16 from .errors import KeepWriteError, AssertionError, ArgumentError
17 from .keep import KeepLocator
18 from ._normalize_stream import normalize_stream
19 from ._ranges import locators_and_ranges, replace_range, Range
20 from .retry import retry_method
25 _logger = logging.getLogger('arvados.arvfile')
28 """split(path) -> streamname, filename
30 Separate the stream name and file name in a /-separated stream path and
31 return a tuple (stream_name, file_name). If no stream name is available,
36 stream_name, file_name = path.rsplit('/', 1)
37 except ValueError: # No / in string
38 stream_name, file_name = '.', path
39 return stream_name, file_name
41 class _FileLikeObjectBase(object):
42 def __init__(self, name, mode):
48 def _before_close(orig_func):
49 @functools.wraps(orig_func)
50 def before_close_wrapper(self, *args, **kwargs):
52 raise ValueError("I/O operation on closed stream file")
53 return orig_func(self, *args, **kwargs)
54 return before_close_wrapper
59 def __exit__(self, exc_type, exc_value, traceback):
70 class ArvadosFileReaderBase(_FileLikeObjectBase):
71 def __init__(self, name, mode, num_retries=None):
72 super(ArvadosFileReaderBase, self).__init__(name, mode)
74 self.num_retries = num_retries
75 self._readline_cache = (None, None)
79 data = self.readline()
84 def decompressed_name(self):
85 return re.sub('\.(bz2|gz)$', '', self.name)
87 @_FileLikeObjectBase._before_close
88 def seek(self, pos, whence=os.SEEK_SET):
89 if whence == os.SEEK_CUR:
91 elif whence == os.SEEK_END:
93 self._filepos = min(max(pos, 0L), self.size())
98 @_FileLikeObjectBase._before_close
100 def readall(self, size=2**20, num_retries=None):
102 data = self.read(size, num_retries=num_retries)
107 @_FileLikeObjectBase._before_close
109 def readline(self, size=float('inf'), num_retries=None):
110 cache_pos, cache_data = self._readline_cache
111 if self.tell() == cache_pos:
113 self._filepos += len(cache_data)
116 data_size = len(data[-1])
117 while (data_size < size) and ('\n' not in data[-1]):
118 next_read = self.read(2 ** 20, num_retries=num_retries)
121 data.append(next_read)
122 data_size += len(next_read)
125 nextline_index = data.index('\n') + 1
127 nextline_index = len(data)
128 nextline_index = min(nextline_index, size)
129 self._filepos -= len(data) - nextline_index
130 self._readline_cache = (self.tell(), data[nextline_index:])
131 return data[:nextline_index]
133 @_FileLikeObjectBase._before_close
135 def decompress(self, decompress, size, num_retries=None):
136 for segment in self.readall(size, num_retries=num_retries):
137 data = decompress(segment)
141 @_FileLikeObjectBase._before_close
143 def readall_decompressed(self, size=2**20, num_retries=None):
145 if self.name.endswith('.bz2'):
146 dc = bz2.BZ2Decompressor()
147 return self.decompress(dc.decompress, size,
148 num_retries=num_retries)
149 elif self.name.endswith('.gz'):
150 dc = zlib.decompressobj(16+zlib.MAX_WBITS)
151 return self.decompress(lambda segment: dc.decompress(dc.unconsumed_tail + segment),
152 size, num_retries=num_retries)
154 return self.readall(size, num_retries=num_retries)
156 @_FileLikeObjectBase._before_close
158 def readlines(self, sizehint=float('inf'), num_retries=None):
161 for s in self.readall(num_retries=num_retries):
164 if data_size >= sizehint:
166 return ''.join(data).splitlines(True)
169 raise NotImplementedError()
171 def read(self, size, num_retries=None):
172 raise NotImplementedError()
174 def readfrom(self, start, size, num_retries=None):
175 raise NotImplementedError()
178 class StreamFileReader(ArvadosFileReaderBase):
179 class _NameAttribute(str):
180 # The Python file API provides a plain .name attribute.
181 # Older SDK provided a name() method.
182 # This class provides both, for maximum compatibility.
186 def __init__(self, stream, segments, name):
187 super(StreamFileReader, self).__init__(self._NameAttribute(name), 'rb', num_retries=stream.num_retries)
188 self._stream = stream
189 self.segments = segments
191 def stream_name(self):
192 return self._stream.name()
195 n = self.segments[-1]
196 return n.range_start + n.range_size
198 @_FileLikeObjectBase._before_close
200 def read(self, size, num_retries=None):
201 """Read up to 'size' bytes from the stream, starting at the current file position"""
206 available_chunks = locators_and_ranges(self.segments, self._filepos, size)
208 lr = available_chunks[0]
209 data = self._stream.readfrom(lr.locator+lr.segment_offset,
211 num_retries=num_retries)
213 self._filepos += len(data)
216 @_FileLikeObjectBase._before_close
218 def readfrom(self, start, size, num_retries=None):
219 """Read up to 'size' bytes from the stream, starting at 'start'"""
224 for lr in locators_and_ranges(self.segments, start, size):
225 data.append(self._stream.readfrom(lr.locator+lr.segment_offset, lr.segment_size,
226 num_retries=num_retries))
229 def as_manifest(self):
231 for r in self.segments:
232 segs.extend(self._stream.locators_and_ranges(r.locator, r.range_size))
233 return " ".join(normalize_stream(".", {self.name: segs})) + "\n"
236 def synchronized(orig_func):
237 @functools.wraps(orig_func)
238 def synchronized_wrapper(self, *args, **kwargs):
240 return orig_func(self, *args, **kwargs)
241 return synchronized_wrapper
244 class StateChangeError(Exception):
245 def __init__(self, message, state, nextstate):
246 super(StateChangeError, self).__init__(message)
248 self.nextstate = nextstate
250 class _BufferBlock(object):
251 """A stand-in for a Keep block that is in the process of being written.
253 Writers can append to it, get the size, and compute the Keep locator.
254 There are three valid states:
260 Block is in the process of being uploaded to Keep, append is an error.
263 The block has been written to Keep, its internal buffer has been
264 released, fetching the block will fetch it via keep client (since we
265 discarded the internal copy), and identifiers referring to the BufferBlock
266 can be replaced with the block locator.
275 def __init__(self, blockid, starting_capacity, owner):
278 the identifier for this block
281 the initial buffer capacity
284 ArvadosFile that owns this block
287 self.blockid = blockid
288 self.buffer_block = bytearray(starting_capacity)
289 self.buffer_view = memoryview(self.buffer_block)
290 self.write_pointer = 0
291 self._state = _BufferBlock.WRITABLE
294 self.lock = threading.Lock()
295 self.wait_for_commit = threading.Event()
299 def append(self, data):
300 """Append some data to the buffer.
302 Only valid if the block is in WRITABLE state. Implements an expanding
303 buffer, doubling capacity as needed to accomdate all the data.
306 if self._state == _BufferBlock.WRITABLE:
307 while (self.write_pointer+len(data)) > len(self.buffer_block):
308 new_buffer_block = bytearray(len(self.buffer_block) * 2)
309 new_buffer_block[0:self.write_pointer] = self.buffer_block[0:self.write_pointer]
310 self.buffer_block = new_buffer_block
311 self.buffer_view = memoryview(self.buffer_block)
312 self.buffer_view[self.write_pointer:self.write_pointer+len(data)] = data
313 self.write_pointer += len(data)
316 raise AssertionError("Buffer block is not writable")
318 STATE_TRANSITIONS = frozenset([
320 (PENDING, COMMITTED),
325 def set_state(self, nextstate, val=None):
326 if (self._state, nextstate) not in self.STATE_TRANSITIONS:
327 raise StateChangeError("Invalid state change from %s to %s" % (self._state, nextstate), self._state, nextstate)
328 self._state = nextstate
330 if self._state == _BufferBlock.PENDING:
331 self.wait_for_commit.clear()
333 if self._state == _BufferBlock.COMMITTED:
335 self.buffer_view = None
336 self.buffer_block = None
337 self.wait_for_commit.set()
339 if self._state == _BufferBlock.ERROR:
341 self.wait_for_commit.set()
348 """The amount of data written to the buffer."""
349 return self.write_pointer
353 """The Keep locator for this buffer's contents."""
354 if self._locator is None:
355 self._locator = "%s+%i" % (hashlib.md5(self.buffer_view[0:self.write_pointer]).hexdigest(), self.size())
359 def clone(self, new_blockid, owner):
360 if self._state == _BufferBlock.COMMITTED:
361 raise AssertionError("Cannot duplicate committed buffer block")
362 bufferblock = _BufferBlock(new_blockid, self.size(), owner)
363 bufferblock.append(self.buffer_view[0:self.size()])
369 self.buffer_block = None
370 self.buffer_view = None
373 class NoopLock(object):
377 def __exit__(self, exc_type, exc_value, traceback):
380 def acquire(self, blocking=False):
387 def must_be_writable(orig_func):
388 @functools.wraps(orig_func)
389 def must_be_writable_wrapper(self, *args, **kwargs):
390 if not self.writable():
391 raise IOError(errno.EROFS, "Collection is read-only.")
392 return orig_func(self, *args, **kwargs)
393 return must_be_writable_wrapper
396 class _BlockManager(object):
397 """BlockManager handles buffer blocks.
399 Also handles background block uploads, and background block prefetch for a
400 Collection of ArvadosFiles.
404 DEFAULT_PUT_THREADS = 2
405 DEFAULT_GET_THREADS = 2
407 def __init__(self, keep, copies=None, put_threads=None):
408 """keep: KeepClient object to use"""
410 self._bufferblocks = collections.OrderedDict()
411 self._put_queue = None
412 self._put_threads = None
413 self._prefetch_queue = None
414 self._prefetch_threads = None
415 self.lock = threading.Lock()
416 self.prefetch_enabled = True
418 self.num_put_threads = put_threads
420 self.num_put_threads = _BlockManager.DEFAULT_PUT_THREADS
421 self.num_get_threads = _BlockManager.DEFAULT_GET_THREADS
423 self._pending_write_size = 0
424 self.threads_lock = threading.Lock()
427 def alloc_bufferblock(self, blockid=None, starting_capacity=2**14, owner=None):
428 """Allocate a new, empty bufferblock in WRITABLE state and return it.
431 optional block identifier, otherwise one will be automatically assigned
434 optional capacity, otherwise will use default capacity
437 ArvadosFile that owns this block
440 return self._alloc_bufferblock(blockid, starting_capacity, owner)
442 def _alloc_bufferblock(self, blockid=None, starting_capacity=2**14, owner=None):
444 blockid = "%s" % uuid.uuid4()
445 bufferblock = _BufferBlock(blockid, starting_capacity=starting_capacity, owner=owner)
446 self._bufferblocks[bufferblock.blockid] = bufferblock
450 def dup_block(self, block, owner):
451 """Create a new bufferblock initialized with the content of an existing bufferblock.
454 the buffer block to copy.
457 ArvadosFile that owns the new block
460 new_blockid = "bufferblock%i" % len(self._bufferblocks)
461 bufferblock = block.clone(new_blockid, owner)
462 self._bufferblocks[bufferblock.blockid] = bufferblock
466 def is_bufferblock(self, locator):
467 return locator in self._bufferblocks
469 def _commit_bufferblock_worker(self):
470 """Background uploader thread."""
474 bufferblock = self._put_queue.get()
475 if bufferblock is None:
478 if self.copies is None:
479 loc = self._keep.put(bufferblock.buffer_view[0:bufferblock.write_pointer].tobytes())
481 loc = self._keep.put(bufferblock.buffer_view[0:bufferblock.write_pointer].tobytes(), copies=self.copies)
482 bufferblock.set_state(_BufferBlock.COMMITTED, loc)
484 except Exception as e:
485 bufferblock.set_state(_BufferBlock.ERROR, e)
487 if self._put_queue is not None:
488 self._put_queue.task_done()
490 def start_put_threads(self):
491 with self.threads_lock:
492 if self._put_threads is None:
493 # Start uploader threads.
495 # If we don't limit the Queue size, the upload queue can quickly
496 # grow to take up gigabytes of RAM if the writing process is
497 # generating data more quickly than it can be send to the Keep
500 # With two upload threads and a queue size of 2, this means up to 4
501 # blocks pending. If they are full 64 MiB blocks, that means up to
502 # 256 MiB of internal buffering, which is the same size as the
503 # default download block cache in KeepClient.
504 self._put_queue = Queue.Queue(maxsize=2)
506 self._put_threads = []
507 for i in xrange(0, self.num_put_threads):
508 thread = threading.Thread(target=self._commit_bufferblock_worker)
509 self._put_threads.append(thread)
513 def _block_prefetch_worker(self):
514 """The background downloader thread."""
517 b = self._prefetch_queue.get()
522 _logger.exception("Exception doing block prefetch")
525 def start_get_threads(self):
526 if self._prefetch_threads is None:
527 self._prefetch_queue = Queue.Queue()
528 self._prefetch_threads = []
529 for i in xrange(0, self.num_get_threads):
530 thread = threading.Thread(target=self._block_prefetch_worker)
531 self._prefetch_threads.append(thread)
537 def stop_threads(self):
538 """Shut down and wait for background upload and download threads to finish."""
540 if self._put_threads is not None:
541 for t in self._put_threads:
542 self._put_queue.put(None)
543 for t in self._put_threads:
545 self._put_threads = None
546 self._put_queue = None
548 if self._prefetch_threads is not None:
549 for t in self._prefetch_threads:
550 self._prefetch_queue.put(None)
551 for t in self._prefetch_threads:
553 self._prefetch_threads = None
554 self._prefetch_queue = None
559 def __exit__(self, exc_type, exc_value, traceback):
563 def repack_small_blocks(self, force=False, sync=False, closed_file_size=0):
564 """Packs small blocks together before uploading"""
565 self._pending_write_size += closed_file_size
567 # Check if there are enough small blocks for filling up one in full
568 if force or (self._pending_write_size >= config.KEEP_BLOCK_SIZE):
570 # Search blocks ready for getting packed together before being committed to Keep.
571 # A WRITABLE block always has an owner.
572 # A WRITABLE block with its owner.closed() implies that it's
573 # size is <= KEEP_BLOCK_SIZE/2.
574 small_blocks = [b for b in self._bufferblocks.values() if b.state() == _BufferBlock.WRITABLE and b.owner.closed()]
576 if len(small_blocks) <= 1:
577 # Not enough small blocks for repacking
580 # Update the pending write size count with its true value, just in case
581 # some small file was opened, written and closed several times.
582 self._pending_write_size = sum([b.size() for b in small_blocks])
583 if self._pending_write_size < config.KEEP_BLOCK_SIZE and not force:
586 new_bb = self._alloc_bufferblock()
587 while len(small_blocks) > 0 and (new_bb.write_pointer + small_blocks[0].size()) <= config.KEEP_BLOCK_SIZE:
588 bb = small_blocks.pop(0)
590 self._pending_write_size -= bb.size()
591 new_bb.append(bb.buffer_view[0:bb.write_pointer].tobytes())
592 arvfile.set_segments([Range(new_bb.blockid,
595 new_bb.write_pointer - bb.size())])
596 self._delete_bufferblock(bb.blockid)
597 self.commit_bufferblock(new_bb, sync=sync)
599 def commit_bufferblock(self, block, sync):
600 """Initiate a background upload of a bufferblock.
603 The block object to upload
606 If `sync` is True, upload the block synchronously.
607 If `sync` is False, upload the block asynchronously. This will
608 return immediately unless the upload queue is at capacity, in
609 which case it will wait on an upload queue slot.
613 # Mark the block as PENDING so to disallow any more appends.
614 block.set_state(_BufferBlock.PENDING)
615 except StateChangeError as e:
616 if e.state == _BufferBlock.PENDING:
618 block.wait_for_commit.wait()
621 if block.state() == _BufferBlock.COMMITTED:
623 elif block.state() == _BufferBlock.ERROR:
630 if self.copies is None:
631 loc = self._keep.put(block.buffer_view[0:block.write_pointer].tobytes())
633 loc = self._keep.put(block.buffer_view[0:block.write_pointer].tobytes(), copies=self.copies)
634 block.set_state(_BufferBlock.COMMITTED, loc)
635 except Exception as e:
636 block.set_state(_BufferBlock.ERROR, e)
639 self.start_put_threads()
640 self._put_queue.put(block)
643 def get_bufferblock(self, locator):
644 return self._bufferblocks.get(locator)
647 def delete_bufferblock(self, locator):
648 self._delete_bufferblock(locator)
650 def _delete_bufferblock(self, locator):
651 bb = self._bufferblocks[locator]
653 del self._bufferblocks[locator]
655 def get_block_contents(self, locator, num_retries, cache_only=False):
658 First checks to see if the locator is a BufferBlock and return that, if
659 not, passes the request through to KeepClient.get().
663 if locator in self._bufferblocks:
664 bufferblock = self._bufferblocks[locator]
665 if bufferblock.state() != _BufferBlock.COMMITTED:
666 return bufferblock.buffer_view[0:bufferblock.write_pointer].tobytes()
668 locator = bufferblock._locator
670 return self._keep.get_from_cache(locator)
672 return self._keep.get(locator, num_retries=num_retries)
674 def commit_all(self):
675 """Commit all outstanding buffer blocks.
677 This is a synchronous call, and will not return until all buffer blocks
678 are uploaded. Raises KeepWriteError() if any blocks failed to upload.
681 self.repack_small_blocks(force=True, sync=True)
684 items = self._bufferblocks.items()
687 if v.state() != _BufferBlock.COMMITTED and v.owner:
688 v.owner.flush(sync=False)
691 if self._put_queue is not None:
692 self._put_queue.join()
696 if v.state() == _BufferBlock.ERROR:
697 err.append((v.locator(), v.error))
699 raise KeepWriteError("Error writing some blocks", err, label="block")
702 # flush again with sync=True to remove committed bufferblocks from
705 v.owner.flush(sync=True)
707 def block_prefetch(self, locator):
708 """Initiate a background download of a block.
710 This assumes that the underlying KeepClient implements a block cache,
711 so repeated requests for the same block will not result in repeated
712 downloads (unless the block is evicted from the cache.) This method
717 if not self.prefetch_enabled:
720 if self._keep.get_from_cache(locator) is not None:
724 if locator in self._bufferblocks:
727 self.start_get_threads()
728 self._prefetch_queue.put(locator)
731 class ArvadosFile(object):
732 """Represent a file in a Collection.
734 ArvadosFile manages the underlying representation of a file in Keep as a
735 sequence of segments spanning a set of blocks, and implements random
738 This object may be accessed from multiple threads.
742 def __init__(self, parent, name, stream=[], segments=[]):
744 ArvadosFile constructor.
747 a list of Range objects representing a block stream
750 a list of Range objects representing segments
754 self._writers = set()
755 self._committed = False
757 self.lock = parent.root_collection().lock
759 self._add_segment(stream, s.locator, s.range_size)
760 self._current_bblock = None
763 return self.parent.writable()
766 def permission_expired(self, as_of_dt=None):
767 """Returns True if any of the segment's locators is expired"""
768 for r in self._segments:
769 if KeepLocator(r.locator).permission_expired(as_of_dt):
775 return copy.copy(self._segments)
778 def clone(self, new_parent, new_name):
779 """Make a copy of this file."""
780 cp = ArvadosFile(new_parent, new_name)
781 cp.replace_contents(self)
786 def replace_contents(self, other):
787 """Replace segments of this file with segments from another `ArvadosFile` object."""
791 for other_segment in other.segments():
792 new_loc = other_segment.locator
793 if other.parent._my_block_manager().is_bufferblock(other_segment.locator):
794 if other_segment.locator not in map_loc:
795 bufferblock = other.parent._my_block_manager().get_bufferblock(other_segment.locator)
796 if bufferblock.state() != _BufferBlock.WRITABLE:
797 map_loc[other_segment.locator] = bufferblock.locator()
799 map_loc[other_segment.locator] = self.parent._my_block_manager().dup_block(bufferblock, self).blockid
800 new_loc = map_loc[other_segment.locator]
802 self._segments.append(Range(new_loc, other_segment.range_start, other_segment.range_size, other_segment.segment_offset))
804 self._committed = False
806 def __eq__(self, other):
809 if not isinstance(other, ArvadosFile):
812 othersegs = other.segments()
814 if len(self._segments) != len(othersegs):
816 for i in xrange(0, len(othersegs)):
817 seg1 = self._segments[i]
822 if self.parent._my_block_manager().is_bufferblock(loc1):
823 loc1 = self.parent._my_block_manager().get_bufferblock(loc1).locator()
825 if other.parent._my_block_manager().is_bufferblock(loc2):
826 loc2 = other.parent._my_block_manager().get_bufferblock(loc2).locator()
828 if (KeepLocator(loc1).stripped() != KeepLocator(loc2).stripped() or
829 seg1.range_start != seg2.range_start or
830 seg1.range_size != seg2.range_size or
831 seg1.segment_offset != seg2.segment_offset):
836 def __ne__(self, other):
837 return not self.__eq__(other)
840 def set_segments(self, segs):
841 self._segments = segs
844 def set_committed(self):
845 """Set committed flag to True"""
846 self._committed = True
850 """Get whether this is committed or not."""
851 return self._committed
854 def add_writer(self, writer):
855 """Add an ArvadosFileWriter reference to the list of writers"""
856 if isinstance(writer, ArvadosFileWriter):
857 self._writers.add(writer)
860 def remove_writer(self, writer, flush):
862 Called from ArvadosFileWriter.close(). Remove a writer reference from the list
863 and do some block maintenance tasks.
865 self._writers.remove(writer)
867 if flush or self.size() > config.KEEP_BLOCK_SIZE / 2:
868 # File writer closed, not small enough for repacking
871 # All writers closed and size is adequate for repacking
872 self.parent._my_block_manager().repack_small_blocks(closed_file_size=self.size())
876 Get whether this is closed or not. When the writers list is empty, the file
877 is supposed to be closed.
879 return len(self._writers) == 0
883 def truncate(self, size):
884 """Shrink the size of the file.
886 If `size` is less than the size of the file, the file contents after
887 `size` will be discarded. If `size` is greater than the current size
888 of the file, an IOError will be raised.
891 if size < self.size():
893 for r in self._segments:
894 range_end = r.range_start+r.range_size
895 if r.range_start >= size:
896 # segment is past the trucate size, all done
898 elif size < range_end:
899 nr = Range(r.locator, r.range_start, size - r.range_start, 0)
900 nr.segment_offset = r.segment_offset
906 self._segments = new_segs
907 self._committed = False
908 elif size > self.size():
909 raise IOError(errno.EINVAL, "truncate() does not support extending the file size")
911 def readfrom(self, offset, size, num_retries, exact=False):
912 """Read up to `size` bytes from the file starting at `offset`.
915 If False (default), return less data than requested if the read
916 crosses a block boundary and the next block isn't cached. If True,
917 only return less data than requested when hitting EOF.
921 if size == 0 or offset >= self.size():
923 readsegs = locators_and_ranges(self._segments, offset, size)
924 prefetch = locators_and_ranges(self._segments, offset + size, config.KEEP_BLOCK_SIZE, limit=32)
929 block = self.parent._my_block_manager().get_block_contents(lr.locator, num_retries=num_retries, cache_only=(bool(data) and not exact))
931 blockview = memoryview(block)
932 data.append(blockview[lr.segment_offset:lr.segment_offset+lr.segment_size].tobytes())
938 if lr.locator not in locs:
939 self.parent._my_block_manager().block_prefetch(lr.locator)
944 def _repack_writes(self, num_retries):
945 """Test if the buffer block has more data than actual segments.
947 This happens when a buffered write over-writes a file range written in
948 a previous buffered write. Re-pack the buffer block for efficiency
949 and to avoid leaking information.
952 segs = self._segments
954 # Sum up the segments to get the total bytes of the file referencing
955 # into the buffer block.
956 bufferblock_segs = [s for s in segs if s.locator == self._current_bblock.blockid]
957 write_total = sum([s.range_size for s in bufferblock_segs])
959 if write_total < self._current_bblock.size():
960 # There is more data in the buffer block than is actually accounted for by segments, so
961 # re-pack into a new buffer by copying over to a new buffer block.
962 contents = self.parent._my_block_manager().get_block_contents(self._current_bblock.blockid, num_retries)
963 new_bb = self.parent._my_block_manager().alloc_bufferblock(self._current_bblock.blockid, starting_capacity=write_total, owner=self)
964 for t in bufferblock_segs:
965 new_bb.append(contents[t.segment_offset:t.segment_offset+t.range_size])
966 t.segment_offset = new_bb.size() - t.range_size
968 self._current_bblock = new_bb
972 def writeto(self, offset, data, num_retries):
973 """Write `data` to the file starting at `offset`.
975 This will update existing bytes and/or extend the size of the file as
982 if offset > self.size():
983 raise ArgumentError("Offset is past the end of the file")
985 if len(data) > config.KEEP_BLOCK_SIZE:
986 # Chunk it up into smaller writes
988 dataview = memoryview(data)
990 self.writeto(offset+n, dataview[n:n + config.KEEP_BLOCK_SIZE].tobytes(), num_retries)
991 n += config.KEEP_BLOCK_SIZE
994 self._committed = False
996 if self._current_bblock is None or self._current_bblock.state() != _BufferBlock.WRITABLE:
997 self._current_bblock = self.parent._my_block_manager().alloc_bufferblock(owner=self)
999 if (self._current_bblock.size() + len(data)) > config.KEEP_BLOCK_SIZE:
1000 self._repack_writes(num_retries)
1001 if (self._current_bblock.size() + len(data)) > config.KEEP_BLOCK_SIZE:
1002 self.parent._my_block_manager().commit_bufferblock(self._current_bblock, sync=False)
1003 self._current_bblock = self.parent._my_block_manager().alloc_bufferblock(owner=self)
1005 self._current_bblock.append(data)
1007 replace_range(self._segments, offset, len(data), self._current_bblock.blockid, self._current_bblock.write_pointer - len(data))
1009 self.parent.notify(WRITE, self.parent, self.name, (self, self))
1014 def flush(self, sync=True, num_retries=0):
1015 """Flush the current bufferblock to Keep.
1018 If True, commit block synchronously, wait until buffer block has been written.
1019 If False, commit block asynchronously, return immediately after putting block into
1022 if self.committed():
1025 if self._current_bblock and self._current_bblock.state() != _BufferBlock.COMMITTED:
1026 if self._current_bblock.state() == _BufferBlock.WRITABLE:
1027 self._repack_writes(num_retries)
1028 self.parent._my_block_manager().commit_bufferblock(self._current_bblock, sync=sync)
1032 for s in self._segments:
1033 bb = self.parent._my_block_manager().get_bufferblock(s.locator)
1035 if bb.state() != _BufferBlock.COMMITTED:
1036 self.parent._my_block_manager().commit_bufferblock(bb, sync=True)
1037 to_delete.add(s.locator)
1038 s.locator = bb.locator()
1040 self.parent._my_block_manager().delete_bufferblock(s)
1042 self.parent.notify(MOD, self.parent, self.name, (self, self))
1046 def add_segment(self, blocks, pos, size):
1047 """Add a segment to the end of the file.
1049 `pos` and `offset` reference a section of the stream described by
1050 `blocks` (a list of Range objects)
1053 self._add_segment(blocks, pos, size)
1055 def _add_segment(self, blocks, pos, size):
1056 """Internal implementation of add_segment."""
1057 self._committed = False
1058 for lr in locators_and_ranges(blocks, pos, size):
1059 last = self._segments[-1] if self._segments else Range(0, 0, 0, 0)
1060 r = Range(lr.locator, last.range_start+last.range_size, lr.segment_size, lr.segment_offset)
1061 self._segments.append(r)
1065 """Get the file size."""
1067 n = self._segments[-1]
1068 return n.range_start + n.range_size
1073 def manifest_text(self, stream_name=".", portable_locators=False, normalize=False):
1076 for segment in self.segments:
1077 loc = segment.locator
1078 if loc.startswith("bufferblock"):
1079 loc = self._bufferblocks[loc].calculate_locator()
1080 if portable_locators:
1081 loc = KeepLocator(loc).stripped()
1082 filestream.append(LocatorAndRange(loc, locator_block_size(loc),
1083 segment.segment_offset, segment.range_size))
1084 buf += ' '.join(normalize_stream(stream_name, {stream_name: filestream}))
1090 def _reparent(self, newparent, newname):
1091 self._committed = False
1092 self.flush(sync=True)
1093 self.parent.remove(self.name)
1094 self.parent = newparent
1096 self.lock = self.parent.root_collection().lock
1099 class ArvadosFileReader(ArvadosFileReaderBase):
1100 """Wraps ArvadosFile in a file-like object supporting reading only.
1102 Be aware that this class is NOT thread safe as there is no locking around
1103 updating file pointer.
1107 def __init__(self, arvadosfile, num_retries=None):
1108 super(ArvadosFileReader, self).__init__(arvadosfile.name, "r", num_retries=num_retries)
1109 self.arvadosfile = arvadosfile
1112 return self.arvadosfile.size()
1114 def stream_name(self):
1115 return self.arvadosfile.parent.stream_name()
1117 @_FileLikeObjectBase._before_close
1119 def read(self, size=None, num_retries=None):
1120 """Read up to `size` bytes from the file and return the result.
1122 Starts at the current file position. If `size` is None, read the
1123 entire remainder of the file.
1127 rd = self.arvadosfile.readfrom(self._filepos, config.KEEP_BLOCK_SIZE, num_retries)
1130 self._filepos += len(rd)
1131 rd = self.arvadosfile.readfrom(self._filepos, config.KEEP_BLOCK_SIZE, num_retries)
1132 return ''.join(data)
1134 data = self.arvadosfile.readfrom(self._filepos, size, num_retries, exact=True)
1135 self._filepos += len(data)
1138 @_FileLikeObjectBase._before_close
1140 def readfrom(self, offset, size, num_retries=None):
1141 """Read up to `size` bytes from the stream, starting at the specified file offset.
1143 This method does not change the file position.
1145 return self.arvadosfile.readfrom(offset, size, num_retries)
1151 class ArvadosFileWriter(ArvadosFileReader):
1152 """Wraps ArvadosFile in a file-like object supporting both reading and writing.
1154 Be aware that this class is NOT thread safe as there is no locking around
1155 updating file pointer.
1159 def __init__(self, arvadosfile, mode, num_retries=None):
1160 super(ArvadosFileWriter, self).__init__(arvadosfile, num_retries=num_retries)
1162 self.arvadosfile.add_writer(self)
1164 @_FileLikeObjectBase._before_close
1166 def write(self, data, num_retries=None):
1167 if self.mode[0] == "a":
1168 self.arvadosfile.writeto(self.size(), data, num_retries)
1170 self.arvadosfile.writeto(self._filepos, data, num_retries)
1171 self._filepos += len(data)
1174 @_FileLikeObjectBase._before_close
1176 def writelines(self, seq, num_retries=None):
1178 self.write(s, num_retries=num_retries)
1180 @_FileLikeObjectBase._before_close
1181 def truncate(self, size=None):
1183 size = self._filepos
1184 self.arvadosfile.truncate(size)
1185 if self._filepos > self.size():
1186 self._filepos = self.size()
1188 @_FileLikeObjectBase._before_close
1190 self.arvadosfile.flush()
1192 def close(self, flush=True):
1194 self.arvadosfile.remove_writer(self, flush)
1195 super(ArvadosFileWriter, self).close()