What I want to do is in the title.

Input:

apple -berry --sun house-roof computer-- orange number-43  

Desired output:

apple house-roof computer-- orange number-43

In other words, only when a word starts with - or -- it should be removed. The closest I've got is:

echo "apple -berry --sun house-roof computer-- orange number-43" | sed -r 's/(-|--)\w //g'
apple   house computer-- orange number

Which is far from what I want. Could be solved with other tools? Cut?

CodePudding user response：

You can use

sed -E 's/(^|[[:space:]] )-{1,2}[^[:space:]] //g'

Details:

• -E enables the POSIX ERE regex syntax
• (^|[[:space:]] )-{1,2}[^[:space:]] - matches
  • (^|[[:space:]] ) - start of string or one or more whitespaces
  • -{1,2} - one or two hyphens
  • [^[:space:]] - one or more non-whitespace chars.

See the online demo:

#!/bin/bash
s='apple -berry --sun house-roof computer-- orange number-43'
sed -E 's/(^|[[:space:]] )-{1,2}[^[:space:]] //g' <<< "$s"

Output:

apple house-roof computer-- orange number-43

CodePudding user response：

You can also do it by converting each token to a line and then greping with a negated pattern.

For example:

xargs -n1 <input.txt | grep -v -E -e '^-[-]?[[:alnum:]]' | xargs

CodePudding user response：

Using sed

$ sed s'/^-[^ ]*\| --\?[^ ]*//g' input_file
apple house-roof computer-- orange number-43

CodePudding user response：

plain bash:

input="apple -berry --sun house-roof computer-- orange number-43"
read -ra words <<< "$input"
ok_words=()
for word in "${words[@]}"; do
    [[ $word == @(-|--)* ]] || ok_words =("$word")
done
echo "${ok_words[*]}"

apple house-roof computer-- orange number-43

@(-|--) is an extended pattern matching either - or --