Show More
@@ -1,156 +1,176 b'' | |||||
1 | # narrowrevlog.py - revlog storing irrelevant nodes as "ellipsis" nodes |
|
1 | # narrowrevlog.py - revlog storing irrelevant nodes as "ellipsis" nodes | |
2 | # |
|
2 | # | |
3 | # Copyright 2017 Google, Inc. |
|
3 | # Copyright 2017 Google, Inc. | |
4 | # |
|
4 | # | |
5 | # This software may be used and distributed according to the terms of the |
|
5 | # This software may be used and distributed according to the terms of the | |
6 | # GNU General Public License version 2 or any later version. |
|
6 | # GNU General Public License version 2 or any later version. | |
7 |
|
7 | |||
8 | from __future__ import absolute_import |
|
8 | from __future__ import absolute_import | |
9 |
|
9 | |||
10 | from mercurial import ( |
|
10 | from mercurial import ( | |
11 | manifest, |
|
11 | manifest, | |
12 | revlog, |
|
12 | revlog, | |
13 | util, |
|
13 | util, | |
14 | ) |
|
14 | ) | |
15 |
|
15 | |||
16 | def readtransform(self, text): |
|
16 | def readtransform(self, text): | |
17 | return text, False |
|
17 | return text, False | |
18 |
|
18 | |||
19 | def writetransform(self, text): |
|
19 | def writetransform(self, text): | |
20 | return text, False |
|
20 | return text, False | |
21 |
|
21 | |||
22 | def rawtransform(self, text): |
|
22 | def rawtransform(self, text): | |
23 | return False |
|
23 | return False | |
24 |
|
24 | |||
25 | revlog.addflagprocessor(revlog.REVIDX_ELLIPSIS, |
|
25 | revlog.addflagprocessor(revlog.REVIDX_ELLIPSIS, | |
26 | (readtransform, writetransform, rawtransform)) |
|
26 | (readtransform, writetransform, rawtransform)) | |
27 |
|
27 | |||
28 | def setup(): |
|
28 | def setup(): | |
29 | # We just wanted to add the flag processor, which is done at module |
|
29 | # We just wanted to add the flag processor, which is done at module | |
30 | # load time. |
|
30 | # load time. | |
31 | pass |
|
31 | pass | |
32 |
|
32 | |||
33 | class excludeddir(manifest.treemanifest): |
|
33 | class excludeddir(manifest.treemanifest): | |
|
34 | """Stand-in for a directory that is excluded from the repository. | |||
|
35 | ||||
|
36 | With narrowing active on a repository that uses treemanifests, | |||
|
37 | some of the directory revlogs will be excluded from the resulting | |||
|
38 | clone. This is a huge storage win for clients, but means we need | |||
|
39 | some sort of pseudo-manifest to surface to internals so we can | |||
|
40 | detect a merge conflict outside the narrowspec. That's what this | |||
|
41 | class is: it stands in for a directory whose node is known, but | |||
|
42 | whose contents are unknown. | |||
|
43 | """ | |||
34 | def __init__(self, dir, node): |
|
44 | def __init__(self, dir, node): | |
35 | super(excludeddir, self).__init__(dir) |
|
45 | super(excludeddir, self).__init__(dir) | |
36 | self._node = node |
|
46 | self._node = node | |
37 | # Add an empty file, which will be included by iterators and such, |
|
47 | # Add an empty file, which will be included by iterators and such, | |
38 | # appearing as the directory itself (i.e. something like "dir/") |
|
48 | # appearing as the directory itself (i.e. something like "dir/") | |
39 | self._files[''] = node |
|
49 | self._files[''] = node | |
40 | self._flags[''] = 't' |
|
50 | self._flags[''] = 't' | |
41 |
|
51 | |||
42 | # Manifests outside the narrowspec should never be modified, so avoid |
|
52 | # Manifests outside the narrowspec should never be modified, so avoid | |
43 | # copying. This makes a noticeable difference when there are very many |
|
53 | # copying. This makes a noticeable difference when there are very many | |
44 | # directories outside the narrowspec. Also, it makes sense for the copy to |
|
54 | # directories outside the narrowspec. Also, it makes sense for the copy to | |
45 | # be of the same type as the original, which would not happen with the |
|
55 | # be of the same type as the original, which would not happen with the | |
46 | # super type's copy(). |
|
56 | # super type's copy(). | |
47 | def copy(self): |
|
57 | def copy(self): | |
48 | return self |
|
58 | return self | |
49 |
|
59 | |||
50 | class excludeddirmanifestctx(manifest.treemanifestctx): |
|
60 | class excludeddirmanifestctx(manifest.treemanifestctx): | |
|
61 | """context wrapper for excludeddir - see that docstring for rationale""" | |||
51 | def __init__(self, dir, node): |
|
62 | def __init__(self, dir, node): | |
52 | self._dir = dir |
|
63 | self._dir = dir | |
53 | self._node = node |
|
64 | self._node = node | |
54 |
|
65 | |||
55 | def read(self): |
|
66 | def read(self): | |
56 | return excludeddir(self._dir, self._node) |
|
67 | return excludeddir(self._dir, self._node) | |
57 |
|
68 | |||
58 | def write(self, *args): |
|
69 | def write(self, *args): | |
59 | raise AssertionError('Attempt to write manifest from excluded dir %s' % |
|
70 | raise AssertionError('Attempt to write manifest from excluded dir %s' % | |
60 | self._dir) |
|
71 | self._dir) | |
61 |
|
72 | |||
62 | class excludedmanifestrevlog(manifest.manifestrevlog): |
|
73 | class excludedmanifestrevlog(manifest.manifestrevlog): | |
|
74 | """Stand-in for excluded treemanifest revlogs. | |||
|
75 | ||||
|
76 | When narrowing is active on a treemanifest repository, we'll have | |||
|
77 | references to directories we can't see due to the revlog being | |||
|
78 | skipped. This class exists to conform to the manifestrevlog | |||
|
79 | interface for those directories and proactively prevent writes to | |||
|
80 | outside the narrowspec. | |||
|
81 | """ | |||
|
82 | ||||
63 | def __init__(self, dir): |
|
83 | def __init__(self, dir): | |
64 | self._dir = dir |
|
84 | self._dir = dir | |
65 |
|
85 | |||
66 | def __len__(self): |
|
86 | def __len__(self): | |
67 | raise AssertionError('Attempt to get length of excluded dir %s' % |
|
87 | raise AssertionError('Attempt to get length of excluded dir %s' % | |
68 | self._dir) |
|
88 | self._dir) | |
69 |
|
89 | |||
70 | def rev(self, node): |
|
90 | def rev(self, node): | |
71 | raise AssertionError('Attempt to get rev from excluded dir %s' % |
|
91 | raise AssertionError('Attempt to get rev from excluded dir %s' % | |
72 | self._dir) |
|
92 | self._dir) | |
73 |
|
93 | |||
74 | def linkrev(self, node): |
|
94 | def linkrev(self, node): | |
75 | raise AssertionError('Attempt to get linkrev from excluded dir %s' % |
|
95 | raise AssertionError('Attempt to get linkrev from excluded dir %s' % | |
76 | self._dir) |
|
96 | self._dir) | |
77 |
|
97 | |||
78 | def node(self, rev): |
|
98 | def node(self, rev): | |
79 | raise AssertionError('Attempt to get node from excluded dir %s' % |
|
99 | raise AssertionError('Attempt to get node from excluded dir %s' % | |
80 | self._dir) |
|
100 | self._dir) | |
81 |
|
101 | |||
82 | def add(self, *args, **kwargs): |
|
102 | def add(self, *args, **kwargs): | |
83 | # We should never write entries in dirlogs outside the narrow clone. |
|
103 | # We should never write entries in dirlogs outside the narrow clone. | |
84 | # However, the method still gets called from writesubtree() in |
|
104 | # However, the method still gets called from writesubtree() in | |
85 | # _addtree(), so we need to handle it. We should possibly make that |
|
105 | # _addtree(), so we need to handle it. We should possibly make that | |
86 | # avoid calling add() with a clean manifest (_dirty is always False |
|
106 | # avoid calling add() with a clean manifest (_dirty is always False | |
87 | # in excludeddir instances). |
|
107 | # in excludeddir instances). | |
88 | pass |
|
108 | pass | |
89 |
|
109 | |||
90 | def makenarrowmanifestrevlog(mfrevlog, repo): |
|
110 | def makenarrowmanifestrevlog(mfrevlog, repo): | |
91 | if util.safehasattr(mfrevlog, '_narrowed'): |
|
111 | if util.safehasattr(mfrevlog, '_narrowed'): | |
92 | return |
|
112 | return | |
93 |
|
113 | |||
94 | class narrowmanifestrevlog(mfrevlog.__class__): |
|
114 | class narrowmanifestrevlog(mfrevlog.__class__): | |
95 | # This function is called via debug{revlog,index,data}, but also during |
|
115 | # This function is called via debug{revlog,index,data}, but also during | |
96 | # at least some push operations. This will be used to wrap/exclude the |
|
116 | # at least some push operations. This will be used to wrap/exclude the | |
97 | # child directories when using treemanifests. |
|
117 | # child directories when using treemanifests. | |
98 | def dirlog(self, dir): |
|
118 | def dirlog(self, dir): | |
99 | if dir and not dir.endswith('/'): |
|
119 | if dir and not dir.endswith('/'): | |
100 | dir = dir + '/' |
|
120 | dir = dir + '/' | |
101 | if not repo.narrowmatch().visitdir(dir[:-1] or '.'): |
|
121 | if not repo.narrowmatch().visitdir(dir[:-1] or '.'): | |
102 | return excludedmanifestrevlog(dir) |
|
122 | return excludedmanifestrevlog(dir) | |
103 | result = super(narrowmanifestrevlog, self).dirlog(dir) |
|
123 | result = super(narrowmanifestrevlog, self).dirlog(dir) | |
104 | makenarrowmanifestrevlog(result, repo) |
|
124 | makenarrowmanifestrevlog(result, repo) | |
105 | return result |
|
125 | return result | |
106 |
|
126 | |||
107 | mfrevlog.__class__ = narrowmanifestrevlog |
|
127 | mfrevlog.__class__ = narrowmanifestrevlog | |
108 | mfrevlog._narrowed = True |
|
128 | mfrevlog._narrowed = True | |
109 |
|
129 | |||
110 | def makenarrowmanifestlog(mfl, repo): |
|
130 | def makenarrowmanifestlog(mfl, repo): | |
111 | class narrowmanifestlog(mfl.__class__): |
|
131 | class narrowmanifestlog(mfl.__class__): | |
112 | def get(self, dir, node, verify=True): |
|
132 | def get(self, dir, node, verify=True): | |
113 | if not repo.narrowmatch().visitdir(dir[:-1] or '.'): |
|
133 | if not repo.narrowmatch().visitdir(dir[:-1] or '.'): | |
114 | return excludeddirmanifestctx(dir, node) |
|
134 | return excludeddirmanifestctx(dir, node) | |
115 | return super(narrowmanifestlog, self).get(dir, node, verify=verify) |
|
135 | return super(narrowmanifestlog, self).get(dir, node, verify=verify) | |
116 | mfl.__class__ = narrowmanifestlog |
|
136 | mfl.__class__ = narrowmanifestlog | |
117 |
|
137 | |||
118 | def makenarrowfilelog(fl, narrowmatch): |
|
138 | def makenarrowfilelog(fl, narrowmatch): | |
119 | class narrowfilelog(fl.__class__): |
|
139 | class narrowfilelog(fl.__class__): | |
120 | def renamed(self, node): |
|
140 | def renamed(self, node): | |
121 | m = super(narrowfilelog, self).renamed(node) |
|
141 | m = super(narrowfilelog, self).renamed(node) | |
122 | if m and not narrowmatch(m[0]): |
|
142 | if m and not narrowmatch(m[0]): | |
123 | return None |
|
143 | return None | |
124 | return m |
|
144 | return m | |
125 |
|
145 | |||
126 | def size(self, rev): |
|
146 | def size(self, rev): | |
127 | # We take advantage of the fact that remotefilelog |
|
147 | # We take advantage of the fact that remotefilelog | |
128 | # lacks a node() method to just skip the |
|
148 | # lacks a node() method to just skip the | |
129 | # rename-checking logic when on remotefilelog. This |
|
149 | # rename-checking logic when on remotefilelog. This | |
130 | # might be incorrect on other non-revlog-based storage |
|
150 | # might be incorrect on other non-revlog-based storage | |
131 | # engines, but for now this seems to be fine. |
|
151 | # engines, but for now this seems to be fine. | |
132 | if util.safehasattr(self, 'node'): |
|
152 | if util.safehasattr(self, 'node'): | |
133 | node = self.node(rev) |
|
153 | node = self.node(rev) | |
134 | # Because renamed() is overridden above to |
|
154 | # Because renamed() is overridden above to | |
135 | # sometimes return None even if there is metadata |
|
155 | # sometimes return None even if there is metadata | |
136 | # in the revlog, size can be incorrect for |
|
156 | # in the revlog, size can be incorrect for | |
137 | # copies/renames, so we need to make sure we call |
|
157 | # copies/renames, so we need to make sure we call | |
138 | # the super class's implementation of renamed() |
|
158 | # the super class's implementation of renamed() | |
139 | # for the purpose of size calculation. |
|
159 | # for the purpose of size calculation. | |
140 | if super(narrowfilelog, self).renamed(node): |
|
160 | if super(narrowfilelog, self).renamed(node): | |
141 | return len(self.read(node)) |
|
161 | return len(self.read(node)) | |
142 | return super(narrowfilelog, self).size(rev) |
|
162 | return super(narrowfilelog, self).size(rev) | |
143 |
|
163 | |||
144 | def cmp(self, node, text): |
|
164 | def cmp(self, node, text): | |
145 | different = super(narrowfilelog, self).cmp(node, text) |
|
165 | different = super(narrowfilelog, self).cmp(node, text) | |
146 | if different: |
|
166 | if different: | |
147 | # Similar to size() above, if the file was copied from |
|
167 | # Similar to size() above, if the file was copied from | |
148 | # a file outside the narrowspec, the super class's |
|
168 | # a file outside the narrowspec, the super class's | |
149 | # would have returned True because we tricked it into |
|
169 | # would have returned True because we tricked it into | |
150 | # thinking that the file was not renamed. |
|
170 | # thinking that the file was not renamed. | |
151 | if super(narrowfilelog, self).renamed(node): |
|
171 | if super(narrowfilelog, self).renamed(node): | |
152 | t2 = self.read(node) |
|
172 | t2 = self.read(node) | |
153 | return t2 != text |
|
173 | return t2 != text | |
154 | return different |
|
174 | return different | |
155 |
|
175 | |||
156 | fl.__class__ = narrowfilelog |
|
176 | fl.__class__ = narrowfilelog |
General Comments 0
You need to be logged in to leave comments.
Login now