annotate mercurial/revlogutils/nodemap.py @ 44367:50ad851efd9b

nodemap: introduce append-only incremental update of the persistent data Rewriting the full nodemap for each transaction has a cost we would like to avoid. We introduce a new way to write persistent nodemap data by adding new information at the end for file. Any new and updated block as added at the end of the file. The last block is the new root node. With this method, some of the block already on disk get "dereferenced" and become dead data. In later changesets, We'll start tracking the amount of dead data to eventually re-generate a full nodemap. Differential Revision: https://phab.mercurial-scm.org/D7886
author Pierre-Yves David <pierre-yves.david@octobus.net>
date Wed, 15 Jan 2020 15:49:54 +0100
parents 76a96e3a2bbb
children e41a164db7a9
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
44035
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
1 # nodemap.py - nodemap related code and utilities
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
2 #
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
3 # Copyright 2019 Pierre-Yves David <pierre-yves.david@octobus.net>
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
4 # Copyright 2019 George Racinet <georges.racinet@octobus.net>
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
5 #
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
6 # This software may be used and distributed according to the terms of the
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
7 # GNU General Public License version 2 or any later version.
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
8
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
9 from __future__ import absolute_import
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
10
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
11 import os
44355
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
12 import re
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
13 import struct
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
14
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
15 from .. import (
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
16 error,
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
17 node as nodemod,
44357
7f4f7ef3133e nodemap: add a optional `nodemap_add_full` method on indexes
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44355
diff changeset
18 util,
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
19 )
44035
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
20
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
21
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
22 class NodeMap(dict):
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
23 def __missing__(self, x):
ab595920de0e revlogutils: move the NodeMap class in a dedicated nodemap module
Pierre-Yves David <pierre-yves.david@octobus.net>
parents:
diff changeset
24 raise error.RevlogError(b'unknown node: %s' % x)
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
25
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
26
44352
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
27 def persisted_data(revlog):
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
28 """read the nodemap for a revlog from disk"""
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
29 if revlog.nodemap_file is None:
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
30 return None
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
31 pdata = revlog.opener.tryread(revlog.nodemap_file)
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
32 if not pdata:
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
33 return None
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
34 offset = 0
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
35 (version,) = S_VERSION.unpack(pdata[offset : offset + S_VERSION.size])
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
36 if version != ONDISK_VERSION:
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
37 return None
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
38 offset += S_VERSION.size
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
39 (uid_size,) = S_HEADER.unpack(pdata[offset : offset + S_HEADER.size])
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
40 offset += S_HEADER.size
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
41 docket = NodeMapDocket(pdata[offset : offset + uid_size])
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
42
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
43 filename = _rawdata_filepath(revlog, docket)
44366
76a96e3a2bbb nodemap: keep track of the docket for loaded data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44365
diff changeset
44 return docket, revlog.opener.tryread(filename)
44352
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
45
6c07480d6659 nodemap: add a function to read the data from disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44351
diff changeset
46
44351
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
47 def setup_persistent_nodemap(tr, revlog):
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
48 """Install whatever is needed transaction side to persist a nodemap on disk
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
49
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
50 (only actually persist the nodemap if this is relevant for this revlog)
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
51 """
44353
daad3aace942 nodemap: only use persistent nodemap for non-inlined revlog
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44352
diff changeset
52 if revlog._inline:
daad3aace942 nodemap: only use persistent nodemap for non-inlined revlog
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44352
diff changeset
53 return # inlined revlog are too small for this to be relevant
44351
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
54 if revlog.nodemap_file is None:
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
55 return # we do not use persistent_nodemap on this revlog
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
56 callback_id = b"revlog-persistent-nodemap-%s" % revlog.nodemap_file
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
57 if tr.hasfinalize(callback_id):
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
58 return # no need to register again
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
59 tr.addfinalize(callback_id, lambda tr: _persist_nodemap(tr, revlog))
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
60
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
61
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
62 def _persist_nodemap(tr, revlog):
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
63 """Write nodemap data on disk for a given revlog
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
64 """
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
65 if getattr(revlog, 'filteredrevs', ()):
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
66 raise error.ProgrammingError(
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
67 "cannot persist nodemap of a filtered changelog"
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
68 )
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
69 if revlog.nodemap_file is None:
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
70 msg = "calling persist nodemap on a revlog without the feature enableb"
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
71 raise error.ProgrammingError(msg)
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
72
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
73 can_incremental = util.safehasattr(revlog.index, "nodemap_data_incremental")
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
74 ondisk_docket = revlog._nodemap_docket
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
75
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
76 # first attemp an incremental update of the data
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
77 if can_incremental and ondisk_docket is not None:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
78 target_docket = revlog._nodemap_docket.copy()
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
79 data = revlog.index.nodemap_data_incremental()
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
80 datafile = _rawdata_filepath(revlog, target_docket)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
81 # EXP-TODO: if this is a cache, this should use a cache vfs, not a
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
82 # store vfs
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
83 with revlog.opener(datafile, b'a') as fd:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
84 fd.write(data)
44357
7f4f7ef3133e nodemap: add a optional `nodemap_add_full` method on indexes
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44355
diff changeset
85 else:
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
86 # otherwise fallback to a full new export
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
87 target_docket = NodeMapDocket()
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
88 datafile = _rawdata_filepath(revlog, target_docket)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
89 if util.safehasattr(revlog.index, "nodemap_data_all"):
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
90 data = revlog.index.nodemap_data_all()
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
91 else:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
92 data = persistent_data(revlog.index)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
93 # EXP-TODO: if this is a cache, this should use a cache vfs, not a
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
94 # store vfs
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
95 with revlog.opener(datafile, b'w') as fd:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
96 fd.write(data)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
97 # EXP-TODO: if this is a cache, this should use a cache vfs, not a
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
98 # store vfs
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
99 with revlog.opener(revlog.nodemap_file, b'w', atomictemp=True) as fp:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
100 fp.write(target_docket.serialize())
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
101 revlog._nodemap_docket = target_docket
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
102 # EXP-TODO: if the transaction abort, we should remove the new data and
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
103 # reinstall the old one.
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
104
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
105 # search for old index file in all cases, some older process might have
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
106 # left one behind.
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
107 olds = _other_rawdata_filepath(revlog, target_docket)
44355
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
108 if olds:
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
109 realvfs = getattr(revlog, '_realopener', revlog.opener)
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
110
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
111 def cleanup(tr):
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
112 for oldfile in olds:
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
113 realvfs.tryunlink(oldfile)
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
114
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
115 callback_id = b"revlog-cleanup-nodemap-%s" % revlog.nodemap_file
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
116 tr.addpostclose(callback_id, cleanup)
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
117
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
118
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
119 ### Nodemap docket file
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
120 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
121 # The nodemap data are stored on disk using 2 files:
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
122 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
123 # * a raw data files containing a persistent nodemap
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
124 # (see `Nodemap Trie` section)
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
125 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
126 # * a small "docket" file containing medatadata
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
127 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
128 # While the nodemap data can be multiple tens of megabytes, the "docket" is
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
129 # small, it is easy to update it automatically or to duplicated its content
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
130 # during a transaction.
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
131 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
132 # Multiple raw data can exist at the same time (The currently valid one and a
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
133 # new one beind used by an in progress transaction). To accomodate this, the
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
134 # filename hosting the raw data has a variable parts. The exact filename is
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
135 # specified inside the "docket" file.
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
136 #
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
137 # The docket file contains information to find, qualify and validate the raw
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
138 # data. Its content is currently very light, but it will expand as the on disk
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
139 # nodemap gains the necessary features to be used in production.
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
140
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
141 # version 0 is experimental, no BC garantee, do no use outside of tests.
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
142 ONDISK_VERSION = 0
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
143
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
144 S_VERSION = struct.Struct(">B")
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
145 S_HEADER = struct.Struct(">B")
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
146
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
147 ID_SIZE = 8
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
148
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
149
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
150 def _make_uid():
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
151 """return a new unique identifier.
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
152
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
153 The identifier is random and composed of ascii characters."""
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
154 return nodemod.hex(os.urandom(ID_SIZE))
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
155
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
156
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
157 class NodeMapDocket(object):
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
158 """metadata associated with persistent nodemap data
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
159
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
160 The persistent data may come from disk or be on their way to disk.
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
161 """
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
162
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
163 def __init__(self, uid=None):
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
164 if uid is None:
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
165 uid = _make_uid()
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
166 self.uid = uid
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
167
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
168 def copy(self):
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
169 return NodeMapDocket(uid=self.uid)
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
170
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
171 def serialize(self):
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
172 """return serialized bytes for a docket using the passed uid"""
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
173 data = []
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
174 data.append(S_VERSION.pack(ONDISK_VERSION))
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
175 data.append(S_HEADER.pack(len(self.uid)))
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
176 data.append(self.uid)
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
177 return b''.join(data)
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
178
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
179
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
180 def _rawdata_filepath(revlog, docket):
44354
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
181 """The (vfs relative) nodemap's rawdata file for a given uid"""
2b72c4ff8ed1 nodemap: use an intermediate "docket" file to carry small metadata
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44353
diff changeset
182 prefix = revlog.nodemap_file[:-2]
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
183 return b"%s-%s.nd" % (prefix, docket.uid)
44351
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
184
5962fd0d1045 nodemap: write nodemap data on disk
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44350
diff changeset
185
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
186 def _other_rawdata_filepath(revlog, docket):
44355
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
187 prefix = revlog.nodemap_file[:-2]
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
188 pattern = re.compile(b"(^|/)%s-[0-9a-f]+\.nd$" % prefix)
44365
72c15641c8b4 nodemap: introduce an explicit class/object for the docket
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44364
diff changeset
189 new_file_path = _rawdata_filepath(revlog, docket)
44355
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
190 new_file_name = revlog.opener.basename(new_file_path)
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
191 dirpath = revlog.opener.dirname(new_file_path)
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
192 others = []
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
193 for f in revlog.opener.listdir(dirpath):
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
194 if pattern.match(f) and f != new_file_name:
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
195 others.append(f)
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
196 return others
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
197
563dfdfd01a4 nodemap: delete older raw data file when creating a new ones
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44354
diff changeset
198
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
199 ### Nodemap Trie
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
200 #
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
201 # This is a simple reference implementation to compute and persist a nodemap
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
202 # trie. This reference implementation is write only. The python version of this
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
203 # is not expected to be actually used, since it wont provide performance
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
204 # improvement over existing non-persistent C implementation.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
205 #
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
206 # The nodemap is persisted as Trie using 4bits-address/16-entries block. each
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
207 # revision can be adressed using its node shortest prefix.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
208 #
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
209 # The trie is stored as a sequence of block. Each block contains 16 entries
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
210 # (signed 64bit integer, big endian). Each entry can be one of the following:
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
211 #
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
212 # * value >= 0 -> index of sub-block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
213 # * value == -1 -> no value
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
214 # * value < -1 -> a revision value: rev = -(value+10)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
215 #
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
216 # The implementation focus on simplicity, not on performance. A Rust
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
217 # implementation should provide a efficient version of the same binary
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
218 # persistence. This reference python implementation is never meant to be
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
219 # extensively use in production.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
220
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
221
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
222 def persistent_data(index):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
223 """return the persistent binary form for a nodemap for a given index
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
224 """
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
225 trie = _build_trie(index)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
226 return _persist_trie(trie)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
227
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
228
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
229 def update_persistent_data(index, root, max_idx, last_rev):
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
230 """return the incremental update for persistent nodemap from a given index
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
231 """
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
232 trie = _update_trie(index, root, last_rev)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
233 return _persist_trie(trie, existing_idx=max_idx)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
234
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
235
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
236 S_BLOCK = struct.Struct(">" + ("l" * 16))
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
237
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
238 NO_ENTRY = -1
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
239 # rev 0 need to be -2 because 0 is used by block, -1 is a special value.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
240 REV_OFFSET = 2
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
241
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
242
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
243 def _transform_rev(rev):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
244 """Return the number used to represent the rev in the tree.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
245
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
246 (or retrieve a rev number from such representation)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
247
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
248 Note that this is an involution, a function equal to its inverse (i.e.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
249 which gives the identity when applied to itself).
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
250 """
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
251 return -(rev + REV_OFFSET)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
252
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
253
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
254 def _to_int(hex_digit):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
255 """turn an hexadecimal digit into a proper integer"""
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
256 return int(hex_digit, 16)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
257
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
258
44358
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
259 class Block(dict):
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
260 """represent a block of the Trie
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
261
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
262 contains up to 16 entry indexed from 0 to 15"""
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
263
44364
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
264 def __init__(self):
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
265 super(Block, self).__init__()
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
266 # If this block exist on disk, here is its ID
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
267 self.ondisk_id = None
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
268
44359
55b12f2593c1 nodemap: move the iteratio inside the Block object
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44358
diff changeset
269 def __iter__(self):
55b12f2593c1 nodemap: move the iteratio inside the Block object
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44358
diff changeset
270 return iter(self.get(i) for i in range(16))
55b12f2593c1 nodemap: move the iteratio inside the Block object
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44358
diff changeset
271
44358
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
272
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
273 def _build_trie(index):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
274 """build a nodemap trie
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
275
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
276 The nodemap stores revision number for each unique prefix.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
277
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
278 Each block is a dictionary with keys in `[0, 15]`. Values are either
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
279 another block or a revision number.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
280 """
44358
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
281 root = Block()
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
282 for rev in range(len(index)):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
283 hex = nodemod.hex(index[rev][7])
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
284 _insert_into_block(index, 0, root, rev, hex)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
285 return root
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
286
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
287
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
288 def _update_trie(index, root, last_rev):
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
289 """consume"""
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
290 for rev in range(last_rev + 1, len(index)):
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
291 hex = nodemod.hex(index[rev][7])
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
292 _insert_into_block(index, 0, root, rev, hex)
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
293 return root
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
294
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
295
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
296 def _insert_into_block(index, level, block, current_rev, current_hex):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
297 """insert a new revision in a block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
298
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
299 index: the index we are adding revision for
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
300 level: the depth of the current block in the trie
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
301 block: the block currently being considered
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
302 current_rev: the revision number we are adding
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
303 current_hex: the hexadecimal representation of the of that revision
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
304 """
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
305 if block.ondisk_id is not None:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
306 block.ondisk_id = None
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
307 hex_digit = _to_int(current_hex[level : level + 1])
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
308 entry = block.get(hex_digit)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
309 if entry is None:
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
310 # no entry, simply store the revision number
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
311 block[hex_digit] = current_rev
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
312 elif isinstance(entry, dict):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
313 # need to recurse to an underlying block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
314 _insert_into_block(index, level + 1, entry, current_rev, current_hex)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
315 else:
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
316 # collision with a previously unique prefix, inserting new
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
317 # vertices to fit both entry.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
318 other_hex = nodemod.hex(index[entry][7])
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
319 other_rev = entry
44358
7762a295fd4d nodemap: use an explicit "Block" object in the reference implementation
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44357
diff changeset
320 new = Block()
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
321 block[hex_digit] = new
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
322 _insert_into_block(index, level + 1, new, other_rev, other_hex)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
323 _insert_into_block(index, level + 1, new, current_rev, current_hex)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
324
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
325
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
326 def _persist_trie(root, existing_idx=None):
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
327 """turn a nodemap trie into persistent binary data
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
328
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
329 See `_build_trie` for nodemap trie structure"""
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
330 block_map = {}
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
331 if existing_idx is not None:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
332 base_idx = existing_idx + 1
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
333 else:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
334 base_idx = 0
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
335 chunks = []
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
336 for tn in _walk_trie(root):
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
337 if tn.ondisk_id is not None:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
338 block_map[id(tn)] = tn.ondisk_id
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
339 else:
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
340 block_map[id(tn)] = len(chunks) + base_idx
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
341 chunks.append(_persist_block(tn, block_map))
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
342 return b''.join(chunks)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
343
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
344
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
345 def _walk_trie(block):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
346 """yield all the block in a trie
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
347
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
348 Children blocks are always yield before their parent block.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
349 """
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
350 for (_, item) in sorted(block.items()):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
351 if isinstance(item, dict):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
352 for sub_block in _walk_trie(item):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
353 yield sub_block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
354 yield block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
355
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
356
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
357 def _persist_block(block_node, block_map):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
358 """produce persistent binary data for a single block
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
359
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
360 Children block are assumed to be already persisted and present in
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
361 block_map.
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
362 """
44359
55b12f2593c1 nodemap: move the iteratio inside the Block object
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44358
diff changeset
363 data = tuple(_to_value(v, block_map) for v in block_node)
44350
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
364 return S_BLOCK.pack(*data)
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
365
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
366
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
367 def _to_value(item, block_map):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
368 """persist any value as an integer"""
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
369 if item is None:
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
370 return NO_ENTRY
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
371 elif isinstance(item, dict):
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
372 return block_map[id(item)]
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
373 else:
c577bb4a04d4 nodemap: have some python code writing a nodemap in persistent binary form
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44035
diff changeset
374 return _transform_rev(item)
44360
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
375
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
376
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
377 def parse_data(data):
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
378 """parse parse nodemap data into a nodemap Trie"""
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
379 if (len(data) % S_BLOCK.size) != 0:
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
380 msg = "nodemap data size is not a multiple of block size (%d): %d"
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
381 raise error.Abort(msg % (S_BLOCK.size, len(data)))
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
382 if not data:
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
383 return Block(), None
44360
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
384 block_map = {}
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
385 new_blocks = []
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
386 for i in range(0, len(data), S_BLOCK.size):
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
387 block = Block()
44364
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
388 block.ondisk_id = len(block_map)
f0862ee1a31e nodemap: keep track of the ondisk id of nodemap blocks
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44362
diff changeset
389 block_map[block.ondisk_id] = block
44360
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
390 block_data = data[i : i + S_BLOCK.size]
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
391 values = S_BLOCK.unpack(block_data)
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
392 new_blocks.append((block, values))
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
393 for b, values in new_blocks:
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
394 for idx, v in enumerate(values):
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
395 if v == NO_ENTRY:
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
396 continue
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
397 elif v >= 0:
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
398 b[idx] = block_map[v]
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
399 else:
78721bbdb2ab nodemap: code to parse the persistent binary nodemap data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44359
diff changeset
400 b[idx] = _transform_rev(v)
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
401 return block, i // S_BLOCK.size
44361
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
402
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
403
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
404 # debug utility
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
405
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
406
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
407 def check_data(ui, index, data):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
408 """verify that the provided nodemap data are valid for the given idex"""
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
409 ret = 0
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
410 ui.status((b"revision in index: %d\n") % len(index))
44367
50ad851efd9b nodemap: introduce append-only incremental update of the persistent data
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44366
diff changeset
411 root, __ = parse_data(data)
44361
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
412 all_revs = set(_all_revisions(root))
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
413 ui.status((b"revision in nodemap: %d\n") % len(all_revs))
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
414 for r in range(len(index)):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
415 if r not in all_revs:
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
416 msg = b" revision missing from nodemap: %d\n" % r
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
417 ui.write_err(msg)
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
418 ret = 1
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
419 else:
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
420 all_revs.remove(r)
44362
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
421 nm_rev = _find_node(root, nodemod.hex(index[r][7]))
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
422 if nm_rev is None:
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
423 msg = b" revision node does not match any entries: %d\n" % r
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
424 ui.write_err(msg)
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
425 ret = 1
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
426 elif nm_rev != r:
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
427 msg = (
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
428 b" revision node does not match the expected revision: "
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
429 b"%d != %d\n" % (r, nm_rev)
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
430 )
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
431 ui.write_err(msg)
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
432 ret = 1
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
433
44361
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
434 if all_revs:
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
435 for r in sorted(all_revs):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
436 msg = b" extra revision in nodemap: %d\n" % r
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
437 ui.write_err(msg)
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
438 ret = 1
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
439 return ret
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
440
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
441
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
442 def _all_revisions(root):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
443 """return all revisions stored in a Trie"""
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
444 for block in _walk_trie(root):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
445 for v in block:
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
446 if v is None or isinstance(v, Block):
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
447 continue
20e125cdd719 nodemap: add basic checking of the on disk nodemap content
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44360
diff changeset
448 yield v
44362
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
449
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
450
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
451 def _find_node(block, node):
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
452 """find the revision associated with a given node"""
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
453 entry = block.get(_to_int(node[0:1]))
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
454 if isinstance(entry, dict):
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
455 return _find_node(entry, node[1:])
d58206b70199 nodemap: all check that revision and nodes match in the nodemap
Pierre-Yves David <pierre-yves.david@octobus.net>
parents: 44361
diff changeset
456 return entry