41 comment

1. perl -p -i -e ‘BEGIN { print “First line\n” }’ originalfile

   should work without needing an explicit temp file

   If you’re going to use these commands in a script, though, “man mktemp” first.

   Sean

2. Even simpler would be:
   sed -i ‘1i Prepended line’ /tmp/newfile

   1. +1 This works very nice and without using temporally files.

      1. Works, but it _does_ use a temporary file behind the covers.
         The following is from the info document:
         >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
         –in-place[=SUFFIX]’
         This option specifies that files are to be edited in-place. GNU
         `sed’ does this by creating a temporary file and sending output to
         this file rather than to the standard output.(1).
         <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<

   2. That works on linux but not in solaris (unless you install gsed package), because de solaris sed command doesnt support “in place option” -i.

      In that case

      echo “text”|cat – yourfile > /tmp/out && mv /tmp/out yourfile

      works for me

      Thanks for help :)

      1. Yes it does. The `-i` option is a mandatory part of POSIX, so it will work on any UNIX that takes itself seriously, Solaris included.

         It just works differently, i.e. GNU sed allows you to omit the argument of the `-i` flag, as was done in the example, whereas most other seds require it.

         Try:

         sed -i.bak ‘1i Prepended line’ /tmp/newfile

         Now hang your head in shame.

         1. Ok, I was wrong on the POSIX part, I will hang my head in shame.

            ..but the missing flag argument is probably still the reason the example didn’t work on Solaris.

   3. Yes, that worked very nicely for me in Linux armel 2.6.32-rc5, sed is great but if you don’t use it regularly for a month it’s RTFM every time. Thanks.

   4. Excellent suggestion with `sed` very good.

3. You can also use tac (cat backwards) to make it work. It will print the file from end to beginning.

4. Thread’s old, I know, but I wanted to note that tac — at least in the version I have in Cygwin, is not fully “cat in reverse” alas.

   As you say, it’ll print from the end — but it won’t take an argument like this:

   tac >> somefilethatyouwanttoPREPENDtextto

   Just thought I’d note that for others who might be searching the same issue as the OP.
   This seems like such an obviously useful utility it’s unbelievable what a pita it is.

5. I tried the perl command (same one suggested to me by someone on IRC) and it didn’t work. It just printed to stdout and left the file unchanged. I’m kind of surprised two different people suggested the exact same non-working command, so I must be doing something wrong. What could I be doing wrong though??

   I like the sed command, but I needed to add more than one line (and my text contained special characters).

   This worked for me:

   sed -i ‘1{h; r headerfile.txt
   D;}
   2{x; G; }
   ‘ filetoedit.txt

   I’m not even really sure how it works, I just combined bits of code from a couple of random scripts I saw elsewhere. I should learn more sed :-P

   this also worked:

   for FILE in `find . -iname “*.php”`
   do
   cat header.php | cat – $FILE > /tmp/mytmpfile && mv /tmp/mytmpfile $FILE
   done

6. tac your_file prepended_text | tac > your_new_file

7. tac does not add a damn thing.
   Given this command:
   tac biglist.txt “thisistext” > biglist5.txt
   — I get the error
   tac: cannot open `thisistext’ for reading: No such file or directory.
   Considering it’s a close cousin of cat, I shouldn’t expect it to do any more, logically.
   Please do the newbie world a favour and remove any references to tac in this thread.

   BZT

   1. You need to put your prepend data into the file “thisistext”. Also, unless you want your file to come out upside down, you’ll have to pipe the output from the first tac into a second tac before redirecting to a file.

   2. With a little creative use of shell commands, you can avoid needing a second input file. Passing ‘-‘ as an argument tells most tools to use standard input or output (whichever applies) as a file.

      echo “my_prepend_text_goes_here” | tac biglist.txt – | tac > biglist5.txt

      There are a couple of tricks where you can avoid needing to create biglist5.txt in the process (storing results to a variable or using “tee” to write your output), but I can’t find anything that will work on large files.

      1. This is clever; even simpler, you can use
         echo “prepend this” | cat – file.txt > newfile.txt

8. How about a command or script to add the same text to every line OF a file, not just to the file itself???

   Believe it or not
   bash prepend same text to beginning of all lines of file
   does not have an exact match on Google. Ask.com and Yahoo! were sloppy with it (read:inexact and wide of the mark ultimately). Just more proof to my assertion that the world is going stupid.

   BZT

   1. This can be done with sed in place (a pretty scary operation):

      sed -i -e ‘s/^/PREFIX/’ file_with_lines_to_prefix.txt

      1. Thanks, Q. I’ve attempted something similar in the past without success.

9. This perl one-liner will prefix (not prepend, which isn’t a word) a line to the beginning of a file.

   perl -pi -e ‘print “Put before first line\n” if $. == 1’ inFile.txt

   (from the perl FAQ):

   1. You got gypped on your dictionary if it doesn’t have prepend.
      http://www.dict.org/bin/Dict?Form=Dict2&Database=*&Query=prepend

      It’s also in the Merriam-Webster Unabridged Dictionary, which is more authoritative at the cost of a subscription, and it’s even a command in the popular jQuery javascript library. However, I agree prefix would be more proper English if one cares and has an educated audience that would not be confused.

      1. You’ll have to copy and paste the url, as the link parser cannot handle it.

      2. ‘prepend’ is a corollary to ‘append’, which is an English word.

10. Ugh, my original message totally got garbled…

    What the above SHOULD look like:

    From http://sed.sourceforge.net/grabbag/tutorials/sedfaq.txt:

    sed '1{h; r file_to_prepend.txt
    D; }
    2{x; G; }' < file_to_be_prepended
    

    This can be used to insert a file into any line but the last one into a second file. For instance, to insert file1 into file2 at line 10:

    sed '10{h; r file1
    D; }
    11{x; G; }' <file2
    

    Edited by Admin – added formatting html tags

11. You could generate a patch file and use patch.

    diff -u <(head myfile) mypatch; patch mypatch

    1. That got completely mangled. Is there any way to protect my comments?

       1. My finally correct version (may I request a preview feature?):

          diff -u <(head myfile) <(echo -e "my data goes here and ends at the period.\n`head myfile`") > mypatch; patch mypatch

    2. Let’s try again.

       diff -u <(head myfile)  mypatch; patch mypatch

    3. diff -u <(head -n 50 votd) <(echo -e "my data goes here and ends at the period.\n`head -n 50 votd`") > mypatch; patch mypatch

12. Another method is using ed:

    echo -e “0a\nmy data goes here and ends at the period.\n.\n,wq” | ed myfile

    1. A variation on this is:

       echo “0r header.txt
       w” | ed myfile.txt

13. Thanks, works well…

14. This is the best (sed) one-liner that I found for this problem:

    sed -i -e ‘1i TEXT’ FILE

    Above command will insert string TEXT into the first line of the file FILE.

15. echo “MAINCARD_MSISDN,SUB_SOC,EXPIRY_DATE,S,U,” | cat – $spoolFile > /tmp/out && mv /tmp/out $spoolFile

    that line above works for. problem is if a have to run the script more than once. it will insert the same line again.. Any way to stop it?

    thanks

    1. if ! head -n 1 “$spoolFile” | fgrep “$myString” &>/dev/null; then
       echo “$myString” | cat – “$spoolFile” > /tmp/out && mv /tmp/out “$spoolFile”
       fi

16. Its worth mentioning that

    echo -e “DATA-Line-1\n$(cat input)” > input

    Is very ineficient and works only for small files, the larger files (e.g. 800Mb) crash the terminal and leave the file empty.

17. 
echo -e "DATA-Line-1\n$(cat input)" > input

    dose not always work

    bash-4.1$ cat fish
    VALUE=${VALUE//[$'\t\r\n']}
    
    bash-4.1$ echo -e "DATA-Line-1\n$(cat fish)" > fish
    bash-4.1$ cat fish
    DATA-Line-1
    VALUE=${VALUE//[$'	
    ']}
    

18. This is the only sed syntax that works for me on OSX Sierra:

    sed -i_bak '1s;^;prepended text ;' test.file

    Thanks, I’d been looking for an hour.

19. Thanks, this pointed me to just what I needed! I wanted to point out that your “Bash Only” way doesn’t need to be bash only – instead of using $(cat input), use `cat input` (use the backtick character). It does the same thing as $() in bash, but it works in most shells :D

20. Thank you
    “sed -i ‘1s;^;DATA-Line-1\n;’ input” worked great!
    Just what I needed

    Have a question? Post it on our forum!