This makes the asynchronous and synchronous versions of
readFeatures work with the same data structures, and leaves
projection handling outside the parsers.
Using the y-coordinate of the polygon's bounding box, this
simple algorithm intersects the polygon with the horizontal
center line of its bounding box. The x-coordinate of the label
point is the center of the longest segment of this intersection
that is inside the polygon.
Instead of just listening for the first change, we now listen
for all changes, and expire tiles by tile range. The outcome is
that the vector layer's addFeatures and removeFeatures methods
now show instant results on the rendered map.
This also adds an assert that cache values are an instance of
ol.Tile in expireCache, instead of doing a type cast. Later we
will want to enforce the ol.Tile type by overriding the set
method.
With this change, getFeaturesObjectForExtent may return null if
the source does not have data loaded for the provided extent.
A callback can be passed to getFeaturesObjectForExtent to get
notified when the requested data is available.
This adds url, data and parser options to the source, and makes
EPSG:4326 the default projection. It also adds a prepareFeatures
method, which is used to load/parse data once the target
projection is known.
As a follow-up to #815, this change makes
ol.tilegrid.createForProjection work in cases where a Proj4
projection does not have units set. For better code readability,
a new variable is used when calculating the resolution.
On Android (both native browser and Chrome), the handleTouchEnd_
method is sometimes reached in a state where this.down_ is null.
This check protects against this.click_ being called without
an event object. To see this issue, open any example in Android,
and tap the zoom + or - button.
Instead, we create a new tile grid whenever renderFrame is
called, no animation is active, and the resolution is not in the
tile grid already. This gives better rendering results because
we get vector tiles at native resolutions.
In JavaScript, keys of object literals are always strings, and
internal type conversions are performed. Now if we tell the
compiler that keys are numbers, we get inconsistent types when
iterating through keys. So instead we set the key type to string
and do a type cast to make the compiler happy. Note that we
could also do toString() instead of a type cast, but it would
add a performance penalty (see
http://jsperf.com/internal-type-conversion-vs-tostring-for-object-keys).
The projection validity extent is used to generate a sensible
set of default resolutions and a sensible default tile grid.
By making it optional, we can still generate defaults - with
zoom levels that are similar to the default web mercator zoom
levels (based on fitting the world on a single tile, even if the
projection is not available for the whole world).
Setting the renderer of the ten-thousand-points example to
WebGL should remove the need for this stub. This is basically
the same approach we took for all vector examples in master,
where we set the renderer to Canvas.
To avoid surprises for application developers, this change
creates a new getFeatures method. So it is clear now beforehand
whether features or feature info markup is returned. The result
is now also grouped by layer, so application developers always
have a link between a layer and the feature info it returns.
To make getFeatureInfo return markup for vector layers, this
change also adds a featureInfoFunction property to the vector
layer, which gives developers full control over how features are
rendered to feature info markup.
This change reverts the RTree API back to the original one (i.e.
insert() instead of put(), search() instead of find()), and
creates a new searchReturningObject() method that returns an
object keyed by UIDs.
It also adds missing tests for type restricted search and
searchReturningObject().
Not using Math.max and Math.min increases performance, and by
using ol.extent functions instead of RTree's rectangle
structures and calculations we can get rid of several functions.
This new implementation is based on
http://github.com/imbcmdth/RTree/, with only a few modifications
to add the optional type and provide the API of the previous
implementation.
There is still room for optimization, but this is such an
improvement over the previous RTree already that it's worth
bringing it in now.
By doing the type check before the intersection check, we can
save he intersection check for cases where we don't care about
type or have the specified type in a node.
On servers that report an incorrect content type for XML
documents, responseXML will be empty. In these cases we need to
use responseText instead. This change also prevents readNode
calls with empty data.
ahocevar