Hi @justinv, every vertex to be removed seems to be on edges of rectangular shapes. If my observation was correct, the Generalizer (Douglas algorithm) might help you. I tested tolerance = 1e-4, but you will have to determine the optimal tolerance according to your requirement.

Thanks, Takashi.

Yes, the generalizer works very well for those mostly rectangular features. I have been using this transformer but I am looking for something with a little more control over what gets removed and what doesn't. Sometimes even the extra vertices along edges are legitimate and I don't want them removed. I have point files with attributes that make it easy for me to separate the good from the bad. The Generalizer can't take this into account. Previously I excluded the polygons that had these legitimate vertices or had more complex geometry, like the area in the southeast (rivers, coastlines, etc.), ran the Generalizer, and then went back and removed the extra vertices later. It leaves a lot of cleanup.

This feature class has about 3,000 features and is manageable. I am working with feature classes with over 2 million features.

At this point, am seeing a combination of the two for a solution. Perhaps run the vertex cleaner. Then run the Generalizer on the failed feature classes that are rectangular. Then manually clean up the rest. I think FME can do it all though.

Thanks again.

I found that each boundary line of the failed polygon starts (and ends) with a middle point on an edge, and also a point read from "DeleteTheseVertices" feature type matches the start/end points. Result, such a boundary will be an unclosed line after removing the start/end nodes, then the PointConnector outputs them via the Line port.

Probably you could restore the unclosed lines to polygons by closing them with LineCloser.

Hi @takashi, I see what you are saying. The failing features are missing the last closing point because it was removed in the change detector. I am wondering if somehow I can just duplicate the first point and add the value of the last point plus 1. Perhaps I could do that with a list.

I'll give it a shot.

I was able to get this to work by adding another start/end point to all of the features. Before the PointConnector, I added a StatisticsCalculator grouped by the shell attribute to get the min and max of the _count. Since the points are sorted, a DuplicateRemover leaves the first/last point. Then the _count gets calculated to the max plus one. One more sort on the _count gets everything ready for the point connector. This creates an extra start/end point for every feature, but I think that doesn't matter since the point connector is grouped by the shell.

Thank you your replies @takashi.