Remove previous line to a match string?

Question

How I can remove the previous line of a match pattern?

Or

the opposite of:

sed -n '/pattern/{g;1!p;};h'

Answer 1

Use tac | sed | tac (Linux/Solaris) then it's next line after a match pattern :)

Answer 2

sed is an excellent tool for simple substitutions on a single line, for anything else just use awk:

$ cat file
here is a
a bad line before
a good line
in a file

$ awk 'NR==FNR{if (/good/) del[NR-1]; next} !(FNR in del)' file file
here is a
a good line
in a file

You can use the above idiom to delete any number of lines before and/or after a given pattern, e.g. to delete the 3 lines before and 2 lines after a given target:

$ cat file
-5
-4
-3
-2
-1
target
+1
+2
+3
+4
+5
$
$ awk 'NR==FNR{if (/target/) for (i=-3;i<=2;i++) del[NR+i]; next} !(FNR in del)' file file
-5
-4
+3
+4
+5

or to leave the target in place and just delete the lines around it:

$ awk 'NR==FNR{if (/target/) for (i=-3;i<=2;i++) if (i!=0) del[NR+i]; next} !(FNR in del)' file file
-5
-4
target
+3
+4
+5

All very clear, trivial, and scalable...

Answer 3

For "relatively complex" navigation around a search expression, ed might be a good solution (comments are not part of the command):

ed testfile << EOF
/r.*o/                        # Search the pattern
-1d                           # delete one line above
w                             # write
EOF

Here is an example (using <<< and \n to write as a single line):

sh$ cat testfile
john
paul
george
ringo
sh$ ed testfile <<< $'/r.*o/\n-1d\nw'
23
ringo
16
sh$ cat testfile
john
paul
ringo

Answer 4

You can revert the file and then delete the line after the matche pattern(which is simple), and then revert the result, here is the code:

tail -r|sed '/pattern/{n;d;}'|tail -r

Answer 5

Here is another awk:

awk '/pattern/ {f=1} !f&&NR>1 {print p} {p=$0;f=0} END {print p}' file

---

A tac awk version:

tac file | awk '1; /pattern/ {getline}' | tac

PS getline should normally be avoided since it has many pitfalls, so then this:

tac file | awk '!p||NR!=p+1; /pattern/ {p=NR}' | tac

Answer 6

This might work for you (GNU sed):

 sed '$!N;/\n.*pattern/!P;D' file

Keep a window of 2 lines and test the second of them for the pattern. If the pattern is present do not print the first line.

Answer 7

This would work but only if line count is more than 1 and pattern is not at the last line.

sed -n '/pattern/ { h; b }; 1 { h; b }; ${ H }; x; p' file

Better use awk instead:

awk '!/pattern/ && NR > 1 { print p } { p = $0 } END { if (NR) print p }' file