[GitHub] [arrow] nealrichardson commented on a change in pull request #8351: ARROW-9870: [R] Friendly interface for filesystems (S3)

[GitBox <gi...@apache.org>]

nealrichardson commented on a change in pull request #8351:
URL: https://github.com/apache/arrow/pull/8351#discussion_r501998550



##########
File path: r/R/feather.R
##########
@@ -24,9 +24,8 @@
 #' and the version 2 specification, which is the Apache Arrow IPC file format.
 #'
 #' @param x `data.frame`, [RecordBatch], or [Table]
-#' @param sink A string file path, URI, or [OutputStream]
-#' @param filesystem A [FileSystem] where `sink` should be written if it is a
-#' string file path; default is the local file system
+#' @param sink A string file path, URI, or [OutputStream], or path in a file
+#' system (`SubTreeFileSystem`)

Review comment:
       It is unintuitive, but here's how I ended up here. (1) STFS is a (fs, path) pair that already exists so why invent something else and have to define all the methods for it; (2) if `fs$path(string)` is how you "point" to a file or directory to pass to a reader, the end user need not know what kind of object that produces; (2b) that same user isn't going to understand the originally intended purpose of STFS anyway, so they wouldn't necessarily see it as being misappropriated, just a funny name for something; (3) for datasets, where I am pointing at a directory--such as an S3 bucket--an STFS actually is an appropriate vehicle. So I figured I'd work with what I had until it didn't make sense (i.e. I could answer the "why" in point 1)




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org