How to Do this List Comparison with Find?

Question

I do and get a list of files where I would like to delete many duplicate backup files

find . -type f -name '._*' 

I would like to find those files which have a corresponding filename

• /home/masi/._test.tex matches /home/masi/test.tex
• /home/masi/math/lorem.png matches /home/masi/math/._lorem.png

Pseudocode about files wanted to be saved filename which has corresponding ._filename but also save filename without ._filename

find . -type f -name '._*' -exec \ 
   find filenameWithoutDotUnderscore, if yes, print the filename

Pseudocode 2 clarification about files wanted to be removed = ._filename if there is a corresponding filename

• If there is filename and ._filename in the same directory, print ._filename such that I can remove the duplicate = ._filename.
• Exclude filenamePart1_.filenamePart2, bok_3A.pdf, ... in ._filename.
• Do not remove ._filename if there is no corresponding filename in the same directory.

Reviewing Wildcard's command

I do find . -type f -name '._*' -exec sh -c 'for a; do f="${a%/*}/${a##*/._}"; [ -e "$f" ] && printf "rm -- %s\n" "$a"; done' find-sh {} + but it returns too many files. I think I need more && conditions beside the existence check ([ -e "$f" ]). It would be great to get here some content comparison and lastly diff if suspicion of much difference.

Systems: Ubuntu 16.04 and Debian 8.25
Bash: 4.3.42
Find: 4.7.0

Answer 1

You can do this with find, but to do it robustly you will need to embed a shell one-liner as well. The proper way to do this is one of the following:

Stuff the looping into the spawned shell:

find . -type f -name '._*' -exec sh -c 'for a in "$@"; do f="${a%/*}/${a##*/._}"; [ -e "$f" ] && printf %s\\n "$f"; done' find-sh {} +

Or, spawn a separate shell for each file to be tested (less efficient, potentially more readable):

find . -type f -name '._*' -exec sh -c 'f="${1%/*}/${1##*/._}"; [ -e "$f" ] && printf %s\\n "$f"' find-sh {} \;

---

To directly remove the backup files, change this to the following for a dry run:

find . -type f -name '._*' -exec sh -c 'for a; do f="${a%/*}/${a##*/._}"; [ -e "$f" ] && printf "rm -- %s\n" "$a"; done' find-sh {} +

Then once you're satisfied with the list of commands that gets printed, use:

find . -type f -name '._*' -exec sh -c 'for a; do f="${a%/*}/${a##*/._}"; [ -e "$f" ] && rm -- "$a"; done' find-sh {} +

---

Notes:

In all of these, the find-sh argument is an arbitrary string; you could put anything there. It gets set as $0 within the spawned shell and is used for error reporting.

for a in "$@"; do is exactly equivalent to for a; do.

printf is better than echo.

Quoting is important.

Answer 2

I made an a.sh (because I don't know how to do this as a function) which is just one line:

test -f "${*/._/}" && echo "$* matches ${*/._/}";

That is, I'm testing for the existence of a file whose name is all the arguments passed into the shell script, with the first instance of ._ removed (technically a search-and-replace with nothing in the 'replace' field). When it finds such a file it echoes the name of the two matching files.

Then I ran it with find ... -exec ./a.sh '{}' \;

About search (and replace) in bash: ${foo/bar/baz} means search the variable foo for the string bar and replace it with baz. You can try this experiment: x=abcdef; echo ${x/c/C}. So the code above is searching for ._ and replacing it with nothing. For more details, in man bash search for Pattern substitution.