bzr branch
http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar
|
0.17.9
by Robert Collins
Initial stab at repository format support. |
1 |
# groupcompress, a bzr plugin providing improved disk utilisation
|
2 |
# Copyright (C) 2008 Canonical Limited.
|
|
3 |
#
|
|
4 |
# This program is free software; you can redistribute it and/or modify
|
|
5 |
# it under the terms of the GNU General Public License version 2 as published
|
|
6 |
# by the Free Software Foundation.
|
|
7 |
#
|
|
8 |
# This program is distributed in the hope that it will be useful,
|
|
9 |
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
10 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
11 |
# GNU General Public License for more details.
|
|
12 |
#
|
|
13 |
# You should have received a copy of the GNU General Public License
|
|
14 |
# along with this program; if not, write to the Free Software
|
|
15 |
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
|
16 |
#
|
|
17 |
||
18 |
"""Repostory formats using B+Tree indices and groupcompress compression."""
|
|
19 |
||
20 |
import md5 |
|
21 |
import time |
|
22 |
||
23 |
from bzrlib import debug, errors, pack, repository |
|
24 |
from bzrlib.index import GraphIndex, GraphIndexBuilder |
|
25 |
from bzrlib.repository import InterPackRepo |
|
26 |
from bzrlib.plugins.groupcompress.groupcompress import ( |
|
27 |
_GCGraphIndex, |
|
28 |
GroupCompressVersionedFiles, |
|
29 |
)
|
|
30 |
from bzrlib.plugins.index2.btree_index import ( |
|
31 |
BTreeBuilder, |
|
32 |
BTreeGraphIndex, |
|
33 |
FixedMemoryGraphIndex, |
|
34 |
)
|
|
35 |
from bzrlib.osutils import rand_chars |
|
36 |
from bzrlib.repofmt.pack_repo import ( |
|
37 |
Pack, |
|
38 |
NewPack, |
|
39 |
KnitPackRepository, |
|
40 |
RepositoryPackCollection, |
|
41 |
RepositoryFormatPackDevelopment0, |
|
42 |
RepositoryFormatPackDevelopment0Subtree, |
|
43 |
RepositoryFormatKnitPack1, |
|
44 |
RepositoryFormatKnitPack3, |
|
45 |
RepositoryFormatKnitPack4, |
|
46 |
Packer, |
|
47 |
ReconcilePacker, |
|
48 |
OptimisingPacker, |
|
49 |
)
|
|
50 |
from bzrlib import ui |
|
51 |
||
52 |
||
53 |
def open_pack(self): |
|
54 |
return self._pack_collection.pack_factory(self._pack_collection._upload_transport, |
|
55 |
self._pack_collection._index_transport, |
|
56 |
self._pack_collection._pack_transport, upload_suffix=self.suffix, |
|
57 |
file_mode=self._pack_collection.repo.bzrdir._get_file_mode()) |
|
58 |
||
59 |
||
60 |
Packer.open_pack = open_pack |
|
61 |
||
62 |
||
63 |
class GCPack(NewPack): |
|
64 |
||
65 |
def __init__(self, upload_transport, index_transport, pack_transport, |
|
66 |
upload_suffix='', file_mode=None): |
|
67 |
"""Create a NewPack instance. |
|
68 |
||
69 |
:param upload_transport: A writable transport for the pack to be
|
|
70 |
incrementally uploaded to.
|
|
71 |
:param index_transport: A writable transport for the pack's indices to
|
|
72 |
be written to when the pack is finished.
|
|
73 |
:param pack_transport: A writable transport for the pack to be renamed
|
|
74 |
to when the upload is complete. This *must* be the same as
|
|
75 |
upload_transport.clone('../packs').
|
|
76 |
:param upload_suffix: An optional suffix to be given to any temporary
|
|
77 |
files created during the pack creation. e.g '.autopack'
|
|
78 |
:param file_mode: An optional file mode to create the new files with.
|
|
79 |
"""
|
|
80 |
# The relative locations of the packs are constrained, but all are
|
|
81 |
# passed in because the caller has them, so as to avoid object churn.
|
|
82 |
Pack.__init__(self, |
|
83 |
# Revisions: parents list, no text compression.
|
|
84 |
BTreeBuilder(reference_lists=1), |
|
85 |
# Inventory: compressed, with graph for compatibility with other
|
|
86 |
# existing bzrlib code.
|
|
87 |
BTreeBuilder(reference_lists=1), |
|
88 |
# Texts: per file graph:
|
|
89 |
BTreeBuilder(reference_lists=1, key_elements=2), |
|
90 |
# Signatures: Just blobs to store, no compression, no parents
|
|
91 |
# listing.
|
|
92 |
BTreeBuilder(reference_lists=0), |
|
93 |
)
|
|
94 |
# where should the new pack be opened
|
|
95 |
self.upload_transport = upload_transport |
|
96 |
# where are indices written out to
|
|
97 |
self.index_transport = index_transport |
|
98 |
# where is the pack renamed to when it is finished?
|
|
99 |
self.pack_transport = pack_transport |
|
100 |
# What file mode to upload the pack and indices with.
|
|
101 |
self._file_mode = file_mode |
|
102 |
# tracks the content written to the .pack file.
|
|
103 |
self._hash = md5.new() |
|
104 |
# a four-tuple with the length in bytes of the indices, once the pack
|
|
105 |
# is finalised. (rev, inv, text, sigs)
|
|
106 |
self.index_sizes = None |
|
107 |
# How much data to cache when writing packs. Note that this is not
|
|
108 |
# synchronised with reads, because it's not in the transport layer, so
|
|
109 |
# is not safe unless the client knows it won't be reading from the pack
|
|
110 |
# under creation.
|
|
111 |
self._cache_limit = 0 |
|
112 |
# the temporary pack file name.
|
|
113 |
self.random_name = rand_chars(20) + upload_suffix |
|
114 |
# when was this pack started ?
|
|
115 |
self.start_time = time.time() |
|
116 |
# open an output stream for the data added to the pack.
|
|
117 |
self.write_stream = self.upload_transport.open_write_stream( |
|
118 |
self.random_name, mode=self._file_mode) |
|
119 |
if 'pack' in debug.debug_flags: |
|
120 |
mutter('%s: create_pack: pack stream open: %s%s t+%6.3fs', |
|
121 |
time.ctime(), self.upload_transport.base, self.random_name, |
|
122 |
time.time() - self.start_time) |
|
123 |
# A list of byte sequences to be written to the new pack, and the
|
|
124 |
# aggregate size of them. Stored as a list rather than separate
|
|
125 |
# variables so that the _write_data closure below can update them.
|
|
126 |
self._buffer = [[], 0] |
|
127 |
# create a callable for adding data
|
|
128 |
#
|
|
129 |
# robertc says- this is a closure rather than a method on the object
|
|
130 |
# so that the variables are locals, and faster than accessing object
|
|
131 |
# members.
|
|
132 |
def _write_data(bytes, flush=False, _buffer=self._buffer, |
|
133 |
_write=self.write_stream.write, _update=self._hash.update): |
|
134 |
_buffer[0].append(bytes) |
|
135 |
_buffer[1] += len(bytes) |
|
136 |
# buffer cap
|
|
137 |
if _buffer[1] > self._cache_limit or flush: |
|
138 |
bytes = ''.join(_buffer[0]) |
|
139 |
_write(bytes) |
|
140 |
_update(bytes) |
|
141 |
_buffer[:] = [[], 0] |
|
142 |
# expose this on self, for the occasion when clients want to add data.
|
|
143 |
self._write_data = _write_data |
|
144 |
# a pack writer object to serialise pack records.
|
|
145 |
self._writer = pack.ContainerWriter(self._write_data) |
|
146 |
self._writer.begin() |
|
147 |
# what state is the pack in? (open, finished, aborted)
|
|
148 |
self._state = 'open' |
|
149 |
||
150 |
def _replace_index_with_readonly(self, index_type): |
|
151 |
setattr(self, index_type + '_index', |
|
152 |
BTreeGraphIndex(self.index_transport, |
|
153 |
self.index_name(index_type, self.name), |
|
154 |
self.index_sizes[self.index_offset(index_type)])) |
|
155 |
||
156 |
||
157 |
RepositoryPackCollection.pack_factory = NewPack |
|
158 |
||
159 |
class GCRepositoryPackCollection(RepositoryPackCollection): |
|
160 |
||
161 |
pack_factory = GCPack |
|
162 |
||
163 |
def _make_index(self, name, suffix): |
|
164 |
"""Overridden to use BTreeGraphIndex objects.""" |
|
165 |
size_offset = self._suffix_offsets[suffix] |
|
166 |
index_name = name + suffix |
|
167 |
index_size = self._names[name][size_offset] |
|
168 |
return BTreeGraphIndex( |
|
169 |
self._index_transport, index_name, index_size) |
|
170 |
||
171 |
def _start_write_group(self): |
|
172 |
# Do not permit preparation for writing if we're not in a 'write lock'.
|
|
173 |
if not self.repo.is_write_locked(): |
|
174 |
raise errors.NotWriteLocked(self) |
|
175 |
self._new_pack = self.pack_factory(self._upload_transport, self._index_transport, |
|
176 |
self._pack_transport, upload_suffix='.pack', |
|
177 |
file_mode=self.repo.bzrdir._get_file_mode()) |
|
178 |
# allow writing: queue writes to a new index
|
|
179 |
self.revision_index.add_writable_index(self._new_pack.revision_index, |
|
180 |
self._new_pack) |
|
181 |
self.inventory_index.add_writable_index(self._new_pack.inventory_index, |
|
182 |
self._new_pack) |
|
183 |
self.text_index.add_writable_index(self._new_pack.text_index, |
|
184 |
self._new_pack) |
|
185 |
self.signature_index.add_writable_index(self._new_pack.signature_index, |
|
186 |
self._new_pack) |
|
187 |
||
188 |
self.repo.inventories._index._add_callback = self.inventory_index.add_callback |
|
189 |
self.repo.revisions._index._add_callback = self.revision_index.add_callback |
|
190 |
self.repo.signatures._index._add_callback = self.signature_index.add_callback |
|
191 |
self.repo.texts._index._add_callback = self.text_index.add_callback |
|
192 |
||
193 |
||
194 |
||
195 |
class GCPackRepository(KnitPackRepository): |
|
196 |
"""GC customisation of KnitPackRepository.""" |
|
197 |
||
198 |
def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class, |
|
199 |
_serializer): |
|
200 |
"""Overridden to change pack collection class.""" |
|
201 |
KnitPackRepository.__init__(self, _format, a_bzrdir, control_files, |
|
202 |
_commit_builder_class, _serializer) |
|
203 |
# and now replace everything it did :)
|
|
204 |
index_transport = self._transport.clone('indices') |
|
205 |
self._pack_collection = GCRepositoryPackCollection(self, |
|
206 |
self._transport, index_transport, |
|
207 |
self._transport.clone('upload'), |
|
208 |
self._transport.clone('packs')) |
|
209 |
self.inventories = GroupCompressVersionedFiles( |
|
210 |
_GCGraphIndex(self._pack_collection.inventory_index.combined_index, |
|
211 |
add_callback=self._pack_collection.inventory_index.add_callback, |
|
212 |
parents=True, is_locked=self.is_locked), |
|
213 |
access=self._pack_collection.inventory_index.data_access) |
|
214 |
self.revisions = GroupCompressVersionedFiles( |
|
215 |
_GCGraphIndex(self._pack_collection.revision_index.combined_index, |
|
216 |
add_callback=self._pack_collection.revision_index.add_callback, |
|
217 |
parents=True, is_locked=self.is_locked), |
|
218 |
access=self._pack_collection.revision_index.data_access, |
|
219 |
delta=False) |
|
220 |
self.signatures = GroupCompressVersionedFiles( |
|
221 |
_GCGraphIndex(self._pack_collection.signature_index.combined_index, |
|
222 |
add_callback=self._pack_collection.signature_index.add_callback, |
|
223 |
parents=False, is_locked=self.is_locked), |
|
224 |
access=self._pack_collection.signature_index.data_access, |
|
225 |
delta=False) |
|
226 |
self.texts = GroupCompressVersionedFiles( |
|
227 |
_GCGraphIndex(self._pack_collection.text_index.combined_index, |
|
228 |
add_callback=self._pack_collection.text_index.add_callback, |
|
229 |
parents=True, is_locked=self.is_locked), |
|
230 |
access=self._pack_collection.text_index.data_access) |
|
231 |
# True when the repository object is 'write locked' (as opposed to the
|
|
232 |
# physical lock only taken out around changes to the pack-names list.)
|
|
233 |
# Another way to represent this would be a decorator around the control
|
|
234 |
# files object that presents logical locks as physical ones - if this
|
|
235 |
# gets ugly consider that alternative design. RBC 20071011
|
|
236 |
self._write_lock_count = 0 |
|
237 |
self._transaction = None |
|
238 |
# for tests
|
|
239 |
self._reconcile_does_inventory_gc = True |
|
240 |
self._reconcile_fixes_text_parents = True |
|
241 |
self._reconcile_backsup_inventory = False |
|
242 |
||
243 |
||
244 |
class RepositoryFormatPackGCPlain(RepositoryFormatPackDevelopment0): |
|
245 |
"""A B+Tree index using pack repository.""" |
|
246 |
||
247 |
repository_class = GCPackRepository |
|
248 |
||
249 |
def get_format_string(self): |
|
250 |
"""See RepositoryFormat.get_format_string().""" |
|
251 |
return ("Bazaar development format - btree+gc " |
|
252 |
"(needs bzr.dev from 1.6)\n") |
|
253 |
||
254 |
def get_format_description(self): |
|
255 |
"""See RepositoryFormat.get_format_description().""" |
|
256 |
return ("Development repository format - btree+groupcompress " |
|
257 |
", interoperates with pack-0.92\n") |
|
258 |
||
259 |
||
260 |
class RepositoryFormatPackGCRichRoot(RepositoryFormatKnitPack4): |
|
261 |
"""A B+Tree index using pack repository.""" |
|
262 |
||
263 |
repository_class = GCPackRepository |
|
264 |
||
265 |
def get_format_string(self): |
|
266 |
"""See RepositoryFormat.get_format_string().""" |
|
267 |
return ("Bazaar development format - btree+gc-rich-root " |
|
268 |
"(needs bzr.dev from 1.6)\n") |
|
269 |
||
270 |
def get_format_description(self): |
|
271 |
"""See RepositoryFormat.get_format_description().""" |
|
272 |
return ("Development repository format - btree+groupcompress " |
|
273 |
", interoperates with rich-root-pack\n") |
|
274 |
||
275 |
||
276 |
class RepositoryFormatPackGCSubtrees(RepositoryFormatPackDevelopment0Subtree): |
|
277 |
"""A B+Tree index using pack repository.""" |
|
278 |
||
279 |
repository_class = GCPackRepository |
|
280 |
||
281 |
def get_format_string(self): |
|
282 |
"""See RepositoryFormat.get_format_string().""" |
|
283 |
return ("Bazaar development format - btree+gc-subtrees " |
|
284 |
"(needs bzr.dev from 1.6)\n") |
|
285 |
||
286 |
def get_format_description(self): |
|
287 |
"""See RepositoryFormat.get_format_description().""" |
|
288 |
return ("Development repository format - btree+groupcompress " |
|
289 |
", interoperates with pack-0.92-subtrees\n") |
|
290 |
||
291 |
||
292 |
def pack_incompatible(source, target, orig_method=InterPackRepo.is_compatible): |
|
293 |
formats = (RepositoryFormatPackGCPlain, RepositoryFormatPackGCRichRoot, |
|
294 |
RepositoryFormatPackGCSubtrees) |
|
|
0.17.10
by Robert Collins
Correct optimiser disabling. |
295 |
if isinstance(source._format, formats) or isinstance(target._format, formats): |
|
0.17.9
by Robert Collins
Initial stab at repository format support. |
296 |
return False |
297 |
else: |
|
298 |
return orig_method(source, target) |
|
299 |
||
300 |
||
301 |
InterPackRepo.is_compatible = staticmethod(pack_incompatible) |