Does it compativle with MVT v2 format?

[nbulaj]

Hi. I have a PG 9.5 database with 2.3.2 PostGIS extension and a table with geo data, converted with ogr2ogr.

I'm trying to use dirt-simple-postgis-http-api as a backend for serving tiles for my map, but Mapbox GL JS can't render the vectors:

Vector tile source "source_name" layer "table_name" does not use vector tile spec v2 and therefore may have some rendering errors.

map.addSource('source', {
   'type': 'vector',
   'tiles': [
     "http://localhost:8123/mvt/v1/table_name/{z}/{x}/{y}?geom_column=wkb_geometry&limit=5000"
   ]
});

Am I doing something wrong or dirt-simple-postgis-http-api does not support MVT v2 format? Thanks

[tobinbradley]

I think that warning message is just GL JS whinging a bit but everything is fine. Here's a working example. Since you're not getting outright errors, my guess is the layer is being loaded but isn't getting styled.

I'm using geojson-vt to render vector tiles at the moment. There is a slightly newer version available (2.4 vs 2.3), so you might try updating that dependency to see if the whinging goes away, but I don't see any indication that 2.4 updated the MVT format from the release notes.

Down the road I'll likely replace the geojson-vt bit with PostGIS 2.4's new asMVT function, as it's a tad faster and drops the geojson-vt dependency. I don't know what MVT version it outputs though, so that may not make the whinging go away either.

Hope that helps!

[nbulaj]

So you mean that this functionality can be replaced with just a PostGIS 2.4 toMVT function?

P.S. There are no problems with the styles, because I just try to replace current Mapbox tile storage to hosted (with dirt-simple-postgis-http-api + PG table with all the data). The data is correct - it was tested with the t-rex tileserver.

[tobinbradley]

I'm saying the code in that service would change from PostGIS->GeoJSON->geojson-vt to just PostGIS (using asMVT). The both routes end up at the same place, but the latter should be faster. See this write up, though it's using mapnik instead of geojson-vt.

I don't have enough information to provide insight into the issue you're seeing, but it sure smells like a styling problem, and I can't simulate the issue (see my previous example that works fine). Note I didn't say the data was wrong - I think it's probably fine. With vector tiles loading (source) and styling (layer) the data are two entirely different things; just specifying a source as per your example won't draw anything on the map.

[nbulaj]

I'm saying the code in that service would change from PostGIS->GeoJSON->geojson-vt to just PostGIS (using asMVT). The both routes end up at the same place, but the latter should be faster. See this write up, though it's using mapnik instead of geojson-vt.

Yep, that is what am I talking about.

I don't have enough information to provide insight into the issue you're seeing, but it sure smells like a styling problem, and I can't simulate the issue (see my previous example that works fine). Note I didn't say the data was wrong - I think it's probably fine. With vector tiles loading (source) and styling (layer) the data are two entirely different things; just specifying a source as per your example won't draw anything on the map.

Yes, I understand what do you mean. I forgot to say that there are no problems with the styles because it is a legacy stable app. I will do some research tomorrow to find the problem. Anyway thanks!

[nbulaj]

Hi @tobinbradley. I found the problem: there were wrong filters for mapbox. Thanks!

Btw, what are the plans about PostGis ST_AsMVT functionality for this project? I found it really helpful and working (cant say this about most of the other tile servers)

[tobinbradley]

Glad you figured it out!

My plan is to move the MVT service to st_asmvt before the end of the year. Hopefully that'll be enough time for people to move to PostGIS 2.4. I'd rather not have pre-2.4 and post-2.4 paths for that service if I can avoid it.

[tobinbradley]

Update - looks like they pushed a patch to vt-pbf allowing the specification of MVT version. I updated the dependency and I think I added the option correctly to the MVT route. Hopefully that'll get rid of the warning message.

[nbulaj]

It's a good news, I will check it tomorrow. I've tested tilestrata postgis mvt - it works much more faster (with ST_AsMVT), but returns all the features without simplification.