CODEWITHSUNDEEP
bashsed
Peter Oram
Peter Oram

Reputation: 6738

How to use sed on mac to delete part of a line and the next line

I have a file which is saved output from an ASCII stream, so it's all text. I'm using a bash script with a sequence of sed statements on a mac to clean it up. There is one aspect of this that I'm stumped on.

In the file there are places where I need to delete part of a line to the end and the next line.

Example section of file: 

abcdefg000

hijk

Should come out to: 

abcdefg
hijk

Tried: 

sed '/000/{N;d;}' FILE

That DOES delete the next blank line, but also deletes the first line. I end up with: 

hijk

Since it's a mac I can't use sed to insert a newline (I've tried), but I have successfully replaced with a character and used tr to switch it out for a newline. Thought if I did that and since tr should take a string, include a special character, then I should be able to do the delete two lines sed and that would work.

sed 's/000/|/' FILE | tr '|' '\n|' | sed '/|/{N;d;}'

However, when I do this, I get only the newline and tr chops off the pipe. sed then doesn't find it and so doesn't delete any lines. I get: 

abcdefg


hijk

man tr says it accepts a string, so not sure why it won't take \n | as a string.

I could redo this in some other language script, but I've spend long enough on it now and looked through enough other questions and answers that I want to get this to work. Either I'm missing something on sed or tr or there's some other simple way to do this.

Upvotes: 0

Views: 933

Answers (2)

anubhava
anubhava

Reputation: 785058

You may use this sed on OSX:

sed '/000$/{s///;n;d;}' file


abcdefg000
hijk


foo

bar

Where original file is:

cat file


abcdefg000

hijk


foo

bar

Upvotes: 1

Greg Nisbet
Greg Nisbet

Reputation: 6994

awk should do a pretty good job of modifying the text file in the way that you want. Conditionally removing a single newline after a line ending with 000 is straightforward. We use a temporary variable w to control how many "lines ahead" we're able to delete blank lines from.

awk '/000$/ { sub(/000$/,""); w = NR + 1; } 
     NF == 0 && w < NR { next; } 
     { print; }'

And here's a way to do it if the script needs to handle the possibility of multiple newlines after a 000. the interpretation of d is whether we're in a state where we're dropping blank lines or not.

awk '/000$/ {d=1;sub(/000$/,"");print;next;} 
     NF == 0 && d { next; } 
    { d = 0; print}'

You can coax sed into cleaning up newlines by swapping newlines with another character and then swapping back. Note that sed will, at least on OS X, add a trailing newline to the stream anyway, so you have to get rid of the stray @ or | or whatever at the very end of the stream.

cat /tmp/data.txt | tr '\n@' '@\n' | sed 's/000@//' | \
    tr '\n@' '@\n' | sed '/^@$/d'

Here's how to compact multiple newlines after 000 to a single newline. Or one way to do it.

cat /tmp/data.txt | tr '\n@' '@\n' | sed 's/000@*/@/' | \
    tr '\n@' '@\n' | sed '/^@$/d'

Upvotes: 1

Related Questions