Re: backup-each-save.el v1.1

Re: backup-each-save.el v1.1

[Kin Cho]

Actually I also wrote some code to backup every file I save, but
to a single directory (with a timestamp appended to each
filename).  The backup directory would fill up with a thousand
files in a month, then I would manually (should write some elisp
to automate eventually) archive (and delete) the oldest 500 to a
single zip file.

I've been doing this for probably 4-5 years, and I have to go
back to this backup directory to recover files quite a few times.
Also I never had trouble finding the right file to recover, even
though the directory structure is not mirrored.

-kin

Benjamin Rutt <rutt.4+news@osu.edu> writes:

> Kin Cho <ignore-this-prefixkin@techie.com> writes:
>
>> I suggest two user specified options to auto purge older
>> versions:
>>
>> A) purge by number of versions
>> B) purge by age of versions
>
> Because of the user-specified timestamping in the current version, I'd
> have no idea how to manage the backup mirror tree, where filenames
> could be named arbitrarily.
>
> I suppose one way to fix this is to specify the suffix that is
> appended to the filename (taking away power to do user-specified
> filename timestamping), using some kind of ~1~, ~2~ type counter like
> the regular backup files.  However, that will scale O(n) as the # of
> backup files increases, since I'll have to look for files with suffix
> ~1~ .. ~n~ to clean up.
>
> Another approach would be to use some kind of registry/database where
> I kept track of which filenames that have been copied.  Is there
> anything convenient to use along these lines in emacs?
>
> I welcome discussions of either approach.  Followup set to
> gnu.emacs.help.
> -- 
> Benjamin Rutt

Re: backup-each-save.el v1.1

[Benjamin Rutt]

Kin Cho <ignore-this-prefixkin@techie.com> writes:

> Actually I also wrote some code to backup every file I save, but
> to a single directory (with a timestamp appended to each
> filename).  The backup directory would fill up with a thousand
> files in a month, then I would manually (should write some elisp
> to automate eventually) archive (and delete) the oldest 500 to a
> single zip file.

I think it is OK to have some other means of managing these backup
files.  For me, I don't mind having oodles of backup files sitting
around, since disk is cheap nowadays and most of the files I save are
relatively small (with the exception of e.g. .newsrc.eld which I took
care of in v.1.1 with a user specified filter).  Also, making a
tarball would probably compress really well with all the inter-file
similarities (at least that should be true if you make a complete .tar
file first).  I don't think the .zip format builds a compression table
across files, but rather compresses each individual file separately.

> I've been doing this for probably 4-5 years, and I have to go
> back to this backup directory to recover files quite a few times.
> Also I never had trouble finding the right file to recover, even
> though the directory structure is not mirrored.

I think that is probably fine, but the extra structure sure won't
hurt, especially if you save a lot of files with the same name
(e.g. Makefile).
-- 
Benjamin Rutt

Re: backup-each-save.el v1.1

[Kin Cho]

Benjamin Rutt <rutt.4+news@osu.edu> writes:

> Kin Cho <ignore-this-prefixkin@techie.com> writes:
>
>> Actually I also wrote some code to backup every file I save, but
>> to a single directory (with a timestamp appended to each
>> filename).  The backup directory would fill up with a thousand
>> files in a month, then I would manually (should write some elisp
>> to automate eventually) archive (and delete) the oldest 500 to a
>> single zip file.
>
> I think it is OK to have some other means of managing these backup
> files.  For me, I don't mind having oodles of backup files sitting
> around, since disk is cheap nowadays and most of the files I save are
> relatively small (with the exception of e.g. .newsrc.eld which I took
> care of in v.1.1 with a user specified filter).  Also, making a
> tarball would probably compress really well with all the inter-file
> similarities (at least that should be true if you make a complete .tar
> file first).  I don't think the .zip format builds a compression table
> across files, but rather compresses each individual file separately.

True, tgz would compress very well.  However, I like zip because
of archive-mode support.

>> I've been doing this for probably 4-5 years, and I have to go
>> back to this backup directory to recover files quite a few times.
>> Also I never had trouble finding the right file to recover, even
>> though the directory structure is not mirrored.
>
> I think that is probably fine, but the extra structure sure won't
> hurt, especially if you save a lot of files with the same name
> (e.g. Makefile).

Consider using find(1) (with -ctime option) to purge old files
from your mirrored directory structure.

-kin