GNU Pattern Match and replace exact number of characters

Question

This question may have been listed but I was not able to find one exact hit.

I am trying to skim through a file, match a pattern and replace it with something else. However there are other occurrences of the pattern but I need to replace only those which are 17 characters in length.

Example:

Content:

dlkfhfd|fedfe|dfwe3f347fde|3745978|dlkfhr**|376663781736102|**fedfe|dfwe3f347fde

Expectation:

dlkfhfd|fedfe|dfwe3f347fde|3745978|dlkfhr**|37xxxxxxxxxxxxx|**fedfe|dfwe3f347fde

Progress: I was able to match the expression with regexp pattern : **\|37[0-9]{13}\|**

However if I put it in an sed, it just replaces everything in the file.

sed -e s/\|37[0-9]{13}\|/\|37xxxxxxxxxxxxx\|/g

My sed version is 4.2.2

Answer 1

Your regular expression is a mix between basic and extended regular expression.

As an extended regular expression (using {13} and \| as a literal pipe):

sed -E 's/\|37[0-9]{13}\|/|37xxxxxxxxxxxxx|/g'

Alternatively, as a basic regular expression (using \{13\} and | as a literal pipe):

sed 's/|37[0-9]\{13\}|/|37xxxxxxxxxxxxx|/g'

This turns your example string into

dlkfhfd|fedfe|dfwe3f347fde|3745978|dlkfhr**|37xxxxxxxxxxxxx|**fedfe|dfwe3f347fde

Also note that there is no need to escape the | in the replacement part of the expression as that part is never interpreted as a regular expression.

---

In awk:

awk -F '|' -vOFS='|' '
    {
        for (i=1; i<=NF; ++i))
            if (length($i)==15 && match($i,"^37[0-9]"))
                $i="37xxxxxxxxxxxxx"
        print 
     }'

One could have used gsub() here, but that would have made it more or less identical to the sed solution, and therefore boring.

This has the benefit that the substitution would also occur in the first or last field even if that field was not delimited by | at both ends.