Fix models repeating in legend on box plot

This commit adds a fix to the legend of the box plot where the legend
can display the same models multiple time. To prevent this, we exclude
the models in `model_names` when computing the single best model and
single worst model for the box plot.

Author: ph-kev

NEWS.md

@@ -31,6 +31,11 @@ julia> reordered_var.dims |> keys |> collect
  "long"
 ```
 
+## Bug fixes
+
+- Fix models repeating in legend of box plots by not considering the models in `model_names`
+  when finding the best and worst models
+
 v0.5.8
 ------
 

docs/src/visualize_rmse_var.md

@@ -4,10 +4,11 @@ Instead of computing summary statistics, it may be more helpful to plot a box pl
 heatmap. `ClimaAnalysis` provides the functions `plot_boxplot!` and `plot_leaderboard!`
 to help visualize the root mean squared errors (RMSEs) in a `RMSEVariable`.
 
-The function [`Visualize.plot_boxplot!`](@ref) makes a box plot for each
-category in the `RMSEVariable`. The best model and worst model and any other models in
-`model_names` are plotted. The category to find the best and worst model defaults to
-"ANN", but can be changed using the parameter `best_and_worst_category_name`.
+The function [`Visualize.plot_boxplot!`](@ref) makes a box plot for each category in the
+`RMSEVariable`. The best model and worst model and any other models in `model_names` are
+plotted. When finding the best and worst single models, any models in `model_names` will be
+excluded. The category to find the best and worst model defaults to "ANN", but can be
+changed using the parameter `best_and_worst_category_name`.
 
 The function [`Visualize.plot_leaderboard!`](@ref) makes a heatmap of the
 RMSEs between the variables of interest and the categories. The best model for each variable
@@ -102,4 +103,4 @@ CairoMakie.save("./assets/boxplot_and_leaderboard.png", fig)
 nothing # hide
 ```
 
-![box plot](./assets/boxplot_and_leaderboard.png)
+![box plot](./assets/boxplot_and_leaderboard.png)

ext/ClimaAnalysisMakieExt.jl

@@ -640,7 +640,8 @@ end
 Plot a Tukey style boxplot for each category in `rmse_var`.
 
 The best and worst single models are found for the category `best_and_worst_category_name`
-and are plotted on the boxplot. Additionally, any model in `model_names` will also be
+and are plotted on the boxplot. When finding the best and worst single models, any models in
+`model_names` will be excluded. Additionally, any model in `model_names` will also be
 plotted on the boxplot.
 
 The parameter `ploc` determines where to place the plot on the figure.
@@ -693,15 +694,19 @@ function Visualize.plot_boxplot!(
         whiskerlinewidth = 1,
     )
 
-    # Plotting best and worst model
+    # Delete any model in model_names to exclude them when finding best and worst models
+    rmse_var_delete =
+        ClimaAnalysis.Leaderboard._delete_model(rmse_var, model_names...)
+
+    # Find and plot best and worst models
     absolute_worst_values, absolute_worst_model_name =
         ClimaAnalysis.find_worst_single_model(
-            rmse_var,
+            rmse_var_delete,
             category_name = best_and_worst_category_name,
         )
     absolute_best_values, absolute_best_model_name =
         ClimaAnalysis.find_best_single_model(
-            rmse_var,
+            rmse_var_delete,
             category_name = best_and_worst_category_name,
         )
     Makie.scatter!(