Ellipse Calculator Processor workaround in GeoEvent?
Yeah Attempting to use this for generating ellipses for lightning data as well, when we run this in ArcMap and compare the output from the processor -- getting two very different outputs. Thinking that there is a bug somewhere in the code for the GEO Ellipse Processor, also noticing that the linear unit is dependent on projected wkid for linear unit. We are using geographic 4326, then to projected 3857, then back to 4326--everything is in kilometer. Have started looking through the code on GitHub. See if anything pokes out at me. 
have tried 4326 with both of these--not even close.
Hey Brian -
First things first, let's make sure we're both looking at the same GitHub repository. The bug filed 02-Sept-2016 with Esri Support (BUG-000099124) was closed Dec-2017 with a remark 'Will Not Be Addressed'. Looking over its notes I see a link to an old Defense Solutions repo (no longer being maintained): solutions.arcgis.com/defense/templates/geoevent-extension
The repository currently being maintained is: defense-solutions-proofs-of-concept/blob/master/geoevent-components
I asked Hanoch Kalmanovich to compile a *.jar for me for the Ellipse Calculator (Defense Solutions is no longer maintaining pre-compiled JAR with the Git resources). He compiled the custom processor using the 10.5.0 GeoEvent SDK. I was able to add the JAR to a 10.6 release of GeoEvent Server without a problem.
When I attempted to reproduce the issue I found that ArcMap's GP Tool (Data Management > Table to Ellipse) produced consistent results with the custom processor I had added to GeoEvent Server. Here are my reproduction steps:
1) Create the following non-spatial table in a file geodatabase and run the 'Table to Ellipse' tool
- Notice that the coordinate values are in a state plane projected coordinate system.
- I chose a location in the northern part of Minnesota to emphasize elliptic distortion, if that was going to be a factor.
2) Publish the Ellipse feature class (above) as a feature service so I could add the feature to a web map for display.
3) Configure the following GeoEvent Service. I'll try to attach an XML of the GeoEvent Server configuration in case you want to take a look at input / output configurations or GeoEvent Definitions ... but GeoNet is making that difficult.
The purpose of this GeoEvent Service is to broadcast a received Point geometry to a stream layer, use the Ellipse Calculator to compute an area-of-uncertainty from received event attributes (semi minor axis, rotation angle, etc.) and then stream the Polygon geometry to a second stream layer. Notice that I have configured the processor to expect the input geometry in the state plane coordinate system appropriate for the area of interest and to use that same SPCS for the spatial analysis. (The output is GCS WGS 1984; the stream service would have projected the geometry to this SR in any case.)
4) Add the two stream services as new layer to the web map displaying the Ellipse feature service (created using ArcMap's GP Tool) ... then HTTP/POST the following JSON to the GeoEvent Server's input to see the geometries processed by GeoEvent overlaid on the web map.
[
{
"center_x": 697182.8,
"center_y": 289173.3,
"wkid": "3595",
"semimajor": 1000,
"semiminor": 375,
"rotation_angle": 60,
"id": "WKID_3595",
"geometry": {
"x": 697182.8,
"y": 289173.3,
"spatialReference": {
"wkid": 3595
}
}
}
]
Notice the coordinate values are expressed in the state plane projected coordinate system (WKID 3595). The only difference between this JSON data and the data in the non-spatial table is that the semimajor and semiminor values have been cut in half ... since the Ellipse Calculator wants the radius and the ArcMap's GP Tool takes the full diameter.
I allow the Ellipse Calculator to overwrite the received Point geometry with the polygon it computes.
Looking at the two ellipse polygon geometries in the web map, they align almost perfectly. Zooming in there is a discrepancy of about 12 meters where the two ellipse's don't align perfectly:
I'm going to chalk that difference up to projection of the geometries. The ellipse created by ArcMap's GP Tool was created with a native spatial reference appropriate for the area of interest (WKID 3595) whereas the ellipse created by GeoEvent Server's custom processor was calculated in that coordinate system, but then projected to GCS WGS 1984 for broadcast by the stream service. Both were then projected, on-the-fly, by the web map to Web Mercator for display.
Hanoch can dig into the Ellipse Calculator's implementation in GitHub if we still think there's a defect there. But it looks to me like the processor is working as its contributor intended. Please let me know if you disagree.
Cheers --
RJ
