I have a file that contains a matrix. For example, I have:
1 a 2 b
2 b 5 b
3 d 4 b
4 b 7 b
I know it's easy to use sed command to delete specific lines with specific strings. But what if I only want to delete those lines where the second field's value is b (i.e., second line and fourth line)?
You can use regex in sed.
sed -i 's/^[0-9]\s+b.*//g' xxx_file
or
sed -i '/^[0-9]\s+b.*/d' xxx_file
The "-i" argument will modify the file's content directly, you can remove "-i" and output the result to other files as you want.
^ marks the start of the line; the [0-9] looks for a single digit; the \s* looks for zero or more space-like characters (if the version of sed you're using supports it, which on Linux is normally the case — beware on non-Linux systems, though), followed by a b (and then the .* matches anything else on the line). The first command replaces all that with 'nothing', leaving you a blank line when there's a b in the second column; the second deletes the line (and really doesn't need the .* at the end of the regex). [...continued...]b in the second column (and doesn't match anything else in the sample data), but it isn't bullet proof. It would delete a line 9b c 7 x (because there can be zero spaces between the digit and the b. It also would delete 9 bad 8 y; it is not entirely clear whether your specification means that should be deleted or not. Fixing those issues is far from impossible; it is not clear whether it is necessary either.
Awk just work fine, just use code as below:
awk '{if ($2 != "b") print $0;}' file
if you want get more usage about awk, just man it!
awk:
cat yourfile.txt | awk '{if($2!="b"){print;}}'
$2 != "b" would print the correct lines; the == condition prints the lines that should be deleted. Also, you get a UUOC Award. You can and should use awk '...' yourfile.txt.