Show More
@@ -1,861 +1,870 b'' | |||||
1 | # revlogdeltas.py - Logic around delta computation for revlog |
|
1 | # revlogdeltas.py - Logic around delta computation for revlog | |
2 | # |
|
2 | # | |
3 | # Copyright 2005-2007 Matt Mackall <mpm@selenic.com> |
|
3 | # Copyright 2005-2007 Matt Mackall <mpm@selenic.com> | |
4 | # Copyright 2018 Octobus <contact@octobus.net> |
|
4 | # Copyright 2018 Octobus <contact@octobus.net> | |
5 | # |
|
5 | # | |
6 | # This software may be used and distributed according to the terms of the |
|
6 | # This software may be used and distributed according to the terms of the | |
7 | # GNU General Public License version 2 or any later version. |
|
7 | # GNU General Public License version 2 or any later version. | |
8 | """Helper class to compute deltas stored inside revlogs""" |
|
8 | """Helper class to compute deltas stored inside revlogs""" | |
9 |
|
9 | |||
10 | from __future__ import absolute_import |
|
10 | from __future__ import absolute_import | |
11 |
|
11 | |||
12 | import collections |
|
12 | import collections | |
13 | import heapq |
|
13 | import heapq | |
14 | import struct |
|
14 | import struct | |
15 |
|
15 | |||
16 | # import stuff from node for others to import from revlog |
|
16 | # import stuff from node for others to import from revlog | |
17 | from ..node import ( |
|
17 | from ..node import ( | |
18 | nullrev, |
|
18 | nullrev, | |
19 | ) |
|
19 | ) | |
20 | from ..i18n import _ |
|
20 | from ..i18n import _ | |
21 |
|
21 | |||
22 | from .constants import ( |
|
22 | from .constants import ( | |
23 | REVIDX_ISCENSORED, |
|
23 | REVIDX_ISCENSORED, | |
24 | REVIDX_RAWTEXT_CHANGING_FLAGS, |
|
24 | REVIDX_RAWTEXT_CHANGING_FLAGS, | |
25 | ) |
|
25 | ) | |
26 |
|
26 | |||
27 | from ..thirdparty import ( |
|
27 | from ..thirdparty import ( | |
28 | attr, |
|
28 | attr, | |
29 | ) |
|
29 | ) | |
30 |
|
30 | |||
31 | from .. import ( |
|
31 | from .. import ( | |
32 | error, |
|
32 | error, | |
33 | mdiff, |
|
33 | mdiff, | |
34 | ) |
|
34 | ) | |
35 |
|
35 | |||
36 | RevlogError = error.RevlogError |
|
36 | RevlogError = error.RevlogError | |
37 | CensoredNodeError = error.CensoredNodeError |
|
37 | CensoredNodeError = error.CensoredNodeError | |
38 |
|
38 | |||
39 | # maximum <delta-chain-data>/<revision-text-length> ratio |
|
39 | # maximum <delta-chain-data>/<revision-text-length> ratio | |
40 | LIMIT_DELTA2TEXT = 2 |
|
40 | LIMIT_DELTA2TEXT = 2 | |
41 |
|
41 | |||
42 | class _testrevlog(object): |
|
42 | class _testrevlog(object): | |
43 | """minimalist fake revlog to use in doctests""" |
|
43 | """minimalist fake revlog to use in doctests""" | |
44 |
|
44 | |||
45 | def __init__(self, data, density=0.5, mingap=0): |
|
45 | def __init__(self, data, density=0.5, mingap=0): | |
46 | """data is an list of revision payload boundaries""" |
|
46 | """data is an list of revision payload boundaries""" | |
47 | self._data = data |
|
47 | self._data = data | |
48 | self._srdensitythreshold = density |
|
48 | self._srdensitythreshold = density | |
49 | self._srmingapsize = mingap |
|
49 | self._srmingapsize = mingap | |
50 |
|
50 | |||
51 | def start(self, rev): |
|
51 | def start(self, rev): | |
52 | if rev == 0: |
|
52 | if rev == 0: | |
53 | return 0 |
|
53 | return 0 | |
54 | return self._data[rev - 1] |
|
54 | return self._data[rev - 1] | |
55 |
|
55 | |||
56 | def end(self, rev): |
|
56 | def end(self, rev): | |
57 | return self._data[rev] |
|
57 | return self._data[rev] | |
58 |
|
58 | |||
59 | def length(self, rev): |
|
59 | def length(self, rev): | |
60 | return self.end(rev) - self.start(rev) |
|
60 | return self.end(rev) - self.start(rev) | |
61 |
|
61 | |||
62 | def __len__(self): |
|
62 | def __len__(self): | |
63 | return len(self._data) |
|
63 | return len(self._data) | |
64 |
|
64 | |||
65 | def slicechunk(revlog, revs, deltainfo=None, targetsize=None): |
|
65 | def slicechunk(revlog, revs, deltainfo=None, targetsize=None): | |
66 | """slice revs to reduce the amount of unrelated data to be read from disk. |
|
66 | """slice revs to reduce the amount of unrelated data to be read from disk. | |
67 |
|
67 | |||
68 | ``revs`` is sliced into groups that should be read in one time. |
|
68 | ``revs`` is sliced into groups that should be read in one time. | |
69 | Assume that revs are sorted. |
|
69 | Assume that revs are sorted. | |
70 |
|
70 | |||
71 | The initial chunk is sliced until the overall density (payload/chunks-span |
|
71 | The initial chunk is sliced until the overall density (payload/chunks-span | |
72 | ratio) is above `revlog._srdensitythreshold`. No gap smaller than |
|
72 | ratio) is above `revlog._srdensitythreshold`. No gap smaller than | |
73 | `revlog._srmingapsize` is skipped. |
|
73 | `revlog._srmingapsize` is skipped. | |
74 |
|
74 | |||
75 | If `targetsize` is set, no chunk larger than `targetsize` will be yield. |
|
75 | If `targetsize` is set, no chunk larger than `targetsize` will be yield. | |
76 | For consistency with other slicing choice, this limit won't go lower than |
|
76 | For consistency with other slicing choice, this limit won't go lower than | |
77 | `revlog._srmingapsize`. |
|
77 | `revlog._srmingapsize`. | |
78 |
|
78 | |||
79 | If individual revisions chunk are larger than this limit, they will still |
|
79 | If individual revisions chunk are larger than this limit, they will still | |
80 | be raised individually. |
|
80 | be raised individually. | |
81 |
|
81 | |||
82 | >>> revlog = _testrevlog([ |
|
82 | >>> revlog = _testrevlog([ | |
83 | ... 5, #00 (5) |
|
83 | ... 5, #00 (5) | |
84 | ... 10, #01 (5) |
|
84 | ... 10, #01 (5) | |
85 | ... 12, #02 (2) |
|
85 | ... 12, #02 (2) | |
86 | ... 12, #03 (empty) |
|
86 | ... 12, #03 (empty) | |
87 | ... 27, #04 (15) |
|
87 | ... 27, #04 (15) | |
88 | ... 31, #05 (4) |
|
88 | ... 31, #05 (4) | |
89 | ... 31, #06 (empty) |
|
89 | ... 31, #06 (empty) | |
90 | ... 42, #07 (11) |
|
90 | ... 42, #07 (11) | |
91 | ... 47, #08 (5) |
|
91 | ... 47, #08 (5) | |
92 | ... 47, #09 (empty) |
|
92 | ... 47, #09 (empty) | |
93 | ... 48, #10 (1) |
|
93 | ... 48, #10 (1) | |
94 | ... 51, #11 (3) |
|
94 | ... 51, #11 (3) | |
95 | ... 74, #12 (23) |
|
95 | ... 74, #12 (23) | |
96 | ... 85, #13 (11) |
|
96 | ... 85, #13 (11) | |
97 | ... 86, #14 (1) |
|
97 | ... 86, #14 (1) | |
98 | ... 91, #15 (5) |
|
98 | ... 91, #15 (5) | |
99 | ... ]) |
|
99 | ... ]) | |
100 |
|
100 | |||
101 | >>> list(slicechunk(revlog, list(range(16)))) |
|
101 | >>> list(slicechunk(revlog, list(range(16)))) | |
102 | [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]] |
|
102 | [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]] | |
103 | >>> list(slicechunk(revlog, [0, 15])) |
|
103 | >>> list(slicechunk(revlog, [0, 15])) | |
104 | [[0], [15]] |
|
104 | [[0], [15]] | |
105 | >>> list(slicechunk(revlog, [0, 11, 15])) |
|
105 | >>> list(slicechunk(revlog, [0, 11, 15])) | |
106 | [[0], [11], [15]] |
|
106 | [[0], [11], [15]] | |
107 | >>> list(slicechunk(revlog, [0, 11, 13, 15])) |
|
107 | >>> list(slicechunk(revlog, [0, 11, 13, 15])) | |
108 | [[0], [11, 13, 15]] |
|
108 | [[0], [11, 13, 15]] | |
109 | >>> list(slicechunk(revlog, [1, 2, 3, 5, 8, 10, 11, 14])) |
|
109 | >>> list(slicechunk(revlog, [1, 2, 3, 5, 8, 10, 11, 14])) | |
110 | [[1, 2], [5, 8, 10, 11], [14]] |
|
110 | [[1, 2], [5, 8, 10, 11], [14]] | |
111 |
|
111 | |||
112 | Slicing with a maximum chunk size |
|
112 | Slicing with a maximum chunk size | |
113 | >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=15)) |
|
113 | >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=15)) | |
114 | [[0], [11], [13], [15]] |
|
114 | [[0], [11], [13], [15]] | |
115 | >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=20)) |
|
115 | >>> list(slicechunk(revlog, [0, 11, 13, 15], targetsize=20)) | |
116 | [[0], [11], [13, 15]] |
|
116 | [[0], [11], [13, 15]] | |
117 | """ |
|
117 | """ | |
118 | if targetsize is not None: |
|
118 | if targetsize is not None: | |
119 | targetsize = max(targetsize, revlog._srmingapsize) |
|
119 | targetsize = max(targetsize, revlog._srmingapsize) | |
120 | # targetsize should not be specified when evaluating delta candidates: |
|
120 | # targetsize should not be specified when evaluating delta candidates: | |
121 | # * targetsize is used to ensure we stay within specification when reading, |
|
121 | # * targetsize is used to ensure we stay within specification when reading, | |
122 | # * deltainfo is used to pick are good delta chain when writing. |
|
122 | # * deltainfo is used to pick are good delta chain when writing. | |
123 | if not (deltainfo is None or targetsize is None): |
|
123 | if not (deltainfo is None or targetsize is None): | |
124 | msg = 'cannot use `targetsize` with a `deltainfo`' |
|
124 | msg = 'cannot use `targetsize` with a `deltainfo`' | |
125 | raise error.ProgrammingError(msg) |
|
125 | raise error.ProgrammingError(msg) | |
126 | for chunk in _slicechunktodensity(revlog, revs, |
|
126 | for chunk in _slicechunktodensity(revlog, revs, | |
127 | deltainfo, |
|
127 | deltainfo, | |
128 | revlog._srdensitythreshold, |
|
128 | revlog._srdensitythreshold, | |
129 | revlog._srmingapsize): |
|
129 | revlog._srmingapsize): | |
130 | for subchunk in _slicechunktosize(revlog, chunk, targetsize): |
|
130 | for subchunk in _slicechunktosize(revlog, chunk, targetsize): | |
131 | yield subchunk |
|
131 | yield subchunk | |
132 |
|
132 | |||
133 | def _slicechunktosize(revlog, revs, targetsize=None): |
|
133 | def _slicechunktosize(revlog, revs, targetsize=None): | |
134 | """slice revs to match the target size |
|
134 | """slice revs to match the target size | |
135 |
|
135 | |||
136 | This is intended to be used on chunk that density slicing selected by that |
|
136 | This is intended to be used on chunk that density slicing selected by that | |
137 | are still too large compared to the read garantee of revlog. This might |
|
137 | are still too large compared to the read garantee of revlog. This might | |
138 | happens when "minimal gap size" interrupted the slicing or when chain are |
|
138 | happens when "minimal gap size" interrupted the slicing or when chain are | |
139 | built in a way that create large blocks next to each other. |
|
139 | built in a way that create large blocks next to each other. | |
140 |
|
140 | |||
141 | >>> revlog = _testrevlog([ |
|
141 | >>> revlog = _testrevlog([ | |
142 | ... 3, #0 (3) |
|
142 | ... 3, #0 (3) | |
143 | ... 5, #1 (2) |
|
143 | ... 5, #1 (2) | |
144 | ... 6, #2 (1) |
|
144 | ... 6, #2 (1) | |
145 | ... 8, #3 (2) |
|
145 | ... 8, #3 (2) | |
146 | ... 8, #4 (empty) |
|
146 | ... 8, #4 (empty) | |
147 | ... 11, #5 (3) |
|
147 | ... 11, #5 (3) | |
148 | ... 12, #6 (1) |
|
148 | ... 12, #6 (1) | |
149 | ... 13, #7 (1) |
|
149 | ... 13, #7 (1) | |
150 | ... 14, #8 (1) |
|
150 | ... 14, #8 (1) | |
151 | ... ]) |
|
151 | ... ]) | |
152 |
|
152 | |||
153 | Cases where chunk is already small enough |
|
153 | Cases where chunk is already small enough | |
154 | >>> list(_slicechunktosize(revlog, [0], 3)) |
|
154 | >>> list(_slicechunktosize(revlog, [0], 3)) | |
155 | [[0]] |
|
155 | [[0]] | |
156 | >>> list(_slicechunktosize(revlog, [6, 7], 3)) |
|
156 | >>> list(_slicechunktosize(revlog, [6, 7], 3)) | |
157 | [[6, 7]] |
|
157 | [[6, 7]] | |
158 | >>> list(_slicechunktosize(revlog, [0], None)) |
|
158 | >>> list(_slicechunktosize(revlog, [0], None)) | |
159 | [[0]] |
|
159 | [[0]] | |
160 | >>> list(_slicechunktosize(revlog, [6, 7], None)) |
|
160 | >>> list(_slicechunktosize(revlog, [6, 7], None)) | |
161 | [[6, 7]] |
|
161 | [[6, 7]] | |
162 |
|
162 | |||
163 | cases where we need actual slicing |
|
163 | cases where we need actual slicing | |
164 | >>> list(_slicechunktosize(revlog, [0, 1], 3)) |
|
164 | >>> list(_slicechunktosize(revlog, [0, 1], 3)) | |
165 | [[0], [1]] |
|
165 | [[0], [1]] | |
166 | >>> list(_slicechunktosize(revlog, [1, 3], 3)) |
|
166 | >>> list(_slicechunktosize(revlog, [1, 3], 3)) | |
167 | [[1], [3]] |
|
167 | [[1], [3]] | |
168 | >>> list(_slicechunktosize(revlog, [1, 2, 3], 3)) |
|
168 | >>> list(_slicechunktosize(revlog, [1, 2, 3], 3)) | |
169 | [[1, 2], [3]] |
|
169 | [[1, 2], [3]] | |
170 | >>> list(_slicechunktosize(revlog, [3, 5], 3)) |
|
170 | >>> list(_slicechunktosize(revlog, [3, 5], 3)) | |
171 | [[3], [5]] |
|
171 | [[3], [5]] | |
172 | >>> list(_slicechunktosize(revlog, [3, 4, 5], 3)) |
|
172 | >>> list(_slicechunktosize(revlog, [3, 4, 5], 3)) | |
173 | [[3], [5]] |
|
173 | [[3], [5]] | |
174 | >>> list(_slicechunktosize(revlog, [5, 6, 7, 8], 3)) |
|
174 | >>> list(_slicechunktosize(revlog, [5, 6, 7, 8], 3)) | |
175 | [[5], [6, 7, 8]] |
|
175 | [[5], [6, 7, 8]] | |
176 | >>> list(_slicechunktosize(revlog, [0, 1, 2, 3, 4, 5, 6, 7, 8], 3)) |
|
176 | >>> list(_slicechunktosize(revlog, [0, 1, 2, 3, 4, 5, 6, 7, 8], 3)) | |
177 | [[0], [1, 2], [3], [5], [6, 7, 8]] |
|
177 | [[0], [1, 2], [3], [5], [6, 7, 8]] | |
178 |
|
178 | |||
179 | Case with too large individual chunk (must return valid chunk) |
|
179 | Case with too large individual chunk (must return valid chunk) | |
180 | >>> list(_slicechunktosize(revlog, [0, 1], 2)) |
|
180 | >>> list(_slicechunktosize(revlog, [0, 1], 2)) | |
181 | [[0], [1]] |
|
181 | [[0], [1]] | |
182 | >>> list(_slicechunktosize(revlog, [1, 3], 1)) |
|
182 | >>> list(_slicechunktosize(revlog, [1, 3], 1)) | |
183 | [[1], [3]] |
|
183 | [[1], [3]] | |
184 | >>> list(_slicechunktosize(revlog, [3, 4, 5], 2)) |
|
184 | >>> list(_slicechunktosize(revlog, [3, 4, 5], 2)) | |
185 | [[3], [5]] |
|
185 | [[3], [5]] | |
186 | """ |
|
186 | """ | |
187 | assert targetsize is None or 0 <= targetsize |
|
187 | assert targetsize is None or 0 <= targetsize | |
188 | if targetsize is None or segmentspan(revlog, revs) <= targetsize: |
|
188 | if targetsize is None or segmentspan(revlog, revs) <= targetsize: | |
189 | yield revs |
|
189 | yield revs | |
190 | return |
|
190 | return | |
191 |
|
191 | |||
192 | startrevidx = 0 |
|
192 | startrevidx = 0 | |
193 | startdata = revlog.start(revs[0]) |
|
193 | startdata = revlog.start(revs[0]) | |
194 | endrevidx = 0 |
|
194 | endrevidx = 0 | |
195 | iterrevs = enumerate(revs) |
|
195 | iterrevs = enumerate(revs) | |
196 | next(iterrevs) # skip first rev. |
|
196 | next(iterrevs) # skip first rev. | |
197 | for idx, r in iterrevs: |
|
197 | for idx, r in iterrevs: | |
198 | span = revlog.end(r) - startdata |
|
198 | span = revlog.end(r) - startdata | |
199 | if span <= targetsize: |
|
199 | if span <= targetsize: | |
200 | endrevidx = idx |
|
200 | endrevidx = idx | |
201 | else: |
|
201 | else: | |
202 | chunk = _trimchunk(revlog, revs, startrevidx, endrevidx + 1) |
|
202 | chunk = _trimchunk(revlog, revs, startrevidx, endrevidx + 1) | |
203 | if chunk: |
|
203 | if chunk: | |
204 | yield chunk |
|
204 | yield chunk | |
205 | startrevidx = idx |
|
205 | startrevidx = idx | |
206 | startdata = revlog.start(r) |
|
206 | startdata = revlog.start(r) | |
207 | endrevidx = idx |
|
207 | endrevidx = idx | |
208 | yield _trimchunk(revlog, revs, startrevidx) |
|
208 | yield _trimchunk(revlog, revs, startrevidx) | |
209 |
|
209 | |||
210 | def _slicechunktodensity(revlog, revs, deltainfo=None, targetdensity=0.5, |
|
210 | def _slicechunktodensity(revlog, revs, deltainfo=None, targetdensity=0.5, | |
211 | mingapsize=0): |
|
211 | mingapsize=0): | |
212 | """slice revs to reduce the amount of unrelated data to be read from disk. |
|
212 | """slice revs to reduce the amount of unrelated data to be read from disk. | |
213 |
|
213 | |||
214 | ``revs`` is sliced into groups that should be read in one time. |
|
214 | ``revs`` is sliced into groups that should be read in one time. | |
215 | Assume that revs are sorted. |
|
215 | Assume that revs are sorted. | |
216 |
|
216 | |||
217 | ``deltainfo`` is a _deltainfo instance of a revision that we would append |
|
217 | ``deltainfo`` is a _deltainfo instance of a revision that we would append | |
218 | to the top of the revlog. |
|
218 | to the top of the revlog. | |
219 |
|
219 | |||
220 | The initial chunk is sliced until the overall density (payload/chunks-span |
|
220 | The initial chunk is sliced until the overall density (payload/chunks-span | |
221 | ratio) is above `targetdensity`. No gap smaller than `mingapsize` is |
|
221 | ratio) is above `targetdensity`. No gap smaller than `mingapsize` is | |
222 | skipped. |
|
222 | skipped. | |
223 |
|
223 | |||
224 | >>> revlog = _testrevlog([ |
|
224 | >>> revlog = _testrevlog([ | |
225 | ... 5, #00 (5) |
|
225 | ... 5, #00 (5) | |
226 | ... 10, #01 (5) |
|
226 | ... 10, #01 (5) | |
227 | ... 12, #02 (2) |
|
227 | ... 12, #02 (2) | |
228 | ... 12, #03 (empty) |
|
228 | ... 12, #03 (empty) | |
229 | ... 27, #04 (15) |
|
229 | ... 27, #04 (15) | |
230 | ... 31, #05 (4) |
|
230 | ... 31, #05 (4) | |
231 | ... 31, #06 (empty) |
|
231 | ... 31, #06 (empty) | |
232 | ... 42, #07 (11) |
|
232 | ... 42, #07 (11) | |
233 | ... 47, #08 (5) |
|
233 | ... 47, #08 (5) | |
234 | ... 47, #09 (empty) |
|
234 | ... 47, #09 (empty) | |
235 | ... 48, #10 (1) |
|
235 | ... 48, #10 (1) | |
236 | ... 51, #11 (3) |
|
236 | ... 51, #11 (3) | |
237 | ... 74, #12 (23) |
|
237 | ... 74, #12 (23) | |
238 | ... 85, #13 (11) |
|
238 | ... 85, #13 (11) | |
239 | ... 86, #14 (1) |
|
239 | ... 86, #14 (1) | |
240 | ... 91, #15 (5) |
|
240 | ... 91, #15 (5) | |
241 | ... ]) |
|
241 | ... ]) | |
242 |
|
242 | |||
243 | >>> list(_slicechunktodensity(revlog, list(range(16)))) |
|
243 | >>> list(_slicechunktodensity(revlog, list(range(16)))) | |
244 | [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]] |
|
244 | [[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]] | |
245 | >>> list(_slicechunktodensity(revlog, [0, 15])) |
|
245 | >>> list(_slicechunktodensity(revlog, [0, 15])) | |
246 | [[0], [15]] |
|
246 | [[0], [15]] | |
247 | >>> list(_slicechunktodensity(revlog, [0, 11, 15])) |
|
247 | >>> list(_slicechunktodensity(revlog, [0, 11, 15])) | |
248 | [[0], [11], [15]] |
|
248 | [[0], [11], [15]] | |
249 | >>> list(_slicechunktodensity(revlog, [0, 11, 13, 15])) |
|
249 | >>> list(_slicechunktodensity(revlog, [0, 11, 13, 15])) | |
250 | [[0], [11, 13, 15]] |
|
250 | [[0], [11, 13, 15]] | |
251 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14])) |
|
251 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14])) | |
252 | [[1, 2], [5, 8, 10, 11], [14]] |
|
252 | [[1, 2], [5, 8, 10, 11], [14]] | |
253 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], |
|
253 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], | |
254 | ... mingapsize=20)) |
|
254 | ... mingapsize=20)) | |
255 | [[1, 2, 3, 5, 8, 10, 11], [14]] |
|
255 | [[1, 2, 3, 5, 8, 10, 11], [14]] | |
256 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], |
|
256 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], | |
257 | ... targetdensity=0.95)) |
|
257 | ... targetdensity=0.95)) | |
258 | [[1, 2], [5], [8, 10, 11], [14]] |
|
258 | [[1, 2], [5], [8, 10, 11], [14]] | |
259 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], |
|
259 | >>> list(_slicechunktodensity(revlog, [1, 2, 3, 5, 8, 10, 11, 14], | |
260 | ... targetdensity=0.95, mingapsize=12)) |
|
260 | ... targetdensity=0.95, mingapsize=12)) | |
261 | [[1, 2], [5, 8, 10, 11], [14]] |
|
261 | [[1, 2], [5, 8, 10, 11], [14]] | |
262 | """ |
|
262 | """ | |
263 | start = revlog.start |
|
263 | start = revlog.start | |
264 | length = revlog.length |
|
264 | length = revlog.length | |
265 |
|
265 | |||
266 | if len(revs) <= 1: |
|
266 | if len(revs) <= 1: | |
267 | yield revs |
|
267 | yield revs | |
268 | return |
|
268 | return | |
269 |
|
269 | |||
270 | nextrev = len(revlog) |
|
270 | nextrev = len(revlog) | |
271 | nextoffset = revlog.end(nextrev - 1) |
|
271 | nextoffset = revlog.end(nextrev - 1) | |
272 |
|
272 | |||
273 | if deltainfo is None: |
|
273 | if deltainfo is None: | |
274 | deltachainspan = segmentspan(revlog, revs) |
|
274 | deltachainspan = segmentspan(revlog, revs) | |
275 | chainpayload = sum(length(r) for r in revs) |
|
275 | chainpayload = sum(length(r) for r in revs) | |
276 | else: |
|
276 | else: | |
277 | deltachainspan = deltainfo.distance |
|
277 | deltachainspan = deltainfo.distance | |
278 | chainpayload = deltainfo.compresseddeltalen |
|
278 | chainpayload = deltainfo.compresseddeltalen | |
279 |
|
279 | |||
280 | if deltachainspan < mingapsize: |
|
280 | if deltachainspan < mingapsize: | |
281 | yield revs |
|
281 | yield revs | |
282 | return |
|
282 | return | |
283 |
|
283 | |||
284 | readdata = deltachainspan |
|
284 | readdata = deltachainspan | |
285 |
|
285 | |||
286 | if deltachainspan: |
|
286 | if deltachainspan: | |
287 | density = chainpayload / float(deltachainspan) |
|
287 | density = chainpayload / float(deltachainspan) | |
288 | else: |
|
288 | else: | |
289 | density = 1.0 |
|
289 | density = 1.0 | |
290 |
|
290 | |||
291 | if density >= targetdensity: |
|
291 | if density >= targetdensity: | |
292 | yield revs |
|
292 | yield revs | |
293 | return |
|
293 | return | |
294 |
|
294 | |||
295 | if deltainfo is not None and deltainfo.deltalen: |
|
295 | if deltainfo is not None and deltainfo.deltalen: | |
296 | revs = list(revs) |
|
296 | revs = list(revs) | |
297 | revs.append(nextrev) |
|
297 | revs.append(nextrev) | |
298 |
|
298 | |||
299 | # Store the gaps in a heap to have them sorted by decreasing size |
|
299 | # Store the gaps in a heap to have them sorted by decreasing size | |
300 | gapsheap = [] |
|
300 | gapsheap = [] | |
301 | heapq.heapify(gapsheap) |
|
301 | heapq.heapify(gapsheap) | |
302 | prevend = None |
|
302 | prevend = None | |
303 | for i, rev in enumerate(revs): |
|
303 | for i, rev in enumerate(revs): | |
304 | if rev < nextrev: |
|
304 | if rev < nextrev: | |
305 | revstart = start(rev) |
|
305 | revstart = start(rev) | |
306 | revlen = length(rev) |
|
306 | revlen = length(rev) | |
307 | else: |
|
307 | else: | |
308 | revstart = nextoffset |
|
308 | revstart = nextoffset | |
309 | revlen = deltainfo.deltalen |
|
309 | revlen = deltainfo.deltalen | |
310 |
|
310 | |||
311 | # Skip empty revisions to form larger holes |
|
311 | # Skip empty revisions to form larger holes | |
312 | if revlen == 0: |
|
312 | if revlen == 0: | |
313 | continue |
|
313 | continue | |
314 |
|
314 | |||
315 | if prevend is not None: |
|
315 | if prevend is not None: | |
316 | gapsize = revstart - prevend |
|
316 | gapsize = revstart - prevend | |
317 | # only consider holes that are large enough |
|
317 | # only consider holes that are large enough | |
318 | if gapsize > mingapsize: |
|
318 | if gapsize > mingapsize: | |
319 | heapq.heappush(gapsheap, (-gapsize, i)) |
|
319 | heapq.heappush(gapsheap, (-gapsize, i)) | |
320 |
|
320 | |||
321 | prevend = revstart + revlen |
|
321 | prevend = revstart + revlen | |
322 |
|
322 | |||
323 | # Collect the indices of the largest holes until the density is acceptable |
|
323 | # Collect the indices of the largest holes until the density is acceptable | |
324 | indicesheap = [] |
|
324 | indicesheap = [] | |
325 | heapq.heapify(indicesheap) |
|
325 | heapq.heapify(indicesheap) | |
326 | while gapsheap and density < targetdensity: |
|
326 | while gapsheap and density < targetdensity: | |
327 | oppgapsize, gapidx = heapq.heappop(gapsheap) |
|
327 | oppgapsize, gapidx = heapq.heappop(gapsheap) | |
328 |
|
328 | |||
329 | heapq.heappush(indicesheap, gapidx) |
|
329 | heapq.heappush(indicesheap, gapidx) | |
330 |
|
330 | |||
331 | # the gap sizes are stored as negatives to be sorted decreasingly |
|
331 | # the gap sizes are stored as negatives to be sorted decreasingly | |
332 | # by the heap |
|
332 | # by the heap | |
333 | readdata -= (-oppgapsize) |
|
333 | readdata -= (-oppgapsize) | |
334 | if readdata > 0: |
|
334 | if readdata > 0: | |
335 | density = chainpayload / float(readdata) |
|
335 | density = chainpayload / float(readdata) | |
336 | else: |
|
336 | else: | |
337 | density = 1.0 |
|
337 | density = 1.0 | |
338 |
|
338 | |||
339 | # Cut the revs at collected indices |
|
339 | # Cut the revs at collected indices | |
340 | previdx = 0 |
|
340 | previdx = 0 | |
341 | while indicesheap: |
|
341 | while indicesheap: | |
342 | idx = heapq.heappop(indicesheap) |
|
342 | idx = heapq.heappop(indicesheap) | |
343 |
|
343 | |||
344 | chunk = _trimchunk(revlog, revs, previdx, idx) |
|
344 | chunk = _trimchunk(revlog, revs, previdx, idx) | |
345 | if chunk: |
|
345 | if chunk: | |
346 | yield chunk |
|
346 | yield chunk | |
347 |
|
347 | |||
348 | previdx = idx |
|
348 | previdx = idx | |
349 |
|
349 | |||
350 | chunk = _trimchunk(revlog, revs, previdx) |
|
350 | chunk = _trimchunk(revlog, revs, previdx) | |
351 | if chunk: |
|
351 | if chunk: | |
352 | yield chunk |
|
352 | yield chunk | |
353 |
|
353 | |||
354 | def _trimchunk(revlog, revs, startidx, endidx=None): |
|
354 | def _trimchunk(revlog, revs, startidx, endidx=None): | |
355 | """returns revs[startidx:endidx] without empty trailing revs |
|
355 | """returns revs[startidx:endidx] without empty trailing revs | |
356 |
|
356 | |||
357 | Doctest Setup |
|
357 | Doctest Setup | |
358 | >>> revlog = _testrevlog([ |
|
358 | >>> revlog = _testrevlog([ | |
359 | ... 5, #0 |
|
359 | ... 5, #0 | |
360 | ... 10, #1 |
|
360 | ... 10, #1 | |
361 | ... 12, #2 |
|
361 | ... 12, #2 | |
362 | ... 12, #3 (empty) |
|
362 | ... 12, #3 (empty) | |
363 | ... 17, #4 |
|
363 | ... 17, #4 | |
364 | ... 21, #5 |
|
364 | ... 21, #5 | |
365 | ... 21, #6 (empty) |
|
365 | ... 21, #6 (empty) | |
366 | ... ]) |
|
366 | ... ]) | |
367 |
|
367 | |||
368 | Contiguous cases: |
|
368 | Contiguous cases: | |
369 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0) |
|
369 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0) | |
370 | [0, 1, 2, 3, 4, 5] |
|
370 | [0, 1, 2, 3, 4, 5] | |
371 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 5) |
|
371 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 5) | |
372 | [0, 1, 2, 3, 4] |
|
372 | [0, 1, 2, 3, 4] | |
373 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 4) |
|
373 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 0, 4) | |
374 | [0, 1, 2] |
|
374 | [0, 1, 2] | |
375 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 2, 4) |
|
375 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 2, 4) | |
376 | [2] |
|
376 | [2] | |
377 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3) |
|
377 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3) | |
378 | [3, 4, 5] |
|
378 | [3, 4, 5] | |
379 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3, 5) |
|
379 | >>> _trimchunk(revlog, [0, 1, 2, 3, 4, 5, 6], 3, 5) | |
380 | [3, 4] |
|
380 | [3, 4] | |
381 |
|
381 | |||
382 | Discontiguous cases: |
|
382 | Discontiguous cases: | |
383 | >>> _trimchunk(revlog, [1, 3, 5, 6], 0) |
|
383 | >>> _trimchunk(revlog, [1, 3, 5, 6], 0) | |
384 | [1, 3, 5] |
|
384 | [1, 3, 5] | |
385 | >>> _trimchunk(revlog, [1, 3, 5, 6], 0, 2) |
|
385 | >>> _trimchunk(revlog, [1, 3, 5, 6], 0, 2) | |
386 | [1] |
|
386 | [1] | |
387 | >>> _trimchunk(revlog, [1, 3, 5, 6], 1, 3) |
|
387 | >>> _trimchunk(revlog, [1, 3, 5, 6], 1, 3) | |
388 | [3, 5] |
|
388 | [3, 5] | |
389 | >>> _trimchunk(revlog, [1, 3, 5, 6], 1) |
|
389 | >>> _trimchunk(revlog, [1, 3, 5, 6], 1) | |
390 | [3, 5] |
|
390 | [3, 5] | |
391 | """ |
|
391 | """ | |
392 | length = revlog.length |
|
392 | length = revlog.length | |
393 |
|
393 | |||
394 | if endidx is None: |
|
394 | if endidx is None: | |
395 | endidx = len(revs) |
|
395 | endidx = len(revs) | |
396 |
|
396 | |||
397 | # If we have a non-emtpy delta candidate, there are nothing to trim |
|
397 | # If we have a non-emtpy delta candidate, there are nothing to trim | |
398 | if revs[endidx - 1] < len(revlog): |
|
398 | if revs[endidx - 1] < len(revlog): | |
399 | # Trim empty revs at the end, except the very first revision of a chain |
|
399 | # Trim empty revs at the end, except the very first revision of a chain | |
400 | while (endidx > 1 |
|
400 | while (endidx > 1 | |
401 | and endidx > startidx |
|
401 | and endidx > startidx | |
402 | and length(revs[endidx - 1]) == 0): |
|
402 | and length(revs[endidx - 1]) == 0): | |
403 | endidx -= 1 |
|
403 | endidx -= 1 | |
404 |
|
404 | |||
405 | return revs[startidx:endidx] |
|
405 | return revs[startidx:endidx] | |
406 |
|
406 | |||
407 | def segmentspan(revlog, revs, deltainfo=None): |
|
407 | def segmentspan(revlog, revs, deltainfo=None): | |
408 | """Get the byte span of a segment of revisions |
|
408 | """Get the byte span of a segment of revisions | |
409 |
|
409 | |||
410 | revs is a sorted array of revision numbers |
|
410 | revs is a sorted array of revision numbers | |
411 |
|
411 | |||
412 | >>> revlog = _testrevlog([ |
|
412 | >>> revlog = _testrevlog([ | |
413 | ... 5, #0 |
|
413 | ... 5, #0 | |
414 | ... 10, #1 |
|
414 | ... 10, #1 | |
415 | ... 12, #2 |
|
415 | ... 12, #2 | |
416 | ... 12, #3 (empty) |
|
416 | ... 12, #3 (empty) | |
417 | ... 17, #4 |
|
417 | ... 17, #4 | |
418 | ... ]) |
|
418 | ... ]) | |
419 |
|
419 | |||
420 | >>> segmentspan(revlog, [0, 1, 2, 3, 4]) |
|
420 | >>> segmentspan(revlog, [0, 1, 2, 3, 4]) | |
421 | 17 |
|
421 | 17 | |
422 | >>> segmentspan(revlog, [0, 4]) |
|
422 | >>> segmentspan(revlog, [0, 4]) | |
423 | 17 |
|
423 | 17 | |
424 | >>> segmentspan(revlog, [3, 4]) |
|
424 | >>> segmentspan(revlog, [3, 4]) | |
425 | 5 |
|
425 | 5 | |
426 | >>> segmentspan(revlog, [1, 2, 3,]) |
|
426 | >>> segmentspan(revlog, [1, 2, 3,]) | |
427 | 7 |
|
427 | 7 | |
428 | >>> segmentspan(revlog, [1, 3]) |
|
428 | >>> segmentspan(revlog, [1, 3]) | |
429 | 7 |
|
429 | 7 | |
430 | """ |
|
430 | """ | |
431 | if not revs: |
|
431 | if not revs: | |
432 | return 0 |
|
432 | return 0 | |
433 | if deltainfo is not None and len(revlog) <= revs[-1]: |
|
433 | if deltainfo is not None and len(revlog) <= revs[-1]: | |
434 | if len(revs) == 1: |
|
434 | if len(revs) == 1: | |
435 | return deltainfo.deltalen |
|
435 | return deltainfo.deltalen | |
436 | offset = revlog.end(len(revlog) - 1) |
|
436 | offset = revlog.end(len(revlog) - 1) | |
437 | end = deltainfo.deltalen + offset |
|
437 | end = deltainfo.deltalen + offset | |
438 | else: |
|
438 | else: | |
439 | end = revlog.end(revs[-1]) |
|
439 | end = revlog.end(revs[-1]) | |
440 | return end - revlog.start(revs[0]) |
|
440 | return end - revlog.start(revs[0]) | |
441 |
|
441 | |||
442 | def _textfromdelta(fh, revlog, baserev, delta, p1, p2, flags, expectednode): |
|
442 | def _textfromdelta(fh, revlog, baserev, delta, p1, p2, flags, expectednode): | |
443 | """build full text from a (base, delta) pair and other metadata""" |
|
443 | """build full text from a (base, delta) pair and other metadata""" | |
444 | # special case deltas which replace entire base; no need to decode |
|
444 | # special case deltas which replace entire base; no need to decode | |
445 | # base revision. this neatly avoids censored bases, which throw when |
|
445 | # base revision. this neatly avoids censored bases, which throw when | |
446 | # they're decoded. |
|
446 | # they're decoded. | |
447 | hlen = struct.calcsize(">lll") |
|
447 | hlen = struct.calcsize(">lll") | |
448 | if delta[:hlen] == mdiff.replacediffheader(revlog.rawsize(baserev), |
|
448 | if delta[:hlen] == mdiff.replacediffheader(revlog.rawsize(baserev), | |
449 | len(delta) - hlen): |
|
449 | len(delta) - hlen): | |
450 | fulltext = delta[hlen:] |
|
450 | fulltext = delta[hlen:] | |
451 | else: |
|
451 | else: | |
452 | # deltabase is rawtext before changed by flag processors, which is |
|
452 | # deltabase is rawtext before changed by flag processors, which is | |
453 | # equivalent to non-raw text |
|
453 | # equivalent to non-raw text | |
454 | basetext = revlog.revision(baserev, _df=fh, raw=False) |
|
454 | basetext = revlog.revision(baserev, _df=fh, raw=False) | |
455 | fulltext = mdiff.patch(basetext, delta) |
|
455 | fulltext = mdiff.patch(basetext, delta) | |
456 |
|
456 | |||
457 | try: |
|
457 | try: | |
458 | res = revlog._processflags(fulltext, flags, 'read', raw=True) |
|
458 | res = revlog._processflags(fulltext, flags, 'read', raw=True) | |
459 | fulltext, validatehash = res |
|
459 | fulltext, validatehash = res | |
460 | if validatehash: |
|
460 | if validatehash: | |
461 | revlog.checkhash(fulltext, expectednode, p1=p1, p2=p2) |
|
461 | revlog.checkhash(fulltext, expectednode, p1=p1, p2=p2) | |
462 | if flags & REVIDX_ISCENSORED: |
|
462 | if flags & REVIDX_ISCENSORED: | |
463 | raise RevlogError(_('node %s is not censored') % expectednode) |
|
463 | raise RevlogError(_('node %s is not censored') % expectednode) | |
464 | except CensoredNodeError: |
|
464 | except CensoredNodeError: | |
465 | # must pass the censored index flag to add censored revisions |
|
465 | # must pass the censored index flag to add censored revisions | |
466 | if not flags & REVIDX_ISCENSORED: |
|
466 | if not flags & REVIDX_ISCENSORED: | |
467 | raise |
|
467 | raise | |
468 | return fulltext |
|
468 | return fulltext | |
469 |
|
469 | |||
470 | @attr.s(slots=True, frozen=True) |
|
470 | @attr.s(slots=True, frozen=True) | |
471 | class _deltainfo(object): |
|
471 | class _deltainfo(object): | |
472 | distance = attr.ib() |
|
472 | distance = attr.ib() | |
473 | deltalen = attr.ib() |
|
473 | deltalen = attr.ib() | |
474 | data = attr.ib() |
|
474 | data = attr.ib() | |
475 | base = attr.ib() |
|
475 | base = attr.ib() | |
476 | chainbase = attr.ib() |
|
476 | chainbase = attr.ib() | |
477 | chainlen = attr.ib() |
|
477 | chainlen = attr.ib() | |
478 | compresseddeltalen = attr.ib() |
|
478 | compresseddeltalen = attr.ib() | |
479 | snapshotdepth = attr.ib() |
|
479 | snapshotdepth = attr.ib() | |
480 |
|
480 | |||
481 | def isgooddeltainfo(revlog, deltainfo, revinfo): |
|
481 | def isgooddeltainfo(revlog, deltainfo, revinfo): | |
482 | """Returns True if the given delta is good. Good means that it is within |
|
482 | """Returns True if the given delta is good. Good means that it is within | |
483 | the disk span, disk size, and chain length bounds that we know to be |
|
483 | the disk span, disk size, and chain length bounds that we know to be | |
484 | performant.""" |
|
484 | performant.""" | |
485 | if deltainfo is None: |
|
485 | if deltainfo is None: | |
486 | return False |
|
486 | return False | |
487 |
|
487 | |||
488 | # - 'deltainfo.distance' is the distance from the base revision -- |
|
488 | # - 'deltainfo.distance' is the distance from the base revision -- | |
489 | # bounding it limits the amount of I/O we need to do. |
|
489 | # bounding it limits the amount of I/O we need to do. | |
490 | # - 'deltainfo.compresseddeltalen' is the sum of the total size of |
|
490 | # - 'deltainfo.compresseddeltalen' is the sum of the total size of | |
491 | # deltas we need to apply -- bounding it limits the amount of CPU |
|
491 | # deltas we need to apply -- bounding it limits the amount of CPU | |
492 | # we consume. |
|
492 | # we consume. | |
493 |
|
493 | |||
494 | if revlog._sparserevlog: |
|
494 | if revlog._sparserevlog: | |
495 | # As sparse-read will be used, we can consider that the distance, |
|
495 | # As sparse-read will be used, we can consider that the distance, | |
496 | # instead of being the span of the whole chunk, |
|
496 | # instead of being the span of the whole chunk, | |
497 | # is the span of the largest read chunk |
|
497 | # is the span of the largest read chunk | |
498 | base = deltainfo.base |
|
498 | base = deltainfo.base | |
499 |
|
499 | |||
500 | if base != nullrev: |
|
500 | if base != nullrev: | |
501 | deltachain = revlog._deltachain(base)[0] |
|
501 | deltachain = revlog._deltachain(base)[0] | |
502 | else: |
|
502 | else: | |
503 | deltachain = [] |
|
503 | deltachain = [] | |
504 |
|
504 | |||
505 | # search for the first non-snapshot revision |
|
505 | # search for the first non-snapshot revision | |
506 | for idx, r in enumerate(deltachain): |
|
506 | for idx, r in enumerate(deltachain): | |
507 | if not revlog.issnapshot(r): |
|
507 | if not revlog.issnapshot(r): | |
508 | break |
|
508 | break | |
509 | deltachain = deltachain[idx:] |
|
509 | deltachain = deltachain[idx:] | |
510 | chunks = slicechunk(revlog, deltachain, deltainfo) |
|
510 | chunks = slicechunk(revlog, deltachain, deltainfo) | |
511 | all_span = [segmentspan(revlog, revs, deltainfo) |
|
511 | all_span = [segmentspan(revlog, revs, deltainfo) | |
512 | for revs in chunks] |
|
512 | for revs in chunks] | |
513 | distance = max(all_span) |
|
513 | distance = max(all_span) | |
514 | else: |
|
514 | else: | |
515 | distance = deltainfo.distance |
|
515 | distance = deltainfo.distance | |
516 |
|
516 | |||
517 | textlen = revinfo.textlen |
|
517 | textlen = revinfo.textlen | |
518 | defaultmax = textlen * 4 |
|
518 | defaultmax = textlen * 4 | |
519 | maxdist = revlog._maxdeltachainspan |
|
519 | maxdist = revlog._maxdeltachainspan | |
520 | if not maxdist: |
|
520 | if not maxdist: | |
521 | maxdist = distance # ensure the conditional pass |
|
521 | maxdist = distance # ensure the conditional pass | |
522 | maxdist = max(maxdist, defaultmax) |
|
522 | maxdist = max(maxdist, defaultmax) | |
523 | if revlog._sparserevlog and maxdist < revlog._srmingapsize: |
|
523 | if revlog._sparserevlog and maxdist < revlog._srmingapsize: | |
524 | # In multiple place, we are ignoring irrelevant data range below a |
|
524 | # In multiple place, we are ignoring irrelevant data range below a | |
525 | # certain size. Be also apply this tradeoff here and relax span |
|
525 | # certain size. Be also apply this tradeoff here and relax span | |
526 | # constraint for small enought content. |
|
526 | # constraint for small enought content. | |
527 | maxdist = revlog._srmingapsize |
|
527 | maxdist = revlog._srmingapsize | |
528 |
|
528 | |||
529 | # Bad delta from read span: |
|
529 | # Bad delta from read span: | |
530 | # |
|
530 | # | |
531 | # If the span of data read is larger than the maximum allowed. |
|
531 | # If the span of data read is larger than the maximum allowed. | |
532 | if maxdist < distance: |
|
532 | if maxdist < distance: | |
533 | return False |
|
533 | return False | |
534 |
|
534 | |||
535 | # Bad delta from new delta size: |
|
535 | # Bad delta from new delta size: | |
536 | # |
|
536 | # | |
537 | # If the delta size is larger than the target text, storing the |
|
537 | # If the delta size is larger than the target text, storing the | |
538 | # delta will be inefficient. |
|
538 | # delta will be inefficient. | |
539 | if textlen < deltainfo.deltalen: |
|
539 | if textlen < deltainfo.deltalen: | |
540 | return False |
|
540 | return False | |
541 |
|
541 | |||
542 | # Bad delta from cumulated payload size: |
|
542 | # Bad delta from cumulated payload size: | |
543 | # |
|
543 | # | |
544 | # If the sum of delta get larger than K * target text length. |
|
544 | # If the sum of delta get larger than K * target text length. | |
545 | if textlen * LIMIT_DELTA2TEXT < deltainfo.compresseddeltalen: |
|
545 | if textlen * LIMIT_DELTA2TEXT < deltainfo.compresseddeltalen: | |
546 | return False |
|
546 | return False | |
547 |
|
547 | |||
548 | # Bad delta from chain length: |
|
548 | # Bad delta from chain length: | |
549 | # |
|
549 | # | |
550 | # If the number of delta in the chain gets too high. |
|
550 | # If the number of delta in the chain gets too high. | |
551 | if (revlog._maxchainlen |
|
551 | if (revlog._maxchainlen | |
552 | and revlog._maxchainlen < deltainfo.chainlen): |
|
552 | and revlog._maxchainlen < deltainfo.chainlen): | |
553 | return False |
|
553 | return False | |
554 |
|
554 | |||
555 | # bad delta from intermediate snapshot size limit |
|
555 | # bad delta from intermediate snapshot size limit | |
556 | # |
|
556 | # | |
557 | # If an intermediate snapshot size is higher than the limit. The |
|
557 | # If an intermediate snapshot size is higher than the limit. The | |
558 | # limit exist to prevent endless chain of intermediate delta to be |
|
558 | # limit exist to prevent endless chain of intermediate delta to be | |
559 | # created. |
|
559 | # created. | |
560 | if (deltainfo.snapshotdepth is not None and |
|
560 | if (deltainfo.snapshotdepth is not None and | |
561 | (textlen >> deltainfo.snapshotdepth) < deltainfo.deltalen): |
|
561 | (textlen >> deltainfo.snapshotdepth) < deltainfo.deltalen): | |
562 | return False |
|
562 | return False | |
563 |
|
563 | |||
564 | # bad delta if new intermediate snapshot is larger than the previous |
|
564 | # bad delta if new intermediate snapshot is larger than the previous | |
565 | # snapshot |
|
565 | # snapshot | |
566 | if (deltainfo.snapshotdepth |
|
566 | if (deltainfo.snapshotdepth | |
567 | and revlog.length(deltainfo.base) < deltainfo.deltalen): |
|
567 | and revlog.length(deltainfo.base) < deltainfo.deltalen): | |
568 | return False |
|
568 | return False | |
569 |
|
569 | |||
570 | return True |
|
570 | return True | |
571 |
|
571 | |||
572 | def _candidategroups(revlog, textlen, p1, p2, cachedelta): |
|
572 | def _candidategroups(revlog, textlen, p1, p2, cachedelta): | |
573 | """Provides group of revision to be tested as delta base |
|
573 | """Provides group of revision to be tested as delta base | |
574 |
|
574 | |||
575 | This top level function focus on emitting groups with unique and worthwhile |
|
575 | This top level function focus on emitting groups with unique and worthwhile | |
576 | content. See _raw_candidate_groups for details about the group order. |
|
576 | content. See _raw_candidate_groups for details about the group order. | |
577 | """ |
|
577 | """ | |
578 | # should we try to build a delta? |
|
578 | # should we try to build a delta? | |
579 | if not (len(revlog) and revlog._storedeltachains): |
|
579 | if not (len(revlog) and revlog._storedeltachains): | |
580 | yield None |
|
580 | yield None | |
581 | return |
|
581 | return | |
582 |
|
582 | |||
583 | deltalength = revlog.length |
|
583 | deltalength = revlog.length | |
584 | deltaparent = revlog.deltaparent |
|
584 | deltaparent = revlog.deltaparent | |
|
585 | good = None | |||
585 |
|
586 | |||
586 | deltas_limit = textlen * LIMIT_DELTA2TEXT |
|
587 | deltas_limit = textlen * LIMIT_DELTA2TEXT | |
587 |
|
588 | |||
588 | tested = set([nullrev]) |
|
589 | tested = set([nullrev]) | |
589 | for temptative in _refinedgroups(revlog, p1, p2, cachedelta): |
|
590 | for temptative in _refinedgroups(revlog, p1, p2, cachedelta): | |
590 | group = [] |
|
591 | group = [] | |
591 | for rev in temptative: |
|
592 | for rev in temptative: | |
592 | # skip over empty delta (no need to include them in a chain) |
|
593 | # skip over empty delta (no need to include them in a chain) | |
593 | while not (rev == nullrev or rev in tested or deltalength(rev)): |
|
594 | while not (rev == nullrev or rev in tested or deltalength(rev)): | |
594 | rev = deltaparent(rev) |
|
595 | rev = deltaparent(rev) | |
595 | tested.add(rev) |
|
596 | tested.add(rev) | |
596 | # filter out revision we tested already |
|
597 | # filter out revision we tested already | |
597 | if rev in tested: |
|
598 | if rev in tested: | |
598 | continue |
|
599 | continue | |
599 | tested.add(rev) |
|
600 | tested.add(rev) | |
600 | # filter out delta base that will never produce good delta |
|
601 | # filter out delta base that will never produce good delta | |
601 | if deltas_limit < revlog.length(rev): |
|
602 | if deltas_limit < revlog.length(rev): | |
602 | continue |
|
603 | continue | |
603 | # no need to try a delta against nullrev, this will be done as a |
|
604 | # no need to try a delta against nullrev, this will be done as a | |
604 | # last resort. |
|
605 | # last resort. | |
605 | if rev == nullrev: |
|
606 | if rev == nullrev: | |
606 | continue |
|
607 | continue | |
607 | # no delta for rawtext-changing revs (see "candelta" for why) |
|
608 | # no delta for rawtext-changing revs (see "candelta" for why) | |
608 | if revlog.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS: |
|
609 | if revlog.flags(rev) & REVIDX_RAWTEXT_CHANGING_FLAGS: | |
609 | continue |
|
610 | continue | |
610 | group.append(rev) |
|
611 | group.append(rev) | |
611 | if group: |
|
612 | if group: | |
612 | # XXX: in the sparse revlog case, group can become large, |
|
613 | # XXX: in the sparse revlog case, group can become large, | |
613 | # impacting performances. Some bounding or slicing mecanism |
|
614 | # impacting performances. Some bounding or slicing mecanism | |
614 | # would help to reduce this impact. |
|
615 | # would help to reduce this impact. | |
615 | yield tuple(group) |
|
616 | good = yield tuple(group) | |
|
617 | if good is not None: | |||
|
618 | break | |||
616 | yield None |
|
619 | yield None | |
617 |
|
620 | |||
618 | def _findsnapshots(revlog, cache, start_rev): |
|
621 | def _findsnapshots(revlog, cache, start_rev): | |
619 | """find snapshot from start_rev to tip""" |
|
622 | """find snapshot from start_rev to tip""" | |
620 | deltaparent = revlog.deltaparent |
|
623 | deltaparent = revlog.deltaparent | |
621 | issnapshot = revlog.issnapshot |
|
624 | issnapshot = revlog.issnapshot | |
622 | for rev in revlog.revs(start_rev): |
|
625 | for rev in revlog.revs(start_rev): | |
623 | if issnapshot(rev): |
|
626 | if issnapshot(rev): | |
624 | cache[deltaparent(rev)].append(rev) |
|
627 | cache[deltaparent(rev)].append(rev) | |
625 |
|
628 | |||
626 | def _refinedgroups(revlog, p1, p2, cachedelta): |
|
629 | def _refinedgroups(revlog, p1, p2, cachedelta): | |
627 | good = None |
|
630 | good = None | |
628 | for candidates in _rawgroups(revlog, p1, p2, cachedelta): |
|
631 | for candidates in _rawgroups(revlog, p1, p2, cachedelta): | |
629 | good = yield candidates |
|
632 | good = yield candidates | |
630 | if good is not None: |
|
633 | if good is not None: | |
631 | break |
|
634 | break | |
632 |
|
635 | |||
633 | def _rawgroups(revlog, p1, p2, cachedelta): |
|
636 | def _rawgroups(revlog, p1, p2, cachedelta): | |
634 | """Provides group of revision to be tested as delta base |
|
637 | """Provides group of revision to be tested as delta base | |
635 |
|
638 | |||
636 | This lower level function focus on emitting delta theorically interresting |
|
639 | This lower level function focus on emitting delta theorically interresting | |
637 | without looking it any practical details. |
|
640 | without looking it any practical details. | |
638 |
|
641 | |||
639 | The group order aims at providing fast or small candidates first. |
|
642 | The group order aims at providing fast or small candidates first. | |
640 | """ |
|
643 | """ | |
641 | gdelta = revlog._generaldelta |
|
644 | gdelta = revlog._generaldelta | |
642 | sparse = revlog._sparserevlog |
|
645 | sparse = revlog._sparserevlog | |
643 | curr = len(revlog) |
|
646 | curr = len(revlog) | |
644 | prev = curr - 1 |
|
647 | prev = curr - 1 | |
645 | deltachain = lambda rev: revlog._deltachain(rev)[0] |
|
648 | deltachain = lambda rev: revlog._deltachain(rev)[0] | |
646 |
|
649 | |||
647 | # First we try to reuse a the delta contained in the bundle. |
|
650 | # First we try to reuse a the delta contained in the bundle. | |
648 | # (or from the source revlog) |
|
651 | # (or from the source revlog) | |
649 | # |
|
652 | # | |
650 | # This logic only applies to general delta repositories and can be disabled |
|
653 | # This logic only applies to general delta repositories and can be disabled | |
651 | # through configuration. Disabling reuse of source delta is useful when |
|
654 | # through configuration. Disabling reuse of source delta is useful when | |
652 | # we want to make sure we recomputed "optimal" deltas. |
|
655 | # we want to make sure we recomputed "optimal" deltas. | |
653 | if cachedelta and gdelta and revlog._lazydeltabase: |
|
656 | if cachedelta and gdelta and revlog._lazydeltabase: | |
654 | # Assume what we received from the server is a good choice |
|
657 | # Assume what we received from the server is a good choice | |
655 | # build delta will reuse the cache |
|
658 | # build delta will reuse the cache | |
656 | yield (cachedelta[0],) |
|
659 | yield (cachedelta[0],) | |
657 |
|
660 | |||
658 | if gdelta: |
|
661 | if gdelta: | |
659 | # exclude already lazy tested base if any |
|
662 | # exclude already lazy tested base if any | |
660 | parents = [p for p in (p1, p2) if p != nullrev] |
|
663 | parents = [p for p in (p1, p2) if p != nullrev] | |
661 |
|
664 | |||
662 | if not revlog._deltabothparents and len(parents) == 2: |
|
665 | if not revlog._deltabothparents and len(parents) == 2: | |
663 | parents.sort() |
|
666 | parents.sort() | |
664 | # To minimize the chance of having to build a fulltext, |
|
667 | # To minimize the chance of having to build a fulltext, | |
665 | # pick first whichever parent is closest to us (max rev) |
|
668 | # pick first whichever parent is closest to us (max rev) | |
666 | yield (parents[1],) |
|
669 | yield (parents[1],) | |
667 | # then the other one (min rev) if the first did not fit |
|
670 | # then the other one (min rev) if the first did not fit | |
668 | yield (parents[0],) |
|
671 | yield (parents[0],) | |
669 | elif len(parents) > 0: |
|
672 | elif len(parents) > 0: | |
670 | # Test all parents (1 or 2), and keep the best candidate |
|
673 | # Test all parents (1 or 2), and keep the best candidate | |
671 | yield parents |
|
674 | yield parents | |
672 |
|
675 | |||
673 | if sparse and parents: |
|
676 | if sparse and parents: | |
674 | snapshots = collections.defaultdict(list) # map: base-rev: snapshot-rev |
|
677 | snapshots = collections.defaultdict(list) # map: base-rev: snapshot-rev | |
675 | # See if we can use an existing snapshot in the parent chains to use as |
|
678 | # See if we can use an existing snapshot in the parent chains to use as | |
676 | # a base for a new intermediate-snapshot |
|
679 | # a base for a new intermediate-snapshot | |
677 | # |
|
680 | # | |
678 | # search for snapshot in parents delta chain |
|
681 | # search for snapshot in parents delta chain | |
679 | # map: snapshot-level: snapshot-rev |
|
682 | # map: snapshot-level: snapshot-rev | |
680 | parents_snaps = collections.defaultdict(set) |
|
683 | parents_snaps = collections.defaultdict(set) | |
681 | for p in parents: |
|
684 | for p in parents: | |
682 | for idx, s in enumerate(deltachain(p)): |
|
685 | for idx, s in enumerate(deltachain(p)): | |
683 | if not revlog.issnapshot(s): |
|
686 | if not revlog.issnapshot(s): | |
684 | break |
|
687 | break | |
685 | parents_snaps[idx].add(s) |
|
688 | parents_snaps[idx].add(s) | |
686 | snapfloor = min(parents_snaps[0]) + 1 |
|
689 | snapfloor = min(parents_snaps[0]) + 1 | |
687 | _findsnapshots(revlog, snapshots, snapfloor) |
|
690 | _findsnapshots(revlog, snapshots, snapfloor) | |
688 | # Test them as possible intermediate snapshot base |
|
691 | # Test them as possible intermediate snapshot base | |
689 | # We test them from highest to lowest level. High level one are more |
|
692 | # We test them from highest to lowest level. High level one are more | |
690 | # likely to result in small delta |
|
693 | # likely to result in small delta | |
691 | floor = None |
|
694 | floor = None | |
692 | for idx, snaps in sorted(parents_snaps.items(), reverse=True): |
|
695 | for idx, snaps in sorted(parents_snaps.items(), reverse=True): | |
693 | siblings = set() |
|
696 | siblings = set() | |
694 | for s in snaps: |
|
697 | for s in snaps: | |
695 | siblings.update(snapshots[s]) |
|
698 | siblings.update(snapshots[s]) | |
696 | # Before considering making a new intermediate snapshot, we check |
|
699 | # Before considering making a new intermediate snapshot, we check | |
697 | # if an existing snapshot, children of base we consider, would be |
|
700 | # if an existing snapshot, children of base we consider, would be | |
698 | # suitable. |
|
701 | # suitable. | |
699 | # |
|
702 | # | |
700 | # It give a change to reuse a delta chain "unrelated" to the |
|
703 | # It give a change to reuse a delta chain "unrelated" to the | |
701 | # current revision instead of starting our own. Without such |
|
704 | # current revision instead of starting our own. Without such | |
702 | # re-use, topological branches would keep reopening new chains. |
|
705 | # re-use, topological branches would keep reopening new chains. | |
703 | # Creating more and more snapshot as the repository grow. |
|
706 | # Creating more and more snapshot as the repository grow. | |
704 |
|
707 | |||
705 | if floor is not None: |
|
708 | if floor is not None: | |
706 | # We only do this for siblings created after the one in our |
|
709 | # We only do this for siblings created after the one in our | |
707 | # parent's delta chain. Those created before has less chances |
|
710 | # parent's delta chain. Those created before has less chances | |
708 | # to be valid base since our ancestors had to create a new |
|
711 | # to be valid base since our ancestors had to create a new | |
709 | # snapshot. |
|
712 | # snapshot. | |
710 | siblings = [r for r in siblings if floor < r] |
|
713 | siblings = [r for r in siblings if floor < r] | |
711 | yield tuple(sorted(siblings)) |
|
714 | yield tuple(sorted(siblings)) | |
712 | # then test the base from our parent's delta chain. |
|
715 | # then test the base from our parent's delta chain. | |
713 | yield tuple(sorted(snaps)) |
|
716 | yield tuple(sorted(snaps)) | |
714 | floor = min(snaps) |
|
717 | floor = min(snaps) | |
715 | # No suitable base found in the parent chain, search if any full |
|
718 | # No suitable base found in the parent chain, search if any full | |
716 | # snapshots emitted since parent's base would be a suitable base for an |
|
719 | # snapshots emitted since parent's base would be a suitable base for an | |
717 | # intermediate snapshot. |
|
720 | # intermediate snapshot. | |
718 | # |
|
721 | # | |
719 | # It give a chance to reuse a delta chain unrelated to the current |
|
722 | # It give a chance to reuse a delta chain unrelated to the current | |
720 | # revisions instead of starting our own. Without such re-use, |
|
723 | # revisions instead of starting our own. Without such re-use, | |
721 | # topological branches would keep reopening new full chains. Creating |
|
724 | # topological branches would keep reopening new full chains. Creating | |
722 | # more and more snapshot as the repository grow. |
|
725 | # more and more snapshot as the repository grow. | |
723 | yield tuple(snapshots[nullrev]) |
|
726 | yield tuple(snapshots[nullrev]) | |
724 |
|
727 | |||
725 | # other approach failed try against prev to hopefully save us a |
|
728 | # other approach failed try against prev to hopefully save us a | |
726 | # fulltext. |
|
729 | # fulltext. | |
727 | yield (prev,) |
|
730 | yield (prev,) | |
728 |
|
731 | |||
729 | class deltacomputer(object): |
|
732 | class deltacomputer(object): | |
730 | def __init__(self, revlog): |
|
733 | def __init__(self, revlog): | |
731 | self.revlog = revlog |
|
734 | self.revlog = revlog | |
732 |
|
735 | |||
733 | def buildtext(self, revinfo, fh): |
|
736 | def buildtext(self, revinfo, fh): | |
734 | """Builds a fulltext version of a revision |
|
737 | """Builds a fulltext version of a revision | |
735 |
|
738 | |||
736 | revinfo: _revisioninfo instance that contains all needed info |
|
739 | revinfo: _revisioninfo instance that contains all needed info | |
737 | fh: file handle to either the .i or the .d revlog file, |
|
740 | fh: file handle to either the .i or the .d revlog file, | |
738 | depending on whether it is inlined or not |
|
741 | depending on whether it is inlined or not | |
739 | """ |
|
742 | """ | |
740 | btext = revinfo.btext |
|
743 | btext = revinfo.btext | |
741 | if btext[0] is not None: |
|
744 | if btext[0] is not None: | |
742 | return btext[0] |
|
745 | return btext[0] | |
743 |
|
746 | |||
744 | revlog = self.revlog |
|
747 | revlog = self.revlog | |
745 | cachedelta = revinfo.cachedelta |
|
748 | cachedelta = revinfo.cachedelta | |
746 | baserev = cachedelta[0] |
|
749 | baserev = cachedelta[0] | |
747 | delta = cachedelta[1] |
|
750 | delta = cachedelta[1] | |
748 |
|
751 | |||
749 | fulltext = btext[0] = _textfromdelta(fh, revlog, baserev, delta, |
|
752 | fulltext = btext[0] = _textfromdelta(fh, revlog, baserev, delta, | |
750 | revinfo.p1, revinfo.p2, |
|
753 | revinfo.p1, revinfo.p2, | |
751 | revinfo.flags, revinfo.node) |
|
754 | revinfo.flags, revinfo.node) | |
752 | return fulltext |
|
755 | return fulltext | |
753 |
|
756 | |||
754 | def _builddeltadiff(self, base, revinfo, fh): |
|
757 | def _builddeltadiff(self, base, revinfo, fh): | |
755 | revlog = self.revlog |
|
758 | revlog = self.revlog | |
756 | t = self.buildtext(revinfo, fh) |
|
759 | t = self.buildtext(revinfo, fh) | |
757 | if revlog.iscensored(base): |
|
760 | if revlog.iscensored(base): | |
758 | # deltas based on a censored revision must replace the |
|
761 | # deltas based on a censored revision must replace the | |
759 | # full content in one patch, so delta works everywhere |
|
762 | # full content in one patch, so delta works everywhere | |
760 | header = mdiff.replacediffheader(revlog.rawsize(base), len(t)) |
|
763 | header = mdiff.replacediffheader(revlog.rawsize(base), len(t)) | |
761 | delta = header + t |
|
764 | delta = header + t | |
762 | else: |
|
765 | else: | |
763 | ptext = revlog.revision(base, _df=fh, raw=True) |
|
766 | ptext = revlog.revision(base, _df=fh, raw=True) | |
764 | delta = mdiff.textdiff(ptext, t) |
|
767 | delta = mdiff.textdiff(ptext, t) | |
765 |
|
768 | |||
766 | return delta |
|
769 | return delta | |
767 |
|
770 | |||
768 | def _builddeltainfo(self, revinfo, base, fh): |
|
771 | def _builddeltainfo(self, revinfo, base, fh): | |
769 | # can we use the cached delta? |
|
772 | # can we use the cached delta? | |
770 | if revinfo.cachedelta and revinfo.cachedelta[0] == base: |
|
773 | if revinfo.cachedelta and revinfo.cachedelta[0] == base: | |
771 | delta = revinfo.cachedelta[1] |
|
774 | delta = revinfo.cachedelta[1] | |
772 | else: |
|
775 | else: | |
773 | delta = self._builddeltadiff(base, revinfo, fh) |
|
776 | delta = self._builddeltadiff(base, revinfo, fh) | |
774 | revlog = self.revlog |
|
777 | revlog = self.revlog | |
775 | header, data = revlog.compress(delta) |
|
778 | header, data = revlog.compress(delta) | |
776 | deltalen = len(header) + len(data) |
|
779 | deltalen = len(header) + len(data) | |
777 | chainbase = revlog.chainbase(base) |
|
780 | chainbase = revlog.chainbase(base) | |
778 | offset = revlog.end(len(revlog) - 1) |
|
781 | offset = revlog.end(len(revlog) - 1) | |
779 | dist = deltalen + offset - revlog.start(chainbase) |
|
782 | dist = deltalen + offset - revlog.start(chainbase) | |
780 | if revlog._generaldelta: |
|
783 | if revlog._generaldelta: | |
781 | deltabase = base |
|
784 | deltabase = base | |
782 | else: |
|
785 | else: | |
783 | deltabase = chainbase |
|
786 | deltabase = chainbase | |
784 | chainlen, compresseddeltalen = revlog._chaininfo(base) |
|
787 | chainlen, compresseddeltalen = revlog._chaininfo(base) | |
785 | chainlen += 1 |
|
788 | chainlen += 1 | |
786 | compresseddeltalen += deltalen |
|
789 | compresseddeltalen += deltalen | |
787 |
|
790 | |||
788 | revlog = self.revlog |
|
791 | revlog = self.revlog | |
789 | snapshotdepth = None |
|
792 | snapshotdepth = None | |
790 | if deltabase == nullrev: |
|
793 | if deltabase == nullrev: | |
791 | snapshotdepth = 0 |
|
794 | snapshotdepth = 0 | |
792 | elif revlog._sparserevlog and revlog.issnapshot(deltabase): |
|
795 | elif revlog._sparserevlog and revlog.issnapshot(deltabase): | |
793 | # A delta chain should always be one full snapshot, |
|
796 | # A delta chain should always be one full snapshot, | |
794 | # zero or more semi-snapshots, and zero or more deltas |
|
797 | # zero or more semi-snapshots, and zero or more deltas | |
795 | p1, p2 = revlog.rev(revinfo.p1), revlog.rev(revinfo.p2) |
|
798 | p1, p2 = revlog.rev(revinfo.p1), revlog.rev(revinfo.p2) | |
796 | if deltabase not in (p1, p2) and revlog.issnapshot(deltabase): |
|
799 | if deltabase not in (p1, p2) and revlog.issnapshot(deltabase): | |
797 | snapshotdepth = len(revlog._deltachain(deltabase)[0]) |
|
800 | snapshotdepth = len(revlog._deltachain(deltabase)[0]) | |
798 |
|
801 | |||
799 | return _deltainfo(dist, deltalen, (header, data), deltabase, |
|
802 | return _deltainfo(dist, deltalen, (header, data), deltabase, | |
800 | chainbase, chainlen, compresseddeltalen, |
|
803 | chainbase, chainlen, compresseddeltalen, | |
801 | snapshotdepth) |
|
804 | snapshotdepth) | |
802 |
|
805 | |||
803 | def _fullsnapshotinfo(self, fh, revinfo): |
|
806 | def _fullsnapshotinfo(self, fh, revinfo): | |
804 | curr = len(self.revlog) |
|
807 | curr = len(self.revlog) | |
805 | rawtext = self.buildtext(revinfo, fh) |
|
808 | rawtext = self.buildtext(revinfo, fh) | |
806 | data = self.revlog.compress(rawtext) |
|
809 | data = self.revlog.compress(rawtext) | |
807 | compresseddeltalen = deltalen = dist = len(data[1]) + len(data[0]) |
|
810 | compresseddeltalen = deltalen = dist = len(data[1]) + len(data[0]) | |
808 | deltabase = chainbase = curr |
|
811 | deltabase = chainbase = curr | |
809 | snapshotdepth = 0 |
|
812 | snapshotdepth = 0 | |
810 | chainlen = 1 |
|
813 | chainlen = 1 | |
811 |
|
814 | |||
812 | return _deltainfo(dist, deltalen, data, deltabase, |
|
815 | return _deltainfo(dist, deltalen, data, deltabase, | |
813 | chainbase, chainlen, compresseddeltalen, |
|
816 | chainbase, chainlen, compresseddeltalen, | |
814 | snapshotdepth) |
|
817 | snapshotdepth) | |
815 |
|
818 | |||
816 | def finddeltainfo(self, revinfo, fh): |
|
819 | def finddeltainfo(self, revinfo, fh): | |
817 | """Find an acceptable delta against a candidate revision |
|
820 | """Find an acceptable delta against a candidate revision | |
818 |
|
821 | |||
819 | revinfo: information about the revision (instance of _revisioninfo) |
|
822 | revinfo: information about the revision (instance of _revisioninfo) | |
820 | fh: file handle to either the .i or the .d revlog file, |
|
823 | fh: file handle to either the .i or the .d revlog file, | |
821 | depending on whether it is inlined or not |
|
824 | depending on whether it is inlined or not | |
822 |
|
825 | |||
823 | Returns the first acceptable candidate revision, as ordered by |
|
826 | Returns the first acceptable candidate revision, as ordered by | |
824 | _candidategroups |
|
827 | _candidategroups | |
825 |
|
828 | |||
826 | If no suitable deltabase is found, we return delta info for a full |
|
829 | If no suitable deltabase is found, we return delta info for a full | |
827 | snapshot. |
|
830 | snapshot. | |
828 | """ |
|
831 | """ | |
829 | if not revinfo.textlen: |
|
832 | if not revinfo.textlen: | |
830 | return self._fullsnapshotinfo(fh, revinfo) |
|
833 | return self._fullsnapshotinfo(fh, revinfo) | |
831 |
|
834 | |||
832 | # no delta for flag processor revision (see "candelta" for why) |
|
835 | # no delta for flag processor revision (see "candelta" for why) | |
833 | # not calling candelta since only one revision needs test, also to |
|
836 | # not calling candelta since only one revision needs test, also to | |
834 | # avoid overhead fetching flags again. |
|
837 | # avoid overhead fetching flags again. | |
835 | if revinfo.flags & REVIDX_RAWTEXT_CHANGING_FLAGS: |
|
838 | if revinfo.flags & REVIDX_RAWTEXT_CHANGING_FLAGS: | |
836 | return self._fullsnapshotinfo(fh, revinfo) |
|
839 | return self._fullsnapshotinfo(fh, revinfo) | |
837 |
|
840 | |||
838 | cachedelta = revinfo.cachedelta |
|
841 | cachedelta = revinfo.cachedelta | |
839 | p1 = revinfo.p1 |
|
842 | p1 = revinfo.p1 | |
840 | p2 = revinfo.p2 |
|
843 | p2 = revinfo.p2 | |
841 | revlog = self.revlog |
|
844 | revlog = self.revlog | |
842 |
|
845 | |||
843 | deltainfo = None |
|
846 | deltainfo = None | |
844 | p1r, p2r = revlog.rev(p1), revlog.rev(p2) |
|
847 | p1r, p2r = revlog.rev(p1), revlog.rev(p2) | |
845 | groups = _candidategroups(self.revlog, revinfo.textlen, |
|
848 | groups = _candidategroups(self.revlog, revinfo.textlen, | |
846 | p1r, p2r, cachedelta) |
|
849 | p1r, p2r, cachedelta) | |
847 | candidaterevs = next(groups) |
|
850 | candidaterevs = next(groups) | |
848 | while candidaterevs is not None: |
|
851 | while candidaterevs is not None: | |
849 | nominateddeltas = [] |
|
852 | nominateddeltas = [] | |
|
853 | if deltainfo is not None: | |||
|
854 | # if we already found a good delta, | |||
|
855 | # challenge it against refined candidates | |||
|
856 | nominateddeltas.append(deltainfo) | |||
850 | for candidaterev in candidaterevs: |
|
857 | for candidaterev in candidaterevs: | |
851 | candidatedelta = self._builddeltainfo(revinfo, candidaterev, fh) |
|
858 | candidatedelta = self._builddeltainfo(revinfo, candidaterev, fh) | |
852 | if isgooddeltainfo(self.revlog, candidatedelta, revinfo): |
|
859 | if isgooddeltainfo(self.revlog, candidatedelta, revinfo): | |
853 | nominateddeltas.append(candidatedelta) |
|
860 | nominateddeltas.append(candidatedelta) | |
854 | if nominateddeltas: |
|
861 | if nominateddeltas: | |
855 | deltainfo = min(nominateddeltas, key=lambda x: x.deltalen) |
|
862 | deltainfo = min(nominateddeltas, key=lambda x: x.deltalen) | |
856 | break |
|
863 | if deltainfo is not None: | |
|
864 | candidaterevs = groups.send(deltainfo.base) | |||
|
865 | else: | |||
857 | candidaterevs = next(groups) |
|
866 | candidaterevs = next(groups) | |
858 |
|
867 | |||
859 | if deltainfo is None: |
|
868 | if deltainfo is None: | |
860 | deltainfo = self._fullsnapshotinfo(fh, revinfo) |
|
869 | deltainfo = self._fullsnapshotinfo(fh, revinfo) | |
861 | return deltainfo |
|
870 | return deltainfo |
General Comments 0
You need to be logged in to leave comments.
Login now