I have an Xml file in this format,

<?xml version="1.0"?>
<!DOCTYPE LCTL_TEXT SYSTEM "ltf.v1.2.dtd">
<LCTL_TEXT lang="BER" source_file="articles1.ltf.xml" source_type="web_news" author="LDC" encoding="UTF-8">
<DOC id="articles1" lang="BER">
<TEXT>
<SEG id="articles1revu-1" start_char="0" end_char="27">
<ORIGINAL_TEXT>I have a book</ORIGINAL_TEXT>
<TOKEN id="articles1revu-1-1" start_char="0" end_char="2">I</TOKEN>
<TOKEN id="articles1revu-1-2" start_char="4" end_char="10">have</TOKEN>
<TOKEN id="articles1revu-1-4" start_char="14" end_char="18">a</TOKEN>
<TOKEN id="articles1revu-1-5" start_char="20" end_char="20">book</TOKEN>

I want the result to be:

I
have
a
book

I have several files in this format, I want to go through them all and extract only tokens;
is there a way to do using grep?

Thank you

Have a look at this thread
http://www.linuxquestions.org/questi...estion-945390/

It's possible to do it in sed (you can adapt my solution) but as other LQ member stated there, you're better off using an XML parser.

Yes, I tried this shell script to extract first the line that contains TOKEN and then use the sed command, but it doesn't work:
I want to do them both at the same time

Okay, if you do want to go the sed way...

1. first try if the sed command does what you want. Don't write a bash script straight away.
2. The sed command is malformed ("sed's"?!)
3. You don't have to create a temporary file for permanent changes (see sed's -i flag)
4. 'grep TOKEN' is not necessary. Sed can do it for you. Besides there should be the pipe (|) symbol between them, not a comma (,)

eg.

Yeah I tried it this way, but still doesn't give the right result:
is there anything wrong?

Thanks

yes, grep needs a filename:
grep TOKEN this | .....

Do you really need to do this using tools like sed and grep? Python's ElementTree, for example, would be perfect for this.

I still couldn't get the correct result!

Thank you Nylex for your suggestion, but I am not familiar with Python :s

Thanks

you could try to fix it yourself:

grep TOKEN $this | sed 's/^<.*>\([^<].*\)<.*>$/\1/' > $this

sed -i cannot be used with pipes.

Thank you very much pan64. Indeed, I tried to fix it, but I still couldn't get the correct result; it just doe nothing to the file.

Thanks

No, no ... heed thou the sage advice to use an XML parsing package, and a real programming language to drive it. This is not a suitable application for "a bash script."

For example, look at this link ... http://search.cpan.org/~shlomif/XML-....98/LibXML.pod ... which happens to be a description of a Perl-language binding for the libxml2 library. My purpose here is to point out, if you follow the related links from this page, (a) what a sturdy library like libxml2 can do, and (b) how you do not need to write a C program to use it. If you follow links about "XPath expressions," and if you google for "XSLT stylesheets," you'll begin to get an idea of just what fully-tested options are at your disposal.

Don't "monkey around" trying to "cobble up" a "solution" that will wind up just wasting you a lot of time while producing a decidedly inferior, if not utterly useless bag of code. (And mind you, I am saying this bluntly but in the nicest and most respectful way possible.) There's a big fat deep river in that direction, and no bridge.

Thank you, but I believe there can be an easy way to d it using sed and grep.

Thanks

I'm not at a linux computer now (still at work) so can't test any sed solutions but perhaps you should listen to the majority of the posters in this thread:

Ok, may be you are right, do you know any easy XML parser that I can use?