How can I check if a directory contains the same files of a TAR archive?

Question

Let's say I have a folder Documents and a TAR file Documents.tar, how to check if the tar file contains the same files that are present in the directory?

The more obvious solution to me would be to do:

$ tar xvf Documents.tar -C untarDocs
$ diff -r Documents untarDocs

Unfortunately this is very slow for large TAR files, is there any other alternative?

Using tar -dvf Documents.tar (or --diff, --compare) doesn't work because it doesn't detect a file that is present in the filesystem but not in the TAR file, it just detects a file present in the TAR file but not in the filesystem e.g.:

$ mkdir new
$ touch new/foo{1..4}
$ tar cvf new.tar new/
$ touch new/bar
$ tar --diff --verbose --file=new.tar       #### doesn't detect new/bar #########
$ rm new/foo1
$ tar --diff --verbose --file=new.tar

Output

new/
new/foo2
new/foo3
new/foo4
new/foo1
tar: new/foo1: Warning: Cannot stat: No such file or directory   ### works ###

Answer 1

If you want only to compare lists of file- and directory-names, the -d option is not helpful. Instead, diff'ing sorted lists from find and tar -tf would do that.

Starting with the names assumed in OP's original example:

$ tar xvf Documents.tar -C untarDocs
$ diff -r Documents untarDocs

here is a suggested script to diff the filenames:

#!/bin/sh
MYDIR=$(mktemp -d)
tar tf Documents.tar |sort >$MYDIR/from-tar
find Documents |sort >$MYDIR/from-dir
(cd $MYDIR && diff -r from-tar from-dir)
rm -rf $MYDIR

This assumes that Documents.tar contains the same top-level "Documents" directory. If that is not a good assumption, then the lists should be filtered to remove the name of the top-level directory. OP did not indicate that this would be a problem, however.

In any case, the lists must be sorted, because there is no guarantee which order the tar and find programs use.

I used mktemp because of the clue that OP is using GNU tar (the -d option), which makes it likely on Linux.

There is of course no POSIX tar for reference with regard to -d. pax does not do diff's either.

Answer 2

With a standard tar archive, this is not possible. This is because a standard tar archive does not include the list of directory contents.

With the star enhacements that are present when star is told to archive everything that is needed for incremental backups, it works:

1) star -c -dump -C somedir . > /tmp/tarfile

2) star -diff -vv -C comparedir < /tmp/tarfile

You may reduce the amount of metadata that is compare using the diffopts= option, see man page.