Question

So I’ve got a txt file where each line is a file path, I would like to:

1. Read this txt file (line-by-line).
2. Delete all lines that do not end with ,-,.txt
3. In the remaining lines, delete everything from after the last / to the ,-,.txt.
4. Write the output to a new txt.

How could this be done with sed?

Input:

/a/b1/
/a/b1/car
/a/b1/car/bil/
/a/b1/car/bil/,-,.txt
/a/b2/
/a/b2/flower
/a/b2/flower/bil/
/a/b2/flower/bil/,-,.txt
/a/b2/
/a/b2/boat
/a/b2/boat/baat/
/a/b2/boat/baat/abc,-,.txt

Second step:

/a/b1/car/bil/,-,.txt
/a/b2/flower/bil/,-,.txt
/a/b2/boat/baat/abc,-,.txt

Third step/desired output:

/a/b1/car/bil/
/a/b2/flower/bil/
/a/b2/boat/baat/

Answer 1

sed -n '/,-,\.txt$/s|/[^/]*$||p' input.txt > output.txt

What it does:

It reads a line at a time from input.txt; -n tells it not to print lines by default. For each line that matches the pattern ,-,\.txt$, everything consisting of a / character followed by zero or more non-/ characters, up to the end of the line, is deleted (i.e., from the last / to the end of the line); I use | as the delimiter so I don’t have to escape the /.

This is a fairly straightforward rendition of your requirements.

Now that you’ve posted sample input and output, I see that you want to keep the final / (which is inconsistent with your requirement “delete everything from the last / to the ,-,.txt“). To do that:

sed -n '/,-,\.txt$/s|/[^/]*$|/|p' input.txt > output.txt

This produces your expected results given your sample input.

If I were doing this on the fly, I might use a simpler approach, combining sed and grep:

grep ',-,\.txt$' input.txt | sed 's|/[^/]*$|/|' > output.txt