mechanism to clear cache?

[cboettig]

Is there a way to adjust the cache expiration for data urls?

In general, in maplibre data resources come from an external url, which can make it a bit more difficult to update maps dynamically. If I understand correctly, the typical solution (provided we can update the remote data source) is to expire the cache so that the new data loads, i.e. set

'Cache-Control: 'max-age=0'

(perhaps as discussed in mapbox/mapbox-gl-js#2633 (comment) ? I'm a bit out of my depth on this one).

[crazycapivara]

@cboettig In most case you do not need to update your data with Map.set_data but use filters instead which makes it very comfortable:

m = Map(...)
m.add_layer(...)

# Update your layer using a filter on the 'growth' property of your source
filter_expression = [">=", ["get", "growth"], 0.5] # only show features where growth >= 0.5

m.set_filter("your_layer_id", filter_expression)

Furthermore, the upcoming release comes with seamless support for geopandas data frames for vector data sources.

See also changelog for unreleased version.

[cboettig]

oh fantastic! native support for geopandas is much simpler than cache exploration.

Thanks for mentioning the filters, I need to spend more time exploring the performance implications there. Do you know if maplibre js is able to execute those as a range request without transferring all the data? For the moment I'm relying on duckdb to do my filters, it's quite good at filtering 150gb parquet file in seconds, and figured I that would be difficult purely client-side. But in at least some cases I'm probably better off using the filters.

That filter syntax is really nice too. @giswqs -- is something like m.set_filter() possible in leafmap.maplibre wrapper? Guess that feels a bit more pythonic than putting it all in the style dict (though I guess something about configuring the map entirely with a dict is also kinda elegant in it's own way).

[crazycapivara]

@cboettig Yes, range requests are possible:

["all", [">=", ["get", "growth"], 0.2], ["<=", ["get", "growth"], 0.5]]

In the next release I also added some helper functions for creating these kind of expressions / filters as you can see in this example.

In general, for large datasets it is good practice to use url sources, because maplibre only loads / queries the features in the current viewport and for the filters set, it never loads all of the data.