How to exclude a directory in find . command

Solution 1:

If -prune doesn't work for you, this will:

find -name "*.js" -not -path "./directory/*"

Caveat: requires traversing all of the unwanted directories.

Solution 2:

Use the -prune primary. For example, if you want to exclude ./misc:

find . -path ./misc -prune -o -name '*.txt' -print

To exclude multiple directories, OR them between parentheses.

find . -type d \( -path ./dir1 -o -path ./dir2 -o -path ./dir3 \) -prune -o -name '*.txt' -print

And, to exclude directories with a specific name at any level, use the -name primary instead of -path.

find . -type d -name node_modules -prune -o -name '*.json' -print

Solution 3:

I find the following easier to reason about than other proposed solutions:

find build -not \( -path build/external -prune \) -name \*.js
# you can also exclude multiple paths
find build -not \( -path build/external -prune \) -not \( -path build/blog -prune \) -name \*.js

Important Note: the paths you type after -path must exactly match what find would print without the exclusion. If this sentence confuses you just make sure to use full paths through out the whole command like this: find /full/path/ -not \( -path /full/path/exclude/this -prune \) .... See note [1] if you'd like a better understanding.

Inside \( and \) is an expression that will match exactly build/external (see important note above), and will, on success, avoid traversing anything below. This is then grouped as a single expression with the escaped parenthesis, and prefixed with -not which will make find skip anything that was matched by that expression.

One might ask if adding -not will not make all other files hidden by -prune reappear, and the answer is no. The way -prune works is that anything that, once it is reached, the files below that directory are permanently ignored.

This comes from an actual use case, where I needed to call yui-compressor on some files generated by wintersmith, but leave out other files that need to be sent as-is.

Note [1]: If you want to exclude /tmp/foo/bar and you run find like this "find /tmp \(..." then you must specify -path /tmp/foo/bar. If on the other hand you run find like this cd /tmp; find . \(... then you must specify -path ./foo/bar.

Solution 4:

There is clearly some confusion here as to what the preferred syntax for skipping a directory should be.

GNU Opinion

To ignore a directory and the files under it, use -prune

From the GNU find man page

Reasoning

-prune stops find from descending into a directory. Just specifying -not -path will still descend into the skipped directory, but -not -path will be false whenever find tests each file.

Issues with -prune

-prune does what it's intended to, but are still some things you have to take care of when using it.

1. find prints the pruned directory.

   • TRUE That's intended behavior, it just doesn't descend into it. To avoid printing the directory altogether, use a syntax that logically omits it.

2. -prune only works with -print and no other actions.

   • NOT TRUE. -prune works with any action except -delete. Why doesn't it work with delete? For -delete to work, find needs to traverse the directory in DFS order, since -deletewill first delete the leaves, then the parents of the leaves, etc... But for specifying -prune to make sense, find needs to hit a directory and stop descending it, which clearly makes no sense with -depth or -delete on.

Performance

I set up a simple test of the three top upvoted answers on this question (replaced -print with -exec bash -c 'echo $0' {} \; to show another action example). Results are below

----------------------------------------------
# of files/dirs in level one directories
.performance_test/prune_me     702702    
.performance_test/other        2         
----------------------------------------------

> find ".performance_test" -path ".performance_test/prune_me" -prune -o -exec bash -c 'echo "$0"' {} \;
.performance_test
.performance_test/other
.performance_test/other/foo
  [# of files] 3 [Runtime(ns)] 23513814

> find ".performance_test" -not \( -path ".performance_test/prune_me" -prune \) -exec bash -c 'echo "$0"' {} \;
.performance_test
.performance_test/other
.performance_test/other/foo
  [# of files] 3 [Runtime(ns)] 10670141

> find ".performance_test" -not -path ".performance_test/prune_me*" -exec bash -c 'echo "$0"' {} \;
.performance_test
.performance_test/other
.performance_test/other/foo
  [# of files] 3 [Runtime(ns)] 864843145

Conclusion

Both f10bit's syntax and Daniel C. Sobral's syntax took 10-25ms to run on average. GetFree's syntax, which doesn't use -prune, took 865ms. So, yes this is a rather extreme example, but if you care about run time and are doing anything remotely intensive you should use -prune.

Note Daniel C. Sobral's syntax performed the better of the two -prune syntaxes; but, I strongly suspect this is the result of some caching as switching the order in which the two ran resulted in the opposite result, while the non-prune version was always slowest.

Test Script

#!/bin/bash

dir='.performance_test'

setup() {
  mkdir "$dir" || exit 1
  mkdir -p "$dir/prune_me/a/b/c/d/e/f/g/h/i/j/k/l/m/n/o/p/q/r/s/t/u/w/x/y/z" \
    "$dir/other"

  find "$dir/prune_me" -depth -type d -exec mkdir '{}'/{A..Z} \;
  find "$dir/prune_me" -type d -exec touch '{}'/{1..1000} \;
  touch "$dir/other/foo"
}

cleanup() {
  rm -rf "$dir"
}

stats() {
  for file in "$dir"/*; do
    if [[ -d "$file" ]]; then
      count=$(find "$file" | wc -l)
      printf "%-30s %-10s\n" "$file" "$count"
    fi
  done
}

name1() {
  find "$dir" -path "$dir/prune_me" -prune -o -exec bash -c 'echo "$0"'  {} \;
}

name2() {
  find "$dir" -not \( -path "$dir/prune_me" -prune \) -exec bash -c 'echo "$0"' {} \;
}

name3() {
  find "$dir" -not -path "$dir/prune_me*" -exec bash -c 'echo "$0"' {} \;
}

printf "Setting up test files...\n\n"
setup
echo "----------------------------------------------"
echo "# of files/dirs in level one directories"
stats | sort -k 2 -n -r
echo "----------------------------------------------"

printf "\nRunning performance test...\n\n"

echo \> find \""$dir"\" -path \""$dir/prune_me"\" -prune -o -exec bash -c \'echo \"\$0\"\'  {} \\\;
name1
s=$(date +%s%N)
name1_num=$(name1 | wc -l)
e=$(date +%s%N)
name1_perf=$((e-s))
printf "  [# of files] $name1_num [Runtime(ns)] $name1_perf\n\n"

echo \> find \""$dir"\" -not \\\( -path \""$dir/prune_me"\" -prune \\\) -exec bash -c \'echo \"\$0\"\' {} \\\;
name2
s=$(date +%s%N)
name2_num=$(name2 | wc -l)
e=$(date +%s%N)
name2_perf=$((e-s))
printf "  [# of files] $name2_num [Runtime(ns)] $name2_perf\n\n"

echo \> find \""$dir"\" -not -path \""$dir/prune_me*"\" -exec bash -c \'echo \"\$0\"\' {} \\\;
name3
s=$(date +%s%N)
name3_num=$(name3 | wc -l)
e=$(date +%s%N)
name3_perf=$((e-s))
printf "  [# of files] $name3_num [Runtime(ns)] $name3_perf\n\n"

echo "Cleaning up test files..."
cleanup