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
|
Darcs compared to Arch
======================
Simpler to use; perhaps harder to completely understand.
Always local; always fast.
Patch commution is slow and perhaps doesn't clearly do what people
want.
Too slow!
Can't reliably get back to any previous point. Explicitly not
addressing source archive/librarian function.
Loads everything into memory.
Written in Haskell.
A really simple pre-commit check hook is remarkably useful.
http://www.scannedinavian.org/DarcsWiki/DifferencesFromArch
Sometimes useful to be able to set email per-branch, for people who
work on different projects under different personas.
Token replace
-------------
Very cute; possibly handy; not absolutely necessary in most places.
Somewhat limited by the requirement that it be reversible.
This is one of very few cases where it does seem necessary that we
store deltas, rather than tree states. But that seems to cause other
problems in terms of being able to reliably sign revisions.
This can perhaps be inferred by a smart 3-way merge tool. Certainly
you could have it do sub-line merges.
Partial commit
--------------
darcs allows you to commit only some of the changes to a single file.
This is like the common feature of commiting only a subset of changed
files, but taken to a higher level.
It is useful more often than one might think: it is common to fix some
documentation 'on the wing' and while strictly it should be in a
separate commit it is not always worth the hassle to back out changes,
fix the docs, then do the real change. Similarly for making a
separate branch.
Although the idea is very good, the current darcs implementation is
limited to selecting by patch hunk, which means that neighbouring
changes cannot be separated. Fixing this probably means having some
kind of pluggable GUI to build the file-to-be-committed or an edited
patch, possibly using something like meld, emacs, or dirdiff. Another
approach some people might like is editing the diff file to chop out
hunks.
I don't think this needs to be on by default, as it is in darcs. It
is usual to commit all the changes.
For this to work safely, it is good to have a commit hook that
builds/tests the tree. Of course this needs to be evaluated against
the tree as it will be committed (taking account of partial commits),
not the working tree.
Schwern
-------
* http://www.scannedinavian.org/DarcsWiki/SchwernLikesDarcs
* http://www.scannedinavian.org/DarcsWiki/SchwernHatesDarcs
I think we hit most of these; more evidence for the hypothesis that
what people really like about darcs is the simple interface not the
patch-commutation model.
|