Visualization going Red on Zoom

[nyjr10025]

Running into an issue where the screen is going red when zooming into the map. No styles have been applied. Has happened consistently with multiple data sets

[nickkieffer]

Makes geo viz nearly unusable for me

[mentin]

Could you share any of the datasets that cause such result?

One hypothesis, judging by the fact that red starts at antimeridian in the image above - there are some inverted polygons in the dataset. Converted to GeoJson used by GeoViz they are represented by +-180/+-90 box, with the polygon you do want to show as a hole. I could not reproduce this with the polygons I've tried though. So having a dataset to repro would be helpful.

[nickkieffer]

Yep here ya go bquxjob_7d574d42_18b67e06b84.json

[mentin]

Thanks!

For some reasons I don't understand, I don't get that kind of fill in GeoViz. But I can confirm the polygons are inverted (if they are supposed to describe small blocks in Virginia):

select 
  st_area(st_geogfromtext(geom, oriented=>TRUE)) area, 
  st_area(st_geogfromtext(geom)) area2,
  geom
from  tmp.gvbug

Results I get:

Row | area | area2 | geom |  
-- | -- | -- | -- | --
1 | 510066073116955.81 | 1032.7522888034407 | POLYGON((-79.9924959415828 37.7939704994434, -79.9923042878136 37.7941072727986, -79.9919581354936 37.7938100261923, -79.9921564035582 37.793670066394, -79.9924959415828 37.7939704994434)) |  
2 | 510066073117479.56 | 508.98898554590954 | POLYGON((-79.9921195896996 37.794239082534, -79.992024820132 37.7943067140553, -79.9916787100328 37.7940072747658, -79.9917752115521 37.7939391538722, -79.9921195896996 37.794239082534)) |  
3 | 510066073117735.62 | 252.93059403002817 | POLYGON((-79.9919262704538 37.7943770429917, -79.9915845989301 37.794073722444, -79.9915913848041 37.7940689180758, -79.9916303878571 37.7940413852345, -79.9919753481127 37.7943420190847, -79.9919262704538 37.7943770429917)) |  
4 | 510066073117728.12 | 260.51798355690715 | POLYGON((-79.9919753481127 37.7943420190847, -79.9916303878571 37.7940413852345, -79.9916787100328 37.7940072747658, -79.992024820132 37.7943067140553, -79.9919753481127 37.7943420190847)) |  
5 |  510066073116963.5 |   1025.10061000995 | POLYGON((-79.9924959415828 37.7939704994434, -79.9921564035582 37.793670066394, -79.9923494443263 37.7935337962376, -79.9926909882889 37.7938313030569, -79.9924959415828 37.7939704994434))

The area of the polygons is almost whole globe - so the red fill of the whole globe is correct. I wonder why it does not happen in my browser - this might be an actual bug.

I suggest fixing the polygons - you are likely getting invalid analysis results any way. When loading the geospatial data BigQuery uses oriented semantics. ST_GeogFromText uses unoriented semantics (smaller of two interpretations) by default. Another option is to subtract (ST_Difference) already loaded geography from the whole globe (ST_GeogFromText('fullglobe')).

[nickkieffer]

Hmm thank you for taking a look and letting me know about my geometries! Extra weird is that the correct polygons look to be where they should be but I just can't zoom to them because of this other backwards polygon coming in at just after this zoom level

[nickkieffer]

They show in the data view as all wonky too I guess, should've noticed that 😬

[nickkieffer]

FWIW inverting my geometries fixed the issue I was seeing. Sorry for the sass and thanks for the help!