shrug-l: RE: ArcGIS Identity bug.....make a note of it.

[Tan, Yurong]

For those of you who are doing a lot of analyses using the ArcGIS
IDENTITY overlay arctool, here is an example you need to look out for.
I just received a confirmation on a BUG (CQ00260773) in the ArcGIS
identity arctool that can "mis-behave" and output more points than the
input data layer.  This bug will be fixed in ArcGIS 10.  If you
encounter this problem, my suggested work around is to go back to
ArcInfo command line IDENTITY tool.

==>ALWAYS CHECK IF THE NUMBER OF OUTPUT POINTS IS THE SAME AS THE INPUT
LAYER.  ==>IF NOT, THEN YOU MUST HIT THE BUG.

Here is the original question:
>I have a puzzle about the IDENTITY tool in ArcGIS 9.0 (SP3) in Windows
XP Prof. >in overlaying a point layer onto a polygon zip.  The tool
consistently 
>output more points than the input and I couldn't explain why.  This
should be >very straight forward processing and both the input files are
normal.  I >checked if it could be due to some hidden "multi-part"
points in the input >layer and all the input points were unique without
any multiparts.
>
>I also tried the input layers in both ArcSDE and shapefile formats and
it had >the same results.
>
....
>
>Basically, I used the (9.0) IDENTITY tool from ArcTool box to overly
the 
>point layer with the polygon layer.  I set the input point layer, the
overlay  >polygon layer, and the output layer as directed in ArcGIS 9.0.
In the >meanwhile, I also noticed that if I varied the tolerance
somewhat (0.001 M to >0.00000000001M), I could get a different number of
output points.  The issue is >not just for the points right on the
polygon boundaries, it also affects those >points that are very close
(e.g. it may be as small as 0.02 mm to a polygon >boundary) to the
polygon lines.
>
>I have to let you know that I did a couple tests using coverages of
those two >input layers using different tolerance values and I could not
replicate the >problem.  
>
>Many years ago (in 1989?), there was a heated discussion in the then
newly >emerged GIS-L discussion list on this particular issue.  If my
memory still >serves correctly, the implementation was that the point in
question would be >assigned to the polygon to the LEFT and it appears to
me that results using >coverages for the same identity processing could
confirm that.  Unfortunately, >I did not see this happening in the 9.0
ArcGIS.  
..

And here is the response from ESRI ...
>Yurong,
>
>A bug has been logged for this incident. This is a planned fix for
ArcGIS 10. >The bug number is CQ00260773. I would recommend contacting
us with that number >in the future to verify that it definitely is fixed
by that point.  
>
>The plan is to set ArcGIS to work as Workstation does work, by
selecting to the >left as you described. 

     

Yurong Tan, Ph.D., Program Analyst/GIS
PSSG - Field Unit of ADUSH for Policy and Planning
Department of Veterans Affairs
Tel:  352-374-6080 (x5382)
Fax: 352-374-6119
___________________________________
They all gave some and some gave all!