PostGIS Dream Journal

Where Ideas Go to Dance

HOW TO JOIN THE DANCE

• Create a google slides presentation
• Add your idea to that presentation
• Just one slide please
• Include an example function name
• Include (lots of) pictures
• Any other necessary explanatory text
• Links to background info if necessary (papers, algorithms, etc)
• Include your contact info
• twitter handle, github handle, something we can credit
• Make your presentation public-with-link
• Add the link in a comment on this slide!

ST_Skeletonize(Polygon, [MultiPoint]) → MultiLinestring

• Given a polygon, and a set of entry points, generate connecting lines that respect a medial axis of the geometry and connect all the entry points.
• Current skeletonize doesn’t go to edges of polygon and doesn’t support providing input points.

ST_Drape(TIN, Linestring, [Densify]) → LineString

• Given a 3D TIN, add the Z values to the linestring, including new vertices as necessary to fit the 3D surface.

ST_FittedSurface(MultiPointM, Tolerance) → TIN

• Given a set of points with measure, generate a fitted surface that corresponds to the measure. (Kriging?)
• Build a temperature surface from discrete weather station data, etc.

ST_Viewshed(Raster, MultiPoint) → Raster

ST_BarnesSurface(MultiPointM, Converge, Passes) → Raster

• Given a set of points, generate a Barnes surface
• Build a heat-map from a collection of occurrence points.

ST_IsoLines(TIN, ARRAY[float]) → MultiLineString�ST_IsoLines(Raster, ARRAY[float]) → MultiLineString

• Given a surface in XYZ or XYM, return a multi-linestring with an isoline for each value provided in the array.
• Convert a temperature surface into a weather map. Convert an elevation TIN into contour lines for a topographic map.

ST_SpanningTree(MultiLineString) → MultiLineString

• Join all the vertices in a graph using the minimum length of edges.
• https://en.m.wikipedia.org/wiki/Minimum_spanning_tree

ST_Raster(TIN) → Raster�ST_TIN(Raster, Tolerance) → TIN

• Given a TIN, produce a one band raster that gives the TIN value at each point in the raster.
• Given a one band raster, fit a TIN, skipping out the raster pixel values that are not outside the tolerance.

ST_MakeClean(Geometry, Tolerance) → Geometry

• For individual features, remove common data errors within a tolerance and enforce OGC or ESRI validity rules.
• Zigzags
• Spikes
• Micro-loops
• Micro-holes
• Ring crossings / grazes
• Repeated points
• Co-linear points

ST_MakeClean(Geometry, Tolerance) → Geometry

• Window function
• As for geometry function, but also look for geometry-to-geometry relationships
• Linear features snap ends to nearby lines, add vertices at snap points
• Polygonal features, close gaps, merge overlaps, ensure all vertices on shared boundaries match, return clean coverage

ST_CoverageUnion, ST_IsCoverageValid

• Aggregate Function
• Valid Polygonal coverages can be unioned very efficiently
• essentially by dropping all interior edges
• Helpful to be able to test a priori for valid polygonal coverage
• ideally report where it is invalid to aid fixing

ST_Intersection(geography, geography)

• Overlay operations �on the sphere
• Add S2 as a (optional?) dependency for geography

ST_Union(geometry[]) in Parallel

• Note number of available CPUs
• Parallelize each pair-wise union operations in the cascaded union code of GEOS
• Take advantage of the OpenMP extensions in clang/gcc modern versions

ST_Combine(Raster, Raster) → Raster and/or Table

• Given 2 value integer rasters create a new raster with the raster value being the unique combination of the input rasters. An attribute table is created with the resulting value, the 2 inputs and their count as columns
• like the ArcGIS raster tool
• See also: 1 and 2

source and (c): ESRI