Print Valid words with _ in between them - sed

Question

I have done my research, but not able to find the solution to my problem.
I am trying to extract all valid words(Starting with a letter) in a string and concatenate them with underscore("_"). I am looking for solution with awk, sed or grep, etc.
Something like:
echo "The string under consideration" | (awk/grep/sed) (pattern match)
Example 1
Input:
1.2.3::L2 Traffic-house seen during ABCD from 2.2.4/5.2.3a to 1.2.3.X11
Desired output:
L2_Traffic_house_seen_during_ABCD_from
Example 2
Input:
XYZ-2-VRECYY_FAIL: Verify failed - Client 0x880016, Reason: Object exi
Desired Output:
XYZ_VRECYY_FAIL_Verify_failed_Client_Reason_Object_exi
Example 3
Input:
ABCMGR-2-SERVICE_CRASHED: Service "abcmgr" (PID 7582) during UPGRADE
Desired Output:
ABCMGR_SERVICE_CRASHED_Service_abcmgr_PID_during_UPGRADE

Answer 1

This might work for you (GNU sed):
sed 's/[[:punct:]]/ /g;s/\<[[:alpha:]]/\n&/g;s/[^\n]*\n//;s/ [^\n]*//g;y/\n/_/' file

Answer 2

A perl one-liner. It searches any alphabetic character followed by any number of word characters enclosed in word boundaries. Use the /g flag to try several matches for each line.
Content of infile:
1.2.3::L2 Traffic-house seen during ABCD from 2.2.4/5.2.3a to 1.2.3.X11
XYZ-2-VRECYY_FAIL: Verify failed - Client 0x880016, Reason: Object exi
ABCMGR-2-SERVICE_CRASHED: Service "abcmgr" (PID 7582) during UPGRADE
Perl command:
perl -ne 'printf qq|%s\n|, join qq|_|, (m/\b([[:alpha:]]\w*)\b/g)' infile
Output:
L2_Traffic_house_seen_during_ABCD_from_to_X11
XYZ_VRECYY_FAIL_Verify_failed_Client_Reason_Object_exi
ABCMGR_SERVICE_CRASHED_Service_abcmgr_PID_during_UPGRADE

Answer 3

One way using awk, with the contents of script.awk:
BEGIN {
FS="[^[:alnum:]_]"
}
{
for (i=1; i<=NF; i++) {
if ($i !~ /^[0-9]/ && $i != "") {
if (i < NF) {
printf "%s_", $i
}
else {
print $i
}
}
}
}
Run like:
awk -f script.awk file.txt
Alternatively, here is the one liner:
awk -F "[^[:alnum:]_]" '{ for (i=1; i<=NF; i++) { if ($i !~ /^[0-9]/ && $i != "") { if (i < NF) printf "%s_", $i; else print $i; } } }' file.txt
Results:
L2_Traffic_house_seen_during_ABCD_from_to_X11
XYZ_VRECYY_FAIL_Verify_failed_Client_Reason_Object_exi
ABCMGR_SERVICE_CRASHED_Service_abcmgr_PID_during_UPGRADE

Answer 4

This solution requires some tuning and I think one needs gawk to have regexp as "record separator"
http://www.gnu.org/software/gawk/manual/html_node/Records.html#Records
gawk -v ORS='_' -v RS='[-: \"()]' '/^[a-zA-Z]/' file.dat