Search and Replace by record position

Shell Programming and Scripting
1

Hi All,
I have a file that I would like to search for data and replace other data by record position number:

Example search..
search for "CLARK KENT" and replace Amt data "000025" with "000155"???

I'm able to search and replace unique data but, came to a stump when wanting to replace data that can be similar throughout other records.
Example code of search routine..

$1="CLARK KENT"
$2="MJ WOO"

awk -v s="$1" -v r="$2" 'function convert(str) {return (length(str)>11)?substr(str,1,11):sprintf("%-11s",str)}
        {if ($0~s) sub(convert(s),convert(r))}1' FILENAME
2

Code tags preserve whitespace, quotes do not.

Try something like:

sed '/CLARK KENT/s/^\(.\{50\}\).\{6\}/\1000155/' infile
3

Hmmm? So I tested the sed command and didn't get any results. Am I doing something wrong??

4

You have to use double quotes on the sed command - the shell won't expand any variables in single quotes.

e.g on bash:

$ cat x.sh
#!/bin/bash

a="CLARK KENT"
b="000099"

sed "/${a}/s/^\(.\{50\}\).\{6\}/\1${b}/" infile
$ ./x.sh
1112223333BRAD PIT   PO BOX 111 PHOENIX    AZ10567000025103111
9994442222CLARK KENT PO BOX 222 LAS VEGAS  NV27407000099102511
7778884444STEVE JOBS PO BOX 444 SUN CITY   AZ10473000015100111
5

Awesome! Thanks CarloM.
Small problem though.... How would I pad data I replaced with spaces. Lest say I would like to search for name "CLARK KENT" and replace the city "LAS VAGAS" with "DEMING". When I do the sed command you showed me it shifts all the data to the left.

6

You can use printf:

#!/bin/bash

a="CLARK KENT"
b="DEMING"
b1="$(printf %-11s $b)"

sed "/${a}/s/^\(.\{32\}\).\{11\}/\1${b1}/" infile

Also, note that you need to change the 2 numbers in the regular expression to match the size of the preceding text and of the text you're replacing.

7

Thanks again CarloM, but now lets say I change "PHOENIX" to "LAS VEGAS" it separates "LAS" as a 20 characters and "VEGAS" as the next 20 characters. Sorry to keep making this harder than what it probably is... :frowning:

a="BRAD PIT"
b="LAS VEGAS"
c="`(printf %-20s $b)`"

sed "/${a}/s/^\(.\{32\}\).\{11\}/\1${c}/" infile
8

Should have been:

b1=$(printf %-11s "$b")

(without quotes, printf will treat $b's expansion to LAS VEGAS as 2 parameters and just re-uses the format arg for both)

9

Thank you so much CarloM! I was wondering though what is it in the printf command that pads the data with space? I was wondering if say I wanted to pad with zero's instead of spaces. :slight_smile:

10

You can pad numbers with a format specifier like %03d , but iirc it doesn't work with strings - check the man page(s).

1 Like
11

Ok thanks again CarloM for all your help. :smiley: Have a good day.

12

To make it easier and safer, maybe you sould consider formating your file with a separator at first (here an example with colon ":" as separator)

sed 's/PO BOX [0-9][0-9]*/:&:/;s/[^ ][^ ]*$/:&/;s/[ ]*:[ ]*/:/g' yourfile

All further processing can then easly be handled like changing the wanted field

nawk -F: -vNAME="BRAD PIT" -vCITY="LAS VEGAS" '$0~NAME{$3=CITY}1'
1 Like