~bzr-pqm/bzr/bzr.dev

1185.1.29 by Robert Collins
merge merge tweaks from aaron, which includes latest .dev
1
Codeville
2
*********
3
4
Documentation on how this actually works is pretty scarce to say the
5
least.
6
7
I *think* I understand their merge algorithm though, and it's pretty
8
clever.  Basically we do a two-way merge between annotated forms of
9
the two files: that is, with each line marked with the revision in
10
which it last changed.  (I am simplifying here by speaking of lines
11
and changes, but I don't think it causes any essential problem.)
12
13
Now we walk through each file, line by line.  If the change that
14
introduced the line state in branch A is already merged into branch B,
15
then we can just take B.
16
17
Now is this actually better?
18
19
It may be better in several ways:
20
21
* Do not need to choose just a single ancestor, but rather can
22
  take advantage of all possible previous changes.
23
24
* Can handle OTHER containing changes which have been merged into
25
  THIS, but have then been overwritten.
26
27
* Can handle cherrypicks(!) by remembering which lines came in from
28
  that cherrypick; then we don't need to merge them again.
29
30
Some questions:
31
32
* Do we actually need to store the annotations, or can we just infer
33
  it at the time we do the merge?
34
35
* Can this be accomodated in something like an SCCS weave format?  I
36
  think something like a weave may work, in as much as it is basically
37
  a concatenation of annotations, but I don't know if it represents
38
  merges well enough to cope.
39
40
Can this handle binaries or type-specific merges, and if so how?
41
Unmergeable binaries are easy: just get the user to pick one.  Things
42
like XML are harder; we probably need to punt out to a type-specific
43
three-way merge.  Of course this approach does not forbid also
44
offering a 3-way merge.
45
46
----
47
48
I suppose this could be accomodated by an annotation cache on top of
49
plain history storage, or by using a storage format such as a weave
50
that can efficiently produce annotation information.
51
52
That is to say there is nothing inherently necessary about remembering
53
the line history at the point when it is committed, except that it
54
might be more efficient to do this once and remember it than to 
55
56
----
57
58
There is another interesting approach that can be used even in a tool
59
that does not inherently remember annotations:
60
61
Given two files to merge, find all regions of difference.  For each
62
such, try to find a common ancestor having the same content for the
63
region.  Subdivide the region if necessary.
64
65
This naive approach is probably infeasible, since it would mean
66
checking every possible predecessor.
67
68
----
69
70
Rather than storing or calculating annotations, we could try using a
71
complex weave, which allows one file version to be represented as a
72
weave of multiple disjoint previous versions.  It sounds complex but
73
it might work.
74
75
Essentially we store each file as a selection of lines that should be
76
turned on in that file.  These files might come from any of the
77
predecessors that were merged into that file.  Complex to get right
78
but it might work.
79
80
This is written in terms of lines, but it might make more sense to
81
just use byte ranges: perhaps more efficient when handling long files,
82
and makes binaries less of a special case.
83
84
codeville in fact does *not* seem to do this, though to me it seems
85
like a fairly natural corollary of their design.
86
87
This seems to imply holding the file text and ancestry of every branch
88
that ever merged into this one, rather than just finding them if we
89
later want them.  Hm.  That is nice in terms of doing smart merges.
90
That possibly causes trouble in terms of having a name for these
91
branches floating around inside our space, and making sure we don't
92
clash with them.  It may make sense in terms of having a working
93
directory be just a view into a shared database, looking at a
94
particular line of development.
95
96
Indeed the main difficulty seems to be of naming branches in this
97
space.  Perhaps we should move back to using repositories and named
98
branches within them, but not rely on branch names being unique out of
99
the context of a single repository.
100
101
Wow, this seems to open a big can of worms.
102
103
----
104
105
So the conclusion is that this is very cool, but it does not require a
106
fundamental change of model and can be implemented later.