Shell Script Help

[Steven Chen_1]

Hope someone can help:

I have the following script:

sed -e 's/EMAIL_ADDR_TXT/AGENCY_LOCATION/'
-e 's/CONT_EMAIL_ADDR_TXT/AGENCY_CONTACT/'
-e 's/EMX_ADDR/WEB_APPLICANT/'
-e 's/SITE_EMAIL_ADDR_TXT/RESOURCE_LOCATION/' $BADMAIL>badmail1

but the result is all words of EMAIL_ADDR_TXT have been changed to AGENCY_LOCATION, the first sed choice.

I need some field protection and change the word to exactly what it means. But using [ ] to include words still not working. Nor I succeeds in using " " to protect.

Any comments are appreciated.

[Peter Godron · Accepted solution]

Steven,
is there a field seperator in the file?

If, for example it is "," you could do
sed -e's/,EMAIL_ADDR_TXT,/,AGENCY_LOCATION,/

[James R. Ferguson]

Hi Steven:

You need to provide a better regular expression -- one that uniquely identifies the boundries of a match/substitution.

For example, anchor EMAIL_ADDR_TXT to the beginning of a line with a caret:

sed -e 's/^EMAIL_ADDR_TXT/

You can anchor at the end of a line with a dollar sign ($). You can use spaces and tabs too to form logical boundries.

Far, far superior to 'sed' is 'perl'. Perl has some of the very best regular expression support you can find, in my opinion.

Regards!

...JRF...

[renarios]

Hi Steven,

I guess you need to use the global option (/g) option, just like in vi :%s///g

So the syntax would be:
sed -e 's/EMAIL_ADDR_TXT/AGENCY_LOCATION/g'\
-e 's/CONT_EMAIL_ADDR_TXT/AGENCY_CONTACT/g'\
-e 's/EMX_ADDR/WEB_APPLICANT/g'\
-e 's/SITE_EMAIL_ADDR_TXT/RESOURCE_LOCATION/g'\
$BADMAIL>badmail1

Cheers,

Renarios

[Arunvijai_4]

Hi,

A very good reading about "sed"

http://www.student.northpark.edu/pemente/sed/sed1line.txt

-Arun

[Peter Nikitka]

Hi,

as a first change, you could change the order of the sed-commands: since the command sequence of sed is in serial order, put the longest strings first: they have been changed already when sed executes the following command(s).
sed -e 's/CONT_EMAIL_ADDR_TXT/AGENCY_CONTACT/'
-e 's/SITE_EMAIL_ADDR_TXT/RESOURCE_LOCATION/'
-e 's/EMAIL_ADDR_TXT/AGENCY_LOCATION/' ...

If there is always at least one character of text before the search pattern, you can anchor it (like the other posters suggested). Pattern starting at the beginning of a line must be added seperately:

sed -e 's/\([^_]\)EMAIL_ADDR_TXT/\1AGENCY_LOCATION/' -e 's/^EMAIL_ADDR_TXT/AGENCY_LOCATION/'
..
and so on. If the overlapping search pattern differ not in the '_' unterline, the anchor pattern has to be refined, perhaps you need a different anchor pattern for the end of the search pattern.

mfG Peter

[Sandman!]

Hello!

I'm not sure if your question has been answered by one of the aforementioned posts, but it'll help if you could attach a portion of the input file on which these sed substitutions have to occur. These would definitely help in providing an accurate solution to your problem.

cheers!

[Arturo Galbiati]

Hi,
try:
sed -e 's/SITE_EMAIL_ADDR_TXT/RESOURCE_LOCATION/'
-e 's/CONT_EMAIL_ADDR_TXT/AGENCY_CONTACT/'
-e 's/EMX_ADDR/WEB_APPLICANT/'
-e 's/EMAIL_ADDR_TXT/AGENCY_LOCATION/' $BADMAIL>badmail1

Since Sed exceutes the commands in order, first chnage the long pattern and after the pattern contained into previous pattern already chnaged.
HTH,
Art

[Steven Chen_1]

I am going to attach original source file "badmail.lst" (=$BADMAIL) and the resulting file using my script. Maybe you could find out the solution.

Thanks a lot and have fun.

[Arunvijai_4]

Hello Steven,

What exactly you want to do ? Its always better to use perl for regular expressions.

http://www.english.uga.edu/humcomp/perl/regex2a.html

-Arun