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
|
Syntax for identifying revisions
********************************
There seem to be two main options: use a separate ``-r`` option, or
put a revision identifier within the filename.
``-r`` options are familiar from CVS and Subversion, and easy to
specify, and cover almost all situations.
Special identifiers in the filename are sometimes more concise, and
allow for more complete specification of filenames in the presence of
renames. They may be more complex to implement, and will cause some
filenames to be either forbidden or to require escaping.
----
How to interpret revisions in the presence of renamed files? Suppose
hello.c was renamed in r3; when we do ``bzr diff -r2 hello.c`` should
we compare to
* the file currently called ``hello.c``, as it was in revision 2, even
if it then had a different name
or
* the file that was called ``hello.c`` in revision 2, even if that is
a different file-id?
Often only one will work, but it is possible to have situations where
in revision 2 there was a file called ``hello.c`` but it is not the
same one that now has that name.
If we can have a simple syntax that accomodates either that would be
good. One might be to mix revision numbers into the path, rather than
a separate ``-r`` option. For example::
hello.c@2
@2/hello.c
src/@2/hello.c
The simplest case is to just append the number, but you can vary
them. It might even be possible to chain them, getting the file id
from one revision and its state from another:
@70/hello.c@31
This means "the file that was called ``hello.c`` in revision 70, give
me its state in revision 31." (If you wanted to do this across
branches, I think you might need to get the file id and then look it
up, etc.)
Some more syntax: you can refer to a revision by its hash just by
specifying that; we can distinguish them from anything else by their
length::
hello.c@09fac8dbfd27bd9b4d23a00eb648aa751789536d
(Is that really safe? Perhaps some punctuation is needed to
distinguish it from a label. But then I think people should rarely or
never use them; perhaps it need not be allowed at all.)
You can also give a date, which is used to find the most recent
revision no later than that date::
hello.c@{2004-12-01}
hello.c@{yesterday}
./@{10:30}/
(Braces can be a bit scary for shell, but if there are no commas they
are somewhat safe.)
Or, once we have labels, you can use a label::
hello.c@release-2.2
|