Uploaded image for project: 'Couchbase Server'
  1. Couchbase Server
  2. MB-52990

[FTS] Differing ES and FTS results for geospatial queries

    XMLWordPrintable

Details

    • Bug
    • Status: Closed
    • Critical
    • Resolution: Not a Bug
    • 7.1.2
    • None
    • fts
    • None
    • Untriaged
    • 1
    • No

    Description

      This ticket is a log of geospatial queries that return differing results with ES and FTS during functional testing. 

      The dataset for this testing is attached as `geoshape2.json`. The plots contain the query shape and the differing hits plotted as a GeoJSON file for convenient visualisation using geojson.io 

      1. A multipolygon 'intersects' query results in FTS returning 1 hit and ES returning none. 

      The executor logs showing the test failure are available on this link

      The query is in `query.json` and the plot is in `map(1).geojson`. 

       

      2. A polygon 'within' query results in FTS returning a point, but ES returns no hits.

      The executor logs showing the test failure are available here: http://qa.sc.couchbase.com/job/test_suite_executor/491922/console.

      The query is in `query2.json`. The plot is in in `map(2).geojson`. 

       

      3. A polygon 'intersects' query results in FTS returning an additional hit compared to the ES hits.

      The query is in `query3.json`. The plot is in `map(3).geojson`.

       

      4. A polygon 'within' query returning an additional hit in FTS compared to the ES hits.

      The query is in `query4.json`. The plot is in `map(4).geojson`.

       

      5. A polygon 'intersects' query returns 2 additional ES hits compared to FTS.

      The query is in `query5.json`. The plot is in `map(5).geojson`. 

       

      6. A polygon 'within' query returns an additional ES hit compared to the FTS hits.

      The query is in `query6.json`. The plot is in `map(6).geojson`. 

       

       

       

       

       

       

       

      Attachments

        1. geoshape2.json
          11 kB
        2. map (1).geojson
          3 kB
        3. map (2).geojson
          0.7 kB
        4. map (3).geojson
          0.5 kB
        5. map (4).geojson
          0.4 kB
        6. map (5).geojson
          0.5 kB
        7. map (6).geojson
          0.6 kB
        8. query.json
          9 kB
        9. query2.json
          2 kB
        10. query3.json
          1 kB
        11. query4.json
          1.0 kB
        12. query5.json
          1.0 kB
        13. query6.json
          1 kB
        14. run-1-plot-geodesic.png
          run-1-plot-geodesic.png
          174 kB
        15. run-3-plot-geodesic.png
          run-3-plot-geodesic.png
          182 kB
        16. run-4-plot-geodesic.png
          run-4-plot-geodesic.png
          211 kB
        17. run-5-plot-geodesic.png
          run-5-plot-geodesic.png
          268 kB
        18. run-6-plot.png
          run-6-plot.png
          668 kB
        No reviews matched the request. Check your Options in the drop-down menu of this sections header.

        Activity

          Aditi Ahuja, thanks for the details.

          One follow-up question,

          -Does all these differing hits in FTS are functionally incorrect as well according to the plottings?  If not, then a mention of that against each of the 6 runs would be helpful too.

           

          For one of the earlier erratic results with multiple polygon vertices along an edge, I have raised a Q&A topic here -

          https://github.com/google/s2geometry/discussions/263

           

          Sreekanth Sivasankaran Sreekanth Sivasankaran added a comment - Aditi Ahuja , thanks for the details. One follow-up question, -Does all these differing hits in FTS are functionally incorrect as well according to the plottings?  If not, then a mention of that against each of the 6 runs would be helpful too.   For one of the earlier erratic results with multiple polygon vertices along an edge, I have raised a Q&A topic here - https://github.com/google/s2geometry/discussions/263  

          Upon inspecting further, all these points are lying within or outside the geodesic boundary and hence FTS is returning the right results.

          Attaching the screenshots for them.

          Run1 geodesic plot => (point is inside/onboundary)

           

          Run 2 => is within the boundary as per the attached geojson plot itself.

          Run 3 geodesic plot => (point is inside)

           

          Run 4 geodesic plot => (point is inside)

           

          Run 5 geodesic plot => (point is outside)

          Run 6 => point is outside as per the geojson plot itself.

           

          As ES is not following geodesics for representing the edges, looks like we need to find out alternate tools for verifying the spatial indexing capability.

          Sreekanth Sivasankaran Sreekanth Sivasankaran added a comment - Upon inspecting further, all these points are lying within or outside the geodesic boundary and hence FTS is returning the right results. Attaching the screenshots for them. Run1 geodesic plot => (point is inside/onboundary)   Run 2 => is within the boundary as per the attached geojson plot itself. Run 3 geodesic plot => (point is inside)   Run 4 geodesic plot => (point is inside)   Run 5 geodesic plot => (point is outside) Run 6 => point is outside as per the geojson plot itself.   As ES is not following geodesics for representing the edges, looks like we need to find out alternate tools for verifying the spatial indexing capability.

          Sreekanth Sivasankaran  - Open since July, can you pls triage/update the fix version on this one or comment as appropriate, thanks! 

          prachi.sikchi Prachi Sikchi (Inactive) added a comment - Sreekanth Sivasankaran   - Open since July, can you pls triage/update the fix version on this one or comment as appropriate, thanks! 

          This isn't a bug and the feature is working as expected. The confusion arose as ES wasn't supporting geodesics and FTS spatial features are based on spherical earth or geodesics.

          So, no fix is planned/needed for the observations made in this ticket.

          Sreekanth Sivasankaran Sreekanth Sivasankaran added a comment - This isn't a bug and the feature is working as expected. The confusion arose as ES wasn't supporting geodesics and FTS spatial features are based on spherical earth or geodesics. So, no fix is planned/needed for the observations made in this ticket.

          People

            Sreekanth Sivasankaran Sreekanth Sivasankaran
            aditi.ahuja Aditi Ahuja
            Votes:
            0 Vote for this issue
            Watchers:
            3 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved:

              Gerrit Reviews

                There are no open Gerrit changes

                PagerDuty