import os
import zlib
import bz2
-from .ranges import *
+from ._ranges import locators_and_ranges, replace_range, Range
from arvados.retry import retry_method
import config
import hashlib
import Queue
import copy
import errno
+from .errors import KeepWriteError, AssertionError
+from .keep import KeepLocator
+from _normalize_stream import normalize_stream
def split(path):
- """split(path) -> streamname, filename
+ """Separate the stream name and file name in a /-separated stream path and
+ return a tuple (stream_name, file_name).
- Separate the stream name and file name in a /-separated stream path.
If no stream name is available, assume '.'.
+
"""
try:
stream_name, file_name = path.rsplit('/', 1)
stream_name, file_name = '.', path
return stream_name, file_name
-class ArvadosFileBase(object):
+class _FileLikeObjectBase(object):
def __init__(self, name, mode):
self.name = name
self.mode = mode
@staticmethod
def _before_close(orig_func):
@functools.wraps(orig_func)
- def wrapper(self, *args, **kwargs):
+ def before_close_wrapper(self, *args, **kwargs):
if self.closed:
raise ValueError("I/O operation on closed stream file")
return orig_func(self, *args, **kwargs)
- return wrapper
+ return before_close_wrapper
def __enter__(self):
return self
self.closed = True
-class ArvadosFileReaderBase(ArvadosFileBase):
+class ArvadosFileReaderBase(_FileLikeObjectBase):
class _NameAttribute(str):
# The Python file API provides a plain .name attribute.
# Older SDK provided a name() method.
def decompressed_name(self):
return re.sub('\.(bz2|gz)$', '', self.name)
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
def seek(self, pos, whence=os.SEEK_CUR):
if whence == os.SEEK_CUR:
pos += self._filepos
def tell(self):
return self._filepos
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readall(self, size=2**20, num_retries=None):
while True:
break
yield data
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readline(self, size=float('inf'), num_retries=None):
cache_pos, cache_data = self._readline_cache
self._readline_cache = (self.tell(), data[nextline_index:])
return data[:nextline_index]
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def decompress(self, decompress, size, num_retries=None):
for segment in self.readall(size, num_retries):
if data:
yield data
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readall_decompressed(self, size=2**20, num_retries=None):
self.seek(0)
else:
return self.readall(size, num_retries=num_retries)
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readlines(self, sizehint=float('inf'), num_retries=None):
data = []
n = self.segments[-1]
return n.range_start + n.range_size
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def read(self, size, num_retries=None):
"""Read up to 'size' bytes from the stream, starting at the current file position"""
self._filepos += len(data)
return data
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readfrom(self, start, size, num_retries=None):
"""Read up to 'size' bytes from the stream, starting at 'start'"""
return ''.join(data)
def as_manifest(self):
- from stream import normalize_stream
segs = []
for r in self.segments:
segs.extend(self._stream.locators_and_ranges(r.locator, r.range_size))
return " ".join(normalize_stream(".", {self.name: segs})) + "\n"
-class BufferBlock(object):
- """
- A BufferBlock is a stand-in for a Keep block that is in the process of being
- written. Writers can append to it, get the size, and compute the Keep locator.
+def synchronized(orig_func):
+ @functools.wraps(orig_func)
+ def synchronized_wrapper(self, *args, **kwargs):
+ with self.lock:
+ return orig_func(self, *args, **kwargs)
+ return synchronized_wrapper
+class _BufferBlock(object):
+ """A BufferBlock is a stand-in for a Keep block that is in the process of being
+ written.
+
+ Writers can append to it, get the size, and compute the Keep locator.
There are three valid states:
WRITABLE
released, fetching the block will fetch it via keep client (since we
discarded the internal copy), and identifiers referring to the BufferBlock
can be replaced with the block locator.
+
"""
+
WRITABLE = 0
PENDING = 1
COMMITTED = 2
:owner:
ArvadosFile that owns this block
+
"""
self.blockid = blockid
self.buffer_block = bytearray(starting_capacity)
self.buffer_view = memoryview(self.buffer_block)
self.write_pointer = 0
- self.state = BufferBlock.WRITABLE
+ self._state = _BufferBlock.WRITABLE
self._locator = None
self.owner = owner
+ self.lock = threading.Lock()
+ @synchronized
def append(self, data):
+ """Append some data to the buffer.
+
+ Only valid if the block is in WRITABLE state. Implements an expanding
+ buffer, doubling capacity as needed to accomdate all the data.
+
"""
- Append some data to the buffer. Only valid if the block is in WRITABLE
- state. Implements an expanding buffer, doubling capacity as needed to
- accomdate all the data.
- """
- if self.state == BufferBlock.WRITABLE:
+ if self._state == _BufferBlock.WRITABLE:
while (self.write_pointer+len(data)) > len(self.buffer_block):
new_buffer_block = bytearray(len(self.buffer_block) * 2)
new_buffer_block[0:self.write_pointer] = self.buffer_block[0:self.write_pointer]
else:
raise AssertionError("Buffer block is not writable")
+ @synchronized
+ def set_state(self, nextstate, loc=None):
+ if ((self._state == _BufferBlock.WRITABLE and nextstate == _BufferBlock.PENDING) or
+ (self._state == _BufferBlock.PENDING and nextstate == _BufferBlock.COMMITTED)):
+ self._state = nextstate
+ if self._state == _BufferBlock.COMMITTED:
+ self._locator = loc
+ self.buffer_view = None
+ self.buffer_block = None
+ else:
+ raise AssertionError("Invalid state change from %s to %s" % (self.state, state))
+
+ @synchronized
+ def state(self):
+ return self._state
+
def size(self):
- """Amount of data written to the buffer"""
+ """The amount of data written to the buffer."""
return self.write_pointer
+ @synchronized
def locator(self):
"""The Keep locator for this buffer's contents."""
if self._locator is None:
return self._locator
-class AsyncKeepWriteErrors(Exception):
- """
- Roll up one or more Keep write exceptions (generated by background
- threads) into a single one.
- """
- def __init__(self, errors):
- self.errors = errors
-
- def __repr__(self):
- return "\n".join(self.errors)
-
-def synchronized(orig_func):
- @functools.wraps(orig_func)
- def wrapper(self, *args, **kwargs):
- with self.lock:
- return orig_func(self, *args, **kwargs)
- return wrapper
-
class NoopLock(object):
def __enter__(self):
return self
def must_be_writable(orig_func):
@functools.wraps(orig_func)
- def wrapper(self, *args, **kwargs):
+ def must_be_writable_wrapper(self, *args, **kwargs):
if self.sync_mode() == SYNC_READONLY:
raise IOError((errno.EROFS, "Collection is read only"))
return orig_func(self, *args, **kwargs)
- return wrapper
+ return must_be_writable_wrapper
-class BlockManager(object):
- """
- BlockManager handles buffer blocks, background block uploads, and
- background block prefetch for a Collection of ArvadosFiles.
+class _BlockManager(object):
+ """BlockManager handles buffer blocks, background block uploads, and background
+ block prefetch for a Collection of ArvadosFiles.
+
"""
def __init__(self, keep):
"""keep: KeepClient object to use"""
@synchronized
def alloc_bufferblock(self, blockid=None, starting_capacity=2**14, owner=None):
- """
- Allocate a new, empty bufferblock in WRITABLE state and return it.
+ """Allocate a new, empty bufferblock in WRITABLE state and return it.
:blockid:
optional block identifier, otherwise one will be automatically assigned
:owner:
ArvadosFile that owns this block
+
"""
if blockid is None:
blockid = "bufferblock%i" % len(self._bufferblocks)
- bb = BufferBlock(blockid, starting_capacity=starting_capacity, owner=owner)
- self._bufferblocks[bb.blockid] = bb
- return bb
+ bufferblock = _BufferBlock(blockid, starting_capacity=starting_capacity, owner=owner)
+ self._bufferblocks[bufferblock.blockid] = bufferblock
+ return bufferblock
@synchronized
- def dup_block(self, blockid, owner):
- """
- Create a new bufferblock in WRITABLE state, initialized with the content of an existing bufferblock.
+ def dup_block(self, block, owner):
+ """Create a new bufferblock in WRITABLE state, initialized with the content of
+ an existing bufferblock.
- :blockid:
- the block to copy. May be an existing buffer block id.
+ :block:
+ the buffer block to copy.
:owner:
ArvadosFile that owns the new block
+
"""
new_blockid = "bufferblock%i" % len(self._bufferblocks)
- block = self._bufferblocks[blockid]
- bb = BufferBlock(new_blockid, len(block), owner)
- bb.append(block)
- self._bufferblocks[bb.blockid] = bb
- return bb
+ with block.lock:
+ if block._state == _BufferBlock.COMMITTED:
+ raise AssertionError("Can only duplicate a writable or pending buffer block")
+
+ bufferblock = _BufferBlock(new_blockid, block.size(), owner)
+ bufferblock.append(block.buffer_view[0:block.size()])
+ self._bufferblocks[bufferblock.blockid] = bufferblock
+ return bufferblock
@synchronized
- def is_bufferblock(self, id):
- return id in self._bufferblocks
+ def is_bufferblock(self, locator):
+ return locator in self._bufferblocks
@synchronized
def stop_threads(self):
- """
- Shut down and wait for background upload and download threads to finish.
- """
+ """Shut down and wait for background upload and download threads to finish."""
+
if self._put_threads is not None:
for t in self._put_threads:
self._put_queue.put(None)
self._prefetch_queue = None
def commit_bufferblock(self, block):
+ """Initiate a background upload of a bufferblock.
+
+ This will block if the upload queue is at capacity, otherwise it will
+ return immediately.
+
"""
- Initiate a background upload of a bufferblock. This will block if the
- upload queue is at capacity, otherwise it will return immediately.
- """
- def worker(self):
- """
- Background uploader thread.
- """
+ def commit_bufferblock_worker(self):
+ """Background uploader thread."""
+
while True:
try:
- b = self._put_queue.get()
- if b is None:
+ bufferblock = self._put_queue.get()
+ if bufferblock is None:
return
- b._locator = self._keep.put(b.buffer_view[0:b.write_pointer].tobytes())
- b.state = BufferBlock.COMMITTED
- b.buffer_view = None
- b.buffer_block = None
+ loc = self._keep.put(bufferblock.buffer_view[0:bufferblock.write_pointer].tobytes())
+ bufferblock.set_state(_BufferBlock.COMMITTED, loc)
+
except Exception as e:
print e
- self._put_errors.put(e)
+ self._put_errors.put((bufferblock.locator(), e))
finally:
if self._put_queue is not None:
self._put_queue.task_done()
self._put_threads = []
for i in xrange(0, self.num_put_threads):
- t = threading.Thread(target=worker, args=(self,))
- self._put_threads.append(t)
- t.daemon = True
- t.start()
+ thread = threading.Thread(target=commit_bufferblock_worker, args=(self,))
+ self._put_threads.append(thread)
+ thread.daemon = True
+ thread.start()
# Mark the block as PENDING so to disallow any more appends.
- block.state = BufferBlock.PENDING
+ block.set_state(_BufferBlock.PENDING)
self._put_queue.put(block)
- def get_block(self, locator, num_retries, cache_only=False):
- """
- Fetch a block. First checks to see if the locator is a BufferBlock and
- return that, if not, passes the request through to KeepClient.get().
+ @synchronized
+ def get_bufferblock(self, locator):
+ return self._bufferblocks.get(locator)
+
+ def get_block_contents(self, locator, num_retries, cache_only=False):
+ """Fetch a block.
+
+ First checks to see if the locator is a BufferBlock and return that, if
+ not, passes the request through to KeepClient.get().
+
"""
with self.lock:
if locator in self._bufferblocks:
- bb = self._bufferblocks[locator]
- if bb.state != BufferBlock.COMMITTED:
- return bb.buffer_view[0:bb.write_pointer].tobytes()
+ bufferblock = self._bufferblocks[locator]
+ if bufferblock.state() != _BufferBlock.COMMITTED:
+ return bufferblock.buffer_view[0:bufferblock.write_pointer].tobytes()
else:
- locator = bb._locator
- return self._keep.get(locator, num_retries=num_retries, cache_only=cache_only)
+ locator = bufferblock._locator
+ if cache_only:
+ return self._keep.get_from_cache(locator)
+ else:
+ return self._keep.get(locator, num_retries=num_retries)
def commit_all(self):
- """
- Commit all outstanding buffer blocks. Unlike commit_bufferblock(), this
- is a synchronous call, and will not return until all buffer blocks are
- uploaded. Raises AsyncKeepWriteErrors() if any blocks failed to
- upload.
+ """Commit all outstanding buffer blocks.
+
+ Unlike commit_bufferblock(), this is a synchronous call, and will not
+ return until all buffer blocks are uploaded. Raises
+ KeepWriteError() if any blocks failed to upload.
+
"""
with self.lock:
items = self._bufferblocks.items()
for k,v in items:
- if v.state == BufferBlock.WRITABLE:
+ if v.state() == _BufferBlock.WRITABLE:
self.commit_bufferblock(v)
with self.lock:
if self._put_queue is not None:
self._put_queue.join()
+
if not self._put_errors.empty():
- e = []
+ err = []
try:
while True:
- e.append(self._put_errors.get(False))
+ err.append(self._put_errors.get(False))
except Queue.Empty:
pass
- raise AsyncKeepWriteErrors(e)
+ raise KeepWriteError("Error writing some blocks", err)
def block_prefetch(self, locator):
- """
- Initiate a background download of a block. This assumes that the
- underlying KeepClient implements a block cache, so repeated requests
- for the same block will not result in repeated downloads (unless the
- block is evicted from the cache.) This method does not block.
+ """Initiate a background download of a block.
+
+ This assumes that the underlying KeepClient implements a block cache,
+ so repeated requests for the same block will not result in repeated
+ downloads (unless the block is evicted from the cache.) This method
+ does not block.
+
"""
if not self.prefetch_enabled:
return
- def worker(self):
- """Background downloader thread."""
+ def block_prefetch_worker(self):
+ """The background downloader thread."""
while True:
try:
b = self._prefetch_queue.get()
self._prefetch_queue = Queue.Queue()
self._prefetch_threads = []
for i in xrange(0, self.num_get_threads):
- t = threading.Thread(target=worker, args=(self,))
- self._prefetch_threads.append(t)
- t.daemon = True
- t.start()
+ thread = threading.Thread(target=block_prefetch_worker, args=(self,))
+ self._prefetch_threads.append(thread)
+ thread.daemon = True
+ thread.start()
self._prefetch_queue.put(locator)
class ArvadosFile(object):
- """
- ArvadosFile manages the underlying representation of a file in Keep as a sequence of
- segments spanning a set of blocks, and implements random read/write access.
+ """ArvadosFile manages the underlying representation of a file in Keep as a
+ sequence of segments spanning a set of blocks, and implements random
+ read/write access.
+
+ This object may be accessed from multiple threads.
+
"""
def __init__(self, parent, stream=[], segments=[]):
"""
+ ArvadosFile constructor.
+
:stream:
a list of Range objects representing a block stream
self.parent = parent
self._modified = True
self._segments = []
- self.lock = parent._root_lock()
+ self.lock = parent.root_collection().lock
for s in segments:
self._add_segment(stream, s.locator, s.range_size)
self._current_bblock = None
def clone(self, new_parent):
"""Make a copy of this file."""
cp = ArvadosFile(new_parent)
-
- map_loc = {}
- for r in self._segments:
- new_loc = r.locator
- if self.parent._my_block_manager().is_bufferblock(r.locator):
- if r.locator not in map_loc:
- map_loc[r.locator] = self.parent._my_block_manager().dup_block(r.locator, cp).blockid
- new_loc = map_loc[r.locator]
-
- cp._segments.append(Range(new_loc, r.range_start, r.range_size, r.segment_offset))
-
+ cp.replace_contents(self)
return cp
@must_be_writable
@synchronized
def replace_contents(self, other):
"""Replace segments of this file with segments from another `ArvadosFile` object."""
- self._segments = other.segments()
+
+ map_loc = {}
+ self._segments = []
+ for other_segment in other.segments():
+ new_loc = other_segment.locator
+ if other.parent._my_block_manager().is_bufferblock(other_segment.locator):
+ if other_segment.locator not in map_loc:
+ bufferblock = other.parent._my_block_manager().get_bufferblock(other_segment.locator)
+ if bufferblock.state() != _BufferBlock.WRITABLE:
+ map_loc[other_segment.locator] = bufferblock.locator()
+ else:
+ map_loc[other_segment.locator] = self.parent._my_block_manager().dup_block(bufferblock, self).blockid
+ new_loc = map_loc[other_segment.locator]
+
+ self._segments.append(Range(new_loc, other_segment.range_start, other_segment.range_size, other_segment.segment_offset))
+
self._modified = True
def __eq__(self, other):
if not isinstance(other, ArvadosFile):
return False
- s = other.segments()
+ othersegs = other.segments()
with self.lock:
- return self._segments == s
+ if len(self._segments) != len(othersegs):
+ return False
+ for i in xrange(0, len(othersegs)):
+ seg1 = self._segments[i]
+ seg2 = othersegs[i]
+ loc1 = seg1.locator
+ loc2 = seg2.locator
+
+ if self.parent._my_block_manager().is_bufferblock(loc1):
+ loc1 = self.parent._my_block_manager().get_bufferblock(loc1).locator()
+
+ if other.parent._my_block_manager().is_bufferblock(loc2):
+ loc2 = other.parent._my_block_manager().get_bufferblock(loc2).locator()
+
+ if (KeepLocator(loc1).stripped() != KeepLocator(loc2).stripped() or
+ seg1.range_start != seg2.range_start or
+ seg1.range_size != seg2.range_size or
+ seg1.segment_offset != seg2.segment_offset):
+ return False
+
+ return True
def __ne__(self, other):
return not self.__eq__(other)
@must_be_writable
@synchronized
def truncate(self, size):
- """
- Adjust the size of the file. If `size` is less than the size of the file,
- the file contents after `size` will be discarded. If `size` is greater
- than the current size of the file, an IOError will be raised.
+ """Shrink the size of the file.
+
+ If `size` is less than the size of the file, the file contents after
+ `size` will be discarded. If `size` is greater than the current size
+ of the file, an IOError will be raised.
+
"""
if size < self.size():
new_segs = []
elif size > self.size():
raise IOError("truncate() does not support extending the file size")
- @synchronized
def readfrom(self, offset, size, num_retries):
- """
- read upto `size` bytes from the file starting at `offset`.
- """
- if size == 0 or offset >= self.size():
- return ''
- data = []
+ """Read upto `size` bytes from the file starting at `offset`."""
+
+ with self.lock:
+ if size == 0 or offset >= self.size():
+ return ''
+ prefetch = locators_and_ranges(self._segments, offset, size + config.KEEP_BLOCK_SIZE)
+ readsegs = locators_and_ranges(self._segments, offset, size)
- for lr in locators_and_ranges(self._segments, offset, size + config.KEEP_BLOCK_SIZE):
+ for lr in prefetch:
self.parent._my_block_manager().block_prefetch(lr.locator)
- for lr in locators_and_ranges(self._segments, offset, size):
- d = self.parent._my_block_manager().get_block(lr.locator, num_retries=num_retries, cache_only=bool(data))
- if d:
- data.append(d[lr.segment_offset:lr.segment_offset+lr.segment_size])
+ data = []
+ for lr in readsegs:
+ block = self.parent._my_block_manager().get_block_contents(lr.locator, num_retries=num_retries, cache_only=bool(data))
+ if block:
+ data.append(block[lr.segment_offset:lr.segment_offset+lr.segment_size])
else:
break
return ''.join(data)
def _repack_writes(self):
- """
- Test if the buffer block has more data than is referenced by actual segments
- (this happens when a buffered write over-writes a file range written in
- a previous buffered write). Re-pack the buffer block for efficiency
+ """Test if the buffer block has more data than is referenced by actual
+ segments.
+
+ This happens when a buffered write over-writes a file range written in
+ a previous buffered write. Re-pack the buffer block for efficiency
and to avoid leaking information.
+
"""
segs = self._segments
@must_be_writable
@synchronized
def writeto(self, offset, data, num_retries):
- """
- Write `data` to the file starting at `offset`. This will update
- existing bytes and/or extend the size of the file as necessary.
+ """Write `data` to the file starting at `offset`.
+
+ This will update existing bytes and/or extend the size of the file as
+ necessary.
+
"""
if len(data) == 0:
return
self._modified = True
- if self._current_bblock is None or self._current_bblock.state != BufferBlock.WRITABLE:
+ if self._current_bblock is None or self._current_bblock.state() != _BufferBlock.WRITABLE:
self._current_bblock = self.parent._my_block_manager().alloc_bufferblock(owner=self)
if (self._current_bblock.size() + len(data)) > config.KEEP_BLOCK_SIZE:
@must_be_writable
@synchronized
def add_segment(self, blocks, pos, size):
- """
- Add a segment to the end of the file, with `pos` and `offset` referencing a
+ """Add a segment to the end of the file, with `pos` and `offset` referencing a
section of the stream described by `blocks` (a list of Range objects)
+
"""
self._add_segment(blocks, pos, size)
def _add_segment(self, blocks, pos, size):
- """
- (Internal version.)
- """
+ """Internal implementation of add_segment."""
self._modified = True
for lr in locators_and_ranges(blocks, pos, size):
last = self._segments[-1] if self._segments else Range(0, 0, 0)
@synchronized
def size(self):
- """Get the file size"""
+ """Get the file size."""
if self._segments:
n = self._segments[-1]
return n.range_start + n.range_size
else:
return 0
+
+ @synchronized
+ def manifest_text(self, stream_name=".", portable_locators=False, normalize=False):
+ buf = ""
+ item = self
+ filestream = []
+ for segment in item.segments:
+ loc = segment.locator
+ if loc.startswith("bufferblock"):
+ loc = item._bufferblocks[loc].calculate_locator()
+ if portable_locators:
+ loc = KeepLocator(loc).stripped()
+ filestream.append(LocatorAndRange(loc, locator_block_size(loc),
+ segment.segment_offset, segment.range_size))
+ stream[stream_name] = filestream
+ buf += ' '.join(normalize_stream(stream_name, stream))
+ buf += "\n"
+ return buf
+
+
class ArvadosFileReader(ArvadosFileReaderBase):
+ """Wraps ArvadosFile in a file-like object supporting reading only.
+
+ Be aware that this class is NOT thread safe as there is no locking around
+ updating file pointer.
+
+ """
+
def __init__(self, arvadosfile, name, mode="r", num_retries=None):
super(ArvadosFileReader, self).__init__(name, mode, num_retries=num_retries)
self.arvadosfile = arvadosfile
def size(self):
return self.arvadosfile.size()
- @ArvadosFileBase._before_close
+ def stream_name(self):
+ return self.arvadosfile.parent.stream_name()
+
+ @_FileLikeObjectBase._before_close
@retry_method
def read(self, size, num_retries=None):
- """Read up to `size` bytes from the stream, starting at the current file position"""
- data = self.arvadosfile.readfrom(self._filepos, size, num_retries=num_retries)
+ """Read up to `size` bytes from the stream, starting at the current file position."""
+ data = self.arvadosfile.readfrom(self._filepos, size, num_retries)
self._filepos += len(data)
return data
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def readfrom(self, offset, size, num_retries=None):
- """Read up to `size` bytes from the stream, starting at the current file position"""
+ """Read up to `size` bytes from the stream, starting at the current file position."""
return self.arvadosfile.readfrom(offset, size, num_retries)
def flush(self):
class ArvadosFileWriter(ArvadosFileReader):
+ """Wraps ArvadosFile in a file-like object supporting both reading and writing.
+
+ Be aware that this class is NOT thread safe as there is no locking around
+ updating file pointer.
+
+ """
+
def __init__(self, arvadosfile, name, mode, num_retries=None):
super(ArvadosFileWriter, self).__init__(arvadosfile, name, mode, num_retries=num_retries)
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def write(self, data, num_retries=None):
if self.mode[0] == "a":
- self.arvadosfile.writeto(self.size(), data)
+ self.arvadosfile.writeto(self.size(), data, num_retries)
else:
self.arvadosfile.writeto(self._filepos, data, num_retries)
self._filepos += len(data)
- @ArvadosFileBase._before_close
+ @_FileLikeObjectBase._before_close
@retry_method
def writelines(self, seq, num_retries=None):
for s in seq:
- self.write(s)
+ self.write(s, num_retries)
def truncate(self, size=None):
if size is None:
self.arvadosfile.truncate(size)
if self._filepos > self.size():
self._filepos = self.size()
+
+ def close(self):
+ if self.arvadosfile.parent.sync_mode() == SYNC_LIVE:
+ self.arvadosfile.parent.root_collection().save()
projects / arvados.git / blobdiff