Tiger extension loading invalid geometries

[jnu]

I discovered a small number of invalid geometries among the data loaded with the Tiger CENSUS data loader. The place table contains 4 invalid entries, per ST_IsInvalid:

SELECT plcidfp, namelsad, St_IsInvalidReason(the_geom) reason
FROM tiger.place
WHERE St_IsInvalid(the_geom);

 plcidfp |       namelsad        |                 reason
 --------+-----------------------+---------------------------------------
 0427820 | Glendale city         | Holes are nested[-112.40957 33.529845]
 4740000 | Knoxville city        | Holes are nested[-84.082463 35.893129]
 1738570 | Joliet city           | Holes are nested[-88.074056 41.510616]
 3771940 | Wesley Chapel village | Holes are nested[-80.682658 34.989184]

There were 120 invalid geometries in the zcta5 table as well. Other tables appeared free from errors (state, county, cousub, tract, bg).

I've run St_MakeValid(the_geom) as a workaround. I don't believe these errors originated in my application, though I don't know for certain they are introduced by tiger extension, either.

[robe]

I'll take a look and see. The places table comes straight from tiger. No massaging is done on that, so I would guess those errors are coming from census.

I assume you are using 2022 data set? You can check the tiger.loader_variables if you are not sure.

Unfortunately I don't have the full set of states loaded, but I can add this step to the processing moving forward.

By chance do you have all states loaded or just: AZ,IL,NC,TN

I'm not seeing any invalids in the MA set, but perhaps there aren't any.

The zcta5 set I am a bit surprised that one has invalids in it as that I do process to split and across state boundaries and thought I had a make valid step in there, but perhaps not.