Shell脚本，使用sed替换字符

Question

In shell scripting, i want to replace 1bc1034gf45dna22 (16 characters total) with 1b:c1:03:4g:f4:5d:na:22 (seperated by colon) using sed.

Edit

I have tried

sed 's/\w{2}/\w{2}:/g' a.txt > b.txt

where a.txt has

1bc1034gf45dna22

Answer 1

Too Many Edge Cases

Sed is not an ideal tool for this job, because there are just too many edge cases. If you have very regular data and a known corpus then that's fine, but in some cases you need something more powerful that can act on pieces of a line independently after matching/capturing the text of interest. For example, given an input file containing:

foo 1bc1034gf45dna22 bar

you might need to extract the second column, transform it, and then substitute it in place. You might do this with Ruby as follows:

$ echo 'foo 1bc1034gf45dna22 bar' |
ruby -pe '
  if /(?<str>\p{Alnum}{16})/ =~ $_
    $_.sub!(str, str.scan(/../).join(?:))
  end'

This correctly yields:

foo 1b:c1:03:4g:f4:5d:na:22 bar

Answer 2

这可能对您有用（GNU sed）：

sed 's/\w\w\B/&:/g' file

Answer 3

Does

sed -e 's/../&:/g' -e 's/:$//' a.txt > b.txt

or

sed -e 's/\(..\)/\1:/g' -e 's/:$//' a.txt > b.txt

work for you?

Answer 4

为什么不sed 's/1bc1034gf45dna22/1b:c1:03:4g:f4:5d:na:2/g' file ？

Answer 5

With GNU/sed on any GNU/Linux distro you can use extentded regular expressions. This will convert a string of any length to a string where each two letters are separated by a colon:

echo 1bc1034gf45dna22 | sed -re 's/(\w{2})/\1:/g' -e 's/:$//'

On your AIX platform, you can leverage awk:

echo 1bc1034gf45dna22 | awk '
  BEGIN { FS=""; ORS=""; } 
  {
    for (i=1; i <=NF; i++) {
      if (i > 1 && (i + 1) % 2 == 0) { 
        print(":"); 
      }  

      print($i);
    }
  }'

Answer 6

You could:

sed -i 's/\(\w\{2\}\)/\1:/g' file

This would also append a : to the end of the string. Since we cannot use positive lookahead in our sed regex syntax, we would then need to:

sed -i 's/:$//g' file

Notice that we used inplace editing for sed therefore the file will in this way altered.

Alternatively, you could use perl to use a regex that would have been /(\\w{2})(?=\\w{2})/$1:/g

Answer 7

Well, there's always the brute-force method:

sed 's/\<'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\([0-9a-z][0-9a-z]\)'\
'\>/\1:\2:\3:\4:\5:\6:\7:\8/g'

With no leading spaces on the continuation lines the shell pastes the strings together properly.