Zsh Mailing List Archive
Messages sorted by: Reverse Date, Date, Thread, Author

Surprising behaviour with numeric glob sort



Some odd behaviour:

$ echo *(n)
a3 a10 a-2
$ rm a10
$ echo *(n)
a-2 a3

(order of a-2 and a3 reversed after a10 is removed)

$ echo $LANG
en_GB.UTF-8
$ /lib/x86_64-linux-gnu/libc.so.6
GNU C Library (Ubuntu GLIBC 2.23-0ubuntu7) stable release version 2.23, by Roland McGrath et al.


I think the problem is that here, zsh uses a non-total order. We have:

a3 < a10 (same prefix, so numeric comparison of 3 < 10)
a10 < a-2 (as per strcoll(). - ignored in first pass. No same prefix, so no numeric comparison)
a-2 < a3 (as per strcoll(). - ignored in first pass. No same prefix, so no numeric comparison)

We've got a circle here. AFAIK, the behaviour is unspecified if
qsort() is called with a comparison function that doesn't
implement a total order, so the result could be random.

Once we remove a10, we're OK.

GNU sort -V and ls -v seem to "handle" the issue by giving up on
strcoll() which is not a lot better:

$ ls
a0  á0  a10  a-2  a3  b0
$ ls -v
a0  a3  a10  a-2  b0  á0
$ ls | sort
a0
á0
a10
a-2
a3
b0
$ ls | sort -V
a0
a3
a10
a-2
b0
á0

Maybe a better approach would be to break down the strings
between non-numeric and numeric parts and use strcoll() on the
non-numeric and number comparison on the numeric parts, stopping
at the first difference.

a3 < a-2 because a < a- as per strcoll (even though a3 > 1-2)

-- 
Stephane



Messages sorted by: Reverse Date, Date, Thread, Author