Ttyjfyk

I have a problem with the output of a program. I need to launch a command in bash and take its output (a string) and split it to add new lines in certain places. The string looks like this:

battery.charge: 90 battery.charge.low: 30 battery.runtime: 3690 battery.voltage: 230.0 device.mfr: MGE UPS SYSTEMS device.model: Pulsar Evolution 500

basically it is an xxx.yy.zz: value, but the value may contain spaces.
Here's the output I'd like to get

battery.charge: 90
battery.charge.low: 30
battery.runtime: 3690
battery.voltage: 230.0
device.mfr: MGE UPS SYSTEMS
device.model: Pulsar Evolution 500 

I have an idea to search for first dot and then look back from that position for space to put a new line there, but I'm not sure how to achieve it in Bash.

Pure bash solution, no external tools used to process the strings, just parameter expansion:

#! /bin/bash
str='battery.charge: 90 battery.charge.low: 30 battery.runtime: 3690 battery.voltage: 230.0 device.mfr: MGE UPS SYSTEMS device.model: Pulsar Evolution 500'

IFS=: read -a fields <<< "$str"

for (( i=0 ; i < $#fields[@] ; i++ )) ; do
 f=$fields[i]

 notfirst=$(( i>0 ))
 last=$(( i+1 == $#fields[@] ))

 (( notfirst )) && echo -n $f% *

 start=('' $'n' ' ')
 colon=('' ': ')
 echo -n "$start[notfirst + last]$f##* $colon[!last]"
done
echo

Explanation: $notfirst and $last are booleans. The part before the last space $f% * isn't printed for the first field, as there is no such thing. $start and $colon hold various strings that separate the fields: at the first item, notfirst + last is 0, so nothing is prepended, for the rest of the lines, $notfirst is 1, so a newline is printed, and for the last line, the addition gives 2, so a space is printed. Then, the part after the last space is printed $f##* . Colon is printed for all lines except the last one.

A perl solution:

$ perl -pe 'sS+:$seen++ ? "n$&" : "$&"ge' file
battery.charge: 90 
battery.charge.low: 30 
battery.runtime: 3690 
battery.voltage: 230.0 
device.mfr: MGE UPS SYSTEMS 
device.model: Pulsar Evolution 500

Explanation

• 
  S+: matches string end with :.


• With all matched strings, we insert the newline before them ("n$&") except the first one ($seen++).

With GNU sed, you can match each contiguous string (i.e. without whitespace) terminated by : and then place a newline before all but the first one:

sed 's/[^[:space:]]+:/n&/g2'

If your version of sed does not support the gn extension, you can use a plain g modifier

sed 's/[^[:space:]]1,:/
&/g'

which will work the same except for printing an additional newline before the first key. You could use perl -pe 's/S+:/n$&/g' with the same proviso (there may be a perl equivalent of the GNU sed g2 but I don't know it).

It's easier using a tool that supports lookarounds:

$ s="battery.charge: 90 battery.charge.low: 30 battery.runtime: 3690 battery.voltage: 230.0 device.mfr: MGE UPS SYSTEMS device.model: Pulsar Evolution 500"
$ grep -oP 'S+:s+.*?(?=s+S+:|$)' <<< "$s"
battery.charge: 90
battery.charge.low: 30
battery.runtime: 3690
battery.voltage: 230.0
device.mfr: MGE UPS SYSTEMS
device.model: Pulsar Evolution 500

If you wanted the result in an array:

$ IFS=$'n' foo=($(grep -oP 'S+:s+.*?(?=s+S+:|$)' <<< "$s"))
$ for i in "$!foo[@]"; do echo "$i<==>$foo[i]"; done
0<==>battery.charge: 90
1<==>battery.charge.low: 30
2<==>battery.runtime: 3690
3<==>battery.voltage: 230.0
4<==>device.mfr: MGE UPS SYSTEMS
5<==>device.model: Pulsar Evolution 500

EDIT: Explanation of the regex:

'S+:s+.*?(?=s+S+:|$)'

• 
  S+ matches one or more non-whitespace characters


• 
  : matches :
  


• 
  s+ matches one or more spaces after the :
  


• 
  .*? denotes a non-greedy match


• 
  (?=s+S+:|$) is a lookahead assertion to determine if there is:
  
  • one or more space followed by a string (non-whitespace charaters) and a colon, or
  
  
  • end of string
  
  

So the string is split into parts like battery.charge: 90, ... device.mfr: MGE UPS SYSTEMS, ...

Below are links to a couple of online regular expression analyzers:

• http://rick.measham.id.au/paste/explain.pl


• http://xenon.stanford.edu/~xusch/regexp/

Here's a naive approach that should work assuming you don't care that tabs and newlines in the input (if any) are converted to plain spaces.

The idea is simple: split the input on whitespace, and print every token except that you prepend tokens that end with : with a newline (and re-add a space in front of the others). The $count variable and related if are only useful to prevent an initial empty line. Could be removed if that's not a problem. (The script assumes the input is in a file called intput in the current directory.)

#! /bin/bash

count=0
for i in $(<input) ; do
 fmt=
 if [[ $i =~ :$ ]] ; then
 if [[ $count -gt 0 ]] ; then
 fmt="n%s"
 else
 fmt="%s"
 fi
 ((count++))
 else
 fmt=" %s"
 fi
 printf "$fmt" "$i"
done
echo
echo "Num items: $count"

I hope someone can come up with a nicer alternative though.

$ cat input
battery.charge: 90 battery.charge.low: 30 battery.runtime: 3690 battery.voltage: 230.0 device.mfr: MGE UPS SYSTEMS device.model: Pulsar Evolution 500
$ ./t.sh
battery.charge: 90
battery.charge.low: 30
battery.runtime: 3690
battery.voltage: 230.0
device.mfr: MGE UPS SYSTEMS
device.model: Pulsar Evolution 500
Num items: 6

You can use awk(1) with the following script split.awk:

BEGIN RS=" "; first=1; 
first first=0; printf "%s", $1; next; 
/[a-z]+.[^:]+:/ printf "n%s", $1; next; 
 printf " %s", $1 
END printf "n" 

When you run

awk -f split.awk input.dat

you will get

battery.charge: 90
battery.charge.low: 30
battery.runtime: 3690
battery.voltage: 230.0
device.mfr: MGE UPS SYSTEMS
device.model: Pulsar Evolution 500

The idea is to let awk split the input when it sees a space (setting record separator RS in line 1). Then it matches xxx.yy.zz: values in line 2 and 3 (distinguishing the very first match from subsequent ones), while line 4 matches whenever line 2 and 3 do not match. Line 5 just print the last newline.