Show More
@@ -1,288 +1,287 b'' | |||||
1 | """\ |
|
1 | """reorder a revlog (the manifest by default) to save space | |
2 | reorder a revlog (the manifest by default) to save space |
|
|||
3 |
|
2 | |||
4 | Specifically, this topologically sorts the revisions in the revlog so that |
|
3 | Specifically, this topologically sorts the revisions in the revlog so that | |
5 | revisions on the same branch are adjacent as much as possible. This is a |
|
4 | revisions on the same branch are adjacent as much as possible. This is a | |
6 | workaround for the fact that Mercurial computes deltas relative to the |
|
5 | workaround for the fact that Mercurial computes deltas relative to the | |
7 | previous revision rather than relative to a parent revision. |
|
6 | previous revision rather than relative to a parent revision. | |
8 |
|
7 | |||
9 | This is *not* safe to run on a changelog. |
|
8 | This is *not* safe to run on a changelog. | |
10 | """ |
|
9 | """ | |
11 |
|
10 | |||
12 | # Originally written by Benoit Boissinot <benoit.boissinot at ens-lyon.org> |
|
11 | # Originally written by Benoit Boissinot <benoit.boissinot at ens-lyon.org> | |
13 | # as a patch to rewrite-log. Cleaned up, refactored, documented, and |
|
12 | # as a patch to rewrite-log. Cleaned up, refactored, documented, and | |
14 | # renamed by Greg Ward <greg at gerg.ca>. |
|
13 | # renamed by Greg Ward <greg at gerg.ca>. | |
15 |
|
14 | |||
16 | # XXX would be nice to have a way to verify the repository after shrinking, |
|
15 | # XXX would be nice to have a way to verify the repository after shrinking, | |
17 | # e.g. by comparing "before" and "after" states of random changesets |
|
16 | # e.g. by comparing "before" and "after" states of random changesets | |
18 | # (maybe: export before, shrink, export after, diff). |
|
17 | # (maybe: export before, shrink, export after, diff). | |
19 |
|
18 | |||
20 | import os, tempfile, errno |
|
19 | import os, tempfile, errno | |
21 | from mercurial import revlog, transaction, node, util |
|
20 | from mercurial import revlog, transaction, node, util | |
22 | from mercurial import changegroup |
|
21 | from mercurial import changegroup | |
23 | from mercurial.i18n import _ |
|
22 | from mercurial.i18n import _ | |
24 |
|
23 | |||
25 |
|
24 | |||
26 | def postorder(start, edges): |
|
25 | def postorder(start, edges): | |
27 | result = [] |
|
26 | result = [] | |
28 | visit = list(start) |
|
27 | visit = list(start) | |
29 | finished = set() |
|
28 | finished = set() | |
30 |
|
29 | |||
31 | while visit: |
|
30 | while visit: | |
32 | cur = visit[-1] |
|
31 | cur = visit[-1] | |
33 | for p in edges[cur]: |
|
32 | for p in edges[cur]: | |
34 | if p not in finished: |
|
33 | if p not in finished: | |
35 | visit.append(p) |
|
34 | visit.append(p) | |
36 | break |
|
35 | break | |
37 | else: |
|
36 | else: | |
38 | result.append(cur) |
|
37 | result.append(cur) | |
39 | finished.add(cur) |
|
38 | finished.add(cur) | |
40 | visit.pop() |
|
39 | visit.pop() | |
41 |
|
40 | |||
42 | return result |
|
41 | return result | |
43 |
|
42 | |||
44 | def toposort_reversepostorder(ui, rl): |
|
43 | def toposort_reversepostorder(ui, rl): | |
45 | # postorder of the reverse directed graph |
|
44 | # postorder of the reverse directed graph | |
46 |
|
45 | |||
47 | # map rev to list of parent revs (p2 first) |
|
46 | # map rev to list of parent revs (p2 first) | |
48 | parents = {} |
|
47 | parents = {} | |
49 | heads = set() |
|
48 | heads = set() | |
50 | ui.status(_('reading revs\n')) |
|
49 | ui.status(_('reading revs\n')) | |
51 | try: |
|
50 | try: | |
52 | for rev in rl: |
|
51 | for rev in rl: | |
53 | ui.progress(_('reading'), rev, total=len(rl)) |
|
52 | ui.progress(_('reading'), rev, total=len(rl)) | |
54 | (p1, p2) = rl.parentrevs(rev) |
|
53 | (p1, p2) = rl.parentrevs(rev) | |
55 | if p1 == p2 == node.nullrev: |
|
54 | if p1 == p2 == node.nullrev: | |
56 | parents[rev] = () # root node |
|
55 | parents[rev] = () # root node | |
57 | elif p1 == p2 or p2 == node.nullrev: |
|
56 | elif p1 == p2 or p2 == node.nullrev: | |
58 | parents[rev] = (p1,) # normal node |
|
57 | parents[rev] = (p1,) # normal node | |
59 | else: |
|
58 | else: | |
60 | parents[rev] = (p2, p1) # merge node |
|
59 | parents[rev] = (p2, p1) # merge node | |
61 | heads.add(rev) |
|
60 | heads.add(rev) | |
62 | for p in parents[rev]: |
|
61 | for p in parents[rev]: | |
63 | heads.discard(p) |
|
62 | heads.discard(p) | |
64 | finally: |
|
63 | finally: | |
65 | ui.progress(_('reading'), None) |
|
64 | ui.progress(_('reading'), None) | |
66 |
|
65 | |||
67 | heads = list(heads) |
|
66 | heads = list(heads) | |
68 | heads.sort(reverse=True) |
|
67 | heads.sort(reverse=True) | |
69 |
|
68 | |||
70 | ui.status(_('sorting revs\n')) |
|
69 | ui.status(_('sorting revs\n')) | |
71 | return postorder(heads, parents) |
|
70 | return postorder(heads, parents) | |
72 |
|
71 | |||
73 | def toposort_postorderreverse(ui, rl): |
|
72 | def toposort_postorderreverse(ui, rl): | |
74 | # reverse-postorder of the reverse directed graph |
|
73 | # reverse-postorder of the reverse directed graph | |
75 |
|
74 | |||
76 | children = {} |
|
75 | children = {} | |
77 | roots = set() |
|
76 | roots = set() | |
78 | ui.status(_('reading revs\n')) |
|
77 | ui.status(_('reading revs\n')) | |
79 | try: |
|
78 | try: | |
80 | for rev in rl: |
|
79 | for rev in rl: | |
81 | ui.progress(_('reading'), rev, total=len(rl)) |
|
80 | ui.progress(_('reading'), rev, total=len(rl)) | |
82 | (p1, p2) = rl.parentrevs(rev) |
|
81 | (p1, p2) = rl.parentrevs(rev) | |
83 | if p1 == p2 == node.nullrev: |
|
82 | if p1 == p2 == node.nullrev: | |
84 | roots.add(rev) |
|
83 | roots.add(rev) | |
85 | children[rev] = [] |
|
84 | children[rev] = [] | |
86 | if p1 != node.nullrev: |
|
85 | if p1 != node.nullrev: | |
87 | children[p1].append(rev) |
|
86 | children[p1].append(rev) | |
88 | if p2 != node.nullrev: |
|
87 | if p2 != node.nullrev: | |
89 | children[p2].append(rev) |
|
88 | children[p2].append(rev) | |
90 | finally: |
|
89 | finally: | |
91 | ui.progress(_('reading'), None) |
|
90 | ui.progress(_('reading'), None) | |
92 |
|
91 | |||
93 | roots = list(roots) |
|
92 | roots = list(roots) | |
94 | roots.sort() |
|
93 | roots.sort() | |
95 |
|
94 | |||
96 | ui.status(_('sorting revs\n')) |
|
95 | ui.status(_('sorting revs\n')) | |
97 | result = postorder(roots, children) |
|
96 | result = postorder(roots, children) | |
98 | result.reverse() |
|
97 | result.reverse() | |
99 | return result |
|
98 | return result | |
100 |
|
99 | |||
101 | def writerevs(ui, r1, r2, order, tr): |
|
100 | def writerevs(ui, r1, r2, order, tr): | |
102 |
|
101 | |||
103 | ui.status(_('writing revs\n')) |
|
102 | ui.status(_('writing revs\n')) | |
104 |
|
103 | |||
105 |
|
104 | |||
106 | order = [r1.node(r) for r in order] |
|
105 | order = [r1.node(r) for r in order] | |
107 |
|
106 | |||
108 | # this is a bit ugly, but it works |
|
107 | # this is a bit ugly, but it works | |
109 | count = [0] |
|
108 | count = [0] | |
110 | def lookup(x): |
|
109 | def lookup(x): | |
111 | count[0] += 1 |
|
110 | count[0] += 1 | |
112 | ui.progress(_('writing'), count[0], total=len(order)) |
|
111 | ui.progress(_('writing'), count[0], total=len(order)) | |
113 | return "%020d" % r1.linkrev(r1.rev(x)) |
|
112 | return "%020d" % r1.linkrev(r1.rev(x)) | |
114 |
|
113 | |||
115 | unlookup = lambda x: int(x, 10) |
|
114 | unlookup = lambda x: int(x, 10) | |
116 |
|
115 | |||
117 | try: |
|
116 | try: | |
118 | group = util.chunkbuffer(r1.group(order, lookup, progress)) |
|
117 | group = util.chunkbuffer(r1.group(order, lookup, progress)) | |
119 | group = changegroup.unbundle10(group, "UN") |
|
118 | group = changegroup.unbundle10(group, "UN") | |
120 | r2.addgroup(group, unlookup, tr) |
|
119 | r2.addgroup(group, unlookup, tr) | |
121 | finally: |
|
120 | finally: | |
122 | ui.progress(_('writing'), None) |
|
121 | ui.progress(_('writing'), None) | |
123 |
|
122 | |||
124 | def report(ui, r1, r2): |
|
123 | def report(ui, r1, r2): | |
125 | def getsize(r): |
|
124 | def getsize(r): | |
126 | s = 0 |
|
125 | s = 0 | |
127 | for fn in (r.indexfile, r.datafile): |
|
126 | for fn in (r.indexfile, r.datafile): | |
128 | try: |
|
127 | try: | |
129 | s += os.stat(fn).st_size |
|
128 | s += os.stat(fn).st_size | |
130 | except OSError, inst: |
|
129 | except OSError, inst: | |
131 | if inst.errno != errno.ENOENT: |
|
130 | if inst.errno != errno.ENOENT: | |
132 | raise |
|
131 | raise | |
133 | return s |
|
132 | return s | |
134 |
|
133 | |||
135 | oldsize = float(getsize(r1)) |
|
134 | oldsize = float(getsize(r1)) | |
136 | newsize = float(getsize(r2)) |
|
135 | newsize = float(getsize(r2)) | |
137 |
|
136 | |||
138 | # argh: have to pass an int to %d, because a float >= 2^32 |
|
137 | # argh: have to pass an int to %d, because a float >= 2^32 | |
139 | # blows up under Python 2.5 or earlier |
|
138 | # blows up under Python 2.5 or earlier | |
140 | ui.write(_('old file size: %12d bytes (%6.1f MiB)\n') |
|
139 | ui.write(_('old file size: %12d bytes (%6.1f MiB)\n') | |
141 | % (int(oldsize), oldsize / 1024 / 1024)) |
|
140 | % (int(oldsize), oldsize / 1024 / 1024)) | |
142 | ui.write(_('new file size: %12d bytes (%6.1f MiB)\n') |
|
141 | ui.write(_('new file size: %12d bytes (%6.1f MiB)\n') | |
143 | % (int(newsize), newsize / 1024 / 1024)) |
|
142 | % (int(newsize), newsize / 1024 / 1024)) | |
144 |
|
143 | |||
145 | shrink_percent = (oldsize - newsize) / oldsize * 100 |
|
144 | shrink_percent = (oldsize - newsize) / oldsize * 100 | |
146 | shrink_factor = oldsize / newsize |
|
145 | shrink_factor = oldsize / newsize | |
147 | ui.write(_('shrinkage: %.1f%% (%.1fx)\n') |
|
146 | ui.write(_('shrinkage: %.1f%% (%.1fx)\n') | |
148 | % (shrink_percent, shrink_factor)) |
|
147 | % (shrink_percent, shrink_factor)) | |
149 |
|
148 | |||
150 | def shrink(ui, repo, **opts): |
|
149 | def shrink(ui, repo, **opts): | |
151 | """shrink a revlog by reordering revisions |
|
150 | """shrink a revlog by reordering revisions | |
152 |
|
151 | |||
153 | Rewrites all the entries in some revlog of the current repository |
|
152 | Rewrites all the entries in some revlog of the current repository | |
154 | (by default, the manifest log) to save space. |
|
153 | (by default, the manifest log) to save space. | |
155 |
|
154 | |||
156 | Different sort algorithms have different performance |
|
155 | Different sort algorithms have different performance | |
157 | characteristics. Use ``--sort`` to select a sort algorithm so you |
|
156 | characteristics. Use ``--sort`` to select a sort algorithm so you | |
158 | can determine which works best for your data. |
|
157 | can determine which works best for your data. | |
159 | """ |
|
158 | """ | |
160 |
|
159 | |||
161 | if not repo.local(): |
|
160 | if not repo.local(): | |
162 | raise util.Abort(_('not a local repository: %s') % repo.root) |
|
161 | raise util.Abort(_('not a local repository: %s') % repo.root) | |
163 |
|
162 | |||
164 | fn = opts.get('revlog') |
|
163 | fn = opts.get('revlog') | |
165 | if not fn: |
|
164 | if not fn: | |
166 | indexfn = repo.sjoin('00manifest.i') |
|
165 | indexfn = repo.sjoin('00manifest.i') | |
167 | else: |
|
166 | else: | |
168 | if not fn.endswith('.i'): |
|
167 | if not fn.endswith('.i'): | |
169 | raise util.Abort(_('--revlog option must specify the revlog index ' |
|
168 | raise util.Abort(_('--revlog option must specify the revlog index ' | |
170 | 'file (*.i), not %s') % opts.get('revlog')) |
|
169 | 'file (*.i), not %s') % opts.get('revlog')) | |
171 |
|
170 | |||
172 | indexfn = os.path.realpath(fn) |
|
171 | indexfn = os.path.realpath(fn) | |
173 | store = repo.sjoin('') |
|
172 | store = repo.sjoin('') | |
174 | if not indexfn.startswith(store): |
|
173 | if not indexfn.startswith(store): | |
175 | raise util.Abort(_('--revlog option must specify a revlog in %s, ' |
|
174 | raise util.Abort(_('--revlog option must specify a revlog in %s, ' | |
176 | 'not %s') % (store, indexfn)) |
|
175 | 'not %s') % (store, indexfn)) | |
177 |
|
176 | |||
178 | sortname = opts['sort'] |
|
177 | sortname = opts['sort'] | |
179 | try: |
|
178 | try: | |
180 | toposort = globals()['toposort_' + sortname] |
|
179 | toposort = globals()['toposort_' + sortname] | |
181 | except KeyError: |
|
180 | except KeyError: | |
182 | raise util.Abort(_('no such toposort algorithm: %s') % sortname) |
|
181 | raise util.Abort(_('no such toposort algorithm: %s') % sortname) | |
183 |
|
182 | |||
184 | if not os.path.exists(indexfn): |
|
183 | if not os.path.exists(indexfn): | |
185 | raise util.Abort(_('no such file: %s') % indexfn) |
|
184 | raise util.Abort(_('no such file: %s') % indexfn) | |
186 | if '00changelog' in indexfn: |
|
185 | if '00changelog' in indexfn: | |
187 | raise util.Abort(_('shrinking the changelog ' |
|
186 | raise util.Abort(_('shrinking the changelog ' | |
188 | 'will corrupt your repository')) |
|
187 | 'will corrupt your repository')) | |
189 |
|
188 | |||
190 | ui.write(_('shrinking %s\n') % indexfn) |
|
189 | ui.write(_('shrinking %s\n') % indexfn) | |
191 | prefix = os.path.basename(indexfn)[:-1] |
|
190 | prefix = os.path.basename(indexfn)[:-1] | |
192 | tmpindexfn = util.mktempcopy(indexfn, emptyok=True) |
|
191 | tmpindexfn = util.mktempcopy(indexfn, emptyok=True) | |
193 |
|
192 | |||
194 | r1 = revlog.revlog(util.opener(os.getcwd(), audit=False), indexfn) |
|
193 | r1 = revlog.revlog(util.opener(os.getcwd(), audit=False), indexfn) | |
195 | r2 = revlog.revlog(util.opener(os.getcwd(), audit=False), tmpindexfn) |
|
194 | r2 = revlog.revlog(util.opener(os.getcwd(), audit=False), tmpindexfn) | |
196 |
|
195 | |||
197 | datafn, tmpdatafn = r1.datafile, r2.datafile |
|
196 | datafn, tmpdatafn = r1.datafile, r2.datafile | |
198 |
|
197 | |||
199 | oldindexfn = indexfn + '.old' |
|
198 | oldindexfn = indexfn + '.old' | |
200 | olddatafn = datafn + '.old' |
|
199 | olddatafn = datafn + '.old' | |
201 | if os.path.exists(oldindexfn) or os.path.exists(olddatafn): |
|
200 | if os.path.exists(oldindexfn) or os.path.exists(olddatafn): | |
202 | raise util.Abort(_('one or both of\n' |
|
201 | raise util.Abort(_('one or both of\n' | |
203 | ' %s\n' |
|
202 | ' %s\n' | |
204 | ' %s\n' |
|
203 | ' %s\n' | |
205 | 'exists from a previous run; please clean up ' |
|
204 | 'exists from a previous run; please clean up ' | |
206 | 'before running again') % (oldindexfn, olddatafn)) |
|
205 | 'before running again') % (oldindexfn, olddatafn)) | |
207 |
|
206 | |||
208 | # Don't use repo.transaction(), because then things get hairy with |
|
207 | # Don't use repo.transaction(), because then things get hairy with | |
209 | # paths: some need to be relative to .hg, and some need to be |
|
208 | # paths: some need to be relative to .hg, and some need to be | |
210 | # absolute. Doing it this way keeps things simple: everything is an |
|
209 | # absolute. Doing it this way keeps things simple: everything is an | |
211 | # absolute path. |
|
210 | # absolute path. | |
212 | lock = repo.lock(wait=False) |
|
211 | lock = repo.lock(wait=False) | |
213 | tr = transaction.transaction(ui.warn, |
|
212 | tr = transaction.transaction(ui.warn, | |
214 | open, |
|
213 | open, | |
215 | repo.sjoin('journal')) |
|
214 | repo.sjoin('journal')) | |
216 |
|
215 | |||
217 | def ignoremissing(func): |
|
216 | def ignoremissing(func): | |
218 | def f(*args, **kw): |
|
217 | def f(*args, **kw): | |
219 | try: |
|
218 | try: | |
220 | return func(*args, **kw) |
|
219 | return func(*args, **kw) | |
221 | except OSError, inst: |
|
220 | except OSError, inst: | |
222 | if inst.errno != errno.ENOENT: |
|
221 | if inst.errno != errno.ENOENT: | |
223 | raise |
|
222 | raise | |
224 | return f |
|
223 | return f | |
225 |
|
224 | |||
226 | try: |
|
225 | try: | |
227 | try: |
|
226 | try: | |
228 | order = toposort(ui, r1) |
|
227 | order = toposort(ui, r1) | |
229 |
|
228 | |||
230 | suboptimal = 0 |
|
229 | suboptimal = 0 | |
231 | for i in xrange(1, len(order)): |
|
230 | for i in xrange(1, len(order)): | |
232 | parents = [p for p in r1.parentrevs(order[i]) |
|
231 | parents = [p for p in r1.parentrevs(order[i]) | |
233 | if p != node.nullrev] |
|
232 | if p != node.nullrev] | |
234 | if parents and order[i - 1] not in parents: |
|
233 | if parents and order[i - 1] not in parents: | |
235 | suboptimal += 1 |
|
234 | suboptimal += 1 | |
236 | ui.note(_('%d suboptimal nodes\n') % suboptimal) |
|
235 | ui.note(_('%d suboptimal nodes\n') % suboptimal) | |
237 |
|
236 | |||
238 | writerevs(ui, r1, r2, order, tr) |
|
237 | writerevs(ui, r1, r2, order, tr) | |
239 | report(ui, r1, r2) |
|
238 | report(ui, r1, r2) | |
240 | tr.close() |
|
239 | tr.close() | |
241 | except: |
|
240 | except: | |
242 | # Abort transaction first, so we truncate the files before |
|
241 | # Abort transaction first, so we truncate the files before | |
243 | # deleting them. |
|
242 | # deleting them. | |
244 | tr.abort() |
|
243 | tr.abort() | |
245 | for fn in (tmpindexfn, tmpdatafn): |
|
244 | for fn in (tmpindexfn, tmpdatafn): | |
246 | ignoremissing(os.unlink)(fn) |
|
245 | ignoremissing(os.unlink)(fn) | |
247 | raise |
|
246 | raise | |
248 | if not opts.get('dry_run'): |
|
247 | if not opts.get('dry_run'): | |
249 | # racy, both files cannot be renamed atomically |
|
248 | # racy, both files cannot be renamed atomically | |
250 | # copy files |
|
249 | # copy files | |
251 | util.os_link(indexfn, oldindexfn) |
|
250 | util.os_link(indexfn, oldindexfn) | |
252 | ignoremissing(util.os_link)(datafn, olddatafn) |
|
251 | ignoremissing(util.os_link)(datafn, olddatafn) | |
253 |
|
252 | |||
254 | # rename |
|
253 | # rename | |
255 | util.rename(tmpindexfn, indexfn) |
|
254 | util.rename(tmpindexfn, indexfn) | |
256 | try: |
|
255 | try: | |
257 | os.chmod(tmpdatafn, os.stat(datafn).st_mode) |
|
256 | os.chmod(tmpdatafn, os.stat(datafn).st_mode) | |
258 | util.rename(tmpdatafn, datafn) |
|
257 | util.rename(tmpdatafn, datafn) | |
259 | except OSError, inst: |
|
258 | except OSError, inst: | |
260 | if inst.errno != errno.ENOENT: |
|
259 | if inst.errno != errno.ENOENT: | |
261 | raise |
|
260 | raise | |
262 | ignoremissing(os.unlink)(datafn) |
|
261 | ignoremissing(os.unlink)(datafn) | |
263 | else: |
|
262 | else: | |
264 | for fn in (tmpindexfn, tmpdatafn): |
|
263 | for fn in (tmpindexfn, tmpdatafn): | |
265 | ignoremissing(os.unlink)(fn) |
|
264 | ignoremissing(os.unlink)(fn) | |
266 | finally: |
|
265 | finally: | |
267 | lock.release() |
|
266 | lock.release() | |
268 |
|
267 | |||
269 | if not opts.get('dry_run'): |
|
268 | if not opts.get('dry_run'): | |
270 | ui.write(_('note: old revlog saved in:\n' |
|
269 | ui.write(_('note: old revlog saved in:\n' | |
271 | ' %s\n' |
|
270 | ' %s\n' | |
272 | ' %s\n' |
|
271 | ' %s\n' | |
273 | '(You can delete those files when you are satisfied that your\n' |
|
272 | '(You can delete those files when you are satisfied that your\n' | |
274 | 'repository is still sane. ' |
|
273 | 'repository is still sane. ' | |
275 | 'Running \'hg verify\' is strongly recommended.)\n') |
|
274 | 'Running \'hg verify\' is strongly recommended.)\n') | |
276 | % (oldindexfn, olddatafn)) |
|
275 | % (oldindexfn, olddatafn)) | |
277 |
|
276 | |||
278 | cmdtable = { |
|
277 | cmdtable = { | |
279 | 'shrink': (shrink, |
|
278 | 'shrink': (shrink, | |
280 | [('', 'revlog', '', _('index (.i) file of the revlog to shrink')), |
|
279 | [('', 'revlog', '', _('index (.i) file of the revlog to shrink')), | |
281 | ('n', 'dry-run', None, _('do not shrink, simulate only')), |
|
280 | ('n', 'dry-run', None, _('do not shrink, simulate only')), | |
282 | ('', 'sort', 'reversepostorder', 'name of sort algorithm to use'), |
|
281 | ('', 'sort', 'reversepostorder', 'name of sort algorithm to use'), | |
283 | ], |
|
282 | ], | |
284 | _('hg shrink [--revlog PATH]')) |
|
283 | _('hg shrink [--revlog PATH]')) | |
285 | } |
|
284 | } | |
286 |
|
285 | |||
287 | if __name__ == "__main__": |
|
286 | if __name__ == "__main__": | |
288 | print "shrink-revlog.py is now an extension (see hg help extensions)" |
|
287 | 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