Find and replace pattern of fileA in fileC by fileB pattern

Question

I have two files, fileA with a list of name :

AAAAA 
BBBBB
CCCCC
DDDDD

and another fileB with another list :

and a third fileC with some text :

Hello AAAAA toto BBBBB dear "AAAAA" trird BBBBBB tuizf AAAAA dfdsf CCCCC

So I need to find and replace every pattern of fileA in fileC by fileB pattern. It works ! But i realised that fileC contains words like "AAAAA" and it isn't replace by "111".

I'm doing this but it doesn't seems to work.

#! /bin/bash
while IFS= read -r lineA && IFS= read -r lineB <&3; do
sed -i -e "s/$lineA/$lineB/g" fileC
done <fileA 3<fileB

I tested your solution and it works for me: `Hello 111 toto 222 dear 111 trird 222B tuizf 111 dfdsf 333` — svante, Oct 15 '13 at 08:41
I need to replace every AAAAA by 111 in fileC then do it for every BBBBB by 222 and so on... — Peter Dev, Oct 15 '13 at 08:42
It works ! But i realised that fileC contains words like "AAAAA" and it isn't replace by "111". — Peter Dev, Oct 15 '13 at 09:03
@PeterDev `AAAAA` in fileC isn't replaced because `fileA` contains `AAAAA ` and not `AAAAA` (notice the trailing space). — devnull, Oct 15 '13 at 09:09

score 3 · Accepted Answer · edited May 23 '17 at 12:15

3

This is a good job for GNU awk:

$ cat replace.awk 
FILENAME=="filea" {
    a[FNR]=$0
    next
}
FILENAME=="fileb" {
    b[a[FNR]]=$0
    next
}
{
    for (i=1;i<=NF;i++) {
        printf "%s%s",(b[$i]?b[$i]:$i),(i==NF?RS:FS)
    }
}

Demo:

$ awk -f replace.awk filea fileb filec
Hello 111 toto 222 dear 111 trird BBBBBB tuizf 111 dfdsf 333

A solution for sehe:

FILENAME==ARGV[1] {              # Read the first file passed in
    find[FNR]=$0                 # Create a hash of words to replace
    next                         # Get the next line in the current file
}
FILENAME==ARGV[2] {              # Read the second file passed in
    replace[find[FNR]]=$0        # Hash find words by the words to replace them 
    next                         # Get the next line in the current file
}
{                                # Read any other file passed in (i.e third)
    for (i=1;i<=NF;i++) {        # Loop over all field & do replacement if needed
        printf "%s%s",(replace[$i]?replace[$i]:$i),(i==NF?RS:FS)
    }
}

For replacements the ignore word boundaries:

$ cat replace.awk 
FILENAME==ARGV[1] {
    find[FNR]=$0
    next
}
FILENAME==ARGV[2] {
    replace[find[FNR]]=$0
    next
}
{
    for (word in find)
        gsub(find[word],replace[find[word]])
    print
}

Demo:

$ awk -f replace.awk filea fileb filec
Hello 111 toto 222 dear "111" trird 222B tuizf 111 dfdsf 333

edited May 23 '17 at 12:15

Community

1
1

answered Oct 15 '13 at 08:38

Chris Seymour

83,387
30
160
202

1

I'm continuously amazed how, after years of exposure, `awk` manages to make zero lasting impression on my brain. I mean, it always looks like _the tool_ for the job, but I really can make heads or tails of it (FNR? NF,RS,FS?) Also, while is `filea` and `fileb` still on the command line when they're also hardcoded in the script? Just - foreign to me. – sehe Oct 15 '13 at 08:41
It's pretty natural to me.. you probably already describe data in terms of records and fields so `FS` for `field separator`, `RS` for `record separator` and `NF` for `number of fields` is pretty sane. You could match files on positions and use `argv` but using the name is more readable *IMO* and of course you still need to pass in the handle of each file. – Chris Seymour Oct 15 '13 at 08:44
My script works ! But i realised that fileC contains words like "AAAAA" and it isn't replace by "111". Any idea ? – Peter Dev Oct 15 '13 at 09:09
@PeterDev I added script that does replacements regardless or word boundaries. – Chris Seymour Oct 15 '13 at 09:25
Thx and nice work but it doens't seems to work on the "" ~/test# awk -f replace.awk fileA fileB fileC 111 toto 222 dear "AAAAA" trird 222B tuizf 111 dfdsf 333 – Peter Dev Oct 15 '13 at 11:11
@PeterDev check that both `filea` and `fileb` do not containing trailing whitespace. – Chris Seymour Oct 15 '13 at 11:27

hipe · Answer 2 · 2013-10-15T09:16:05.683

2

sed 's/.*/s/' fileA | paste -d/ - fileA fileB | sed 's/$/\//' | sed -f - fileC

and the correct and faster version would be

paste -d/ fileA fileB | sed 's/^/s\//;s/$/\/g/' | sed -f - fileC

edited Oct 15 '13 at 09:16

answered Oct 15 '13 at 08:35

hipe

802
6
8

score 1 · Answer 3 · answered Oct 15 '13 at 08:37

1

A two-phase rocket:

sed -e "$(paste file[AB] | sed 's/\(.*\)\t\(.*\)/s\/\1\/\2\/g;/')" fileC

What this does is create an adhoc sed script using paste file[AB] | sed 's/\(.*\)\t\(.*\)/s\/\1\/\2\/g;/':

s/AAAAA/111/g;
s/BBBBB/222/g;
s/CCCCC/333/g;
s/DDDDD/444/g;

And then runs it with fileC as the input

answered Oct 15 '13 at 08:37

sehe

374,641
47
450
633

@hipe I didn't notice. Anyways, my version is limited as well (fileA/fileB cannot contain tabs) – sehe Oct 15 '13 at 08:51

Find and replace pattern of fileA in fileC by fileB pattern

3 Answers3