add time series and comment code

lib/ClimaCoreMakie/examples/file.jl

@@ -52,84 +52,149 @@ sproj = [
  "+proj=sinu",
 ]
 
+# Downloading and unpacking time series needs to happen manually from:
+# https://caltech.box.com/shared/static/a65jnfdfpwqt9ka9nszvoedcg412c6q1.gz
+
+time_series = map(readdir(joinpath(pwd(), "time-series"); join = true)) do file
+ reader = ClimaCore.InputOutput.HDF5Reader(file)
+ return ClimaCore.InputOutput.read_field(reader, "diagnostics")
+end
+field = time_series[1].temperature
+nlevels = ClimaCore.Spaces.nlevels(axes(field))
+# Calculate a global color range per field type, so that the colorrange stays static when moving time/level.
+color_range_per_field = Dict(
+ map(propertynames(time_series[1])) do name
+ vecs = [
+ extrema(
+ hcat(
+ vec.(
+ parent.(
+ ClimaCore.level.(
+ (getproperty(slice, name),),
+ 1:nlevels,
+ )
+ )
+ )...,
+ ),
+ ) for slice in time_series
+ ]
+ name => reduce(
+ ((amin, amax), (bmin, bmax)) ->
+ (min(amin, bmin), max(amax, bmax)),
+ vecs,
+ )
+ end,
+)
+
 begin
  fig = Figure(resolution = (1600, 800))
+ ####
+ ### Setup the grid layout for the menu and different plots
+ ####
+
+ # One slot for menus + toggles + colorbar
  toggles = fig[1, 1] = GridLayout()
  plots = fig[2, 1] = GridLayout()
- colgap!(plots, 0)
- projection = Observable(sproj[1])
- fields = collect(string.(propertynames(diagnostics)))
+ colgap!(plots, 0) # leave no space between plots
 
- toggles[1, 1] = Label(fig, "earth overlay:", tellwidth = true)
- toggles[1, 2] = overlay_toggle = Toggle(fig, active = true)
- toggles[1, 3] =
+ # Define a few observables we need to create entries, which get connected to the widgets later
+ projection = Observable(sproj[1])
+ fields = collect(string.(propertynames(time_series[1])))
+ # Two sliders for setting Time + Label
+ toggles[1, :] =
+ sliders = SliderGrid(
+ fig,
+ (label = "Time", range = 1:length(time_series)),
+ (label = "Level", range = 1:nlevels),
+ )
+ time_slider, level_slider = sliders.sliders
+ # Need a sub gridlayout, to get the columns aligned with the sliders
+ menu_items = toggles[2, :] = GridLayout()
+ # Use labels to label toggle + menu
+ menu_items[1, 1] = Label(fig, "earth overlay:", tellwidth = true)
+ menu_items[1, 2] = overlay_toggle = Toggle(fig, active = true)
+
+ menu_items[1, 3] = Label(fig, "field:", tellwidth = true)
+ menu_items[1, 4] =
  field_selector = Menu(fig, options = fields, default = "temperature")
- toggles[1, 4] =
- dest_proj = Menu(fig, options = sproj, default = projection[])
 
+ menu_items[1, 5] = Label(fig, "projection:", tellwidth = true)
+ menu_items[1, 6] =
+ dest_proj = Menu(fig, options = sproj, default = projection[])
+ # Have a global color range we
  colorrange = Observable((0.0, 1.0))
-
- toggles[2, :] =
+ toggles[3, :] =
  level_s = Colorbar(
  fig,
  colorrange = colorrange,
  colormap = :balance,
  vertical = false,
  )
 
+ # Create a geo axis to handle the projection
+ # Note, that Makie still needs some work in the event propagation of menu + friends,
+ # so when selecting e.g. a field, it may happen that it changes the axis zoom level
+ # use ctrl + double mouse click to reset it.
  plots[1, 1] = ax = GeoAxis(fig; dest = projection)
+ # Create a toggable land overlay
  earth_overlay =
  poly!(ax, GeoMakie.land(), color = (:white, 0.2), transparency = true)
  translate!(earth_overlay, 0, 0, 10)
  connect!(earth_overlay.visible, overlay_toggle.active)
-
+ # create a 3D plot without an axis for mapping the simulation on the sphere
  plots[1, 2] = ax2 = LScene(fig; show_axis = false)
 
- colsize!(plots, 1, Relative(3 / 4))
-
+ colsize!(plots, 1, Relative(3 / 4)) # Give the GeoAxis more space
 
+ # Now, connect all the observables to the widgets
  on(dest_proj.selection) do proj
  projection[] = proj
  end
 
- field_observable = Observable{Any}(diagnostics.temperature)
- on(field_selector.selection; update = true) do fieldname
- # Don't block menu when updating
- @async begin
- field_observable[] = getproperty(diagnostics, Symbol(fieldname))
- end
+ field_observable = Observable{Any}()
+ #=
+ map!(f, new_observable, observable) works like:
+ new_observable = map(f, observable)
+ which works like:
+ # run everytime observable updates
+ on(observable) do new_value
+ new_observable[] = new_value
  end
-
- on(field_observable; update = true) do field
- @async begin
- colorrange[] =
- extrema(hcat(vec.(parent.(ClimaCore.level.((field,), 1:n)))...))
- end
+ =#
+ map!(
+ field_observable,
+ field_selector.selection,
+ time_slider.value,
+ ) do fieldname, idx
+ return getproperty(time_series[idx], Symbol(fieldname))
+ end
+ # update=true, runs f immediately, otherwise f only runs when the input observable is triggered the first time
+ on(field_selector.selection; update = true) do field_name
+ # select the correct color range from the globally calcualted color ranges
+ colorrange[] = color_range_per_field[Symbol(field_name)]
+ return
  end
 
- field = field_observable[]
- n = ClimaCore.Spaces.nlevels(axes(field))
-
- toggles[3, :] = height_slider = Slider(fig, range = 1:n)
+ # manually create the mesh for the 3D Sphere plot:
 
- field_slice = ClimaCore.level(T, 1)
+ field_slice = ClimaCore.level(field, 1)
  space = axes(field_slice)
-
  a, b, c = ClimaCore.Spaces.triangulate(space)
  triangles = GLTriangleFace.(a, b, c)
  cf = ClimaCore.Fields.coordinate_field(space)
  long, lat = vec.(parent.((cf.long, cf.lat)))
  vertices = Point2f.(long, lat)
 
- # plot level at slider
+ # plot level at slider, needs to be any since the type changes (also the reason why we use map! instead of map)
  field_slice_observable = Observable{Any}()
  map!(
  ClimaCore.level,
  field_slice_observable,
  field_observable,
- height_slider.value,
+ level_slider.value,
  )
 
+ # extract the scalar field
  scalars = map(field_slice_observable) do field_slice
  Float32.(vec(parent(field_slice)))
  end