Find two lines and replace with one

Question

I am looking for a solution that would allow me to search text files on a linux server that would look a file and find a pattern such as:

Text 123
Blue Green

And then replaces it with one line, every time it finds it in a file...

Order Blue Green

I am not sure what would be the easiest way to solve this. I have seen many guides using SED but only for finding one line and replacing it.

Answer 1

You ask about sed, here is an answer in sed.
Let me mention however, that while sed is fun for this kind of exercise, you probably should choose something else, more flexible and easier to learn; perl for example.

• look for first line /Text 123/
• when found start a loop :a
  • concat next line N
  • replace twins of searched text with single copy and print it
    s/Text 123\nText 123/Text 123/p;
  • loop while that replaces ta;
• try to replace s///
• rely on concat being printed unchanged if replace does not trigger

Code:

sed "/Text 123/{:a;N;s/Text 123\nText 123/Text 123/p;ta;s/Text 123\nBlue Green/Order Blue Green/}"

Test input:

Text 123
Do not replace
Lala
Text 123
Blue Green
lulu
Text 123
Do not replace either
Text 123
Text 123
Blue Green
preceding should be replaced

Output:

Text 123
Do not replace
Lala
Order Blue Green
lulu
Text 123
Do not replace either
Text 123
Order Blue Green
preceding should be replaced

Platform: Windows and GNU sed version 4.2.1

Note: On that platform the sed line allows to use the environment variables for the two text fragments, which you probably want to do:

sed "/%EnvVar2%/{:a;N;s/%EnvVar2%\n%EnvVar2%/%EnvVar2%/p;ta;s/%EnvVar2%\n%EnvVar%/Order %EnvVar%/}"

Platform2:
still Windows
using bash GNU bash, version 3.1.17(1)-release (i686-pc-msys)
GNU sed version 4.2.1 (same)

On this platform, variables can e.g. be used like:

sed "/${EnvVar2}/{:a;N;s/${EnvVar2}\n${EnvVar2}/${EnvVar2}/p;ta;s/${EnvVar2}\n${EnvVar}/Order ${EnvVar}/}"

On this platform it is important to use "..." in order to be able to use variables,
it does not work with '...'.

As @edMorton has hinted, on all platforms be careful however with trying to replace (using variables) text which looks like using a variable. E.g. with "Text $123" in bash. In that case, not using variables but trying to replace text which looks like variables, using '...' instead of "..." is the way to go.

Answer 2

sed is for simple substitutions on individual lines, that is all. If you find yourself trying to use constructs other than s, g, and p (with -n) then you are on the wrong track as all other sed constructs became obsolete in the mid-1970s when awk was invented.

Your problem is not doing substitutions on individual lines, it's on a multi-line record and to do that with GNU awk for multi-char RS is:

$ awk -v RS='^$' -v ORS= '{gsub(/Text 123\nBlue Green/,"Order Blue Green")}1' file
Order Blue Green

but there are several other approaches depending on your real needs.