Include replace_NAs option no_conversion and combinations

.buildlibrary

@@ -1,4 +1,4 @@
-ValidationKey: '1332170'
+ValidationKey: '1523040'
 AutocreateReadme: no
 AcceptedWarnings:
 - 'Warning: package ''.*'' was built under R version'

.zenodo.json

@@ -1,6 +1,6 @@
 {
   "title": "GDPuc: Easily Convert GDP Data",
-  "version": "0.7.0",
+  "version": "0.8.0",
   "description": "<p>A function to convert GDP time series from one unit to another.\n    All common GDP units are included, i.e. current and constant local currency\n    units, US$ via market exchange rates and international dollars via\n    purchasing power parities. Conversion factors can easily be customized. \n    Conversion at a regional/aggregated level is also possible.<\/p>",
   "creators": [
     {

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: GDPuc
 Title: Easily Convert GDP Data
-Version: 0.7.0
+Version: 0.8.0
 Authors@R: 
     c(person(given = "Johannes",
              family = "Koch",
@@ -43,6 +43,6 @@ Suggests:
     zoo
 Config/testthat/edition: 3
 Encoding: UTF-8
-Date: 2022-02-08
+Date: 2022-02-15
 Roxygen: list(markdown = TRUE)
 VignetteBuilder: knitr

NEWS.md

@@ -1,3 +1,9 @@
+# GDPuc 0.8.0
+
+* Add options "no_conversion" for the `replace_NAs` argument.
+
+* The `replace_NAs` argument can now take a vector allowing for a combination of options.
+
 # GDPuc 0.7.0
 
 * Add argument `return_cfs` to return the conversion factors used.

R/adapt_source.R

@@ -14,8 +14,8 @@ adapt_source <- function(gdp, source, with_regions, replace_NAs) {
                        dplyr::anti_join(source, by = c("iso3c", "year"))) %>%
     tidyr::complete(.data$iso3c, .data$year)
 
-  if (replace_NAs %in% c("linear", "linear_regional_average")) {
-    # Make sure that source contains obersvations for every year between min and max years.
+  if (replace_NAs[1] == "linear") {
+    # Make sure that source contains observations for every year between min and max years.
     # This is important for the function lin_int_ext, which works with indices, to compute the
     # correct values
     source_adapted <- source_adapted %>%
@@ -33,7 +33,7 @@ adapt_source <- function(gdp, source, with_regions, replace_NAs) {
         lin_int_ext))
   }
 
-  if (replace_NAs %in% c("regional_average", "linear_regional_average")) {
+  if ("regional_average" %in% replace_NAs) {
     # Get GDP variable from source object, with its unit
     regex_var <- "GDP, PPP \\(constant .... international \\$\\)"
     weight_var <- grep(regex_var, colnames(source), value = TRUE)[1]
@@ -68,7 +68,7 @@ adapt_source <- function(gdp, source, with_regions, replace_NAs) {
       dplyr::ungroup()
   }
 
-  if (replace_NAs == 1) {
+  if (1 %in% replace_NAs) {
     source_adapted <- source_adapted %>%
       # Mutate the 3 important columns
       dplyr::rowwise() %>%

R/check_user_input.R

@@ -120,14 +120,24 @@ check_with_regions <- function(unit_in, unit_out, source, with_regions) {
 # Check input parameter 'replace_NAs'
 check_replace_NAs <- function(with_regions, replace_NAs) {
   if (!is.null(replace_NAs)) {
-    if (replace_NAs == 1) {
-      lifecycle::deprecate_warn("0.7.0", "convertGDP(replace_NAs = 'was deprecated')")
+    if (setequal(replace_NAs, 1)) {
+      lifecycle::deprecate_warn("0.7.0", "convertGDP(replace_NAs = 'should not be 1')")
     }
-    if (!replace_NAs %in% c(0, 1, "linear", "regional_average", "linear_regional_average")) {
-      abort("Invalid 'replace_NAs' argument. Has to be either NULL, 0, 1, linear, regional_average or \\
-            linear_regional_average.")
+    if ("linear_regional_average" %in% replace_NAs) {
+      lifecycle::deprecate_stop(
+        "0.8.0",
+        "convertGDP(replace_NAs = '\"linear_regional_average\" has been replaced by c(\"linear\", \"regional_average\")')"
+      )
     }
-    if (replace_NAs == "regional_average" && is.null(with_regions)) {
+    if (!all(replace_NAs %in% c(0, 1, "no_conversion", "linear", "regional_average"))) {
+      abort("Invalid 'replace_NAs' argument. Has to be either NULL, 0, 1, no_conversion, linear, regional_average or \\
+            a combination of the above.")
+    }
+    if (length(replace_NAs) > 1 && replace_NAs[1] != "linear") {
+      abort("Invalid 'replace_NAs' argument. The only accepted combinations of arguments start with 'linear', e.g. \\
+            c('linear', 'no_conversion').")
+    }
+    if ("regional_average" %in% replace_NAs && is.null(with_regions)) {
       abort("Using 'regional_average' requires a region mapping. The 'with_regions' argument can't be NULL.")
     }
   }

R/convertGDP.R

@@ -64,16 +64,15 @@
 #' @param replace_NAs NULL by default, meaning no NA replacement. Can be set to one of the following:
 #'   \itemize{
 #'     \item 0: resulting NAs are simply replaced with 0.
+#'     \item "no_conversion": resulting NAs are simply replaced with the values from the gdp argument.
 #'     \item "linear": missing conversion factors in the source object are inter- and extrapolated linearly.
 #'     For the extrapolation, the closest 5 data points are used.
 #'     \item "regional_average": missing conversion factors in the source object are replaced with
 #'     the regional average of the region to which the country belongs. This requires a region-mapping to
 #'     be passed to the function, see the with_regions argument.
-#'     \item "linear_regional_average": missing conversion factors in the source object will be linearly
-#'     inter- and extrapolated, and when impossible (e.g. when no data at all is available for a country) set
-#'     to the regional GDP-weighted averages. This also requires a region-mapping to
-#'     be passed to the function, see the with_regions argument.
 #'   }
+#'   Can also be a vector with "linear" as first element, e.g. c("linear", 0) or c("linear", "no_conversion"),
+#'   in which case, the operations are done in sequence.
 #' @param verbose TRUE or FALSE. A flag to turn verbosity on or off. Be default it is equal to the
 #'   GDPuc.verbose option, which is FALSE if not set to TRUE by the user.
 #' @param return_cfs TRUE or FALSE. Set to TRUE to additionally return a tibble with the conversion factors
@@ -133,9 +132,14 @@ convertGDP <- function(gdp,
   # Call function
   x <- do.call(f, a)
 
-  if (!is.null(replace_NAs) && replace_NAs == 0) x[is.na(x)] <- 0
+  # Handle NAs
+  if (!is.null(replace_NAs) && 0 %in% replace_NAs) x[is.na(x)] <- 0
   if (any(is.na(x$value) & !is.na(internal$gdp$value))) {
-    warn("NAs have been generated for countries lacking conversion factors!")
+    if (!is.null(replace_NAs) && "no_conversion" %in% replace_NAs) {
+      x$value[is.na(x$value)] <- internal$gdp$value[is.na(x$value)]
+    } else {
+      warn("NAs have been generated for countries lacking conversion factors!")
+    }
   }
 
   # Return with original type and names

R/transform_user_input.R

@@ -67,7 +67,7 @@ transform_user_input <- function(gdp, unit_in, unit_out, source, with_regions, r
   }
 
   # Use different source if required
-  if (!is.null(replace_NAs) && replace_NAs != 0) {
+  if (!is.null(replace_NAs) && !setequal(replace_NAs, 0)  && !setequal(replace_NAs, "no_conversion")) {
     source <- adapt_source(gdp, source, with_regions, replace_NAs)
     source_name <- paste0(source_name, "_adapted")
   }

tests/testthat/test-check_user_input.R

@@ -33,8 +33,8 @@ test_that("source argument", {
                "Invalid 'source' argument. 'source' is neither a data frame nor a string.")
   s <- "blabla"
   expect_error(check_user_input(gdp, unit_in, unit_out, source = s),
-               glue::glue("Invalid 'source' argument. If 'source' is a string, it must be one of the internal sources. \\
-                          Use print_source_info\\(\\) for information on available sources. \\
+               glue::glue("Invalid 'source' argument. If 'source' is a string, it must be one of the internal \\
+                          sources. Use print_source_info\\(\\) for information on available sources. \\
                           If you are trying to pass a custom source, pass the data frame directly, not its name."))
 
   # Following doesn't work in covr for some reason
@@ -43,8 +43,7 @@ test_that("source argument", {
                "Invalid 'source' argument. Required columns are:(.*)")
 })
 
-test_that("Abort with bad input", {
-
+test_that("with_regions argument", {
   gdp <- tibble::tibble("iso3c" = "EUR", "year" = 2010, "value" = 100)
   unit_in = "current Int$PPP"
   unit_out = "current US$MER"
@@ -64,9 +63,34 @@ test_that("Abort with bad input", {
   )
   s <- my_bad_source
   expect_error(check_user_input(gdp, unit_in, unit_out,  source = s, with_regions = with_regions))
+})
+
+test_that("replace_NAs argument", {
 
+  gdp <- tibble::tibble("iso3c" = "EUR", "year" = 2010, "value" = 100)
+  unit_in = "current Int$PPP"
+  unit_out = "current US$MER"
+  s <- wb_wdi
+
+  expect_error(check_user_input(gdp, unit_in, unit_out,  source = s, replace_NAs = 2, with_regions = NULL),
+               glue::glue("Invalid 'replace_NAs' argument. Has to be either NULL, 0, 1, no_conversion, linear, \\
+                          regional_average or a combination of the above."))
+  expect_error(check_user_input(gdp, unit_in, unit_out,  source = s, replace_NAs = c(0, 1), with_regions = NULL),
+               glue::glue("Invalid 'replace_NAs' argument. The only accepted combinations of arguments start with \\
+                          'linear', e.g. c\\('linear', 'no_conversion'\\)."))
+
+  expect_error(
+    check_user_input(gdp, unit_in, unit_out,  source = s, replace_NAs = "regional_average", with_regions = NULL),
+    glue::glue("Using 'regional_average' requires a region mapping. The 'with_regions' argument can't be NULL.")
+  )
+})
+
+test_that("boolean arguments", {
+
+  gdp <- tibble::tibble("iso3c" = "EUR", "year" = 2010, "value" = 100)
+  unit_in = "current Int$PPP"
+  unit_out = "current US$MER"
   s <- wb_wdi
-  expect_error(check_user_input(gdp, unit_in, unit_out,  source = s, replace_NAs = 2, with_regions = NULL))
 
   expect_error(check_user_input(gdp, unit_in, unit_out,  source = s,
                                 with_regions = NULL, replace_NAs = NULL, verbose = "blabla"))
@@ -81,3 +105,4 @@ test_that("Abort with bad input", {
     verbose = TRUE,
     return_cfs = "blabla"))
 })
+

tests/testthat/test-replace_NAs.R

@@ -52,9 +52,26 @@ test_that("convertGDP replace missing conversion factors", {
   expect_equal(gdp2$iso3c, gdp_conv5$iso3c)
 })
 
+test_that("convertGDP replace_NAs = 'no_conversion'", {
+  # wb_wi does not have info for ABW in 2019
+  gdp <- tidyr::expand_grid("iso3c" = c("ABW", "DEU", "USA"),
+                            "year" = c(2010, 2015, 2025),
+                            "SSP" = c("SSP1", "SSP2"), "value" = 100)
+
+  expect_warning(convertGDP(gdp,
+                            unit_in = "constant 2005 Int$PPP",
+                            unit_out = "constant 2019 US$MER"))
+
+  gdp_conv <- convertGDP(gdp,
+                         unit_in = "constant 2005 Int$PPP",
+                         unit_out = "constant 2019 US$MER",
+                         replace_NAs = "no_conversion")
+
+  expect_identical(gdp[1:6,], gdp_conv[1:6,])
+})
 
 test_that("convertGDP replace_NAs = linear", {
-  # wb_wi does not have info for AIA, so AIA is used for testing here
+  # wb_wi does not have info for ABW in 2019
   gdp <- tidyr::expand_grid("iso3c" = c("ABW", "DEU", "USA"),
                             "year" = c(2010, 2015, 2025),
                             "SSP" = c("SSP1", "SSP2"), "value" = 100)
@@ -65,14 +82,40 @@ test_that("convertGDP replace_NAs = linear", {
 
   gdp_conv <- convertGDP(gdp,
                           unit_in = "constant 2005 Int$PPP",
-                          unit_out = "constant 2005 US$MER",
+                          unit_out = "constant 2019 US$MER",
                           replace_NAs = "linear")
 
   expect_true(!any(is.na(gdp_conv$value)))
 })
 
-
 test_that("lin_int_ext", {
   x <- c(NA,NA,NA,NA,NA,NA,2,3,4,5,NA,7,8,NA,NA,NA,NA,NA,NA)
   expect_equal(lin_int_ext(x), -4:14)
 })
+
+
+test_that("convertGDP replace_NAs = c('linear', 'no_conversion')", {
+  # wb_wi does not have info for ABW in 2019
+  gdp <- tidyr::expand_grid("iso3c" = c("ABW", "DEU", "USA", "JJJ"),
+                            "year" = c(2010, 2015, 2025),
+                            "SSP" = c("SSP1", "SSP2"), "value" = 100)
+
+  expect_warning(convertGDP(gdp,
+                            unit_in = "constant 2005 Int$PPP",
+                            unit_out = "constant 2019 US$MER"))
+
+  gdp_conv <- convertGDP(gdp,
+                         unit_in = "constant 2005 Int$PPP",
+                         unit_out = "constant 2019 US$MER",
+                         replace_NAs = c("linear", "no_conversion"))
+
+  expect_true(!any(is.na(gdp_conv$value)))
+  expect_identical(gdp[19:24,], gdp_conv[19:24,])
+
+  gdp_conv <- convertGDP(gdp,
+                         unit_in = "constant 2005 Int$PPP",
+                         unit_out = "constant 2019 US$MER",
+                         replace_NAs = c("linear", 0))
+  expect_identical(dplyr::pull(gdp_conv[19:24, "value"]), rep(0, 6))
+})
+

vignettes/handle_NAs.Rmd

@@ -65,6 +65,31 @@ x$result
 x$cfs
 ```
 
+
+### `replace_NAs` = "no_conversion"  
+
+If set to "no_conversion", NAs are replaced with the values in the gdp argument.
+
+```{r}
+my_gdp <- tibble::tibble(
+  iso3c = "ABW", 
+  year = 2010:2014, 
+  value = 100:104
+)
+
+x <- convertGDP(
+  gdp = my_gdp, 
+  unit_in = "constant 2005 Int$PPP", 
+  unit_out = "constant 2019 Int$PPP",
+  replace_NAs = "no_conversion",
+  return_cfs = TRUE
+)
+x$result
+
+x$cfs
+```
+
+
 ### `replace_NAs` = "linear"
 
 If set to "linear", missing conversion factors are inter- and extrapolated linearly. For the extrapolation, the closest 5 data points are used.
@@ -120,9 +145,11 @@ x$cfs
 # Obviously, being a part of the same region, does not mean the currencies are of the same strength.
 ```
 
-### `replace_NAs` = "linear_regional_average"
+### `replace_NAs` = c("linear", "...")
 
-If set to "linear_regional_average", conversion factors will be linearly inter- and extrapolated, and when impossible (e.g. when no data at all is available for a country) set to the regional GDP-weighted averages.
+If a vector is passed, with "linear" as first element, then the operations are done in sequence. 
+For example for c("linear", 0), missing conversion factors are first inter- and extrapolated linearly but 
+if any missing conversion factors still lead to NAs, these are replaced with 0.
 
 ```{r}
 # Create an imaginary country XXX, and add it to the Latin America region
@@ -141,7 +168,7 @@ x <- convertGDP(
   gdp = my_gdp, 
   unit_in = "constant 2005 Int$PPP", 
   unit_out = "constant 2019 Int$PPP",
-  replace_NAs = "linear_regional_average",
+  replace_NAs = c("linear", 0),
   with_regions = my_mapping_data_frame,
   return_cfs = TRUE
 )

vignettes/verbose.Rmd

@@ -1,8 +1,8 @@
 ---
-title: "Geting information on the conversion process"
+title: "Getting information on the conversion process"
 output: rmarkdown::html_vignette
 vignette: >
-  %\VignetteIndexEntry{Geting information on the conversion process}
+  %\VignetteIndexEntry{Getting information on the conversion process}
   %\VignetteEngine{knitr::rmarkdown}
   %\VignetteEncoding{UTF-8}
 ---