Home

Awesome

GeoTrellis Server

CI Maven Central Snapshots

GeoTrellis Server is a set of components designed to simplify viewing, processing, and serving raster data from arbitrary sources with an emphasis on doing so in a functional style. It aims to ease the pains related to constructing complex raster processing workflows which result in TMS-based (z, x, y) and extent-based products.

In addition to providing a story about how sources of imagery can be displayed or returned, this project aims to simplify the creation of dynamic, responsive layers whose transformations can be described in MAML (Map Algebra Modeling Language).

Getting Started with GeoTrellis Server

GeoTrellis Server is currently available for Scala 2.13 and 2.12.

To get started with SBT, simply add the following to your build.sbt file:

libraryDependencies += "com.azavea.geotrellis" %% "geotrellis-server-core" % "<latest version>"

High level concepts

Imagine you've got a simple case class which is sufficient to identify layers of imagery for an application you're working on.

import java.net.URI

case class ImageryLayer(location: URI)

Imagine further that you'd like to enable your application to compose these layers together via map algebra to produce derived layers and that the combinations can't be known at compile-time (users will be deciding which - if any - map algebra to run). This is a job that GeoTrellis server can radically simplify. Because we're dealing with behavior specified at runtime, we need to evaluate program descriptions rather than simple, first-order parameters - the task for which MAML was written.

// This node describes a `LocalAdd` on the values eventually
// bound to `RasterVar("test1")` and `RasterVar("test2")
val simpleAdd = Addition(List(RasterVar("test1"), RasterVar("test2")))

// This describes incrementing every value in the eventually bound raster by 1
val plusOne = Addition(List(RasterVar("additionRaster1"), IntLit(1)))

Because MAML encodes map alagebra transformations in data, actually executing a MAML program for practical purposes can be difficult and unintuitive. GeoTrellis server bridges this gap by providing a typeclass-based means of extending source types in client applications with the behaviors necessary to actually evaluate a MAML AST.

The following example demonstrates what is required to generate functions which produce extents provided MAML program descriptions. For a more complete example, check out CogNode.

import io.circe._
import io.circe.syntax._
import io.circe.generic.semiauto._
import cats._
import cats.effect._
import com.azavea.maml.ast._
import com.azavea.maml.eval.BufferingInterpreter
import geotrellis.server._

// This class points to a COG and specifies a band of interest
case class RasterRef(uri: URI, band: Int)

// We need to provide some implicit evidence. Most applications can do this within companion objects
object RasterRef {
  // reification means 'thingification', and that's what we're proving we can do here
  implicit val rasterRefExtentReification: ExtentReification[RasterRef] = new ExtentReification[RasterRef] {
    def extentReification(self: CogNode, buffer: Int)(implicit contextShift: ContextShift[IO]): (Extent, CellSize) => IO[Literal] = ???
  }
  // We can lean on circe's automatic derivation to provide an encoder
  implicit val rasterRefEncoding: Encoder[RasterRef] = deriveEncoder[RasterRef]
}

// A source from which MAML evaluation will be able to derive necessary artifacts
val reference = RasterRef("http://some.url.com", 1)

// We need to key provided references based on the ID of the Var they'll replace
val parameters = Map("additionRaster1" -> reference)

// This is an interpeter GT Server will use to roll up the tree + params to some result
val interpreter = BufferingInterpreter.DEFAULT

// Not yet a result: we can use the result here to produce artifacts for different extent inputs
val tileEval = LayerExtent.apply(IO.pure(plusOne), IO.pure(parameters), interpreter)

val targetExtent: Extent = ??? // Where should the tile come from?
val targetCellSize: CellSize = ??? // What resolution should the tile be?

// Branch on Valid/Invalid and print some info about which branch we're on
tileEval(targetExtent, targetCellSize) map {
  case Valid(tile) =>
    println("we did it, a tile: (rows: ${tile.rows}, cols: ${tile.cols})")
  case Invalid(err) =>
    println("Ran into an error (${err.asJson}) during MAML evaluation of AST (${plusOne.asJson}) with params (${params.asJson})")
}

LayerExtent is joined by two other objects which organize evaluation strategies for different products:

Each of these objects is a response to distinct needs encountered when writing raster-based applications. Included in each object are methods which encode several strategies for evaluating their products. The strategies currently available are:

Running an example

Three example servers are available which can be run through the provided makefile. These examples have been implemented via http4s, which has a pleasant, clean API and plays nicely with the cats and cats-effect libraries.

  1. A server and simple UI that evaluates weighted overlays between arbitrary COGs. This demo includes a simple UI, available at http://localhost:9000/ for a
./scripts/server --overlay
  1. Integrates GTServer components with application-specific persistence needs.
./scripts/server --persistence
  1. Illustrates GTServer evaluating a remote-sensing classic, the NDVI.
./scripts/server --ndvi