Depend on juliabridge, which juliaready was renamed to

DESCRIPTION

@@ -18,17 +18,16 @@ Imports:
     checkmate,
     distspec (>= 0.2.0),
     ggplot2,
-    juliaready,
+    juliabridge,
     posterior,
     rlang,
     stats
 Suggests:
-    withr,
     knitr,
     rmarkdown,
     testthat (>= 3.2.0)
 Remotes:
-    sbfnk/juliaready
+    epiforecasts/juliabridge
 VignetteBuilder: knitr
 Roxygen: list(markdown = TRUE)
 RoxygenNote: 7.3.3

NAMESPACE

@@ -5,13 +5,12 @@ S3method(as_draws_df,cidm_fit)
 S3method(as_draws_list,cidm_fit)
 S3method(as_draws_matrix,cidm_fit)
 S3method(as_draws_rvars,cidm_fit)
+S3method(as_julia_value,dist_spec)
 S3method(plot,cidm_fit)
 S3method(predict,cidm_fit)
-S3method(print,cidm_component)
 S3method(print,cidm_fit)
 S3method(print,cidm_nuts)
-S3method(simulate,cidm_julia)
-S3method(simulate,cidm_model)
+S3method(simulate,julia_spec)
 S3method(summary,cidm_fit)
 export(AR)
 export(Ascertainment)
@@ -42,9 +41,9 @@ export(Weibull)
 export(as_julia)
 export(cidm_available)
 export(cidm_setup_julia)
-export(component)
 export(fit)
 export(julia)
+export(julia_spec)
 export(nuts)
 export(truncated)
 importFrom(distspec,Beta)
@@ -55,6 +54,10 @@ importFrom(distspec,LogNormal)
 importFrom(distspec,NonParametric)
 importFrom(distspec,Normal)
 importFrom(distspec,Weibull)
+importFrom(juliabridge,as_julia)
+importFrom(juliabridge,as_julia_value)
+importFrom(juliabridge,julia)
+importFrom(juliabridge,julia_spec)
 importFrom(posterior,as_draws_array)
 importFrom(posterior,as_draws_df)
 importFrom(posterior,as_draws_list)

NEWS.md

@@ -23,9 +23,21 @@
 * `fit(model, y, method = nuts())` takes a numeric vector of observations and
   replaces `nuts_sampler()` with `nuts()`. Pathfinder initialisation has been
   removed.
-* `epiaware_call()` has been replaced by `component()` and `julia()`.
+* `epiaware_call()` has been replaced by `julia_spec()` and `julia()`.
 * `plot(type = "posterior")` has been removed; use bayesplot on
   `posterior::as_draws_array(fit)` instead.
+* The Julia setup dependency is now
+  [juliabridge](https://github.com/epiforecasts/juliabridge), which juliaready
+  was renamed to. An install from GitHub picks it up through `Remotes`; an
+  existing checkout needs the dependency reinstalled.
+* `julia_spec()`, `julia()` and `as_julia()` come from juliabridge, which
+  composableIDModelR re-exports. The code that renders a component to Julia
+  source moved there so that other packages can build Julia calls the same
+  way. The role of a component is given as `.role`, leaving every Julia
+  keyword name free; components carry classes such as `julia_latent` and
+  `julia_model`; and one prints as `<julia model spec>`. A component built
+  without a role is accepted wherever one is expected, since the role is what
+  the caller says the component is.
 
 ## New features
 
@@ -43,6 +55,11 @@
 * `plot()` gains `type = "infections"`, `type = "latent"` and a `horizon`
   argument for forecasts.
 
+## Bug fixes
+
+* Forecasting from a fit whose Julia session has gone says so, rather than
+  setting Julia up again and then failing inside Julia.
+
 # EpiAwareR 0.2.0
 
 ## New Features

R/dist-convert.R

@@ -36,7 +36,7 @@
 #'
 #' @param x A distspec distribution or a composableIDModelR component.
 #' @param arg_name Name used in error messages.
-#' @return A component of class `cidm_distribution`.
+#' @return A component of class `julia_distribution`.
 #' @keywords internal
 .as_prior <- function(x, arg_name = deparse(substitute(x))) {
   if (!inherits(x, "dist_spec")) {
@@ -56,19 +56,19 @@
   if (!distribution %in% names(.julia_distributions)) {
     stop(
       "`", arg_name, "` is a ", distribution, " distribution, which cannot be ",
-      "used as a prior. Use a parametric distribution, or `component()` for a ",
-      "Distributions.jl distribution with no distspec equivalent.",
+      "used as a prior. Use a parametric distribution, or `julia_spec()` ",
+      "for a Distributions.jl distribution with no distspec equivalent.",
       call. = FALSE
     )
   }
   spec <- .julia_distributions[[distribution]]
   args <- lapply(spec$args(distspec::get_parameters(x)), as.numeric)
   prior <- do.call(
-    component, c(spec$fn, args, role = "distribution")
+    julia_spec, c(spec$fn, args, .role = "distribution")
   )
   max <- .dist_max(x)
   if (is.finite(max)) {
-    prior <- component("truncated", prior, -Inf, max, role = "distribution")
+    prior <- julia_spec("truncated", prior, -Inf, max, .role = "distribution")
   }
   prior
 }
@@ -191,9 +191,9 @@
   priors <- lapply(names(parameters), function(name) {
     prior <- .as_prior(parameters[[name]], arg_name)
     if (is.finite(bounds[[name]])) {
-      prior <- component(
+      prior <- julia_spec(
         "truncated", prior, as.numeric(bounds[[name]]), Inf,
-        role = "distribution"
+        .role = "distribution"
       )
     }
     prior
@@ -202,7 +202,9 @@
                unname(priors), D = as.numeric(max_delay))
   args[["\u0394d"]] <- if (!is.null(delta_d)) as.numeric(delta_d)
   list(
-    dist = do.call(component, c("UncertainDelay", args, role = "distribution")),
+    dist = do.call(
+      julia_spec, c("UncertainDelay", args, .role = "distribution")
+    ),
     max_delay = NULL, delta_d = NULL
   )
 }
@@ -260,7 +262,7 @@
   if (is.null(x)) {
     return(NULL)
   }
-  if (is.list(x) && !inherits(x, c("cidm_component", "dist_spec"))) {
+  if (is.list(x) && !inherits(x, c("julia_spec", "dist_spec"))) {
     return(lapply(seq_along(x), function(i) {
       .as_prior(x[[i]], paste0(arg_name, "[[", i, "]]"))
     }))

R/distributions.R

@@ -30,7 +30,7 @@
 #' @param lower,upper Numeric truncation bounds. Use `-Inf` or `Inf` to leave
 #'   a side unbounded.
 #'
-#' @return A component of class `cidm_distribution`.
+#' @return A component of class `julia_distribution`.
 #'
 #' @family components
 #' @name distributions
@@ -50,7 +50,7 @@ NULL
 # nolint start: object_name_linter.
 HalfNormal <- function(mean = 1) {
   checkmate::assert_number(mean, lower = 0, finite = TRUE)
-  component("HalfNormal", as.numeric(mean), role = "distribution")
+  julia_spec("HalfNormal", as.numeric(mean), .role = "distribution")
 }
 # nolint end
 
@@ -63,8 +63,8 @@ truncated <- function(dist, lower = -Inf, upper = Inf) {
   if (lower >= upper) {
     stop("`lower` must be less than `upper`.", call. = FALSE)
   }
-  component("truncated", dist, as.numeric(lower), as.numeric(upper),
-            role = "distribution")
+  julia_spec("truncated", dist, as.numeric(lower), as.numeric(upper),
+             .role = "distribution")
 }
 
 #' @importFrom distspec Normal

R/fit.R

@@ -154,7 +154,9 @@ fit <- function(model, y, method = nuts(), dates = NULL, seed = NULL) {
       y = y,
       dates = dates,
       method = method,
-      julia = .julia_handle(result$handle, result$session)
+      julia = juliabridge::julia_handle(
+        result$handle, result$session, .cidm_env
+      )
     ),
     class = "cidm_fit"
   )
@@ -190,9 +192,9 @@ fit <- function(model, y, method = nuts(), dates = NULL, seed = NULL) {
 #' }
 #' @importFrom stats simulate
 #' @export
-simulate.cidm_model <- function(object, nsim = 1, seed = NULL, n, ...) {
-  # An untyped `julia()` expression dispatches here too, as do typed ones of
-  # every role, so the role is checked rather than left to dispatch.
+simulate.julia_spec <- function(object, nsim = 1, seed = NULL, n, ...) {
+  # Every specification dispatches here, whatever role it carries and whether
+  # it carries one, so the role is checked here rather than left to dispatch.
   .assert_role(object, "model")
   checkmate::assert_count(nsim, positive = TRUE)
   checkmate::assert_count(n, positive = TRUE)
@@ -204,10 +206,6 @@ simulate.cidm_model <- function(object, nsim = 1, seed = NULL, n, ...) {
   .with_rt(.generated_list(result), object)
 }
 
-#' @rdname simulate.cidm_model
-#' @export
-simulate.cidm_julia <- simulate.cidm_model
-
 #' Convert bridge output to a draws_df
 #'
 #' @param values Matrix of draws (rows) by variables (columns).
@@ -253,7 +251,9 @@ simulate.cidm_julia <- simulate.cidm_model
 #'   as Julia code has no components to read, so it gets none.
 #' @keywords internal
 .with_rt <- function(generated, model) {
-  if (inherits(model, "cidm_julia")) {
+  # `IDModel()` passes both models positionally, so a specification with no
+  # positional arguments was written by hand and says nothing about infections.
+  if (inherits(model, "julia_code") || length(model$args) == 0) {
     return(generated)
   }
   infection <- model$args[[1]]

R/id-model.R

@@ -8,7 +8,7 @@
 #' @param observation An observation model, e.g. from
 #'   [NegativeBinomialError()].
 #'
-#' @return An object of class `cidm_model`.
+#' @return An object of class `julia_model`.
 #'
 #' @family components
 #' @examples
@@ -28,5 +28,5 @@
 IDModel <- function(infection, observation) { # nolint: object_name_linter.
   .assert_role(infection, "infection")
   .assert_role(observation, "observation")
-  component("IDModel", infection, observation, role = "model")
+  julia_spec("IDModel", infection, observation, .role = "model")
 }

R/infection-models.R

@@ -26,7 +26,7 @@
 #'   distribution, taken from the distribution's `max` when it has one.
 #' @param delta_d Numeric. Discretisation interval width.
 #'
-#' @return An object of class `cidm_infection`.
+#' @return An object of class `julia_infection`.
 #'
 #' @family components
 #' @name infection-models
@@ -52,14 +52,14 @@ Renewal <- function(generation_time, rt = NULL, initialisation = NULL,
                    max_name = "D_gen")
   rt <- .as_prior_slot(rt)
   initialisation <- .as_prior_slot(initialisation)
-  .assert_role(transformation, "julia", null_ok = TRUE)
+  .assert_role(transformation, "code", null_ok = TRUE)
   args <- list(
     generation_time = gen$dist, rt = rt,
     initialisation = initialisation, transformation = transformation,
     D_gen = if (!is.null(gen$max_delay)) as.numeric(gen$max_delay)
   )
   args[["\u0394d"]] <- if (!is.null(gen$delta_d)) as.numeric(gen$delta_d)
-  do.call(component, c("Renewal", args, role = "infection"))
+  do.call(julia_spec, c("Renewal", args, .role = "infection"))
 }
 
 #' @rdname infection-models
@@ -68,10 +68,10 @@ DirectInfections <- function(Z = NULL, initialisation = NULL,
                              transformation = NULL) {
   Z <- .as_prior_slot(Z)
   initialisation <- .as_prior_slot(initialisation)
-  .assert_role(transformation, "julia", null_ok = TRUE)
-  component(
+  .assert_role(transformation, "code", null_ok = TRUE)
+  julia_spec(
     "DirectInfections", Z = Z, initialisation = initialisation,
-    transformation = transformation, role = "infection"
+    transformation = transformation, .role = "infection"
   )
 }
 
@@ -81,10 +81,10 @@ ExpGrowthRate <- function(rt = NULL, initialisation = NULL,
                           transformation = NULL) {
   rt <- .as_prior_slot(rt)
   initialisation <- .as_prior_slot(initialisation)
-  .assert_role(transformation, "julia", null_ok = TRUE)
-  component(
+  .assert_role(transformation, "code", null_ok = TRUE)
+  julia_spec(
     "ExpGrowthRate", rt = rt, initialisation = initialisation,
-    transformation = transformation, role = "infection"
+    transformation = transformation, .role = "infection"
   )
 }