Reusing the iterator in a nested loop is not a good idea. And in
this case it is better to not have nested loops at all, because
we only have to create one array.
This is another attempt to bring master closer to the
vector-api branch. In anticipation of the ability to keep track
of modifications on ol.Object through a beforechange event
(d7e4be0), we will be able to manage originals on the
application level or in a separate component outside of
ol.Feature.
This separates the action of requesting an extent to be loaded from the action of requesting cached features. The renderer (or any other consumer of a vector source) calls load to request a data extent. A `featureload` event fires when new features are loaded. The renderer (or any other consumer) separately asks for cached features given an extent. This vector source only loads features once, but this separation will also work with sources that make multiple requests for data in different extents.
This also removes the `data` option from the vector source in favor of a `features` option. Since we no longer have shared data structures for geometries, people can manually create features and pass them to a vector source. The `addFeatures` method is exported as well. This is used to add features to a source that don't have a representation on the "remote" (or server).
This moves the feature cache from ol.layer.Vector to ol.source.Vector. These are the minimum changes required to maintain the existing functionality and make tests pass. More refactoring to come.
This is one more step to bring the master and vector-api
branches closer together. I am well aware that the current
Modify interaction will probably be rewritten as part of the
vector-api effort, but with changes like this we will eventually
be able to remove ol.structs.RTree from master.
By providing a segment even for points, we have less distinction
points between points and other geometry types, and are able to
do the segment distance calculation also for points.
The only feature on the temporary layer is now the
vertexFeature, and the temporary layer's style is dynamically
changed to the style of the layer whose segment is being edited
(the topmost layer if more than one segment are being edited).
With this simplification, we can also put all segments on a
single RTree. Finally, we no longer rely on structures set on
layers - all we need is now on the instance itself.
This refactoring also changes the way we define layers to
participate in modification - by using an array or a filter
function, or assuming all layers if no layers property is set.
The RTree cannot work with NaN coordinates. So instead of
preparing a vertexFeature before we know a coordinate, we now
lazily create the vertexFeature.
In combination with the additional squaredDistanceToSegment
check we can ensure that a future vertex is only created on one
segment per feature, and only on the segment that it appears on.
This change also renames the misleading dragVertices_ member to
dragSegments_.
This is a first draft. The way geometry changes are handled for
now is a bit clumsy. Both updating the feature cache RTree and
making the layer aware of feature and geometry changes could be
handled in a smarter way if these changes would be made through
the layer instead of directly on the geometry or feature.
The tests revealed that polygons can be drawn with counter-clockwise exterior ring order. In the polygon constructor, we enforce clockwise winding for exterior rings. So after drawing, we pass polygon coordinates back to the constructor before adding feature to the target layer.
This allows us to cast single-part geometries to multi-part types before adding features to the target layer. This doesn't yet allow for drawing multi-part geometries with multiple parts. That can be handled separately.
