Coverage for python/lsst/resources/http.py: 20%

559 statements  

« prev     ^ index     » next       coverage.py v6.5.0, created at 2023-04-12 02:04 -0700

1# This file is part of lsst-resources. 

2# 

3# Developed for the LSST Data Management System. 

4# This product includes software developed by the LSST Project 

5# (https://www.lsst.org). 

6# See the COPYRIGHT file at the top-level directory of this distribution 

7# for details of code ownership. 

8# 

9# Use of this source code is governed by a 3-clause BSD-style 

10# license that can be found in the LICENSE file. 

11 

12from __future__ import annotations 

13 

14__all__ = ("HttpResourcePath",) 

15 

16import contextlib 

17import functools 

18import io 

19import logging 

20import math 

21import os 

22import os.path 

23import random 

24import re 

25import stat 

26import tempfile 

27import xml.etree.ElementTree as eTree 

28from typing import TYPE_CHECKING, BinaryIO, Iterator, List, Optional, Tuple, Union, cast 

29 

30import requests 

31from astropy import units as u 

32from lsst.utils.timer import time_this 

33from requests.adapters import HTTPAdapter 

34from requests.auth import AuthBase 

35from urllib3.util.retry import Retry 

36 

37from ._resourceHandles import ResourceHandleProtocol 

38from ._resourceHandles._httpResourceHandle import HttpReadResourceHandle 

39from ._resourcePath import ResourcePath 

40 

41if TYPE_CHECKING: 

42 from .utils import TransactionProtocol 

43 

44log = logging.getLogger(__name__) 

45 

46 

47class HttpResourcePathConfig: 

48 """Configuration class to encapsulate the configurable items used by class 

49 HttpResourcePath. 

50 """ 

51 

52 # Default timeouts for all HTTP requests (seconds). 

53 DEFAULT_TIMEOUT_CONNECT = 30.0 

54 DEFAULT_TIMEOUT_READ = 1_500.0 

55 

56 # Default lower and upper bounds for the backoff interval (seconds). 

57 # A value in this interval is randomly selected as the backoff factor when 

58 # requests need to be retried. 

59 DEFAULT_BACKOFF_MIN = 1.0 

60 DEFAULT_BACKOFF_MAX = 3.0 

61 

62 # Default number of connections to persist with both the front end and 

63 # back end servers. 

64 DEFAULT_FRONTEND_PERSISTENT_CONNECTIONS = 2 

65 DEFAULT_BACKEND_PERSISTENT_CONNECTIONS = 1 

66 

67 # Accepted digest algorithms 

68 ACCEPTED_DIGESTS = ("adler32", "md5", "sha-256", "sha-512") 

69 

70 _front_end_connections: Optional[int] = None 

71 _back_end_connections: Optional[int] = None 

72 _digest_algorithm: Optional[str] = None 

73 _send_expect_on_put: Optional[bool] = None 

74 _timeout: Optional[tuple[float, float]] = None 

75 _collect_memory_usage: Optional[bool] = None 

76 _backoff_min: Optional[float] = None 

77 _backoff_max: Optional[float] = None 

78 

79 @property 

80 def front_end_connections(self) -> int: 

81 """Number of persistent connections to the front end server.""" 

82 

83 if self._front_end_connections is not None: 83 ↛ 84line 83 didn't jump to line 84, because the condition on line 83 was never true

84 return self._front_end_connections 

85 

86 try: 

87 self._front_end_connections = int( 

88 os.environ.get( 

89 "LSST_HTTP_FRONTEND_PERSISTENT_CONNECTIONS", self.DEFAULT_FRONTEND_PERSISTENT_CONNECTIONS 

90 ) 

91 ) 

92 except ValueError: 

93 self._front_end_connections = self.DEFAULT_FRONTEND_PERSISTENT_CONNECTIONS 

94 

95 return self._front_end_connections 

96 

97 @property 

98 def back_end_connections(self) -> int: 

99 """Number of persistent connections to the back end servers.""" 

100 

101 if self._back_end_connections is not None: 101 ↛ 102line 101 didn't jump to line 102, because the condition on line 101 was never true

102 return self._back_end_connections 

103 

104 try: 

105 self._back_end_connections = int( 

106 os.environ.get( 

107 "LSST_HTTP_BACKEND_PERSISTENT_CONNECTIONS", self.DEFAULT_BACKEND_PERSISTENT_CONNECTIONS 

108 ) 

109 ) 

110 except ValueError: 

111 self._back_end_connections = self.DEFAULT_BACKEND_PERSISTENT_CONNECTIONS 

112 

113 return self._back_end_connections 

114 

115 @property 

116 def digest_algorithm(self) -> str: 

117 """Algorithm to ask the server to use for computing and recording 

118 digests of each file contents in PUT requests. 

119 

120 Returns 

121 ------- 

122 digest_algorithm: `str` 

123 The name of a digest algorithm or the empty string if no algotihm 

124 is configured. 

125 """ 

126 

127 if self._digest_algorithm is not None: 

128 return self._digest_algorithm 

129 

130 digest = os.environ.get("LSST_HTTP_DIGEST", "").lower() 

131 if digest not in self.ACCEPTED_DIGESTS: 

132 digest = "" 

133 

134 self._digest_algorithm = digest 

135 return self._digest_algorithm 

136 

137 @property 

138 def send_expect_on_put(self) -> bool: 

139 """Return True if a "Expect: 100-continue" header is to be sent to 

140 the server on each PUT request. 

141 

142 Some servers (e.g. dCache) uses this information as an indication that 

143 the client knows how to handle redirects to the specific server that 

144 will actually receive the data for PUT requests. 

145 """ 

146 

147 if self._send_expect_on_put is not None: 

148 return self._send_expect_on_put 

149 

150 self._send_expect_on_put = "LSST_HTTP_PUT_SEND_EXPECT_HEADER" in os.environ 

151 return self._send_expect_on_put 

152 

153 @property 

154 def timeout(self) -> tuple[float, float]: 

155 """Return a tuple with the values of timeouts for connecting to the 

156 server and reading its response, respectively. Both values are in 

157 seconds. 

158 """ 

159 

160 if self._timeout is not None: 

161 return self._timeout 

162 

163 self._timeout = (self.DEFAULT_TIMEOUT_CONNECT, self.DEFAULT_TIMEOUT_READ) 

164 try: 

165 timeout = ( 

166 float(os.environ.get("LSST_HTTP_TIMEOUT_CONNECT", self.DEFAULT_TIMEOUT_CONNECT)), 

167 float(os.environ.get("LSST_HTTP_TIMEOUT_READ", self.DEFAULT_TIMEOUT_READ)), 

168 ) 

169 if not math.isnan(timeout[0]) and not math.isnan(timeout[1]): 

170 self._timeout = timeout 

171 except ValueError: 

172 pass 

173 

174 return self._timeout 

175 

176 @property 

177 def collect_memory_usage(self) -> bool: 

178 """Return true if we want to collect memory usage when timing 

179 operations against the remote server via the `lsst.utils.time_this` 

180 context manager. 

181 """ 

182 

183 if self._collect_memory_usage is not None: 

184 return self._collect_memory_usage 

185 

186 self._collect_memory_usage = "LSST_HTTP_COLLECT_MEMORY_USAGE" in os.environ 

187 return self._collect_memory_usage 

188 

189 @property 

190 def backoff_min(self) -> float: 

191 """Lower bound of the interval from which a backoff factor is randomly 

192 selected when retrying requests (seconds). 

193 """ 

194 

195 if self._backoff_min is not None: 

196 return self._backoff_min 

197 

198 self._backoff_min = self.DEFAULT_BACKOFF_MIN 

199 try: 

200 backoff_min = float(os.environ.get("LSST_HTTP_BACKOFF_MIN", self.DEFAULT_BACKOFF_MIN)) 

201 if not math.isnan(backoff_min): 201 ↛ 206line 201 didn't jump to line 206, because the condition on line 201 was never false

202 self._backoff_min = backoff_min 

203 except ValueError: 

204 pass 

205 

206 return self._backoff_min 

207 

208 @property 

209 def backoff_max(self) -> float: 

210 """Upper bound of the interval from which a backoff factor is randomly 

211 selected when retrying requests (seconds). 

212 """ 

213 

214 if self._backoff_max is not None: 

215 return self._backoff_max 

216 

217 self._backoff_max = self.DEFAULT_BACKOFF_MAX 

218 try: 

219 backoff_max = float(os.environ.get("LSST_HTTP_BACKOFF_MAX", self.DEFAULT_BACKOFF_MAX)) 

220 if not math.isnan(backoff_max): 220 ↛ 225line 220 didn't jump to line 225, because the condition on line 220 was never false

221 self._backoff_max = backoff_max 

222 except ValueError: 

223 pass 

224 

225 return self._backoff_max 

226 

227 

228@functools.lru_cache 

229def _is_webdav_endpoint(path: Union[ResourcePath, str]) -> bool: 

230 """Check whether the remote HTTP endpoint implements WebDAV features. 

231 

232 Parameters 

233 ---------- 

234 path : `ResourcePath` or `str` 

235 URL to the resource to be checked. 

236 Should preferably refer to the root since the status is shared 

237 by all paths in that server. 

238 

239 Returns 

240 ------- 

241 _is_webdav_endpoint : `bool` 

242 True if the endpoint implements WebDAV, False if it doesn't. 

243 """ 

244 log.debug("Detecting HTTP endpoint type for '%s'...", path) 

245 try: 

246 ca_cert_bundle = os.getenv("LSST_HTTP_CACERT_BUNDLE") 

247 verify: Union[bool, str] = ca_cert_bundle if ca_cert_bundle else True 

248 resp = requests.options(str(path), verify=verify, stream=False) 

249 if resp.status_code not in (requests.codes.ok, requests.codes.created): 

250 raise ValueError( 

251 f"Unexpected response to OPTIONS request for {path}, status: {resp.status_code} " 

252 f"{resp.reason}" 

253 ) 

254 

255 # Check that "1" is part of the value of the "DAV" header. We don't 

256 # use locks, so a server complying to class 1 is enough for our 

257 # purposes. All webDAV servers must advertise at least compliance 

258 # class "1". 

259 # 

260 # Compliance classes are documented in 

261 # http://www.webdav.org/specs/rfc4918.html#dav.compliance.classes 

262 # 

263 # Examples of values for header DAV are: 

264 # DAV: 1, 2 

265 # DAV: 1, <http://apache.org/dav/propset/fs/1> 

266 if "DAV" not in resp.headers: 

267 return False 

268 else: 

269 # Convert to str to keep mypy happy 

270 compliance_class = str(resp.headers.get("DAV")) 

271 return "1" in compliance_class.replace(" ", "").split(",") 

272 except requests.exceptions.SSLError as e: 

273 log.warning( 

274 "Environment variable LSST_HTTP_CACERT_BUNDLE can be used to " 

275 "specify a bundle of certificate authorities you trust which are " 

276 "not included in the default set of trusted authorities of your " 

277 "system." 

278 ) 

279 raise e 

280 

281 

282# Tuple (path, block_size) pointing to the location of a local directory 

283# to save temporary files and the block size of the underlying file system. 

284_TMPDIR: Optional[tuple[str, int]] = None 

285 

286 

287def _get_temp_dir() -> tuple[str, int]: 

288 """Return the temporary directory path and block size. 

289 

290 This function caches its results in _TMPDIR. 

291 """ 

292 global _TMPDIR 

293 if _TMPDIR: 

294 return _TMPDIR 

295 

296 # Use the value of environment variables 'LSST_RESOURCES_TMPDIR' or 

297 # 'TMPDIR', if defined. Otherwise use current working directory. 

298 tmpdir = os.getcwd() 

299 for dir in (os.getenv(v) for v in ("LSST_RESOURCES_TMPDIR", "TMPDIR")): 

300 if dir and os.path.isdir(dir): 

301 tmpdir = dir 

302 break 

303 

304 # Compute the block size as 256 blocks of typical size 

305 # (i.e. 4096 bytes) or 10 times the file system block size, 

306 # whichever is higher. This is a reasonable compromise between 

307 # using memory for buffering and the number of system calls 

308 # issued to read from or write to temporary files. 

309 fsstats = os.statvfs(tmpdir) 

310 return (_TMPDIR := (tmpdir, max(10 * fsstats.f_bsize, 256 * 4096))) 

311 

312 

313class BearerTokenAuth(AuthBase): 

314 """Attach a bearer token 'Authorization' header to each request. 

315 

316 Parameters 

317 ---------- 

318 token : `str` 

319 Can be either the path to a local protected file which contains the 

320 value of the token or the token itself. 

321 """ 

322 

323 def __init__(self, token: str): 

324 self._token = self._path = None 

325 self._mtime: float = -1.0 

326 if not token: 

327 return 

328 

329 self._token = token 

330 if os.path.isfile(token): 

331 self._path = os.path.abspath(token) 

332 if not _is_protected(self._path): 

333 raise PermissionError( 

334 f"Bearer token file at {self._path} must be protected for access only by its owner" 

335 ) 

336 self._refresh() 

337 

338 def _refresh(self) -> None: 

339 """Read the token file (if any) if its modification time is more recent 

340 than the last time we read it. 

341 """ 

342 if not self._path: 

343 return 

344 

345 if (mtime := os.stat(self._path).st_mtime) > self._mtime: 

346 log.debug("Reading bearer token file at %s", self._path) 

347 self._mtime = mtime 

348 with open(self._path) as f: 

349 self._token = f.read().rstrip("\n") 

350 

351 def __call__(self, req: requests.PreparedRequest) -> requests.PreparedRequest: 

352 if self._token: 

353 self._refresh() 

354 req.headers["Authorization"] = f"Bearer {self._token}" 

355 return req 

356 

357 

358class SessionStore: 

359 """Cache a reusable HTTP client session per endpoint.""" 

360 

361 def __init__( 

362 self, 

363 num_pools: int = 10, 

364 max_persistent_connections: int = 1, 

365 backoff_min: float = 1.0, 

366 backoff_max: float = 3.0, 

367 ) -> None: 

368 # Dictionary to store the session associated to a given URI. The key 

369 # of the dictionary is a root URI and the value is the session. 

370 self._sessions: dict[str, requests.Session] = {} 

371 

372 # Number of connection pools to keep: there is one pool per remote 

373 # host. See documentation of urllib3 PoolManager class: 

374 # https://urllib3.readthedocs.io 

375 self._num_pools: int = num_pools 

376 

377 # Maximum number of connections per remote host to persist in each 

378 # connection pool. See urllib3 Advanced Usage documentation: 

379 # https://urllib3.readthedocs.io/en/stable/advanced-usage.html 

380 self._max_persistent_connections: int = max_persistent_connections 

381 

382 # Minimum and maximum values of the inverval to compute the exponential 

383 # backoff factor when retrying requests (seconds). 

384 self._backoff_min: float = backoff_min 

385 self._backoff_max: float = backoff_max if backoff_max > backoff_min else backoff_min + 1.0 

386 

387 def clear(self) -> None: 

388 """Destroy all previously created sessions and attempt to close 

389 underlying idle network connections. 

390 """ 

391 

392 # Close all sessions and empty the store. Idle network connections 

393 # should be closed as a consequence. We don't have means through 

394 # the API exposed by Requests to actually force closing the 

395 # underlying open sockets. 

396 for session in self._sessions.values(): 

397 session.close() 

398 

399 self._sessions.clear() 

400 

401 def get(self, rpath: ResourcePath) -> requests.Session: 

402 """Retrieve a session for accessing the remote resource at rpath. 

403 

404 Parameters 

405 ---------- 

406 rpath : `ResourcePath` 

407 URL to a resource at the remote server for which a session is to 

408 be retrieved. 

409 

410 Notes 

411 ----- 

412 Once a session is created for a given endpoint it is cached and 

413 returned every time a session is requested for any path under that same 

414 endpoint. For instance, a single session will be cached and shared 

415 for paths "https://www.example.org/path/to/file" and 

416 "https://www.example.org/any/other/path". 

417 

418 Note that "https://www.example.org" and "https://www.example.org:12345" 

419 will have different sessions since the port number is not identical. 

420 

421 In order to configure the session, some environment variables are 

422 inspected: 

423 

424 - LSST_HTTP_CACERT_BUNDLE: path to a .pem file containing the CA 

425 certificates to trust when verifying the server's certificate. 

426 

427 - LSST_HTTP_AUTH_BEARER_TOKEN: value of a bearer token or path to a 

428 local file containing a bearer token to be used as the client 

429 authentication mechanism with all requests. 

430 The permissions of the token file must be set so that only its 

431 owner can access it. 

432 If initialized, takes precedence over LSST_HTTP_AUTH_CLIENT_CERT 

433 and LSST_HTTP_AUTH_CLIENT_KEY. 

434 

435 - LSST_HTTP_AUTH_CLIENT_CERT: path to a .pem file which contains the 

436 client certificate for authenticating to the server. 

437 If initialized, the variable LSST_HTTP_AUTH_CLIENT_KEY must also be 

438 initialized with the path of the client private key file. 

439 The permissions of the client private key must be set so that only 

440 its owner can access it, at least for reading. 

441 """ 

442 root_uri = str(rpath.root_uri()) 

443 if root_uri not in self._sessions: 

444 # We don't have yet a session for this endpoint: create a new one. 

445 self._sessions[root_uri] = self._make_session(rpath) 

446 

447 return self._sessions[root_uri] 

448 

449 def _make_session(self, rpath: ResourcePath) -> requests.Session: 

450 """Make a new session configured from values from the environment.""" 

451 session = requests.Session() 

452 root_uri = str(rpath.root_uri()) 

453 log.debug("Creating new HTTP session for endpoint %s ...", root_uri) 

454 retries = Retry( 

455 # Total number of retries to allow. Takes precedence over other 

456 # counts. 

457 total=6, 

458 # How many connection-related errors to retry on. 

459 connect=3, 

460 # How many times to retry on read errors. 

461 read=3, 

462 # Backoff factor to apply between attempts after the second try 

463 # (seconds). Compute a random jitter to prevent all the clients 

464 # to overwhelm the server by sending requests at the same time. 

465 backoff_factor=self._backoff_min + (self._backoff_max - self._backoff_min) * random.random(), 

466 # How many times to retry on bad status codes. 

467 status=5, 

468 # Set of uppercased HTTP method verbs that we should retry on. 

469 # We only automatically retry idempotent requests. 

470 allowed_methods=frozenset( 

471 [ 

472 "COPY", 

473 "DELETE", 

474 "GET", 

475 "HEAD", 

476 "MKCOL", 

477 "OPTIONS", 

478 "PROPFIND", 

479 "PUT", 

480 ] 

481 ), 

482 # HTTP status codes that we should force a retry on. 

483 status_forcelist=frozenset( 

484 [ 

485 requests.codes.too_many_requests, # 429 

486 requests.codes.internal_server_error, # 500 

487 requests.codes.bad_gateway, # 502 

488 requests.codes.service_unavailable, # 503 

489 requests.codes.gateway_timeout, # 504 

490 ] 

491 ), 

492 # Whether to respect Retry-After header on status codes defined 

493 # above. 

494 respect_retry_after_header=True, 

495 ) 

496 

497 # Persist the specified number of connections to the front end server. 

498 session.mount( 

499 root_uri, 

500 HTTPAdapter( 

501 pool_connections=self._num_pools, 

502 pool_maxsize=self._max_persistent_connections, 

503 pool_block=False, 

504 max_retries=retries, 

505 ), 

506 ) 

507 

508 # Do not persist the connections to back end servers which may vary 

509 # from request to request. Systematically persisting connections to 

510 # those servers may exhaust their capabilities when there are thousands 

511 # of simultaneous clients. 

512 session.mount( 

513 f"{rpath.scheme}://", 

514 HTTPAdapter( 

515 pool_connections=self._num_pools, 

516 pool_maxsize=0, 

517 pool_block=False, 

518 max_retries=retries, 

519 ), 

520 ) 

521 

522 # If the remote endpoint doesn't use secure HTTP we don't include 

523 # bearer tokens in the requests nor need to authenticate the remote 

524 # server. 

525 if rpath.scheme != "https": 

526 return session 

527 

528 # Should we use a specific CA cert bundle for authenticating the 

529 # server? 

530 session.verify = True 

531 if ca_bundle := os.getenv("LSST_HTTP_CACERT_BUNDLE"): 

532 session.verify = ca_bundle 

533 

534 # Should we use bearer tokens for client authentication? 

535 if token := os.getenv("LSST_HTTP_AUTH_BEARER_TOKEN"): 

536 log.debug("... using bearer token authentication") 

537 session.auth = BearerTokenAuth(token) 

538 return session 

539 

540 # Should we instead use client certificate and private key? If so, both 

541 # LSST_HTTP_AUTH_CLIENT_CERT and LSST_HTTP_AUTH_CLIENT_KEY must be 

542 # initialized. 

543 client_cert = os.getenv("LSST_HTTP_AUTH_CLIENT_CERT") 

544 client_key = os.getenv("LSST_HTTP_AUTH_CLIENT_KEY") 

545 if client_cert and client_key: 

546 if not _is_protected(client_key): 

547 raise PermissionError( 

548 f"Private key file at {client_key} must be protected for access only by its owner" 

549 ) 

550 log.debug("... using client certificate authentication.") 

551 session.cert = (client_cert, client_key) 

552 return session 

553 

554 if client_cert: 

555 # Only the client certificate was provided. 

556 raise ValueError( 

557 "Environment variable LSST_HTTP_AUTH_CLIENT_KEY must be set to client private key file path" 

558 ) 

559 

560 if client_key: 

561 # Only the client private key was provided. 

562 raise ValueError( 

563 "Environment variable LSST_HTTP_AUTH_CLIENT_CERT must be set to client certificate file path" 

564 ) 

565 

566 log.debug( 

567 "Neither LSST_HTTP_AUTH_BEARER_TOKEN nor (LSST_HTTP_AUTH_CLIENT_CERT and " 

568 "LSST_HTTP_AUTH_CLIENT_KEY) are initialized. Client authentication is disabled." 

569 ) 

570 return session 

571 

572 

573class HttpResourcePath(ResourcePath): 

574 """General HTTP(S) resource. 

575 

576 Notes 

577 ----- 

578 In order to configure the behavior of instances of this class, the 

579 environment variables below are inspected: 

580 

581 - LSST_HTTP_PUT_SEND_EXPECT_HEADER: if set (with any value), a 

582 "Expect: 100-Continue" header will be added to all HTTP PUT requests. 

583 This header is required by some servers to detect if the client 

584 knows how to handle redirections. In case of redirection, the body 

585 of the PUT request is sent to the redirected location and not to 

586 the front end server. 

587 

588 - LSST_HTTP_TIMEOUT_CONNECT and LSST_HTTP_TIMEOUT_READ: if set to a 

589 numeric value, they are interpreted as the number of seconds to wait 

590 for establishing a connection with the server and for reading its 

591 response, respectively. 

592 

593 - LSST_HTTP_FRONTEND_PERSISTENT_CONNECTIONS and 

594 LSST_HTTP_BACKEND_PERSISTENT_CONNECTIONS: contain the maximum number 

595 of connections to attempt to persist with both the front end servers 

596 and the back end servers. 

597 Default values: DEFAULT_FRONTEND_PERSISTENT_CONNECTIONS and 

598 DEFAULT_BACKEND_PERSISTENT_CONNECTIONS. 

599 

600 - LSST_HTTP_DIGEST: case-insensitive name of the digest algorithm to 

601 ask the server to compute for every file's content sent to the server 

602 via a PUT request. No digest is requested if this variable is not set 

603 or is set to an invalid value. 

604 Valid values are those in ACCEPTED_DIGESTS. 

605 """ 

606 

607 _is_webdav: Optional[bool] = None 

608 

609 # Configuration items for this class instances. 

610 _config = HttpResourcePathConfig() 

611 

612 # The session for metadata requests is used for interacting with 

613 # the front end servers for requests such as PROPFIND, HEAD, etc. Those 

614 # interactions are typically served by the front end servers. We want to 

615 # keep the connection to the front end servers open, to reduce the cost 

616 # associated to TCP and TLS handshaking for each new request. 

617 _metadata_session_store = SessionStore( 

618 num_pools=5, 

619 max_persistent_connections=_config.front_end_connections, 

620 backoff_min=_config.backoff_min, 

621 backoff_max=_config.backoff_max, 

622 ) 

623 

624 # The data session is used for interaction with the front end servers which 

625 # typically redirect to the back end servers for serving our PUT and GET 

626 # requests. We attempt to keep a single connection open with the front end 

627 # server, if possible. This depends on how the server behaves and the 

628 # kind of request. Some servers close the connection when redirecting 

629 # the client to a back end server, for instance when serving a PUT 

630 # request. 

631 _data_session_store = SessionStore( 

632 num_pools=25, 

633 max_persistent_connections=_config.back_end_connections, 

634 backoff_min=_config.backoff_min, 

635 backoff_max=_config.backoff_max, 

636 ) 

637 

638 # Process ID which created the sessions above. We need to store this 

639 # to replace sessions created by a parent process and inherited by a 

640 # child process after a fork, to avoid confusing the SSL layer. 

641 _pid: int = -1 

642 

643 @property 

644 def metadata_session(self) -> requests.Session: 

645 """Client session to send requests which do not require upload or 

646 download of data, i.e. mostly metadata requests. 

647 """ 

648 

649 if hasattr(self, "_metadata_session") and self._pid == os.getpid(): 

650 return self._metadata_session 

651 

652 # Reset the store in case it was created by another process and 

653 # retrieve a session. 

654 self._metadata_session_store.clear() 

655 self._pid = os.getpid() 

656 self._metadata_session: requests.Session = self._metadata_session_store.get(self) 

657 return self._metadata_session 

658 

659 @property 

660 def data_session(self) -> requests.Session: 

661 """Client session for uploading and downloading data.""" 

662 

663 if hasattr(self, "_data_session") and self._pid == os.getpid(): 

664 return self._data_session 

665 

666 # Reset the store in case it was created by another process and 

667 # retrieve a session. 

668 self._data_session_store.clear() 

669 self._pid = os.getpid() 

670 self._data_session: requests.Session = self._data_session_store.get(self) 

671 return self._data_session 

672 

673 def _clear_sessions(self) -> None: 

674 """Internal method to close the socket connections still open. Used 

675 only in test suites to avoid warnings. 

676 """ 

677 self._metadata_session_store.clear() 

678 self._data_session_store.clear() 

679 

680 if hasattr(self, "_metadata_session"): 

681 delattr(self, "_metadata_session") 

682 

683 if hasattr(self, "_data_session"): 

684 delattr(self, "_data_session") 

685 

686 @property 

687 def is_webdav_endpoint(self) -> bool: 

688 """Check if the current endpoint implements WebDAV features. 

689 

690 This is stored per URI but cached by root so there is 

691 only one check per hostname. 

692 """ 

693 if self._is_webdav is not None: 

694 return self._is_webdav 

695 

696 self._is_webdav = _is_webdav_endpoint(self.root_uri()) 

697 return self._is_webdav 

698 

699 def exists(self) -> bool: 

700 """Check that a remote HTTP resource exists.""" 

701 log.debug("Checking if resource exists: %s", self.geturl()) 

702 if not self.is_webdav_endpoint: 

703 # The remote is a plain HTTP server. Let's attempt a HEAD 

704 # request, even if the behavior for such a request against a 

705 # directory is not specified, so it depends on the server 

706 # implementation. 

707 resp = self.metadata_session.head( 

708 self.geturl(), timeout=self._config.timeout, allow_redirects=True, stream=False 

709 ) 

710 return resp.status_code == requests.codes.ok # 200 

711 

712 # The remote endpoint is a webDAV server: send a PROPFIND request 

713 # to determine if it exists. 

714 resp = self._propfind() 

715 if resp.status_code == requests.codes.multi_status: # 207 

716 prop = _parse_propfind_response_body(resp.text)[0] 

717 return prop.exists 

718 else: # 404 Not Found 

719 return False 

720 

721 def size(self) -> int: 

722 """Return the size of the remote resource in bytes.""" 

723 if self.dirLike: 

724 return 0 

725 

726 if not self.is_webdav_endpoint: 

727 # The remote is a plain HTTP server. Send a HEAD request to 

728 # retrieve the size of the resource. 

729 resp = self.metadata_session.head( 

730 self.geturl(), timeout=self._config.timeout, allow_redirects=True, stream=False 

731 ) 

732 if resp.status_code == requests.codes.ok: # 200 

733 if "Content-Length" in resp.headers: 

734 return int(resp.headers["Content-Length"]) 

735 else: 

736 raise ValueError( 

737 f"Response to HEAD request to {self} does not contain 'Content-Length' header" 

738 ) 

739 elif resp.status_code == requests.codes.not_found: 

740 raise FileNotFoundError( 

741 f"Resource {self} does not exist, status: {resp.status_code} {resp.reason}" 

742 ) 

743 else: 

744 raise ValueError( 

745 f"Unexpected response for HEAD request for {self}, status: {resp.status_code} " 

746 f"{resp.reason}" 

747 ) 

748 

749 # The remote is a webDAV server: send a PROPFIND request to retrieve 

750 # the size of the resource. Sizes are only meaningful for files. 

751 resp = self._propfind() 

752 if resp.status_code == requests.codes.multi_status: # 207 

753 prop = _parse_propfind_response_body(resp.text)[0] 

754 if prop.is_file: 

755 return prop.size 

756 elif prop.is_directory: 

757 raise IsADirectoryError( 

758 f"Resource {self} is reported by server as a directory but has a file path" 

759 ) 

760 else: 

761 raise FileNotFoundError(f"Resource {self} does not exist") 

762 else: # 404 Not Found 

763 raise FileNotFoundError( 

764 f"Resource {self} does not exist, status: {resp.status_code} {resp.reason}" 

765 ) 

766 

767 def mkdir(self) -> None: 

768 """Create the directory resource if it does not already exist.""" 

769 # Creating directories is only available on WebDAV back ends. 

770 if not self.is_webdav_endpoint: 

771 raise NotImplementedError( 

772 f"Creation of directory {self} is not implemented by plain HTTP servers" 

773 ) 

774 

775 if not self.dirLike: 

776 raise NotADirectoryError(f"Can not create a 'directory' for file-like URI {self}") 

777 

778 # Check if the target directory already exists. 

779 resp = self._propfind() 

780 if resp.status_code == requests.codes.multi_status: # 207 

781 prop = _parse_propfind_response_body(resp.text)[0] 

782 if prop.exists: 

783 if prop.is_directory: 

784 return 

785 else: 

786 # A file exists at this path 

787 raise NotADirectoryError( 

788 f"Can not create a directory for {self} because a file already exists at that path" 

789 ) 

790 

791 # Target directory does not exist. Create it and its ancestors as 

792 # needed. We need to test if parent URL is different from self URL, 

793 # otherwise we could be stuck in a recursive loop 

794 # where self == parent. 

795 if self.geturl() != self.parent().geturl(): 

796 self.parent().mkdir() 

797 

798 log.debug("Creating new directory: %s", self.geturl()) 

799 self._mkcol() 

800 

801 def remove(self) -> None: 

802 """Remove the resource.""" 

803 self._delete() 

804 

805 def read(self, size: int = -1) -> bytes: 

806 """Open the resource and return the contents in bytes. 

807 

808 Parameters 

809 ---------- 

810 size : `int`, optional 

811 The number of bytes to read. Negative or omitted indicates 

812 that all data should be read. 

813 """ 

814 

815 # Use the data session as a context manager to ensure that the 

816 # network connections to both the front end and back end servers are 

817 # closed after downloading the data. 

818 log.debug("Reading from remote resource: %s", self.geturl()) 

819 stream = True if size > 0 else False 

820 with self.data_session as session: 

821 with time_this(log, msg="GET %s", args=(self,)): 

822 resp = session.get(self.geturl(), stream=stream, timeout=self._config.timeout) 

823 

824 if resp.status_code != requests.codes.ok: # 200 

825 raise FileNotFoundError( 

826 f"Unable to read resource {self}; status: {resp.status_code} {resp.reason}" 

827 ) 

828 if not stream: 

829 return resp.content 

830 else: 

831 return next(resp.iter_content(chunk_size=size)) 

832 

833 def write(self, data: bytes, overwrite: bool = True) -> None: 

834 """Write the supplied bytes to the new resource. 

835 

836 Parameters 

837 ---------- 

838 data : `bytes` 

839 The bytes to write to the resource. The entire contents of the 

840 resource will be replaced. 

841 overwrite : `bool`, optional 

842 If `True` the resource will be overwritten if it exists. Otherwise 

843 the write will fail. 

844 """ 

845 log.debug("Writing to remote resource: %s", self.geturl()) 

846 if not overwrite: 

847 if self.exists(): 

848 raise FileExistsError(f"Remote resource {self} exists and overwrite has been disabled") 

849 

850 # Ensure the parent directory exists. 

851 self.parent().mkdir() 

852 

853 # Upload the data. 

854 log.debug("Writing data to remote resource: %s", self.geturl()) 

855 self._put(data=data) 

856 

857 def transfer_from( 

858 self, 

859 src: ResourcePath, 

860 transfer: str = "copy", 

861 overwrite: bool = False, 

862 transaction: Optional[TransactionProtocol] = None, 

863 ) -> None: 

864 """Transfer the current resource to a Webdav repository. 

865 

866 Parameters 

867 ---------- 

868 src : `ResourcePath` 

869 Source URI. 

870 transfer : `str` 

871 Mode to use for transferring the resource. Supports the following 

872 options: copy. 

873 transaction : `~lsst.resources.utils.TransactionProtocol`, optional 

874 Currently unused. 

875 """ 

876 # Fail early to prevent delays if remote resources are requested. 

877 if transfer not in self.transferModes: 

878 raise ValueError(f"Transfer mode {transfer} not supported by URI scheme {self.scheme}") 

879 

880 # Existence checks cost time so do not call this unless we know 

881 # that debugging is enabled. 

882 if log.isEnabledFor(logging.DEBUG): 

883 log.debug( 

884 "Transferring %s [exists: %s] -> %s [exists: %s] (transfer=%s)", 

885 src, 

886 src.exists(), 

887 self, 

888 self.exists(), 

889 transfer, 

890 ) 

891 

892 # Short circuit immediately if the URIs are identical. 

893 if self == src: 

894 log.debug( 

895 "Target and destination URIs are identical: %s, returning immediately." 

896 " No further action required.", 

897 self, 

898 ) 

899 return 

900 

901 if not overwrite and self.exists(): 

902 raise FileExistsError(f"Destination path {self} already exists.") 

903 

904 if transfer == "auto": 

905 transfer = self.transferDefault 

906 

907 # We can use webDAV 'COPY' or 'MOVE' if both the current and source 

908 # resources are located in the same server. 

909 if isinstance(src, type(self)) and self.root_uri() == src.root_uri() and self.is_webdav_endpoint: 

910 log.debug("Transfer from %s to %s directly", src, self) 

911 return self._move(src) if transfer == "move" else self._copy(src) 

912 

913 # For resources of different classes or for plain HTTP resources we can 

914 # perform the copy or move operation by downloading to a local file 

915 # and uploading to the destination. 

916 self._copy_via_local(src) 

917 

918 # This was an explicit move, try to remove the source. 

919 if transfer == "move": 

920 src.remove() 

921 

922 def walk( 

923 self, file_filter: Optional[Union[str, re.Pattern]] = None 

924 ) -> Iterator[Union[List, Tuple[ResourcePath, List[str], List[str]]]]: 

925 """Walk the directory tree returning matching files and directories. 

926 Parameters 

927 ---------- 

928 file_filter : `str` or `re.Pattern`, optional 

929 Regex to filter out files from the list before it is returned. 

930 Yields 

931 ------ 

932 dirpath : `ResourcePath` 

933 Current directory being examined. 

934 dirnames : `list` of `str` 

935 Names of subdirectories within dirpath. 

936 filenames : `list` of `str` 

937 Names of all the files within dirpath. 

938 """ 

939 if not self.dirLike: 

940 raise ValueError("Can not walk a non-directory URI") 

941 

942 # Walking directories is only available on WebDAV back ends. 

943 if not self.is_webdav_endpoint: 

944 raise NotImplementedError(f"Walking directory {self} is not implemented by plain HTTP servers") 

945 

946 if isinstance(file_filter, str): 

947 file_filter = re.compile(file_filter) 

948 

949 resp = self._propfind(depth="1") 

950 if resp.status_code == requests.codes.multi_status: # 207 

951 files: List[str] = [] 

952 dirs: List[str] = [] 

953 

954 for prop in _parse_propfind_response_body(resp.text): 

955 if prop.is_file: 

956 files.append(prop.name) 

957 elif not prop.href.rstrip("/").endswith(self.path.rstrip("/")): 

958 # Only include the names of sub-directories not the name of 

959 # the directory being walked. 

960 dirs.append(prop.name) 

961 

962 if file_filter is not None: 

963 files = [f for f in files if file_filter.search(f)] 

964 

965 if not dirs and not files: 

966 return 

967 else: 

968 yield type(self)(self, forceAbsolute=False, forceDirectory=True), dirs, files 

969 

970 for dir in dirs: 

971 new_uri = self.join(dir, forceDirectory=True) 

972 yield from new_uri.walk(file_filter) 

973 

974 def _as_local(self) -> Tuple[str, bool]: 

975 """Download object over HTTP and place in temporary directory. 

976 

977 Returns 

978 ------- 

979 path : `str` 

980 Path to local temporary file. 

981 temporary : `bool` 

982 Always returns `True`. This is always a temporary file. 

983 """ 

984 

985 # Use the session as a context manager to ensure that connections 

986 # to both the front end and back end servers are closed after the 

987 # download operation is finished. 

988 with self.data_session as session: 

989 resp = session.get(self.geturl(), stream=True, timeout=self._config.timeout) 

990 if resp.status_code != requests.codes.ok: 

991 raise FileNotFoundError( 

992 f"Unable to download resource {self}; status: {resp.status_code} {resp.reason}" 

993 ) 

994 

995 tmpdir, buffering = _get_temp_dir() 

996 with tempfile.NamedTemporaryFile( 

997 suffix=self.getExtension(), buffering=buffering, dir=tmpdir, delete=False 

998 ) as tmpFile: 

999 expected_length = int(resp.headers.get("Content-Length", "-1")) 

1000 with time_this( 

1001 log, 

1002 msg="GET %s [length=%d] to local file %s [chunk_size=%d]", 

1003 args=(self, expected_length, tmpFile.name, buffering), 

1004 mem_usage=self._config.collect_memory_usage, 

1005 mem_unit=u.mebibyte, 

1006 ): 

1007 content_length = 0 

1008 for chunk in resp.iter_content(chunk_size=buffering): 

1009 tmpFile.write(chunk) 

1010 content_length += len(chunk) 

1011 

1012 # Check that the expected and actual content lengths match. Perform 

1013 # this check only when the contents of the file was not encoded by 

1014 # the server. 

1015 if "Content-Encoding" not in resp.headers: 

1016 if expected_length >= 0 and expected_length != content_length: 

1017 raise ValueError( 

1018 f"Size of downloaded file does not match value in Content-Length header for {self}: " 

1019 f"expecting {expected_length} and got {content_length} bytes" 

1020 ) 

1021 

1022 return tmpFile.name, True 

1023 

1024 def _send_webdav_request( 

1025 self, 

1026 method: str, 

1027 url: Optional[str] = None, 

1028 headers: dict[str, str] = {}, 

1029 body: Optional[str] = None, 

1030 session: Optional[requests.Session] = None, 

1031 timeout: Optional[tuple[float, float]] = None, 

1032 ) -> requests.Response: 

1033 """Send a webDAV request and correctly handle redirects. 

1034 

1035 Parameters 

1036 ---------- 

1037 method : `str` 

1038 The mthod of the HTTP request to be sent, e.g. PROPFIND, MKCOL. 

1039 headers : `dict`, optional 

1040 A dictionary of key-value pairs (both strings) to include as 

1041 headers in the request. 

1042 body: `str`, optional 

1043 The body of the request. 

1044 

1045 Notes 

1046 ----- 

1047 This way of sending webDAV requests is necessary for handling 

1048 redirection ourselves, since the 'requests' package changes the method 

1049 of the redirected request when the server responds with status 302 and 

1050 the method of the original request is not HEAD (which is the case for 

1051 webDAV requests). 

1052 

1053 That means that when the webDAV server we interact with responds with 

1054 a redirection to a PROPFIND or MKCOL request, the request gets 

1055 converted to a GET request when sent to the redirected location. 

1056 

1057 See `requests.sessions.SessionRedirectMixin.rebuild_method()` in 

1058 https://github.com/psf/requests/blob/main/requests/sessions.py 

1059 

1060 This behavior of the 'requests' package is meant to be compatible with 

1061 what is specified in RFC 9110: 

1062 

1063 https://www.rfc-editor.org/rfc/rfc9110#name-302-found 

1064 

1065 For our purposes, we do need to follow the redirection and send a new 

1066 request using the same HTTP verb. 

1067 """ 

1068 if url is None: 

1069 url = self.geturl() 

1070 

1071 if session is None: 

1072 session = self.metadata_session 

1073 

1074 if timeout is None: 

1075 timeout = self._config.timeout 

1076 

1077 with time_this( 

1078 log, 

1079 msg="%s %s", 

1080 args=( 

1081 method, 

1082 url, 

1083 ), 

1084 mem_usage=self._config.collect_memory_usage, 

1085 mem_unit=u.mebibyte, 

1086 ): 

1087 for _ in range(max_redirects := 5): 

1088 resp = session.request( 

1089 method, 

1090 url, 

1091 data=body, 

1092 headers=headers, 

1093 stream=False, 

1094 timeout=timeout, 

1095 allow_redirects=False, 

1096 ) 

1097 if resp.is_redirect: 

1098 url = resp.headers["Location"] 

1099 else: 

1100 return resp 

1101 

1102 # We reached the maximum allowed number of redirects. 

1103 # Stop trying. 

1104 raise ValueError( 

1105 f"Could not get a response to {method} request for {self} after " 

1106 f"{max_redirects} redirections" 

1107 ) 

1108 

1109 def _propfind(self, body: Optional[str] = None, depth: str = "0") -> requests.Response: 

1110 """Send a PROPFIND webDAV request and return the response. 

1111 

1112 Parameters 

1113 ---------- 

1114 body : `str`, optional 

1115 The body of the PROPFIND request to send to the server. If 

1116 provided, it is expected to be a XML document. 

1117 depth : `str`, optional 

1118 The value of the 'Depth' header to include in the request. 

1119 

1120 Returns 

1121 ------- 

1122 response : `requests.Response` 

1123 Response to the PROPFIND request. 

1124 

1125 Notes 

1126 ----- 

1127 It raises `ValueError` if the status code of the PROPFIND request 

1128 is different from "207 Multistatus" or "404 Not Found". 

1129 """ 

1130 if body is None: 

1131 # Request only the DAV live properties we are explicitly interested 

1132 # in namely 'resourcetype', 'getcontentlength', 'getlastmodified' 

1133 # and 'displayname'. 

1134 body = ( 

1135 """<?xml version="1.0" encoding="utf-8" ?>""" 

1136 """<D:propfind xmlns:D="DAV:"><D:prop>""" 

1137 """<D:resourcetype/><D:getcontentlength/><D:getlastmodified/><D:displayname/>""" 

1138 """</D:prop></D:propfind>""" 

1139 ) 

1140 headers = { 

1141 "Depth": depth, 

1142 "Content-Type": 'application/xml; charset="utf-8"', 

1143 "Content-Length": str(len(body)), 

1144 } 

1145 resp = self._send_webdav_request("PROPFIND", headers=headers, body=body) 

1146 if resp.status_code in (requests.codes.multi_status, requests.codes.not_found): 

1147 return resp 

1148 else: 

1149 raise ValueError( 

1150 f"Unexpected response for PROPFIND request for {self}, status: {resp.status_code} " 

1151 f"{resp.reason}" 

1152 ) 

1153 

1154 def _options(self) -> requests.Response: 

1155 """Send a OPTIONS webDAV request for this resource.""" 

1156 resp = self._send_webdav_request("OPTIONS") 

1157 if resp.status_code in (requests.codes.ok, requests.codes.created): 

1158 return resp 

1159 

1160 raise ValueError( 

1161 f"Unexpected response to OPTIONS request for {self}, status: {resp.status_code} " f"{resp.reason}" 

1162 ) 

1163 

1164 def _head(self) -> requests.Response: 

1165 """Send a HEAD webDAV request for this resource.""" 

1166 

1167 return self._send_webdav_request("HEAD") 

1168 

1169 def _mkcol(self) -> None: 

1170 """Send a MKCOL webDAV request to create a collection. The collection 

1171 may already exist. 

1172 """ 

1173 resp = self._send_webdav_request("MKCOL") 

1174 if resp.status_code == requests.codes.created: # 201 

1175 return 

1176 

1177 if resp.status_code == requests.codes.method_not_allowed: # 405 

1178 # The remote directory already exists 

1179 log.debug("Can not create directory: %s may already exist: skipping.", self.geturl()) 

1180 else: 

1181 raise ValueError(f"Can not create directory {self}, status: {resp.status_code} {resp.reason}") 

1182 

1183 def _delete(self) -> None: 

1184 """Send a DELETE webDAV request for this resource.""" 

1185 

1186 log.debug("Deleting %s ...", self.geturl()) 

1187 

1188 # If this is a directory, ensure the remote is a webDAV server because 

1189 # plain HTTP servers don't support DELETE requests on non-file 

1190 # paths. 

1191 if self.dirLike and not self.is_webdav_endpoint: 

1192 raise NotImplementedError( 

1193 f"Deletion of directory {self} is not implemented by plain HTTP servers" 

1194 ) 

1195 

1196 # Deleting non-empty directories may take some time, so increase 

1197 # the timeout for getting a response from the server. 

1198 timeout = self._config.timeout 

1199 if self.dirLike: 

1200 timeout = (timeout[0], timeout[1] * 100) 

1201 resp = self._send_webdav_request("DELETE", timeout=timeout) 

1202 if resp.status_code in ( 

1203 requests.codes.ok, 

1204 requests.codes.accepted, 

1205 requests.codes.no_content, 

1206 requests.codes.not_found, 

1207 ): 

1208 # We can get a "404 Not Found" error when the file or directory 

1209 # does not exist or when the DELETE request was retried several 

1210 # times and a previous attempt actually deleted the resource. 

1211 # Therefore we consider that a "Not Found" response is not an 

1212 # error since we reached the state desired by the user. 

1213 return 

1214 else: 

1215 # TODO: the response to a DELETE request against a webDAV server 

1216 # may be multistatus. If so, we need to parse the reponse body to 

1217 # determine more precisely the reason of the failure (e.g. a lock) 

1218 # and provide a more helpful error message. 

1219 raise ValueError(f"Unable to delete resource {self}; status: {resp.status_code} {resp.reason}") 

1220 

1221 def _copy_via_local(self, src: ResourcePath) -> None: 

1222 """Replace the contents of this resource with the contents of a remote 

1223 resource by using a local temporary file. 

1224 

1225 Parameters 

1226 ---------- 

1227 src : `HttpResourcePath` 

1228 The source of the contents to copy to `self`. 

1229 """ 

1230 with src.as_local() as local_uri: 

1231 log.debug("Transfer from %s to %s via local file %s", src, self, local_uri) 

1232 with open(local_uri.ospath, "rb") as f: 

1233 self._put(data=f) 

1234 

1235 def _copy_or_move(self, method: str, src: HttpResourcePath) -> None: 

1236 """Send a COPY or MOVE webDAV request to copy or replace the contents 

1237 of this resource with the contents of another resource located in the 

1238 same server. 

1239 

1240 Parameters 

1241 ---------- 

1242 method : `str` 

1243 The method to perform. Valid values are "COPY" or "MOVE" (in 

1244 uppercase). 

1245 

1246 src : `HttpResourcePath` 

1247 The source of the contents to move to `self`. 

1248 """ 

1249 headers = {"Destination": self.geturl()} 

1250 resp = self._send_webdav_request(method, url=src.geturl(), headers=headers, session=self.data_session) 

1251 if resp.status_code in (requests.codes.created, requests.codes.no_content): 

1252 return 

1253 

1254 if resp.status_code == requests.codes.multi_status: 

1255 tree = eTree.fromstring(resp.content) 

1256 status_element = tree.find("./{DAV:}response/{DAV:}status") 

1257 status = status_element.text if status_element is not None else "unknown" 

1258 error = tree.find("./{DAV:}response/{DAV:}error") 

1259 raise ValueError(f"{method} returned multistatus reponse with status {status} and error {error}") 

1260 else: 

1261 raise ValueError( 

1262 f"{method} operation from {src} to {self} failed, status: {resp.status_code} {resp.reason}" 

1263 ) 

1264 

1265 def _copy(self, src: HttpResourcePath) -> None: 

1266 """Send a COPY webDAV request to replace the contents of this resource 

1267 (if any) with the contents of another resource located in the same 

1268 server. 

1269 

1270 Parameters 

1271 ---------- 

1272 src : `HttpResourcePath` 

1273 The source of the contents to copy to `self`. 

1274 """ 

1275 # Neither dCache nor XrootD currently implement the COPY 

1276 # webDAV method as documented in 

1277 # http://www.webdav.org/specs/rfc4918.html#METHOD_COPY 

1278 # (See issues DM-37603 and DM-37651 for details) 

1279 # 

1280 # For the time being, we use a temporary local file to 

1281 # perform the copy client side. 

1282 # TODO: when those 2 issues above are solved remove the 3 lines below. 

1283 must_use_local = True 

1284 if must_use_local: 

1285 return self._copy_via_local(src) 

1286 

1287 return self._copy_or_move("COPY", src) 

1288 

1289 def _move(self, src: HttpResourcePath) -> None: 

1290 """Send a MOVE webDAV request to replace the contents of this resource 

1291 with the contents of another resource located in the same server. 

1292 

1293 Parameters 

1294 ---------- 

1295 src : `HttpResourcePath` 

1296 The source of the contents to move to `self`. 

1297 """ 

1298 return self._copy_or_move("MOVE", src) 

1299 

1300 def _put(self, data: Union[BinaryIO, bytes]) -> None: 

1301 """Perform an HTTP PUT request and handle redirection. 

1302 

1303 Parameters 

1304 ---------- 

1305 data : `Union[BinaryIO, bytes]` 

1306 The data to be included in the body of the PUT request. 

1307 """ 

1308 # Retrieve the final URL for this upload by sending a PUT request with 

1309 # no content. Follow a single server redirection to retrieve the 

1310 # final URL. 

1311 headers = {"Content-Length": "0"} 

1312 if self._config.send_expect_on_put: 

1313 headers["Expect"] = "100-continue" 

1314 

1315 url = self.geturl() 

1316 

1317 # Use the session as a context manager to ensure the underlying 

1318 # connections are closed after finishing uploading the data. 

1319 with self.data_session as session: 

1320 # Send an empty PUT request to get redirected to the final 

1321 # destination. 

1322 log.debug("Sending empty PUT request to %s", url) 

1323 with time_this( 

1324 log, 

1325 msg="PUT (no data) %s", 

1326 args=(url,), 

1327 mem_usage=self._config.collect_memory_usage, 

1328 mem_unit=u.mebibyte, 

1329 ): 

1330 resp = session.request( 

1331 "PUT", 

1332 url, 

1333 data=None, 

1334 headers=headers, 

1335 stream=False, 

1336 timeout=self._config.timeout, 

1337 allow_redirects=False, 

1338 ) 

1339 if resp.is_redirect: 

1340 url = resp.headers["Location"] 

1341 

1342 # Upload the data to the final destination. 

1343 log.debug("Uploading data to %s", url) 

1344 

1345 # Ask the server to compute and record a checksum of the uploaded 

1346 # file contents, for later integrity checks. Since we don't compute 

1347 # the digest ourselves while uploading the data, we cannot control 

1348 # after the request is complete that the data we uploaded is 

1349 # identical to the data recorded by the server, but at least the 

1350 # server has recorded a digest of the data it stored. 

1351 # 

1352 # See RFC-3230 for details and 

1353 # https://www.iana.org/assignments/http-dig-alg/http-dig-alg.xhtml 

1354 # for the list of supported digest algorithhms. 

1355 # In addition, note that not all servers implement this RFC so 

1356 # the checksum may not be computed by the server. 

1357 put_headers: Optional[dict[str, str]] = None 

1358 if digest := self._config.digest_algorithm: 

1359 put_headers = {"Want-Digest": digest} 

1360 

1361 with time_this( 

1362 log, 

1363 msg="PUT %s", 

1364 args=(url,), 

1365 mem_usage=self._config.collect_memory_usage, 

1366 mem_unit=u.mebibyte, 

1367 ): 

1368 resp = session.request( 

1369 "PUT", 

1370 url, 

1371 data=data, 

1372 headers=put_headers, 

1373 stream=False, 

1374 timeout=self._config.timeout, 

1375 allow_redirects=False, 

1376 ) 

1377 if resp.status_code in ( 

1378 requests.codes.ok, 

1379 requests.codes.created, 

1380 requests.codes.no_content, 

1381 ): 

1382 return 

1383 else: 

1384 raise ValueError(f"Can not write file {self}, status: {resp.status_code} {resp.reason}") 

1385 

1386 @contextlib.contextmanager 

1387 def _openImpl( 

1388 self, 

1389 mode: str = "r", 

1390 *, 

1391 encoding: Optional[str] = None, 

1392 ) -> Iterator[ResourceHandleProtocol]: 

1393 resp = self._head() 

1394 accepts_range = resp.status_code == requests.codes.ok and resp.headers.get("Accept-Ranges") == "bytes" 

1395 handle: ResourceHandleProtocol 

1396 if mode in ("rb", "r") and accepts_range: 

1397 handle = HttpReadResourceHandle( 

1398 mode, log, url=self.geturl(), session=self.data_session, timeout=self._config.timeout 

1399 ) 

1400 if mode == "r": 

1401 # cast because the protocol is compatible, but does not have 

1402 # BytesIO in the inheritance tree 

1403 yield io.TextIOWrapper(cast(io.BytesIO, handle), encoding=encoding) 

1404 else: 

1405 yield handle 

1406 else: 

1407 with super()._openImpl(mode, encoding=encoding) as http_handle: 

1408 yield http_handle 

1409 

1410 

1411def _dump_response(resp: requests.Response) -> None: 

1412 """Log the contents of a HTTP or webDAV request and its response. 

1413 

1414 Parameters 

1415 ---------- 

1416 resp : `requests.Response` 

1417 The response to log. 

1418 

1419 Notes 

1420 ----- 

1421 Intended for development purposes only. 

1422 """ 

1423 log.debug("-----------------------------------------------") 

1424 log.debug("Request") 

1425 log.debug(" method=%s", resp.request.method) 

1426 log.debug(" URL=%s", resp.request.url) 

1427 log.debug(" headers=%s", resp.request.headers) 

1428 if resp.request.method == "PUT": 

1429 log.debug(" body=<data>") 

1430 elif resp.request.body is None: 

1431 log.debug(" body=<empty>") 

1432 else: 

1433 log.debug(" body=%r", resp.request.body[:120]) 

1434 

1435 log.debug("Response:") 

1436 log.debug(" status_code=%d", resp.status_code) 

1437 log.debug(" headers=%s", resp.headers) 

1438 if not resp.content: 

1439 log.debug(" body=<empty>") 

1440 elif "Content-Type" in resp.headers and resp.headers["Content-Type"] == "text/plain": 

1441 log.debug(" body=%r", resp.content) 

1442 else: 

1443 log.debug(" body=%r", resp.content[:80]) 

1444 

1445 

1446def _is_protected(filepath: str) -> bool: 

1447 """Return true if the permissions of file at filepath only allow for access 

1448 by its owner. 

1449 

1450 Parameters 

1451 ---------- 

1452 filepath : `str` 

1453 Path of a local file. 

1454 """ 

1455 if not os.path.isfile(filepath): 

1456 return False 

1457 mode = stat.S_IMODE(os.stat(filepath).st_mode) 

1458 owner_accessible = bool(mode & stat.S_IRWXU) 

1459 group_accessible = bool(mode & stat.S_IRWXG) 

1460 other_accessible = bool(mode & stat.S_IRWXO) 

1461 return owner_accessible and not group_accessible and not other_accessible 

1462 

1463 

1464def _parse_propfind_response_body(body: str) -> List[DavProperty]: 

1465 """Parse the XML-encoded contents of the response body to a webDAV PROPFIND 

1466 request. 

1467 

1468 Parameters 

1469 ---------- 

1470 body : `str` 

1471 XML-encoded response body to a PROPFIND request 

1472 

1473 Returns 

1474 ------- 

1475 responses : `List[DavProperty]` 

1476 

1477 Notes 

1478 ----- 

1479 Is is expected that there is at least one reponse in `body`, otherwise 

1480 this function raises. 

1481 """ 

1482 # A response body to a PROPFIND request is of the form (indented for 

1483 # readability): 

1484 # 

1485 # <?xml version="1.0" encoding="UTF-8"?> 

1486 # <D:multistatus xmlns:D="DAV:"> 

1487 # <D:response> 

1488 # <D:href>path/to/resource</D:href> 

1489 # <D:propstat> 

1490 # <D:prop> 

1491 # <D:resourcetype> 

1492 # <D:collection xmlns:D="DAV:"/> 

1493 # </D:resourcetype> 

1494 # <D:getlastmodified> 

1495 # Fri, 27 Jan 2 023 13:59:01 GMT 

1496 # </D:getlastmodified> 

1497 # <D:getcontentlength> 

1498 # 12345 

1499 # </D:getcontentlength> 

1500 # </D:prop> 

1501 # <D:status> 

1502 # HTTP/1.1 200 OK 

1503 # </D:status> 

1504 # </D:propstat> 

1505 # </D:response> 

1506 # <D:response> 

1507 # ... 

1508 # </D:response> 

1509 # <D:response> 

1510 # ... 

1511 # </D:response> 

1512 # </D:multistatus> 

1513 

1514 # Scan all the 'response' elements and extract the relevant properties 

1515 responses = [] 

1516 multistatus = eTree.fromstring(body.strip()) 

1517 for response in multistatus.findall("./{DAV:}response"): 

1518 responses.append(DavProperty(response)) 

1519 

1520 if responses: 

1521 return responses 

1522 else: 

1523 # Could not parse the body 

1524 raise ValueError(f"Unable to parse response for PROPFIND request: {body}") 

1525 

1526 

1527class DavProperty: 

1528 """Helper class to encapsulate select live DAV properties of a single 

1529 resource, as retrieved via a PROPFIND request. 

1530 """ 

1531 

1532 # Regular expression to compare against the 'status' element of a 

1533 # PROPFIND response's 'propstat' element. 

1534 _status_ok_rex = re.compile(r"^HTTP/.* 200 .*$", re.IGNORECASE) 

1535 

1536 def __init__(self, response: Optional[eTree.Element]): 

1537 self._href: str = "" 

1538 self._displayname: str = "" 

1539 self._collection: bool = False 

1540 self._getlastmodified: str = "" 

1541 self._getcontentlength: int = -1 

1542 

1543 if response is not None: 

1544 self._parse(response) 

1545 

1546 def _parse(self, response: eTree.Element) -> None: 

1547 # Extract 'href'. 

1548 if (element := response.find("./{DAV:}href")) is not None: 

1549 # We need to use "str(element.text)"" instead of "element.text" to 

1550 # keep mypy happy. 

1551 self._href = str(element.text).strip() 

1552 else: 

1553 raise ValueError( 

1554 f"Property 'href' expected but not found in PROPFIND response: " 

1555 f"{eTree.tostring(response, encoding='unicode')}" 

1556 ) 

1557 

1558 for propstat in response.findall("./{DAV:}propstat"): 

1559 # Only extract properties of interest with status OK. 

1560 status = propstat.find("./{DAV:}status") 

1561 if status is None or not self._status_ok_rex.match(str(status.text)): 

1562 continue 

1563 

1564 for prop in propstat.findall("./{DAV:}prop"): 

1565 # Parse "collection". 

1566 if (element := prop.find("./{DAV:}resourcetype/{DAV:}collection")) is not None: 

1567 self._collection = True 

1568 

1569 # Parse "getlastmodified". 

1570 if (element := prop.find("./{DAV:}getlastmodified")) is not None: 

1571 self._getlastmodified = str(element.text) 

1572 

1573 # Parse "getcontentlength". 

1574 if (element := prop.find("./{DAV:}getcontentlength")) is not None: 

1575 self._getcontentlength = int(str(element.text)) 

1576 

1577 # Parse "displayname". 

1578 if (element := prop.find("./{DAV:}displayname")) is not None: 

1579 self._displayname = str(element.text) 

1580 

1581 # Some webDAV servers don't include the 'displayname' property in the 

1582 # response so try to infer it from the value of the 'href' property. 

1583 # Depending on the server the href value may end with '/'. 

1584 if not self._displayname: 

1585 self._displayname = os.path.basename(self._href.rstrip("/")) 

1586 

1587 # Force a size of 0 for collections. 

1588 if self._collection: 

1589 self._getcontentlength = 0 

1590 

1591 @property 

1592 def exists(self) -> bool: 

1593 # It is either a directory or a file with length of at least zero 

1594 return self._collection or self._getcontentlength >= 0 

1595 

1596 @property 

1597 def is_directory(self) -> bool: 

1598 return self._collection 

1599 

1600 @property 

1601 def is_file(self) -> bool: 

1602 return not self._collection 

1603 

1604 @property 

1605 def size(self) -> int: 

1606 return self._getcontentlength 

1607 

1608 @property 

1609 def name(self) -> str: 

1610 return self._displayname 

1611 

1612 @property 

1613 def href(self) -> str: 

1614 return self._href