fpart - Man Page

Sort and pack files into partitions

Synopsis

fpart[-h] [-V] -n num | -f files | -s size [-i infile] [-a] [-o outfile] [-0] [-e] [-P] [-v] [-l] [-b] [-y pattern] [-Y pattern] [-x pattern] [-X pattern] [-z] [-zz] [-zzz] [-Z] [-d depth] [-D] [-E] [-L] [-S] [-w cmd] [-W cmd] [-R cmd] [-p num] [-q num] [-r num] [FILE or DIR...]

Description

The fpart utility helps you sort file trees and pack them into bags (called "partitions").

General Options

-h, --help

Print help

-V, --version

Print version

Partition Control

-n num, --parts num

Create exactly num partitions and try to generate partitions with the same size and number of files. This option cannot be used in conjunction with -f, -s or -L.

-f files, --files files

Create partitions containing at most files files or directories. This option can be used in conjunction with -s and -L.

-s size, --size size

Create partitions with a maximum size of size bytes. With this option, a special partition 0 may be used to handle files that do not fit in a regular partition, given the provided size limit. This option can be used in conjunction with -f and -L. You can use a human-friendly unit suffix here (k, m, g, t, p).

Input Control

-i infile

Read file list from infile. If infile is “-”, then list is read from stdin.

-a, --arbitrary

Input contains arbitrary values; just sort them (do not crawl filesystem). Input must follow the “size(blank)path” scheme. This option is incompatible with crawling-related options.

Output Control

-o outfile

Output partitions' contents to outfile template. Multiple files will be generated given that template. Each outfile will get partition number as a suffix. If outfile is “-”, then partitions will be printed to stdout, with partition number used as a prefix (so you can grep partitions you are interested in, or do whatever you want).

-0

End filenames with a null (’\0’) character when using option -o.

-e

When adding directories (see Directory Handling ), add a trailing “/” to each directory entry.

-P

Add parent directories when closing partitions. That option can be used in conjunction with -zzz to produce partitions that can be synchronized in parallel with tools such as cpio(1), pax(1) or tar(1). Adding parent directories at the end of each partition ensures that modification times get reapplied to directories whatever the processing order of partitions is. Directories added that way are 0-sized and not subject to partition counters (for example, a file limit given using option -f will be exceeded by n parent directories). Also, they are not subject to inclusion/exclusion options ( -y, -Y, -x, -X) and they always have a trailing “/” even if option -e has not been used (this simplifies symlinks handling as we always want to add targets here, never the links themselves). That option may lead to creating duplicate directory entries when next partition begins with a directory entry that has already been added as a parent when closing the previous partition. Only intermediate partitions will get parent directories added, not the very last one which gets its parents through option -zzz when fts(3) crawling finishes. Requires live mode (option -L).

-v, --verbose

Verbose mode (may be specified more than once).

Filesystem Crawling Control

-l

Follow symbolic links (default: do not follow).

-b

Do not cross filesystem boundaries (default: cross).

-y pattern, --include pattern

Include files or directories matching pattern only (and discard all other files). This option may be specified several times. Pattern may be a leaf (file or directory) name or a specific path. Shell pattern matching characters (“[”, “]”, “*”, “?”) may be used. Include patterns are ignored when computing size of directories.

-Y pattern

Same as -y but case insensitive. This option may not be available on your platform (at least FreeBSD and GNU/Linux support it, Solaris does not).

-x pattern, --exclude pattern

Exclude files or directories matching pattern. This option can be used in conjunction with -y and -Y. In this case, exclusion is performed after. This option may be specified several times. Pattern may be a leaf (file or directory) name or a specific path. Shell pattern matching characters (“[”, “]”, “*”, “?”) may be used. Exclude patterns also apply when computing size of directories.

-X pattern

Same as -x but case insensitive. This option may not be available on your platform (at least FreeBSD and GNU/Linux support it, Solaris does not).

Directory Handling

-z

Pack empty directories. By default, fpart will pack files only (except when using the -d or -D options). This option can be useful for tools such as rsync(1) to be able to recreate a full file tree when used with fpart (e.g. using rsync's --files-from option). See the -zz option to also pack un-readable directories.

-zz

Treat un-readable or erroneous (partly-read) directories as empty, causing them to be packed anyway. Partly-read (non-empty) directories can end up being packed while some of their children have already been packed.

-zzz

Pack all directories (as empty). Useful when 3rd party tools need directory entries to update them (e.g. cpio(1), pax(1) or tar(1) ).

-Z

Pack un-readable/erroneous directories in dedicated partitions. This option helps isolating erroneous parts of a filesystem. Used in conjunction with FPART_PARTERRNO variable, hooks can try to handle or work around the error. Requires live mode (option -L) and option -zz ( or -zzz)

-d depth

After a certain depth, pack directories instead of files (directories themselves will be added to partitions, instead of their content). You can force a specific file to be packed anyway by listing it on the command line explicitly.

-D, --leaf-dirs

Implies -z. Pack leaf directories: if a directory contains files only, it will be packed as a single entry. You can force a specific file to be packed anyway by listing it on the command line explicitly.

-E, --dirs-only

Implies -D. Pack directories only (work on a per-directory basis): in that mode, no file will be packed. Instead, each directory will be packed as a single entry with a size being the sum of all top-level files' sizes. You can force a specific file to be packed anyway by listing it on the command line explicitly.

Live Mode

-L, --live

Live mode (default: disabled). When using this mode, partitions will be generated while crawling filesystem. This option saves time and memory but will never produce special partition 0 (see options -s and -S ). As a consequence, it will generate partitions slightly larger than the size specified with option -s. This option can be used in conjunction with options -f and -s, but not with option -n.

-S

Skip big files (default: disabled). In live mode, no special partition 0 can be produced and big files are added to the current partition as they are found while crawling the filesystem. That can lead to huge partitions. That option makes fpart skip files bigger than the specified maximum partition size (option -s ) and print them to stdout (even when using option -o ) as belonging to a pseudo-partition S (as in 'S'kipped). It allows a consumer to handle them immediately through a separate process (no fpart hook will be executed for skipped files). That option can only be used in Live mode (option -L ), when a maximum partition size has been given (option -s ).

-w cmd, --pre-part-cmd cmd

When using live mode, execute cmd when starting a new partition (before having opened next output file, if any). cmd is run in a specific environment that provides several variables describing the state of the program: FPART_HOOKTYPE ("pre-part", "post-part" or "post-run"), FPART_PARTFILENAME (current partition's output file name), FPART_PARTNUMBER (current partition number), FPART_PARTSIZE (current partition's size), FPART_TOTALSIZE (total partitions' size), FPART_PARTNUMFILES (number of files packed in current partition), FPART_TOTALNUMFILES (total number of files packed so far), FPART_PARTERRNO (0 if every single partition's entry has been read without error, else last erroneous entry's errno. For error detection to work properly, you may need to rebuild fpart using embedded fts(3) library, depending on the version shipped with your OS), FPART_PID (PID of fpart), FPART_TOTALNUMPARTS (total number of partitions generated so far). Variables may or may not be defined, depending on requested options and current partition's state when the hook is triggered. Hooks are executed in a synchronous way while crawling filesystem, so 1) avoid executing commands that take a long time to return as it slows down filesystem crawling and 2) do not presume cwd (PWD) is the one fpart has been started in, as it is regularly changed to speed up crawling (i.e. use absolute paths within hooks). Size-related variables are affected by preloading, overloading and rounding options.

-W cmd, --post-part-cmd cmd

Same as -w, but executes cmd when finishing a partition (after having closed last output file, if any).

-R cmd, --post-run-cmd cmd

Same as -w and -W but executes cmd just once before exiting. Note that only FPART_TOTALSIZE and FPART_TOTALNUMFILES environment variables are available in post-run hooks.

Size Handling

-p num

Preload each partition with num bytes. You can use a human-friendly unit suffix here (k, m, g, t, p).

-q num

Overload each file size with num bytes. You can use a human-friendly unit suffix here (k, m, g, t, p).

-r num

Round each file size up to next num bytes multiple. This option can be used in conjunction with overloading, which is done *before* rounding. You can use a human-friendly unit suffix here (k, m, g, t, p).

Examples

Here are some examples:

fpart -n 3 -o var-parts /var

Produce 3 partitions, with (hopefully) the same size and number of files. Three files: var-parts.1, var-parts.2 and var-parts.3 are generated as output.

fpart -s 4724464025 -o music-parts /path/to/music ./*.mp3

Produce partitions of 4.4 GB, containing music files from /path/to/music as well as MP3 files from current directory; with such a partition size, each partition content will be ready to be burnt to a DVD. Files music-parts.0 to music-parts.n, are generated as output.

find /usr ! -type d | fpart -f 10000 -i - /home | grep '^1 '

Produce partitions containing 10000 files each by examining /usr first and then /home and display only partition 1 on stdout.

du * | fpart -n 2 -a

Produce two partitions by using du(1) output. Fpart will not examine the file system but instead use arbitrary values printed by du(1) and sort them.

See Also

du(1), find(1), fpsync(1), grep(1), rsync(1)

Author, Availability

Fpart has been written by Ganaël LAPLANCHE and is available under the BSD license on http://contribs.martymac.org

Bugs

No bug known (yet).

Referenced By

fpsync(1).

November 18, 2011