4
The status command is used to provide a pithy listing of the changes between
5
two trees. Its common case is between the working tree and the basis tree, but
6
it can be used between any two arbitrary trees.
13
Status shows several things in parallel (for the paths the user supplied mapped
14
across the from and to tree, and any pending merges in the to tree).
16
1. Single line summary of all new revisions - the pending merges and their
18
2. Changes to the tree shape - adds/deletes/renames.
19
3. Changes to versioned content - kind changes and content changes.
20
4. Unknown files in the to tree.
21
5. Files with conflicts in the to tree.
24
Ideal work for working tree to historical status
25
------------------------------------------------
27
We need to do the following things at a minimum:
29
1. Determine new revisions - the pending merges and history.
31
1. Retrieve the first line of the commit message for the new revisions.
33
1. Determine the tree differences between the two trees using the users paths
34
to limit the scope, and resolving paths in the trees for any pending merges.
35
We arguably don't care about tracking metadata for this - only the value of
36
the tree the user commited.
38
1. The entire contents of directories which are versioned when showing
41
1. Whether a given unversioned path is unknown or ignored.
43
1. The list conflicted paths in the tree (which match the users path
47
Expanding on the tree difference case we will need to:
49
1. Stat every path in working trees which is included by the users path
50
selection to ascertain kind and execute bit.
52
1. For paths which have the same kind in both trees and have content, read
53
that content or otherwise determine whether the content has changed. Using
54
our hash cache from the dirstate allows us to avoid reading the file in the
55
common case. There are alternative ways to achieve this - we could record
56
a pointer to a revision which contained this fileid with the current content
57
rather than storing the content's hash; but this seems to be a pointless
58
double-indirection unless we save enough storage in the working tree. A
59
variation of this is to not record an explicit pointer but instead
60
define an implicit pointer as being to the left-hand-parent tree.
66
- We should stat files in the same directory without reading or statting
67
files in other directories. That is we should do all the statting we
68
intend to do within a given directory without doing any other IO, to
69
minimise pressure on the drive heads to seek.
71
- We should read files in the same directory without reading or writing
72
files in other directories - and note this is separate to statting (file
73
data is usually physically disjoint to metadata).
79
- The stat operation clearly involves every versioned path in the common case.
80
- Expanding out the users path selection in a naive manner involves reading the
81
entire tree shape information for both trees and for all pending-merge trees.
82
(Dirstate makes this tolerably cheap for now, but we're still scaling
84
- The amount of effort required to generate tree differences between the
85
working tree and the basis tree is interesting: with a tree-like structure
86
and some generatable name for child nodes we use the working tree data to
87
eliminate accessing or considering subtrees regardless of historival
88
age. However, if we have had to access the historical tree shape to
89
perform path selection this rather reduces the win we can obtain here.
90
If we can cause path expansion to not require historical shape access
91
(perhaps by performing the expansion after calculating the tree
92
difference for the top level of the selected path) then we can gain a
93
larger win. This strongly suggests that path expansion and tree
94
difference generation should be linked in terms of API.