I have a directory (Directory A) with files.
Directories B and C contain files from directory A (no files in B are in C and vice versa).
How can I list files in Directory A that are not present in Directories B nor C?
Asked
Active
Viewed 38 times
0
-
What kind of volume? Are we dealing with tens of thousands of files here, or only a few tens/hundreds? – Charles Duffy Sep 05 '18 at 15:31
-
..and are these recursive directories? (Does Directory A have subdirectories containing files, etc.?) – Ben DeVries Sep 05 '18 at 15:35
-
Also, which platform? Do you have GNU `find`? Are you trying to support all POSIX-compliant platforms? – Charles Duffy Sep 05 '18 at 15:36
-
@CharlesDuffy Platform is Linux or Windows. I prefer Linux answers but can do it either way. We're dealing with about 100K files in total. Additionally, all files are PDFs. – Robco Sep 05 '18 at 16:12
-
@BenDeVries - These are non-recursive directories. – Robco Sep 05 '18 at 16:14
-
100k files is enough that the `comm` approach will be considerably faster than the native-bash one. – Charles Duffy Sep 05 '18 at 17:10
1 Answers
3
If you don't need to be fast in very-high-volume cases:
for f_path in a/*; do f=${f_path#a/}
[[ -e "b/$f" || -e "c/$f" ]] && continue
printf '%s\n' "$f"
done
If you do, and have GNU comm, find and sort, see the following -- of course, replace the tr at the end with code that actually reads a NUL-delimited list properly if you want to be able to safely handle all possible filenames:
comm -z23 <(find a -maxdepth 1 -printf '%P\0' | sort -z) \
<(find b c -maxdepth 1 -printf '%P\0' | sort -z) \
| tr '\0' '\n'
For more on this use of comm, see BashFAQ #36.
Charles Duffy
- 280,126
- 43
- 390
- 441
-
1With the first solution (globbing), use `shopt -s nullglob` to handle an empty `a` directory correctly. Also, use `shopt -s dotglob` if the files to be checked may have names that begin with a dot. – pjh Sep 05 '18 at 16:14