Using FME Server 2016.0 (build 16177), a NeighborFinder fails to do its job properly. Not all features input features result in output features in any of the output ports. Testing this on FME Desktop 2016.0 (build 16177), the issue - of course - remains. Testing this on FME Desktop 2016.1 (20160419), the issue is no more. It seems to me that this is a bug in either the NeighborFinder or the FME build. However, as both 2016.0 and 2016.1 use the NeighborFinder version 14, the latter seems to be the only logical cause for this behaviour.
My question is this: Can I solve this problem without having to upgrade the FME Server version to a newer version? I need a workaround for this bug in FME 2016.0, using FME 2016.0.
Best answer by conch
Hi @bakkermans
I am using 2016.1 (build 20160419) and have found a similar bug. I have tried all manner of values in the Distance and Number of Neighbour to Find parameters and still have the same result. I also used the 'GeometryValidator' to clean the data but nothing improved in that I was still lacking features after the NeighborFinder.
To get around it and include all the features in subsequent processes used a 'Matcher' to compare the UnmatchedCandidates to the original input features. See attached image, it's pretty self explanatory. The image shows the operation on a small subset of data where my 441 missing features re-appear after the matching process. The full data set has over 1 million features and this is the only way I can assure all UmatchedCandidates truly get through to the next steps.
Hope this helps, ConCh.neighbor-bug.png
This post is closed to further activity.
It may be an old question, an answered question, an implemented idea, or a notification-only post.
Please check post dates before relying on any information in a question or answer.
For follow-up or related questions, please post a new question or idea.
If there is a genuine update to be made, please contact us and request that the post is reopened.
Can you eloborate a more on the circumstances in which this is happening?
If you have a reproducible case you could maybe try and workbench your way around it, but that would not guarantee it'll always work. If you have the option to upgrade your FME Server I would highly recommend that.
First I would inspect the data and find out exactly what features are missing.
To do that you can use the ChangeDetector (lead all features into the original and the features coming out of the NeighborFinder into the revised, and check only the ID field). The deleted output will be the features that were lost in the NeighborFinder. See if you can find out why.
If that does not resolve the case, I would submit a call to Support@Safe.com
The neighborfinder is using a Group By attribute and the input is set at Candidates Only. Only 1 neighbor is to be found and a maximum distance of 191m is applied.
I already reproduced the error by exporting all the input before the NeighborFinder to an .fss, and next create a workbench with only this input and the NeighborFinder. The result is the same.:
Using FME 2016.0, 351 features insert the NeighborFinder and only 131 (Matched) + 1 (UnmatchedCandidate) leave it. The other output ports have no output.
Using FME 2016.1, 351 features insert the NeighborFinder and 131 (Matched) + 224 (UnmatchedCandidate) leave it, making up for all inserted features. The other output ports have no output. Note: this is actually resulting in more output compared to input, is this possible?
Upgrading FME Server is something I only want to consider when other options are out, since we are working with the FME Server of a client.
I am using 2016.1 (build 20160419) and have found a similar bug. I have tried all manner of values in the Distance and Number of Neighbour to Find parameters and still have the same result. I also used the 'GeometryValidator' to clean the data but nothing improved in that I was still lacking features after the NeighborFinder.
To get around it and include all the features in subsequent processes used a 'Matcher' to compare the UnmatchedCandidates to the original input features. See attached image, it's pretty self explanatory. The image shows the operation on a small subset of data where my 441 missing features re-appear after the matching process. The full data set has over 1 million features and this is the only way I can assure all UmatchedCandidates truly get through to the next steps.
I am using 2016.1 (build 20160419) and have found a similar bug. I have tried all manner of values in the Distance and Number of Neighbour to Find parameters and still have the same result. I also used the 'GeometryValidator' to clean the data but nothing improved in that I was still lacking features after the NeighborFinder.
To get around it and include all the features in subsequent processes used a 'Matcher' to compare the UnmatchedCandidates to the original input features. See attached image, it's pretty self explanatory. The image shows the operation on a small subset of data where my 441 missing features re-appear after the matching process. The full data set has over 1 million features and this is the only way I can assure all UmatchedCandidates truly get through to the next steps.
Hope this helps, ConCh.neighbor-bug.png
Thank you for mentioning the workaround! However, since this issue was urgent we already moved on with the installation of a more recent version of FME Server.
