Show More
@@ -1,403 +1,430 b'' | |||
|
1 | 1 | // revlog.rs |
|
2 | 2 | // |
|
3 | 3 | // Copyright 2019-2020 Georges Racinet <georges.racinet@octobus.net> |
|
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 | use crate::{ |
|
9 | 9 | cindex, |
|
10 | 10 | utils::{node_from_py_bytes, node_from_py_object}, |
|
11 | 11 | }; |
|
12 | 12 | use cpython::{ |
|
13 | 13 | exc::{IndexError, ValueError}, |
|
14 | 14 | ObjectProtocol, PyBytes, PyClone, PyDict, PyErr, PyModule, PyObject, |
|
15 | 15 | PyResult, PyString, PyTuple, Python, PythonObject, ToPyObject, |
|
16 | 16 | }; |
|
17 | 17 | use hg::{ |
|
18 | nodemap::{NodeMapError, NodeTree}, | |
|
18 | nodemap::{Block, NodeMapError, NodeTree}, | |
|
19 | 19 | revlog::{nodemap::NodeMap, RevlogIndex}, |
|
20 | 20 | NodeError, Revision, |
|
21 | 21 | }; |
|
22 | 22 | use std::cell::RefCell; |
|
23 | 23 | |
|
24 | 24 | /// Return a Struct implementing the Graph trait |
|
25 | 25 | pub(crate) fn pyindex_to_graph( |
|
26 | 26 | py: Python, |
|
27 | 27 | index: PyObject, |
|
28 | 28 | ) -> PyResult<cindex::Index> { |
|
29 | 29 | match index.extract::<MixedIndex>(py) { |
|
30 | 30 | Ok(midx) => Ok(midx.clone_cindex(py)), |
|
31 | 31 | Err(_) => cindex::Index::new(py, index), |
|
32 | 32 | } |
|
33 | 33 | } |
|
34 | 34 | |
|
35 | 35 | py_class!(pub class MixedIndex |py| { |
|
36 | 36 | data cindex: RefCell<cindex::Index>; |
|
37 | 37 | data nt: RefCell<Option<NodeTree>>; |
|
38 | data docket: RefCell<Option<PyObject>>; | |
|
38 | 39 | |
|
39 | 40 | def __new__(_cls, cindex: PyObject) -> PyResult<MixedIndex> { |
|
40 | 41 | Self::new(py, cindex) |
|
41 | 42 | } |
|
42 | 43 | |
|
43 | 44 | /// Compatibility layer used for Python consumers needing access to the C index |
|
44 | 45 | /// |
|
45 | 46 | /// Only use case so far is `scmutil.shortesthexnodeidprefix`, |
|
46 | 47 | /// that may need to build a custom `nodetree`, based on a specified revset. |
|
47 | 48 | /// With a Rust implementation of the nodemap, we will be able to get rid of |
|
48 | 49 | /// this, by exposing our own standalone nodemap class, |
|
49 | 50 | /// ready to accept `MixedIndex`. |
|
50 | 51 | def get_cindex(&self) -> PyResult<PyObject> { |
|
51 | 52 | Ok(self.cindex(py).borrow().inner().clone_ref(py)) |
|
52 | 53 | } |
|
53 | 54 | |
|
54 | 55 | // Index API involving nodemap, as defined in mercurial/pure/parsers.py |
|
55 | 56 | |
|
56 | 57 | /// Return Revision if found, raises a bare `error.RevlogError` |
|
57 | 58 | /// in case of ambiguity, same as C version does |
|
58 | 59 | def get_rev(&self, node: PyBytes) -> PyResult<Option<Revision>> { |
|
59 | 60 | let opt = self.get_nodetree(py)?.borrow(); |
|
60 | 61 | let nt = opt.as_ref().unwrap(); |
|
61 | 62 | let idx = &*self.cindex(py).borrow(); |
|
62 | 63 | let node = node_from_py_bytes(py, &node)?; |
|
63 | 64 | nt.find_bin(idx, (&node).into()).map_err(|e| nodemap_error(py, e)) |
|
64 | 65 | } |
|
65 | 66 | |
|
66 | 67 | /// same as `get_rev()` but raises a bare `error.RevlogError` if node |
|
67 | 68 | /// is not found. |
|
68 | 69 | /// |
|
69 | 70 | /// No need to repeat `node` in the exception, `mercurial/revlog.py` |
|
70 | 71 | /// will catch and rewrap with it |
|
71 | 72 | def rev(&self, node: PyBytes) -> PyResult<Revision> { |
|
72 | 73 | self.get_rev(py, node)?.ok_or_else(|| revlog_error(py)) |
|
73 | 74 | } |
|
74 | 75 | |
|
75 | 76 | /// return True if the node exist in the index |
|
76 | 77 | def has_node(&self, node: PyBytes) -> PyResult<bool> { |
|
77 | 78 | self.get_rev(py, node).map(|opt| opt.is_some()) |
|
78 | 79 | } |
|
79 | 80 | |
|
80 | 81 | /// find length of shortest hex nodeid of a binary ID |
|
81 | 82 | def shortest(&self, node: PyBytes) -> PyResult<usize> { |
|
82 | 83 | let opt = self.get_nodetree(py)?.borrow(); |
|
83 | 84 | let nt = opt.as_ref().unwrap(); |
|
84 | 85 | let idx = &*self.cindex(py).borrow(); |
|
85 | 86 | match nt.unique_prefix_len_node(idx, &node_from_py_bytes(py, &node)?) |
|
86 | 87 | { |
|
87 | 88 | Ok(Some(l)) => Ok(l), |
|
88 | 89 | Ok(None) => Err(revlog_error(py)), |
|
89 | 90 | Err(e) => Err(nodemap_error(py, e)), |
|
90 | 91 | } |
|
91 | 92 | } |
|
92 | 93 | |
|
93 | 94 | def partialmatch(&self, node: PyObject) -> PyResult<Option<PyBytes>> { |
|
94 | 95 | let opt = self.get_nodetree(py)?.borrow(); |
|
95 | 96 | let nt = opt.as_ref().unwrap(); |
|
96 | 97 | let idx = &*self.cindex(py).borrow(); |
|
97 | 98 | |
|
98 | 99 | let node_as_string = if cfg!(feature = "python3-sys") { |
|
99 | 100 | node.cast_as::<PyString>(py)?.to_string(py)?.to_string() |
|
100 | 101 | } |
|
101 | 102 | else { |
|
102 | 103 | let node = node.extract::<PyBytes>(py)?; |
|
103 | 104 | String::from_utf8_lossy(node.data(py)).to_string() |
|
104 | 105 | }; |
|
105 | 106 | |
|
106 | 107 | nt.find_hex(idx, &node_as_string) |
|
107 | 108 | // TODO make an inner API returning the node directly |
|
108 | 109 | .map(|opt| opt.map( |
|
109 | 110 | |rev| PyBytes::new(py, idx.node(rev).unwrap().as_bytes()))) |
|
110 | 111 | .map_err(|e| nodemap_error(py, e)) |
|
111 | 112 | |
|
112 | 113 | } |
|
113 | 114 | |
|
114 | 115 | /// append an index entry |
|
115 | 116 | def append(&self, tup: PyTuple) -> PyResult<PyObject> { |
|
116 | 117 | if tup.len(py) < 8 { |
|
117 | 118 | // this is better than the panic promised by tup.get_item() |
|
118 | 119 | return Err( |
|
119 | 120 | PyErr::new::<IndexError, _>(py, "tuple index out of range")) |
|
120 | 121 | } |
|
121 | 122 | let node_bytes = tup.get_item(py, 7).extract(py)?; |
|
122 | 123 | let node = node_from_py_object(py, &node_bytes)?; |
|
123 | 124 | |
|
124 | 125 | let mut idx = self.cindex(py).borrow_mut(); |
|
125 | 126 | let rev = idx.len() as Revision; |
|
126 | 127 | |
|
127 | 128 | idx.append(py, tup)?; |
|
128 | 129 | self.get_nodetree(py)?.borrow_mut().as_mut().unwrap() |
|
129 | 130 | .insert(&*idx, &node, rev) |
|
130 | 131 | .map_err(|e| nodemap_error(py, e))?; |
|
131 | 132 | Ok(py.None()) |
|
132 | 133 | } |
|
133 | 134 | |
|
134 | 135 | def __delitem__(&self, key: PyObject) -> PyResult<()> { |
|
135 | 136 | // __delitem__ is both for `del idx[r]` and `del idx[r1:r2]` |
|
136 | 137 | self.cindex(py).borrow().inner().del_item(py, key)?; |
|
137 | 138 | let mut opt = self.get_nodetree(py)?.borrow_mut(); |
|
138 | 139 | let mut nt = opt.as_mut().unwrap(); |
|
139 | 140 | nt.invalidate_all(); |
|
140 | 141 | self.fill_nodemap(py, &mut nt)?; |
|
141 | 142 | Ok(()) |
|
142 | 143 | } |
|
143 | 144 | |
|
144 | 145 | // |
|
145 | 146 | // Reforwarded C index API |
|
146 | 147 | // |
|
147 | 148 | |
|
148 | 149 | // index_methods (tp_methods). Same ordering as in revlog.c |
|
149 | 150 | |
|
150 | 151 | /// return the gca set of the given revs |
|
151 | 152 | def ancestors(&self, *args, **kw) -> PyResult<PyObject> { |
|
152 | 153 | self.call_cindex(py, "ancestors", args, kw) |
|
153 | 154 | } |
|
154 | 155 | |
|
155 | 156 | /// return the heads of the common ancestors of the given revs |
|
156 | 157 | def commonancestorsheads(&self, *args, **kw) -> PyResult<PyObject> { |
|
157 | 158 | self.call_cindex(py, "commonancestorsheads", args, kw) |
|
158 | 159 | } |
|
159 | 160 | |
|
160 | 161 | /// clear the index caches |
|
161 | 162 | def clearcaches(&self, *args, **kw) -> PyResult<PyObject> { |
|
162 | 163 | self.call_cindex(py, "clearcaches", args, kw) |
|
163 | 164 | } |
|
164 | 165 | |
|
165 | 166 | /// get an index entry |
|
166 | 167 | def get(&self, *args, **kw) -> PyResult<PyObject> { |
|
167 | 168 | self.call_cindex(py, "get", args, kw) |
|
168 | 169 | } |
|
169 | 170 | |
|
170 | 171 | /// compute phases |
|
171 | 172 | def computephasesmapsets(&self, *args, **kw) -> PyResult<PyObject> { |
|
172 | 173 | self.call_cindex(py, "computephasesmapsets", args, kw) |
|
173 | 174 | } |
|
174 | 175 | |
|
175 | 176 | /// reachableroots |
|
176 | 177 | def reachableroots2(&self, *args, **kw) -> PyResult<PyObject> { |
|
177 | 178 | self.call_cindex(py, "reachableroots2", args, kw) |
|
178 | 179 | } |
|
179 | 180 | |
|
180 | 181 | /// get head revisions |
|
181 | 182 | def headrevs(&self, *args, **kw) -> PyResult<PyObject> { |
|
182 | 183 | self.call_cindex(py, "headrevs", args, kw) |
|
183 | 184 | } |
|
184 | 185 | |
|
185 | 186 | /// get filtered head revisions |
|
186 | 187 | def headrevsfiltered(&self, *args, **kw) -> PyResult<PyObject> { |
|
187 | 188 | self.call_cindex(py, "headrevsfiltered", args, kw) |
|
188 | 189 | } |
|
189 | 190 | |
|
190 | 191 | /// True if the object is a snapshot |
|
191 | 192 | def issnapshot(&self, *args, **kw) -> PyResult<PyObject> { |
|
192 | 193 | self.call_cindex(py, "issnapshot", args, kw) |
|
193 | 194 | } |
|
194 | 195 | |
|
195 | 196 | /// Gather snapshot data in a cache dict |
|
196 | 197 | def findsnapshots(&self, *args, **kw) -> PyResult<PyObject> { |
|
197 | 198 | self.call_cindex(py, "findsnapshots", args, kw) |
|
198 | 199 | } |
|
199 | 200 | |
|
200 | 201 | /// determine revisions with deltas to reconstruct fulltext |
|
201 | 202 | def deltachain(&self, *args, **kw) -> PyResult<PyObject> { |
|
202 | 203 | self.call_cindex(py, "deltachain", args, kw) |
|
203 | 204 | } |
|
204 | 205 | |
|
205 | 206 | /// slice planned chunk read to reach a density threshold |
|
206 | 207 | def slicechunktodensity(&self, *args, **kw) -> PyResult<PyObject> { |
|
207 | 208 | self.call_cindex(py, "slicechunktodensity", args, kw) |
|
208 | 209 | } |
|
209 | 210 | |
|
210 | 211 | /// stats for the index |
|
211 | 212 | def stats(&self, *args, **kw) -> PyResult<PyObject> { |
|
212 | 213 | self.call_cindex(py, "stats", args, kw) |
|
213 | 214 | } |
|
214 | 215 | |
|
215 | 216 | // index_sequence_methods and index_mapping_methods. |
|
216 | 217 | // |
|
217 | 218 | // Since we call back through the high level Python API, |
|
218 | 219 | // there's no point making a distinction between index_get |
|
219 | 220 | // and index_getitem. |
|
220 | 221 | |
|
221 | 222 | def __len__(&self) -> PyResult<usize> { |
|
222 | 223 | self.cindex(py).borrow().inner().len(py) |
|
223 | 224 | } |
|
224 | 225 | |
|
225 | 226 | def __getitem__(&self, key: PyObject) -> PyResult<PyObject> { |
|
226 | 227 | // this conversion seems needless, but that's actually because |
|
227 | 228 | // `index_getitem` does not handle conversion from PyLong, |
|
228 | 229 | // which expressions such as [e for e in index] internally use. |
|
229 | 230 | // Note that we don't seem to have a direct way to call |
|
230 | 231 | // PySequence_GetItem (does the job), which would be better for |
|
231 | 232 | // for performance |
|
232 | 233 | let key = match key.extract::<Revision>(py) { |
|
233 | 234 | Ok(rev) => rev.to_py_object(py).into_object(), |
|
234 | 235 | Err(_) => key, |
|
235 | 236 | }; |
|
236 | 237 | self.cindex(py).borrow().inner().get_item(py, key) |
|
237 | 238 | } |
|
238 | 239 | |
|
239 | 240 | def __setitem__(&self, key: PyObject, value: PyObject) -> PyResult<()> { |
|
240 | 241 | self.cindex(py).borrow().inner().set_item(py, key, value) |
|
241 | 242 | } |
|
242 | 243 | |
|
243 | 244 | def __contains__(&self, item: PyObject) -> PyResult<bool> { |
|
244 | 245 | // ObjectProtocol does not seem to provide contains(), so |
|
245 | 246 | // this is an equivalent implementation of the index_contains() |
|
246 | 247 | // defined in revlog.c |
|
247 | 248 | let cindex = self.cindex(py).borrow(); |
|
248 | 249 | match item.extract::<Revision>(py) { |
|
249 | 250 | Ok(rev) => { |
|
250 | 251 | Ok(rev >= -1 && rev < cindex.inner().len(py)? as Revision) |
|
251 | 252 | } |
|
252 | 253 | Err(_) => { |
|
253 | 254 | cindex.inner().call_method( |
|
254 | 255 | py, |
|
255 | 256 | "has_node", |
|
256 | 257 | PyTuple::new(py, &[item]), |
|
257 | 258 | None)? |
|
258 | 259 | .extract(py) |
|
259 | 260 | } |
|
260 | 261 | } |
|
261 | 262 | } |
|
262 | 263 | |
|
263 | 264 | def nodemap_data_all(&self) -> PyResult<PyBytes> { |
|
264 | 265 | self.inner_nodemap_data_all(py) |
|
265 | 266 | } |
|
266 | 267 | |
|
268 | def nodemap_data_incremental(&self) -> PyResult<PyObject> { | |
|
269 | self.inner_nodemap_data_incremental(py) | |
|
270 | } | |
|
267 | 271 | |
|
268 | 272 | }); |
|
269 | 273 | |
|
270 | 274 | impl MixedIndex { |
|
271 | 275 | fn new(py: Python, cindex: PyObject) -> PyResult<MixedIndex> { |
|
272 | 276 | Self::create_instance( |
|
273 | 277 | py, |
|
274 | 278 | RefCell::new(cindex::Index::new(py, cindex)?), |
|
275 | 279 | RefCell::new(None), |
|
280 | RefCell::new(None), | |
|
276 | 281 | ) |
|
277 | 282 | } |
|
278 | 283 | |
|
279 | 284 | /// This is scaffolding at this point, but it could also become |
|
280 | 285 | /// a way to start a persistent nodemap or perform a |
|
281 | 286 | /// vacuum / repack operation |
|
282 | 287 | fn fill_nodemap( |
|
283 | 288 | &self, |
|
284 | 289 | py: Python, |
|
285 | 290 | nt: &mut NodeTree, |
|
286 | 291 | ) -> PyResult<PyObject> { |
|
287 | 292 | let index = self.cindex(py).borrow(); |
|
288 | 293 | for r in 0..index.len() { |
|
289 | 294 | let rev = r as Revision; |
|
290 | 295 | // in this case node() won't ever return None |
|
291 | 296 | nt.insert(&*index, index.node(rev).unwrap(), rev) |
|
292 | 297 | .map_err(|e| nodemap_error(py, e))? |
|
293 | 298 | } |
|
294 | 299 | Ok(py.None()) |
|
295 | 300 | } |
|
296 | 301 | |
|
297 | 302 | fn get_nodetree<'a>( |
|
298 | 303 | &'a self, |
|
299 | 304 | py: Python<'a>, |
|
300 | 305 | ) -> PyResult<&'a RefCell<Option<NodeTree>>> { |
|
301 | 306 | if self.nt(py).borrow().is_none() { |
|
302 | 307 | let readonly = Box::new(Vec::new()); |
|
303 | 308 | let mut nt = NodeTree::load_bytes(readonly, 0); |
|
304 | 309 | self.fill_nodemap(py, &mut nt)?; |
|
305 | 310 | self.nt(py).borrow_mut().replace(nt); |
|
306 | 311 | } |
|
307 | 312 | Ok(self.nt(py)) |
|
308 | 313 | } |
|
309 | 314 | |
|
310 | 315 | /// forward a method call to the underlying C index |
|
311 | 316 | fn call_cindex( |
|
312 | 317 | &self, |
|
313 | 318 | py: Python, |
|
314 | 319 | name: &str, |
|
315 | 320 | args: &PyTuple, |
|
316 | 321 | kwargs: Option<&PyDict>, |
|
317 | 322 | ) -> PyResult<PyObject> { |
|
318 | 323 | self.cindex(py) |
|
319 | 324 | .borrow() |
|
320 | 325 | .inner() |
|
321 | 326 | .call_method(py, name, args, kwargs) |
|
322 | 327 | } |
|
323 | 328 | |
|
324 | 329 | pub fn clone_cindex(&self, py: Python) -> cindex::Index { |
|
325 | 330 | self.cindex(py).borrow().clone_ref(py) |
|
326 | 331 | } |
|
327 | 332 | |
|
328 | 333 | /// Returns the full nodemap bytes to be written as-is to disk |
|
329 | 334 | fn inner_nodemap_data_all(&self, py: Python) -> PyResult<PyBytes> { |
|
330 | 335 | let nodemap = self.get_nodetree(py)?.borrow_mut().take().unwrap(); |
|
331 | 336 | let (readonly, bytes) = nodemap.into_readonly_and_added_bytes(); |
|
332 | 337 | |
|
333 | 338 | // If there's anything readonly, we need to build the data again from |
|
334 | 339 | // scratch |
|
335 | 340 | let bytes = if readonly.len() > 0 { |
|
336 | 341 | let mut nt = NodeTree::load_bytes(Box::new(vec![]), 0); |
|
337 | 342 | self.fill_nodemap(py, &mut nt)?; |
|
338 | 343 | |
|
339 | 344 | let (readonly, bytes) = nt.into_readonly_and_added_bytes(); |
|
340 | 345 | assert_eq!(readonly.len(), 0); |
|
341 | 346 | |
|
342 | 347 | bytes |
|
343 | 348 | } else { |
|
344 | 349 | bytes |
|
345 | 350 | }; |
|
346 | 351 | |
|
347 | 352 | let bytes = PyBytes::new(py, &bytes); |
|
348 | 353 | Ok(bytes) |
|
349 | 354 | } |
|
355 | ||
|
356 | /// Returns the last saved docket along with the size of any changed data | |
|
357 | /// (in number of blocks), and said data as bytes. | |
|
358 | fn inner_nodemap_data_incremental( | |
|
359 | &self, | |
|
360 | py: Python, | |
|
361 | ) -> PyResult<PyObject> { | |
|
362 | let docket = self.docket(py).borrow(); | |
|
363 | let docket = match docket.as_ref() { | |
|
364 | Some(d) => d, | |
|
365 | None => return Ok(py.None()), | |
|
366 | }; | |
|
367 | ||
|
368 | let node_tree = self.get_nodetree(py)?.borrow_mut().take().unwrap(); | |
|
369 | let masked_blocks = node_tree.masked_readonly_blocks(); | |
|
370 | let (_, data) = node_tree.into_readonly_and_added_bytes(); | |
|
371 | let changed = masked_blocks * std::mem::size_of::<Block>(); | |
|
372 | ||
|
373 | Ok((docket, changed, PyBytes::new(py, &data)) | |
|
374 | .to_py_object(py) | |
|
375 | .into_object()) | |
|
376 | } | |
|
350 | 377 | } |
|
351 | 378 | |
|
352 | 379 | fn revlog_error(py: Python) -> PyErr { |
|
353 | 380 | match py |
|
354 | 381 | .import("mercurial.error") |
|
355 | 382 | .and_then(|m| m.get(py, "RevlogError")) |
|
356 | 383 | { |
|
357 | 384 | Err(e) => e, |
|
358 | 385 | Ok(cls) => PyErr::from_instance(py, cls), |
|
359 | 386 | } |
|
360 | 387 | } |
|
361 | 388 | |
|
362 | 389 | fn rev_not_in_index(py: Python, rev: Revision) -> PyErr { |
|
363 | 390 | PyErr::new::<ValueError, _>( |
|
364 | 391 | py, |
|
365 | 392 | format!( |
|
366 | 393 | "Inconsistency: Revision {} found in nodemap \ |
|
367 | 394 | is not in revlog index", |
|
368 | 395 | rev |
|
369 | 396 | ), |
|
370 | 397 | ) |
|
371 | 398 | } |
|
372 | 399 | |
|
373 | 400 | /// Standard treatment of NodeMapError |
|
374 | 401 | fn nodemap_error(py: Python, err: NodeMapError) -> PyErr { |
|
375 | 402 | match err { |
|
376 | 403 | NodeMapError::MultipleResults => revlog_error(py), |
|
377 | 404 | NodeMapError::RevisionNotInIndex(r) => rev_not_in_index(py, r), |
|
378 | 405 | NodeMapError::InvalidNodePrefix(s) => invalid_node_prefix(py, &s), |
|
379 | 406 | } |
|
380 | 407 | } |
|
381 | 408 | |
|
382 | 409 | fn invalid_node_prefix(py: Python, ne: &NodeError) -> PyErr { |
|
383 | 410 | PyErr::new::<ValueError, _>( |
|
384 | 411 | py, |
|
385 | 412 | format!("Invalid node or prefix: {:?}", ne), |
|
386 | 413 | ) |
|
387 | 414 | } |
|
388 | 415 | |
|
389 | 416 | /// Create the module, with __package__ given from parent |
|
390 | 417 | pub fn init_module(py: Python, package: &str) -> PyResult<PyModule> { |
|
391 | 418 | let dotted_name = &format!("{}.revlog", package); |
|
392 | 419 | let m = PyModule::new(py, dotted_name)?; |
|
393 | 420 | m.add(py, "__package__", package)?; |
|
394 | 421 | m.add(py, "__doc__", "RevLog - Rust implementations")?; |
|
395 | 422 | |
|
396 | 423 | m.add_class::<MixedIndex>(py)?; |
|
397 | 424 | |
|
398 | 425 | let sys = PyModule::import(py, "sys")?; |
|
399 | 426 | let sys_modules: PyDict = sys.get(py, "modules")?.extract(py)?; |
|
400 | 427 | sys_modules.set_item(py, dotted_name, &m)?; |
|
401 | 428 | |
|
402 | 429 | Ok(m) |
|
403 | 430 | } |
General Comments 0
You need to be logged in to leave comments.
Login now