Geomtric errors being created whilst editing in ArcSDE 9.3.1/Oracle 10g environment

[TimLangner]

Hi there

I have been digitising in ArcGIS 10.0.3 and earlier, with the data stored in a geodatabase in ArcSDE 9.3.1, hosted on an Oracle 10g database.

Errors are being created in the geometries of polygons, whenever I do the below steps. I sent some sample data to ESRI (UK), who were able to reproduce the same error themselves. I would be interested to know if anyone else has experienced this or other geometry issues, when working with ArcSDE and Oracle.

Steps taken:
1. I digitise a polygon, polygon 1, which is only made up of straight line segments.
2. I digitise 3 polygons of varying sizes using the ellipse tool.
3. Selecting each of the 3 ellipse polygons in turn, I use the clip tool to create 3 holes inside polygon 1.
4. Finally I should be left with 4 polygons which do not overlap

However doing these steps is causing geometric errors in the polygon I am clipping. I am getting incorrect ring order. I also have other errors in my feature class, such as incorrect segment orientation and self intersections.

Although not reproduced in the same sample I sent ESRI, I have been able to make large polygons change shape complete, when clipping them using a variety of polygons, some with a mixture of arcs and straight lines and others either ellipse polygons and or circle polygons. If the polygon shape doesn't change completely, then the boundaries do not align. Even tracing around the edges doesn't help because I've tried that too.

As I am working with an archive attached to my feature class, I cannot export the data to a file geodatabase, to correct the errors and then reimport it back in, since complete archives cannot be exported.

If anyone else has any experiences of this or anything similar, I'd be grateful if they could post them here because it might help aid finding a solution to the problem.

Kind regards

Tim

[TimLangner]

ESRI Inc. in America have issued a bug number for this problem. It is NIM074901.

I'm not sure if this covers the self intersections geometry errors I was getting or just the incorrect ring ordering. I'm stil in contact with ESRI(UK) Inc. so if I get anything else I will post it here.

Kind regards

Tim

[WeifengHe]

The bug number you got covers the ring orientation problem.

Can you please specify the self-intersecting problem you have encountered?

[TimLangner]

Thanks for your reply.

I did raise this issue with UK support who are looking into it. However some examples of what I did to cause this are:

A)
A road, split up into a variety of feature types, required amending based on newer aerial photography. One section required to have a curve put in. From memory I did the following:1. Firstly I tried to reshape a road feature by creating a new curve but it  kept stating it wasn't possible to reshape the feature once I went to finish the reshape;
2. Therefore I changed tact and created the grass feature I wished to create, including creating a curve using the End Point Arc Segment tool;
3. At this stage I may have adding some additional points to the  curve I had just created, in order to make the polygon curve segment transition to the straight  line segments smoother. However this may have been done further down in the  process, once the road had then modified to match the grass feature.
4. After doing that I then selected the road;
5. Now I selected the reshape tool and the trace tool;
6. Clicking on the road feature I wished to reshape, I then traced the edge  of the grass. This now worked with no error generated. Where as at first I was getting the reshape error.

Before working on this road, there was no self intersections error.

Since then it has transpired that for another feature at a different site, the following error message I was getting when tracing, was due to a self intersecting feature:
"The Reshape task could not be completed. An internal error has occurred in the geometry system."

B)
Another example of generating a self intersecting feature was another road:
With this all I believe I did was move a point using the following process:
1. An adjacent polygon, which was some grass, required to be moved down. This I did by moving it's vertices;
2. Then I either move two vertices of the polygon it now overlapped or clipped it to that polygon;
3. Finally I moved one point from the polygon which now has the self intersections error;

I do have some screenshots but I'm not sure I should post them for data license and copyright reasons.

This lead to the following reply, which partly came from Redlands. They referred me to the following article:
http://support.esri.com/en/knowledgebase/techarticles/detail/29465

Note the article states it's for older versions than the version I'm using.

"If the points you have digitised are taken strictly at face value there does not appear to be any overlap or intersection. But you have chosen a Tolerance of 0.001 for this Feature Class so the difference between the points is at, or very close to, that value. Therefore the points can be interpreted as being in the same place.

Is there really a feature on the ground at this location that is just a couple of mm long that you are trying to represent? If not then the answer is to use the Check Geometry as a warning of poor (although not strictly incorrect) digitising and correct the defect. If this is a genuine feature then you need to use a smaller tolerance in order to allow this."

This lead to looking into the use of topology to perhaps try and prevent the situation arising. Whilst I believe that could stop me being able to create the conditions that cause the self intersection error to be created, I don't feel it will actually fix the reason it can be created.

Of course I can't do a check geometry on an SDE feature class and I have an archive attached to my data, so repairing these geometric errors isn't a trivial process. That;s due to the fact a complete archive can't be exported.

It since transpires that the xy resolution of the polygon feature class was set to: 0.000000002219847
The data originally came from an external supplier and I was unaware that the resolution wasn't set to the default 0.0001, which the other point and line feature classes are. I don't know if this impacts on the issue at all.

However if I'm moving points, I would expect them to snap whatever resolution is used, as they are in the same feature class and I have all the snapping turned on.

I also have a large number of polygons where I'd traced existing features in the same feature class and another feature and these have quite a few self intersection errors, where as the data they were traced from does not. That may also involved the moving of some points.

Kind regards

[TimLangner]

Thanks for your reply.

I did raise this issue with UK support who are looking into it. However some examples of what I did to cause this are:

A)
A road, split up into a variety of feature types, required amending  based on newer aerial photography. One section required to have a curve  put in. From memory I did the following:1. Firstly I tried to reshape a  road feature by creating a new curve but it  kept stating it wasn't  possible to reshape the feature once I went to finish the reshape;
2. Therefore I changed tact and created the grass feature I wished to  create, including creating a curve using the End Point Arc Segment tool;
3. At this stage I may have adding some additional points to the  curve  I had just created, in order to make the polygon curve segment  transition to the straight  line segments smoother. However this may  have been done further down in the  process, once the road had then  modified to match the grass feature.
4. After doing that I then selected the road;
5. Now I selected the reshape tool and the trace tool;
6. Clicking on the road feature I wished to reshape, I then traced the  edge  of the grass. This now worked with no error generated. Where as at  first I was getting the reshape error.

Before working on this road, there was no self intersections error.

Since then it has transpired that for another feature at a different  site, the following error message I was getting when tracing, was due to  a self intersecting feature:
"The Reshape task could not be completed. An internal error has occurred in the geometry system."

B)
Another example of generating a self intersecting feature was another road:
With this all I believe I did was move a point using the following process:
1. An adjacent polygon, which was some grass, required to be moved down. This I did by moving it's vertices;
2. Then I either move two vertices of the polygon it now overlapped or clipped it to that polygon;
3. Finally I moved one point from the polygon which now has the self intersections error;

I do have some screenshots but I'm not sure I should post them for data license and copyright reasons.

This lead to the following reply, which partly came from Redlands. They referred me to the following article:
http://support.esri.com/en/knowledgebase/techarticles/detail/29465

Note the article states it's for older versions than the version I'm using.

"If the points you have digitised are taken strictly at face value there  does not appear to be any overlap or intersection. But you have chosen a  Tolerance of 0.001 for this Feature Class so the difference between the  points is at, or very close to, that value. Therefore the points can be  interpreted as being in the same place.

Is there really a feature on the ground at this location that is just a  couple of mm long that you are trying to represent? If not then the  answer is to use the Check Geometry as a warning of poor (although not  strictly incorrect) digitising and correct the defect. If this is a  genuine feature then you need to use a smaller tolerance in order to  allow this."

This lead to looking into the use of topology to perhaps try and prevent  the situation arising. Whilst I believe that could stop me being able  to create the conditions that cause the self intersection error to be  created, I don't feel it will actually fix the reason it can be created.

Of course I can't do a check geometry on an SDE feature class and I have  an archive attached to my data, so repairing these geometric errors  isn't a trivial process. That;s due to the fact a complete archive can't  be exported.

It since transpires that the xy resolution of the polygon feature class was set to: 0.000000002219847
The data originally came from an external supplier and I was unaware  that the resolution wasn't set to the default 0.0001, which the other  point and line feature classes are. I don't know if this impacts on the  issue at all.

However if I'm moving points, I would expect them to snap whatever  resolution is used, as they are in the same feature class and I have all  the snapping turned on.

I also have a large number of polygons where I'd traced existing  features in the same feature class and another feature and these have  quite a few self intersection errors, where as the data they were traced  from does not. That may also involved the moving of some points.

Kind regards

[TimLangner]

I just wish to add that if I get an error saying "A geometry error has occurred" when reshaping a feature, then fairly often I will just draw the reshape and merge it with the existing feature. This usually solves gets round the reshape problem but whether it deals with the issue that caused the "A geometry error has occurred" message is another matter. Possibly not.

Kind regards

Tim