Add continuous benchmarking scripts (#207)

* Add touchstone infrastructure, WIP #197

* Add benchmarking setup and script, WIP #197

* Update OS, R version, RSPM source

.github/workflows/touchstone-receive.yaml

@@ -1,3 +1,4 @@
+# Modified to follow workflow in EpiNow2
 name: Continuous Benchmarks (Receive)
 
 concurrency:
@@ -6,17 +7,20 @@ concurrency:
 
 on: 
   pull_request:
+    paths:
+      # Where model code is defined: R/, src/, inst/include/
+      - "inst/include/**"
+      - "src/**"
+      - "R/**"
+      # Benchmarking code, config file, and manual trigger
+      - "touchstone/**"
+      - ".github/workflows/touchstone-*.yaml"
+      - ".benchmark"
 
 jobs:
   prepare:
-    runs-on: ubuntu-latest 
-    if:
-      true &&
-      (
-        github.event.comment.author_association == 'OWNER' ||
-        github.event.comment.author_association == 'MEMBER' ||
-        github.event.comment.author_association == 'COLLABORATOR' 
-      )
+    runs-on: ubuntu-latest
+    # Allow benchmarking workflow to run for external contribs
     outputs:
       config: ${{ steps.read_touchstone_config.outputs.config }}
     steps:
@@ -52,5 +56,4 @@ jobs:
           cache-version: 1
           benchmarking_repo: ${{ matrix.config.benchmarking_repo }}
           benchmarking_ref: ${{ matrix.config.benchmarking_ref }}
-          benchmarking_path: ${{ matrix.config.benchmarking_path }} 
-          force_upstream: true
+          benchmarking_path: ${{ matrix.config.benchmarking_path }}

touchstone/.gitignore

@@ -4,3 +4,4 @@
 !.gitignore
 !header.R
 !footer.R
+!setup*.R

touchstone/config.json

@@ -1,5 +1,5 @@
 {
     "os": "ubuntu-22.04",
     "r": "4.4.0",
-    "https://packagemanager.rstudio.com/all/__linux__/jammy/latest"
+    "rspm": "https://packagemanager.rstudio.com/all/__linux__/jammy/latest"
 }

touchstone/script.R

@@ -1,30 +1,71 @@
 # see `help(run_script, package = 'touchstone')` on how to run this
 # interactively
 
-# TODO OPTIONAL Add directories you want to be available in this file or during the
-# benchmarks.
+# TODO OPTIONAL Add directories you want to be available in this file or during
+# the benchmarks.
 # touchstone::pin_assets("some/dir")
 
 # installs branches to benchmark
 touchstone::branch_install()
 
-# benchmark a function call from your package (two calls per branch)
+#### Simple final size calculation on uniform population ####
+# run for 1000 values of R0
+# using iterative solver
 touchstone::benchmark_run(
-  # expr_before_benchmark = source("dir/data.R"), #<-- TODO OTPIONAL setup before benchmark
-  random_test = yourpkg::f(), #<- TODO put the call you want to benchmark here
-  n = 2
+  expr_before_benchmark = {
+    source("touchstone/setup.R")
+  },
+  default_iterative = {
+    lapply(r0_samples, final_size, solver = "iterative")
+  },
+  n = 100
 )
 
-# TODO OPTIONAL benchmark any R expression (six calls per branch)
-# touchstone::benchmark_run(
-#   more = {
-#     if (TRUE) {
-#       y <- yourpkg::f2(x = 3)
-#     }
-#   }, #<- TODO put the call you want to benchmark here
-#   n = 6
-# )
+# using newton solver
+touchstone::benchmark_run(
+  expr_before_benchmark = {
+    source("touchstone/setup.R")
+  },
+  default_newton = {
+    lapply(r0_samples, final_size, solver = "newton")
+  },
+  n = 100
+)
 
+#### Final size calculation on heterogeneous population ####
+# run for 1000 values of R0
+# using iterative solver
+touchstone::benchmark_run(
+  expr_before_benchmark = {
+    source("touchstone/setup.R")
+  },
+  complex_iterative = {
+    lapply(
+      r0_samples, final_size,
+      contact_matrix = contact_matrix,
+      demography_vector = demography_vector,
+      susceptibility = susceptibility,
+      p_susceptibility = p_susceptibility,
+      solver = "iterative"
+    )
+  },
+  n = 100
+)
 
-# create artifacts used downstream in the GitHub Action
-touchstone::benchmark_analyze()
+# using newton solver
+touchstone::benchmark_run(
+  expr_before_benchmark = {
+    source("touchstone/setup.R")
+  },
+  complex_newton = {
+    lapply(
+      r0_samples, final_size,
+      contact_matrix = contact_matrix,
+      demography_vector = demography_vector,
+      susceptibility = susceptibility,
+      p_susceptibility = p_susceptibility,
+      solver = "newton"
+    )
+  },
+  n = 100
+)

touchstone/setup.R

@@ -0,0 +1,46 @@
+library(finalsize)
+library("socialmixr")
+library("withr")
+
+#### Set up population characteristics ####
+# load contact and population data from socialmixr::polymod
+polymod <- socialmixr::polymod
+contact_data <- socialmixr::contact_matrix(
+  polymod,
+  countries = "United Kingdom",
+  age.limits = seq(0, 80, 10), # final age bin is 70+
+  symmetric = TRUE
+)
+
+# get the contact matrix and demography data
+contact_matrix <- t(contact_data$matrix)
+demography_vector <- contact_data$demography$population
+
+# scale the contact matrix so the largest eigenvalue is 1.0
+contact_matrix <- contact_matrix / max(Re(eigen(contact_matrix)$values))
+
+# divide each row of the contact matrix by the corresponding demography
+contact_matrix <- contact_matrix / demography_vector
+
+n_demo_grps <- length(demography_vector)
+
+# get 1000 R0 samples
+r0_mean <- 1.5
+r0_sd <- 0.01
+samples <- 1000
+r0_samples <- withr::with_seed(1, rnorm(1000, r0_mean, r0_sd))
+
+# define susceptibility values for age groups, greater for older ages
+susc_variable <- matrix(
+  data = seq_along(demography_vector) / length(demography_vector)
+)
+susceptibility <- cbind(
+  susc_variable, susc_variable * 0.8, susc_variable * 0.6,
+  susc_variable * 0.5
+)
+
+# define proportion of each demographic group in each susceptibility group
+# assume uniform distribution
+n_demo_grps <- nrow(contact_matrix)
+p_susceptibility <- matrix(1.0, nrow = n_demo_grps, ncol = ncol(susceptibility))
+p_susceptibility <- p_susceptibility / rowSums(p_susceptibility)