Allow to enable/disable automatic detection of numeric variables

NEWS.md

@@ -4,6 +4,9 @@
 * Add `group()`, `ungroup()`, `group_levels()`, `group_names()`, `group_indices()`, `group_rows()`, `group_length()`, `group_size()`, `group_extract()`, `group_split()` and `is_grouped()` to work with grouped data.
 * Add `transform_lr()`, `transform_clr()`, `transform_alr()`, `transform_ilr()`, `transform_plr()` and `transform_inverse()` methods for `GroupedComposition` and `GroupedLogRatio` objects.
 
+## Enhancements
+* `as_composition()` gained a new `autodetect` argument to enable/disable automatic detection of numeric variables.
+
 ## Breaking changes
 * Redesign the internal mechanism for grouping data.
 * `hist()` now produces a single histogram.

R/AllGenerics.R

@@ -19,10 +19,12 @@ NULL
 #' Coerces an object to a `CompositionMatrix` object.
 #' @param from A [`matrix`] or [`data.frame`] to be coerced.
 #' @param parts A `vector` giving the index of the column to be used a
-#'  compositional parts. If `NULL` (the default), all [`double`] columns will be
-#'  used.
+#'  compositional parts. If `NULL` and `autodetect` is `TRUE` (the default),
+#'  all `numeric` columns will be used.
 #' @param groups An [`integer`] giving the index of the column to be used to
 #'  group the samples. If `NULL` (the default), no grouping is stored.
+#' @param autodetect A [`logical`] scalar: should `numeric` variables be
+#'  automatically used as compositional parts?
 #' @param verbose A [`logical`] scalar: should \R report extra information
 #'  on progress?
 #' @param ... Currently not used.

R/coerce.R

@@ -22,6 +22,9 @@ setMethod(
   f = "as_composition",
   signature = c(from = "matrix"),
   definition = function(from) {
+    ## Validation
+    arkhe::assert_type(from, "numeric")
+
     ## Make row/column names
     lab <- make_names(x = NULL, n = nrow(from), prefix = "S")
     rownames(from) <- if (has_rownames(from)) rownames(from) else lab
@@ -41,7 +44,7 @@ setMethod(
 setMethod(
   f = "as_composition",
   signature = c(from = "data.frame"),
-  definition = function(from, parts = NULL, groups = NULL,
+  definition = function(from, parts = NULL, groups = NULL, autodetect = TRUE,
                         verbose = getOption("nexus.verbose")) {
     ## Clean row/column names
     lab <- make_names(x = NULL, n = nrow(from), prefix = "S")
@@ -50,12 +53,16 @@ setMethod(
 
     ## Remove non-numeric columns
     if (is.null(parts)) {
-      parts <- arkhe::detect(from, f = is.double, margin = 2) # Logical
-      if (isTRUE(verbose)) {
-        n <- sum(parts)
-        what <- ngettext(n, "Found %g part (%s)", "Found %g parts (%s)")
-        cols <- paste0(colnames(from)[parts], collapse = ", ")
-        message(sprintf(what, n, cols))
+      if (isTRUE(autodetect)) {
+        parts <- arkhe::detect(from, f = is.numeric, margin = 2)
+        if (isTRUE(verbose)) {
+          n <- sum(parts)
+          what <- ngettext(n, "Found %g part (%s)", "Found %g parts (%s)")
+          cols <- paste0(colnames(from)[parts], collapse = ", ")
+          message(sprintf(what, n, cols))
+        }
+      } else {
+        arkhe::assert_filled(parts)
       }
     } else {
       if (is.numeric(parts)) parts <- seq_len(ncol(from)) %in% parts

vignettes/groups.Rmd

@@ -40,7 +40,7 @@ When coercing a `data.frame` to a `CompositionMatrix` object, **nexus** allows t
 data("bronze", package = "folio")
 
 ## Use the third column (dynasties) for grouping
-coda <- as_composition(bronze, groups = 3)
+coda <- as_composition(bronze, parts = 4:11, groups = 3)
 ```
 
 `groups(x)` and `groups(x) <- value` allow to retrieve or set groups of an existing `CompositionMatrix`. Missing values (`NA`) or empty strings can be used to specify that a sample does not belong to any group.