replace a pattern using sed

Question

I have a file containing a lot of text and some digits describing numbers < 1 with three digits of accuracy. I'd like to replace those numbers with roughly equivalent integer percentages (numbers 0-99).

0.734 -> 73
0.063 -> 6
0.979 -> 97

It would be great to round properly, but not required.

I've tried the following and several variants and I can't seem to get even one match:

sed -e 's/0\.(\d\d)/&/' myfile.txt

Which I understand to mean, match the digit 0, then the decimal, capture the next to digits and have sed replace the whole match with the captured part?

Even if I got this to work, I don't know how to handle the 0.063 -> 6 case. Sure would apprecaite a helping hand on this.

Answer 1

sed does support character class but use the longer POSIX name. Digits are [[:digit:]] . It's shorter to just write [0-9] .

Try this:

sed -E 's/0\.([0-9][0-9]).*/\1/;s/^0//' myfile.txt

The -E flag tells it to use modern regex. There are actually 2 commands here, separated by ; :

s/0\\.([0-9][0-9]).*/\\1/ : put the two digits following 0 and a dot into a capture group and replace the whole string with this capture group.

s/^0// : remove the leading zero from the string after the above.

Answer 2

Apart from sed answer Zoff gave, you can use awk for better implementation (with roundings):

#round down
awk '{print int($1*100)}' myfile.txt
#0.979 -> 97

#round up
awk '{printf "%.0f\n",$1*100}' myfile.txt
#0.979 -> 98

A bash only implementation:

#round down
while read n; do result=$(bc <<< "$n*100"); echo ${result%%.*}; done < myfile.txt
#round up
while read n; do result=$(bc <<< "$n*100"); printf "%.f\n" $result; done < myfile.txt

Answer 3

This might work for you (GNU sed):

sed -r 's/0\.(([1-9][0-9])|0([0-9])).*/\2\3/' file

This uses alternation and backreferences (BR) to match the required pattern. If the first pattern matches the second BR will return that value and the third (BR) will be empty. Likewise if the second pattern in the alternation matches the the second BR will be empty and the third BR will return the required value.