locationtech / geotrellis Goto Github PK

We should create a "getting started" tutorial that walks a new user through the process of creating a sample REST geoprocessing service, along the lines of tutorials such as http://typesafe.com/resources/getting-started/tutorials/getting-started-first-scala.html.

It would be useful to have operations which compress and decompress their value ranges. CompressedArrayHistogram currently implements something similar, but this functionality really belongs in an operation and not in a histogram.

Currently we support PNG via a single transparent color value.

We should instead provide support for an alpha channel. This would allow us to support partial transparency.

We should still maintain support for 24-bit color because if we don't need alpha transparency then it's a bit more efficient.

Right now we don't really have code to do this, so you do a lot of it manually.

I think that we want to stick with manually composing SQL (at least for now) but we should at least have an operation like:

GetPointsFromPostgis(sql:Op[String]):Op[Array[Point]] or something.

These are redundant with the cols/rows stored in RasterExtent.

If necessary we can write methods which call-through to RasterExtent, although ultimately we should just rewrite all uses to use the RasterExtent directly.

We're no longer using Configgy for logging, as that project is no longer being supported. slf4j (simple logging facade) seems like a good choice, and would let us combine our logging configuration with akka's logging configuration.

Bordered tiles -- tiles with a variable width border -- would allow for tiled parallelization of neighborhood raster operations.

We need to enhance and adapt the existing command line tool to be the general purpose data import tool.

Right now there's a lot of overlap, confusion, and ugliness with PNG rendering. Here's what we should aim for:

1. Java PNGWriter class

This is an optimized class for taking a ByteBuffer of RGB data and writing it to an OutputStream using static methods.

1. Scala PNGWriter class

This is a wrapper around the Java class that will convert a raster into a ByteBuffer.

(OPTIONAL: #1 and #2 could be combined if necesary)

1. Code for mapping data rasters into RGB rasters based on ColorBreaks, etc
2. Operations for doing doing #3
3. High-level operations that given a raster and some params will generate breaks, do the mapping ,etc.

Right now we have a weird mixture of this stuff. The goal would be to move knowledge about colors and such up to step #3 from step #2.

Currently ArrayHistogram maps values directly as array indices. It would be good to add an offset parameter to modify this addressing, so that if you have data ranging from e.g. 145-316 you won't end up wasting the first 145 buckets of the histogram.

The offset can default to 0 to preserve the existing behavior. Operations involving ArrayHistogram should also be changed.

Select a serialization protocol for fast remote transmission of raster data (e.g. protocol buffers, avro, thrift) and then integrate the library with Akka. The default Java serialization works but is not fast.

All resampling code should be classified as transform (no change to underlying data) or warp (interpolation) operations, and where appropriate alternatives should be provided.

Operations (or some Operations) should automatically and asynchronously run the input operations specified in their constructors. One possibility is that they will extend an Input and Output class with type parameters that describe their input and output, or we might consider using a technique from the shapeless library.

The tests should use a catalog.json file designed for the test collateral data. The tests should load rasters by name, instead of using LoadFile with file path.s

It may not end up in this repo, but it would be good to update the GeoServer plugin to the new API.