Enable distribution vectors in tibbles (#108)

* Enabled the inclusion of  vectors as columns in  data objects, which needs vctrs and tibble as Suggests dependencies

* add internal keyword to vec_proxy/vec_restore documentation

DESCRIPTION

@@ -36,11 +36,13 @@ Suggests:
     PoissonBinomial,
     revdbayes (>= 1.3.5),
     rmarkdown,
-    testthat (>= 3.0.0)
+    testthat (>= 3.0.0),
+    tibble,
+    vctrs
 VignetteBuilder: 
     knitr
 Encoding: UTF-8
 Roxygen: list(markdown = TRUE)
-RoxygenNote: 7.3.1
+RoxygenNote: 7.3.2
 Config/testthat/edition: 3
 Config/testthat/parallel: true

NAMESPACE

@@ -431,6 +431,8 @@ S3method(variance,ZINegativeBinomial)
 S3method(variance,ZIPoisson)
 S3method(variance,ZTNegativeBinomial)
 S3method(variance,ZTPoisson)
+S3method(vctrs::vec_proxy,distribution)
+S3method(vctrs::vec_restore,distribution)
 export(Bernoulli)
 export(Beta)
 export(Binomial)

NEWS.md

@@ -25,6 +25,8 @@
   using `match()` for distribution objects.
 - Added a `duplicated()` method which relies on the corresponding method for the `data.frame`
   of parameters in a distribution.
+- Enabled the inclusion of `distribution` vectors as columns in `tibble` data objects, see
+  `?vec_proxy.distribution` for further details and a practical example.
 - Fixed errors in notation of cumulative distribution function in the documentation of
   `HurdlePoisson()` and `HurdleNegativeBinomial()` (by @dkwhu in #94 and #96).
 - The `prodist()` method for `glm` objects can now also handle `family` specifications from

R/tibble-vctrs.R

@@ -0,0 +1,61 @@
+#' Methods for including distributions as vctrs in tibbles
+#'
+#' Methods for \code{\link[vctrs]{vec_proxy}} and \code{\link[vctrs]{vec_restore}}
+#' from \pkg{vctrs} in order to include \code{distribution} objects in
+#' \code{\link[tibble]{tibble}} objects.
+#'
+#' @details The methods for \code{\link[vctrs]{vec_proxy}} and
+#'   \code{\link[vctrs]{vec_restore}} from \pkg{vctrs} are needed so that
+#'   \code{distribution} objects can be included as a vector column in
+#'   (and extracted from) \code{\link[tibble]{tibble}} data frames.
+#'   \code{vec_proxy} simply adds the class \code{data.frame} which is the
+#'   actual underlying data structure used by \code{distribution} objects.
+#'   This way the number of rows etc. can be correctly determined. Conversely,
+#'   \code{vec_restore} strips off the additional \code{data.frame} class and
+#'   restores the original \code{distribution} classes. Users typically do not
+#'   need to call \code{vec_proxy} and \code{vec_restore} directly.
+#'
+#' @param x,to Objects inheriting from \code{distribution}.
+#' @param ... Currently not used.
+#'
+#' @return The `vec_proxy` method returns a `distribution` object which
+#'   additionally inherits of `data.frame` while the `vec_restore` method
+#'   restores the original `distribution` classes.
+#'
+#' @examples
+#' \dontshow{ if(!requireNamespace("tibble")) {
+#'   if(interactive() || is.na(Sys.getenv("_R_CHECK_PACKAGE_NAME_", NA))) {
+#'     stop("not all packages required for the example are installed")
+#'   } else q() }
+#' }
+#' ## Poisson GLM for FIFA 2018 goals data
+#' data("FIFA2018", package = "distributions3")
+#' m <- glm(goals ~ difference, data = FIFA2018, family = poisson)
+#' 
+#' ## Predict fitted Poisson distributions for teams with ability differences
+#' ## of -1, 0, 1 (out-of-sample) using the new data as a data.frame
+#' nd <- data.frame(difference = -1:1)
+#' nd$dist <- prodist(m, newdata = nd)
+#' nd
+#'
+#' ## Do the same using the new data as a tibble
+#' library("tibble")
+#' nt <- tibble(difference = -1:1)
+#' nt$dist <- prodist(m, newdata = nt)
+#' nt
+#'
+#' @keywords internal
+
+#' @rdname vec_proxy.distribution
+#' @exportS3Method vctrs::vec_proxy distribution
+vec_proxy.distribution <- function(x, ...) {
+  class(x) <- c(class(x), "data.frame")
+  return(x)
+}
+
+#' @rdname vec_proxy.distribution
+#' @exportS3Method vctrs::vec_restore distribution
+vec_restore.distribution <- function(x, to, ...) {
+  class(x) <- class(to)
+  return(x)
+}