spData | Datasets for spatial analysis | Map library

You should be able to as long as you can express it in your query.

https://docs.microsoft.com/dotnet/machine-learning/how-to-guides/load-data-ml-net

One other way to do it would be to execute a query with your stored procedure and map the results to an IEnumerable. Then, load the IEnumerable into an IDataView.

https://docs.microsoft.com/dotnet/machine-learning/how-to-guides/load-data-ml-net#load-data-from-other-sources

It looks like the issue is with the use of summarise() instead of mutate(); when you use summarise() you only keep the variable of interest, e.g.

Interesting question. As you point out, in normal use it is unlikely that one would need to display multiple compasses for the same map and that is probably why the default behavior of the tmap library does not handle this case.

That said, it is still possible to add all five tmap compasses to the same map using some workarounds! So, please find below the general "strategy":

1. Building 5 maps, each with one of the tmap compasses. Then convert these maps into 'grob' objects using the tmap::tmap_grob() function to extract the compasses with the help of the getGrob() function from the base R library grid.

2. Visualizing the compasses (without and with labels) using the cowplot library

3. Building the final map with the five compasses using the cowplot library

NB: when running the reprex just below, don't worry about the rendering of the different plots that will be displayed in your plotting device (as the rendering depends on the aspect ratio of the device); what matters is the rendering of the maps saved in the .png files.

STEP 1 - EXTRACT EACH COMPASS TYPE FROM FIVE 'DUMMY' MAPS AS 'GROB' OBJECTS

This is probably because you have a version or raster that is ahead of the version of terra that it requires, caused by a temporary glitch at CRAN with the binary versions for Windows. The easiest way around this problem is to install the development version of terra, and then reinstall raster:

Can share more about what you're trying to accomplish? You say you want to remove whitespace, but you also want the plot to be at the bottom left of whitespace. What is your ultimate goal here? Let me know if I missed something.

I believe the whitespace is added outside of your ggplot based on your graphic device and ggplot margins. You can control the ggplot margins in theme() (like in this question/answer). You could remove the whitespace by defining your graphic device to the same dimensions as your ggplot object.

If you wanted to place the plot at the bottom left of whitespace, you could define your graphic device space and then plot within that area with one of your favorite plot layout tools. For example:

It turns out that I am just an idiot. exact_extract() expects the functions to be quoted because they are run internally and not taken from the external functions. This solved the issue entirely:

Drawing a shared border is kind of tricky in context of {sf} > 1.0, as it uses the s2 dependency for spherical operations, and the s2 library introduced a new concept of semi-closed polygons (which was not there in the GEOS years).

See https://r-spatial.github.io/s2/reference/s2_options.html#model for more information.

Anyhow, consider this piece of code, built on {giscoR} package to access EU NUTS regions, and sf::st_intersection() to find the shared border. Note the use of model = "closed" (i.e. all polygons contain their entire boundaries) which may not be immediately obvious, and is necessary for the code to work as intended.

If you make the weights sum to 1, the model converges.

The developers intentionally disabled computing the marginals to make the model faster.

To enable it, you can add these to the inla arguments: