I have a directory that looks a little like this:
drw-r--r-- 1 root root 0 Jan 24 17:26 -=1=-directoryname
drw-r--r-- 1 root root 0 Jan 24 17:26 -=2=-directoryname
drw-r--r-- 1 root root 0 Jan 24 17:26 -=3=-directoryname
drw-r--r-- 1 root root 0 Jan 24 17:26 -=4=-directoryname
drw-r--r-- 1 root root 0 Jan 24 17:26 -=5=-directoryname
I am trying to write a script to change these folders from
-=1=- Folder#1
to strip off the "-=1=-" section, but alas I am having no luck.
Can anyone help me find a solution to this?
So far my script below has failed me.
#!/bin/bash
for i in {1..250}
do
rename "-=$i=-" ""*
i=i+1
done
I have used the 1..250 because there are 250 folders.
Given the number, you can manufacture the names and use the mv command:
#!/bin/bash
for i in {1..250}
do
mv "-=$i=- Folder#$i" "Folder#$i"
done
With the Perl-based rename command (sometimes called prename), you could use:
rename 's/-=\d+=- //' -=*=-*Folder#*
or, given the revised question (the information after the pattern isn't fixed):
rename 's/-=\d+=- //' -=*=-*
This worked! Can you please explain how it worked? What's the \d+ for?
The \d is Perl regex notation for a digit 0..9. The + modifier indicates 'one or more'. So, the regex part of s/-=\d+=- // looks for a minus, an equals, one or more digits, an equals, a minus and a space. The replace part converts all of the matched material into an empty string. It's all surrounded by single quotes so the shell leaves it alone (though there's only the backslash that's a shell metacharacter in that substitute command, but the backslash and space would need protecting if you omitted the quotes).
I'm not sure how you'd use the C-based rename command for this job; it is much less powerful than the Perl-based version.
Related
Running rm projects/artwork/My Project (543893)/Images/*.png at the command line in Debian does not work because of the spaces and the parenthesis. I need to escape them. But I thought quotation marks "" were an alternative to escaping. It works with other commands, such as cd, but not with rm. Why is that?
Because globbing does not work with quoting:
$ ll /tmp/"d i r"/*.png
-rw-r--r--. 1 postgres postgres 0 May 26 14:02 /tmp/d i r/a.png
-rw-r--r--. 1 postgres postgres 0 May 26 14:02 /tmp/d i r/p.png
$ ll "/tmp/d i r/*.png"
ls: cannot access /tmp/d i r/*.png: No such file or directory
$ rm "/tmp/d i r/*.png"
rm: cannot remove ‘/tmp/d i r/*.png’: No such file or directory
$ rm /tmp/"d i r"/*.png
$ ll /tmp/"d i r"/*.png
ls: cannot access /tmp/d i r/*.png: No such file or directory
You should ensure that:
the spaces and parentheses are inside the quotation marks, so that they are treated as literals, because otherwise these characters will have special meanings for the shell (spaces to separate command-line arguments, and parentheses for a subshell command)
but for the * you want the exact opposite: it must be outside the quotation marks so that it is indeed given its special meaning as a wildcard, not a literal *.
Remember that you do not need to apply quotation marks to the whole string, only the part that needs treating as a literal and otherwise would be given a special meaning by the shell.
So for example you could use:
rm projects/artwork/"My Project (543893)"/Images/*.png
Similarly, if you use escapes, escape the spaces and parentheses, but not the wildcard:
rm projects/artwork/My\ Project\ \(543893\)/Images/*.png
(In other words, you would not use \*.)
I've written a script (foo) which makes a simple sed replacement on text in the input file. I have a directory (a) containing a large number of subdirectories (a/b1, a/b2 etc) which all have the same subdirs (c, etc) and contain a file with the same name (d). So the rough structure is:
a/
-b1/
--c/
---d
-b2/
--c/
---d
-b3/
--c/
---d
I want to run my script on every file (d) in the tree. Unfortunately the following doesn't work:
sudo sh foo a/*/c/d
how do I use wildcards in a bash command like this? Do I have to use find with specific max and mindepth, or is there a more elegant solution?
The wildcard expansion in your example should work, and no find should be needed. I assume a b and c are just some generic file names to simplify the question. Do any of your folders/files contain spaces?
If you do:
ls -l a/*/d/c
are you getting the files you need listed? If so, then it is how you handle the $* in your script file. Mind sharing it with us?
As you can see, wildcard expansion works
$ ls -l a/*/c/d
-rw-r--r-- 1 user wheel 0 15 Apr 08:05 a/b1/c/d
-rw-r--r-- 1 user wheel 0 15 Apr 08:05 a/b2/c/d
-rw-r--r-- 1 user wheel 0 15 Apr 08:05 a/b3/c/d
In Linux how do I move files without replacing if a particular file already exists in the destination?
I tried the following command:
mv --backup=t <source> <dest>
The file doesn't get replaced but the issue is the extension gets changed because it puts "~" at the back of the filename.
Is there any other way to preserve the extension but only the filename gets changed when moving?
E.g.
test~1.txt instead of test.txt~1
When the extension gets replaced, subsequently you can't just view a file by double clicking on it.
If you want to make it in shell, without requiring atomicity (so if two shell processes are running the same code at the same time, you could be in trouble), you simply can (using the builtin test(1) feature of your shell)
[ -f destfile.txt ] || mv srcfile.txt destfile.txt
If you require atomicity (something that works when two processes are simultaneously running it), things are quite difficult, and you'll need to call some system calls in C. Look into renameat2(2)
Perhaps you should consider using some version control system like git ?
mv has an option:
-S, --suffix=SUFFIX
override the usual backup suffix
which you might use; however afaik mv doesn't have a functionality to change part of the filename but not the extension. If you just want to be able to open the backup file with a text editor, you might consider something like:
mv --suffix=.backup.txt <source> <dest>
how this would work: suppose you have
-rw-r--r-- 1 chris users 2 Jan 25 11:43 test2.txt
-rw-r--r-- 1 chris users 0 Jan 25 11:42 test.txt
then after the command mv --suffix=.backup.txt test.txt test2.txt you get:
-rw-r--r-- 1 chris users 0 Jan 25 11:42 test2.txt
-rw-r--r-- 1 chris users 2 Jan 25 11:43 test2.txt.backup.txt
#aandroidtest: if you are able to rely upon a Bash shell script and the source directory (where the files reside presently) and the target directory (where you want to them to move to) are same file system, I suggest you try out a script that I wrote. You can find it at https://github.com/jmmitchell/movestough
In short, the script allows you to move files from a source directory to a target directory while taking into account new files, duplicate (same file name, same contents) files, file collisions (same file name, different contents), as well as replicating needed subdirectory structures. In addition, the script handles file collision renaming in three forms. As an example if, /some/path/somefile.name.ext was found to be a conflicting file. It would be moved to the target directory with a name like one of the following, depending on the deconflicting style chosen (via the -u= or --unique-style= flag):
default style : /some/path/somefile.name.ext-< unique string here >
style 1 : /some/path/somefile.name.< unique string here >.ext
style 2 : /some/path/somefile.< unique string here >.name.ext
Let me know if you have any questions.
Guess mv command is quite limited if moving files with same filename.
Below is the bash script that can be used to move and if the file with the same filename exists it will append a number to the filename and the extension is also preserved for easier viewing.
I modified the script that can be found here:
https://superuser.com/a/313924
#!/bin/bash
source=$1
dest=$2
file=$(basename $source)
basename=${file%.*}
ext=${file##*.}
if [[ ! -e "$dest/$basename.$ext" ]]; then
mv "$source" "$dest"
else
num=1
while [[ -e "$dest/$basename$num.$ext" ]]; do
(( num++ ))
done
mv "$source" "$dest/$basename$num.$ext"
fi
Here's an output of diff -u "temp temp/docs 1.txt" "temp temp/docs 2.txt":
--- temp temp/docs 1.txt Mon Apr 7 16:15:08 2014
+++ temp temp/docs 2.txt Mon Apr 7 16:18:45 2014
## -2,6 +2,6 ##
22
333
4444
-555555
+55555
666666
7777777
However, feeding this diff to patch -u fails with following message:
can't find file to patch at input line 3
Perhaps you should have used the -p or --strip option?
The text leading up to this was:
--------------------------
|--- temp temp/docs 1.txt Mon Apr 7 16:15:08 2014
|+++ temp temp/docs 2.txt Mon Apr 7 16:18:45 2014
--------------------------
Apparently, the spaces are the problem; is there a way to make patch to work on files with spaces in names?
No, GNU patch doesn't support this. Here's the official statement: http://www.gnu.org/software/diffutils/manual/html_node/Unusual-File-Names.html#Unusual%20File%20Names
Gnu patch 2.6.1 (linux) seems to obey at least 1 space (not tried with more) if the filename is separated from the date with tab.
YYMV
I encountered the same problem when trying to establish conventions how to do manual version control with diff and patch.
I found out that GNU "diff" creates quoted path names in the patch headers if they contain spaces, while BusyBox "diff" doesn't.
Neither GNU nor BusyBox "patch" accepts quoted path names.
If the problem is just embedded spaces within filenames, it can therefore be avoided by using "busybox patch" rather than GNU "patch".
Another solution is to postprocess the output of GNU "diff" before feeding it into "patch":
sed 's,^\([-+]\{3\} \)"\([^"]*\)",\1\2,' $PATCHFILE | patch -p1
This works whether $PATCHFILE was created with GNU or busybox diff, but will only work with unified diff format.
Unfortunately, it turns out that leading or trailing spaces in filenames cannot be preserved with this method, as "patch" will skip them when parsing the path names from the patch instructions.
The approach will neither work if the filename starts with a literal double quote - but then, who uses such file names?
Most of the time, however, the above approach works just fine.
Finally a note of other approaches I have also tried but which did not work:
First I tried to replace the quotation of the whole path names by individually quoted path name components. This failed because "patch" does not use double quotes as meta-characters at all. It considers them to be normal literal characters.
Then I tried to replace all spaces by "\040" like CVS does - but "patch" does not seem to accept octal-escapes either, and this failed too.
What does the "->" notation mean in Linux .
eg. When I do ls -l in a particular folder, I get the following.
lrwxrwxrwx 1 root root 29 Feb 27 12:23 ojdbc.jar -> /apps/hadoop/sqoop/ojdbc6.jar
Is the first file a placeholder of the second one?
Kind Regards.
It means the file is not a physical file, but a symbolic link pointing to the file to the right of the arrow.
The command "ls -l" uses "->" to denote a symbolic-link (that is, a psuedo-file which only points to another file).
In your example ojdbc.jar is a symbolic-link to /apps/hadoop/sqoop/ojdbc6.jar.
I'm not aware that this meaning holds beyond ls, however.