<!-- workaround ios -->
<svg class="wka-ios">
<g transform="rotate(-90.1 30 30)">
<circle cx="25" cy="25" r="15"/>
</g>
</svg>
I've experienced this painfully on iOS 10.1 and Safari 10.0.1. The bug is definitely triggered by any rotate
value which computes to a value divisible by 90 degrees.
But it gets weirder: the bug's presence is affected by the current zoom level.
See this demo/series of minimal test cases I put together (jsFiddle version here). Best to run the snippet then expand to full page:
svg {
height: 80px;
width: 80px;
}
circle {
fill: none;
stroke-dasharray: 150;
stroke-width: 4px;
stroke: #6fdb6f;
transform-origin: center center;
}
.degrot {
transform: rotate(-90deg);
}
.degrot-offset {
transform: rotate(-90.1deg);
}
.degrot-offset-more {
transform: rotate(-92deg);
}
.turnrot {
transform: rotate(-0.25turn);
}
.turnrot-offset {
transform: rotate(-0.251turn);
}
svg[viewBox] circle {
stroke-dasharray: 300;
stroke-width: 8px;
}
svg[viewBox].scaledown circle {
stroke-dasharray: 300;
stroke-width: 8px;
}
svg[viewBox].noscale circle {
stroke-dasharray: 150;
stroke-width: 4px;
}
svg[viewBox].scaleup circle {
stroke-dasharray: 75;
stroke-width: 2px;
}
.wc {
will-change: transform;
}
/* Demo prettification */
p:last-child {
margin-bottom: 0;
}
td {
padding: 10px;
}
tr td:first-of-type {
width: 80px;
min-height: 80px;
}
tr + tr td {
border-top: 1px solid #dcdcdc;
}
<table>
<tr><td colspan="2">In Safari 10.0.1 and iOS 10.1, strange behavior can be observed on SVG shapes with <code>rotate</code> values not divisible by 90 degrees, when <code>transform-origin: center center;</code></td></tr>
<tr>
<td>
<svg xmlns="http://www.w3.org/2000/svg">
<circle class="degrot" r="35" cy="40" cx="40" />
</svg>
</td>
<td>
<code>transform: rotate(-90deg);</code>
<p>The stroke improperly begins <a href="https://www.w3.org/TR/SVG11/shapes.html#CircleElement">at 3:00</a>, as if the <code>transform</code> rule hadn't been applied.</p>
</td>
</tr>
<tr>
<td>
<svg xmlns="http://www.w3.org/2000/svg">
<circle class="degrot-offset" r="35" cy="40" cx="40" />
</svg>
</td>
<td>
<code>transform: rotate(-90.1deg);</code>
<p>The stroke begins at (twelve seconds before) 12:00, as expected.</p>
</td>
</tr>
<tr>
<td>
<svg xmlns="http://www.w3.org/2000/svg">
<circle class="turnrot" r="35" cy="40" cx="40" />
</svg>
</td>
<td>
<code>transform: rotate(-0.25turn);</code>
<p>The same bug applies to any <code>rotate</code> value which computes to a multiple of 90 degrees.</p>
</td>
</tr>
<tr>
<td>
<svg xmlns="http://www.w3.org/2000/svg">
<circle class="turnrot-offset" r="35" cy="40" cx="40" />
</svg>
</td>
<td>
<code>transform: rotate(-0.251turn);</code>
<p>43 seconds before noon.</p>
</td>
</tr>
<tr><td colspan="2">But when the SVG element specifies a <code>viewBox</code> which is being scaled down, things can get weird:</td></tr>
<tr>
<td>
<svg viewBox="0 0 160 160" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot" r="70" cy="80" cx="80" />
</svg>
</td>
<td>
<code>transform: rotate(-90deg);</code>
<p>So far, so the same.</p>
</td>
</tr>
<tr>
<td>
<svg viewBox="0 0 160 160" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot-offset" r="70" cy="80" cx="80" />
</svg>
</td>
<td>
<code>transform: rotate(-90.1deg);</code>
<p>But now, offsetting by a little bit doesn't work, <em>unless</em> you zoom in the page in past a certain zoom threshold (either via pinching, or <code>View > Zoom</code> and/or keyboard shortcut). Try it; it's unsetting!</p>
<p>This is probably because of some rounding of that the zooming engine performs, because...</p>
</td>
</tr>
<tr>
<td>
<svg viewBox="0 0 160 160" class="scaledown" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot-offset-more" r="70" cy="80" cx="80" />
</svg>
</td>
<td>
<code>transform: rotate(-92deg);</code>
<p>offsetting by a larger amount restores expected behavior.</p>
</td>
</tr>
<tr><td colspan="2">If the SVG element is not being scaled <em>down</em>, behavior identical to the first section resumes. Zooming has no effect:</td></tr>
<tr>
<td>
<svg viewBox="0 0 80 80" class="noscale" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot" r="35" cy="40" cx="40" />
</svg>
<svg viewBox="0 0 40 40" class="scaleup" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot" r="17.5" cy="20" cx="20" />
</svg>
</td>
<td>
<code>transform: rotate(-90deg);</code>
<p>
Top: No scaling (viewBox dimensions match parent element's)<br><br>
Bottom: Scaling up (viewBox dimensions half of parent element's)
</p>
</td>
</tr>
<tr>
<td>
<svg viewBox="0 0 80 80" class="noscale" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot-offset" r="35" cy="40" cx="40" />
</svg>
<svg viewBox="0 0 40 40" class="scaleup" xmlns="http://www.w3.org/2000/svg">
<circle class="degrot-offset" r="17.5" cy="20" cx="20" />
</svg>
</td>
<td>
<code>transform: rotate(-90.1deg);</code>
<p>
Top: No scaling (viewBox dimensions match parent element's)<br><br>
Bottom: Scaling up (viewBox dimensions half of parent element's)
</p>
</td>
</tr>
<tr><td colspan="2">But there is one exception:</td></tr>
<tr>
<td>
<svg class="degrot wc" xmlns="http://www.w3.org/2000/svg">
<circle r="35" cy="40" cx="40" />
</svg>
</td>
<td>
<p>On the parent <code>svg</code> element:</p>
<code>transform: rotate(-90deg);<br>will-change: transform;</code>
<p>Iff the the the rotation is applied to a <em>parent</em> of the SVG shape (including the SVG element itself) along with the rule <code>will-change: transform</code>, all rotation values work as expected.</p>
</td>
</tr>
<tr><td colspan="2">All these behaviors have been observed in Safari 10.0.1 and iOS 10.1. They appear to be fixed as of iOS 10.2 Beta 2.</td></tr>
</table>
As stated in the demo, it appears to be fixed in iOS 10.2, at least in the public beta version I just downloaded. Presumably, a Safari fix will also be arriving in due time.
This is happening to me as well, I settled on using a rotation just shy of being divisible by 90 degrees to get around this in the interim.