feat: add excel serializer/parser (#975)

Author: r2evans

DESCRIPTION

@@ -54,7 +54,10 @@ Suggests:
     rapidoc,
     sf,
     ragg,
-    svglite
+    svglite,
+    readxl,
+    writexl,
+    utils
 RoxygenNote: 7.3.2
 Collate:
     'async.R'

NAMESPACE

@@ -27,6 +27,7 @@ export(include_rmd)
 export(is_plumber)
 export(options_plumber)
 export(parser_csv)
+export(parser_excel)
 export(parser_feather)
 export(parser_form)
 export(parser_geojson)
@@ -80,6 +81,7 @@ export(serializer_cat)
 export(serializer_content_type)
 export(serializer_csv)
 export(serializer_device)
+export(serializer_excel)
 export(serializer_feather)
 export(serializer_format)
 export(serializer_geojson)

NEWS.md

@@ -6,6 +6,12 @@
 * `parse_rds()`, `parse_feather()`, and `parse_parquet()` no longer writes data to disk during parsing (@thomasp85, #942)
 * Returning error messages are now turned off by default rather than being turned on if running interactively and turned off if not (@thomasp85, #962)
 
+* New serializers
+  * `serializer_excel()`: Return an object serialized by `writexl::write_xlsx` (@r2evans, #973).
+
+* New request body parsers
+  * `parser_excel()`: Parse request body as an excel workbook using `readxl::read_excel` (@r2evans, #973). This defaults to loading in the first worksheet only, you can use `@parse excel list(sheet=NA)` to import all worksheets. This always returns a list of frames, even if just one worksheet.
+
 # plumber 1.2.2
 
 * Allow to set plumber options using environment variables `?options_plumber`. (@meztez #934)

R/parse-body.R

@@ -510,6 +510,41 @@ parser_parquet <- function(...) {
   }
 }
 
+# readxl's default behavior is to read only one worksheet at a time; in order for an endpoint to
+# read multiple worksheets, its documentation suggests to iterate over discovered names (c.f.,
+# https://readxl.tidyverse.org/articles/readxl-workflows.html#iterate-over-multiple-worksheets-in-a-workbook);
+# for this reason, this parser detects an NA in the 'sheet=' argument and replaces it with all
+# worksheet names found in the workbook
+
+#' @describeIn parsers excel parser. See [readxl::read_excel()] for more details. (Defaults to reading in the first worksheet only, use `@parser excel list(sheet=NA)` to read in all worksheets.)
+#' @param sheet Sheet to read. Either a string (the name of a sheet), or an
+#' integer (the position of the sheet). Defaults to the first sheet. To read all
+#' sheets, use `NA`.
+#' @export
+parser_excel <- function(..., sheet = NULL) {
+  if (!requireNamespace("readxl", quietly = TRUE)) {
+    stop("`readxl` must be installed for `parser_excel` to work")
+  }
+  parse_fn <- parser_read_file(function(tmpfile) {
+    if (is.null(sheet)) {
+      # we have to hard-code this since lapply won't iterate if NULL
+      sheet <- 1L
+    } else if (anyNA(sheet)) {
+      sheet <- readxl::excel_sheets(tmpfile)
+    }
+    if (is.character(sheet)) names(sheet) <- sheet
+    out <- suppressWarnings(
+      lapply(sheet, function(sht) {
+        readxl::read_excel(path = tmpfile, sheet = sht, ...)
+      })
+    )
+    out
+  })
+  function(value, ...) {
+    parse_fn(value)
+  }
+}
+
 #' @describeIn parsers Octet stream parser. Returns the raw content.
 #' @export
 parser_octet <- function() {
@@ -588,6 +623,7 @@ register_parsers_onLoad <- function() {
   register_parser("rds",     parser_rds,     fixed = "application/rds")
   register_parser("feather", parser_feather, fixed = c("application/vnd.apache.arrow.file", "application/feather"))
   register_parser("parquet", parser_parquet, fixed = "application/vnd.apache.parquet")
+  register_parser("excel",   parser_excel,   fixed = c("application/vnd.openxmlformats-officedocument.spreadsheetml.sheet", "application/vnd.ms-excel"))
   register_parser("text",    parser_text,    fixed = "text/plain", regex = "^text/")
   register_parser("tsv",     parser_tsv,     fixed = c("application/tab-separated-values", "text/tab-separated-values"))
   # yaml types: https://stackoverflow.com/a/38000954/591574

R/serializer.R

@@ -306,6 +306,22 @@ serializer_parquet <- function(type = "application/vnd.apache.parquet") {
   )
 }
 
+#' @describeIn serializers excel serializer. See also: [writexl::write_xlsx()]
+#' @export
+serializer_excel <- function(..., type = "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet") {
+  if (!requireNamespace("writexl", quietly = TRUE)) {
+    stop("The writexl package is not available but is required in order to use the writexl serializer",
+         call. = FALSE)
+  }
+
+  serializer_write_file(
+    fileext = ".xlsx",
+    type = type,
+    write_fn = function(val, tmpfile) {
+      writexl::write_xlsx(x = val, path = tmpfile, ...)
+    }
+  )
+}
 
 #' @describeIn serializers YAML serializer. See also: [yaml::as.yaml()]
 #' @export
@@ -693,6 +709,7 @@ add_serializers_onLoad <- function() {
   register_serializer("tsv",         serializer_tsv)
   register_serializer("feather",     serializer_feather)
   register_serializer("parquet",     serializer_parquet)
+  register_serializer("excel",       serializer_excel)
   register_serializer("yaml",        serializer_yaml)
   register_serializer("geojson",     serializer_geojson)
 

man/parsers.Rd

@@ -134,6 +134,32 @@ test_that("Test parquet parser", {
   expect_equal(parsed, r_object)
 })
 
+test_that("Test excel parser", {
+  skip_if_not_installed("readxl")
+  skip_if_not_installed("writexl")
+
+  tmp <- tempfile(fileext = ".xlsx")
+  on.exit({
+    file.remove(tmp)
+  }, add = TRUE)
+
+  # note: factors will fail the round-trip test
+  r_object <- data.frame(chr = LETTERS[1:3], int = 1:3, num = pi+1:3, lgl = c(TRUE, FALSE, NA))
+  res <- try(writexl::write_xlsx(r_object, tmp), silent = TRUE)
+  skip_if(
+    inherits(res, "try-error"),
+    "writexl::write_xlsx() isn't working."
+  )
+
+  val <- readBin(tmp, "raw", 10000)
+
+  parsed <- parse_body(val, "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet", make_parser("excel"))
+  # convert from tibble to data.frame
+  parsed <- as.data.frame(parsed[[1]], stringsAsFactors = FALSE)
+
+  expect_equal(parsed, r_object)
+})
+
 test_that("Test geojson parser", {
   skip_if_not_installed("geojsonsf")
   skip_if_not_installed("sf")

man/serializers.Rd

@@ -0,0 +1,34 @@
+context("excel serializer")
+
+test_that("excel serializes properly", {
+  skip_if_not_installed("writexl")
+
+  d <- data.frame(a=1, b=2, c="hi")
+  val <- serializer_excel()(d, data.frame(), PlumberResponse$new(), stop)
+  expect_equal(val$status, 200L)
+  expect_equal(val$headers$`Content-Type`, "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet")
+
+  # the remaining relies on the fact that xlsx files start as zip files
+  # https://en.wikipedia.org/wiki/List_of_file_signatures
+  expect_equal(val$body[1:4], as.raw(c(0x50, 0x4b, 0x03, 0x04)))
+  tf <- tempfile()
+  on.exit(unlink(tf), add = TRUE)
+  writeBin(val$body, tf)
+  zipcontents <- expect_silent(utils::unzip(tf, list = TRUE))
+  expect_s3_class(zipcontents, "data.frame")
+  expect_true("xl/workbook.xml" %in% zipcontents$Name)
+
+})
+
+test_that("Errors call error handler", {
+  skip_if_not_installed("writexl")
+
+  errors <- 0
+  errHandler <- function(req, res, err){
+    errors <<- errors + 1
+  }
+
+  expect_equal(errors, 0)
+  serializer_excel()(parse(text="hi"), data.frame(), PlumberResponse$new("csv"), errorHandler = errHandler)
+  expect_equal(errors, 1)
+})