Added a saveObject,data.frame-method to avoid unexpected saving as a …

…list.

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: alabaster.base
 Title: Save Bioconductor Objects To File
-Version: 1.5.9
-Date: 2024-09-22
+Version: 1.5.10
+Date: 2024-10-17
 Authors@R: person("Aaron", "Lun", role=c("aut", "cre"), email="[email protected]")
 License: MIT + file LICENSE
 Description: 

R/saveDataFrame.R

@@ -2,12 +2,12 @@
 #'
 #' Stage a DataFrame by saving it to a HDF5 file.
 #'
-#' @param x A \linkS4class{DataFrame}.
+#' @param x A \linkS4class{DataFrame} or data.frame.
 #' @inheritParams saveObject
 #'
 #' @return
 #' A named list containing the metadata for \code{x}.
-#' \code{x} itself is written to a CSV or HDF5 file inside \code{path}.
+#' \code{x} itself is written to a HDF5 file inside \code{path}.
 #' Additional files may also be created inside \code{path} and referenced from the metadata.
 #'
 #' @details
@@ -21,6 +21,9 @@
 #' If \code{\link{metadata}} or \code{\link{mcols}} are present, 
 #' they are saved to the \code{other_annotations} and \code{column_annotations} subdirectories, respectively, via \code{\link{saveObject}}.
 #'
+#' In the on-disk representation, no distinction is made between \linkS4class{DataFrame} and data.frame instances of \code{x}.
+#' Calling \code{readDataFrame} will always produce a \linkS4class{DFrame} regardless of the class of \code{x}.
+#'
 #' @author Aaron Lun
 #'
 #' @examples
@@ -48,7 +51,7 @@ setMethod("saveObject", "DataFrame", function(x, path, ...) {
 })
 
 #' @importFrom rhdf5 h5write h5createGroup h5createFile H5Gopen H5Gclose H5Acreate H5Aclose H5Awrite H5Fopen H5Fclose H5Dopen H5Dclose
-.write_hdf5_new <- function(x, path, ...) {
+.write_hdf5_new <- function(x, path, row.names=rownames(x), ...) {
     subpath <- "basic_columns.h5"
     ofile <- paste0(path, "/", subpath)
 
@@ -133,11 +136,23 @@ setMethod("saveObject", "DataFrame", function(x, path, ...) {
     }
 
     h5_write_vector(ghandle, "column_names", colnames(x))
-    if (!is.null(rownames(x))) {
-        h5_write_vector(ghandle, "row_names", rownames(x))
+    if (!is.null(row.names)) {
+        h5_write_vector(ghandle, "row_names", row.names)
     }
 }
 
+#' @export
+#' @rdname stageDataFrame
+setMethod("saveObject", "data.frame", function(x, path, ...) {
+    dir.create(path, showWarnings=FALSE)
+    rn <- attr(x, "row.names")
+    if (is.integer(rn)) {
+        rn <- NULL
+    }
+    .write_hdf5_new(x, path, row.names=rn, ...)
+    saveObjectFile(path, "data_frame", list(data_frame=list(version="1.0")))
+})
+
 #######################################
 ########### OLD STUFF HERE ############
 #######################################

inst/NEWS.Rd

@@ -20,6 +20,8 @@ to allow developers to declare that custom subclasses satisfy an interface or ha
 
 \item Updated \code{validateDirectory()} so that it works with a directory of objects saved via \code{saveObject()}.
 Objects saved under the old regime (i.e., \code{stageObject()}) are auto-detected but can also be explicitly validated by setting \code{legacy=FALSE}.
+
+\item Added a data.frame method for \code{saveObject()}, to avoid fallback to the list method.
 }}
 
 \section{Version 1.4.0}{\itemize{

tests/testthat/test-DataFrame.R

@@ -677,3 +677,27 @@ test_that("staging of arrays continues to work with character matrices", {
     roundtrip$Z <- as.matrix(roundtrip$Z)
     expect_identical(roundtrip, input)
 })
+
+test_that("saving works for base data.frames", {
+    nrows <- 123
+    df <- data.frame(
+        stuff = rep(LETTERS[1:3], length.out=nrows),
+        foo = seq_len(nrows),
+        whee = as.numeric(10 + seq_len(nrows))
+    )
+    df$blah <- factor(df$stuff, LETTERS[10:1])
+    df$rabbit <- factor(df$stuff, LETTERS[1:3], ordered=TRUE)
+
+    tmp <- tempfile()
+    saveObject(df, tmp)
+    roundtrip <- readObject(tmp)
+    expect_null(rownames(roundtrip))
+    expect_identical(as.data.frame(roundtrip), df)
+
+    # Respects row names.
+    rownames(df) <- sprintf("GENE_%i", seq_len(nrows))
+    tmp <- tempfile()
+    saveObject(df, tmp)
+    roundtrip <- readObject(tmp)
+    expect_identical(as.data.frame(roundtrip), df)
+})