Add as copy?

[Thomas Singer via dev <de...@subversion.apache.org>]

Hi all,

We use our SVN repository mostly for binary files and it happens quite 
often to have the same file contents at different paths/names 
(intentionally). We do not use SVN commands to create copies, but just 
copy them using normal file operations.

Is there a command that compares the hash of the file to add with the 
ones of existing tracked files and instead of adding a new file, just 
marking it as a copy of another?

-- 
Thomas

Re: Add as copy?

[Thomas Singer via dev <de...@subversion.apache.org>]

Hi Stefan,

Thanks. OK, that's even better.

--
Thomas


On Wed, 08-Jul-26 11:42, Stefan Sperling wrote:
> On Wed, Jul 08, 2026 at 11:13:18AM +0200, Thomas Singer via dev wrote:
>> Hi all,
>>
>> We use our SVN repository mostly for binary files and it happens quite often
>> to have the same file contents at different paths/names (intentionally). We
>> do not use SVN commands to create copies, but just copy them using normal
>> file operations.
>>
>> Is there a command that compares the hash of the file to add with the ones
>> of existing tracked files and instead of adding a new file, just marking it
>> as a copy of another?
> 
> De-duplication of file content is already handled internally in two ways:
> 
> There will only be one copy of the file's content in the working copy's
> pristine store (where content is indexed by SHA1 checksum rather than by
> path), and only one copy of the file's content in the repository (unless
> you have explicitly disabled representation sharing in the fsfs.conf file
> on the server).
> 
> However, there will be multiple on-disk copies of file content in the
> working copy since Subversion was designed to enable independent
> modifications at each path in the working copy. If you need de-duplication
> of the working files as well then consider storing your working copies on
> a filesystem which supports de-duplication and copy-on-write internally,
> if your operating system offers such a feature. See this fairly recent
> discussion thread for some hints: https://marc.info/?t=178057784400005&r=1&w=2

Re: Add as copy?

[Stefan Sperling <st...@apache.org>]

On Wed, Jul 08, 2026 at 11:13:18AM +0200, Thomas Singer via dev wrote:
> Hi all,
> 
> We use our SVN repository mostly for binary files and it happens quite often
> to have the same file contents at different paths/names (intentionally). We
> do not use SVN commands to create copies, but just copy them using normal
> file operations.
> 
> Is there a command that compares the hash of the file to add with the ones
> of existing tracked files and instead of adding a new file, just marking it
> as a copy of another?

De-duplication of file content is already handled internally in two ways:

There will only be one copy of the file's content in the working copy's
pristine store (where content is indexed by SHA1 checksum rather than by
path), and only one copy of the file's content in the repository (unless
you have explicitly disabled representation sharing in the fsfs.conf file
on the server).

However, there will be multiple on-disk copies of file content in the
working copy since Subversion was designed to enable independent
modifications at each path in the working copy. If you need de-duplication
of the working files as well then consider storing your working copies on
a filesystem which supports de-duplication and copy-on-write internally,
if your operating system offers such a feature. See this fairly recent
discussion thread for some hints: https://marc.info/?t=178057784400005&r=1&w=2