[SOLVED] Replacing a string of text that includes "/"

Mike_V · 09-29-2010, 11:11 AM

Suppose I have a file named temp.txt

cat temp.txt
word/one/two/three

If I would like to replace "one" with "six" I would do this:

cat temp.txt | sed 's/one/six/'
word/six/two/three

Now I want to replace "one/two" with "six/seven" and these options don't work:
cat temp.txt | sed 's/one/two/six/seven/' (OFCOURSE)
cat temp.txt | sed 's/"one/two"/"six/seven"/'

I even want to be able to replace "one/two" with "six/seven/eight"

Is there a simple way do do this?

Thanks!

unSpawn · 09-29-2010, 11:23 AM

'sed -i "s|one/two|six/seven|g" temp.txt'?

Mike_V · 09-29-2010, 11:41 AM

unSpawn: thanks a lot, yes, that did the trick!

quick question: why did you add the "g" at the end? In my case it doesn't seem to matter if I include it or not... I'm just curious....

GrapefruiTgirl · 09-29-2010, 11:51 AM

Since you're working with regular expressions where your search query contains a slash, unSpawn changed to using a pipe symbol (|) as the delimiter character in sed, instead of using slashes. This makes the regex simpler, because you don't have to escape the slashes in your search/replacement text.

The 'g' makes sed continue processing the input even after it has made a replacement (to see if there is more to replace), rather than quitting after one replacement. Try doing multiple replacements on a single line, and see the difference with and without the 'g' on the end:

Code:

root@reactor: echo "one one one" | sed 's/one/two/'
two one one
root@reactor: echo "one one one" | sed 's/one/two/g'
two two two
root@reactor:

unSpawn · 09-29-2010, 11:52 AM

I usually want to delete all instances of something which is what "g" does (as far as I know sed). Sure you don't need to use it.

Mike_V · 09-29-2010, 11:57 AM

@GrapefruiTgirl: thanks for explaining that!

kurumi · 09-29-2010, 07:21 PM

Code:

$ ruby -pne 'gsub("one/two" ,"six/seven")' file
word/six/seven/three