How do I remove every number that's surrounded by <>

This is really easy to do with sed, actually. You just get as many as you can in one go, then try, try again:

sed -e :t -e 's/\(<[^<]*\)[0-9]\{1,\}\([^>]*>\)/\1\2/g;tt' 

I tried it with the following random bits of input:

<Number1> 234234 </Nu994845mb6er8>' 234234 <000000000000000000000000000000000000>> <a1> 2 <34b5c> 6 7 def 

And the results were:

<Number> 234234 </Number> 234234 <>> <a> 2 <bc> 6 7 def 

The regex just matches at least one number between a < and a >. It continues to replace that number sequence with nothing at all until it can no longer successfully do so. This is the purpose of the test command.

Else you can do it without a loop like:

sed 's/^/>/;s/\(>[^<>]*\)*[0-9]*/\1/g;s/.//' <<\INPUT <Number1> 234234 </Nu994845mb6er8>' 234234 <000000000000000000000000000000000000>> <a1> 2 <34b5c> 6 7 def INPUT 

OUTPUT

<Number> 234234 </Number>' 234234 <>> <a> 2 <bc> 6 7 def 

It will always skip any > until it encounters a < - so it only affects <[^<>]*> groups. See this if you're interested in why.

You either need a loop around a substitution command (possible in both sed and perl), or a nested substitution command (perl only). I prefer the latter approach; it's a bit more general:

perl -pe 's/\<([^>]*)\>/do{$a = $1; $a =~ s,\d,,g; "\<" . $a . "\>"}/ge;' 

Example input:

<a1> 2 <34b5c> 6 7 def 

Output:

<a> 2 <bc> 6 7 def 

Explanation: The -p option says that we want to read the file line by line, execute the script for each line, and print the result (like in sed); -e means that the next argument is the script to be executed.

Essentially, the script is just a substitution command: We look for <, followed by any number of non->-characters, followed by >. The e modifier after the trailing / indicates a special feature of the substitution command: Its replacement part is not a string to be printed, but again a command sequence to be executed. In this command sequence, we first assign the string between < and > (i.e., $1) to a new variable $a, then execute another substitution command on $a that simply replaces every digit (\d) by nothing, and finally return <, followed by the modified string, followed by >. The g modifier (both after the trailing / and the trailing ,) means that the substitution commands should be executed for every matching string, not just for the first one.

If the opening < and the corresponding > can be in different lines, say,

<abc1 opt="def"> 

add the option -0777 (i.e., perl -0777 -pe '...'), so that perl reads the entire file before processing it instead of working line-by-line (slurp mode).

short sed way

sed 's/<\([^>]\+\)[0-9]\+>/<\1>/g' file