OpenBSD find command -- exclude dot directory

[gustaf]

How do I tell the OpenBSD find(1) command to exclude the dot directory, but not its contents, from the list of directories and files it generates?

It seems like this should be possible with the not (!) operator, but after searching the Web, reading (and re-reading) the find(1) man page, and trying out numerous combinations of arguments and operators, I still have not found a solution.

Using -prune doesn't work for me in this situation; -prune will cause find to not descend into the pruned directory. If the pruned directory is the dot directory, the contents of the directory will be missing. Effectively, find will generate an empty list. This is obviously not the desired behavior.

I have found ways to exclude:

1. individual files plus the dot directory, or
2. individual files and named directories,

but not all three types (the dot directory, named directories, and individual files) in the same command.

I've written a sample script to demonstrate the problem because it's easier to see than on the command line.

If, for example, my home directory is the current directory and contains the following items:

./
../
.Xdefaults
.cache/
.profile
dirA/
dirA/dirB/
dirA/dirB/file1
dirC/
dirC/file2
file3.txt
file4.pdf

and I want the generated list to look like this:

.profile
dirA/
dirA/dirB
dirA/dirB/file1
file3.txt

I can run this script to test the find command -- the results are explained in the comments:

#!/bin/sh
# findtest

cd $HOME

# Method 1 works to exclude the dot directory and specified files; 
# contents of named directories are ignored by find and end up in the printed list.
METHOD1='( 
            ! -name .
            ! -path ./.Xdefaults
            ! -path ./.cache
            ! -path ./dirC
            ! -path ./file4.pdf
         )'
find . $METHOD1 -print | sort > method1.txt

# Method 2 works to exclude named directories and specified files;
# adding the dot directory to this command results in an empty list.
METHOD2='(
            -path ./.Xdefaults -or
            -path ./.cache -or
            -path ./dirC -or
            -path ./file4.pdf
         )'
find . $METHOD2 -prune -or -print | sort > method2.txt

#Method 3 should exclude the dot directory, named directories, and specified files.
## METHOD3=...
## find . $METHOD3...

Is this an impossible task for OpenBSD find or have I just not figured out the logic?

Any help would be appreciated.

[kl3]

To exclude "the dot directory" (I assume you mean either of ./ ../) in general, tell find to always go one step down the hierarchy:

 $ find . -mindepth 1
.Xdefaults
.cache/
.profile
dirA/
dirA/dirB/
dirA/dirB/file1
dirC/
dirC/file2
file3.txt
file4.pdf

Exlcuding hidden directories but not their content:

$ find . ! \( -type d -name .\* \)
.Xdefaults
.profile
dirA/
dirA/dirB/
dirA/dirB/file1
dirC/
dirC/file2
file3.txt
file4.pdf

In order to omit arbitrary files or directories, just add the respective expressions to filter them out:

$ find . ! \( -type f \( -name .Xdefaults -o -name file4.pdf \) \) ! \( -type d -name dirC -prune \)
./
../
.cache/
.profile
dirA/
dirA/dirB/
dirA/dirB/file1
file3.txt

Combining all examples (the second one covers the first one, but still), you'll eventually get your desired list:

$ #find . -mindepth 1 ! \( -type d -name .\* \) ! \( -type f \( -name .Xdefaults -o -name file4.pdf \) \) ! \( -type d -name dirC -prune \) # Group negations
$ #find . -mindepth 1 ! \( \( -type d -name .\* \) -o \( -type f \( -name .Xdefaults -o -name file4.pdf \) \) -o \( -type d -name dirC -prune \) \) # Group types
$ find . -mindepth 1 ! \( \( -type d \( -name .\* -o \( -name dirC -prune \) \) \) -o \( -type f \( -name .Xdefaults -o -name file4.pdf \) \) \)
.cache/
.profile
dirA/
dirA/dirB/
dirA/dirB/file1
file3.txt

I expect this to be what you wanted in the end:

1. Skip hidden directories themselves but not their content
2. Skip certain individual files
3. Skip certain directories entirely including their content

Node how the expressions regarding directories comes first to improve execution time. I'll leave the Why? up to you, the OPERATORS section in find(1) might be helpful.

[gustaf]

After more searching and trial-and-error, I finally figured it out. I also realized that the original question didn't completely address the problem. I want to tell find(1) to generate a list of files and directories that:

• includes everything in the current directory

• EXCEPT

• the directories and files specified in the command AND

• the ./ directory

(The original post has the example directory to be searched and the search results.)

The find utility doesn't have a proper exclude argument for nested directories, so I had to use -prune with -path and -or to exclude directories and their contents:

find . \
     -path ./.Xdefaults -prune -or \
     -path ./.cache -prune -or \
     -path ./dirC -prune -or \
      ! -name file4.pdf -and \
      ! -name . \
     -print

Once I unlocked the mysteries of -prune, the problem became a lot less insurmountable:

According to the man page, find is by default recursive: "...find recursively descends the directory tree for each path listed..." (find(1), DESCRIPTION).

Using -prune "...causes find to not descend into the current file..." (find(1), PRIMARIES). That explains what -prune DOESN'T do, but what does -prune actually DO?

Re-phrased as a positive statement, -prune TURNS OFF RECURSION. It's that simple. When find encounters a pruned directory, it will skip over the entire contents of that directory and move on to the next top-level directory in the tree.

Excluding both the top-level directory and the directory's contents involves taking advantage of the side-effect of -prune combined with -or.

The GNU documentation for find (— Action: -prune) gives a detailed explanation of -prune:

If the file is a directory, do not descend into it. The result is true. For example, to skip the directory src/emacs and all files and directories under it, and print the names of the other files found:

find . -wholename './src/emacs' -prune -o -print

The above command will not print ./src/emacs among its list of results. This however is not due to the effect of the '-prune' action (which only prevents further descent, it doesn't make sure we ignore that item). Instead, this effect is due to the use of '-o'. Since the left hand side of the 'or' condition has succeeded for ./src/emacs, it is not necessary to evaluate the right-hand-side ('-print') at all for this particular file...

Greg's Wiki also has a good explanation of why -prune is confusing and counterintuitive.

The POSIX documentation for find provides some additional examples.