Rowan Winsemius

I try to be open to using both - have them both in the toolbox, and use the right one depending on the project requirements.

In my experience ESRI works very well when everything and everyone is using it, but the real world is messier than that.

And the other is big wow is what no licensing costs unlocks for certain environments/projects - cost can eventually become prohibitive in an esri-based setup.

Depends a bit on the desired use case as to what might wow them.
But for me one of the big ones of data interoperability - support for different formats, and connectability of different systems.

I haven't released the Claude implementation yet as it's pretty slow. It seems to generate a set of simplifications up front, and then you can send a subsequent query for the level of simplification you want which is then very fast. I need more time to poke at the paper.

I was reading a recently released paper last week on line simplification - no fancy videos though!
So I asked ClaudeAI to implement it for me and it did 🤯
drops.dagstuhl.de/storage/00li...

Shapely probably is the closest I think
shapely.readthedocs.io/en/stable/ma...

I previously arranged sponsorship through an ex-employer and it was easier than I thought it would be, so i'd encourage anyone who knows what value their org gets out of GDAL to give it a go and have that conversation or write that paper, the worst that can happen is they say no!

So good to see a government dept getting behind this - wild to think yours might be the first given the vast majority would rely on GDAL in a significant fashion either directly or indirectly.

GitHub - rowanwins/point-in-polygon-hao: A point in polygon library based on the paper "Optimal Reliable Point-in-Polygon Test and Differential Coding Boolean Operations on Polygons" by Hao A point in polygon library based on the paper "Optimal Reliable Point-in-Polygon Test and Differential Coding Boolean Operations on Polygons" by Hao - rowanwins/point-in-polygon-hao

Just released an update to my point-in-polygon library.
Now smaller and faster, but most importantly more accurate with robust floating point maths thanks to the excellent robust-predicates library by @agafonkin.com

github.com/rowanwins/po...

To be fair the actual netcdf part wasn't too bad, it was post-processing we did that was the hard part, particularly with the World Ocean Database, which is a massive point dataset of water data used to generate the Atlas raster.

A similar product is the World Ocean Atlas from NOAA - it has info like water temperature at diff depths over time.
It's netcdf and I once wrote a horribly long python script to download and process it 😂
www.ncei.noaa.gov/products/wor...

That is the most detailed anyone had ever digitised a palm tree 🌴 !

Published a long overdue update to my js sweepline-intersections library to make use of the excellent robust-predicates library for improved intersection detection.
github.com/rowanwins/sw...
This library is downloaded >100k times a week thanks to TurfJS which is a tad scary when doing a release 😅

🎉 awesome. FWIW there is no perfect solution for this stuff, lots of pros and cons with diff formats, but my general experience is that flatgeobuf is a bit easier to use, esp for non-developers, but hopefully with time geoparquet support will improve & APIs will get simpler.

For anyone interested in displaying a flatgeobuf file efficiently on a maplibre-gl / mapbox-gl map I've created a little library that helps make tiled requests to your flatgeobuf datasource, this is particularly handy if your datasource is very large.
github.com/rowanwins/ma...

Looks like a spatial index creation is opt-out via GDAL, so geopandas probably follows suite.
gdal.org/en/latest/dr...

I'm not sure you can create a flatgeobuf file that doesn't have a spatial index

I believe yes, but that's a guess... If it's loading quickly in qgis that's probably a good sign that it has

Depending on intended use flatgeobuf is perhaps a smidge easier to use in a web browser. Geoparquet requires some more complicated dependencies, but I'm probably about 1 year behind in my last muck about with it so things may have improved.

Either flatgeobuf or geoparquet would be pretty good options, both support getting data in a bounding box, and have gdal support

In the good ol' days we had an #augischat at a sensible time, most weeks probably had 10-20ish participants. I'm a bit less spatial at the moment but happy to contribute a thumbs up for what works in Aus are doing 😂

This is a neat idea! Did you any particular use case in mind? One that came to mind for me was crowd-sourcing validation tasks, or a pre-check on geometry edits (eg hey these changes look wildly different, are you sure you want to proceed)