🎉 News:
- Paths rendering has been improved, and paths are automatically simplified depending on the scale to improve performance
- New gestures added (zoom fling, double tap to zoom, two fingers tap)
- Marker clustering and lazy-loading. New examples added to the demo app
MapCompose is a fast, memory efficient Jetpack compose library to display tiled maps with minimal effort. It shows the visible part of a tiled map with support of markers and paths, and various gestures (flinging, dragging, scaling, and rotating).
An example of setting up:
/* Inside your view-model */
val tileStreamProvider = TileStreamProvider { row, col, zoomLvl ->
FileInputStream(File("path/{$zoomLvl}/{$row}/{$col}.jpg")) // or it can be a remote HTTP fetch
}
val state: MapState by mutableStateOf(
MapState(4, 4096, 4096).apply {
addLayer(tileStreamProvider)
enableRotation()
}
)
/* Inside a composable */
@Composable
fun MapContainer(
modifier: Modifier = Modifier, viewModel: YourViewModel
) {
MapUI(modifier, state = viewModel.state)
}
This project holds the source code of this library, plus a demo app - which is useful to get started. To test the demo, just clone the repo and launch the demo app from Android Studio.
Marker clustering regroups markers of close proximity into clusters. The video below shows how it works.
clustering.mp4
The sample below shows the relevant part of the code. We can still add regular markers (not managed by a clusterer), such as the red marker in the video. See the full code.
/* Add clusterer */
state.addClusterer("default") { ids ->
{ Cluster(size = ids.size) }
}
/* Add marker managed by the clusterer */
state.addMarker(
id = "marker",
x = 0.2,
y = 0.3,
renderingStrategy = RenderingStrategy.Clustering("default"),
) {
Marker()
}
There's an example in the demo app.
Add this to your module's build.gradle
implementation 'ovh.plrapps:mapcompose:2.8.1'
Starting with v.2.4.1, the library is using the compose BOM. The version of the BOM is specified in the release notes. The demo app shows an example of how to use it.
MapCompose is optimized to display maps that have several levels, like this:
Each next level is twice bigger than the former, and provides more details. Overall, this looks like a pyramid. Another common name is "deep-zoom" map. This library comes with a demo app featuring various use-cases such as using markers, paths, map rotation, etc. All examples use the same map stored in the assets, which is a great example of deep-zoom map.
MapCompose can also be used with single level maps.
With Jetpack Compose, we have to change the way we think about views. In the previous View
system, we had references on views and mutated their state directly. While that could be done right,
the state often ended-up scattered between views own state and application state. Sometimes, it was
difficult to predict how views were rendered because there were so many things to take into account.
Now, the rendering is a function of a state. If that state changes, the "view" updates accordingly.
In a typical application, you create a MapState
instance inside a ViewModel
(or whatever
component which survives device rotation). Your MapState
should then be passed to the MapUI
composable. The code sample at the top of this readme shows an example. Then, whenever you need to
update the map (add a marker, a path, change the scale, etc.), you invoke APIs on your MapState
instance. As its name suggests, MapState
also owns the state. Therefore, composables will always
render consistently - even after a device rotation.
All public APIs are located under the api
package. The following sections provide details on the MapState
class, and give examples of how to
add markers, callouts, and paths.
The MapState
class expects three parameters for its construction:
levelCount
: The number of levels of the map,fullWidth
: The width of the map at scale 1.0, which is the width of last level,fullHeight
: The height of the map at scale 1.0, which is the height of last level
MapCompose supports layers - e.g it's possible to add several tile pyramids. Each level is made of the superposition of tiles from all pyramids at the given level. For example, at the second level (starting from the lowest scale), tiles would look like the image below when three layers are added.
Your implementation of the TileStreamProvider
interface (see below) is what defines a tile
pyramid. It provides InputStream
s of image files (png, jpg). MapCompose will request tiles using
the convention that the origin is at the top-left corner. For example, the tile requested with
row
= 0, and col = 0
will be positioned at the top-left corner.
fun interface TileStreamProvider {
suspend fun getTileStream(row: Int, col: Int, zoomLvl: Int): InputStream?
}
Depending on your configuration, your TileStreamProvider
implementation might fetch local files,
as well as performing remote HTTP requests - it's up to you. You don't have to worry about threading,
MapCompose takes care of that (the main thread isn't blocked by getTileStream
calls). However, in
case of HTTP requests, it's advised to create a MapState
with a higher than default workerCount
.
That optional parameter defines the size of the dedicated thread pool for fetching tiles, and defaults
to the number of cores minus one. Typically, you would want to set workerCount
to 16 when performing
HTTP requests. Otherwise, you can safely leave it to its default.
To add a layer, use the addLayer
on your MapState
instance. There are others APIs for reordering,
removing, setting alpha - all dynamically.
To add a marker, use the addMarker API, like so:
/* Add a marker at the center of the map */
mapState.addMarker("id", x = 0.5, y = 0.5) {
Icon(
painter = painterResource(id = R.drawable.map_marker),
contentDescription = null,
modifier = Modifier.size(50.dp),
tint = Color(0xCC2196F3)
)
}
A marker is a composable that you supply (in the example above, it's an Icon
). It can be
whatever composable you like. A marker does not scale, but it's position updates as the map scales,
so it's always attached to the original position. A marker has an anchor point defined - the point
which is fixed relatively to the map. This anchor point is defined using relative offsets, which are
applied to the width and height of the marker. For example, to have a marker centered horizontally
and aligned at the bottom edge (like a typical map pin would do), you'd pass -0.5f and -1.0f as
relative offsets (left position is offset by half the width, and top is offset by the full height).
If necessary, an absolute offset expressed in pixels can be applied, in addition to the
relative offset.
Markers can be moved, removed, and be draggable. See the following APIs: moveMarker, removeMarker, enableMarkerDrag.
Callouts are typically message popups which are, like markers, attached to a specific position. However, they automatically dismiss on touch down. This default behavior can be changed. To add a callout, use addCallout.
Callouts can be programmatically removed (if automatic dismiss was disabled).
To add a path, use the addPath
api:
mapState.addPath("pathId", color = Color(0xFF448AFF)) {
addPoints(points)
}
The demo app shows a complete example.
It's pretty common to programmatically animate the scroll and/or the scale, or even the rotation of the map.
scroll and/or scale animation
When animating the scale, we generally do so while maintaining the center of the screen at a specific position. Likewise, when animating the scroll position, we can do so with or without animating the scale altogether, using scrollTo and snapScrollTo.
rotation animation
For animating the rotation while keeping the current scale and scroll, use the rotateTo API.
Both scrollTo
and rotateTo
are suspending functions. Therefore, you know exactly when
an animation finishes, and you can easily chain animations inside a coroutine.
// Inside a ViewModel
viewModelScope.launch {
mapState.scrollTo(0.8, 0.8, destScale = 2f)
mapState.rotateTo(180f, TweenSpec(2000, easing = FastOutSlowInEasing))
}
For a detailed example, see the "AnimationDemo".
-
In MapView, you had to define bounds before you could add markers. There's no such concept in MapCompose anymore. Now, coordinates are normalized. For example, (x=0.5, y=0.5) is a point located at the center of the map. Normalized coordinates are easier to reason about, and application code can still translate this coordinate system to a custom one.
-
In MapView, you had to build a configuration and use that configuration to create a
MapView
instance. There's no such thing in MapCompose. Now, you create aMapState
object with required parameters. -
A lot of things which couldn't change after MapView configuration can now be changed dynamically in MapCompose. For example, the
zIndex
of a marker, or the minimum scale mode can be changed at runtime.
-
There's now a way to set initial values for various properties such as scroll, scale, etc using the
InitialValuesBuilder
in theMapState
constructor. To produce similar behavior in 1.x, one had to launch a coroutine right afterMapState
creation - which wasn't perfect since some undesired tile loading could happen between the initialization and the destination state. -
Having a
TileStreamProvider
atMapState
construction is no longer mandatory.TileStreamProvider
s are now added using theaddLayer
api, which is completely dynamic. -
While 1.x version had a non-suspending
TileStreamProvider
, 2.x greatly benefits from the new suspend version. If you're using a library like Retrofit to perform remote http fetch (and suspend calls), tile loading will be optimal since all layers are fetched concurrently. That was already the case in 1.x, but not thanks to suspending calls.
Marcin (@Nohus) has contributed and fixed some issues. He also thoroughly tested the new layers
feature – which made MapCompose
better.