Shell Programming and Scripting

You did not show the desired output for the sample data, so I'm assuming that your goal is to print out each line in file2 that contains a line in file1 as a substring.

Code:

grep -Ff file1 file2

Regards,
Alister

---------- Post updated at 12:02 PM ---------- Previous update was at 11:58 AM ----------

To use a different delimiter for an address, the opening delimiter must be backslash-escaped (this is not necessary when using a different delimiter to delimit the components of a substitute, s, command).

However, any sed solution will treat literal text as a regular expression. That is a very poor approach. Each . in file1 will match any character at all. Each * will allow the preceding character to occur an arbitrary number of times, or to not occur at all. And so it goes with all regular expression metacharacters.

Regards,
Alister

OW yes you went back to using single quotes, double quotes is what you would need..
try:

Code:

sed -i "\%${line}%d"

It would be inefficient though..

The awk is not producing any output since every record of file2 is contained in file1 so the output should be empty. Otherwise, what do you expect the output (you did not explicitly specify) to be?

I remind it should be sed -i "%${line}%d" (not work) but I see you add \ before the first % and it works

Lei


---------- Post updated at 10:20 AM ---------- Previous update was at 10:06 AM ----------

this sed still can't work in some cases.
I want to delete the file2'line when the line contains the content of file1

file2

Code:

3gdgdgd python: regenerate plat-linux2/*.py
34sfddf python: depend on bzip2
fdererf the good test

file1

Code:

python: regenerate plat-linux2/*.py

i want to get below for file2

Code:

34sfddf python: depend on bzip2
fdererf the good test

with below code it still not work

Code:

#!/bin/bash
#s=`cat file1 |awk -F "zzzzzz" '{print $2}' |sed -n "${n}p"`
while read line
do
sed -i "\%${line}%d" file2
done < file1

sed cannot be made to work correctly, unless you escape every single basic regular expression metacharacter. And, since some characters are special in some situations and not in others, you cannot naively place a backslash before every one of them; you'll have to build a basic regular expression lexer and parser. Or ...

You should stop treating literal text as regular expressions. Use awk, grep, or some other tool that can treat arbitrary input as a literal string.

Here's another line in your sample data that will cause a problem:

Code:

[e]glibc-package.inc: fix providers for libsegfault and other

As a regular expression, it won't match itself (because the [e] won't match the brackets themselves).

Regards,
Alister