Implement create/delete dirs for blob storage (#59)

* blob create/delete

* implement generics

* works

* update news

* better wording

* also update readme

* update readme again

NAMESPACE

@@ -119,6 +119,7 @@ export(create_adls_dir)
 export(create_adls_filesystem)
 export(create_azure_dir)
 export(create_blob_container)
+export(create_blob_dir)
 export(create_file_share)
 export(create_storage_container)
 export(create_storage_dir)
@@ -129,6 +130,7 @@ export(delete_azure_dir)
 export(delete_azure_file)
 export(delete_blob)
 export(delete_blob_container)
+export(delete_blob_dir)
 export(delete_file_share)
 export(delete_storage_container)
 export(delete_storage_dir)

NEWS.md

@@ -1,7 +1,9 @@
 # AzureStor 3.2.2
 
-- Implement recursive/non-recursive directory listings for `list_blobs`, thanks to @cantpitch. Note that since blob storage doesn't have true directories, there are some warts to be aware of; see the help for `list_blobs` for more details.
-- Fixes to the directory detection logic of `list_blobs`. Again, since blob storage doesn't have true directories, the `isdir` column of the `list_blobs` output should be treated as a best guess.
+- Extended support for directories in blob storage. Note that since blob storage doesn't have true directories, there are some warts to be aware of; see `?blob` for more details.
+  - Implement non-recursive directory listings for `list_blobs`, thanks to @cantpitch.
+  - Fixes to the directory detection logic of `list_blobs`.
+  - Implement `create_blob_dir` and `delete_blob_dir`.
 - Remove broken implementation for recursive deleting of subdirectory contents for file storage.
 
 # AzureStor 3.2.1

R/blob_client_funcs.R

@@ -220,7 +220,7 @@ delete_blob_container.blob_endpoint <- function(endpoint, name, confirm=TRUE, le
 
 #' Operations on a blob container or blob
 #'
-#' Upload, download, or delete a blob; list blobs in a container; check blob availability.
+#' Upload, download, or delete a blob; list blobs in a container; create or delete directories; check blob availability.
 #'
 #' @param container A blob container object.
 #' @param blob A string naming a blob.
@@ -235,7 +235,7 @@ delete_blob_container.blob_endpoint <- function(endpoint, name, confirm=TRUE, le
 #' @param use_azcopy Whether to use the AzCopy utility from Microsoft to do the transfer, rather than doing it in R.
 #' @param max_concurrent_transfers For `multiupload_blob` and `multidownload_blob`, the maximum number of concurrent file transfers. Each concurrent file transfer requires a separate R process, so limit this if you are low on memory.
 #' @param prefix For `list_blobs`, an alternative way to specify the directory.
-#' @param recursive For the multiupload/download functions, whether to recursively transfer files in subdirectories. For `list_blobs`, whether to include the contents of any subdirectories in the listing.
+#' @param recursive For the multiupload/download functions, whether to recursively transfer files in subdirectories. For `list_blobs`, whether to include the contents of any subdirectories in the listing. For `delete_blob_dir`, whether to recursively delete subdirectory contents as well (not yet supported).
 #'
 #' @details
 #' `upload_blob` and `download_blob` are the workhorse file transfer functions for blobs. They each take as inputs a _single_ filename as the source for uploading/downloading, and a single filename as the destination. Alternatively, for uploading, `src` can be a [textConnection] or [rawConnection] object; and for downloading, `dest` can be NULL or a `rawConnection` object. If `dest` is NULL, the downloaded data is returned as a raw vector, and if a raw connection, it will be placed into the connection. See the examples below.
@@ -256,8 +256,8 @@ delete_blob_container.blob_endpoint <- function(endpoint, name, confirm=TRUE, le
 #'
 #' - The `isdir` column in the data frame output of `list_blobs` is a best guess as to whether an object represents a file or directory, and may not always be correct. Currently, `list_blobs` assumes that any object with a file size of zero is a directory.
 #' - Zero-length files can cause problems for the blob storage service as a whole (not just AzureStor). Try to avoid uploading such files.
-#' - The output of `list_blobs(recursive=TRUE)` can vary based on whether the storage account has hierarchical namespaces enabled.
-#' - `create_storage_dir` and `delete_storage_dir` currently do not have methods for blob containers.
+#' - `create_blob_dir` and `delete_blob_dir` function as expected only for accounts with hierarchical namespaces enabled. When this feature is disabled, directories do not exist as objects in their own right: to create a directory, simply upload a blob to that directory. To delete a directory, delete all the blobs within it; as far as the blob storage service is concerned, the directory then no longer exists.
+#' - Similarly, the output of `list_blobs(recursive=TRUE)` can vary based on whether the storage account has hierarchical namespaces enabled.
 #'
 #' @return
 #' For `list_blobs`, details on the blobs in the container. For `download_blob`, if `dest=NULL`, the contents of the downloaded blob as a raw vector. For `blob_exists` a flag whether the blob exists.
@@ -472,6 +472,45 @@ delete_blob <- function(container, blob, confirm=TRUE)
     invisible(do_container_op(container, blob, http_verb="DELETE"))
 }
 
+#' @rdname blob
+#' @export
+create_blob_dir <- function(container, dir)
+{
+    # workaround: upload a zero-length file to the desired dir, then delete the file
+    destfile <- file.path(dir, basename(tempfile()))
+
+    opts <- options(azure_storage_progress_bar=FALSE)
+    on.exit(options(opts))
+
+    upload_blob(container, rawConnection(raw(0)), destfile)
+    delete_blob(container, destfile, confirm=FALSE)
+    invisible(NULL)
+}
+
+#' @rdname blob
+#' @export
+delete_blob_dir <- function(container, dir, recursive=FALSE, confirm=TRUE)
+{
+    if(dir %in% c("/", "."))
+        return(invisible(NULL))
+
+    if(!delete_confirmed(confirm, paste0(container$endpoint$url, container$name, "/", dir), "directory"))
+        return(invisible(NULL))
+
+    if(recursive)
+        stop("Recursive deleting of subdirectory contents not yet supported", call.=FALSE)
+
+    parent <- dirname(dir)
+    if(parent == ".")
+        parent <- "/"
+    lst <- list_blobs(container, parent, recursive=FALSE)
+    whichrow <- which(lst$name == paste0(dir, "/"))
+    if(is_empty(whichrow) || !lst$isdir[whichrow])
+        stop("Not a directory", call.=FALSE)
+
+    delete_blob(container, dir, confirm=FALSE)
+}
+
 #' @rdname blob
 #' @export
 blob_exists <- function(container, blob)
@@ -483,3 +522,4 @@ blob_exists <- function(container, blob)
     httr::stop_for_status(res, storage_error_message(res))
     return(TRUE)
 }
+

R/client_generics.R

@@ -222,8 +222,8 @@ UseMethod("create_storage_dir")
 
 #' @rdname generics
 #' @export
-create_storage_dir.blob_container <- function(container, ...)
-stop("Blob storage does not support directories")
+create_storage_dir.blob_container <- function(container, dir, ...)
+create_blob_dir(container, dir, ...)
 
 #' @rdname generics
 #' @export
@@ -245,8 +245,8 @@ UseMethod("delete_storage_dir")
 
 #' @rdname generics
 #' @export
-delete_storage_dir.blob_container <- function(container, ...)
-stop("Blob storage does not support directories")
+delete_storage_dir.blob_container <- function(container, dir, ...)
+delete_blob_dir(container, dir, ...)
 
 #' @rdname generics
 #' @export

README.md

@@ -53,9 +53,8 @@ delete_storage_container(newcont)
 
 These functions for working with objects within a storage container:
 
-- `list_storage_files`: list files/blobs in a directory (for ADLSgen2 and file storage) or blob container
-- `create_storage_dir`: for ADLSgen2 and file storage, create a directory
-- `delete_storage_dir`: for ADLSgen2 and file storage, delete a directory
+- `list_storage_files`: list files/blobs in a directory (defaults to the root directory)
+- `create_storage_dir`/`delete_storage_dir`: create or delete a directory
 - `delete_storage_file`: delete a file or blob
 - `storage_file_exists`: check that a file or blob exists
 - `storage_upload`/`storage_download`: transfer a file to or from a storage container
@@ -65,7 +64,7 @@ These functions for working with objects within a storage container:
 
 ```r
 # example of working with files and directories (ADLSgen2)
-cont <- storage_container(ad_end_tok, "myfilesystem")
+cont <- storage_container(ad_endp_tok, "myfilesystem")
 list_storage_files(cont)
 create_storage_dir(cont, "newdir")
 storage_download(cont, "/readme.txt")

man/blob.Rd

@@ -7,6 +7,8 @@
 \alias{download_blob}
 \alias{multidownload_blob}
 \alias{delete_blob}
+\alias{create_blob_dir}
+\alias{delete_blob_dir}
 \alias{blob_exists}
 \alias{copy_url_to_blob}
 \alias{multicopy_url_to_blob}
@@ -31,6 +33,10 @@ multidownload_blob(container, src, dest, recursive = FALSE,
 
 delete_blob(container, blob, confirm = TRUE)
 
+create_blob_dir(container, dir)
+
+delete_blob_dir(container, dir, recursive = FALSE, confirm = TRUE)
+
 blob_exists(container, blob)
 
 copy_url_to_blob(container, src, dest, lease = NULL, async = FALSE)
@@ -47,7 +53,7 @@ multicopy_url_to_blob(container, src, dest, lease = NULL, async = FALSE,
 
 \item{prefix}{For \code{list_blobs}, an alternative way to specify the directory.}
 
-\item{recursive}{For the multiupload/download functions, whether to recursively transfer files in subdirectories. For \code{list_blobs}, whether to include the contents of any subdirectories in the listing.}
+\item{recursive}{For the multiupload/download functions, whether to recursively transfer files in subdirectories. For \code{list_blobs}, whether to include the contents of any subdirectories in the listing. For \code{delete_blob_dir}, whether to recursively delete subdirectory contents as well (not yet supported).}
 
 \item{src, dest}{The source and destination files for uploading and downloading. See 'Details' below.}
 
@@ -73,7 +79,7 @@ multicopy_url_to_blob(container, src, dest, lease = NULL, async = FALSE,
 For \code{list_blobs}, details on the blobs in the container. For \code{download_blob}, if \code{dest=NULL}, the contents of the downloaded blob as a raw vector. For \code{blob_exists} a flag whether the blob exists.
 }
 \description{
-Upload, download, or delete a blob; list blobs in a container; check blob availability.
+Upload, download, or delete a blob; list blobs in a container; create or delete directories; check blob availability.
 }
 \details{
 \code{upload_blob} and \code{download_blob} are the workhorse file transfer functions for blobs. They each take as inputs a \emph{single} filename as the source for uploading/downloading, and a single filename as the destination. Alternatively, for uploading, \code{src} can be a \link{textConnection} or \link{rawConnection} object; and for downloading, \code{dest} can be NULL or a \code{rawConnection} object. If \code{dest} is NULL, the downloaded data is returned as a raw vector, and if a raw connection, it will be placed into the connection. See the examples below.
@@ -101,8 +107,8 @@ Blob storage does not have true directories, instead using filenames containing
 \itemize{
 \item The \code{isdir} column in the data frame output of \code{list_blobs} is a best guess as to whether an object represents a file or directory, and may not always be correct. Currently, \code{list_blobs} assumes that any object with a file size of zero is a directory.
 \item Zero-length files can cause problems for the blob storage service as a whole (not just AzureStor). Try to avoid uploading such files.
-\item The output of \code{list_blobs(recursive=TRUE)} can vary based on whether the storage account has hierarchical namespaces enabled.
-\item \code{create_storage_dir} and \code{delete_storage_dir} currently do not have methods for blob containers.
+\item \code{create_blob_dir} and \code{delete_blob_dir} function as expected only for accounts with hierarchical namespaces enabled. When this feature is disabled, directories do not exist as objects in their own right: to create a directory, simply upload a blob to that directory. To delete a directory, delete all the blobs within it; as far as the blob storage service is concerned, the directory then no longer exists.
+\item Similarly, the output of \code{list_blobs(recursive=TRUE)} can vary based on whether the storage account has hierarchical namespaces enabled.
 }
 }
 

man/generics.Rd

@@ -101,7 +101,7 @@ list_storage_files(container, ...)
 
 create_storage_dir(container, ...)
 
-\method{create_storage_dir}{blob_container}(container, ...)
+\method{create_storage_dir}{blob_container}(container, dir, ...)
 
 \method{create_storage_dir}{file_share}(container, dir, ...)
 
@@ -109,7 +109,7 @@ create_storage_dir(container, ...)
 
 delete_storage_dir(container, ...)
 
-\method{delete_storage_dir}{blob_container}(container, ...)
+\method{delete_storage_dir}{blob_container}(container, dir, ...)
 
 \method{delete_storage_dir}{file_share}(container, dir, ...)
 

tests/testthat/test05_generics.R

@@ -22,7 +22,41 @@ stor2 <- sub$get_resource_group(rgname)$get_storage_account(storname2)
 options(azure_storage_progress_bar=FALSE)
 
 
-test_that("Blob dispatch works",
+test_that("Blob dispatch works, HNS",
+{
+    endpname <- stor1$properties$primaryEndpoints$blob
+    expect_type(endpname, "character")
+    key <- stor1$list_keys()[[1]]
+
+    contname <- paste(sample(letters, 10, TRUE), collapse="")
+    dirname <- "newdir"
+    filename <- "iris.csv"
+
+    # working with a container
+    expect_is(endp <- storage_endpoint(endpname, key=key), "blob_endpoint")
+    expect_silent(cont <- storage_container(endp, contname))
+    expect_silent(create_storage_container(cont))
+
+    # working with objects within container
+    expect_silent(list_storage_files(cont))
+    expect_silent(create_storage_dir(cont, dirname))
+
+    # file transfer
+    expect_silent(storage_upload(cont, file.path("../resources", filename), filename))
+    expect_silent(storage_download(cont, filename, tempfile()))
+
+    # file existence
+    expect_false(storage_file_exists(cont, "nonexistent"))
+    expect_true(storage_file_exists(cont, filename))
+
+    # delete the objects
+    expect_silent(delete_storage_file(cont, filename, confirm=FALSE))
+    expect_silent(delete_storage_dir(cont, dirname, confirm=FALSE))
+    expect_silent(delete_storage_container(cont, confirm=FALSE))
+})
+
+
+test_that("Blob dispatch works, no HNS",
 {
     endpname <- stor2$properties$primaryEndpoints$blob
     expect_type(endpname, "character")
@@ -39,7 +73,7 @@ test_that("Blob dispatch works",
 
     # working with objects within container
     expect_silent(list_storage_files(cont))
-    expect_error(create_storage_dir(cont, dirname))
+    expect_silent(create_storage_dir(cont, dirname))
 
     # file transfer
     expect_silent(storage_upload(cont, file.path("../resources", filename), filename))
@@ -51,7 +85,7 @@ test_that("Blob dispatch works",
 
     # delete the objects
     expect_silent(delete_storage_file(cont, filename, confirm=FALSE))
-    expect_error(delete_storage_dir(cont, dirname, confirm=FALSE))
+    expect_error(delete_storage_dir(cont, dirname, confirm=FALSE)) # deleting object also deletes dir
     expect_silent(delete_storage_container(cont, confirm=FALSE))
 })