Busybox filter string list

Question

Consider I have a list like this:

list='item1.service item2.service item3.target item4.service'

I need to filter this list to get something as:

item1 item2 item4

So, there are two things to notice:

1. I need keep only the .service items.
2. I need only the "base" names, without the .service suffix.

And one more important information: I am running on busybox, where tools are often crippled (e.g.: my grep has no support for Perl regexes).

I have been struggling against combinations of sed and grep and the best I could get is:

$ echo $list | grep -io '[a-z0-9\-\_@]*.service' | sed 's/\.service//'
item1
item2
item4

but it needs to perform essentially the same match twice for each input, what doesn't look very efficient. Could anyone suggest any better solution, please?

Thanks in advance.

Answer 1

list='item1.service item2.service item3.target item4.service'
echo "$list" | sed -r '

# Since the input is only one line, all commands will scan all the pattern space,
# so the commands order matters.

# replace for nothing unwanted text
s/[a-z0-9@_-]+\.target//

# replace for nothing unwanted suffix
# (with 'g' flag the command will replace all occurrences)
s/\.service//g

# squeeze double spaces
s/ +/ /g

# replace space for new line character
s/ /\n/g'

I think this will work in all versions of BusyBox's sed (could be awk too if you have it in your BusyBox).

I just downloaded the latest BusyBox version and ran make menuconfig but couldn't find any reference to Perl regex.