Paul Ramsey

Two confusing things. None of the functions are marked as `PARALLEL SAFE`, I wonder why you get a parallel plan? The http calls happen in distinct worker processes, I wonder...

So, thank you for that gist. I tried it out, and it worked fine... parallel plan, and the 7-wide parallel query ran and returned 7 results nice and fast. Could...

For people interested in this topic, it's possible to get async HTTP queries by combining this extension with the [pg_background](https://github.com/vibhorkum/pg_background) extension.

No, looks like it is not possible. OGR FDW will to open options, but it seems like the only way to get at OGR_STYLE is vis the [OGRFeature::GetStyleString](http://www.gdal.org/classOGRFeature.html#a54c179e53eb6eddb657dd392a87cccab) method. That...

Interesting. This feels like it's drifting into a pgsql-mime extension. There's a lot of stuff inside once you start unwrapping this onion.

So, thinking about this and wondering what the API should look like... feels kind of like a key/value thing, maybe like the http_header[] tuple array? so optionally pass the data...

Probably you want the "structure" option for the [make valid parameters](http://libgeos.org/doxygen/geos__c_8h.html#acb2987c643bda31b1fcaecff8b62ce98). That should give a result that is "much like buffer(0)" without the failure modes.

@rsbivand I merged into main, so hopefully that is the last re-naming (sorry, did one more on you) you have to handle. Thanks for taking an early look at the...

This is kind of odd, there's been very little change in that code line, but there is one commit since 3.9, could this have had the effect you see? https://github.com/libgeos/geos/commit/e6c6fa39e32caf8bc15022cd17f47334a510197b#diff-97c48336edbb9a6f90de3e7b4424c22c8a97c9337631ac5cab30f8ff44ad6e3e

We can certainly strip NaN from WKT and WKB and GeoJSON inputs, but yeah, testing the underlying coordinates in a bulk buffer load could potentially get quite costly.