Batch mode scenario ("use case")

[Suresh Govindachar]

Matt wrote: 
>Suresh Govindachar wrote:
>
>> How about the following scheme?
>>
>>   open /e/cmds/foo
>>   <do till end-of-file>:
>>       <begin-continue>
>>
>>           discard lines till there is a match to some/path/projects/c_a
>>
>>           keep lines till there is a match to a directory outside c_a
>>                     Note that the lines being kept could be
>>                     data/commands corresponding to files inside c_a
>>
>>       <end-continue>
>>   <end-do>
>>   write kept lines to new batch file
>
> This is a clever idea, but I think it would be very difficult to
> subset the batch file correctly since the rsync protocol is quite
> complex (and binary).  For one thing, the new batch file would
> need to contain the protocol initialization and at least the
> relevant portion of the file list.  You might be able to
> accomplish this by keeping everything until the first file
> transfer is logged, but that won't work if the sender used
> incremental recursion because the relevant portion of the file
> list will be randomly interspersed with file transfers.
> 
> If you need to restore a subtree from a batch file, another thing
> you could do is make a modified copy of rsync just for the job
> with a strncmp thrown in to skip all file transfers except the
> ones in the subtree.  Also, instead of a batch file, you might
> consider copying to a new destination with --compare-dest to the
> original; that gives you a tree of changed files from which you
> can easily restore files or subtrees.
  
  Thanks for pointing out --compare-dest:  I verified it on a test
  directory and this feature does indeed offer a resolution to the
  situation described in the original post.  

  (I have a follow up question which belongs to another thread.)

  --Suresh