Identify, Select fail on geojson mapsources when property is null

[brentfraser]

At least I think that is the problem. To test use geojson files in https://github.com/geomoose/gm3-demo-data/pull/8

[theduckylittle]

Can you get me some more details here? I've tried corrupting a few files in order try to recreate. I tried:

• A null property value, that seems to just render as the string "null".
• Uneven properties (some features with some properties and some with others) and they seem to work okay.
• A false property value, again, just seems to render the string "false"

When I eliminate all properties, the feature does not render at all, which is maybe expected?

[brentfraser]

Hmm, yes I should have supplied more details. So here's what i did today:

1. load a GeoMoose site with a sketch layer with an upload tool.
2. Upload gm3-demo-data/demo/census/census_cities.geojson file
3. Do Identify on "Farmington" polygon
   • Note no feature is found (bad, I expected to find one feature and show properties) The other city polygons work as expected.

[brentfraser]

Looking at the census_cities.geojson in a text editor , all the properties for Farmington look ok (and no nulls), so I'm not sure why it fails.

[brentfraser]

Looks like it is bad (?) geometry for the holes in the Farminton polygon. Looking at the ogr2ogr command I used (https://github.com/geomoose/gm3-demo-data/tree/main/demo/census), I may need to specify -nlt MULTIPOLYGON

[theduckylittle]

Digging in a little, I'm seeing the queryVectorLayer function return no features from OpenLayer's forEachFeatureAtCoordinateDirect call. Since other features work, I don't believe there to be a mismatched projection issue, and could be due to the malformed Geometry. But it's a bug in OpenLayers that we would need to investigate (locally I have version 6.9.0 installed).

[brentfraser]

In the mean time, I'm OK with just saying to users "make sure your geometry is valid"

[brentfraser]

and I thought I had filed an Issue on "Can't do identify on point or line in Drawing and Markup". Maybe it's buried in another Issue...

[theduckylittle]

It looks as if we drop the -nlt Polygon from the creation script, all is well.

Should we create a separate MR to fix our data? I suspect that this is garbage-in-is-garbage-out and there won't be much of a way for us to assist the user in that case.

[brentfraser]

Yes, I need to fix it and do a PR.

[brentfraser]

Looking at the history of the files in gm3-demo-data, i had originally used MULTIPOLYGON, but after loading the data into GeoServer inserts were a problem.

[theduckylittle]

:thinking: That's interesting. I would think it would have an issue with mixed types (say Polygons and MultiePolygons) but usually a single time doesn't phase it.

[brentfraser]

It could have been GM was telling it to insert a POLYGON and GeoServer expected MULTIPOLYGON only? Maybe there is a "promote geometry" switch in GeoServer...