sed: replace spaces within quotes with underscores

Question

I have input (for example, from ifconfig run0 scan on OpenBSD) that has some fields that are separated by spaces, but some of the fields themselves contain spaces (luckily, such fields that contain spaces are always enclosed in quotes).

I need to distinguish between the spaces within the quotes, and the separator spaces. The idea is to replace spaces within quotes with underscores.

Sample data:

%cat /tmp/ifconfig_scan | fgrep nwid | cut -f3
nwid Websense chan 6 bssid 00:22:7f:xx:xx:xx 59dB 54M short_preamble,short_slottime
nwid ZyXEL chan 8 bssid cc:5d:4e:xx:xx:xx 5dB 54M privacy,short_slottime
nwid "myTouch 4G Hotspot" chan 11 bssid d8:b3:77:xx:xx:xx 49dB 54M privacy,short_slottime

Which doesn't end up processed the way I want, since I haven't replaced the spaces within the quotes with the underscores yet:

%cat /tmp/ifconfig_scan | fgrep nwid | cut -f3 |\
    cut -s -d ' ' -f 2,4,6,7,8 | sort -n -k4
"myTouch Hotspot" 11 bssid d8:b3:77:xx:xx:xx
ZyXEL 8 cc:5d:4e:xx:xx:xx 5dB 54M
Websense 6 00:22:7f:xx:xx:xx 59dB 54M

Answer 1

For a sed-only solution (which I don't necessarily advocate), try:

echo 'a b "c d e" f g "h i"' |\
sed ':a;s/^\(\([^"]*"[^"]*"[^"]*\)*[^"]*"[^"]*\) /\1_/;ta'
a b "c_d_e" f g "h_i"

Translation:

• Start at the beginning of the line.
• Look for the pattern junk"junk", repeated zero or more times, where junk doesn't have a quote, followed by junk"junk space.
• Replace the final space with _.
• If successful, jump back to the beginning.

Answer 2

try this:

awk -F'"' '{for(i=2;i<=NF;i++)if(i%2==0)gsub(" ","_",$i);}1' OFS="\"" file

it works for multi quotation parts in a line:

echo '"first part" foo "2nd part" bar "the 3rd part comes" baz'| awk -F'"' '{for(i=2;i<=NF;i++)if(i%2==0)gsub(" ","_",$i);}1' OFS="\"" 
"first_part" foo "2nd_part" bar "the_3rd_part_comes" baz

EDIT alternative form:

awk 'BEGIN{FS=OFS="\""} {for(i=2;i<NF;i+=2)gsub(" ","_",$i)} 1' file