git-annex tries to ensure that the configured number of copies of your data always exist, and leaves it up to you to use commands like git annex get and git annex drop to move the content to the repositories you want to contain it. But often, it can be good to have more fine-grained control over which content is wanted by which repositories. Configuring this allows the git-annex assistant as well as git annex get --auto, git annex drop --auto, git annex sync --content, etc to do smarter things.

Preferred content settings can be edited using git annex vicfg, or viewed and set at the command line with git annex wanted. Each repository can have its own settings, and other repositories will try to honor those settings when interacting with it. (So there's no local .git/config for preferred content settings.)


Rather than writing your own preferred content expression, you can use several standard ones included in git-annex that are tuned to cover different common use cases.

You do this by putting a repository in a group, and simply setting its preferred content to "standard" to match whatever is standard for that group. See standard groups for a list.

The idea is that you write an expression that files are matched against. If a file matches, the repository wants to store its content. If it doesn't, the repository wants to drop its content (if there are enough copies elsewhere to allow removing it).

To check at the command line which files are matched by preferred content settings, you can use the --want-get and --want-drop options.

For example, git annex find --want-get --not --in . will find all the files that git annex get --auto will want to get, and git annex find --want-drop --in . will find all the files that git annex drop --auto will want to drop.

The expressions are very similar to the matching options documented on the git-annex man page. At the command line, you can use those options in commands like this:

git annex get --include='*.mp3' --and -'(' --not --largerthan=100mb -')'

The equivalent preferred content expression looks like this:

include=*.mp3 and (not largerthan=100mb)

So, just remove the dashes, basically. However, there are some differences from the command line options to keep in mind:

difference: file matching

While --include and --exclude match files relative to the current directory, preferred content expressions always match files relative to the top of the git repository. Perhaps you put files into archive directories when you're done with them. Then you could configure your laptop to prefer to not retain those files, like this:


difference: no "in="

Preferred content expressions have no direct equivalent to --in.

Often, it's best to add repositories to groups, and match against the groups in a preferred content expression. So rather than --in=usbdrive, put all the USB drives into a "transfer" group, and use "copies=transfer:1"

difference: dropping

To decide if content should be dropped, git-annex evaluates the preferred content expression under the assumption that the content has already been dropped. If the content would not be wanted then, the drop can be done. So, for example, copies=2 in a preferred content expression lets content be dropped only when there are currently 3 copies of it, including the repo it's being dropped from. This is different than running git annex drop --copies=2, which will drop files that currently have 2 copies.

difference: "present"

There's a special "present" keyword you can use in a preferred content expression. This means that content is wanted if it's present, and not otherwise. This leaves it up to you to use git-annex manually to move content around. You can use this to avoid preferred content settings from affecting a subdirectory. For example:

auto/* or (include=ad-hoc/* and present)

Note that not present is a very bad thing to put in a preferred content expression. It'll make it want to get content that's not present, and drop content that is present! Don't go there..

difference: "inpreferreddir"

There's a special "inpreferreddir" keyword you can use in a preferred content expression of a special remote. This means that the content is preferred if it's in a directory (located anywhere in the tree) with a special name.

The name of the directory can be configured using git annex enableremote $remote preferreddir=$dirname

(If no directory name is configured, it uses "public" by default.)

difference: "standard"

git-annex comes with some built-in preferred content expressions, that can be used with repositories that are in some standard groups.

When a repository is in exactly one such group, you can use the "standard" keyword in its preferred content expression, to match whatever content the group's expression matches. (If a repository is put into multiple standard groups, "standard" will match anything.. so don't do that!)

Most often, the whole preferred content expression is simply "standard". But, you can do more complicated things, for example: "standard or include=otherdir/*"

difference: "groupwanted"

The "groupwanted" keyword can be used to refer to a preferred content expression that is associated with a group. This is like the "standard" keyword, but you can configure the preferred content expressions using git annex groupwanted.

Note that when writing a groupwanted preferred content expression, you can use all of the keywords listed above, including "standard". (But not "groupwanted".)

For example, to make a variant of the standard client preferred content expression that does not want files in the "out" directory, you could run: git annex groupwanted client "standard and exclude=out/*"

Then repositories that are in the client group and have their preferred content expression set to "groupwanted" will use that, while other client repositories that have their preferred content expression set to "standard" will use the standard expression.

Or, you could make a new group, with your own custom preferred content expression tuned for your needs, and every repository you put in this group and make its preferred content be "groupwanted" will use it.

For example, the archive group only wants to archive 1 copy of each file, spread among every repository in the group. Here's how to configure a group named redundantarchive, that instead wants to contain 3 copies of each file:

git annex groupwanted redundantarchive "not (copies=redundantarchive:3)"
for repo in foo bar baz; do
    git annex group $repo redundantarchive
    git annex wanted $repo groupwanted

difference: metadata matching


git annex get --metadata tag=done



difference: unused

The --unused option makes git-annex operate on every key that git annex unused has determined to be unused. The corresponding unused keyword in a preferred content expression also matches those keys.

However, the latter doesn't make git-annex consider those keys. So when git-annex is only checking preferred content expressions against files in the repository (which are obviously used), unused in a preferred content expression won't match anything.

So when is unused useful in a preferred content expression?

The git-annex assistant periodically scans for unused files, and moves them to some repository whose preferred content expression matches "unused". (Or, if annex.expireunused is set, it may just delete them.)


It's important that all clones of a repository can understand one-another's preferred content expressions, especially when using the git-annex assistant. So using newly added keywords can cause a problem if an older version of git-annex is in use elsewhere.

Before git-annex version 5.20140320, when git-annex saw a keyword it did not understand, it defaulted to assuming all files were preferred content. From version 5.20140320, git-annex has a nicer fallback behavior: When it is unable to parse a preferred content expression, it assumes all files that are currently present are preferred content.

Here are recent changes to preferred content expressions, and the version they were added in.

  • "standard" 5.20140314
    (only when used in a more complicated expression; "standard" by itself has been supported for a long time)
  • "groupwanted=" 5.20140314
  • "metadata=" 5.20140221
  • "lackingcopies=", "approxlackingcopies=", "unused=" 5.20140127
  • "inpreferreddir=" 4.20130501

How does the preferred content settings interfere with the numcopies setting?

I could not get behind it. E.g. a case I do not unterstand:

I have a preferred setting evaluating to true and still

git annex get --auto

does nothing, if the number of copies produced would surpass the numcopies setting.


Comment by Sehr Wed Dec 5 20:41:26 2012
Yeah, that didn't make sense. I've fixed it, so it gets files if needed for either numcopies or preferred content.
Comment by Thu Dec 6 17:24:29 2012

Built a new copy of git-annex yesterday. I have a "client" on my macbook, and two "backup"s, one on an external HD, one on an ssh git remote.

git annex get --auto works beautifully!

It doesn't seem to work for copying content to a place where it's needed, though.

If I drop a file from my "backup" USB drive, and then go back to my macbook and do a "git annex sync" and "git annex copy --to=usbdrive --auto" it does not send the file out to the USB drive, even though by preferred content settings, the USB drive should "want" the file because it's a backup drive and it wants all content.

Similarly, if I add a new file on my macbook and then do a "git annex copy --to=usbdrive auto" it does not get copied to the USB drive.

Is this missing functionality, or should the preferred content setting for remotes only affect the assistant?

Comment by edheil [] Fri Dec 7 20:24:18 2012
It was a bug in the backup group's preferred content pagespec, introduced by the changes I made to fix the previous problem. Now fixed.
Comment by Mon Dec 10 19:46:01 2012
Comment by edheil [] Tue Dec 11 16:03:04 2012

Is there a way to change these definitions for a given annex?

ie: in this repo make "client" mean

present and exclude=*/archive/* and exclude=archive/*
Comment by Andrew Thu Jan 10 03:00:52 2013
Sorry, there's not. The expressions used for "standard" are built in.
Comment by Thu Jan 10 03:51:38 2013

By way of a feature request: Maybe the way to do this is to have an additional keyword like "config" or "repo" that allows you to use vicfg and/or git config to set alternative rules and even additional group names.

In git config:

annex.groups.<groupname> = present and exclude=*/archive/* and exclude=archive/*

in vicfg:

# (for passport)
#trust A0637025-ED47-4F95-A887-346121F1B4A0 = semitrusted

# (for passport)
group A0637025-ED47-4F95-A887-346121F1B4A0 = transfer

# (for passport)
preferred-content A0637025-ED47-4F95-A887-346121F1B4A0 = repo

# (for transfer)
group-content transfer = present and exclude=*/archive/* and exclude=archive/*
Comment by Andrew Thu Jan 10 10:24:28 2013

Is there a way to drop only the files that are located in an "archive" directory? I want to drop all files when calling

git annex drop --auto

if I move them to the archive. But I want to keep the files that are outside of the archive, even if they are already present in other repos. As far as I have seen and tested, as soon as I have the files in an other repo all files get dropped, also the ones outside the archive directory. Or do I have to increase "numcopies" in order to circumvent the "(not copies=semitrusted+:1)" case?

Comment by Georg Tue Oct 8 17:53:08 2013
Georg, drop --auto will only drop files that are not preferred content. I'd need to know what preferred content expression you're using to say more.
Comment by Fri Oct 11 17:12:00 2013