[Bf-committers] Re: Bf-committers Digest, Vol 18, Issue 15

[Joe Eagar]

Stephen Swaney wrote:

>On Sun, Jan 08, 2006 at 03:13:37PM +0100, Erik Unger wrote:
>  
>
>>>Certainly there are exceptions that
>>>handle binary data.
>>>      
>>>
>>Subversion uses binary diffs for efficient transfer and storage
>>of binary content.
>>But there is no visual diff viewer (because it doesn't make
>>much sense for generic binary content).
>>    
>>
>
>Since the same confusion came up on IRC, I should clarify
>that when I said some revision control systems can diff
>binary data, I meant being able to do a meaningful diff in
>the context of the data.
>
>An example would be being able to see a paragraph has
>changed in a word processor file rather than simply seeing
>that a range of bytes is different.  A Blender example is being
>able to see that Material.001 has been modified.
>
>The very expensive ClearCase is able to do this for certain
>application binary files.
>
>  
>
That may be possible, actually.  You'd write a program (or even a python 
script) that analyzes each Library Block in a .blend file for differences.

A straight binary comparison probably wouldn't work, as saved pointer 
data would change in different saves (especially on different operating 
systems).  You'd likely have to check each Library Block pointer 
variable in each SDNA struct, to see if in both files the blocks they 
point to have the same names.  Pointers that point to other data  (like 
non-block SDNA structs or simply allocated data) would have to be dealt 
with somehow, too.

So this would probably be pretty complicated. . .

joeedh