Bash command to search for a string in file and replace the value with value from another file

Question

I have two files file a.) xmlFile.xml b.) emails.txt

xmlFile.xml has the following structure repeated multiple times

<gname>Office</gname>
<uname>person</uname>

emails.txt has list of email addresses

[email protected]
[email protected]
...

What I want to accomplish is to replace "person" in xmlFile.xml with subsequent value taken from emails.txt

I have tried

# while read email ; do sed  "s/person/$email/g" xmlFile.xml > xmlFile.new; done < emails.txt

However I endup with file that has all "person" values replaced with the last email from emails.txt

Thanks, Filip

You need to have sed only replace the first occurrence, then repeat once for each line. This will be slow, of course. The sane answer is, I believe, use Perl. :-P Alternatively, you could transform your emails.txt into a giant sed script and run it once. — derobert
– derobert, Commented Feb 25, 2011 at 19:35

SiegeX · Accepted Answer · 2011-02-25 20:36:16Z

awk 'NR==FNR{e[i++]=$0;next} /person/{sub("person",e[j++])}1' emails.txt xmlFile.xml

Explanation

NR==FNR: This is only true when awk is reading the first file. It essentially tests total number of records seen (NR) vs the input record in the current file (FNR).
e[i++]=$0: Create an array named e who's index increments by 1 (i++)and who's value is equal to the current record $0. This array will hold our emails
next: Ignore the rest of the script if this is reached, start over with a new input record
/person/: Only perform the subsequent code if the current record matches the regex "person"
sub("person",e[j++]): Substitute the literal value "person" for a value in our array e that we created earlier. Increment this array j++ for the next record we match
1: Always returns true, essentially a shortcut for {print $0}, or output our current record

Proof Of Concept

$ cat emails.txt
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]
[email protected]

$ cat xmlFile.xml
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>
<gname>Office</gname>
<uname>person</uname>

$ awk 'NR==FNR{e[i++]=$0;next} /person/{sub("person",e[j++])}1' emails.txt xmlFile.xml
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>
<gname>Office</gname>
<uname>[email protected]</uname>

The above script assumes that person is a literal value. If it is not, then..

Replace: /person/{sub("person",emails[j++])}
With: /<uname>/{sub(".*","<uname>"emails[j++]"</uname>")}

The solution seems perfect, but to be a good answer some explanation for awk novices should be given.

Dennis Williamson · Accepted Answer · 2011-02-25 20:59:12Z

1

One way to accomplish this would be to use in-place editing:

while read email ; do sed -i "s/person/$email/;q" xmlFile.xml; done < emails.txt

If there's little or nothing more to the XML file than what you've show, just reconstruct it:

sed -e 'i <gname>Office</gname>' -e 's|.*|<uname>&</uname>|' emails.txt > newxmlFile.xml

without even touching the existing xmlFile.xml.

However, you should probably use an XML parser such as xmlstarlet.

edited Feb 25, 2011 at 20:59

answered Feb 25, 2011 at 20:19

Dennis Williamson

364k95 gold badges386 silver badges446 bronze badges

1 Comment

Filip Over a year ago

Thanks Dennis, The first example resulted each "person" being changed to the first email in the emails.txt, however the second example did work. Thanks!

mitma · Accepted Answer · 2011-03-16 15:23:28Z

0

Here's how to do it using bash & xmlstarlet!

IFS=$'\n' read -r -d "" -a array < emails.txt                   # read file with email addresses into array
n=$(xmlstarlet sel -T -t -v "count(//uname)" -n xmlFile.xml)    # count "uname" nodes in XML file
xmlFileStr="$(< xmlFile.xml)"                                   # read XML file into variable


if [[ $n -eq ${#array[@]} ]]; then   # if the number of nodes & email addresses is equal ...
   for ((i=1; i <= ${n}; i+=1)); do
      xmlFileStr="$(printf '%s' "$xmlFileStr" | xmlstarlet ed -P -t -u "//uname[${i}]" -v "${array[$((i-1))]}")"
   done
fi

printf '%s\n' "$xmlFileStr" > xmlFile.xml
cat xmlFile.xml

answered Mar 16, 2011 at 15:23

mitma

1

Collectives™ on Stack Overflow

Bash command to search for a string in file and replace the value with value from another file

3 Answers 3

Explanation

Proof Of Concept

1 Comment

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Explanation

Proof Of Concept

1 Comment

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related