fix issues with NA values in recodes

DESCRIPTION

@@ -1,7 +1,7 @@
 Type: Package
 Package: datawizard
 Title: Easy Data Wrangling and Statistical Transformations
-Version: 0.8.0.8
+Version: 0.8.0.9
 Authors@R: c(
     person("Indrajeet", "Patil", , "[email protected]", role = "aut",
            comment = c(ORCID = "0000-0003-1995-6531", Twitter = "@patilindrajeets")),

NEWS.md

@@ -28,6 +28,9 @@ BUG FIXES
 * Fixed issue in `recode_into()` with probably wrong case number printed in the
   warning when several recode patterns match to one case.
 
+* Fixed issue in `recode_into()` when original data contained `NA` values and
+  `NA` was not included in the recode pattern.
+
 * Fixed issue in `data_filter()` where functions containing a `=` (e.g. when
   naming arguments, like `grepl(pattern, x = a)`) were mistakenly seen as
   faulty syntax. 

R/recode_into.r

@@ -135,6 +135,11 @@ recode_into <- function(..., data = NULL, default = NA, overwrite = TRUE, verbos
       index <- with(data, eval(dots[[i]][[2]]))
       value <- with(data, eval(dots[[i]][[3]]))
     }
+    # remember missing values, so we can add back later
+    missing_index <- is.na(index)
+    # make sure index has no missing values. when we have missing values in
+    # original expression, these are considered as "no match" and set to FALSE
+    index[is.na(index)] <- FALSE
     # overwriting values? do more recode-patterns match the same case?
     if (is.na(default)) {
       already_exists <- !is.na(out[index])
@@ -144,7 +149,7 @@ recode_into <- function(..., data = NULL, default = NA, overwrite = TRUE, verbos
     # save indices of overwritten cases
     overwritten_cases <- which(index)[already_exists]
     # tell user...
-    if (any(already_exists) && verbose) {
+    if (any(already_exists, na.rm = TRUE) && verbose) {
       if (overwrite) {
         msg <- paste(
           "Several recode patterns apply to the same cases.",
@@ -164,7 +169,12 @@ recode_into <- function(..., data = NULL, default = NA, overwrite = TRUE, verbos
     if (!overwrite) {
       index[overwritten_cases] <- FALSE
     }
+    # write new values into output vector
     out[index] <- value
+    # set back missing values
+    if (any(missing_index) && !is.na(default)) {
+      out[missing_index & out == default] <- NA
+    }
   }
 
   out

tests/testthat/test-recode_into.R

@@ -170,7 +170,7 @@ test_that("recode_into, check differen input length", {
   )
 })
 
-test_that("recode_into, check differen input length", {
+test_that("recode_into, check different input length", {
   x <- 1:5
   y <- c(5, 2, 3, 1, 4)
   expect_warning(
@@ -184,3 +184,35 @@ test_that("recode_into, check differen input length", {
     regexp = "Several recode patterns"
   )
 })
+
+test_that("recode_into, make sure recode works with missing in original variable", {
+  mtcars$mpg[c(3, 10, 12, 15, 16)] <- NA
+  mtcars$cyl[c(2, 15, 16)] <- NA
+  out1 <- recode_into(
+    mtcars$mpg > 20 & mtcars$cyl == 6 ~ 1,
+    mtcars$mpg <= 20 ~ 2,
+    mtcars$cyl == 4 ~ 3,
+    default = 0
+  )
+  out2 <- recode_into(
+    mtcars$mpg > 20 & mtcars$cyl == 6 ~ 1,
+    mtcars$mpg <= 20 ~ 2,
+    default = 0
+  )
+  # one NA in mpg is overwritten by valid value from cyl, total 5 NA
+  expect_identical(
+    out1,
+    c(
+      1, NA, 3, 1, 2, 2, 2, 3, 3, NA, 2, NA, 2, 2, NA, NA, 2, 3,
+      3, 3, 3, 2, 2, 2, 2, 3, 3, 3, 2, 2, 2, 3
+    )
+  )
+  # total 6 NA
+  expect_identical(
+    out2,
+    c(
+      1, NA, NA, 1, 2, 2, 2, 0, 0, NA, 2, NA, 2, 2, NA, NA, 2, 0,
+      0, 0, 0, 2, 2, 2, 2, 0, 0, 0, 2, 2, 2, 0
+    )
+  )
+})