recursive ls - Ignore files only in specific sub-directories

Question

So, ls -R is a command I use a lot to quickly verify the file structure of my projects, or check for any leftover files, as it's simple enough to use and read, and is part of coreutils (which means I don't need to install anything, as it's already available on all distros).

However, lately, my projects have grown more complex, having more and more files and sub-directories, and sub-directories of sub-directories, which I don't care about at all. Using ls -RI "PATTERN" used to work well, but now I'm facing an issue where there are files matching a specific pattern that I want to ignore in one specific directory, but not in the other, and vice versa.

The problem is that ls -I "PATTERN" seems to look for the shell PATTERN in name of the file, rather than its full path, which makes it impossible to do something like ls -R "subdir/file.a", as the name of the file I'm trying to ignore here is "file.a", and not "subdir/file.a" (which is its path).

In short, is there any way to make ls -I look for the pattern in the full path of the file, instead of only its name, or is there any coreutils equivalent command that would allow me to do that without losing much human readability?

---

Here's a quick example to illustrate my problem:

1. File structure:

   .
   ├── subdir1
   │  ├── file.a
   │  └── file.b
   └── subdir2
    ├── file.a
    └── file.b
   

2. What I want:

   .:
   subdir1 subdir2
   ./subdir1:
   file.a
   ./subdir2:
   file.b
   

3. What I get with ls -ARIsubdir1/file.b -Isubdir2/file.a:

   .:
   subdir1 subdir2
   ./subdir1:
   file.a file.b
   ./subdir2:
   file.a file.b
   

• Chris Davies: "is part of coreutils" - I think you simply mean is installed as standard
• rexkogitans: It is important to understand that ls * does not mean to pass * to ls, same goes with any other wildcard pattern. Instead, the shell first replaces * before passing the resulting arguments to ls (this is called globbing). Instead, ls '*' (with quotes) would pass *, but ls won't handle this, find is the right tool you are looking for.

Answer 1

While ls can't do it, find can. And while find isn't coreutils, it is generally available.

$ find . -path ./subdir1/file.b -o -path ./subdir2/file.a -o -print
.
./subdir2
./subdir2/file.b
./subdir1
./subdir1/file.a
$ 

In any case, you will probably want to pipe it through sort.

There are multiple techniques for ignoring things through find. You might be better off making a script to run find, then sort, then sed (or other tools) to delete things you don't want shown.

• Stéphane Chazelas: GNU coreutils and GNU findutils are just two bundles of basic command line software for the GNU system. coreutils merges what used to be sh-utils (formerly shellutils) fileutils and textutils, all of which include GNU implementations of standard POSIX utilities such as find as well as some more non-standard utilities (such as locate also found in findutils). In any case, GNU systems will have the find from GNU findutils like they'll have ls from GNU coreutils, or GNU sed or grep which are in separate packages. More importantly find like ls is a standard POSIX commands.
• David G.: @StéphaneChazelas right. but OP raised about coreutils, and liking that its already available. find (one of two tools in findutils) is much the same: available. Which I said without going into painful details.
• Stéphane Chazelas: (though ls -I or you not giving any path for find to find in are GNU extensions; changing find -path... to find . -path... would make it standard)
• Stéphane Chazelas: Yes, my initial comment was more for the OP who may be confused as to what "coreutils" is.
• Stéphane Chazelas: To get a sorted list of file paths, that would have to be something like find . -path ./subdir1/file.b -o -path ./subdir2/file.a -o -print0 | sort -z | tr '0円' '\n' (with sort -z being a GNU extension). Sorting the output of find -print would break file paths that contain newline characters.

Answer 2

With zsh:

set -o extendedglob
print -rC1 -- **/*~(subdir2/file.a|subdir1/file.b)(N)

Would print raw on 1 Column the sorted list of all the non-hidden files excluding subdir2/file.a and subdir1/file.b. To also exclude any file under subdir2/file.a and subdir1/file.b in case those are of type directory (like ls -RI file.a -I file.b would), you can change it to:

print -rC1 -- **/*~(subdir2/file.a|subdir1/file.b)(|/*)(N)

(though note that the ~ operator to exclude files applies after directory traversal, so zsh would still look into those directories recursively even if it wouldn't print the corresponding paths).

With the globstartshort option enabled, **/* can be shortened to **.

You can replace print -rC1 with ls -ld to get a long listing for those files.

A GNU find + GNU sort equivalent could look like:

LC_ALL=C find . -mindepth 1 '(' \
 -name '.?*' -o -path ./subdir2/file.a -o -path ./subdir1/file.b \
 ')' -prune -o -printf '%P0円' | sort -z | tr '0円' '\n'

Or to pass to ls -l:

LC_ALL=C find . -mindepth 1 '(' \
 -name '.?*' -o -path ./subdir2/file.a -o -path ./subdir1/file.b \
 ')' -prune -o -printf '%P0円' | sort -z | xargs -r0 ls -ld --