SED to remove a Line with REGEX Pattern

Question

i've got a hundreds of files with thousands of lines, which i need to delete some lines that follows a pattern,so i went to SED with regex .The struct of files is something like this

A,12121212121212,foo,bar,lorem
C,32JL,JL
C,32JL,JL
C,32JL,JL
C,32JL,JL
A,21212121212121,foo,bar,lorem
C,32JL,JL
C,32JL,JL
C,32JL,JL
A,9999,88888,77777

I need to delete All the lines that starts with "A" and ends with "lorem"

Expected output-

C,32JL,JL
C,32JL,JL
C,32JL,JL
C,32JL,JL
C,32JL,JL
C,32JL,JL
C,32JL,JL
A,9999,88888,77777

I've made the Regex :

^(A).*(lorem)

And it match in my text editor (Sublime,UltraEdit)

In the UNIX shell

sed '/^(A).*(lorem)/d' file.txt

But somehow it doesn't work,it shows the whole file, and i can't figure out why.

Can someone help me please?

Answer 1

$ sed '/^A.*lorem$/d' file.txt

• ^A: starts with an A
• .*: stuff in the middle
• lorem$: ends with lorem

Answer 2

The others gave you correct solutions but didn't explain why your regex didn't work. The () surely were useless, but if you had used the regex with other tools/languages, you might very well have had the expected result.

It didn't work with sed because it will by default use POSIX's basic regular expressions, where the characters for grouping are \( and \), while ( and ) will match literal characters. There were no such brackets in your input text, so it didn't match.

Your regular expression would have worked if you had used GNU's sed -r or BSD's sed -E, the flag switching to POSIX's extended regular expressions where ( and ) are used to group and \( \) match the literal brackets.

In conclusion, the following commands will do the same thing :

• sed '/^A.*lorem$/d' file.txt
• sed -r '/^(A).*(lorem)$/d' file.txt (with GNU sed)
• sed -E '/^(A).*(lorem)$/d' file.txt (with BSD sed and modern GNU sed)
• sed '/^\(A\).*\(lorem\)$/d' file.txt

Answer 3

Remove the brackets.

Using your code, the appropriate one-liner becomes-

sed '/^A.*lorem/d' file.txt

If you want to be more rigourous, you can look at James's answer which more correctly terminates the regex as-

sed '/^A.*lorem$/d' file.txt

Both will work.