Show More
@@ -1,411 +1,410 b'' | |||
|
1 | 1 | # v2.py - Pure-Python implementation of the dirstate-v2 file format |
|
2 | 2 | # |
|
3 | 3 | # Copyright Mercurial Contributors |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import struct |
|
11 | 11 | |
|
12 | 12 | from ..thirdparty import attr |
|
13 | 13 | from .. import error, policy |
|
14 | 14 | |
|
15 | 15 | parsers = policy.importmod('parsers') |
|
16 | 16 | |
|
17 | 17 | |
|
18 | 18 | # Must match the constant of the same name in |
|
19 | 19 | # `rust/hg-core/src/dirstate_tree/on_disk.rs` |
|
20 | 20 | TREE_METADATA_SIZE = 44 |
|
21 | 21 | NODE_SIZE = 43 |
|
22 | 22 | |
|
23 | 23 | |
|
24 | 24 | # Must match the `TreeMetadata` Rust struct in |
|
25 | 25 | # `rust/hg-core/src/dirstate_tree/on_disk.rs`. See doc-comments there. |
|
26 | 26 | # |
|
27 | 27 | # * 4 bytes: start offset of root nodes |
|
28 | 28 | # * 4 bytes: number of root nodes |
|
29 | 29 | # * 4 bytes: total number of nodes in the tree that have an entry |
|
30 | 30 | # * 4 bytes: total number of nodes in the tree that have a copy source |
|
31 | 31 | # * 4 bytes: number of bytes in the data file that are not used anymore |
|
32 | 32 | # * 4 bytes: unused |
|
33 | 33 | # * 20 bytes: SHA-1 hash of ignore patterns |
|
34 | 34 | TREE_METADATA = struct.Struct('>LLLLL4s20s') |
|
35 | 35 | |
|
36 | 36 | |
|
37 | 37 | # Must match the `Node` Rust struct in |
|
38 | 38 | # `rust/hg-core/src/dirstate_tree/on_disk.rs`. See doc-comments there. |
|
39 | 39 | # |
|
40 | 40 | # * 4 bytes: start offset of full path |
|
41 | 41 | # * 2 bytes: length of the full path |
|
42 | 42 | # * 2 bytes: length within the full path before its "base name" |
|
43 | 43 | # * 4 bytes: start offset of the copy source if any, or zero for no copy source |
|
44 | 44 | # * 2 bytes: length of the copy source if any, or unused |
|
45 | 45 | # * 4 bytes: start offset of child nodes |
|
46 | 46 | # * 4 bytes: number of child nodes |
|
47 | 47 | # * 4 bytes: number of descendant nodes that have an entry |
|
48 | 48 | # * 4 bytes: number of descendant nodes that have a "tracked" state |
|
49 | 49 | # * 1 byte: flags |
|
50 | 50 | # * 4 bytes: expected size |
|
51 | 51 | # * 4 bytes: mtime seconds |
|
52 | 52 | # * 4 bytes: mtime nanoseconds |
|
53 | 53 | NODE = struct.Struct('>LHHLHLLLLBlll') |
|
54 | 54 | |
|
55 | 55 | |
|
56 | 56 | assert TREE_METADATA_SIZE == TREE_METADATA.size |
|
57 | 57 | assert NODE_SIZE == NODE.size |
|
58 | 58 | |
|
59 | 59 | |
|
60 | 60 | def parse_dirstate(map, copy_map, data, tree_metadata): |
|
61 | 61 | """parse a full v2-dirstate from a binary data into dictionnaries: |
|
62 | 62 | |
|
63 | 63 | - map: a {path: entry} mapping that will be filled |
|
64 | 64 | - copy_map: a {path: copy-source} mapping that will be filled |
|
65 | 65 | - data: a binary blob contains v2 nodes data |
|
66 | 66 | - tree_metadata:: a binary blob of the top level node (from the docket) |
|
67 | 67 | """ |
|
68 | 68 | ( |
|
69 | 69 | root_nodes_start, |
|
70 | 70 | root_nodes_len, |
|
71 | 71 | _nodes_with_entry_count, |
|
72 | 72 | _nodes_with_copy_source_count, |
|
73 | 73 | _unreachable_bytes, |
|
74 | 74 | _unused, |
|
75 | 75 | _ignore_patterns_hash, |
|
76 | 76 | ) = TREE_METADATA.unpack(tree_metadata) |
|
77 | 77 | parse_nodes(map, copy_map, data, root_nodes_start, root_nodes_len) |
|
78 | 78 | |
|
79 | 79 | |
|
80 | 80 | def parse_nodes(map, copy_map, data, start, len): |
|
81 | 81 | """parse <len> nodes from <data> starting at offset <start> |
|
82 | 82 | |
|
83 | 83 | This is used by parse_dirstate to recursively fill `map` and `copy_map`. |
|
84 | 84 | """ |
|
85 | 85 | for i in range(len): |
|
86 | 86 | node_start = start + NODE_SIZE * i |
|
87 | 87 | node_bytes = slice_with_len(data, node_start, NODE_SIZE) |
|
88 | 88 | ( |
|
89 | 89 | path_start, |
|
90 | 90 | path_len, |
|
91 | 91 | _basename_start, |
|
92 | 92 | copy_source_start, |
|
93 | 93 | copy_source_len, |
|
94 | 94 | children_start, |
|
95 | 95 | children_count, |
|
96 | 96 | _descendants_with_entry_count, |
|
97 | 97 | _tracked_descendants_count, |
|
98 | 98 | flags, |
|
99 | 99 | size, |
|
100 | 100 | mtime_s, |
|
101 | 101 | _mtime_ns, |
|
102 | 102 | ) = NODE.unpack(node_bytes) |
|
103 | 103 | |
|
104 | 104 | # Parse child nodes of this node recursively |
|
105 | 105 | parse_nodes(map, copy_map, data, children_start, children_count) |
|
106 | 106 | |
|
107 | 107 | item = parsers.DirstateItem.from_v2_data(flags, size, mtime_s) |
|
108 | 108 | if not item.any_tracked: |
|
109 | 109 | continue |
|
110 | 110 | path = slice_with_len(data, path_start, path_len) |
|
111 | 111 | map[path] = item |
|
112 | 112 | if copy_source_start: |
|
113 | 113 | copy_map[path] = slice_with_len( |
|
114 | 114 | data, copy_source_start, copy_source_len |
|
115 | 115 | ) |
|
116 | 116 | |
|
117 | 117 | |
|
118 | 118 | def slice_with_len(data, start, len): |
|
119 | 119 | return data[start : start + len] |
|
120 | 120 | |
|
121 | 121 | |
|
122 | 122 | @attr.s |
|
123 | 123 | class Node(object): |
|
124 | 124 | path = attr.ib() |
|
125 | 125 | entry = attr.ib() |
|
126 | 126 | parent = attr.ib(default=None) |
|
127 | 127 | children_count = attr.ib(default=0) |
|
128 | 128 | children_offset = attr.ib(default=0) |
|
129 | 129 | descendants_with_entry = attr.ib(default=0) |
|
130 | 130 | tracked_descendants = attr.ib(default=0) |
|
131 | 131 | |
|
132 | 132 | def pack(self, copy_map, paths_offset): |
|
133 | 133 | path = self.path |
|
134 | 134 | copy = copy_map.get(path) |
|
135 | 135 | entry = self.entry |
|
136 | 136 | |
|
137 | 137 | path_start = paths_offset |
|
138 | 138 | path_len = len(path) |
|
139 | 139 | basename_start = path.rfind(b'/') + 1 # 0 if rfind returns -1 |
|
140 | 140 | if copy is not None: |
|
141 | 141 | copy_source_start = paths_offset + len(path) |
|
142 | 142 | copy_source_len = len(copy) |
|
143 | 143 | else: |
|
144 | 144 | copy_source_start = 0 |
|
145 | 145 | copy_source_len = 0 |
|
146 | 146 | if entry is not None: |
|
147 | 147 | flags, size, mtime_s = entry.v2_data() |
|
148 | 148 | mtime_ns = 0 |
|
149 | 149 | else: |
|
150 | 150 | # There are no mtime-cached directories in the Python implementation |
|
151 | 151 | flags = 0 |
|
152 | mode = 0 | |
|
153 | 152 | size = 0 |
|
154 | 153 | mtime_s = 0 |
|
155 | 154 | mtime_ns = 0 |
|
156 | 155 | return NODE.pack( |
|
157 | 156 | path_start, |
|
158 | 157 | path_len, |
|
159 | 158 | basename_start, |
|
160 | 159 | copy_source_start, |
|
161 | 160 | copy_source_len, |
|
162 | 161 | self.children_offset, |
|
163 | 162 | self.children_count, |
|
164 | 163 | self.descendants_with_entry, |
|
165 | 164 | self.tracked_descendants, |
|
166 | 165 | flags, |
|
167 | 166 | size, |
|
168 | 167 | mtime_s, |
|
169 | 168 | mtime_ns, |
|
170 | 169 | ) |
|
171 | 170 | |
|
172 | 171 | |
|
173 | 172 | def pack_dirstate(map, copy_map, now): |
|
174 | 173 | """ |
|
175 | 174 | Pack `map` and `copy_map` into the dirstate v2 binary format and return |
|
176 | 175 | the bytearray. |
|
177 | 176 | `now` is a timestamp of the current filesystem time used to detect race |
|
178 | 177 | conditions in writing the dirstate to disk, see inline comment. |
|
179 | 178 | |
|
180 | 179 | The on-disk format expects a tree-like structure where the leaves are |
|
181 | 180 | written first (and sorted per-directory), going up levels until the root |
|
182 | 181 | node and writing that one to the docket. See more details on the on-disk |
|
183 | 182 | format in `mercurial/helptext/internals/dirstate-v2`. |
|
184 | 183 | |
|
185 | 184 | Since both `map` and `copy_map` are flat dicts we need to figure out the |
|
186 | 185 | hierarchy. This algorithm does so without having to build the entire tree |
|
187 | 186 | in-memory: it only keeps the minimum number of nodes around to satisfy the |
|
188 | 187 | format. |
|
189 | 188 | |
|
190 | 189 | # Algorithm explanation |
|
191 | 190 | |
|
192 | 191 | This explanation does not talk about the different counters for tracked |
|
193 | 192 | descendents and storing the copies, but that work is pretty simple once this |
|
194 | 193 | algorithm is in place. |
|
195 | 194 | |
|
196 | 195 | ## Building a subtree |
|
197 | 196 | |
|
198 | 197 | First, sort `map`: this makes it so the leaves of the tree are contiguous |
|
199 | 198 | per directory (i.e. a/b/c and a/b/d will be next to each other in the list), |
|
200 | 199 | and enables us to use the ordering of folders to have a "cursor" of the |
|
201 | 200 | current folder we're in without ever going twice in the same branch of the |
|
202 | 201 | tree. The cursor is a node that remembers its parent and any information |
|
203 | 202 | relevant to the format (see the `Node` class), building the relevant part |
|
204 | 203 | of the tree lazily. |
|
205 | 204 | Then, for each file in `map`, move the cursor into the tree to the |
|
206 | 205 | corresponding folder of the file: for example, if the very first file |
|
207 | 206 | is "a/b/c", we start from `Node[""]`, create `Node["a"]` which points to |
|
208 | 207 | its parent `Node[""]`, then create `Node["a/b"]`, which points to its parent |
|
209 | 208 | `Node["a"]`. These nodes are kept around in a stack. |
|
210 | 209 | If the next file in `map` is in the same subtree ("a/b/d" or "a/b/e/f"), we |
|
211 | 210 | add it to the stack and keep looping with the same logic of creating the |
|
212 | 211 | tree nodes as needed. If however the next file in `map` is *not* in the same |
|
213 | 212 | subtree ("a/other", if we're still in the "a/b" folder), then we know that |
|
214 | 213 | the subtree we're in is complete. |
|
215 | 214 | |
|
216 | 215 | ## Writing the subtree |
|
217 | 216 | |
|
218 | 217 | We have the entire subtree in the stack, so we start writing it to disk |
|
219 | 218 | folder by folder. The way we write a folder is to pop the stack into a list |
|
220 | 219 | until the folder changes, revert this list of direct children (to satisfy |
|
221 | 220 | the format requirement that children be sorted). This process repeats until |
|
222 | 221 | we hit the "other" subtree. |
|
223 | 222 | |
|
224 | 223 | An example: |
|
225 | 224 | a |
|
226 | 225 | dir1/b |
|
227 | 226 | dir1/c |
|
228 | 227 | dir2/dir3/d |
|
229 | 228 | dir2/dir3/e |
|
230 | 229 | dir2/f |
|
231 | 230 | |
|
232 | 231 | Would have us: |
|
233 | 232 | - add to the stack until "dir2/dir3/e" |
|
234 | 233 | - realize that "dir2/f" is in a different subtree |
|
235 | 234 | - pop "dir2/dir3/e", "dir2/dir3/d", reverse them so they're sorted and |
|
236 | 235 | pack them since the next entry is "dir2/dir3" |
|
237 | 236 | - go back up to "dir2" |
|
238 | 237 | - add "dir2/f" to the stack |
|
239 | 238 | - realize we're done with the map |
|
240 | 239 | - pop "dir2/f", "dir2/dir3" from the stack, reverse and pack them |
|
241 | 240 | - go up to the root node, do the same to write "a", "dir1" and "dir2" in |
|
242 | 241 | that order |
|
243 | 242 | |
|
244 | 243 | ## Special case for the root node |
|
245 | 244 | |
|
246 | 245 | The root node is not serialized in the format, but its information is |
|
247 | 246 | written to the docket. Again, see more details on the on-disk format in |
|
248 | 247 | `mercurial/helptext/internals/dirstate-v2`. |
|
249 | 248 | """ |
|
250 | 249 | now = int(now) |
|
251 | 250 | data = bytearray() |
|
252 | 251 | root_nodes_start = 0 |
|
253 | 252 | root_nodes_len = 0 |
|
254 | 253 | nodes_with_entry_count = 0 |
|
255 | 254 | nodes_with_copy_source_count = 0 |
|
256 | 255 | # Will always be 0 since this implementation always re-writes everything |
|
257 | 256 | # to disk |
|
258 | 257 | unreachable_bytes = 0 |
|
259 | 258 | unused = b'\x00' * 4 |
|
260 | 259 | # This is an optimization that's only useful for the Rust implementation |
|
261 | 260 | ignore_patterns_hash = b'\x00' * 20 |
|
262 | 261 | |
|
263 | 262 | if len(map) == 0: |
|
264 | 263 | tree_metadata = TREE_METADATA.pack( |
|
265 | 264 | root_nodes_start, |
|
266 | 265 | root_nodes_len, |
|
267 | 266 | nodes_with_entry_count, |
|
268 | 267 | nodes_with_copy_source_count, |
|
269 | 268 | unreachable_bytes, |
|
270 | 269 | unused, |
|
271 | 270 | ignore_patterns_hash, |
|
272 | 271 | ) |
|
273 | 272 | return data, tree_metadata |
|
274 | 273 | |
|
275 | 274 | sorted_map = sorted(map.items(), key=lambda x: x[0]) |
|
276 | 275 | |
|
277 | 276 | # Use a stack to not have to only remember the nodes we currently need |
|
278 | 277 | # instead of building the entire tree in memory |
|
279 | 278 | stack = [] |
|
280 | 279 | current_node = Node(b"", None) |
|
281 | 280 | stack.append(current_node) |
|
282 | 281 | |
|
283 | 282 | for index, (path, entry) in enumerate(sorted_map, 1): |
|
284 | 283 | if entry.need_delay(now): |
|
285 | 284 | # The file was last modified "simultaneously" with the current |
|
286 | 285 | # write to dirstate (i.e. within the same second for file- |
|
287 | 286 | # systems with a granularity of 1 sec). This commonly happens |
|
288 | 287 | # for at least a couple of files on 'update'. |
|
289 | 288 | # The user could change the file without changing its size |
|
290 | 289 | # within the same second. Invalidate the file's mtime in |
|
291 | 290 | # dirstate, forcing future 'status' calls to compare the |
|
292 | 291 | # contents of the file if the size is the same. This prevents |
|
293 | 292 | # mistakenly treating such files as clean. |
|
294 | 293 | entry.set_possibly_dirty() |
|
295 | 294 | nodes_with_entry_count += 1 |
|
296 | 295 | if path in copy_map: |
|
297 | 296 | nodes_with_copy_source_count += 1 |
|
298 | 297 | current_folder = get_folder(path) |
|
299 | 298 | current_node = move_to_correct_node_in_tree( |
|
300 | 299 | current_folder, current_node, stack |
|
301 | 300 | ) |
|
302 | 301 | |
|
303 | 302 | current_node.children_count += 1 |
|
304 | 303 | # Entries from `map` are never `None` |
|
305 | 304 | if entry.tracked: |
|
306 | 305 | current_node.tracked_descendants += 1 |
|
307 | 306 | current_node.descendants_with_entry += 1 |
|
308 | 307 | stack.append(Node(path, entry, current_node)) |
|
309 | 308 | |
|
310 | 309 | should_pack = True |
|
311 | 310 | next_path = None |
|
312 | 311 | if index < len(sorted_map): |
|
313 | 312 | # Determine if the next entry is in the same sub-tree, if so don't |
|
314 | 313 | # pack yet |
|
315 | 314 | next_path = sorted_map[index][0] |
|
316 | 315 | should_pack = not get_folder(next_path).startswith(current_folder) |
|
317 | 316 | if should_pack: |
|
318 | 317 | pack_directory_children(current_node, copy_map, data, stack) |
|
319 | 318 | while stack and current_node.path != b"": |
|
320 | 319 | # Go up the tree and write until we reach the folder of the next |
|
321 | 320 | # entry (if any, otherwise the root) |
|
322 | 321 | parent = current_node.parent |
|
323 | 322 | in_parent_folder_of_next_entry = next_path is not None and ( |
|
324 | 323 | get_folder(next_path).startswith(get_folder(stack[-1].path)) |
|
325 | 324 | ) |
|
326 | 325 | if parent is None or in_parent_folder_of_next_entry: |
|
327 | 326 | break |
|
328 | 327 | pack_directory_children(parent, copy_map, data, stack) |
|
329 | 328 | current_node = parent |
|
330 | 329 | |
|
331 | 330 | # Special case for the root node since we don't write it to disk, only its |
|
332 | 331 | # children to the docket |
|
333 | 332 | current_node = stack.pop() |
|
334 | 333 | assert current_node.path == b"", current_node.path |
|
335 | 334 | assert len(stack) == 0, len(stack) |
|
336 | 335 | |
|
337 | 336 | tree_metadata = TREE_METADATA.pack( |
|
338 | 337 | current_node.children_offset, |
|
339 | 338 | current_node.children_count, |
|
340 | 339 | nodes_with_entry_count, |
|
341 | 340 | nodes_with_copy_source_count, |
|
342 | 341 | unreachable_bytes, |
|
343 | 342 | unused, |
|
344 | 343 | ignore_patterns_hash, |
|
345 | 344 | ) |
|
346 | 345 | |
|
347 | 346 | return data, tree_metadata |
|
348 | 347 | |
|
349 | 348 | |
|
350 | 349 | def get_folder(path): |
|
351 | 350 | """ |
|
352 | 351 | Return the folder of the path that's given, an empty string for root paths. |
|
353 | 352 | """ |
|
354 | 353 | return path.rsplit(b'/', 1)[0] if b'/' in path else b'' |
|
355 | 354 | |
|
356 | 355 | |
|
357 | 356 | def move_to_correct_node_in_tree(target_folder, current_node, stack): |
|
358 | 357 | """ |
|
359 | 358 | Move inside the dirstate node tree to the node corresponding to |
|
360 | 359 | `target_folder`, creating the missing nodes along the way if needed. |
|
361 | 360 | """ |
|
362 | 361 | while target_folder != current_node.path: |
|
363 | 362 | if target_folder.startswith(current_node.path): |
|
364 | 363 | # We need to go down a folder |
|
365 | 364 | prefix = target_folder[len(current_node.path) :].lstrip(b'/') |
|
366 | 365 | subfolder_name = prefix.split(b'/', 1)[0] |
|
367 | 366 | if current_node.path: |
|
368 | 367 | subfolder_path = current_node.path + b'/' + subfolder_name |
|
369 | 368 | else: |
|
370 | 369 | subfolder_path = subfolder_name |
|
371 | 370 | next_node = stack[-1] |
|
372 | 371 | if next_node.path == target_folder: |
|
373 | 372 | # This folder is now a file and only contains removed entries |
|
374 | 373 | # merge with the last node |
|
375 | 374 | current_node = next_node |
|
376 | 375 | else: |
|
377 | 376 | current_node.children_count += 1 |
|
378 | 377 | current_node = Node(subfolder_path, None, current_node) |
|
379 | 378 | stack.append(current_node) |
|
380 | 379 | else: |
|
381 | 380 | # We need to go up a folder |
|
382 | 381 | current_node = current_node.parent |
|
383 | 382 | return current_node |
|
384 | 383 | |
|
385 | 384 | |
|
386 | 385 | def pack_directory_children(node, copy_map, data, stack): |
|
387 | 386 | """ |
|
388 | 387 | Write the binary representation of the direct sorted children of `node` to |
|
389 | 388 | `data` |
|
390 | 389 | """ |
|
391 | 390 | direct_children = [] |
|
392 | 391 | |
|
393 | 392 | while stack[-1].path != b"" and get_folder(stack[-1].path) == node.path: |
|
394 | 393 | direct_children.append(stack.pop()) |
|
395 | 394 | if not direct_children: |
|
396 | 395 | raise error.ProgrammingError(b"no direct children for %r" % node.path) |
|
397 | 396 | |
|
398 | 397 | # Reverse the stack to get the correct sorted order |
|
399 | 398 | direct_children.reverse() |
|
400 | 399 | packed_children = bytearray() |
|
401 | 400 | # Write the paths to `data`. Pack child nodes but don't write them yet |
|
402 | 401 | for child in direct_children: |
|
403 | 402 | packed = child.pack(copy_map=copy_map, paths_offset=len(data)) |
|
404 | 403 | packed_children.extend(packed) |
|
405 | 404 | data.extend(child.path) |
|
406 | 405 | data.extend(copy_map.get(child.path, b"")) |
|
407 | 406 | node.tracked_descendants += child.tracked_descendants |
|
408 | 407 | node.descendants_with_entry += child.descendants_with_entry |
|
409 | 408 | # Write the fixed-size child nodes all together |
|
410 | 409 | node.children_offset = len(data) |
|
411 | 410 | data.extend(packed_children) |
General Comments 0
You need to be logged in to leave comments.
Login now