Jtdylktuy

Given a file type .txt, find all directories that has only that type of file.

For example

a--

   b-- 1.txt

   c--

      |

       ---- 2.jpg

       ---- 3.txt

The command should output only b not c.

find . -type d -execdir sh -c '

   [ "$(find "$1" -maxdepth 1 -type f -name "*.txt" -print -quit | wc -l)" -gt 0 ] &&

   [ "$(find "$1" -maxdepth 1 -type f ! -name "*.txt" -print -quit | wc -l)" -eq 0 ]

   ' find-sh {} ; -print

The outer find supplies directories to be examined. Two inner find-s check if there is at least one .txt file and no non-.txt files in the directory. sh shell implements logic with [ … ] and &&.

Notes:

• 
  -maxdepth is not required by POSIX. For POSIX approach see this question.


• 
  -quit is not required by POSIX. This action makes find quit as soon as any matching file is reported. It's useful because we need at most one matching file to obtain the result with wc -l and [ … ], so quitting early saves time. Without -quit the whole command will work, it will just be slower when there are many files. Alternatively you can use find … | head -n 1 | wc -l; in this case head will terminate the pipe after the first file found, wc will yield the result right away, but find will only notice the broken pipe when (if) it tries to write yet another line. And it's a trade-off: head may save you some time and resources but (as a separate process) it needs time and resources to be spawned twice in every directory.


• Paths with newlines (if any) will fool wc -l but it doesn't matter because these extra newlines may add to the count only if the "right" count is non-zero anyway, and we only need to know whether the result is zero or not.