Add keyword remove_destination to mv function

[peter1000]

This is a follow up issue from comments:
#11024 (comment)

julia could add the kwarg: remove_destination::Bool=false to mv to be in line with cp

cp(src::AbstractString, dst::AbstractString; remove_destination::Bool=false, follow_symlinks::Bool=false)

If that is wanted I could do it.

reply StefanKarpinski
@StefanKarpinski > Yes please, @peter1000!

---

I just had a fast look and I think there are some other issues too:

1. current docs say Move a file from src to dst. Related to second question: is it suppose to work also on directories?
2. The real issue is here again similar we had in the cp discussion. symlinks. I tried a fast mv on a directory and obviously any absolute symlinks within the rootdir get broken if the folder is renamed.

I can fix this but the question is: should we also add the kw: follow_symlinks similar to copy.

cp(src::AbstractString, dst::AbstractString; remove_destination::Bool=false, follow_symlinks::Bool=false)

Relevant code portion: fs.jl
file.jl

[peter1000]

according to my reading of the src: mv with the current implementation was only supposed for files: uv_fs_rename lbuv

Equivalent to rename(2).

rename(2)

Name
rename - change the name or location of a file
Synopsis
#include <stdio.h>

int rename(const char *oldpath, const char *newpath);
Description
rename() renames a file, moving it between directories if required.

[simonster]

remove_destination seems reasonable, but I don't think we should try to do anything about absolute symlinks when moving files. I don't know of anything else that tries to handle this case.

rename should work just fine for directories. From the man page you linked:

oldpath can specify a directory. In this case, newpath must either not exist, or it must specify an empty directory.

But I noticed that mv does not presently perform a recursive copy when the rename throws an error (i.e., when the source and destination are on different file systems). It seems like it should.

[peter1000]

just something I noticed: currently the test suit uses: Base.samefile to test drectories stat info. src

I just wondered if the name Base.samefile is not a bit misleading in such a case. Would it be worth a consideration to have a samedir or just a samepath.

[simonster]

samefile is the Python name. I think having separate functions for files and directories would be annoying and not very useful. samepath doesn't seem like it would do the same thing since the same file can reside at multiple paths due to hard links. I think sameinode would be more correct (except that NTFS doesn't call them inodes) but "inode" may be too technical for us.

[StefanKarpinski]

I think sticking with the Python name here is probably best and making sure the documentation is clear.

[peter1000]

There is no official documentation as it is not exported: I could add a small comment in the srcfile that it is supposed to work for files and directories when I send this PR.

[jakebolewski]

closed by #11172