In using btrfs-restore, I've found that much of the time the interface is a
poor match for what I'm trying to do. It works fine in the simplest cases
where you want to extract everything, but wanting to extract a subset of the
files in a directory is more difficult, and it gets more complicated from
there.
My solutions was to hack up (not very well) a btrfs-list-files program, feed
it to a Perl script to pick out the files I wanted and turn them into
(sanitized) regexes, and feed the results to btrfs-restore in a loop.
I think that, while my implementation of this was subpar, the idea deserves
wider consideration.
The idea is to split the restore utility into two components:
A 'find'-style listing utility
Supports various matching operations such as -name, -regex, etc.
Supports -print and -print0
Does not support -exec, -execdir, -delete, etc.
An extraction tool
Accepts a list of files in argv[] or on STDIN
STDIN supports both newline- and null-terminated operation
Does simple string matching
Extracts the named files
Benefits:
Simplicity:
The find utility only needs to care about walking dentries and
matching patterns
The extraction utility just needs to compare strings and extract
data on success
Usability:
Find has a reasonably well-known and well-understood interface
A simple extractor you just feed a list of files and directories to
is easier to grok for a new user.
Eliminate the annoying way the current interface requires splitting
up the path components of files into alternations
Efficiency:
The extractor can build up a trie of files to be matched
A near-arbitrary-size list of files can be extracted in two passes
(find names, extract files) without undue complexity, negating the need to
either build truly horrendous regexes or extract in a loop and take the
performance hit.
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
