GeoParquet Downloader for QGIS

This repo contains a QGIS plugin for downloading GeoParquet data from cloud sources, including Overture Maps, Source Cooperative, and the ability to enter the location of any online GeoParquet file or partition. Just the user's current viewport then gets downloaded, as GeoParquet, DuckDB or GeoPackage.

The core idea is that GeoParquet can act more like a 'server', letting users download only the data they need, if you add a bit more smarts to the client. So this plugin uses DuckDB but abstracts all the details of forming the right queries to external sources, so users can just pick the data they want and pull it down with ease. And with GeoPackage output users don't even need to know anything about GeoParquet. More info is on the plugin homepage.

Installation

The easiest way to install the plugin file is to use the QGIS plugin manager. Just go to Plugins > Manage and Install Plugins, click the 'install' tab and search for 'GeoParquet Downloader'. Click on 'Install Plugin' and it will install. Alternatively you can download the zip file from one of the releases and 'install from zip' in QGIS.

The plugin needs the DuckDB Python package (1.1.0 or newer). Most more recent QGIS distributions should ship with it. If your qgis distro does not ship with it then QGIS should shows a message with an Install DuckDB button, which installs it with pip into QGIS's Python. This auto-install has been tricky to get right and difficult to test properly, so if it doesn't work for you please raise an issue and we'll try again.

One thing to try, if that install doesn't work (for example on Linux, where the system Python may refuse pip installs), you can use the QDuckDB plugin which includes a precompiled DuckDB binary. They also document how to install DuckDB on Linux and Mac OS/X.

See metadata.txt for more installation notes.

Usage

The plugin will install 1 button on the "Plugin" QGIS toolbar, that you might have to enable through View > Toolbars > Plugins:

It opens a dialog box, that lets you select Overture and Source Cooperative, OpenStreetMap or 'custom' - where you can enter the location of any GeoParquet or partition file online.

To use it move to an area where you'd like to download data and then select which layer you'd like to download. From there you can choose the output format (GeoParquet, GeoPackage, DuckDB, GeoJSON or FlatGeobuf) and the location to download the data to.

Downloads can sometimes take awhile, especially if the data provider hasn't optimized their GeoParquet files very well, or if you're downloading an area with a lot of data. Overture is one of the faster ones for now, others may take a minute or two. But it should most always be faster than trying to figure out exactly which files you need and downloading them manually.

By default the plugin downloads the current map view. Check the 'Area of Interest' box for more options. These include the ability to draw a polygon or a bounding box, or to use the features selected on a layer. By default Features that extend past the area, like a large lake or land cover polygon, are downloaded whole, but you can check 'Clip features to the area' to cut them at its edge.

If your QGIS doesn't have GeoParquet support you'll get a warning dialog after the data downloads completes. The GeoParquet will be there, but it won't automatically open on the map. We definitely recommend getting your QGIS working with GeoParquet, as the format is faster and handles nested attributes better. See Installing GeoParquet Support in QGIS for more details.

Contributing

This plugin has been made entirely with AI coding tools (primarily Cursor with claude-3.5-sonnet in the early days, claude with Opus in recent days). Contributions are very welcome, both from more experienced python developers who can help clean up the code and add missing features, and from anyone who wants a place to do AI-assisted coding that (hopefully) actually gets widely used.

I'm interested in exploring open source collaboration in the age of AI coding tools, especially working with less experienced developers who'd like to contribute, so don't hesitate to jump in with AI-assisted pull requests.

And any help on ideas/feedback, documentation, testing, promoting, etc. is very welcome!