os: Remove documentation does not explain how it differs from RemoveAll #26507
Comments
|
Not a helpful comment: I wish that "rm" just removed an empty directory, that I didn't need to use "rmdir". More helpful: "rm" and "rmdir" both use the underlying "unlink" system call on Unix; the different experiences they provide is part of their design. The OS just has one system call, which we wrap with Consider permissions: Do we need to add a paragraph to every I have no idea what Windows does here, but I know that It might be OK to say,
adding the parenthetical. I would protest at doing more than that. |
|
I think the distinction I'd draw with the permissions case is that the permissions behavior is fairly consistent in general. Directory removal has been fairly inconsistent bbetween programming languages. Users who aren't used to thinking at the syscall level are quite possibly used to "drag file to trash" or something similar, and may not expect os.Remove() to fail even on non-empty directories; users who are used to thinking in syscalls may think "directories need a different remove operation than files". My first thought was that if a thing removes a file-or-directory, it presumably removes directories recursively; I only guessed otherwise because the adjacent RemoveAll implied a distinction. I agree that the (empty) qualifier is enough to make the intent unambiguous. (But in fact, I wasn't even 100% sure that this was the intent, as opposed to how it happened to work on existing targets.) (And I agree about rm/rmdir, but I think we may be >45 years late to win that one.) |
ianlancetaylor
changed the title
ianlancetaylor
added
the
NeedsFix
|
Change https://golang.org/cl/127835 mentions this issue: |
What version of Go are you using (
go version)?1.10
Does this issue reproduce with the latest release?
yes
What operating system and processor architecture are you using (
go env)?N/A (documentation)
What did you do?
Read documentation.
What did you expect to see?
A description of what os.Remove does that might differ from os.RemoveAll. :)
What did you see instead?
The blank assertion that os.Remove "removes a file".
ISO C's
removewill remove a file or directory, but fail if the directory is empty. On the other hand,rmwill remove only files, if not given the-rflag, and you needrmdirto remove a directory. I have heard horror stories about filesystems where you could remove a directory that wasn't empty, and nightmarish things occurred. (I hope this isn't real.)My point is: I think it would be good for the docs for
os.Removeto explicitly state behavior for likely edge cases, such as "can remove both files and directories" or "will not remove directories that have other contents." A quick straw poll revealed at least one person who assumed it would definitely not work on empty directories, and I suspect there are people out there assuming it will remove directories even if they are empty (probably recursively). A sentence outlining the intended behavior would reduce confusion.The text was updated successfully, but these errors were encountered: