Coverage for python/lsst/resources/_resourcePath.py: 21%
Shortcuts on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
Shortcuts on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
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.
12from __future__ import annotations
14import concurrent.futures
15import contextlib
16import copy
17import io
18import locale
19import logging
20import os
21import posixpath
22import re
23import shutil
24import tempfile
25import urllib.parse
26from pathlib import Path, PurePath, PurePosixPath
27from random import Random
29__all__ = ("ResourcePath", "ResourcePathExpression")
31from typing import (
32 IO,
33 TYPE_CHECKING,
34 Any,
35 Dict,
36 Iterable,
37 Iterator,
38 List,
39 Literal,
40 Optional,
41 Tuple,
42 Type,
43 Union,
44 overload,
45)
47if TYPE_CHECKING: 47 ↛ 48line 47 didn't jump to line 48, because the condition on line 47 was never true
48 from .utils import TransactionProtocol
51log = logging.getLogger(__name__)
53# Regex for looking for URI escapes
54ESCAPES_RE = re.compile(r"%[A-F0-9]{2}")
56# Precomputed escaped hash
57ESCAPED_HASH = urllib.parse.quote("#")
59# Maximum number of worker threads for parallelized operations.
60# If greater than 10, be aware that this number has to be consistent
61# with connection pool sizing (for example in urllib3).
62MAX_WORKERS = 10
65ResourcePathExpression = Union[str, urllib.parse.ParseResult, "ResourcePath", Path]
66"""Type-annotation alias for objects that can be coerced to ResourcePath.
67"""
70class ResourcePath:
71 """Convenience wrapper around URI parsers.
73 Provides access to URI components and can convert file
74 paths into absolute path URIs. Scheme-less URIs are treated as if
75 they are local file system paths and are converted to absolute URIs.
77 A specialist subclass is created for each supported URI scheme.
79 Parameters
80 ----------
81 uri : `str`, `Path`, `urllib.parse.ParseResult`, or `ResourcePath`.
82 URI in string form. Can be scheme-less if referring to a local
83 filesystem path.
84 root : `str` or `ResourcePath`, optional
85 When fixing up a relative path in a ``file`` scheme or if scheme-less,
86 use this as the root. Must be absolute. If `None` the current
87 working directory will be used. Can be a file URI.
88 forceAbsolute : `bool`, optional
89 If `True`, scheme-less relative URI will be converted to an absolute
90 path using a ``file`` scheme. If `False` scheme-less URI will remain
91 scheme-less and will not be updated to ``file`` or absolute path.
92 forceDirectory: `bool`, optional
93 If `True` forces the URI to end with a separator, otherwise given URI
94 is interpreted as is.
95 isTemporary : `bool`, optional
96 If `True` indicates that this URI points to a temporary resource.
97 The default is `False`, unless ``uri`` is already a `ResourcePath`
98 instance and ``uri.isTemporary is True``.
99 """
101 _pathLib: Type[PurePath] = PurePosixPath
102 """Path library to use for this scheme."""
104 _pathModule = posixpath
105 """Path module to use for this scheme."""
107 transferModes: Tuple[str, ...] = ("copy", "auto", "move")
108 """Transfer modes supported by this implementation.
110 Move is special in that it is generally a copy followed by an unlink.
111 Whether that unlink works depends critically on whether the source URI
112 implements unlink. If it does not the move will be reported as a failure.
113 """
115 transferDefault: str = "copy"
116 """Default mode to use for transferring if ``auto`` is specified."""
118 quotePaths = True
119 """True if path-like elements modifying a URI should be quoted.
121 All non-schemeless URIs have to internally use quoted paths. Therefore
122 if a new file name is given (e.g. to updatedFile or join) a decision must
123 be made whether to quote it to be consistent.
124 """
126 isLocal = False
127 """If `True` this URI refers to a local file."""
129 # This is not an ABC with abstract methods because the __new__ being
130 # a factory confuses mypy such that it assumes that every constructor
131 # returns a ResourcePath and then determines that all the abstract methods
132 # are still abstract. If they are not marked abstract but just raise
133 # mypy is fine with it.
135 # mypy is confused without these
136 _uri: urllib.parse.ParseResult
137 isTemporary: bool
138 dirLike: bool
140 def __new__(
141 cls,
142 uri: ResourcePathExpression,
143 root: Optional[Union[str, ResourcePath]] = None,
144 forceAbsolute: bool = True,
145 forceDirectory: bool = False,
146 isTemporary: Optional[bool] = None,
147 ) -> ResourcePath:
148 """Create and return new specialist ResourcePath subclass."""
149 parsed: urllib.parse.ParseResult
150 dirLike: bool = False
151 subclass: Optional[Type[ResourcePath]] = None
153 if isinstance(uri, os.PathLike):
154 uri = str(uri)
156 # Record if we need to post process the URI components
157 # or if the instance is already fully configured
158 if isinstance(uri, str):
159 # Since local file names can have special characters in them
160 # we need to quote them for the parser but we can unquote
161 # later. Assume that all other URI schemes are quoted.
162 # Since sometimes people write file:/a/b and not file:///a/b
163 # we should not quote in the explicit case of file:
164 if "://" not in uri and not uri.startswith("file:"):
165 if ESCAPES_RE.search(uri):
166 log.warning("Possible double encoding of %s", uri)
167 else:
168 uri = urllib.parse.quote(uri)
169 # Special case hash since we must support fragments
170 # even in schemeless URIs -- although try to only replace
171 # them in file part and not directory part
172 if ESCAPED_HASH in uri:
173 dirpos = uri.rfind("/")
174 # Do replacement after this /
175 uri = uri[: dirpos + 1] + uri[dirpos + 1 :].replace(ESCAPED_HASH, "#")
177 parsed = urllib.parse.urlparse(uri)
178 elif isinstance(uri, urllib.parse.ParseResult):
179 parsed = copy.copy(uri)
180 # If we are being instantiated with a subclass, rather than
181 # ResourcePath, ensure that that subclass is used directly.
182 # This could lead to inconsistencies if this constructor
183 # is used externally outside of the ResourcePath.replace() method.
184 # S3ResourcePath(urllib.parse.urlparse("file://a/b.txt"))
185 # will be a problem.
186 # This is needed to prevent a schemeless absolute URI become
187 # a file URI unexpectedly when calling updatedFile or
188 # updatedExtension
189 if cls is not ResourcePath:
190 parsed, dirLike = cls._fixDirectorySep(parsed, forceDirectory)
191 subclass = cls
193 elif isinstance(uri, ResourcePath):
194 # Since ResourcePath is immutable we can return the argument
195 # unchanged if it already agrees with forceDirectory, isTemporary,
196 # and forceAbsolute.
197 # We invoke __new__ again with str(self) to add a scheme for
198 # forceAbsolute, but for the others that seems more likely to paper
199 # over logic errors than do something useful, so we just raise.
200 if forceDirectory and not uri.dirLike:
201 raise RuntimeError(
202 f"{uri} is already a file-like ResourcePath; cannot force it to directory."
203 )
204 if isTemporary is not None and isTemporary is not uri.isTemporary:
205 raise RuntimeError(
206 f"{uri} is already a {'temporary' if uri.isTemporary else 'permanent'} "
207 f"ResourcePath; cannot make it {'temporary' if isTemporary else 'permanent'}."
208 )
209 if forceAbsolute and not uri.scheme:
210 return ResourcePath(
211 str(uri),
212 root=root,
213 forceAbsolute=True,
214 forceDirectory=uri.dirLike,
215 isTemporary=uri.isTemporary,
216 )
217 return uri
218 else:
219 raise ValueError(
220 f"Supplied URI must be string, Path, ResourcePath, or ParseResult but got '{uri!r}'"
221 )
223 if subclass is None:
224 # Work out the subclass from the URI scheme
225 if not parsed.scheme:
226 from .schemeless import SchemelessResourcePath
228 subclass = SchemelessResourcePath
229 elif parsed.scheme == "file":
230 from .file import FileResourcePath
232 subclass = FileResourcePath
233 elif parsed.scheme == "s3":
234 from .s3 import S3ResourcePath
236 subclass = S3ResourcePath
237 elif parsed.scheme.startswith("http"):
238 from .http import HttpResourcePath
240 subclass = HttpResourcePath
241 elif parsed.scheme == "resource":
242 # Rules for scheme names disallow pkg_resource
243 from .packageresource import PackageResourcePath
245 subclass = PackageResourcePath
246 elif parsed.scheme == "mem":
247 # in-memory datastore object
248 from .mem import InMemoryResourcePath
250 subclass = InMemoryResourcePath
251 else:
252 raise NotImplementedError(
253 f"No URI support for scheme: '{parsed.scheme}' in {parsed.geturl()}"
254 )
256 parsed, dirLike = subclass._fixupPathUri(
257 parsed, root=root, forceAbsolute=forceAbsolute, forceDirectory=forceDirectory
258 )
260 # It is possible for the class to change from schemeless
261 # to file so handle that
262 if parsed.scheme == "file":
263 from .file import FileResourcePath
265 subclass = FileResourcePath
267 # Now create an instance of the correct subclass and set the
268 # attributes directly
269 self = object.__new__(subclass)
270 self._uri = parsed
271 self.dirLike = dirLike
272 if isTemporary is None:
273 isTemporary = False
274 self.isTemporary = isTemporary
275 return self
277 @property
278 def scheme(self) -> str:
279 """Return the URI scheme.
281 Notes
282 -----
283 (``://`` is not part of the scheme).
284 """
285 return self._uri.scheme
287 @property
288 def netloc(self) -> str:
289 """Return the URI network location."""
290 return self._uri.netloc
292 @property
293 def path(self) -> str:
294 """Return the path component of the URI."""
295 return self._uri.path
297 @property
298 def unquoted_path(self) -> str:
299 """Return path component of the URI with any URI quoting reversed."""
300 return urllib.parse.unquote(self._uri.path)
302 @property
303 def ospath(self) -> str:
304 """Return the path component of the URI localized to current OS."""
305 raise AttributeError(f"Non-file URI ({self}) has no local OS path.")
307 @property
308 def relativeToPathRoot(self) -> str:
309 """Return path relative to network location.
311 Effectively, this is the path property with posix separator stripped
312 from the left hand side of the path.
314 Always unquotes.
315 """
316 p = self._pathLib(self.path)
317 relToRoot = str(p.relative_to(p.root))
318 if self.dirLike and not relToRoot.endswith("/"):
319 relToRoot += "/"
320 return urllib.parse.unquote(relToRoot)
322 @property
323 def is_root(self) -> bool:
324 """Return whether this URI points to the root of the network location.
326 This means that the path components refers to the top level.
327 """
328 relpath = self.relativeToPathRoot
329 if relpath == "./":
330 return True
331 return False
333 @property
334 def fragment(self) -> str:
335 """Return the fragment component of the URI."""
336 return self._uri.fragment
338 @property
339 def params(self) -> str:
340 """Return any parameters included in the URI."""
341 return self._uri.params
343 @property
344 def query(self) -> str:
345 """Return any query strings included in the URI."""
346 return self._uri.query
348 def geturl(self) -> str:
349 """Return the URI in string form.
351 Returns
352 -------
353 url : `str`
354 String form of URI.
355 """
356 return self._uri.geturl()
358 def root_uri(self) -> ResourcePath:
359 """Return the base root URI.
361 Returns
362 -------
363 uri : `ResourcePath`
364 root URI.
365 """
366 return self.replace(path="", forceDirectory=True)
368 def split(self) -> Tuple[ResourcePath, str]:
369 """Split URI into head and tail.
371 Returns
372 -------
373 head: `ResourcePath`
374 Everything leading up to tail, expanded and normalized as per
375 ResourcePath rules.
376 tail : `str`
377 Last `self.path` component. Tail will be empty if path ends on a
378 separator. Tail will never contain separators. It will be
379 unquoted.
381 Notes
382 -----
383 Equivalent to `os.path.split()` where head preserves the URI
384 components.
385 """
386 head, tail = self._pathModule.split(self.path)
387 headuri = self._uri._replace(path=head)
389 # The file part should never include quoted metacharacters
390 tail = urllib.parse.unquote(tail)
392 # Schemeless is special in that it can be a relative path
393 # We need to ensure that it stays that way. All other URIs will
394 # be absolute already.
395 forceAbsolute = self._pathModule.isabs(self.path)
396 return ResourcePath(headuri, forceDirectory=True, forceAbsolute=forceAbsolute), tail
398 def basename(self) -> str:
399 """Return the base name, last element of path, of the URI.
401 Returns
402 -------
403 tail : `str`
404 Last part of the path attribute. Trail will be empty if path ends
405 on a separator.
407 Notes
408 -----
409 If URI ends on a slash returns an empty string. This is the second
410 element returned by `split()`.
412 Equivalent of `os.path.basename()``.
413 """
414 return self.split()[1]
416 def dirname(self) -> ResourcePath:
417 """Return the directory component of the path as a new `ResourcePath`.
419 Returns
420 -------
421 head : `ResourcePath`
422 Everything except the tail of path attribute, expanded and
423 normalized as per ResourcePath rules.
425 Notes
426 -----
427 Equivalent of `os.path.dirname()`.
428 """
429 return self.split()[0]
431 def parent(self) -> ResourcePath:
432 """Return a `ResourcePath` of the parent directory.
434 Returns
435 -------
436 head : `ResourcePath`
437 Everything except the tail of path attribute, expanded and
438 normalized as per `ResourcePath` rules.
440 Notes
441 -----
442 For a file-like URI this will be the same as calling `dirname()`.
443 """
444 # When self is file-like, return self.dirname()
445 if not self.dirLike:
446 return self.dirname()
447 # When self is dir-like, return its parent directory,
448 # regardless of the presence of a trailing separator
449 originalPath = self._pathLib(self.path)
450 parentPath = originalPath.parent
451 return self.replace(path=str(parentPath), forceDirectory=True)
453 def replace(self, forceDirectory: bool = False, isTemporary: bool = False, **kwargs: Any) -> ResourcePath:
454 """Return new `ResourcePath` with specified components replaced.
456 Parameters
457 ----------
458 forceDirectory : `bool`, optional
459 Parameter passed to ResourcePath constructor to force this
460 new URI to be dir-like.
461 isTemporary : `bool`, optional
462 Indicate that the resulting URI is temporary resource.
463 **kwargs
464 Components of a `urllib.parse.ParseResult` that should be
465 modified for the newly-created `ResourcePath`.
467 Returns
468 -------
469 new : `ResourcePath`
470 New `ResourcePath` object with updated values.
472 Notes
473 -----
474 Does not, for now, allow a change in URI scheme.
475 """
476 # Disallow a change in scheme
477 if "scheme" in kwargs:
478 raise ValueError(f"Can not use replace() method to change URI scheme for {self}")
479 return self.__class__(
480 self._uri._replace(**kwargs), forceDirectory=forceDirectory, isTemporary=isTemporary
481 )
483 def updatedFile(self, newfile: str) -> ResourcePath:
484 """Return new URI with an updated final component of the path.
486 Parameters
487 ----------
488 newfile : `str`
489 File name with no path component.
491 Returns
492 -------
493 updated : `ResourcePath`
495 Notes
496 -----
497 Forces the ResourcePath.dirLike attribute to be false. The new file
498 path will be quoted if necessary.
499 """
500 if self.quotePaths:
501 newfile = urllib.parse.quote(newfile)
502 dir, _ = self._pathModule.split(self.path)
503 newpath = self._pathModule.join(dir, newfile)
505 updated = self.replace(path=newpath)
506 updated.dirLike = False
507 return updated
509 def updatedExtension(self, ext: Optional[str]) -> ResourcePath:
510 """Return a new `ResourcePath` with updated file extension.
512 All file extensions are replaced.
514 Parameters
515 ----------
516 ext : `str` or `None`
517 New extension. If an empty string is given any extension will
518 be removed. If `None` is given there will be no change.
520 Returns
521 -------
522 updated : `ResourcePath`
523 URI with the specified extension. Can return itself if
524 no extension was specified.
525 """
526 if ext is None:
527 return self
529 # Get the extension
530 current = self.getExtension()
532 # Nothing to do if the extension already matches
533 if current == ext:
534 return self
536 # Remove the current extension from the path
537 # .fits.gz counts as one extension do not use os.path.splitext
538 path = self.path
539 if current:
540 path = path[: -len(current)]
542 # Ensure that we have a leading "." on file extension (and we do not
543 # try to modify the empty string)
544 if ext and not ext.startswith("."):
545 ext = "." + ext
547 return self.replace(path=path + ext)
549 def getExtension(self) -> str:
550 """Return the file extension(s) associated with this URI path.
552 Returns
553 -------
554 ext : `str`
555 The file extension (including the ``.``). Can be empty string
556 if there is no file extension. Usually returns only the last
557 file extension unless there is a special extension modifier
558 indicating file compression, in which case the combined
559 extension (e.g. ``.fits.gz``) will be returned.
560 """
561 special = {".gz", ".bz2", ".xz", ".fz"}
563 # Get the file part of the path so as not to be confused by
564 # "." in directory names.
565 basename = self.basename()
566 extensions = self._pathLib(basename).suffixes
568 if not extensions:
569 return ""
571 ext = extensions.pop()
573 # Multiple extensions, decide whether to include the final two
574 if extensions and ext in special:
575 ext = f"{extensions[-1]}{ext}"
577 return ext
579 def join(
580 self, path: Union[str, ResourcePath], isTemporary: Optional[bool] = None, forceDirectory: bool = False
581 ) -> ResourcePath:
582 """Return new `ResourcePath` with additional path components.
584 Parameters
585 ----------
586 path : `str`, `ResourcePath`
587 Additional file components to append to the current URI. Assumed
588 to include a file at the end. Will be quoted depending on the
589 associated URI scheme. If the path looks like a URI with a scheme
590 referring to an absolute location, it will be returned
591 directly (matching the behavior of `os.path.join()`). It can
592 also be a `ResourcePath`.
593 isTemporary : `bool`, optional
594 Indicate that the resulting URI represents a temporary resource.
595 Default is ``self.isTemporary``.
596 forceDirectory : `bool`, optional
597 If `True` forces the URI to end with a separator, otherwise given
598 URI is interpreted as is.
600 Returns
601 -------
602 new : `ResourcePath`
603 New URI with any file at the end replaced with the new path
604 components.
606 Notes
607 -----
608 Schemeless URIs assume local path separator but all other URIs assume
609 POSIX separator if the supplied path has directory structure. It
610 may be this never becomes a problem but datastore templates assume
611 POSIX separator is being used.
613 If an absolute `ResourcePath` is given for ``path`` is is assumed that
614 this should be returned directly. Giving a ``path`` of an absolute
615 scheme-less URI is not allowed for safety reasons as it may indicate
616 a mistake in the calling code.
618 Raises
619 ------
620 ValueError
621 Raised if the ``path`` is an absolute scheme-less URI. In that
622 situation it is unclear whether the intent is to return a
623 ``file`` URI or it was a mistake and a relative scheme-less URI
624 was meant.
625 RuntimeError
626 Raised if this attempts to join a temporary URI to a non-temporary
627 URI.
628 """
629 if isTemporary is None:
630 isTemporary = self.isTemporary
631 elif not isTemporary and self.isTemporary:
632 raise RuntimeError("Cannot join temporary URI to non-temporary URI.")
633 # If we have a full URI in path we will use it directly
634 # but without forcing to absolute so that we can trap the
635 # expected option of relative path.
636 path_uri = ResourcePath(
637 path, forceAbsolute=False, forceDirectory=forceDirectory, isTemporary=isTemporary
638 )
639 if path_uri.scheme:
640 # Check for scheme so can distinguish explicit URIs from
641 # absolute scheme-less URIs.
642 return path_uri
644 if path_uri.isabs():
645 # Absolute scheme-less path.
646 raise ValueError(f"Can not join absolute scheme-less {path_uri!r} to another URI.")
648 # If this was originally a ResourcePath extract the unquoted path from
649 # it. Otherwise we use the string we were given to allow "#" to appear
650 # in the filename if given as a plain string.
651 if not isinstance(path, str):
652 path = path_uri.unquoted_path
654 new = self.dirname() # By definition a directory URI
656 # new should be asked about quoting, not self, since dirname can
657 # change the URI scheme for schemeless -> file
658 if new.quotePaths:
659 path = urllib.parse.quote(path)
661 newpath = self._pathModule.normpath(self._pathModule.join(new.path, path))
663 # normpath can strip trailing / so we force directory if the supplied
664 # path ended with a /
665 return new.replace(
666 path=newpath,
667 forceDirectory=(forceDirectory or path.endswith(self._pathModule.sep)),
668 isTemporary=isTemporary,
669 )
671 def relative_to(self, other: ResourcePath) -> Optional[str]:
672 """Return the relative path from this URI to the other URI.
674 Parameters
675 ----------
676 other : `ResourcePath`
677 URI to use to calculate the relative path. Must be a parent
678 of this URI.
680 Returns
681 -------
682 subpath : `str`
683 The sub path of this URI relative to the supplied other URI.
684 Returns `None` if there is no parent child relationship.
685 Scheme and netloc must match.
686 """
687 # Scheme-less absolute other is treated as if it's a file scheme.
688 # Scheme-less relative other can only return non-None if self
689 # is also scheme-less relative and that is handled specifically
690 # in a subclass.
691 if not other.scheme and other.isabs():
692 other = other.abspath()
694 # Scheme-less self is handled elsewhere.
695 if self.scheme != other.scheme or self.netloc != other.netloc:
696 return None
698 enclosed_path = self._pathLib(self.relativeToPathRoot)
699 parent_path = other.relativeToPathRoot
700 subpath: Optional[str]
701 try:
702 subpath = str(enclosed_path.relative_to(parent_path))
703 except ValueError:
704 subpath = None
705 else:
706 subpath = urllib.parse.unquote(subpath)
707 return subpath
709 def exists(self) -> bool:
710 """Indicate that the resource is available.
712 Returns
713 -------
714 exists : `bool`
715 `True` if the resource exists.
716 """
717 raise NotImplementedError()
719 @classmethod
720 def mexists(cls, uris: Iterable[ResourcePath]) -> Dict[ResourcePath, bool]:
721 """Check for existence of multiple URIs at once.
723 Parameters
724 ----------
725 uris : iterable of `ResourcePath`
726 The URIs to test.
728 Returns
729 -------
730 existence : `dict` of [`ResourcePath`, `bool`]
731 Mapping of original URI to boolean indicating existence.
732 """
733 exists_executor = concurrent.futures.ThreadPoolExecutor(max_workers=MAX_WORKERS)
734 future_exists = {exists_executor.submit(uri.exists): uri for uri in uris}
736 results: Dict[ResourcePath, bool] = {}
737 for future in concurrent.futures.as_completed(future_exists):
738 uri = future_exists[future]
739 try:
740 exists = future.result()
741 except Exception:
742 exists = False
743 results[uri] = exists
744 return results
746 def remove(self) -> None:
747 """Remove the resource."""
748 raise NotImplementedError()
750 def isabs(self) -> bool:
751 """Indicate that the resource is fully specified.
753 For non-schemeless URIs this is always true.
755 Returns
756 -------
757 isabs : `bool`
758 `True` in all cases except schemeless URI.
759 """
760 return True
762 def abspath(self) -> ResourcePath:
763 """Return URI using an absolute path.
765 Returns
766 -------
767 abs : `ResourcePath`
768 Absolute URI. For non-schemeless URIs this always returns itself.
769 Schemeless URIs are upgraded to file URIs.
770 """
771 return self
773 def _as_local(self) -> Tuple[str, bool]:
774 """Return the location of the (possibly remote) resource as local file.
776 This is a helper function for `as_local` context manager.
778 Returns
779 -------
780 path : `str`
781 If this is a remote resource, it will be a copy of the resource
782 on the local file system, probably in a temporary directory.
783 For a local resource this should be the actual path to the
784 resource.
785 is_temporary : `bool`
786 Indicates if the local path is a temporary file or not.
787 """
788 raise NotImplementedError()
790 @contextlib.contextmanager
791 def as_local(self) -> Iterator[ResourcePath]:
792 """Return the location of the (possibly remote) resource as local file.
794 Yields
795 ------
796 local : `ResourcePath`
797 If this is a remote resource, it will be a copy of the resource
798 on the local file system, probably in a temporary directory.
799 For a local resource this should be the actual path to the
800 resource.
802 Notes
803 -----
804 The context manager will automatically delete any local temporary
805 file.
807 Examples
808 --------
809 Should be used as a context manager:
811 .. code-block:: py
813 with uri.as_local() as local:
814 ospath = local.ospath
815 """
816 if self.dirLike:
817 raise IsADirectoryError(f"Directory-like URI {self} cannot be fetched as local.")
818 local_src, is_temporary = self._as_local()
819 local_uri = ResourcePath(local_src, isTemporary=is_temporary)
821 try:
822 yield local_uri
823 finally:
824 # The caller might have relocated the temporary file.
825 # Do not ever delete if the temporary matches self
826 # (since it may have been that a temporary file was made local
827 # but already was local).
828 if self != local_uri and is_temporary and local_uri.exists():
829 local_uri.remove()
831 @classmethod
832 @contextlib.contextmanager
833 def temporary_uri(
834 cls, prefix: Optional[ResourcePath] = None, suffix: Optional[str] = None
835 ) -> Iterator[ResourcePath]:
836 """Create a temporary file-like URI.
838 Parameters
839 ----------
840 prefix : `ResourcePath`, optional
841 Prefix to use. Without this the path will be formed as a local
842 file URI in a temporary directory. Ensuring that the prefix
843 location exists is the responsibility of the caller.
844 suffix : `str`, optional
845 A file suffix to be used. The ``.`` should be included in this
846 suffix.
848 Yields
849 ------
850 uri : `ResourcePath`
851 The temporary URI. Will be removed when the context is completed.
852 """
853 use_tempdir = False
854 if prefix is None:
855 prefix = ResourcePath(tempfile.mkdtemp(), forceDirectory=True, isTemporary=True)
856 # Record that we need to delete this directory. Can not rely
857 # on isTemporary flag since an external prefix may have that
858 # set as well.
859 use_tempdir = True
861 # Need to create a randomized file name. For consistency do not
862 # use mkstemp for local and something else for remote. Additionally
863 # this method does not create the file to prevent name clashes.
864 characters = "abcdefghijklmnopqrstuvwxyz0123456789_"
865 rng = Random()
866 tempname = "".join(rng.choice(characters) for _ in range(16))
867 if suffix:
868 tempname += suffix
869 temporary_uri = prefix.join(tempname, isTemporary=True)
870 if temporary_uri.dirLike:
871 # If we had a safe way to clean up a remote temporary directory, we
872 # could support this.
873 raise NotImplementedError("temporary_uri cannot be used to create a temporary directory.")
874 try:
875 yield temporary_uri
876 finally:
877 if use_tempdir:
878 shutil.rmtree(prefix.ospath, ignore_errors=True)
879 else:
880 try:
881 # It's okay if this does not work because the user removed
882 # the file.
883 temporary_uri.remove()
884 except FileNotFoundError:
885 pass
887 def read(self, size: int = -1) -> bytes:
888 """Open the resource and return the contents in bytes.
890 Parameters
891 ----------
892 size : `int`, optional
893 The number of bytes to read. Negative or omitted indicates
894 that all data should be read.
895 """
896 raise NotImplementedError()
898 def write(self, data: bytes, overwrite: bool = True) -> None:
899 """Write the supplied bytes to the new resource.
901 Parameters
902 ----------
903 data : `bytes`
904 The bytes to write to the resource. The entire contents of the
905 resource will be replaced.
906 overwrite : `bool`, optional
907 If `True` the resource will be overwritten if it exists. Otherwise
908 the write will fail.
909 """
910 raise NotImplementedError()
912 def mkdir(self) -> None:
913 """For a dir-like URI, create the directory resource if needed."""
914 raise NotImplementedError()
916 def isdir(self) -> bool:
917 """Return True if this URI looks like a directory, else False."""
918 return self.dirLike
920 def size(self) -> int:
921 """For non-dir-like URI, return the size of the resource.
923 Returns
924 -------
925 sz : `int`
926 The size in bytes of the resource associated with this URI.
927 Returns 0 if dir-like.
928 """
929 raise NotImplementedError()
931 def __str__(self) -> str:
932 """Convert the URI to its native string form."""
933 return self.geturl()
935 def __repr__(self) -> str:
936 """Return string representation suitable for evaluation."""
937 return f'ResourcePath("{self.geturl()}")'
939 def __eq__(self, other: Any) -> bool:
940 """Compare supplied object with this `ResourcePath`."""
941 if not isinstance(other, ResourcePath):
942 return NotImplemented
943 return self.geturl() == other.geturl()
945 def __hash__(self) -> int:
946 """Return hash of this object."""
947 return hash(str(self))
949 def __lt__(self, other: ResourcePath) -> bool:
950 return self.geturl() < other.geturl()
952 def __le__(self, other: ResourcePath) -> bool:
953 return self.geturl() <= other.geturl()
955 def __gt__(self, other: ResourcePath) -> bool:
956 return self.geturl() > other.geturl()
958 def __ge__(self, other: ResourcePath) -> bool:
959 return self.geturl() >= other.geturl()
961 def __copy__(self) -> ResourcePath:
962 """Copy constructor.
964 Object is immutable so copy can return itself.
965 """
966 # Implement here because the __new__ method confuses things
967 return self
969 def __deepcopy__(self, memo: Any) -> ResourcePath:
970 """Deepcopy the object.
972 Object is immutable so copy can return itself.
973 """
974 # Implement here because the __new__ method confuses things
975 return self
977 def __getnewargs__(self) -> Tuple:
978 """Support pickling."""
979 return (str(self),)
981 @classmethod
982 def _fixDirectorySep(
983 cls, parsed: urllib.parse.ParseResult, forceDirectory: bool = False
984 ) -> Tuple[urllib.parse.ParseResult, bool]:
985 """Ensure that a path separator is present on directory paths.
987 Parameters
988 ----------
989 parsed : `~urllib.parse.ParseResult`
990 The result from parsing a URI using `urllib.parse`.
991 forceDirectory : `bool`, optional
992 If `True` forces the URI to end with a separator, otherwise given
993 URI is interpreted as is. Specifying that the URI is conceptually
994 equivalent to a directory can break some ambiguities when
995 interpreting the last element of a path.
997 Returns
998 -------
999 modified : `~urllib.parse.ParseResult`
1000 Update result if a URI is being handled.
1001 dirLike : `bool`
1002 `True` if given parsed URI has a trailing separator or
1003 forceDirectory is True. Otherwise `False`.
1004 """
1005 # assume we are not dealing with a directory like URI
1006 dirLike = False
1008 # Directory separator
1009 sep = cls._pathModule.sep
1011 # URI is dir-like if explicitly stated or if it ends on a separator
1012 endsOnSep = parsed.path.endswith(sep)
1013 if forceDirectory or endsOnSep:
1014 dirLike = True
1015 # only add the separator if it's not already there
1016 if not endsOnSep:
1017 parsed = parsed._replace(path=parsed.path + sep)
1019 return parsed, dirLike
1021 @classmethod
1022 def _fixupPathUri(
1023 cls,
1024 parsed: urllib.parse.ParseResult,
1025 root: Optional[Union[str, ResourcePath]] = None,
1026 forceAbsolute: bool = False,
1027 forceDirectory: bool = False,
1028 ) -> Tuple[urllib.parse.ParseResult, bool]:
1029 """Correct any issues with the supplied URI.
1031 Parameters
1032 ----------
1033 parsed : `~urllib.parse.ParseResult`
1034 The result from parsing a URI using `urllib.parse`.
1035 root : `str` or `ResourcePath`, ignored
1036 Not used by the this implementation since all URIs are
1037 absolute except for those representing the local file system.
1038 forceAbsolute : `bool`, ignored.
1039 Not used by this implementation. URIs are generally always
1040 absolute.
1041 forceDirectory : `bool`, optional
1042 If `True` forces the URI to end with a separator, otherwise given
1043 URI is interpreted as is. Specifying that the URI is conceptually
1044 equivalent to a directory can break some ambiguities when
1045 interpreting the last element of a path.
1047 Returns
1048 -------
1049 modified : `~urllib.parse.ParseResult`
1050 Update result if a URI is being handled.
1051 dirLike : `bool`
1052 `True` if given parsed URI has a trailing separator or
1053 forceDirectory is True. Otherwise `False`.
1055 Notes
1056 -----
1057 Relative paths are explicitly not supported by RFC8089 but `urllib`
1058 does accept URIs of the form ``file:relative/path.ext``. They need
1059 to be turned into absolute paths before they can be used. This is
1060 always done regardless of the ``forceAbsolute`` parameter.
1062 AWS S3 differentiates between keys with trailing POSIX separators (i.e
1063 `/dir` and `/dir/`) whereas POSIX does not neccessarily.
1065 Scheme-less paths are normalized.
1066 """
1067 return cls._fixDirectorySep(parsed, forceDirectory)
1069 def transfer_from(
1070 self,
1071 src: ResourcePath,
1072 transfer: str,
1073 overwrite: bool = False,
1074 transaction: Optional[TransactionProtocol] = None,
1075 ) -> None:
1076 """Transfer the current resource to a new location.
1078 Parameters
1079 ----------
1080 src : `ResourcePath`
1081 Source URI.
1082 transfer : `str`
1083 Mode to use for transferring the resource. Generically there are
1084 many standard options: copy, link, symlink, hardlink, relsymlink.
1085 Not all URIs support all modes.
1086 overwrite : `bool`, optional
1087 Allow an existing file to be overwritten. Defaults to `False`.
1088 transaction : `~lsst.resources.utils.TransactionProtocol`, optional
1089 A transaction object that can (depending on implementation)
1090 rollback transfers on error. Not guaranteed to be implemented.
1092 Notes
1093 -----
1094 Conceptually this is hard to scale as the number of URI schemes
1095 grow. The destination URI is more important than the source URI
1096 since that is where all the transfer modes are relevant (with the
1097 complication that "move" deletes the source).
1099 Local file to local file is the fundamental use case but every
1100 other scheme has to support "copy" to local file (with implicit
1101 support for "move") and copy from local file.
1102 All the "link" options tend to be specific to local file systems.
1104 "move" is a "copy" where the remote resource is deleted at the end.
1105 Whether this works depends on the source URI rather than the
1106 destination URI. Reverting a move on transaction rollback is
1107 expected to be problematic if a remote resource was involved.
1108 """
1109 raise NotImplementedError(f"No transfer modes supported by URI scheme {self.scheme}")
1111 def walk(
1112 self, file_filter: Optional[Union[str, re.Pattern]] = None
1113 ) -> Iterator[Union[List, Tuple[ResourcePath, List[str], List[str]]]]:
1114 """Walk the directory tree returning matching files and directories.
1116 Parameters
1117 ----------
1118 file_filter : `str` or `re.Pattern`, optional
1119 Regex to filter out files from the list before it is returned.
1121 Yields
1122 ------
1123 dirpath : `ResourcePath`
1124 Current directory being examined.
1125 dirnames : `list` of `str`
1126 Names of subdirectories within dirpath.
1127 filenames : `list` of `str`
1128 Names of all the files within dirpath.
1129 """
1130 raise NotImplementedError()
1132 @overload
1133 @classmethod
1134 def findFileResources(
1135 cls,
1136 candidates: Iterable[ResourcePathExpression],
1137 file_filter: Optional[Union[str, re.Pattern]],
1138 grouped: Literal[True],
1139 ) -> Iterator[Iterator[ResourcePath]]:
1140 ...
1142 @overload
1143 @classmethod
1144 def findFileResources(
1145 cls,
1146 candidates: Iterable[ResourcePathExpression],
1147 *,
1148 grouped: Literal[True],
1149 ) -> Iterator[Iterator[ResourcePath]]:
1150 ...
1152 @overload
1153 @classmethod
1154 def findFileResources(
1155 cls,
1156 candidates: Iterable[ResourcePathExpression],
1157 file_filter: Optional[Union[str, re.Pattern]] = None,
1158 grouped: Literal[False] = False,
1159 ) -> Iterator[ResourcePath]:
1160 ...
1162 @classmethod
1163 def findFileResources(
1164 cls,
1165 candidates: Iterable[ResourcePathExpression],
1166 file_filter: Optional[Union[str, re.Pattern]] = None,
1167 grouped: bool = False,
1168 ) -> Iterator[Union[ResourcePath, Iterator[ResourcePath]]]:
1169 """Get all the files from a list of values.
1171 Parameters
1172 ----------
1173 candidates : iterable [`str` or `ResourcePath`]
1174 The files to return and directories in which to look for files to
1175 return.
1176 file_filter : `str` or `re.Pattern`, optional
1177 The regex to use when searching for files within directories.
1178 By default returns all the found files.
1179 grouped : `bool`, optional
1180 If `True` the results will be grouped by directory and each
1181 yielded value will be an iterator over URIs. If `False` each
1182 URI will be returned separately.
1184 Yields
1185 ------
1186 found_file: `ResourcePath`
1187 The passed-in URIs and URIs found in passed-in directories.
1188 If grouping is enabled, each of the yielded values will be an
1189 iterator yielding members of the group. Files given explicitly
1190 will be returned as a single group at the end.
1192 Notes
1193 -----
1194 If a value is a file it is yielded immediately without checking that it
1195 exists. If a value is a directory, all the files in the directory
1196 (recursively) that match the regex will be yielded in turn.
1197 """
1198 fileRegex = None if file_filter is None else re.compile(file_filter)
1200 singles = []
1202 # Find all the files of interest
1203 for location in candidates:
1204 uri = ResourcePath(location)
1205 if uri.isdir():
1206 for found in uri.walk(fileRegex):
1207 if not found:
1208 # This means the uri does not exist and by
1209 # convention we ignore it
1210 continue
1211 root, dirs, files = found
1212 if not files:
1213 continue
1214 if grouped:
1215 yield (root.join(name) for name in files)
1216 else:
1217 for name in files:
1218 yield root.join(name)
1219 else:
1220 if grouped:
1221 singles.append(uri)
1222 else:
1223 yield uri
1225 # Finally, return any explicitly given files in one group
1226 if grouped and singles:
1227 yield iter(singles)
1229 @contextlib.contextmanager
1230 def open(
1231 self,
1232 mode: str = "r",
1233 *,
1234 encoding: Optional[str] = None,
1235 prefer_file_temporary: bool = False,
1236 ) -> Iterator[IO]:
1237 """Return a context manager that wraps an object that behaves like an
1238 open file at the location of the URI.
1240 Parameters
1241 ----------
1242 mode : `str`
1243 String indicating the mode in which to open the file. Values are
1244 the same as those accepted by `builtins.open`, though intrinsically
1245 read-only URI types may only support read modes, and
1246 `io.IOBase.seekable` is not guaranteed to be `True` on the returned
1247 object.
1248 encoding : `str`, optional
1249 Unicode encoding for text IO; ignored for binary IO. Defaults to
1250 ``locale.getpreferredencoding(False)``, just as `builtins.open`
1251 does.
1252 prefer_file_temporary : `bool`, optional
1253 If `True`, for implementations that require transfers from a remote
1254 system to temporary local storage and/or back, use a temporary file
1255 instead of an in-memory buffer; this is generally slower, but it
1256 may be necessary to avoid excessive memory usage by large files.
1257 Ignored by implementations that do not require a temporary.
1259 Returns
1260 -------
1261 cm : `contextlib.ContextManager`
1262 A context manager that wraps a file-like object.
1264 Notes
1265 -----
1266 The default implementation of this method uses a local temporary buffer
1267 (in-memory or file, depending on ``prefer_file_temporary``) with calls
1268 to `read`, `write`, `as_local`, and `transfer_from` as necessary to
1269 read and write from/to remote systems. Remote writes thus occur only
1270 when the context manager is exited. `ResourcePath` implementations
1271 that can return a more efficient native buffer should do so whenever
1272 possible (as is guaranteed for local files). `ResourcePath`
1273 implementations for which `as_local` does not return a temporary are
1274 required to reimplement `open`, though they may delegate to `super`
1275 when `prefer_file_temporary` is `False`.
1276 """
1277 if self.dirLike:
1278 raise IsADirectoryError(f"Directory-like URI {self} cannot be opened.")
1279 if "x" in mode and self.exists():
1280 raise FileExistsError(f"File at {self} already exists.")
1281 if prefer_file_temporary:
1282 if "r" in mode or "a" in mode:
1283 local_cm = self.as_local()
1284 else:
1285 local_cm = self.temporary_uri(suffix=self.getExtension())
1286 with local_cm as local_uri:
1287 assert local_uri.isTemporary, (
1288 "ResourcePath implementations for which as_local is not "
1289 "a temporary must reimplement `open`."
1290 )
1291 with open(local_uri.ospath, mode=mode, encoding=encoding) as file_buffer:
1292 if "a" in mode:
1293 file_buffer.seek(0, io.SEEK_END)
1294 yield file_buffer
1295 if "r" not in mode or "+" in mode:
1296 self.transfer_from(local_uri, transfer="copy", overwrite=("x" not in mode))
1297 else:
1298 if "r" in mode or "a" in mode:
1299 in_bytes = self.read()
1300 else:
1301 in_bytes = b""
1302 if "b" in mode:
1303 bytes_buffer = io.BytesIO(in_bytes)
1304 if "a" in mode:
1305 bytes_buffer.seek(0, io.SEEK_END)
1306 yield bytes_buffer
1307 out_bytes = bytes_buffer.getvalue()
1308 else:
1309 if encoding is None:
1310 encoding = locale.getpreferredencoding(False)
1311 str_buffer = io.StringIO(in_bytes.decode(encoding))
1312 if "a" in mode:
1313 str_buffer.seek(0, io.SEEK_END)
1314 yield str_buffer
1315 out_bytes = str_buffer.getvalue().encode(encoding)
1316 if "r" not in mode or "+" in mode:
1317 self.write(out_bytes, overwrite=("x" not in mode))