Shell scripting - Modifying a files contents with sed

[ampakine]

Lets say I want write a script that will find every occurrence of the pattern "PATTERN" in a file called "FILE". If I run this command:

Code: Select all

sed 's/PATTERN//' FILE

it will output text in which the pattern has been deleted but it doesn't actually modify the file so when I run cat FILE the pattern is still there. I tried redirecting the output to the file itself like this:

Code: Select all

sed 's/PATTERN//' FILE > FILE

but that deleted all text in the file for some reason. I can solve the problem by redirected the output to an intermediate file then outputting the intermediate files contents to the original file like this:

Code: Select all

sed 's/PATTERN//' FILE > INTERMEDIATEFILE
cat INTERMEDIATEFILE > FILE


but I'm guessing thats a completely roundabout way to do it and I wouldn't need to do that if I knew how to use the sed command properly.

[tgoe]

AFAIK, the FreeBSD version of sed is the only one able to modify files in place. You can avoid creating the temp file by first reading all of the data into memory.

edit:
GNU sed can edit in place: -i option