3

I have 1,000,000 files in some folders and subfolders. I want to rename them from lowercase to uppercase using shell commands. I don't want to modify the extension. only filename part.

I have found this one:

rename 's/^([^.]*)\.(.*)$/\U$1\E.$2/' *

but it is not recursive and only works on files in current folder.

Then I tried this one:

find . -depth  -execdir rename 's/^([^.]*)\.(.*)$/\U$1\E.$2/' {} \;

But no files changed.

How can I use it recursively?

Improve this question
4
  • And your shell is…? See your shell's manual to find out whether it supports ** syntax in globbing. bash (with globstar turned on) and zsh does. Commented Jul 24, 2013 at 10:37
  • If possible, you should backup your files before running anything. Commented Jul 24, 2013 at 14:17
  • 1
    See also Renaming files to have lower case extensions with 'rename' and Lowercasing all directories under a directory for other methods of changing the case of file names. Commented Jul 25, 2013 at 2:00
  • Obligatory 'Some people, when confronted with a problem, think “I know, I'll use regular expressions.” Now they have two problems.' but with a smattering of history that I'd not seen regex.info/blog/2006-09-15/247 Commented Jul 25, 2013 at 5:19

4 Answers 4

Reset to default
3

Assuming that you have the Perl rename provided by Debian and derived distributions such as Debian and Ubuntu, you're almost there. The problem is that -execdir passes a file name prefixed with ./ to the command. (The reason for that is that some commands treat arguments starting with some characters specially; this way, if you have a file called -foo, it's passed as ./-foo and therefore treated as a file and not as an option.) With your regex, this results in $1 being always empty, and hence the new name is identical to the old name.

Accommodate for this ./ in your regular expression.

find . -depth  -execdir rename 's/^(\.\/[^.]*)\.(.*)$/\U$1\E.$2/' {} \;
Improve this answer
3
  • The above command fails on directory name with a dot(.) separator. ex : dir.xyz/abc.txt is being renamed to DIR.xyz/ABC.txt Commented Jul 25, 2013 at 5:00
  • It is ambiguous in the OP whether files or {files and directories} were to be changed. I read -type f and Giles infers the opposite. We're both wrong and both right. Commented Jul 25, 2013 at 5:21
  • @Ameer rename will never be called with an argument of the form dir.xyz/abc.txt, because -execdir was used and not -exec. All the arguments to rename will be of the form ./SOMETHING where SOMETHING doesn't contain any slash. So dir.xyz will be renamed to DIR.xyz, which is consistent with the expressed requirements. Thanks to -depth, this happens after dir.xyz/abc.txt has been renamed to dir.xyz/ABC.txt, so there won't be a problem with find attempting to move into a directory that's been renamed without it knowing. Commented Jul 25, 2013 at 11:18
0

find . -type f -exec rename -v 's/(\w+).(\w+)$/\U$1\E.$2/' {} \;

Improve this answer
4
  • This is close to correct but it will fail on filenames with spaces or punctuation in them. It would also be a better answer if you explained why you are using -type. Commented Jul 24, 2013 at 10:04
  • @msw What problem do you see with file names with whitespace? This will, however, fail on files in subdirectories that contain lowercase letters, as it will attempt to rename e.g. foo/bar.ext to FOO/BAR.ext (hd. used -execdir to work around this problem). Commented Jul 24, 2013 at 21:56
  • @Gilles , there is no fail with directory as -type f only gives the file path. and renames $1 & $2 only touches the files inside the dir. Commented Jul 25, 2013 at 5:01
  • @Giles I think you know the answer as you used the more inclusive pattern in your answer, but \w+ will not match test file or test-file and many other such likely names. It also uses . when it should have \.. Also since @Ameer commented while I was writing this, without -execdir you need to ensure that only the last path component is matched yielding (untested) s!/([^/]*)\.([^/]*)$!—! Commented Jul 25, 2013 at 5:08
0

You can use this little bash script.

#!/bin/bash

find . -type f -exec sh -c  '
    dn=`dirname "$1"`
    bn=`basename "$1"`
    fn=`echo "${bn%.*}"`
    ext=`echo "${bn##*.}"`
    FN=`echo "${fn^^}"`
    newfn="${FN}"."${ext}"
    #echo ${dn}/$newfn
    if [ "${fn}" = "${FN}" ] ;then : ;else echo "${bn} renamed to ${newfn}";mv "$1" "${dn}/${newfn}";fi
' _ {} \;
Improve this answer
3
  • And how will that traverse the directory structure recursively, as requested in the question? Commented Jul 24, 2013 at 12:22
  • @manatwork: This new script takes into consideration the subdir part of the question that I overlooked before. Commented Jul 24, 2013 at 12:36
  • 1
    Ok, it works fine now. But why you use fn=`echo "${bn%.*}"` instead of fn="${bn%.*}"? Commented Jul 24, 2013 at 12:46
-1 
ls -r | rename 's/^([^.]*)\.(.*)$/\U$1\E.$2/'

Is this OK?

Improve this answer
3
  • never use ls for anything Commented Jul 24, 2013 at 10:29
  • 3
    Except listing the contents of filesystems? I find find -maxdepth 1 -ls to be a bit of a nuisance for that. ;) But joking temporarily aside, parsing ls is a baaaad idea unless you're doing something very constrained and well-defined. And probably not even then. Commented Jul 24, 2013 at 11:14
  • Wondering who upvoted a ls call with --reverse instead of --recursive? Commented Jul 24, 2013 at 12:43

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.