AWK print all regex matches on every line

With GNU awk you could use its OOTB variable named FPAT and could try following awk code.

awk -v FPAT='<[^>]*>' ' NF{ val="" for(i=1;i<=NF;i++){ val=(val?val OFS:"") $i } print val } ' Input_file 

Assuming there are no stray angle brackets, use either < or > as a field separator and print every second field:

awk -F'[<>]' '{for (i=2; i <= NF; i += 2) {printf "<%s> ", $i}; print ""}' data 

match() doesn't work the way you think it does; to find a variable number of matches you would need to first match() the first pattern, strip off that pattern, then match() the remainder of the input for the next pattern, and repeat until no more matches in the current line; eg:

awk ' { out=sep="" # init variables for new line while (match($0,/<[^>]+>/)) { # find 1st match out=out sep substr($0,RSTART,RLENGTH) # build up output line $0=substr($0,RSTART+RLENGTH) # strip off 1st match and prep for next while() check sep=OFS # set field separator for follow-on matches } if (out) print out }' somefile.txt 

Another idea uses the split() function, eg:

awk ' { n=split($0,a,/[<>]/) # split line on dual delimiters "<" and ">" out=sep="" for (i=2;i<=n;i=i+2) { # step through even numbered array entries; assumes line does not contain any standalone "<" or ">" characters !!! out=out sep "<" a[i] ">" # build output line sep=OFS } if (out) print out } ' somefile.txt 

Both of these generate:

<a> <b> <c> <d> <e> <f> 

I would harness GNU AWK for this task following way, let file.txt content be

lorem <a> ipsum <b> dolor <c> sit amet, consectetur <d> adipiscing elit <e>, sed do eiusmod <f> tempor incididunt ut 

then

awk 'BEGIN{FPAT="<[^>]*>"}{$1=$1;print}' file.txt 

gives output

<a> <b> <c> <d> <e> <f> 

Explanation: I inform GNU AWK that field is < followed by zero-or-more (*) non(^)-> followed by >. For each line I do $1=$1 to provoke rebuilt, so now line are found fields joined by space, which I then print.

(tested in gawk 4.2.1)

Here's a simple awk solution based on regexps:

awk '{ gsub(/^[^<]*|[^>]*$/,""); gsub(/>[^<]*</,"> <") } NF' 

^{edit: using NF instead of $0 != ""; thanks @EdMorton}

For each line:

• strip all chars from the left up to the first < (excluded) or up to the end-of-line when < isn't found.
• strip all chars from the right up to the first > (excluded) or up to the start-of-line when > isn't found.
• replace what's between each > and < pair with a space character.
• print the result when it isn't empty

example

lorem <a a> ipsum <b> dolor <c> sit amet, consectetur <d> adipiscing elit <e>, sed do eiusmod <f> tempor <g>incididunt ut<h><i> h>ell<o <j> 

output

<a a> <b> <c> <d> <e> <f> <g> <h> <i> <j> 

Remark: With exactly the same logic you can use sed:

sed 's/^[^<]*//; s/[^>]*$//; s/>[^<]*</> </g; /^$/d' 

INPUT

lorem <a> ipsum <b> dolor <c> sit amet, consectetur <d> adipiscing elit <e>, sed do eiusmod <f> tempor incididunt ut 

CODE

mawk -F'^[^<]+|[^>]+$' 'gsub(">[^<]*<","> <",$!(NF=NF))^_*/./' OFS= 

OUTPUT

<a> <b> <c> <d> <e> <f> 

Another option is to use gnu awk with gensub. You can capture the angle brackets with optional surrounding spaces and match the rest.

In the replacement use group 1 surrounded with a single space.

awk '{$0 = gensub(/ *(<[^>]*>) *|[^<>]+/, " \\1 ", "g"); $1=$1}1' file 

Output

<a> <b> <c> <d> <e> <f> 

if you really wanna do it the patmatch() way, here's how to emulate that effect in other awks :

echo 'lorem <a> ipsum <b> dolor <c> sit amet, consectetur <d> adipiscing elit <e>, sed do eiusmod <f> tempor incididunt ut' | 

awk ' BEGIN { RS = "^$" } _ = gsub(/[<][^>]*[>]/, "\4&\5") { split($!_, __, /((^|\5)[^\4]*)\4|\5[^\4]*$/) for (_ in __) print _, __[_] }' 

1 2 <a> 3 <b> 4 <c> 5 <d> 6 <e> 7 <f> 8