2

Is my understanding correct, namely that two polygons touching at a common point cannot be merged because the result would be a self-intersecting polygon? I know there are dozens of Q&A threads on self-intersecting polygons and I’ve looked through them. This post: Detecting invalid polygons (self-intersecting or touching) seems to indicate merging 2 polygons with a single common point into one would create a self-intersecting polygon, and would thus be invalid. Likewise, the answer to this post Is it possible to have self intersecting polygons in shapefile? suggests such a geometry could not be stored in a shapefile. However, this post seems to indicate it is possible to accomplish it: Combining corner-touching polygons into a single feature.

The reason I ask is that this post: Merging polygons that share one vertex (node neighbors) using ArcGIS Desktop? suggests the user is trying to do exactly that and encountering errors. I’m posting it as a separate question here because my comment on that last post (poorly framed as a subsidiary question) was deemed invalid. Apologies in advance if I have not done the linking correctly in this question or I’ve deviated from protocol by referring to a deleted comment. I’m trying to improve my understanding of valid and invalid geometry.

Vince
  • 20,017
  • 15
  • 45
  • 64
Leigh Bettenay
  • 1,401
  • 8
  • 16
  • The only database that will refuse to store an 'hourglass polygon' (A.K.A. Bowtie polygon) is enterprise; shapefiles, file and personal geodatabase etc. will happily store invalid geometries: Self intersecting, self overlapping, incorrect ring orientation etc are considered non simple geometries. As you might imagine the first step when receiving data from an untrustworthy source is to validate the geometries as these non simple geometries cause havoc with overlay operators (clip, erase, select by location etc). What software are you working with? – Michael Stimson May 15 '19 at 02:26
  • It's not invalid for a multipart polygon to have parts that touch at a point. Single part polygons can even have a hollow inclusion (which ArcGIS will accept; QGIS requires multipart). – Vince May 15 '19 at 05:03
  • I'm using QGIS and Global Mapper but my question is about shapefiles rather than software. If I understand what you are both saying, a "bow-tie" polygon is perfectly OK but would be considered a "non-simple geometry". It is not coming from an untrustworthy source but being created by a merge of two polygons with a common point. If I have this right then I'm answered and will close this. – Leigh Bettenay May 15 '19 at 06:16

0 Answers0