How to replace an entire line in a text file by line number

Question

I have a situation where I want a bash script to replace an entire line in a file. The line number is always the same, so that can be a hard-coded variable.

I'm not trying to replace some sub-string in that line, I just want to replace that line entirely with a new line.

Are there any bash methods for doing this (or something simple that can be thrown into a .sh script).

Solution 1

Not the greatest, but this should work:

sed -i 'Ns/.*/replacement-line/' file.txt

where N should be replaced by your target line number. This replaces the line in the original file. To save the changed text in a different file, drop the -i option:

sed 'Ns/.*/replacement-line/' file.txt > new_file.txt

Solution 2

I actually used this script to replace a line of code in the cron file on our company's UNIX servers awhile back. We executed it as normal shell script and had no problems:

#Create temporary file with new line in place
cat /dir/file | sed -e "s/the_original_line/the_new_line/" > /dir/temp_file
#Copy the new file over the original file
mv /dir/temp_file /dir/file

This doesn't go by line number, but you can easily switch to a line number based system by putting the line number before the s/ and placing a wildcard in place of the_original_line.

Solution 3

Let's suppose you want to replace line 4 with the text "different". You can use AWK like so:

awk '{ if (NR == 4) print "different"; else print $0}' input_file.txt > output_file.txt

AWK considers the input to be "records" divided into "fields". By default, one line is one record. NR is the number of records seen. $0 represents the current complete record (while $1 is the first field from the record and so on; by default the fields are words from the line).

So, if the current line number is 4, print the string "different" but otherwise print the line unchanged.

In AWK, program code enclosed in { } runs once on each input record.

You need to quote the AWK program in single-quotes to keep the shell from trying to interpret things like the $0.

EDIT: A shorter and more elegant AWK program from @chepner in the comments below:

awk 'NR==4 {$0="different"} { print }' input_file.txt

Only for record (i.e. line) number 4, replace the whole record with the string "different". Then for every input record, print the record.

Clearly my AWK skills are rusty! Thank you, @chepner.

EDIT: and see also an even shorter version from @Dennis Williamson:

awk 'NR==4 {$0="different"} 1' input_file.txt

How this works is explained in the comments: the 1 always evaluates true, so the associated code block always runs. But there is no associated code block, which means AWK does its default action of just printing the whole line. AWK is designed to allow terse programs like this.

Solution 4

Given this test file (test.txt)

Lorem ipsum dolor sit amet,
consectetur adipiscing elit. 
Duis eu diam non tortor laoreet 
bibendum vitae et tellus.

the following command will replace the first line to "newline text"

$ sed '1 c\
> newline text' test.txt

Result:

newline text
consectetur adipiscing elit. 
Duis eu diam non tortor laoreet 
bibendum vitae et tellus.

more information can be found here

http://www.thegeekstuff.com/2009/11/unix-sed-tutorial-append-insert-replace-and-count-file-lines/

Solution 5

in bash, replace N,M by the line numbers and xxx yyy by what you want

i=1
while read line;do
  if((i==N));then
    echo 'xxx'
  elif((i==M));then
    echo 'yyy'
  else
    echo "$line"
  fi
  ((i++))
done  < orig-file > new-file

EDIT

In fact in this solution there are some problems, with characters "\0" "\t" and "\"

"\t", can be solve by putting IFS= before read: "\", at end of line with -r

IFS= read -r line

but for "\0", the variable is truncated, there is no a solution in pure bash : Assign string containing null-character (\0) to a variable in Bash But in normal text file there is no nul character \0

perl would be a better choice

perl -ne 'if($.==N){print"xxx\n"}elsif($.==M){print"yyy\n"}else{print}' < orig-file > new-file

View more solutions

Comments

[steveha]

I didn't even think of trying for a pure BASH solution!

[chepner]

Useless use of cat: sed -e "s/.../.../" /dir/file > /dir/temp_file

[chepner]

A little shorter: awk 'NR==4 {$0="different"} { print }' input_file.txt.

[SourceSeeker]

@chepner: A little shorter: awk 'NR==4 {$0="different"}1' input_file.txt

[user788171]

Is there a way to make sed modify the file in question instead of dumping to screen?

[user788171]

I am now hit with sed: illegal option -- i is there a fix for this?

[steveha]

@DennisWilliamson, I don't even know how that works! What does that trailing 1 do? (Note: I tested it and it does indeed work! I just don't understand how.)

[chepner]

I figured there'd be a way to abbreviate the unconditional print! @stevaha: the 1 is just a true value, meaning a "pattern" that always matches. And the default action for a match is to print the current line.

[Clayton Stanley]

sudo port install sed? are you on a mac? if so, just use the *nix version

[Clayton Stanley]

Useless for the interpreter/compiler maybe, but not for a human reading it. @Kyle's code reads nicely left to right; the idiom you used IMO does not, due to the fact that the verb is before the noun.

[Danijel]

For me it says: sed: -e expression #1, char 26: unknown option to ``s' and my line is: sed -i '7s/.*/<param-value>http://localhost:8080/ASDF/services/REWS‌​.REWSHttpSoap12Endpo‌​int/</param-value>/' $TCE_SVN_HOME\trunk\tce\EWC\WebContent\WEB-INF\web.xml. Any idea?

[chepner]

Each / in the replacement text would be interpreted as the closing slash of the s command unless escaped (\/). The easiest thing to do, though, is to pick a different, unused character as the delimiter. For example, sed -i '7s{.*}{<param-value>http://...}' $TCE_SVN_HOME/trunk....

[adelphus]

This should be the correct answer. The c flag does exactly what is required (replaces a numbered line with given text).

[func0der]

That explanation is a little confusing and does not seem to work. If you guys have problems with sed and its delimiter setting capabilities you may want to have a look at this: grymoire.com/Unix/Sed.html#uh-2 It says the first character behind the s determines the delimiter. So to change your command to use for example the # it would be like this: sed -i '7s#.*#<param-value>http://localhost:8080/ASDF/services/REWS‌​.REWSHttpSoap12Endpo‌​‌​int/</param-value>‌​#'

[sikisis]

but it my bash why the result only shows ${newline}?

[Leon S.]

If the -i command doesn't work on your machine, check the man-page. For me the command required a backup file extension after the flag, which you can leave empty: -i '', and in some other version of sed the syntax is --in-place=''.

[ELLIOTTCABLE]

To expand on @meonlol's, macOS requires an -e if -i is given, as well; thus, the correct command becomes: sed -e 'Ns/.*/replacement-line/' -i '' file.txt

[Gabriel Devillers]

Alternative syntax with sed that does not echo the file on stdout: stackoverflow.com/a/13438118/4669135

[Rob]

why do you need a loop for this example, just obfuscates the solution? line=5; sed -i "${line}s/.*/replaced by '$i'!/" output.dat

[Pubudu Dodangoda]

Is there a way to merge those two lines, for example by something like, sed -e "s/.../.../" /dir/file > /dir/file

[Victoria Stuart]

sed streams the entire file, but as noted in this answer, specifying the line number (if known) helps: in my case, a ~2-fold increase in execution speed (GNU sed 4.5). You can grep -n or ripgrep (rg) to find line numbers, based on pattern searches.

[gsamaras]

Use double quotes if the replacement-line is not fixed, e.g. it's a variable, like this: sed -i "Ns/.*/foo=$myVar" file.txt.

[chepner]

@gsamaras You have to be careful with myVar in that case, though. Its value needs to be properly escaped, e.g. if myVar=3/4, you'll get an error from sed; the value would have to be myVar='3\/4' instead.

[stackunderflow]

It's the right answer, but why the ugly line break? sed '1 cnewline text' test.txt would do it as well.

[Chandu]

Your don't need a loop for this

[SeanFromIT]

@PubuduDodangoda sed -i would accomplish that

[li ki]

For anyone who wants to know why -i '': BSD sed: extra characters at the end of d command - Stack Overflow

[a.t.]

This was, for me, the easiest answer to convert into a function with a variable line number, variable new line, and a variable filepath. And for my use case, I did not need to escape any characters even though my new line contained /,-,= and ;. Thank you! Ps. after having posted this comment, I scrolled down Daniel Bigham's answer.

[Nahuel Fouilleul]

This answer is quite old. Now, I would use the perl command at the end, possibly passing the variables by environment.

[Admin]

As it’s currently written, your answer is unclear. Please edit to add additional details that will help others understand how this addresses the question asked. You can find more information on how to write good answers in the help center.

[sapanoia]

The question is specifically about replacing a line with a given number. grep is not needed here.

[justin.m.chase]

This is the best answer by far, lol

[Erwann]

What about the last line? I guess N=$(wc -l file | awk '{print $1;}', but cumbersome.

[chepner]

sed uses $ as the address of the last line: $s/.../.../.