Shell Programming and Scripting

The weird self-generated test construct doesn't seem relevant to how you copy the files.

tar could be used, since it preserves full paths. How about this:

Code:

# Find all files named .xml under /path/to/src/, including things like
# /path/to/subdir/name.xml
find /path/to/src -iname '*.xml' |
# Feed them into tar.  It will archive them with full paths (excluding the leading /)
xargs tar -cf - |
# Extract them under /path/to/dest, including full paths.
tar -C /path/to/dest -xf -

Or if you just want to replicate the entire folder no matter what its contents:

Code:

tar -cf - /path/to/src | tar -C /path/to/dest -xf -

Beware that the leading slash exclusion isn't standard and possibly only provided by GNU tar.

I haven't used it in a while, but pax is probably better than tar for copying files while preserving their original hierarchy. Better in the sense that it's simpler to use since it can read the pathnames on stdin. Also, it can do the job with a single invocation, instead of requiring one instance to create an archive and another to extract it. For example, to move all jpg files while preserving their locations relative to each other:

Code:

find source -name '*.jpg' | pax -rw destination

If I'm not mistaken, the same goes for cpio (although with different options).

Another advantage is that the number of archive members is not limited by ARG_MAX. Even if you use xargs to invoke tar, you are still bound by that limit. Should xargs invoke tar more than once, it would be as if multiple archives had been concatenated and piped into the extracting tar process: cat 1.tar 2.tar | tar -xf - won't work (at least not with the implementations of tar that I use, which only succeed in untarring the first archive).

Availability, however, could be a potential drawback. While you can count on pax and cpio to be present on *BSD and Linux systems, I'm not certain if you can on the proprietary unices. Interestingly, of the three, pax is the only utility that POSIX attempts to standardize.

Regards,
Alister

pax is likely to be available on current proprietary Unix releases, precisely given the fact it is standardized. It is certainly on Solaris, HP-UX and AIX.

One less known feature of this less known utility is its ability to modify on the fly the file path using regular expressions (-s /old/new/[g]). This can be very useful in some situations.