Split a string to print first two characters delimited by “-” In Bash

Question

I am listing the AWS region names.

us-east-1
ap-southeast-1

I want to split the string to print specific first characters delimited by - ie 'two characters'-'one character'-'one character'. So us-east-1 should be printed as use1 and ap-southeast-1 should be printed as aps1

I have tried this and it's giving me expected results. I was thinking if there is a shorter way to achieve this.

region=us-east-1 
regionlen=$(echo -n $region | wc -m) 
echo $region | sed 's/-//' | cut -c 1-3,expr $regionlen - 2-expr $regionlen - 1 

Answer 1

How about using sed :

echo "$region" | sed -E 's/^(.[^-]?)[^-]*-(.)[^-]*-(.).*$/\1\2\3/'

Explanation: the s/pattern/replacement/ command picks out the relevant parts of the region name, replacing the entire name with just the relevant bits. The pattern is:

^         - the beginning of the string
(.[^-]?)  - the first character, and another (if it's not a dash)
[^-]*     - any more things up to a dash
-         - a dash (the first one)
(.)       - The first character of the second word
[^-]*-    - the rest of the second word, then the dash
(.)       - The first character of the third word
.*$       - Anything remaining through the end

The bits in parentheses get captured, so \\1\\2\\3 pulls them out and replaces the whole thing with just those.

Answer 2

IFS influencing field splitting step of parameter expansion:

$ str=us-east-2
$ IFS=- eval 'set -- $str'
$ echo $#
3
$ echo $1
us
$ echo $2
east
$ echo $3

No external utilities; just processing in the language.

This is how smartly written build configuration scripts parse version numbers like 1.13.4 and architecture strings like i386-gnu-linux .

The eval can be avoided, if we save and restore IFS .

$ save_ifs=$IFS; set -- $str; IFS=$save_ifs

Answer 3

Using bash, and assuming that you need to distinguish between things like southwest and southeast:

s=ap-southwest-1

a=${s:0:2}
b=${s#*-}
b=${b%-*}
c=${s##*-}

bb=
case "$b" in
south*) bb+=s ;;&
north*) bb+=n ;;&
*east*) bb+=e ;;
*west*) bb+=w ;;
esac

echo "$a$bb$c"

Answer 4

How about:

region="us-east-1"
echo "$region" | (IFS=- read -r a b c; echo "$a${b:0:1}${c:0:1}")
use1

Answer 5

A simple sed -

$: printf "us-east-1\nap-southeast-1\n" |
     sed -E 's/-(.)[^-]*/\1/g'

To keep noncardinal specifications like southeast distinct from south at the cost of adding an optional additional character -

$: printf "us-east-1\nap-southeast-1\n" |
   sed -E '
    s/north/n/;
    s/south/s/;
    s/east/e/;
    s/west/w/;
    s/-//g;'

If you could have south-southwest , add g to those directional reductions.

if you MUST have exactly 4 characters of output, I recommend mapping the eight or 16 map directions to specific characters, so that north is N, northeast is maybe O and northwest M... that sort of thing.