~bzr-pqm/bzr/bzr.dev

1002 by Martin Pool
- notes from discussion on splitting and joining files
1
Split and join files
2
====================
3
4
Version control systems tend to have trouble when files are split or
5
joined.  Here is one possibility for handling that.
6
7
8
Splitting files
9
---------------
10
11
This can work like 'svn cp': copy with history::
12
13
  % bzr cp a b
14
15
This creates a new file b, whose file-history is the same as a, but
16
with a different file-id, and present at a different name.
17
18
This creates the slightly strange situation that b claims to have text
19
versions for revisions where b was not actually present.
20
21
22
Joining files
23
-------------
24
25
We might want to also join files in various cases:
26
27
 * Multiple imports of the same file
28
29
 * Two source files are being refactored into a single one
30
31
 * Two people apply an external patch that creates the same file
32
33
I'm not aware of any other system that handles this.
34
35
  % bzr join a b
36
37
a and b must be versioned files.  b is removed from the working tree
38
and inventory.  b's file id is added as a secondary id as a, so that
39
later merges between them will detect them as being the same.  
40
41
b's weave is (???) incorporated into a.  
42
43
This means that we have a file which has multiple history paths.  For
44
example, asking to compare that file to its state in a previous
45
version has to arbitrarily pick one to use.  That's not necessarily
46
fatal if we assign one as primary, but it might be a bit confusing.
47
Trying to merge from another tree might can't really choose a good
48
basis.
49
50
51
52
File aliases
53
------------
54
55
Rather than trying to join files we might just add notes that files
56
are also known by a different file-id in other trees.
57
58
59