[Solved] Search & Replace exact string with sed

pedropt · 06-25-2017, 12:50 PM

Lets say that you have a file called sample.txt
and inside that file are these texts :

john
paul
#will
smith

Now the normal procedure to replace an exact string with sed is :

Quote:

sed 's/\<find\>/replace/g' < oldfile > newfile

Now if i want to change in that text above the #will to will with sed , i am unable to do it the normal way because the # symbol .

code that do not work :

Quote:

sed 's/\<#will\>/will/g' < sample.txt > sample.tmp

Anyone knows how to bypass this issue ?

TB0ne · 06-25-2017, 01:01 PM

Quote:

Originally Posted by pedropt

Lets say that you have a file called sample.txt
and inside that file are these texts :

john
paul
#will
smith

Now the normal procedure to replace an exact string with sed is :

Code:

sed 's/\<find\>/replace/g' < oldfile > newfile

Now if i want to change in that text above the #will to will with sed , i am unable to do it the normal way because the # symbol . code that do not work :

Code:

sed 's/\<#will\>/will/g' < sample.txt > sample.tmp

Anyone knows how to bypass this issue ?

You do, since you posted it already. Put a "" in front of the #, just like you did for the <>'s.

scasey · 06-25-2017, 01:02 PM

Quote:

Originally Posted by pedropt

Now if i want to change in that text above the #will to will with sed , i am unable to do it the normal way because the # symbol .

code that do not work :

Anyone knows how to bypass this issue ?

Escape the # ?

Code:

sed 's/\<\#will\>/will/g' < sample.txt > sample.tmp

Laserbeak · 06-25-2017, 01:08 PM

I'm not that familiar with using sed, but can't you just use a backslash to escape the '#' sign?

It works in Perl:

Code:

#!/usr/bin/perl

foreach (<>) {

  s/\#//;
  print;
}

Usage:

Code:

./sometext.pl < sometext.txt
john
paul
will
smith

danielbmartin · 06-25-2017, 01:12 PM

With this InFile ...

Code:

john
paul
#will
smith

... this sed ...

Code:

sed 's/#will/will/' $InFile >$OutFile

... produced this OutFile ...

Code:

john
paul
will
smith

Daniel B. Martin

Laserbeak · 06-25-2017, 01:13 PM

sed should work too:

Code:

echo '#will' | sed 's/\#//'
will

pedropt · 06-25-2017, 01:36 PM

Quote:

With this InFile ...
Code:

john
paul
#will
smith

... this sed ...
Code:

sed 's/#will/will/' $InFile >$OutFile

... produced this OutFile ...
Code:

john
paul
will
smith

eventually it will work if you only got #will in the text .
but if you get in that list also #williams and you only want to change #will then it will not work .
The thing is that "exact match" must be used in this case.

Quote:

sed 's/\<\#will\>/will/g' < sample.txt > sample.tmp

this one does not change the file

TB0ne · 06-25-2017, 01:38 PM

Quote:

Originally Posted by pedropt

eventually it will work if you only got #will in the text .
but if you get in that list also #williams and you only want to change #will then it will not work .

this one does not change the file

Why not, exactly??? Changing the "will" in william to "will" in william is a zero-sum change. You're replacing the exact string...unless you tell it to replace "william" with "will", it won't look one bit different, but just remove the #. Based on what you posted, you've got your solution.

If you don't...then why don't you post an actual piece of your input file, along with detailed requirements and what you've written so far. We can't guess. If the big issue here is the # sign, just remove it....

Code:

sed 's/\#//g'

Done.

pedropt · 06-25-2017, 01:41 PM

Quote:

You do, since you posted it already. Put a "" in front of the #, just like you did for the <>'s.

Post an example of the code .

astrogeek · 06-25-2017, 01:50 PM

Quote:

Originally Posted by pedropt

Code:

sed 's/\<\#will\>/will/g' < sample.txt > sample.tmp

this one does not change the file

No it doesn't, because there is no string matching '<#will>' in the file.

Why are you wrapping '#will' with '<...>'? You seem to think this has some special universal significance other than simply matching those characters - it does not (maybe GNU extension?). It does not mean "exact match" in all regular expressions as you seem to imply.

To match only '#will' and not '#william', use regular expression syntax, this:

Code:

sed 's/^#will$/will/' < sample.txt > sample.tmp

BW-userx · 06-25-2017, 02:09 PM

Code:

userx%slackwhere ⚡ ~ ⚡> echo "#will" | sed 's/\#//'
will
userx%slackwhere ⚡ ~ ⚡> echo "#will" | sed 's/\#will/will/'
will
userx%slackwhere ⚡ ~ ⚡>

seems these two ways work then if you add the $ to keep it just 'will' well then there ya go.

Code:

userx%slackwhere ⚡ ~ ⚡> echo "#william" | sed 's/\#will$/will/'
#william
userx%slackwhere ⚡ ~ ⚡>

Laserbeak · 06-25-2017, 02:18 PM

Oh, I didn't understand exactly what you meant.

Then, in Perl, you'd do this:

Code:

#!/usr/bin/perl

foreach (<>) {

  s/^\#(will)$/$1/;
  print;
}

so using this file:

Code:

john
paul
#will
smith
#william
will#iam
#otherwills
#wil

You'd get this output:

Code:

 ./sometext.pl < sometext.txt
john
paul
will
smith
#william
will#iam
#otherwills
#wil

Is that not what you want?

NevemTeve · 06-25-2017, 02:22 PM

An universal solution would escape every characters of 'From' and 'To' This could be achieved with sed:

Code:

FromE=$(printf '%s' "$From" | sed 's/./\\&/g')
ToE=$(printf '%s' "$To" | sed 's/./\\&/g')
sed "s/$FromE/ToE/"

Laserbeak · 06-25-2017, 02:27 PM

So to make it short, you could make it a one-liner:

Code:

root@sunshine:~# perl -e  'for (<>) {s/^\#(will)$/$1/; print}' < sometext.txt
john
paul
will
smith
#william
will#iam
#otherwills
#wil
root@sunshine:~#

pedropt · 06-25-2017, 02:32 PM

Quote:

sed 's/^\#will$/will/' < sample.txt > sample.tmp

Nice one Astrogeek , this one works perfectly .
thanks for the help on this one .

Thanks to all other alternatives , i forgot to mention , but i wanted this code using only bash language .