Replace Strings Using Sed And Regex

Question

I'm trying to uncomment file content using sed but with regex (for example: [0-9]{1,5})

# one two 12
# three four 34
# five six 56

The following is working:

sed -e 's/# one two 12/one two 12/g' /file

However, what I would like is to use regex pattern to replace all matches without entering numbers but keep the numbers in the result.

Solution 1

For complying sample question, simply

sed 's/^# //' file

will suffice, but if there is a need to remove the comment only on some lines containing a particular regex, then you could use conditionnal address:

sed '/regex/s/^# //' file

So every lines containing regex will be uncomented (if line begin with a #)

... where regex could be [0-9] as:

sed '/[0-9]/s/^# //' file

will remove # at begin of every lines containing a number, or

sed '/[0-9]/s/^# \?//' file

to make first space not needed: #one two 12, or even

sed '/[0-9]$/s/^# //' file

will remove # at begin of lines containing a number as last character. Then

sed '/12$/s/^# //' file

will remove # at begin of lines ended by 12. Or

sed '/\b\(two\|three\)\b/s/^# //' file

will remove # at begin of lines containing word two or three.

Solution 2

sed -e 's/^#\s*\(.*[0-9].*\)$/\1/g' filename

should do it.

Solution 3

If you only want those lines uncommented which contain numbers, you can use this:

sed -e 's/^#\s*\(.*[0-9]+.*\)/\1/g' file

Comments

[Lee Meador]

How do you define "matches"? Which lines do you want to uncomment and how do you decide that?

[tangi]

it should only uncomment text witch contains numbers.

[tangi]

the issue here is that this will uncomment everything including text without numbers.

[cHao]

I like this; it's a bit clearer about the intent. (You don't want to "replace # + some stuff + a digit + other stuff, with everything but the #"; you want to "remove the # on lines that contain digits".) For regex, you could say [0-9] to have it only affect lines with digits in them, or [0-9]$ if the number has to be at the end.

[Jonathan Leffler]

These solutions delete lines which don't contain digits but that were not already uncommented. Maybe there are no such lines (not even any blank lines?), but it arguably isn't wholly appropriate behaviour.

[F. Hauri - Give Up GitHub]

Could by written: sed -e '/[0-9]/s/^# *//' filename! See my answer

[Seraf]

Don't forget to add -i to replace in place, directly in the file, but without that just to check the output