sed replace number with number-1

Question

Can sed handle a substitution and simultaneously evaluate an expression to substitute somehow?

For example, I have the following entry in a text file:

##################################### topd Tree1 - Tree14 #######################################

For reasons I won't go in to, each Tree's number is N+1 relative to another numbering scheme I've been using so far.

Could sed (or any util really, sed is just my go to for find-and-replace operations), find every instance of TreeN and replace them with TreeN-1 such that the above line would then look like:

##################################### topd Tree0 - Tree13 #######################################

(Using Ubuntu 12.02, so pretty much anything goes.)

I think they are (there is the same run of ### space, topd.... etc etc, though that seems less robust to me that finding the string? — Joe Healey, Jul 13 '16 at 10:12

123 · Accepted Answer · 2016-07-13T10:20:44.790

4

Using perl

perl -pe 's/Tree\K\d+/$&-1/ge' file

Output

##################################### topd Tree0 - Tree13 #######################################

edited Jul 13 '16 at 10:20

answered Jul 13 '16 at 10:15

123

10,778
2
22
45

using perl seems to be the easiest solution. – sjsam Jul 13 '16 at 10:33
Nice, I ended up trying this one first so accepted as the answer. To make it a bit more informative, could you elaborate on which bits of syntax are doing what? I can pick apart elements but not all of it. – Joe Healey Jul 13 '16 at 10:59
@JoeHealey `\K` cause the replacement to only affect the right hand side, so only the digits `\d+` are matched. `$&` is matched string. `g` is global, as in make as many matches as many as possible. `e` evaluate the replacement part of the regex as though it were a command. – 123 Jul 13 '16 at 12:22

Tom Fenech · Answer 2 · 2016-07-13T10:16:48.810

1

I'd use perl in this case, with the e modifier to evaluate an expression in the replacement part:

perl -pe 's/(Tree)(\d+)/$1 . ($2 - 1)/ge'

You can also take advantage of more advanced regular expression support, e.g. by using a lookbehind assertion:

perl -pe 's/(?<=Tree)(\d+)/$1 - 1/ge'

edited Jul 13 '16 at 10:16

answered Jul 13 '16 at 10:12

Tom Fenech

72,334
12
107
141

2

You will want `g` flag as well and no `#` – 123 Jul 13 '16 at 10:15
1

@123 yeah I didn't notice that both numbers were supposed to be decremented, edited thanks! – Tom Fenech Jul 13 '16 at 10:21

Mustafa DOGRU · Answer 3 · 2016-07-13T11:20:34.957

0

You can try this;

sed 's/[^0-9,-]*//g' filename | sed -e "s/-/ /g" | awk '{print "##################################### topd Tree"$1-1 " - Tree" $2-1 " #######################################"}'

awk '{gsub(/Tree/,"Tree ",$0);print $1" "$2" "$3$4-1" "$5" "$6$7-1" "$8}' filename

edited Jul 13 '16 at 11:20

answered Jul 13 '16 at 10:29

Mustafa DOGRU

3,994
1
16
24

1

A single awk solution is possible for this. This looks two `sed`s and `awk` seems redundant – sjsam Jul 13 '16 at 10:35

score 0 · Answer 4 · answered Jul 13 '16 at 10:59

0

Using gnu awk below

awk '{
    for(i=1;i<=NF;i++){
     $i ~ /^Tree[[:digit:]]+$/
     {
     n=(gensub(/Tree/,"",1,$i)-1);
     $i=gensub(/Tree([[:digit:]]+)/,"Tree"n,1,$i);
     }
     }
     print}' file

would do the trick

answered Jul 13 '16 at 10:59

sjsam

21,411
5
55
102

sed replace number with number-1

4 Answers4

Linked