AWK recognises the field value "b" in this example:

$ printf "a ab    c d b" | awk '{for (i=1;i<=NF;i++) print $i}'
a
ab
c
d
b
$ printf "a ab    c d b" | awk '{for (i=1;i<=NF;i++) if ($i=="b") print $i}'
b

Note the 3 spaces before the "c". If I try to replace the field with value "b" with "X", the sub replacement happens on the first appearance of "b":

$ printf "a ab c d b" | awk '{for (i=1;i<=NF;i++) if ($i=="b") sub($i,"X"); print}'
a aX   c d b

Is it possible to replace "b" with "X" in the field containing only "b", and also not change the spacing between fields? (Keep the 3 spaces before the "c".)

AWK recognises the field value "b" in this example:

$ printf "a ab    c d b" | awk '{for (i=1;i<=NF;i++) print $i}'
a
ab
c
d
b
$ printf "a ab    c d b" | awk '{for (i=1;i<=NF;i++) if ($i=="b") print $i}'
b

Note the 3 spaces before the "c". If I try to replace the field with value "b" with "X", the sub replacement happens on the first appearance of "b":

$ printf "a ab c d b" | awk '{for (i=1;i<=NF;i++) if ($i=="b") sub($i,"X"); print}'
a aX   c d b

Is it possible to replace "b" with "X" in the field containing only "b", and also not change the spacing between fields? (Keep the 3 spaces before the "c".)

• awk

3 Answers 3

Your code:

$ printf "a ab    c d b" |
awk '{for (i=1;i<=NF;i++) if ($i=="b") sub($i,"X"); print}'
a aX    c d b

fails to change the field you want because you didn't tell it the field name to change when you called sub(), e.g. if ($i=="b") sub($i,"X",$i), and so it's operating on the whole record, but that would fail anyway if the string you wanted to match/modify was a regexp metachar so you should have been trying to do if ($i=="b") sub(/.*/,"X",$i) or better simply if ($i=="b") $i="X".

None of that would address the other, harder to handle, part of your question, though:

and also not change the spacing between fields? (Keep the 3 spaces before the "c".)

Read understanding-how-ofs-works-in-awk for background but you have 2 choices:

1. Change the record, not 1 field of the record, or
2. Save the spacing between fields, then change the field, then restore the spacing.

The first one would be this using any POSIX awk:

$ printf "a ab    c d b" |
awk '
    match(" "$0" ", /[[:space:]]b[[:space:]]/) {
        $0 = substr($0,1,RSTART-1) "X" substr($0,RSTART+RLENGTH-2)
    }
    { print }
'
a ab    c d X

while the second would be this using GNU awk for the 4th arg to split() (you can do the same in any POSIX awk with a while(match($0,[^[:space:]]+)) or similar loop finding the fields but it takes more code):

$ printf "a ab    c d b" |
awk '
    {
        nf = split($0, flds, FS, seps)
        rec = seps[0]
        for (i=1; i<=nf; i++) {
            rec = rec (flds[i]=="b" ? "X" : flds[i]) seps[i]
        }
        $0 = rec
        print
    }
'
a ab    c d X

FWIW for clarity, simplicity, robustness (it's using all literal string operations on the input data), and flexibility (easy to make a regexp instead of string comparison if necessary) I'd use that last script if I really had to do this job.

Is it possible to replace "b" with "X" in the field containing only "b", and also not change the spacing between fields? (Keep the 3 spaces before the "c".)

If you change any field value GNU AWK will use single OFS (default: space) to join them, which result in replacement of one-or-more white-space characters with OFS. In your case I would exploit \y (word boundary) following way

printf "a ab    c d b" | awk '{sub(/\yb\y/,"X");print}'

gives output

a ab    c d X

caveat: while white-space-character letter border does constitute boundary, other combination also do, for example punct letter border, that is

printf "a ab    c d .b." | awk '{sub(/\yb\y/,"X");print}'

gives output

a ab    c d .X.

So you should consider if such values might appear in your input data.

(tested in GNU Awk 5.3.1)

The answer to my question is that sub was not looking in the right place, so to speak! This works:

$ printf "a ab c d b" | awk '{for (i=1;i<=NF;i++) if ($i=="b") sub("b","X",$i); print}'
a ab c d X

The third sub argument is the target for replacement.