Show More
@@ -1,222 +1,218 | |||||
1 | #!/usr/bin/env python |
|
1 | #!/usr/bin/env python | |
2 |
|
2 | |||
3 | """\ |
|
3 | """\ | |
4 | reorder a revlog (the manifest by default) to save space |
|
4 | reorder a revlog (the manifest by default) to save space | |
5 |
|
5 | |||
6 | Specifically, this topologically sorts the revisions in the revlog so that |
|
6 | Specifically, this topologically sorts the revisions in the revlog so that | |
7 | revisions on the same branch are adjacent as much as possible. This is a |
|
7 | revisions on the same branch are adjacent as much as possible. This is a | |
8 | workaround for the fact that Mercurial computes deltas relative to the |
|
8 | workaround for the fact that Mercurial computes deltas relative to the | |
9 | previous revision rather than relative to a parent revision. |
|
9 | previous revision rather than relative to a parent revision. | |
10 |
|
10 | |||
11 | This is *not* safe to run on a changelog. |
|
11 | This is *not* safe to run on a changelog. | |
12 | """ |
|
12 | """ | |
13 |
|
13 | |||
14 | # Originally written by Benoit Boissinot <benoit.boissinot at ens-lyon.org> |
|
14 | # Originally written by Benoit Boissinot <benoit.boissinot at ens-lyon.org> | |
15 | # as a patch to rewrite-log. Cleaned up, refactored, documented, and |
|
15 | # as a patch to rewrite-log. Cleaned up, refactored, documented, and | |
16 | # renamed by Greg Ward <greg at gerg.ca>. |
|
16 | # renamed by Greg Ward <greg at gerg.ca>. | |
17 |
|
17 | |||
18 | # XXX would be nice to have a way to verify the repository after shrinking, |
|
18 | # XXX would be nice to have a way to verify the repository after shrinking, | |
19 | # e.g. by comparing "before" and "after" states of random changesets |
|
19 | # e.g. by comparing "before" and "after" states of random changesets | |
20 | # (maybe: export before, shrink, export after, diff). |
|
20 | # (maybe: export before, shrink, export after, diff). | |
21 |
|
21 | |||
22 |
import |
|
22 | import os, tempfile | |
23 | import optparse |
|
23 | from mercurial import revlog, transaction, node, util | |
24 | from mercurial import ui as ui_, hg, revlog, transaction, node, util |
|
|||
25 | from mercurial import changegroup |
|
24 | from mercurial import changegroup | |
26 | from mercurial.i18n import _ |
|
25 | from mercurial.i18n import _ | |
27 |
|
26 | |||
28 | def toposort(ui, rl): |
|
27 | def toposort(ui, rl): | |
29 |
|
28 | |||
30 | children = {} |
|
29 | children = {} | |
31 | root = [] |
|
30 | root = [] | |
32 | # build children and roots |
|
31 | # build children and roots | |
33 | ui.status(_('reading revs\n')) |
|
32 | ui.status(_('reading revs\n')) | |
34 | try: |
|
33 | try: | |
35 | for i in rl: |
|
34 | for i in rl: | |
36 | ui.progress(_('reading'), i, total=len(rl)) |
|
35 | ui.progress(_('reading'), i, total=len(rl)) | |
37 | children[i] = [] |
|
36 | children[i] = [] | |
38 | parents = [p for p in rl.parentrevs(i) if p != node.nullrev] |
|
37 | parents = [p for p in rl.parentrevs(i) if p != node.nullrev] | |
39 | # in case of duplicate parents |
|
38 | # in case of duplicate parents | |
40 | if len(parents) == 2 and parents[0] == parents[1]: |
|
39 | if len(parents) == 2 and parents[0] == parents[1]: | |
41 | del parents[1] |
|
40 | del parents[1] | |
42 | for p in parents: |
|
41 | for p in parents: | |
43 | assert p in children |
|
42 | assert p in children | |
44 | children[p].append(i) |
|
43 | children[p].append(i) | |
45 |
|
44 | |||
46 | if len(parents) == 0: |
|
45 | if len(parents) == 0: | |
47 | root.append(i) |
|
46 | root.append(i) | |
48 | finally: |
|
47 | finally: | |
49 | ui.progress(_('reading'), None, total=len(rl)) |
|
48 | ui.progress(_('reading'), None, total=len(rl)) | |
50 |
|
49 | |||
51 | # XXX this is a reimplementation of the 'branchsort' topo sort |
|
50 | # XXX this is a reimplementation of the 'branchsort' topo sort | |
52 | # algorithm in hgext.convert.convcmd... would be nice not to duplicate |
|
51 | # algorithm in hgext.convert.convcmd... would be nice not to duplicate | |
53 | # the algorithm |
|
52 | # the algorithm | |
54 | ui.status(_('sorting revs\n')) |
|
53 | ui.status(_('sorting revs\n')) | |
55 | visit = root |
|
54 | visit = root | |
56 | ret = [] |
|
55 | ret = [] | |
57 | while visit: |
|
56 | while visit: | |
58 | i = visit.pop(0) |
|
57 | i = visit.pop(0) | |
59 | ret.append(i) |
|
58 | ret.append(i) | |
60 | if i not in children: |
|
59 | if i not in children: | |
61 | # This only happens if some node's p1 == p2, which can |
|
60 | # This only happens if some node's p1 == p2, which can | |
62 | # happen in the manifest in certain circumstances. |
|
61 | # happen in the manifest in certain circumstances. | |
63 | continue |
|
62 | continue | |
64 | next = [] |
|
63 | next = [] | |
65 | for c in children.pop(i): |
|
64 | for c in children.pop(i): | |
66 | parents_unseen = [p for p in rl.parentrevs(c) |
|
65 | parents_unseen = [p for p in rl.parentrevs(c) | |
67 | if p != node.nullrev and p in children] |
|
66 | if p != node.nullrev and p in children] | |
68 | if len(parents_unseen) == 0: |
|
67 | if len(parents_unseen) == 0: | |
69 | next.append(c) |
|
68 | next.append(c) | |
70 | visit = next + visit |
|
69 | visit = next + visit | |
71 | return ret |
|
70 | return ret | |
72 |
|
71 | |||
73 | def writerevs(ui, r1, r2, order, tr): |
|
72 | def writerevs(ui, r1, r2, order, tr): | |
74 |
|
73 | |||
75 | ui.status(_('writing revs\n')) |
|
74 | ui.status(_('writing revs\n')) | |
76 |
|
75 | |||
77 | count = [0] |
|
76 | count = [0] | |
78 | def progress(*args): |
|
77 | def progress(*args): | |
79 | ui.progress(_('writing'), count[0], total=len(order)) |
|
78 | ui.progress(_('writing'), count[0], total=len(order)) | |
80 | count[0] += 1 |
|
79 | count[0] += 1 | |
81 |
|
80 | |||
82 | order = [r1.node(r) for r in order] |
|
81 | order = [r1.node(r) for r in order] | |
83 |
|
82 | |||
84 | # this is a bit ugly, but it works |
|
83 | # this is a bit ugly, but it works | |
85 | lookup = lambda x: "%020d" % r1.linkrev(r1.rev(x)) |
|
84 | lookup = lambda x: "%020d" % r1.linkrev(r1.rev(x)) | |
86 | unlookup = lambda x: int(x, 10) |
|
85 | unlookup = lambda x: int(x, 10) | |
87 |
|
86 | |||
88 | try: |
|
87 | try: | |
89 | group = util.chunkbuffer(r1.group(order, lookup, progress)) |
|
88 | group = util.chunkbuffer(r1.group(order, lookup, progress)) | |
90 | chunkiter = changegroup.chunkiter(group) |
|
89 | chunkiter = changegroup.chunkiter(group) | |
91 | r2.addgroup(chunkiter, unlookup, tr) |
|
90 | r2.addgroup(chunkiter, unlookup, tr) | |
92 | finally: |
|
91 | finally: | |
93 | ui.progress(_('writing'), None, len(order)) |
|
92 | ui.progress(_('writing'), None, len(order)) | |
94 |
|
93 | |||
95 | def report(ui, olddatafn, newdatafn): |
|
94 | def report(ui, olddatafn, newdatafn): | |
96 | oldsize = float(os.stat(olddatafn).st_size) |
|
95 | oldsize = float(os.stat(olddatafn).st_size) | |
97 | newsize = float(os.stat(newdatafn).st_size) |
|
96 | newsize = float(os.stat(newdatafn).st_size) | |
98 |
|
97 | |||
99 | # argh: have to pass an int to %d, because a float >= 2^32 |
|
98 | # argh: have to pass an int to %d, because a float >= 2^32 | |
100 | # blows up under Python 2.5 or earlier |
|
99 | # blows up under Python 2.5 or earlier | |
101 | ui.write(_('old file size: %12d bytes (%6.1f MiB)\n') |
|
100 | ui.write(_('old file size: %12d bytes (%6.1f MiB)\n') | |
102 | % (int(oldsize), oldsize / 1024 / 1024)) |
|
101 | % (int(oldsize), oldsize / 1024 / 1024)) | |
103 | ui.write(_('new file size: %12d bytes (%6.1f MiB)\n') |
|
102 | ui.write(_('new file size: %12d bytes (%6.1f MiB)\n') | |
104 | % (int(newsize), newsize / 1024 / 1024)) |
|
103 | % (int(newsize), newsize / 1024 / 1024)) | |
105 |
|
104 | |||
106 | shrink_percent = (oldsize - newsize) / oldsize * 100 |
|
105 | shrink_percent = (oldsize - newsize) / oldsize * 100 | |
107 | shrink_factor = oldsize / newsize |
|
106 | shrink_factor = oldsize / newsize | |
108 | ui.write(_('shrinkage: %.1f%% (%.1fx)\n') |
|
107 | ui.write(_('shrinkage: %.1f%% (%.1fx)\n') | |
109 | % (shrink_percent, shrink_factor)) |
|
108 | % (shrink_percent, shrink_factor)) | |
110 |
|
109 | |||
111 | def shrink(ui, repo, **opts): |
|
110 | def shrink(ui, repo, **opts): | |
112 | """ |
|
111 | """ | |
113 | Shrink revlog by re-ordering revisions. Will operate on manifest for |
|
112 | Shrink revlog by re-ordering revisions. Will operate on manifest for | |
114 | the given repository if no other revlog is specified.""" |
|
113 | the given repository if no other revlog is specified.""" | |
115 |
|
114 | |||
116 | # Unbuffer stdout for nice progress output. |
|
|||
117 | sys.stdout = os.fdopen(sys.stdout.fileno(), 'w', 0) |
|
|||
118 |
|
||||
119 | if not repo.local(): |
|
115 | if not repo.local(): | |
120 | raise util.Abort(_('not a local repository: %s') % repo.root) |
|
116 | raise util.Abort(_('not a local repository: %s') % repo.root) | |
121 |
|
117 | |||
122 | fn = opts.get('revlog') |
|
118 | fn = opts.get('revlog') | |
123 | if not fn: |
|
119 | if not fn: | |
124 | indexfn = repo.sjoin('00manifest.i') |
|
120 | indexfn = repo.sjoin('00manifest.i') | |
125 | else: |
|
121 | else: | |
126 | if not fn.endswith('.i'): |
|
122 | if not fn.endswith('.i'): | |
127 | raise util.Abort(_('--revlog option must specify the revlog index ' |
|
123 | raise util.Abort(_('--revlog option must specify the revlog index ' | |
128 | 'file (*.i), not %s') % opts.get('revlog')) |
|
124 | 'file (*.i), not %s') % opts.get('revlog')) | |
129 |
|
125 | |||
130 | indexfn = os.path.realpath(fn) |
|
126 | indexfn = os.path.realpath(fn) | |
131 | store = repo.sjoin('') |
|
127 | store = repo.sjoin('') | |
132 | if not indexfn.startswith(store): |
|
128 | if not indexfn.startswith(store): | |
133 | raise util.Abort(_('--revlog option must specify a revlog in %s, ' |
|
129 | raise util.Abort(_('--revlog option must specify a revlog in %s, ' | |
134 | 'not %s') % (store, indexfn)) |
|
130 | 'not %s') % (store, indexfn)) | |
135 |
|
131 | |||
136 | datafn = indexfn[:-2] + '.d' |
|
132 | datafn = indexfn[:-2] + '.d' | |
137 | if not os.path.exists(indexfn): |
|
133 | if not os.path.exists(indexfn): | |
138 | raise util.Abort(_('no such file: %s') % indexfn) |
|
134 | raise util.Abort(_('no such file: %s') % indexfn) | |
139 | if '00changelog' in indexfn: |
|
135 | if '00changelog' in indexfn: | |
140 | raise util.Abort(_('shrinking the changelog ' |
|
136 | raise util.Abort(_('shrinking the changelog ' | |
141 | 'will corrupt your repository')) |
|
137 | 'will corrupt your repository')) | |
142 | if not os.path.exists(datafn): |
|
138 | if not os.path.exists(datafn): | |
143 | # This is just a lazy shortcut because I can't be bothered to |
|
139 | # This is just a lazy shortcut because I can't be bothered to | |
144 | # handle all the special cases that entail from no .d file. |
|
140 | # handle all the special cases that entail from no .d file. | |
145 | raise util.Abort(_('%s does not exist: revlog not big enough ' |
|
141 | raise util.Abort(_('%s does not exist: revlog not big enough ' | |
146 | 'to be worth shrinking') % datafn) |
|
142 | 'to be worth shrinking') % datafn) | |
147 |
|
143 | |||
148 | oldindexfn = indexfn + '.old' |
|
144 | oldindexfn = indexfn + '.old' | |
149 | olddatafn = datafn + '.old' |
|
145 | olddatafn = datafn + '.old' | |
150 | if os.path.exists(oldindexfn) or os.path.exists(olddatafn): |
|
146 | if os.path.exists(oldindexfn) or os.path.exists(olddatafn): | |
151 | raise util.Abort(_('one or both of\n' |
|
147 | raise util.Abort(_('one or both of\n' | |
152 | ' %s\n' |
|
148 | ' %s\n' | |
153 | ' %s\n' |
|
149 | ' %s\n' | |
154 | 'exists from a previous run; please clean up ' |
|
150 | 'exists from a previous run; please clean up ' | |
155 | 'before running again') % (oldindexfn, olddatafn)) |
|
151 | 'before running again') % (oldindexfn, olddatafn)) | |
156 |
|
152 | |||
157 | ui.write(_('shrinking %s\n') % indexfn) |
|
153 | ui.write(_('shrinking %s\n') % indexfn) | |
158 | prefix = os.path.basename(indexfn)[:-1] |
|
154 | prefix = os.path.basename(indexfn)[:-1] | |
159 | (tmpfd, tmpindexfn) = tempfile.mkstemp(dir=os.path.dirname(indexfn), |
|
155 | (tmpfd, tmpindexfn) = tempfile.mkstemp(dir=os.path.dirname(indexfn), | |
160 | prefix=prefix, |
|
156 | prefix=prefix, | |
161 | suffix='.i') |
|
157 | suffix='.i') | |
162 | tmpdatafn = tmpindexfn[:-2] + '.d' |
|
158 | tmpdatafn = tmpindexfn[:-2] + '.d' | |
163 | os.close(tmpfd) |
|
159 | os.close(tmpfd) | |
164 |
|
160 | |||
165 | r1 = revlog.revlog(util.opener(os.getcwd(), audit=False), indexfn) |
|
161 | r1 = revlog.revlog(util.opener(os.getcwd(), audit=False), indexfn) | |
166 | r2 = revlog.revlog(util.opener(os.getcwd(), audit=False), tmpindexfn) |
|
162 | r2 = revlog.revlog(util.opener(os.getcwd(), audit=False), tmpindexfn) | |
167 |
|
163 | |||
168 | # Don't use repo.transaction(), because then things get hairy with |
|
164 | # Don't use repo.transaction(), because then things get hairy with | |
169 | # paths: some need to be relative to .hg, and some need to be |
|
165 | # paths: some need to be relative to .hg, and some need to be | |
170 | # absolute. Doing it this way keeps things simple: everything is an |
|
166 | # absolute. Doing it this way keeps things simple: everything is an | |
171 | # absolute path. |
|
167 | # absolute path. | |
172 | lock = repo.lock(wait=False) |
|
168 | lock = repo.lock(wait=False) | |
173 | tr = transaction.transaction(ui.warn, |
|
169 | tr = transaction.transaction(ui.warn, | |
174 | open, |
|
170 | open, | |
175 | repo.sjoin('journal')) |
|
171 | repo.sjoin('journal')) | |
176 |
|
172 | |||
177 | try: |
|
173 | try: | |
178 | try: |
|
174 | try: | |
179 | order = toposort(ui, r1) |
|
175 | order = toposort(ui, r1) | |
180 | writerevs(ui, r1, r2, order, tr) |
|
176 | writerevs(ui, r1, r2, order, tr) | |
181 | report(ui, datafn, tmpdatafn) |
|
177 | report(ui, datafn, tmpdatafn) | |
182 | tr.close() |
|
178 | tr.close() | |
183 | except: |
|
179 | except: | |
184 | # Abort transaction first, so we truncate the files before |
|
180 | # Abort transaction first, so we truncate the files before | |
185 | # deleting them. |
|
181 | # deleting them. | |
186 | tr.abort() |
|
182 | tr.abort() | |
187 | if os.path.exists(tmpindexfn): |
|
183 | if os.path.exists(tmpindexfn): | |
188 | os.unlink(tmpindexfn) |
|
184 | os.unlink(tmpindexfn) | |
189 | if os.path.exists(tmpdatafn): |
|
185 | if os.path.exists(tmpdatafn): | |
190 | os.unlink(tmpdatafn) |
|
186 | os.unlink(tmpdatafn) | |
191 | raise |
|
187 | raise | |
192 | if not opts.get('dry_run'): |
|
188 | if not opts.get('dry_run'): | |
193 | # Racy since both files cannot be renamed atomically |
|
189 | # Racy since both files cannot be renamed atomically | |
194 | util.os_link(indexfn, oldindexfn) |
|
190 | util.os_link(indexfn, oldindexfn) | |
195 | util.os_link(datafn, olddatafn) |
|
191 | util.os_link(datafn, olddatafn) | |
196 | util.rename(tmpindexfn, indexfn) |
|
192 | util.rename(tmpindexfn, indexfn) | |
197 | util.rename(tmpdatafn, datafn) |
|
193 | util.rename(tmpdatafn, datafn) | |
198 | else: |
|
194 | else: | |
199 | os.unlink(tmpindexfn) |
|
195 | os.unlink(tmpindexfn) | |
200 | os.unlink(tmpdatafn) |
|
196 | os.unlink(tmpdatafn) | |
201 | finally: |
|
197 | finally: | |
202 | lock.release() |
|
198 | lock.release() | |
203 |
|
199 | |||
204 | if not opts.get('dry_run'): |
|
200 | if not opts.get('dry_run'): | |
205 | ui.write(_('note: old revlog saved in:\n' |
|
201 | ui.write(_('note: old revlog saved in:\n' | |
206 | ' %s\n' |
|
202 | ' %s\n' | |
207 | ' %s\n' |
|
203 | ' %s\n' | |
208 | '(You can delete those files when you are satisfied that your\n' |
|
204 | '(You can delete those files when you are satisfied that your\n' | |
209 | 'repository is still sane. ' |
|
205 | 'repository is still sane. ' | |
210 | 'Running \'hg verify\' is strongly recommended.)\n') |
|
206 | 'Running \'hg verify\' is strongly recommended.)\n') | |
211 | % (oldindexfn, olddatafn)) |
|
207 | % (oldindexfn, olddatafn)) | |
212 |
|
208 | |||
213 | cmdtable = { |
|
209 | cmdtable = { | |
214 | 'shrink': (shrink, |
|
210 | 'shrink': (shrink, | |
215 | [('', 'revlog', '', _('index (.i) file of the revlog to shrink')), |
|
211 | [('', 'revlog', '', _('index (.i) file of the revlog to shrink')), | |
216 | ('n', 'dry-run', None, _('do not shrink, simulate only')), |
|
212 | ('n', 'dry-run', None, _('do not shrink, simulate only')), | |
217 | ], |
|
213 | ], | |
218 | _('hg shrink [--revlog PATH]')) |
|
214 | _('hg shrink [--revlog PATH]')) | |
219 | } |
|
215 | } | |
220 |
|
216 | |||
221 | if __name__ == "__main__": |
|
217 | if __name__ == "__main__": | |
222 | print "shrink-revlog.py is now an extension (see hg help extensions)" |
|
218 | print "shrink-revlog.py is now an extension (see hg help extensions)" |
General Comments 0
You need to be logged in to leave comments.
Login now