Day: November 24, 2021

The `sed` command uncommon behaviors

The sed command uncommon behaviors

The sed command is used in Unix, some strange behaviors can let time waste.

Escape char in regex

Normally, the \ is escape character, but it wasn't in some cases.

For example, . is to match any character, it needs to have \ as escape character if need it to be a dot character.

$ echo "test (111) help . 1" | sed -e "s/.//"
est (111) help . 1
$ echo "test (111) help . 1" | sed -e "s/\.//"
test (111) help  1

But this is not for (), without \, they are (), with \, they are indicating subpattern.

$ echo "test (111) help . 1" | sed -e 's/(111)//'
test  help . 1
$ echo "test (111) help . 1" | sed -e 's/\(111\)//'
test () help . 1
$ 

Same for {}

echo "test (111) help . 1" | sed -e 's/hel{1}p//'
test (111) help . 1
$ echo "test (111) help . 1" | sed -e 's/hel\{1\}p//'
test (111)  . 1

and ?

$ echo "test (111) help . 1" | sed -e 's/he?lp//'
test (111) help . 1
$ echo "test (111) help . 1" | sed -e 's/h?lp//'
test (111) help . 1
$ echo "test (111) help . 1" | sed -e 's/he\?lp//'
test (111)  . 1

* and +

The sed understands the meaning of *, but not for +.

$ echo "test (111) help . 1" | sed -e 's/hel*p//'
test (111)  . 1
$ echo "test (111) help . 1" | sed -e 's/hel+p//'
test (111) help . 1

\* and \+

The sed understands the meaning of \+, but not for \*.

$ echo "test (111) help . 1" | sed -e 's/hel\*p//'
test (111) help . 1
$ echo "test (111) help . 1" | sed -e 's/hel\+p//'
test (111)  . 1