sed以某种方式写一行

Question

I have this text file that have lines made in a certain format just like this next line

bla bla name1=WORD1 bla    bla name2=WORD2 bla bla name3=WORD2

I want to extract WORD1 WORD2 WORD3 without all the bla bla and printing them with semicolon

WORD1;WORD2;WORD3

can this be done using only sed ?

Answer 1

One way that handles a different key=value string.

Assuming infile with content:

bla bla name1=WORD1 bla   noname=WORD4 bla name2=WORD2 bla bla name3=WORD3

And script.sed with content:

## Add a newline character just before each word.
s/name[1-3]=\([^ ]*\)/\n\1/g;

## Remove all characters until each newline appended in previous command, so only
## words will be left, and insert a ';' between them.
s/[^\n]*\n\([^ ]*\)/\1;/g;

## Remove last ';'.
s/;[ ]*$//;

Run it like:

sed -f script.sed infile

That yields:

WORD1;WORD2;WORD3

Answer 2

If WORD* always occur in this manner, you can use these two patterns [^=]*= and [^ ]* to match before and the word respectively. ^ at the beginning of a group inverts the matching. Something like this works in GNU sed:

sed -r 's/[^=]*=([^ ]*)[^=]*=([^ ]*)[^=]*=(.*)/\1;\2;\3/' infile

Answer 3

If you are interseted in awk, below would work.

awk '{for(i=1;i<=NF;i++){if($i~/\=/){split($i,a,"=");if(p){p=p";"a[2]}else{p=a[2]}}}}END{print p}'

tested below:

> echo "bla bla name1=WORD1 bla    bla name2=WORD2 bla bla name3=WORD2" | awk '{for(i=1;i<=NF;i++){if($i~/\=/){split($i,a,"=");if(p){p=p";"a[2]}else{p=a[2]}}}}END{print p}'
WORD1;WORD2;WORD2
>