[SOLVED] Extracting pathnames from a string morass

andrew.comly · 06-10-2016, 11:34 PM

Can't grep pathnames out of string
I have a string:

Code:

tmp=Only in /home/a/.config/cairo-dock/third-party/Clear/icons: pcmanfm.svg Only in /home/a/.config/cairo-dock/third-party/Clear/icons: system-file-manager.svg Only in /home/a/.config/cairo-dock/third-party/Clear/icons: thunar.svg Only in /home/a/.config/cairo-dock/third-party/Clear/icons: thunderbird.svg Only in /home/a/.config/cairo-dock/third-party/Clear/icons: utilities-system-monitor.svg Only in /home/a/.config/cairo-dock/third-party/Clear/icons: utilities-terminal.svg

I would like to extract anything that is a path. "grep" comes to mind and I utilize the pattern that any 'path' will start with a "/". My attempts below:

Code:

$ echo $tmp | grep /
$ echo $tmp | grep ^/
$ echo $tmp | grep -w ^/
$ echo $tmp | grep -w /
$ echo $tmp | grep -w "/"
$ echo $tmp | grep -w '\</'
$ echo $tmp | grep -w '\:\>'
$ echo $tmp | grep -w ':\>'
$ echo $tmp | grep -w '\</*:\>'
$ echo $tmp | grep '\</*:\>'

Results
{fail}

Unfortunately the above attempts yeild no results. It was interesting that when I tried

Code:

$ echo $tmp | grep /

it printed out the entire string, which is why I used -w quite often.

Lastly I wondered how the above commands could be printing out to standard error, so I tried:

Code:

$ echo $tmp 2>&1 | grep /
$ echo $tmp | grep / 2>&1
$ echo $tmp | grep -w / 2>&1

Alas, still I couldn't extract the path.

Can the paths in the above string be extracted? Should I use a 'for in 'echo list'; do * ; done' loop?

firstfire · 06-11-2016, 12:42 AM

Code:

$ echo $tmp | grep -oE '(/[^/: ]+)+'
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons

PS: Not sure about part after colon -- is that belong to a path?

grail · 06-11-2016, 01:25 AM

I went slightly simpler:

Code:

$ grep -oE '/[^:]+' <<< "$tmp"
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons
/home/a/.config/cairo-dock/third-party/Clear/icons

I presume the original setting of the variable is simply missing the quotes around the string

andrew.comly · 06-12-2016, 08:58 AM

Quote:

Originally Posted by firstfire

Code:

$ echo $tmp | grep -oE '(/[^/: ]+)+'

Thanks firstfire, you delivered an answer with pinpoint accuracy!!

Any online learning sources you recommend for learning grep?

firstfire · 06-12-2016, 09:23 AM

Hi.

I believe grail's solution is cleaner for this particular task.

Any online tutorial on regular expressions will do.
I use man grep, info grep, ditto for sed, awk and perl.
Just type

Code:

info grep regular expressions

into your terminal and read.

andrew.comly · 06-12-2016, 09:24 AM

Quote:

Originally Posted by grail

I went slightly simpler:

Code:

$ grep -oE '/[^:]+' <<< "$tmp"

Actually this yields:

Code:

$ grep -oE '/[^:]+' <<< "$tmp"
/home/a/.config/cairo-dock
/home/a/.config/cairo-dock
/home/a/.config/cairo-dock/current_theme and /media/a/REFIND/recent/AC/bckup/Install
...
/home/a/.config/cairo-dock/third-party and /media/a/REFIND/recent/AC/bckup/Install/3_UbuntuLXDE14.04/home/a/.config/cairo-dock/third-party

The "and" above between two paths isn't desirable, Of course this could be a typo, bash version incompatibility or maybe I misread it since the following command yields choice results (merely insert a space between ":" and "]") :

Code:

$ grep -oE '/[^: ]+' <<< "$tmp"

I found the ending <<< "$tmp" interesting and your statement's brevity gave this command a lighter feeling. THANKS!

grail · 06-12-2016, 10:40 AM

I had no space in my original due to none in the example