Add function 'verify.data.frame.columns'

Add the 'verify.data.frame.column' to the miscellaneous utilities to check the existence and welltypedness of columns in a dataframe.

This works towards fixing #208.

Signed-off-by: Maximilian Löffler <s8maloef@stud.uni-saarland.de>

tests/test-misc.R

@@ -107,6 +107,73 @@ test_that("Match argument or take default.", {
     expect_equal(actual.result, expected.result, info = "Multiple choices with ignored default, two choices")
 })
 
+##
+## Check presence and datatype of column.
+##
+
+test_that("Check presence and datatype of column.", {
+
+    user.names = c("John", "Peter", "Maria", "Susanne")
+
+    ## contains NaN to verify functionality does not break
+    age = c(42, 50, NaN, 66)
+
+    ## contains NA to verify functionality does not break
+    is.male = c(TRUE, TRUE, FALSE, NA)
+
+    ## construct simple testing dataframe
+    data.frame = data.frame(user.names, age, is.male)
+
+    ## 1) Check base functionality (benign use-case)
+    expect_no_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age", "is.male"), c("character", "numeric", "logical")),
+        message = "All columns present and well-typed.")
+
+    ## 2) Base test with reordered columns
+    expect_no_error(verify.data.frame.columns(
+        data.frame, c("is.male", "age", "user.names"), c("logical", "numeric", "character")),
+        message = "Order of columns does not matter.")
+
+    ## 3) Specify less columns than present (Allow optional columns)
+    expect_no_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age"), c("character", "numeric")),
+        message = "Optional columns are allowed.")
+
+    ## 4) Unequal amount of column names and datatypes
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age", "is.male"), c("character", "numeric")),
+        message = "More coloumn names specified than datatypes.")
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age"), c("character", "numeric", "logical")),
+        message = "More coloumn names specified than datatypes.")
+
+    ## 5) Datatypes do not match column names
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age", "is.male"), c("logical", "character", "numeric")),
+        message = "Column names do not match datatypes.")
+
+    ## 6) Invalid column / Column not present in dataframe (Typo)
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.name"), c("character")),
+        message = "Column names do not match datatypes.")
+
+    ## 7) No datatypes specified and column names are present
+    expect_no_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age", "is.male")),
+        message = "Column names do not match datatypes.")
+
+    ## 8) No datatypes specified but column names are not present (Typo)
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.name")),
+        message = "Column names do not match datatypes.")
+
+    ## 9) To many column names and no datatypes specified
+    expect_error(verify.data.frame.columns(
+        data.frame, c("user.names", "age", "is.male", "job.orientation")),
+        message = "Column names do not match datatypes.")
+
+})
+
 ## / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /
 ## Date handling -----------------------------------------------------------
 

util-misc.R

@@ -139,6 +139,76 @@ match.arg.or.default = function(arg, choices, default = NULL, several.ok = FALSE
     }
 }
 
+#' Check if a dataframe matches a given structure. This includes the dataframe to contain columns
+#' which must match the column names in \code{columns} and the datatypes in \code{data.types}
+#'
+#' @param data the dataframe under investigation for structural conformity
+#' @param columns a character vector containing the column names the data frame should include
+#' @param data.types an ordered vector containing the data types corresponding to the columns.
+#'                   This vector must be of same length of the vector of \code{columns}
+#'                   [default: NULL]
+verify.data.frame.columns = function(data, columns, data.types = NULL) {
+
+    ## every column of the data frame must be one to one mapped to a datatype expected in the column
+    ## therefore if there aren't as many datatypes provided in \code{data.types} as column names have
+    ## been provided in \code{columns} we can stop here already.
+    if (!is.null(data.types) && length(columns) != length(data.types)) {
+        error.message = sprintf("If specified, the length of the two given vectors columns and data.types must match.")
+        logging::logerror(error.message)
+        stop(error.message)
+    }
+
+    ## obtain vector of all column names included in the dataframe to ease further checks.
+    data.frame.columns = colnames(data)
+
+    ## iterate over all columns in \code{columns}
+    for (i in seq_along(columns)) {
+
+        ## obtain the column.
+        column = columns[i]
+
+        ## stop verification process early if column is not present in the dataframe.
+        if (!(column %in% data.frame.columns)) {
+            error.message = sprintf("Column '%s' is missing from the dataframe", column)
+            logging::logerror(error.message)
+            stop(error.message)
+        }
+
+        if (!is.null(data.types)) {
+
+            ## obtain the datatype that should be present in the dataframe column c
+            ## which is currently under investigation.
+            expected.type = data.types[i]
+
+            ## necessary case distinction for special case list where calling \code{base::class}
+            ## removes information about the listing property
+            if (expected.type == "list()") {
+
+                ## column is not a list
+                if (!is.list(data[[column]])) {
+                    error.message = sprintf("Column '%s' is expected to be a list but it '%s'",
+                                            column, class(received.type))
+                    logging::logerror(error.message)
+                    stop(error.message)
+                }
+
+            } else {
+                ## obtain the datatype that elements of the current column hold in the dataframe.
+                received.type = class(data[[column]])
+
+                ## stop verification process early if column type in the dataframe is not matching
+                ## the expected datatype.
+                if (!(expected.type %in% received.type)) {
+                    error.message = sprintf("Column '%s' has type '%s' in dataframe, expected '%s'",
+                                            column, received.type, expected.type)
+                    logging::logerror(error.message)
+                    stop(error.message)
+                }
+            }
+        }
+    }
+}
+
 
 ## / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /
 ## Empty dataframe creation-------------------------------------------------