A backup rotation filter for the Unix shell

&gt; <i>To list backups that will should be kept use the --invert option.</i><p>May I suggest changing the name to something more direct? Calling it &quot;--invert&quot; means the user must think through what the default sense of the test is, then negate that in their mind. Not exactly a tough mental task, but people do make careless errors.<p>Perhaps something like &quot;--list=keep&quot; and &quot;--list=discard&quot; (with the default being &quot;discard&quot;).<p>Also, typos:<p>&quot;will make prunef to keep&quot; --&gt; &quot;will make prunef keep&quot;<p>&quot;list backups that will should be&quot; --&gt; &quot;list backups that should be&quot;

Handy looking tool:)<p>Meta: I&#x27;m excited to see sr.ht starting to pop up in the wild like this:) I hope this is part of it starting to take off.

I use rotate-backups with temporary files and your script is an interesting alternative.<p><a href="https:&#x2F;&#x2F;pypi.org&#x2F;project&#x2F;rotate-backups&#x2F;" rel="nofollow">https:&#x2F;&#x2F;pypi.org&#x2F;project&#x2F;rotate-backups&#x2F;</a>

FYI the interface format seems to be inspired by borg-prune[1].<p>[1] <a href="https:&#x2F;&#x2F;borgbackup.readthedocs.io&#x2F;en&#x2F;stable&#x2F;usage&#x2F;prune.html" rel="nofollow">https:&#x2F;&#x2F;borgbackup.readthedocs.io&#x2F;en&#x2F;stable&#x2F;usage&#x2F;prune.html</a>

The original title reads &quot;... for your Unix shell&quot;, suggesting that the code is portable to many shells. I have not verified that claim.<p>&quot;... for the Unix shell&quot; makes little sense. When I used Unix my shell was csh, later tcsh. Nowadays my shell is bash in most cases and dash in some more limited environments. Either case, &quot;the Unix shell&quot; does not exist.

Did something similar but employed a Fibonacci sequence, using the hour as the unit (but could be a minute, or a second), to groom a collection of snapshots maintaining a sensible timeline: <a href="https:&#x2F;&#x2F;epxx.co&#x2F;logbook&#x2F;entries&#x2F;fibo_en.html" rel="nofollow">https:&#x2F;&#x2F;epxx.co&#x2F;logbook&#x2F;entries&#x2F;fibo_en.html</a>

If anybody else is using a similar method, I think it&#x27;s also good to encrypt the backups:<p><pre><code> tar -zcv &lt;SRC_BACKUP&gt; | gpg -c --batch --passphrase &lt;PASSWORD&gt; -o &lt;DEST_BACKUP&gt;.gz.gpg </code></pre> In my experience the encryption part doesn&#x27;t add any extra time on a modern machine, with the spinning disk being the slowest cog.

This is awesome. I currently use duplicity for backups to BackBlaze, but my DB backups I just place a files and I&#x27;ve just been pruning old one&#x27;s manually for a while. Something like this, that lets me specify the file format, would be perfect!

Could you leverage information in zfs snapshots metadata to drive things like this? The copy on write semantics inherent seem plausibly related.<p>(Zfs snaps make awesome backups too, but unlike tar are inherently tied to zfs)

Similar concept, in Ruby, acting as a filter in a pipeline, by me:<p><a href="https:&#x2F;&#x2F;github.com&#x2F;francois&#x2F;surrender" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;francois&#x2F;surrender</a>

PSA: Be sure to test every backup completely before succeeding a backup job, or it&#x27;s not a backup.

tarsnap desperately needs something like this included with it.