R/layering.R

In Tplyr: A Traceability Focused Grammar of Clinical Data Summary

### Layering Interfaces

#' Attach a layer to a \code{tplyr_table} object
#'
#' @description
#' \code{add_layer} attaches a \code{tplyr_layer} to a \code{tplyr_table} object. This allows
#' for a tidy style of programming (using \code{magrittr} piping, i.e. \code{\%>\%}) with a
#' secondary advantage - the construction of the layer object may consist of a series of piped
#' functions itself.
#'
#' \code{Tplyr} encourages a user to view the construction of a table as a series of "layers".
#' The construction of each of these layers are isolated and independent of one another - but
#' each of these layers are children of the table itself. \code{add_layer} isolates the construction
#' of an individual layer and allows the user to construct that layer and insert it back into the
#' parent. The syntax for this is intuitive and allows for tidy piping. Simply pipe the current
#' table object in, and write the code to construct your layer within the \code{layer} parameter.
#'
#' \code{add_layers} is another approach to attaching layers to a \code{tplyr_table}. Instead of
#' constructing the entire table at once, \code{add_layers} allows you to construct layers as
#' different objects. These layers can then be attached into the \code{tplyr_table} all at
#' once.
#'
#' \code{add_layer} and \code{add_layers} both additionally allow you to name the layers as you
#' attach them. This is helpful when using functions like \code{\link{get_numeric_data}} or
#' \code{\link{get_stats_data}} when you can access information from a layer directly.
#' \code{add_layer} has a name parameter, and layers can be named in \code{add_layers} by
#' submitting the layer as a named argument.
#'
#'
#' @param parent A \code{tplyr_table} or \code{tplyr_layer}/\code{tplyr_subgroup_layer} object
#' @param layer A layer construction function and associated modifier functions
#' @param name A name to provide the layer in the table layers container
#'
#' @family Layer attachment
#' @rdname layer_attachment
#'
#' @return A \code{tplyr_table} or \code{tplyr_layer}/\code{tplyr_subgroup_layer} with a new layer inserted into the \code{layer}
#'   binding
#'
#' @seealso [tplyr_table(), tplyr_layer(), group_count(), group_desc(), group_shift()]
#'
#' @export
#'
#' @examples
#' # Load in pipe
#' library(magrittr)
#'
#' ## Single layer
#' t <- tplyr_table(mtcars, cyl) %>%
#'   add_layer(
#'     group_desc(target_var=mpg)
#'   )
#'
#' ## Single layer with name
#' t <- tplyr_table(mtcars, cyl) %>%
#'   add_layer(name='mpg',
#'     group_desc(target_var=mpg)
#'   )
#'
#' # Using add_layers
#' t <- tplyr_table(mtcars, cyl)
#' l1 <- group_desc(t, target_var=mpg)
#' l2 <- group_count(t, target_var=cyl)
#'
#' t <- add_layers(t, l1, 'cyl' = l2)
#'
add_layer <- function(parent, layer, name=NULL) {

  assert_that(!missing(parent), msg = "`parent` parameter must be provided")
  assert_that(!missing(layer), msg = "`layer` parameter must be provided")

  # Capture the layer code as a quosure
  layer <- enquo(layer)

  # Insert the `parent` argument into the topmost call of the layer code
  # (i.e. if any pipes %>% then pull out the left most call and modify it)
  l <- quo_get_expr(modify_nested_call(layer, parent=parent))

  # Evaluate the layer and grab `tplyr_layer` or `tplyr_subgroup_layer` object
  executed_layer <- list(eval(l, envir=caller_env()))

  # Attach the name
  names(executed_layer) <- name

  # Insert the layer into the parent object
  parent$layers <- append(parent$layers, executed_layer)
  parent
}

#' @param parent A \code{tplyr_table} or \code{tplyr_layer}/\code{tplyr_subgroup_layer} object
#' @param ... Layers to be added
#'
#' @export
#'
#' @family Layer attachment
#' @rdname layer_attachment
#'
#' @export
add_layers <- function(parent, ...) {
  # Parent exists
  assert_that(!missing(parent), msg = "`parent` parameter must be provided")
  # all objects are Tplyr layers
  map(list2(...), assert_is_layer)

  # Insert the layer into the parent object
  parent$layers <- append(parent$layers, list2(...))
  parent
}

#' Create a \code{count}, \code{desc}, or \code{shift} layer for discrete count
#' based summaries, descriptive statistics summaries, or shift count summaries
#'
#' @description This family of functions specifies the type of summary that is
#'   to be performed within a layer. \code{count} layers are used to create
#'   summary counts of some discrete variable. \code{desc} layers create summary
#'   statistics, and \code{shift} layers summaries the counts of different
#'   changes in states. See the "details" section below for more information.
#'
#' @param parent Required. The parent environment of the layer. This must be the
#'   \code{tplyr_table} object that the layer is contained within.
#' @param target_var Symbol. Required, The variable name(s) on which the summary
#'   is to be performed. Must be a variable within the target dataset. Enter
#'   unquoted - i.e. target_var = AEBODSYS. You may also provide multiple
#'   variables with \code{\link[dplyr]{vars}}.
#' @param by A string, a variable name, or a list of variable names supplied
#'   using \code{\link[dplyr]{vars}}
#' @param where Call. Filter logic used to subset the target data when
#'   performing a summary.
#' @param ... Additional arguments to pass forward
#'
#' @details \describe{ \item{Count Layers}{Count layers allow you to create
#'   summaries based on counting values with a variable. Additionally, this
#'   layer allows you to create n (\%) summaries where you're also summarizing
#'   the proportion of instances a value occurs compared to some denominator.
#'   Count layers are also capable of producing counts of nested relationships.
#'   For example, if you want to produce counts of an overall outside group, and
#'   then the subgroup counts within that group, you can specify the target
#'   variable as vars(OutsideVariable, InsideVariable). This allows you to do
#'   tables like Adverse Events where you want to see the Preferred Terms within
#'   Body Systems, all in one layer. Further control over denominators is
#'   available using the function \code{\link{set_denoms_by}} and distinct
#'   counts can be set using \code{\link{set_distinct_by}}} \item{Descriptive
#'   Statistics Layers}{Descriptive statistics layers perform summaries on
#'   continuous variables. There are a number of summaries built into Tplyr
#'   already that you can perform, including n, mean, median, standard
#'   deviation, variance, min, max, inter-quartile range, Q1, Q3, and missing
#'   value counts. From these available summaries, the default presentation of a
#'   descriptive statistic layer will output 'n', 'Mean (SD)', 'Median', 'Q1, Q3',
#'   'Min, Max', and 'Missing'. You can change these summaries using
#'   \code{\link{set_format_strings}}, and you can also add your own summaries
#'   using \code{\link{set_custom_summaries}}. This allows you to implement any
#'   additional summary statistics you want presented.} \item{Shift Layers}{A
#'   shift layer displays an endpoint's 'shift' throughout the duration of the
#'   study. It is an abstraction over the count layer, however we have provided
#'   an interface that is more efficient and intuitive. Targets are passed as
#'   named symbols using \code{dplyr::vars}. Generally the baseline is passed
#'   with the name 'row' and the shift is passed with the name 'column'. Both
#'   counts (n) and percentages (pct) are supported and can be specified with
#'   the \code{\link{set_format_strings}} function. To allow for flexibility
#'   when defining percentages, you can define the denominator using the
#'   \code{\link{set_denoms_by}} function. This function takes variable names and
#'   uses those to determine the denominator for the counts.} }
#'
#' @return An \code{tplyr_layer} environment that is a child of the specified
#'   parent. The environment contains the object as listed below.
#'
#' @return A \code{tplyr_layer} object
#'
#' @family Layer Construction Functions
#'
#' @rdname layer_constructors
#'
#' @seealso [\link{add_layer}, \link{add_layers}, \link{tplyr_table}, \link{tplyr_layer}]
#'
#' @export
#'
#' @examples
#' # Load in pipe
#' library(magrittr)
#'
#' t <- tplyr_table(iris, Species) %>%
#'   add_layer(
#'     group_desc(target_var=Sepal.Width)
#'   )
#'
#' t <- tplyr_table(iris, Species) %>%
#'   add_layer(
#'     group_desc(target_var=Sepal.Width)
#'   )
#'
#' t <- tplyr_table(mtcars, am) %>%
#'   add_layer(
#'     group_shift(vars(row=gear, column=carb), by=cyl)
#'   )
group_count <- function(parent, target_var, by=vars(), where=TRUE, ...) {
  tplyr_layer(parent, type='count', by=enquos(by), target_var=enquos(target_var), where=enquo(where), ...)
}

#' @rdname layer_constructors
#' @family Layer Construction Functions
#' @export
group_desc <- function(parent, target_var, by=vars(), where=TRUE, ...) {
  tplyr_layer(parent, type='desc', by=enquos(by), target_var=enquos(target_var), where=enquo(where), ...)
}

#' @rdname layer_constructors
#' @family Layer Construction Functions
#' @export
group_shift <- function(parent, target_var, by=vars(), where=TRUE, ...) {
  tplyr_layer(parent, type='shift', by=enquos(by), target_var=enquos(target_var), where=enquo(where), ...)
}