Hi Peter,
On Mar 18, 2012, at 9:11 PM, Peter K wrote:
>
>> Thanks for your email! We don't have any unit tests written yet. We would
>> love to produce some.
>> We have a large geolocation dataset here [1]
>
> I more meant single unit tests which simply ensures the raw
> functionality of one a tiny part of the system and are often written
> when or before the piece of code is written. I think, what you have in
> mind are more integration tests (not sure though :) !) which tests
> performance and more complicated bugs when the units play together...
Yeah that's part of it, but I was also thinking that we could do something
like the following:
1. Abstract the storage interface to load the quad tree, e.g.,
public interface Loadable{
public QuadTreeNode load(SpatialDataset dataset)
}
Then have:
public class GeoRSSData extends SpatialDataset{
//... magic here
}
public class GeoRSSLoader implements Loadable{
public QuadTreeNode load(SpatialDataset dataset){
GeoRSSData data = new GeoRSSData(dataset);
//...load it into the QuadTree
}
public class ExcelGeoData extends SpatialDataset{
//...magic here
}
public class ExcelLoader implements Loadable{
//...
}
You follow me? Then we could define a chunked load interface as well
(e.g., multiple, chunked call backs to load to reduce mem overhead) and
then use e.g., the CSV data as a unit test dataset to both test the storage
interface, and as well, test that the Quad Tree is doing the right thing,
spatially. Does that make sense? You interested in helping?
>
> The 'problem' without unit- and only integration-tests would be that you
> see a test is failing, but you'll need a lot of time to dig into the
> real reason of the bug. Also executing the unit tests should not take a
> long time to load values, also IMO unit tests are a great resource for
> other developers or could even replace documentation ...
Agreed, we need them :) You're preaching to the choir here.
>
>> Do you have any ideas in mind that you'd like to help us out on?
>
> I'm especially interested in an memory efficient datastructure for a Map ala
> Map<lat+lon, AnyValue>
>
> I've already implemented a very rough prototype of such a Map using a
> ByteBuffer but now I think a datastructure with 'getNeighbour' would give a
> huge benefit and I'm looking about possibilities.
Sounds like a SpatialDataset type :)
>
>
>> Help and patches are most welcome!
>
> Yeah, I'll definitely have a look into SIS as I'm more digging into graphs
> and geo stuff now.
Awesome! Welcome aboard and looking forward to having you part of the community!
Cheers,
Chris
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Chris Mattmann, Ph.D.
Senior Computer Scientist
NASA Jet Propulsion Laboratory Pasadena, CA 91109 USA
Office: 171-266B, Mailstop: 171-246
Email: [email protected]
WWW: http://sunset.usc.edu/~mattmann/
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Adjunct Assistant Professor, Computer Science Department
University of Southern California, Los Angeles, CA 90089 USA
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
