[SOLVED] Extract lines matching specific pattern + additional manipulation

Issue

Given a list of directories and/or file paths, I want to extract the paths of the directories containing files with names matching a given pattern, e.g. files having a .txt extension.

Sample input file:

/a/b/c/d.txt
/a/b/c/d/e.txt
f
g
h/i
/a/b/c.txt

Expected output:

/a/b/c/
/a/b/c/d/
/a/b/

How do I do this with Linux command line tools? I'm wondering if the following is a good start:

grep "\.txt" foo | <what else?>

Solution

(Assuming the desired output has a trailing / on the 2nd and 3rd lines, otherwise it's inconsistent with the 1st line of output...)

Look at this:

sed -En '/\.txt$/s!(.*/).*!\1!p' foo

where

• -E is to use ( and ) instead of \( and \) for grouping
• -n tells Sed not to print the pattern space by default
• /\.txt$/ only matches those lines ending by .txt
• s run a substitution on those lines we matched with /\.txt$/
• ! is used instead of / as a separator, because so we don't have to escape /
• (.*/).* matches the whole line, but captures only up to and including the last /
• \1 replaces the line with the part that we catpured
• the p flag tells Sed to print the line

Answered By - Enlico

Answer Checked By - David Marino (WPSolving Volunteer)