InitialDLab / Simba

Spatial In-Memory Big data Analytics
Apache License 2.0
120 stars 62 forks source link

simba perform join #88

Open geoHeil opened 7 years ago

geoHeil commented 7 years ago

In the examples I find ps.knn("p", Array(1.0, 1.0), 4).show() how can I actually perform a join operation in Simba? and not just intersect with an array but rather a column of polygons in a dataframe?

dongx-psu commented 7 years ago

Join examples are located at here:

https://github.com/InitialDLab/Simba/blob/standalone-2.1/src/main/scala/org/apache/spark/sql/simba/examples/BasicSpatialOps.scala#L48

geoHeil commented 7 years ago

What are the Array("x", "y"),Array("x", "y") in the file you linked? Are these supposed to indicate the geospatial columns? if yes, how can these support polygons i.e. it would look like only 2 df of points would be joinable.

dongx-psu commented 7 years ago

You need to use arrays if there is no shape type in your data source. If you are joining two attributes of ShapeType, you can simply put the join attribute names as strings without Array.

Examples for ShapeType attributes are located:

https://github.com/InitialDLab/Simba/blob/standalone-2.1/src/main/scala/org/apache/spark/sql/simba/examples/SpatialClassInference.scala

More about Dataset APIs we supported:

https://github.com/InitialDLab/Simba/blob/standalone-2.1/src/main/scala/org/apache/spark/sql/simba/DataSet.scala

To make it easier to understand, one API joining two ShapeType attributes:

def distanceJoin(right: Dataset[_], leftKeys: Array[String], rightKeys: Array[String], r: Double) : DataFrame

geoHeil commented 7 years ago

So pointsDf. distanceJoin(polygonDf, Array("point"), Array("polygon"), 3).show should work? Do I understand correctly that I should build an index beforehand? simba.indexTable("a", RTreeType, "testqtree", Array("x", "y") )

so here with points and polygons how can I make sure points and polygons are bot partitioned in the same way?

dongx-psu commented 7 years ago

Sorry, it should be:

def distanceJoin(right: Dataset[_], leftKey: String, rightKey: String, r: Double) : DataFrame

Array is not needed. And I should say, it may not work since I haven't try any queries for Polygons since it is still under development. I don't think you can do distance join between Polygons and Points in this version. Only things I can guarantee are point operations.