Atlas Search: Unable to Tessellate shape

[Ikar_Pohorsky]

Hi all,

trying to create an atlas search index with following definition:

{
  "mappings": {
    "dynamic": false,
    "fields": {
      "geometry": {
        "indexShapes": true,
        "type": "geo"
      }
    }
  }
}

sadly the index build fails with following error:

Your index could not be built: Unexpected error: Unable to Tessellate shape [[49.47748652086282, -123.79858384565688] [49.47748631290387, -123.79833537632288] [49.47748610439212, -123.79808690974527] [49.47748589536565, -123.79783844040281] [49.477485761488396, -123.79767969655772] [49.47750369510057, -123.79761404608594] [49.47750367636922, -123.79758989051233] [49.47750365585305, -123.79756573214883] [49.477485610071895, -123.79750024279329] [49.477485475730475, -123.79734149894456] [49.4774852650837, -123.79709303235002] [49.4774850539222, -123.79684456299061] [49.477484914050514, -123.79667891676817] [49.47753880850618, -123.7965958535766] [49.47785707753759, -123.79659445750491] [49.47785658451906, -123.7964562714891] [49.47786416290044, -123.79858393595389] [49.47748652086282, -123.79858384565688] ]. Possible malformed shape detected.

The polygon looks like this: problematic_polygon.geojson · GitHub

I know we could try to fix the concrete polygon, but this is quite impossible for size of the dataset like ours - there will be many of such polygons.

Interesting fact is that the collection contains MongoDB geometry_2dsphere index and there is no problem in inserting the polygon into the MongoDB itself.

Maybe this is related to this bug report?

Unable to Tessellate shape error on indexing for valid geo shape

Any idea how to fix / workaround this?

[Marcus (Marcus Eagan)]

We will release a fix before autumn and I will update this ticket when we do. Thank you for reporting it. I need to test your coordinates to be absolutely sure, but that is my hunch. It can be difficult to eyeball collinearity. I can move it to a spreadsheet, though.

[Ikar_Pohorsky]

That sounds fantastic, thanks @Marcus!

It can be difficult to eyeball collinearity.

…that’s right, but my guess here is: if you manage to store the polygon in MongoDB collection with the geometry_2dsphere index then the polygon is valid

[Marcus (Marcus Eagan)]

@Ikar_Pohorsky

I believe the issue is fixed in the most recent release of Atlas Search but not documented. I’ll circle back with the team to change that fact.

Feel free to check it out. If it doesn’t work out, please let me know and I will investigate.

[Ikar_Pohorsky]

Will try when I have a chance and confirm. Hopefully next Monday.

[Ikar_Pohorsky]

It works! This is brilliant
Thank you so much