How to recursively copy a directory using cpio?

Question

How do I copy recursively a directory within the list at my-files-to-copy.txt?

cd my/source/dir
cpio --create < my-files-to-copy.txt | (cd my/dest/dir && cpio --extract)

my-files-to-copy.txt

src/main.js
src/routes/index.js
src/my-app/ # a directory with it's own child tree I want to recursively copy

Answer 1

I'm afraid cpio can't really do this, unless you do some extra scripting to detect the directory pathname and walk through it to extract the pathnames of all files below it, as a way of "pre-processing" your list file.

It would be far easier to use another tool, such as rsync (note the extra -r after --files-from!):

rsync -a --files-from=my-files-to-copy.txt -r my/source/dir my/dest/dir

... or pax (a standard utility):

( cd my/source/dir && pax -w ) <my-files-to-copy.txt |
( cd my/dest/dir   && pax -r )

... or with pax if you can give an absolute path to the destination directory:

( cd my/source/dir && pax -rw /my/dest/dir ) <my-files-to-copy.txt

... or BSD tar (use -T or --files-from= instead of -I with GNU tar):

tar -c -f - -C my/source/dir -I my-files-to-copy.txt |
tar -x -f - -C my/dest/dir

This is all assuming that my-files-to-copy.txt lives in the current directory and that this directory might be different from the my/source/dir directory.

Answer 2

As mentioned before in the comments, cpio supports what you like. Use e.g.:

find src/main.js src/routes/index.js src/my-app/ | cpio -pdm my/dest/dir

The pitfall of this method is that you need to run two processes in a way that does not speed up things since find needs to stat(2) all files and cpio also needs to stat(2) all file some time later and cpio only uses a single process to do the actual copy.

...or you could do it the most efficient way using my own tool, star (the oldest free tar implementation):

star -copy -p -find src/main.js src/routes/index.js src/my-app/ my/dest/dir

This is efficient because star uses libfind in the star process, only needs to call stat(2) once per file and forks a second process that creates the new files concurrently to the main star process that reads the files.

In order to avoid a slow down from this method, both star processes share the file content and file meta data via shared memory.

BTW: you may of course replace the list of files from my examples by something like:

` cat my-file-list.txt `