Add stroking to polygons #597
Conversation
This is done by creating a sub-feature of lines for the polygons if the polygon feature has any value other than false for the stroke style (either true or a function). Lines are created for each exterior and interior loop in the polygons, and are not set for interaction. Lines are created with the same starting and ending point. Strokes can have all of the properties that lines get -- strokeWidth, strokeColor, strokeOpacity, strokeStyle. Lines are rendered with the same renderer as the polygons (and on the same level). Polygons now have a flag to turn off filling the polygon (though the polygon is still created and still earcut, so this doesn't speed things up for polygons with fill = false). It would be better to add a flag to the line feature to indicate a closed loop. Changed the gl.object class to not re-subclass sceneObject, as all gl features are already subclasses of sceneObject. This requires a vgl update to allow unbinding vertex data. That change requires an update to the quad feature. Fixed a typo in the lineFeature.
|
This is only a WIP until a new VGL version is available. |
Current coverage is 80.63%@@ master #597 diff @@
==========================================
Files 82 82
Lines 7354 7417 +63
Methods 0 0
Messages 0 0
Branches 0 0
==========================================
+ Hits 5796 5981 +185
+ Misses 1558 1436 -122
Partials 0 0
|
|
This looks good, I just need to fix a couple of the selenium tests before it goes in. |
|
I found out the changes here broke styling in the Anyway, I updated the |
|
One other thing I noticed that we should look into after this goes in is the quality of the stroking, which at the moment is pretty bad. See the screenshot attached, which is at display resolution (not zoomed in).
|
|
I filed issue #596 about the way lines are rendered. They are rendered after the fill, so should never underneath it. You'll see the exact same lines (and same errors) with a fill opacity of 0. It appears like the vertices of some of the triangles are wrong due to an error in how mitering is calculated. |
In calculating angles related to lines, two angles were averaged using (a + b) / 2, which doesn't work in the general case.
|
@jbeezley Can you check if the selenium tests need to be updated after the line fix? |
|
Done. LGTM. |
| @@ -96,6 +94,8 @@ var gl_lineFeature = function (arg) { | |||
| ' else anglePrev = atan(deltaPrev.y / aspect, deltaPrev.x);', | |||
| ' if (deltaNext.xy == vec2(0.0, 0.0)) angleNext = anglePrev;', | |||
| ' float angle = (anglePrev + angleNext) / 2.0;', | |||
| ' if (abs(anglePrev - angleNext) >= PI)', | |||
| ' angle += PI;', | |||
This is done by creating a sub-feature of lines for the polygons if the polygon feature has any value other than false for the stroke style (either true or a function). Lines are created for each exterior and interior loop in the polygons, and are not set for interaction. Lines are created with the same starting and ending point.
Strokes can have all of the properties that lines get -- strokeWidth, strokeColor, strokeOpacity, strokeStyle. Lines are rendered with the same renderer as the polygons (and on the same level).
Polygons now have a flag to turn off filling the polygon (though the polygon is still created and still earcut, so this doesn't speed things up for polygons with fill = false).
It would be better to add a flag to the line feature to indicate a closed loop.
Changed the gl.object class to not re-subclass sceneObject, as all gl features are already subclasses of sceneObject.
This requires a vgl update to allow unbinding vertex data. That change requires an update to the quad feature.
Fixed a typo in the lineFeature.
This resolves issue #494.