Colab notebook deps

GeneralisedFilters/docs/literate.jl

@@ -55,39 +55,18 @@ end
 
 const SCRIPTJL = joinpath(EXAMPLEPATH, "script.jl")
 
-# The example's declared dependencies (Literate is build-only), so the notebook can install
-# everything it needs on Colab — including packages used only by included helper files.
-function example_packages()
-    deps = get(Pkg.TOML.parsefile(joinpath(EXAMPLEPATH, "Project.toml")), "deps", Dict())
-    return sort(filter(!=("Literate"), collect(keys(deps))))
-end
-
-# Notebook postprocess: use a generic Julia kernel (so Colab picks its default Julia
-# runtime, with no pinned version) and prepend a single Pkg.add for the example's
-# dependencies so the notebook runs on a fresh Colab runtime.
+# Notebook postprocess: only what can't be expressed as `#nb` source lines — a generic Julia
+# kernel (so Colab picks its default Julia runtime, with no pinned version) and dropping the
+# Documenter-only `#hide` lines Literate otherwise leaves in the notebook. The Colab
+# dependency setup lives in the example script as `#nb` lines.
 function prepare_notebook(nb)
     nb["metadata"]["kernelspec"] = Dict(
         "display_name" => "Julia", "language" => "julia", "name" => "julia"
     )
-    # `#hide` is a Documenter-only directive that Literate leaves in the notebook; drop
-    # those lines (e.g. the repo-relative paths that only work during the docs build).
     for cell in nb["cells"]
         cell["cell_type"] == "code" || continue
         cell["source"] = filter(line -> !endswith(rstrip(line), "#hide"), cell["source"])
     end
-    packages = example_packages()
-    if !isempty(packages)
-        setup = Dict(
-            "cell_type" => "code",
-            "execution_count" => nothing,
-            "metadata" => Dict(),
-            "outputs" => [],
-            "source" =>
-                ["import Pkg\n", string("Pkg.add([", join(repr.(packages), ", "), "])")],
-        )
-        idx = findfirst(cell -> cell["cell_type"] == "code", nb["cells"])
-        insert!(nb["cells"], something(idx, lastindex(nb["cells"]) + 1), setup)
-    end
     return nb
 end
 

GeneralisedFilters/examples/trend-inflation/script.jl

@@ -4,6 +4,21 @@
 # Watson, 2016) using GeneralisedFilters to define a heirarchical model for use in Rao-
 # Blackwellised particle filtering.
 
+#nb # Install dependencies so the notebook runs on a fresh Colab runtime. GeneralisedFilters
+#nb # and SSMProblems are added from the repo's main branch (registered versions may be too
+#nb # old, e.g. lack ReferenceTrajectory):
+#nb import Pkg, Downloads
+#nb Downloads.download(
+#nb     "https://raw.githubusercontent.com/TuringLang/SSMProblems.jl/main/GeneralisedFilters/examples/trend-inflation/Project.toml",
+#nb     "Project.toml",
+#nb )
+#nb Pkg.activate(".")
+#nb Pkg.add([
+#nb     Pkg.PackageSpec(; url="https://github.com/TuringLang/SSMProblems.jl", subdir="GeneralisedFilters", rev="main"),
+#nb     Pkg.PackageSpec(; url="https://github.com/TuringLang/SSMProblems.jl", subdir="SSMProblems", rev="main"),
+#nb ])
+#nb Pkg.instantiate()
+
 using GeneralisedFilters
 using SSMProblems
 using Distributions

SSMProblems/docs/literate.jl

@@ -54,39 +54,18 @@ end
 
 const SCRIPTJL = joinpath(EXAMPLEPATH, "script.jl")
 
-# The example's declared dependencies (Literate is build-only), so the notebook can install
-# everything it needs on Colab — including packages used only by included helper files.
-function example_packages()
-    deps = get(Pkg.TOML.parsefile(joinpath(EXAMPLEPATH, "Project.toml")), "deps", Dict())
-    return sort(filter(!=("Literate"), collect(keys(deps))))
-end
-
-# Notebook postprocess: use a generic Julia kernel (so Colab picks its default Julia
-# runtime, with no pinned version) and prepend a single Pkg.add for the example's
-# dependencies so the notebook runs on a fresh Colab runtime.
+# Notebook postprocess: only what can't be expressed as `#nb` source lines — a generic Julia
+# kernel (so Colab picks its default Julia runtime, with no pinned version) and dropping the
+# Documenter-only `#hide` lines Literate otherwise leaves in the notebook. The Colab
+# dependency setup lives in the example script as `#nb` lines.
 function prepare_notebook(nb)
     nb["metadata"]["kernelspec"] = Dict(
         "display_name" => "Julia", "language" => "julia", "name" => "julia"
     )
-    # `#hide` is a Documenter-only directive that Literate leaves in the notebook; drop
-    # those lines (e.g. the repo-relative paths that only work during the docs build).
     for cell in nb["cells"]
         cell["cell_type"] == "code" || continue
         cell["source"] = filter(line -> !endswith(rstrip(line), "#hide"), cell["source"])
     end
-    packages = example_packages()
-    if !isempty(packages)
-        setup = Dict(
-            "cell_type" => "code",
-            "execution_count" => nothing,
-            "metadata" => Dict(),
-            "outputs" => [],
-            "source" =>
-                ["import Pkg\n", string("Pkg.add([", join(repr.(packages), ", "), "])")],
-        )
-        idx = findfirst(cell -> cell["cell_type"] == "code", nb["cells"])
-        insert!(nb["cells"], something(idx, lastindex(nb["cells"]) + 1), setup)
-    end
     return nb
 end
 

SSMProblems/examples/kalman-filter/script.jl

@@ -1,7 +1,18 @@
 # # Kalman Filter
 #
 # This example implements a Kalman filter for a linear Gaussian state space model using the
-# SSMProblems interface. 
+# SSMProblems interface.
+
+#nb # Install dependencies so the notebook runs on a fresh Colab runtime. SSMProblems is
+#nb # added from the repo's main branch (the registered version may be too old):
+#nb import Pkg, Downloads
+#nb Downloads.download(
+#nb     "https://raw.githubusercontent.com/TuringLang/SSMProblems.jl/main/SSMProblems/examples/kalman-filter/Project.toml",
+#nb     "Project.toml",
+#nb )
+#nb Pkg.activate(".")
+#nb Pkg.add(Pkg.PackageSpec(; url="https://github.com/TuringLang/SSMProblems.jl", subdir="SSMProblems", rev="main"))
+#nb Pkg.instantiate()
 
 using AbstractMCMC
 using Distributions