1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
|
Wiggle - wiggle a mis-match patch into a file.
Given
1/ a file
2/ a patch - which is two file fragments
find the minimal differences between the fragments in the patch
and apply those to the file.
This requires us to do a word-diff of file with frag-A, and
frag-A with frag-B, and the merge the result.
We read in the file and 2 frags and break them into words and keeping
an index and hash for each.
We then perform the two diffs producing lists of inserts and deletes.
ToDo
implement --replace
describe and implement correct replacement procedure
Reject matches that have a dis-proportionate cost
implement testing structure. DONE
Testing:
A directory tree containing tests. We look for key files
and run the appropriate test.
Key files are:
script : run that script in that directory
diff : if new exists, diff orig with new
else diff 'orig' with -1 of 'patch'
ldiff : as above, but lines
rediff : rediff 'patch'
merge : if 'patch' merge 'orig' with 'patch'
else merge 'orig' 'new' 'new2'
Replacement procedure:
Goal: Every change between A' and B' must be merged into
A somehow to produce B.
We can think of changes as additions, deletions, or replacements.
Every addition must be inserted somewhere, at the site of
best match for the context. If there is no good match...
I guess we insert at start or finish.
Every deletion is merged either by deleting matching text,
or inserting the string <<<---deleted-text--->>> and some
reasonably appropriate location.
Every replacement is merged either by removing the original
and replacing by the new, or by inserting
<<<---oldtext///newtext+++>>>
For each difference b->c between B and C:
if b precisely aligns with a in A, then replace a with c
else find some set of lines that b maybe is in and produce:
<<<<<<<<<<
segment from A
||||||||||
b, upto newlines
==========
c, upto newlines
>>>>>>>>>>
Maybe several (two?) passes.
-mw orig new new2 in tests/test dies. - FIXED
in test5, -dw orig new
produces strange output FIXED
if no matches are found, core is domps as lcsl is NULL FIXED
wdiff to look more like udiff
unchanged
+addition
-deletion
|change<<<+++additions+++>>> and <<<---deletions--->>>>
@@ line,numbers @@ in diff output
Speed: us aproxword for pdiff lineup.DONE
"refine" takes a diff and refines it, sortof
return a lcsl when reading a patch and refine that
rather than computing from scratch.
FIXME: pdiff should pick best bit, and rediff the two sides. DONE
|
