[R] Bug in list.files(full.names=T)

Martin Maechler m@ech|er @end|ng |rom @t@t@m@th@ethz@ch
Mon Dec 20 09:46:23 CET 2021

>>>>> Mario Reutter 
>>>>>     on Sat, 18 Dec 2021 15:55:37 +0100 writes:

    > Dear everybody,

    > I'm a researcher in the field of psychology and a
    > passionate R user. After having updated to the newest
    > version, I experienced a problem with list.files() if the
    > parameter full.names is set to TRUE.  A path separator "/"
    > is now always appended to path in the output even if path
    > %>% endsWith("/"). This breaks backwards compatibility in
    > case path ends with a path separator. The problem occurred
    > somewhere between R version 3.6.1 (2019-07-05) and 4.1.2
    > (2021-11-01).

    > Example:
    >>> list.files("C:/Data/", full.names=T)
    > C:/Data//file.csv

    > Expected behavior: Either a path separator should never be
    > appended in accordance with the documentation: "full.names
    > a logical value. If TRUE, the directory path is prepended
    > to the file names to give a relative file path."  Or it
    > could only be appended if path doesn't already end with a
    > path separator.

This expected behavior has never been documented AFAIK.
I tried R 3.6.1 on Linux  and it already behaved like that:  If
you append a path separator  it is kept in addition to the new
one even though it's not needed:

> list.files("/tmp", "^[abc]", full.names = TRUE)
[1] "/tmp/check_proc__localuser"

> list.files("/tmp/", "^[abc]", full.names = TRUE)
[1] "/tmp//check_proc__localuser"

Why would one ever *add* a final unneeded path separator, unless
one wanted it?

Note that the default is  ".",  not "./"  ..

I think the change you see made R-for-Windows compatible
to the rest of the univeRse where list.files() aka dir() always
behaved like this.

I agree that ideally this would have been mentioned in some of
the NEWS; maybe it *is* mentioned in the rw-faw (R for Windows
FAQ) or other R for Windows news.. ?

    > My question would now be if this warrants a bug report?

I don't think so.
As I'm saying above, I think this has rather been a bug fix,
making R more universal / less platform-dependent.

Last but not least: You'd ideally update more than every 2.5 years...

Martin Maechler

    > And if you agree, could someone issue the report since I'm
    > not a member on Bugzilla?

    > Thank you and best regards, Mario Reutter

More information about the R-help mailing list