Bash regex finding and replacing

I don't know if th开发者_运维百科is is possible, but can you dynamically alter a find/replace? Basically I have something like this

<3 digit number> <data>

and what I want to do is if data matches the pattern

<word>:<4 digit number>

replace all instances (in the entire file) of <word>: with the line's 3 digit number I.E:

020 Word
021 Word:0001
Replace with 
020 021
021 0210001

Is this doable with AWK or Sed? If not, is it doable in C?

---

I know this isn't what you asked, but I think the best way to solve this is with a simple Perl script.

#!/usr/bin/perl

$in= "input.txt";
$out= "output.txt";

# Buffer the whole file for replacing:
open(INFILE, $in);
@lines = <INFILE>;
open(INFILE, $in);

# Iterate through each line:
while(<INFILE>) {
  # If the line matches "word:number", replace all instances in the file
  if (/^(\d{3}) (\w+:)\d{4}$/) {
    $num = $1; word = $2;
    s/$word/$num/ foreach @lines;
  }
}

open(OUTFILE, $out);
print OUTFILE foreach @lines;

It looks a lot longer than it really needs to be, because I made it nice and easy-to-read for you.

---

I hope this time I got you right.

try the stuff below:

#file name:t
kent$  cat t
020 Word
021 Word:0001

#first we find out the replacement, 021 in this case:
kent$  v=$(grep -oP "(\d{3})(?= Word:\d{4})" t|head -n1)

#do replace by sed:
kent$  sed -r "s/Word[:]?/$v/g" t                                                                                                        
020 021 
021 0210001

---

number=$(gawk --posix '/[0-9]{3} '${word}':[0-9]{4}/ { print $1; exit }' $file)

if [ "$number" != "" ]; then
    sed -r "s/${word}:?/${number}/" $file
fi

---

This awk solution takes 2 passes through your file: once to find all the Words needing replacement, and once to actually do the replacing:

gawk '
    NR == FNR {
        if (match($2, /^([^:]+):[0-9][0-9][0-9][0-9]$/, a)) 
            repl[a[1] ":?"] = $1
        next
    }
    {
        for (word in repl)
            if ($2 ~ word) {
                sub(word, repl[word], $2)
                break
            }
        print
    }
' filename filename > new.file

Requires gawk for capturing parentheses.

---

Here's another sed solution:

# sweep the file and make a lookup table variable

lookup=$(sed -nr 's/(.*) (.*:).*/\2\1/p' <source_file |tr '\n' ' ')

# append the lookup to each line and substitute using a backreference
# N.B. remove the lookup whatever!

     sed -r "s/\$/@@${lookup}/;
             s/^(... )(.*)$@@.*\2:(\S*).*/\1\3/;
             s/^(... )(.*:)(.*)@@.*\2(\S*).*/\1\4\3/;
             s/@@.*//" <source_file