Allow expressions in data_summary() that return > 1 column summaries (#673)

* Allow `data_summary()` to return expressions with > 1 rows

* typo

* fixes

* add tests

* fix for grouped data frames

* update comment

* add examples in docs

* fix, add tests

* address comments

* Update R/data_summary.R

Co-authored-by: Etienne Bacher <52219252+etiennebacher@users.noreply.github.com>

* whitespace

* comments

* rename `strict`

* use back ticks

* revise

* add missing snapshot

* ...

* docs

* add test

* automatic suffixes

* add comments

* clarify

* Update NEWS.md

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update R/data_summary.R

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* Update R/data_summary.R

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

* update RD

* allow named list in suffix

* wording

* minor

* update

* Update R/data_summary.R

Co-authored-by: Etienne Bacher <52219252+etiennebacher@users.noreply.github.com>

* update RD

* remove trailing whitespace

---------

Co-authored-by: Etienne Bacher <52219252+etiennebacher@users.noreply.github.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>

Author: 3 people

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: datawizard
 Title: Easy Data Wrangling and Statistical Transformations
-Version: 1.3.0
+Version: 1.3.0.1
 Authors@R: c(
     person("Indrajeet", "Patil", , "patilindrajeet.science@gmail.com", role = "aut",
            comment = c(ORCID = "0000-0003-1995-6531")),

NEWS.md

@@ -2,6 +2,11 @@
 
 CHANGES
 
+* `data_summary()` now allows expressions to return more than one summary
+  value. For each value, a new column is created. Additionally, the optional
+  `suffix` argument controls the naming of these columns; if `suffix = NULL`,
+  column names are auto-generated (e.g., with numeric suffixes).
+
 * `standardize()` now works on `fixest` estimations (#665).
 
 # datawizard 1.3.0

R/data_summary.R

@@ -11,6 +11,20 @@
 #' @param remove_na Logical. If `TRUE`, missing values are omitted from the
 #' grouping variable. If `FALSE` (default), missing values are included as a
 #' level in the grouping variable.
+#' @param suffix Optional, suffixes to be added to the new variable names,
+#' especially useful when a function returns several values (e.g. `quantile()`).
+#' Can be:
+#' * a character vector: all expressions in `...` must return the same number
+#'    of values as elements in `suffix`.
+#' * a list of named character vectors: the names of elements in `suffix` must
+#'    match the names of the expressions. It is also allowed to specify suffixes
+#'    for selected expressions only.
+#'
+#' The new column names are a combination of the left-hand side (i.e.,
+#' the name) of the expression and the related suffixes. If `suffix = NULL` (the
+#' default), and a summary expression returns multiple values, either the names
+#' of the returned values (if any) or automatically numbered suffixes such as
+#' `_1`, `_2`, etc. are used. See 'Examples'.
 #' @param ... One or more named expressions that define the new variable name
 #' and the function to compute the summary statistic. Example:
 #' `mean_sepal_width = mean(Sepal.Width)`. The expression can also be provided
@@ -50,15 +64,72 @@
 #'   last = mpg[length(mpg)],
 #'   by = c("am", "gear")
 #' )
+#'
+#' # allow more than one-column-summaries for expressions
+#' d <- data.frame(
+#'   x = rnorm(100, 1, 1),
+#'   y = rnorm(100, 2, 2),
+#'   groups = rep(1:4, each = 25)
+#' )
+#'
+#' # since we have multiple columns for one expression, the names of the
+#' # returned summary results are used as suffix by default
+#' data_summary(
+#'   d,
+#'   quant_x = quantile(x, c(0.25, 0.75)),
+#'   mean_x = mean(x),
+#'   quant_y = quantile(y, c(0.25, 0.5, 0.75))
+#' )
+#'
+#' # if a summary function, like `fivenum()`, returns no named vector, suffixes
+#' # are automatically numbered
+#' data_summary(
+#'   d,
+#'   quant_x = quantile(x, c(0.25, 0.75)),
+#'   mean_x = mean(x),
+#'   fivenum_y = fivenum(y)
+#' )
+#'
+#' # specify column suffix for expressions, matching by names
+#' data_summary(
+#'   d,
+#'   quant_x = quantile(x, c(0.25, 0.75)),
+#'   mean_x = mean(x),
+#'   quant_y = quantile(y, c(0.25, 0.5, 0.75)),
+#'   suffix = list(quant_y = c("_Q1", "_Q2", "_Q3"))
+#' )
+#'
+#' # name multiple expression suffixes, grouped by variable
+#' data_summary(
+#'   d,
+#'   quant_x = quantile(x, c(0.25, 0.75)),
+#'   mean_x = mean(x),
+#'   quant_y = quantile(y, c(0.25, 0.5, 0.75)),
+#'   suffix = list(quant_x = c("Q1", "Q3"), quant_y = c("_Q1", "_Q2", "_Q3")),
+#'   by = "groups"
+#' )
+#'
 #' @export
 data_summary <- function(x, ...) {
   UseMethod("data_summary")
 }
 
 
 #' @export
-data_summary.matrix <- function(x, ..., by = NULL, remove_na = FALSE) {
-  data_summary(as.data.frame(x), ..., by = by, remove_na = remove_na)
+data_summary.matrix <- function(
+  x,
+  ...,
+  by = NULL,
+  remove_na = FALSE,
+  suffix = NULL
+) {
+  data_summary(
+    as.data.frame(x),
+    ...,
+    by = by,
+    remove_na = remove_na,
+    suffix = suffix
+  )
 }
 
 
@@ -72,7 +143,13 @@ data_summary.default <- function(x, ...) {
 
 #' @rdname data_summary
 #' @export
-data_summary.data.frame <- function(x, ..., by = NULL, remove_na = FALSE) {
+data_summary.data.frame <- function(
+  x,
+  ...,
+  by = NULL,
+  remove_na = FALSE,
+  suffix = NULL
+) {
   dots <- eval(substitute(alist(...)))
 
   # do we have any expression at all?
@@ -84,9 +161,10 @@ data_summary.data.frame <- function(x, ..., by = NULL, remove_na = FALSE) {
 
   if (is.null(by)) {
     # when we have no grouping, just compute a one-row summary
-    summarise <- .process_datasummary_dots(dots, x)
-    out <- data.frame(summarise)
-    colnames(out) <- vapply(summarise, names, character(1))
+    summarise <- .process_datasummary_dots(dots, x, suffix)
+    # coerce to data frame
+    out <- as.data.frame(t(summarise))
+    colnames(out) <- names(summarise)
   } else {
     # sanity check - is "by" a character string?
     if (!is.character(by)) {
@@ -122,15 +200,23 @@ data_summary.data.frame <- function(x, ..., by = NULL, remove_na = FALSE) {
         return(NULL)
       }
       # summarize data
-      summarise <- .process_datasummary_dots(dots, s)
+      summarise <- .process_datasummary_dots(dots, s, suffix)
       # coerce to data frame
-      summarised_data <- data.frame(summarise)
+      summarised_data <- as.data.frame(t(summarise))
       # bind grouping-variables and values
       summarised_data <- cbind(s[1, by], summarised_data)
       # make sure we have proper column names
-      colnames(summarised_data) <- c(by, unlist(lapply(summarise, names)))
+      colnames(summarised_data) <- c(by, names(summarise))
       summarised_data
     })
+    # check for correct number of columns. If one expression returns different
+    # number of values (which now means, we have different number of columns
+    # to bind) for each group, tell user
+    if (!all(lengths(out) == lengths(out)[1])) {
+      insight::format_error(
+        "Each expression must return the same number of values for each group. Some of the expressions seem to return varying numbers of values."
+      )
+    }
     out <- do.call(rbind, out)
   }
   # sort data
@@ -143,7 +229,13 @@ data_summary.data.frame <- function(x, ..., by = NULL, remove_na = FALSE) {
 
 
 #' @export
-data_summary.grouped_df <- function(x, ..., by = NULL, remove_na = FALSE) {
+data_summary.grouped_df <- function(
+  x,
+  ...,
+  by = NULL,
+  remove_na = FALSE,
+  suffix = NULL
+) {
   # extract group variables
   grps <- attr(x, "groups", exact = TRUE)
   group_variables <- data_remove(grps, ".rows")
@@ -154,13 +246,13 @@ data_summary.grouped_df <- function(x, ..., by = NULL, remove_na = FALSE) {
   # remove information specific to grouped df's
   attr(x, "groups") <- NULL
   class(x) <- "data.frame"
-  data_summary(x, ..., by = by, remove_na = remove_na)
+  data_summary(x, ..., by = by, remove_na = remove_na, suffix = suffix)
 }
 
 
 # helper -----------------------------------------------------------------------
 
-.process_datasummary_dots <- function(dots, data) {
+.process_datasummary_dots <- function(dots, data, suffix = NULL) {
   out <- NULL
   if (length(dots)) {
     # we check for character vector of expressions, in which case
@@ -199,39 +291,113 @@ data_summary.grouped_df <- function(x, ..., by = NULL, remove_na = FALSE) {
       }
     }
 
+    # sanity check: check the input for the `suffix` argument
+    # `suffix` can be NULL, or must be a (named) list
+    if (!is.null(suffix)) {
+      # if `suffix` is a character vector, we transform it into a list,
+      # matching the names of the expressions
+      if (is.character(suffix)) {
+        suffix <- rep(list(suffix), length(dots))
+        names(suffix) <- names(dots)
+      }
+      # no list? error
+      if (!is.list(suffix)) {
+        insight::format_error(
+          "Argument `suffix` must be a list of (named) character vectors, where the names match the names of the expressions, e.g.:",
+          paste0(
+            "`suffix = list(",
+            names(dots)[1],
+            " = c(\"_suffix1\", \"_suffix2\")`."
+          )
+        )
+      }
+      # not all elements named? error
+      if (!length(which(nzchar(names(suffix), keepNA = TRUE)))) {
+        insight::format_error("All elements of `suffix` must have names.")
+      }
+      # names of suffix do not match names of expressions? error
+      if (!all(names(suffix) %in% names(dots))) {
+        wrong_name <- which(!names(suffix) %in% names(dots))[1]
+        insight::format_error(
+          paste0(
+            "Names of `suffix` must match the names of the expressions. Suffix `",
+            names(suffix)[wrong_name],
+            "` has no corresponding expression."
+          )
+        )
+      }
+      # identical suffixes for one expression? error
+      identical_suffix <- vapply(
+        suffix,
+        function(i) insight::n_unique(i) != length(i),
+        logical(1)
+      )
+      if (any(identical_suffix)) {
+        insight::format_error(
+          paste0(
+            "All suffixes for a single expression must be unique. Suffix for element `",
+            names(identical_suffix)[which(identical_suffix)][1],
+            "` has duplicate values."
+          )
+        )
+      }
+    }
+
     out <- lapply(seq_along(dots), function(i) {
       new_variable <- .get_new_dots_variable(dots, i, data)
+      # check special case here - we want bayestestR::ci to work with
+      # data summary, to easily create CIs for, say, posterior draws
       if (inherits(new_variable, c("bayestestR_ci", "bayestestR_eti"))) {
         stats::setNames(new_variable, c("CI", "CI_low", "CI_high"))
       } else {
-        stats::setNames(new_variable, names(dots)[i])
+        # init
+        current_suffix <- NULL
+        # find matches and set use suffix if found
+        matching_names <- which(names(suffix) == names(dots)[i])
+        # either use suffixes based on matching names, or try to extract
+        # names from the returned summary expression (saved in "new_variable"),
+        # if the summary function returned a named vector
+        if (length(matching_names) > 0) {
+          current_suffix <- suffix[[matching_names]]
+        } else if (
+          length(new_variable) > 1 &&
+            all(nzchar(names(new_variable), keepNA = TRUE))
+        ) {
+          current_suffix <- names(new_variable)
+        }
+        # if we don't have suffixes for multiple columns, but expression
+        # returns multiple columns, we get NA column names - we use
+        # automatically numbered suffixes in this case
+        if (is.null(current_suffix) && length(new_variable) > 1) {
+          current_suffix <- paste0("_", seq_along(new_variable))
+        }
+
+        # if number of suffixes does not match the number of returned values
+        # by the expression, error
+        if (
+          !is.null(current_suffix) &&
+            length(current_suffix) != length(new_variable)
+        ) {
+          insight::format_error(
+            paste0(
+              "Argument `suffix` must have the same length as the result of the corresponding summary expression. `suffix` has ",
+              length(current_suffix),
+              " elements (",
+              text_concatenate(current_suffix, enclose = "`"),
+              ") for the expression `",
+              insight::safe_deparse(dots[[i]]),
+              "`, which returned ",
+              length(new_variable),
+              " values."
+            )
+          )
+        }
+        stats::setNames(new_variable, paste0(names(dots)[i], current_suffix))
       }
     })
   }
 
-  # check for correct length of output - must be a single value!
-  # Exception: bayestestR::ci()
-  wrong_length <- !sapply(
-    out,
-    inherits,
-    what = c("bayestestR_ci", "bayestestR_eti")
-  ) &
-    lengths(out) != 1 # nolint
-  if (any(wrong_length)) {
-    insight::format_error(
-      paste0(
-        "Each expression must return a single value. Following expression",
-        ifelse(sum(wrong_length) > 1, "s", " "),
-        " returned more than one value: ",
-        text_concatenate(
-          vapply(dots[wrong_length], insight::safe_deparse, character(1)),
-          enclose = "\""
-        )
-      )
-    )
-  }
-
-  out
+  unlist(out)
 }