carriage return or funky character stuck in my variables - help :)

My variables contain carriage returns... or something.

Here is the line in my xml file of interest:

<tmp>72</tmp>

And here is the line that extracts the 72 from it:

REAL=`grep '<tmp>' test.xml | sed -e 's/^ *<tmp>//' -e 's/<\/tmp>//' -re 's/(N|n|A|a|\/)/U/g'`

If echo the variable it looks just fine, but if I string a few together, the output is all messed up suggesting to me that there is an extra character in there such as a carriage return or something.

REAL=`grep '<tmp>' test.xml | sed -e 's/^ *<tmp>//' -e 's/<\/tmp>//' -re 's/(N|n|A|a|\/)/U/g'`
echo ${REAL},${REAL},${REAL}

When I run that, it returns simply:

,72

If I try to pipe it into a comma separated file, it's equally messed-up.

REAL=`grep '<tmp>' test.xml | sed -e 's/^ *<tmp>//' -e 's/<\/tmp>//' -re 's/(N|n|A|a|\/)/U/g'`
echo ${REAL},${REAL},${REAL} > test.csv

Here are the contents of test.csv:

72
,72
,72

What am I doing wrong?

Seems to work for me on Linux. I get errors on OS X and OpenBSD though (doesn't understand the -re option).

What are you doing your work on?

Carl

Running Debian LINUX.... the output shouldn't contain any linefeeds/returns though...

In the example above, it should return a file with a single line like this:

72,72,72

Did it do that for you?

GOT IT!

REAL=`grep '<tmp>' $1 | sed -e 's/^ *<tmp>//' -e 's/<\/tmp>//' -e 's/.$//g' -re 's/(N|n|A|a|\/)/U/g'`

...you gotta love google. The solution was contained in this thread.

Just FYI, it did do that for me (I see you have an answer but thought I'd answer you ).

I had 15 or so of the lines so I had 72 72 72 72 72, 72 72 72 72 72, 72 72 72 72 72 but all in a line.

Carl

The grep is quite redundant.

sed -e '/<tmp>/!d' -e 's/^ *<tmp>//' -e 's/<\/tmp>//' -e 's/.$//g' -re 's/(N|n|A|a|\/)/U/g' "$1"