ansaurus

Question

With sed or awk, how do I match from the end of the current line back to a specified character?

Answer 1

+4 A:

Use command basename

$~hawk] basename /var/lib/mlocate
mlocate

Neeraj 2010-02-07 19:45:23

Thanks alot, so simple

paultop6 2010-02-07 19:48:54

basename is good for one file at a time; to process the names in a file, sed or awk works as well if not better.

Jonathan Leffler 2010-02-07 20:39:08

Answer 2

+1 A:

There's really no need to use sed or awk here, simply us basename

IFS=$'\n'
for file in $(cat filelist); do
   basename $file;
done

If you want the directory part instead use dirname.

honk 2010-02-07 19:50:15

Won't work if filenames contain spaces.

Dennis Williamson 2010-02-07 20:55:16

@dennis who has spaces in filenames anyway??? ;) fixed

honk 2010-02-07 21:16:02

a lot a out there has spaces in file names

ghostdog74 2010-02-08 00:05:30

no need to call external command basename.

ghostdog74 2010-02-08 00:20:43

@ghost I added the `IFS` line. That changes the character that separates entries in that list I iterate over from space to `\n`.

honk 2010-02-08 00:21:16

+1 for $'...' syntax. I'd never encountered that (after all these years of bashing)

Adrian Pronk 2010-02-10 12:14:49

Answer 3

+4 A:

$ sed -e 's!^.*/!!' locations.txt
mlocate
mlocate.conffiles
mlocate.list
mlocate.md5sums
mlocate.postinst
mlocate.postrm
mlocate.prerm

Regular-expression quantifiers are greedy, which means .* matches as much of the input as possible. Read a pattern of the form .*X as "the last X in the string." In this case, we're deleting everything up through the final / in each line.

I used bangs rather than the usual forward-slash delimiters to avoid a need for escaping the literal forward slash we want to match. Otherwise, an equivalent albeit less readable command is

$ sed -e 's/^.*\///' locations.txt

Greg Bacon 2010-02-07 20:10:59

Answer 4

+3 A:

I am for "basename" too, but for the sake of completeness, here is an awk one-liner:

awk -F/ 'NF>0{print $NF}' <file.txt

marco 2010-02-07 20:23:39

for completeness get rid of the empty lines too `{if(NF>0) print $NF}` ;) +1

honk 2010-02-07 20:30:08

you are right, I have edited my code ;)

marco 2010-02-07 23:33:44

Answer 5

+1 A:

Pure Bash:

while read -r line
do
    [[ ${#line} != 0 ]] && echo "${line##*/}"
done < files.txt

Edit: Excludes blank lines.

Dennis Williamson 2010-02-07 20:58:11

Cool. Any chance to squeeze a test for empty `line` into the `{}`?

honk 2010-02-07 21:21:57

Answer 6

A:

@OP, you can use awk

awk -F"/" 'NF{ print $NF }' file

NF mean number of fields, $NF means get the value of last field

or with the shell

while read -r line
do
    line=${line##*/} # means longest match from the front till the "/" 
    [ ! -z  "$line" ] && echo $line
done <"file"

NB: if you have big files, use awk.

ghostdog74 2010-02-08 00:14:51

Answer 7

A:

Thius would do the trick too if file contains the list of paths

 $ xargs -d '\n' -n 1 -a file basename

dtmilano 2010-02-08 01:57:40

this will break on files with spaces

ghostdog74 2010-02-08 02:23:15

-d '\n' prevents breaking the line on whitespaces

dtmilano 2010-02-08 18:45:22

Answer 8

A:

This is a less-clever, plodding version of gbacon's:

sed -e 's/^.*\/\([^\/]*\)$/\1/'

13ren 2010-02-10 12:04:29

ansaurus

tags:

views:

answers:

With sed or awk, how do I match from the end of the current line back to a specified character?

related questions