4
Documentation on how this actually works is pretty scarce to say the
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.)
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.
17
Now is this actually better?
19
It may be better in several ways:
21
* Do not need to choose just a single ancestor, but rather can
22
take advantage of all possible previous changes.
24
* Can handle OTHER containing changes which have been merged into
25
THIS, but have then been overwritten.
27
* Can handle cherrypicks(!) by remembering which lines came in from
28
that cherrypick; then we don't need to merge them again.
32
* Do we actually need to store the annotations, or can we just infer
33
it at the time we do the merge?
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.
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.
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.
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
58
There is another interesting approach that can be used even in a tool
59
that does not inherently remember annotations:
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.
65
This naive approach is probably infeasible, since it would mean
66
checking every possible predecessor.
70
So the conclusion is that this is very cool, but it does not require a
71
fundamental change of model and can be implemented later.
added
removed
doc/compared-codeville.txt
