vector_tiler null output

[avanetten]

vector_tiler is giving null output when tiling a geojson using the tiles created by raster_tiler

[nrweir]

It's possible that `VectorTiler` isn't set up to split linestrings, only polygons. Might need to be checked/implemented if it's not.

…

On Mon, Jul 15, 2019 at 12:28 PM Adam Van Etten ***@***.***> wrote: vector_tiler is giving null output when tiling a geojson using the tiles created by raster_tiler — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#197?email_source=notifications&email_token=ADDLNIYO3NKYH4MRW44K2UDP7SQSVA5CNFSM4IDYS75KYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4G7IQFDQ>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADDLNI7UDSEM7TDDU2PGQHLP7SQSVANCNFSM4IDYS75A> .

[daveluo]

First off, much thanks for the Solaris toolkit and kudos on the initial release...really helpful!

I ran into this same problem of getting null outputs from VectorTiler. It worked as expected with the examples from the tutorials but encountered when I tried tiling my own data samples.

At least in my case, the problem seems to stem from my raster input not being in epsg:4326. When raster_tiler.tile_bounds is generated by RasterTiler, they contain non-4326 coordinates like (537722.4375, 9350190.01551792, 537782.2604974508, 9350249.838515371)

When passing this into vector_tiler.tile(tile_bounds=), clip_gdf() isn't able to find the bounds to clip the input geojson/gpd.

I confirmed and fixed the problem by converting the generated raster_tiler.tile_bounds to 4326 first and then passing that into vector_tiler.tile().

Here's a Colab notebook showing the issue and my quick fix: https://nbviewer.jupyter.org/gist/daveluo/7c54d0fbb1d1f32307d2af1a5b640254#Make-tiles-for-znz029

[nrweir]

Thanks a lot for doing the diagnosis and exploration @daveluo! This may have been fixed with #201 - if you feel like checking, you could pip install the version on the dev branch and see if it fixes it.

I'll dig into it myself later tonight and make sure everything's resolved.

[daveluo]

Thanks for the heads up! I pip installed the dev branch but the latest changes did not fix the non-4326 tile_bounds issue.

[nrweir]

OK thanks @daveluo, I'll check back into it in a little bit.

[nrweir]

Alright yeah, there wasn't anything written into the code to handle cases where the raster and vector have different CRSs. I've added an argument to VectorTiler.tile() and VectorTiler.tile_generator(), tile_bounds_crs, where the user can specify the EPSG code for the raster CRS; solaris will check if it differs from the vector CRS and transform the bounds as needed. See #203.

Thanks for bringing this to our attention @daveluo!

[nrweir]

Resolved with #203.