Issue in Metacharacter

arun natarajan · 11-02-2015, 12:30 AM

Hi,

In 1st eg: Am trying to filter the files starting with small letters in first position and numeric in 2nd position. But it also retrieves capital letters. I dunno why.

In 2eg: Am trying to filter the files starting with capital letters in first position and numeric in 2nd position. But it also retrieves small letters in 1st position. I dunno why.

Let me know, if anyone knows the root cause of this.

[root@localhost ~]# ls -ltr [a-z][0-9]*
----rwxrwx 1 root root 10 Oct 29 23:03 a2.link
-rwxr--r-- 1 root root 19 Oct 29 23:04 a3.op
-rw-r--r-- 1 root root 31478 Oct 30 03:08 a20
-rw-r--r-- 1 root root 0 Nov 1 19:45 b123
-rw-r--r-- 1 root root 0 Nov 1 19:55 a2Bcat
-rw-r--r-- 1 root root 0 Nov 1 19:55 A2bcat
-rw-r--r-- 1 root root 0 Nov 1 20:06 a1
-rw-r--r-- 1 root root 0 Nov 1 20:32 a123
-rw-r--r-- 1 root root 0 Nov 1 20:32 a12
-rw-r--r-- 1 root root 0 Nov 1 20:32 a12as
-rw-r--r-- 1 root root 0 Nov 1 20:32 a12as!
-rw-r--r-- 1 root root 0 Nov 1 20:32 a122222

[root@localhost ~]# ls -ltr [A-Z][0-9]*
-rw-r--r-- 1 root root 0 Nov 1 19:45 b123
-rw-r--r-- 1 root root 0 Nov 1 19:55 A2bcat
[root@localhost ~]#

chrism01 · 11-02-2015, 04:06 AM

See http://www.tldp.org/LDP/abs/html/globbingref.html & http://tldp.org/LDP/GNU-Linux-Tools-...tml/x11655.htm for 'globbing'.
This similar but less powerful than regexes.
If you want to use regexes, pipe the o/p through grep which does understand regexes.

allend · 11-02-2015, 05:15 AM

My guess is that you have LC_COLLATE (check the output of locale) set to something other than LC_COLLATE=C.
http://teaching.idallen.com/net2003/...acter_sets.txt

grail · 11-02-2015, 08:50 AM

allend - I must admit this caught me out as my computer also is returning odd info, but I did the following and still get erroneous data back?

Code:

$ LC_ALL=C ls [a-z]*
A2bcat  b123

So I too am looking forward to see why this suddenly does not work

chrism01 - I read through both pages supplied and one would think the above should behave differently??

On a little further testing, it would seem I have an ignore case set somewhere, however, I have also used my unaliased ls and get the same messed up results:

Code:

$ LC_ALL=C /bin/ls a*
A2bcat

allend · 11-02-2015, 09:25 AM

Hmm - curious. I see

Code:

bash-4.3$ locale
LANG=en_AU
LC_CTYPE="en_AU"
LC_NUMERIC="en_AU"
LC_TIME="en_AU"
LC_COLLATE=C
LC_MONETARY="en_AU"
LC_MESSAGES="en_AU"
LC_PAPER="en_AU"
LC_NAME="en_AU"
LC_ADDRESS="en_AU"
LC_TELEPHONE="en_AU"
LC_MEASUREMENT="en_AU"
LC_IDENTIFICATION="en_AU"
LC_ALL=
bash-4.3$ ls
A123  B123  a123  b123
bash-4.3$ ls [A-Z]*
A123  B123
bash-4.3$ ls [a-z]*
a123  b123

and unaliased, just to be sure.

Code:

bash-4.3$ \ls
A123  B123  a123  b123
bash-4.3$ \ls [a-z]*
a123  b123
bash-4.3$ \ls [A-Z]*
A123  B123

chrism01 · 11-02-2015, 07:04 PM

Well, I tried '\ls', but it made no difference; it still lists all the files.
(Centos 6.7 if that helps anyone)

Code:

 ls
a  a1  b  t.sh  t.t  T.T

 \ls [a-z]*
a  a1  b  t.sh	t.t  T.T

 \ls [A-Z]*
b  t.sh  t.t  T.T         # hmmm now THAT is odd; didn't list 'a' files

grail · 11-20-2015, 03:13 AM

This one seemed to have fallen off the radar, but I am still in need of a solution myself as it is now affecting scripts and the like when using some of these conjunctions

I have an older machine in vm running bash 4.3.30 and it does not seem to have this issue. Do we think it might be a bash issue??

grail · 12-13-2015, 11:39 PM

Thought I would give this one another touch to see if anyone has thought / heard of anything?

Here is yet another example of this weird behaviour I just ran into:

Code:

$ ls [D]*
d2.awk*  d2.rb*  d2.sh*  d.awk*  dd_mem_test.sh*  d.html  d.pl*  d.py*  d.rb*  d.sh*

# Also tried unaliased
$ \ls [D]*
d2.awk	d2.rb  d2.sh  d.awk  dd_mem_test.sh  d.html  d.pl  d.py  d.rb  d.sh

chrism01 · 12-14-2015, 02:27 AM

Well, it is quite an interesting one... iirc David_The_H (?) used to be a bit of a whiz at this stuff, but I haven't seen a post from him in a long time.

Whilst I'd love to see a definitive explanation, if you're using this for real (grail), then like I said above, I'd pipe it through eg grep and use real regexes.

grail · 12-14-2015, 03:27 AM

Well if for real you mean on a regular basis, yes, as part of using my system I often use ls and with globbing. To then have to change to also include grep would seem overkill on such a simple task

Guess I will have to keep searching. It does surprise me that more people have not come across this issue (assuming it is a bash thing and not some sill setting I have somewhere)

Will let you know what I find

grail · 12-14-2015, 05:33 AM

I have a vm of my own design which does not have the issue and has latest bash. It would seem my distro provider (Manjaro) took it upon themselves to add the following to .bashrc:

Code:

shopt -s nocaseglob

Turning this off has resolved the most recent example:

Code:

$ ls
ruby/  tmp/  d2.awk*  d2.rb*  d2.sh*  d.awk*  dd_mem_test.sh*  d.html  d.pl*  d.py*  d.rb*  d.sh*  f1  f2  f3  f4  lxapp-fix-gtk3.sh*  mount_iso.sh*  pwd_gen.awk*  system_info.sh*  vimrc.txt
$ ls [D]*
ls: cannot access [D]*: No such file or directory

Unfortunately, it did not help for the ranges