def write(self, data): """Write data to the object.""" if 'r' in self.mode: raise IOSError(EPERM, "File is opened for read") # large file support if self.split_size: # data can be of any size, so we need to split it in split_size chunks offs = 0 while offs < len(data): if self.part_size + len(data) - offs > self.split_size: current_size = self.split_size-self.part_size logging.debug("data is to large (%r), using %s" % (len(data), current_size)) else: current_size = len(data)-offs self.part_size += current_size if not self.obj: self.obj = ChunkObject(self.conn, self.container, self.part_name, content_type=self.content_type) self.obj.send_chunk(data[offs:offs+current_size]) offs += current_size if self.part_size == self.split_size: logging.debug("current size is %r, split_file is %r" % (self.part_size, self.split_size)) self.obj.finish_chunk() # this obj is not valid anymore, will create a new one if a new part is required self.obj = None # make it the first part if self.part == 0: self._start_copy_task() self.part_size = 0 self.part += 1 else: self.obj.send_chunk(data)
def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.part_size = 0 self.part = 0 self.headers = dict() self.content_type = mimetypes.guess_type(self.name)[0] self.pending_copy_task = None self.obj = None # this is only used by `seek`, so we delay the HEAD request until is required self.size = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') logging.debug("ObjectStorageFD object: %r (mode: %r)" % (obj, mode)) if 'r' in self.mode: logging.debug("read fd %r" % self.name) else: # write logging.debug("write fd %r" % self.name) self.obj = ChunkObject(self.conn, self.container, self.name, content_type=self.content_type)
def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.headers = dict() self.obj = None # this is only used by `seek`, so we delay the HEAD request until is required self.size = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') logging.debug("ObjectStorageFD object: %r (mode: %r)" % (obj, mode)) if 'r' in self.mode: logging.debug("read fd %r" % self.name) else: # write logging.debug("write fd %r" % self.name) self.obj = ChunkObject(self.conn, self.container, self.name, content_type=mimetypes.guess_type(self.name)[0])
def __init__(self, connection, container, obj, mode): self.conn = connection self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.part_size = 0 self.part = 0 self.part_collision = 0 self.headers = dict() self.content_type = mimetypes.guess_type(self.name)[0] self.pending_copy_task = None self.obj = None # this is only used by `seek`, so we delay the HEAD request until is required self.size = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') logging.debug("ObjectStorageFD object: %r (mode: %r)" % (obj, mode)) if 'r' in self.mode: logging.debug("read fd %r" % self.name) else: # write logging.debug("write fd %r" % self.name) # check for collisions in case this is a multi-part file if self.split_size: self._find_collisions() self.obj = ChunkObject(self.conn, self.container, self.name, content_type=self.content_type)
class CloudFilesFD(object): '''Acts like a file() object, but attached to a cloud files object''' def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.stream = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object requred') self.container = self.cffs._get_container(self.container) if 'r' in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: #write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk() def write(self, data): '''Write data to the object''' if 'r' in self.mode: raise IOSError(EPERM, "Can't write to stream opened for read") self.obj.send_chunk(data) def close(self): '''Close the object and finish the data transfer''' if 'r' in self.mode: return self.obj.finish_chunk() def read(self, size=65536): '''Read data from the object. We can use just one request because 'seek' is not supported. NB: It uses the size passed into the first call for all subsequent calls''' if not self.stream: self.stream = self.obj.stream(size) logging.debug("read size=%r, total_size=%r, obj.size=%r" % (size, self.total_size, self.obj.size)) try: buff = self.stream.next() self.total_size += len(buff) except StopIteration: return "" else: return buff def seek(self, *kargs, **kwargs): '''Seek in the object: FIXME doesn't work and raises an error''' logging.debug("seek args=%s, kargs=%s" % (str(kargs), str(kwargs))) raise IOSError(EPERM, "Seek not implemented")
class CloudFilesFD(object): """Acts like a file() object, but attached to a cloud files object""" def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 if not all([container, obj]): self.closed = True raise IOSError(EPERM, "Container and object requred") self.container = self.cffs._get_container(self.container) if "r" in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: # write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk() def write(self, data): """Write data to the object""" if "r" in self.mode: raise IOSError(EPERM, "Can't write to stream opened for read") self.obj.send_chunk(data) def close(self): """Close the object and finish the data transfer""" if "r" in self.mode: return self.obj.finish_chunk() def read(self, size=65536): """Read data from the object""" logging.debug("read size=%r, total_size=%r, obj.size=%r" % (size, self.total_size, self.obj.size)) readsize = size if (self.total_size + size) > self.obj.size: readsize = self.obj.size - self.total_size if self.total_size >= self.obj.size: return "" else: offset = self.total_size self.total_size += size return self.obj.read(size=readsize, offset=offset) def seek(self, *kargs, **kwargs): """Seek in the object: FIXME doesn't work and raises an error""" logging.debug("seek args=%s, kargs=%s" % (str(kargs), str(kwargs))) raise IOSError(EPERM, "Seek not implemented")
def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.stream = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') self.container = self.cffs._get_container(self.container) if 'r' in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: #write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk()
def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 if not all([container, obj]): self.closed = True raise IOSError(EPERM, "Container and object requred") self.container = self.cffs._get_container(self.container) if "r" in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: # write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk()
class ObjectStorageFD(object): """File alike object attached to the Object Storage.""" split_size = 0 def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.part_size = 0 self.part = 0 self.headers = dict() self.content_type = mimetypes.guess_type(self.name)[0] self.pending_copy_task = None self.obj = None # this is only used by `seek`, so we delay the HEAD request until is required self.size = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') logging.debug("ObjectStorageFD object: %r (mode: %r)" % (obj, mode)) if 'r' in self.mode: logging.debug("read fd %r" % self.name) else: # write logging.debug("write fd %r" % self.name) self.obj = ChunkObject(self.conn, self.container, self.name, content_type=self.content_type) @property def part_base_name(self): return u"%s.part" % self.name @property def part_name(self): return u"%s/%.6d" % (self.part_base_name, self.part) @property def conn(self): """Connection to the storage.""" return self.cffs.conn def _start_copy_task(self): """ Copy the first part of a multi-part file to its final location and create the manifest file. This happens in the background, pending_copy_task must be cleaned up at the end. """ def copy_task(conn, container, name, part_name, part_base_name): # open a new connection conn = ProxyConnection(None, preauthurl=conn.url, preauthtoken=conn.token) headers = { 'x-copy-from': "/%s/%s" % (container, name) } logging.debug("copying first part %r/%r, %r" % (container, part_name, headers)) try: conn.put_object(container, part_name, headers=headers, contents=None) except ClientException as ex: logging.error("Failed to copy %s: %s" % (name, ex.http_reason)) sys.exit(1) # setup the manifest headers = { 'x-object-manifest': "%s/%s" % (container, part_base_name) } logging.debug("creating manifest %r/%r, %r" % (container, name, headers)) try: conn.put_object(container, name, headers=headers, contents=None) except ClientException as ex: logging.error("Failed to store the manifest %s: %s" % (name, ex.http_reason)) sys.exit(1) logging.debug("copy task done") self.pending_copy_task = multiprocessing.Process(target=copy_task, args=(self.conn, self.container, self.name, self.part_name, self.part_base_name, ), ) self.pending_copy_task.start() @translate_objectstorage_error def write(self, data): """Write data to the object.""" if 'r' in self.mode: raise IOSError(EPERM, "File is opened for read") # large file support if self.split_size: # data can be of any size, so we need to split it in split_size chunks offs = 0 while offs < len(data): if self.part_size + len(data) - offs > self.split_size: current_size = self.split_size-self.part_size logging.debug("data is to large (%r), using %s" % (len(data), current_size)) else: current_size = len(data)-offs self.part_size += current_size if not self.obj: self.obj = ChunkObject(self.conn, self.container, self.part_name, content_type=self.content_type) self.obj.send_chunk(data[offs:offs+current_size]) offs += current_size if self.part_size == self.split_size: logging.debug("current size is %r, split_file is %r" % (self.part_size, self.split_size)) self.obj.finish_chunk() # this obj is not valid anymore, will create a new one if a new part is required self.obj = None # make it the first part if self.part == 0: self._start_copy_task() self.part_size = 0 self.part += 1 else: self.obj.send_chunk(data) @translate_objectstorage_error def close(self): """Close the object and finish the data transfer.""" if 'r' in self.mode: return if self.pending_copy_task: logging.debug("waiting for a pending copy task...") self.pending_copy_task.join() logging.debug("wait is over") if self.pending_copy_task.exitcode != 0: raise IOSError(EIO, 'Failed to store the file') if self.obj is not None: self.obj.finish_chunk() def read(self, size=65536): """ Read data from the object. We can use just one request because 'seek' is not fully supported. NB: It uses the size passed into the first call for all subsequent calls. """ if self.obj is None: if self.total_size > 0: self.conn.range_from = self.total_size # we need to open a new connection to inject the `Range` header if self.conn.http_conn: self.conn.http_conn[1].close() self.conn.http_conn = None _, self.obj = self.conn.get_object(self.container, self.name, resp_chunk_size=size) logging.debug("read size=%r, total_size=%r (range_from: %s)" % (size, self.total_size, self.conn.range_from)) try: buff = self.obj.next() self.total_size += len(buff) except StopIteration: return "" else: return buff def seek(self, offset, whence=None): """ Seek in the object. It's supported only for read operations because of object storage limitations. """ logging.debug("seek offset=%s, whence=%s" % (str(offset), str(whence))) if 'r' in self.mode: if self.size is None: meta = self.conn.head_object(self.container, self.name) try: self.size = int(meta["content-length"]) except ValueError: raise IOSError(EPERM, "Invalid file size") if not whence: offs = offset elif whence == 1: offs = self.total_size + offset elif whence == 2: offs = self.size - offset else: raise IOSError(EPERM, "Invalid file offset") if offs < 0 or offs > self.size: raise IOSError(EPERM, "Invalid file offset") # we need to start over after a seek call if self.obj is not None: del self.obj # GC the generator self.obj = None self.total_size = offs else: raise IOSError(EPERM, "Seek not available for write operations")
class ObjectStorageFD(object): """File alike object attached to the Object Storage.""" def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.headers = dict() self.obj = None # this is only used by `seek`, so we delay the HEAD request until is required self.size = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') logging.debug("ObjectStorageFD object: %r (mode: %r)" % (obj, mode)) if 'r' in self.mode: logging.debug("read fd %r" % self.name) else: # write logging.debug("write fd %r" % self.name) self.obj = ChunkObject(self.conn, self.container, self.name, content_type=mimetypes.guess_type(self.name)[0]) @property def conn(self): """Connection to the storage.""" return self.cffs.conn def write(self, data): """Write data to the object.""" if 'r' in self.mode: raise IOSError(EPERM, "File is opened for read") self.obj.send_chunk(data) def close(self): """Close the object and finish the data transfer.""" if 'r' in self.mode: return self.obj.finish_chunk() def read(self, size=65536): """ Read data from the object. We can use just one request because 'seek' is not fully supported. NB: It uses the size passed into the first call for all subsequent calls. """ if self.obj is None: if self.total_size > 0: self.conn.range_from = self.total_size # we need to open a new connection to inject the `Range` header if self.conn.http_conn: self.conn.http_conn[1].close() self.conn.http_conn = None _, self.obj = self.conn.get_object(self.container, self.name, resp_chunk_size=size) logging.debug("read size=%r, total_size=%r (range_from: %s)" % (size, self.total_size, self.conn.range_from)) try: buff = self.obj.next() self.total_size += len(buff) except StopIteration: return "" else: return buff def seek(self, offset, whence=None): """ Seek in the object. It's supported only for read operations because of object storage limitations. """ logging.debug("seek offset=%s, whence=%s" % (str(offset), str(whence))) if 'r' in self.mode: if self.size is None: meta = self.conn.head_object(self.container, self.name) try: self.size = int(meta["content-length"]) except ValueError: raise IOSError(EPERM, "Invalid file size") if not whence: offs = offset elif whence == 1: offs = self.total_size + offset elif whence == 2: offs = self.size - offset else: raise IOSError(EPERM, "Invalid file offset") if offs < 0 or offs > self.size: raise IOSError(EPERM, "Invalid file offset") # we need to start over after a seek call if self.obj is not None: del self.obj # GC the generator self.obj = None self.total_size = offs else: raise IOSError(EPERM, "Seek not available for write operations")
class CloudFilesFD(object): '''Acts like a file() object, but attached to a cloud files object''' def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.stream = None if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') self.container = self.cffs._get_container(self.container) if 'r' in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: #write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk() def write(self, data): '''Write data to the object''' if 'r' in self.mode: raise IOSError(EPERM, "Can't write to stream opened for read") self.obj.send_chunk(data) def close(self): '''Close the object and finish the data transfer''' if 'r' in self.mode: return self.obj.finish_chunk() def read(self, size=65536): '''Read data from the object. We can use just one request because 'seek' is not supported. NB: It uses the size passed into the first call for all subsequent calls''' if not self.stream: self.stream = self.obj.stream(size) logging.debug("read size=%r, total_size=%r, obj.size=%r" % (size, self.total_size, self.obj.size)) try: buff = self.stream.next() self.total_size += len(buff) except StopIteration: return "" else: return buff def seek(self, *kargs, **kwargs): '''Seek in the object: FIXME doesn't work and raises an error''' logging.debug("seek args=%s, kargs=%s" % (str(kargs), str(kwargs))) raise IOSError(EPERM, "Seek not implemented")
class CloudFilesFD(object): '''Acts like a file() object, but attached to a cloud files object''' def __init__(self, cffs, container, obj, mode): self.cffs = cffs self.container = container self.name = obj self.mode = mode self.closed = False self.total_size = 0 self.stream = None self.headers = dict() if not all([container, obj]): self.closed = True raise IOSError(EPERM, 'Container and object required') self.container = self.cffs._get_container(self.container) if 'r' in self.mode: self.obj = self.container.get_object(self.name) logging.debug("read fd obj.name=%r obj.size=%r" % (self.obj.name, self.obj.size)) else: #write self.obj = ChunkObject(self.container, obj) self.obj.content_type = mimetypes.guess_type(obj)[0] self.obj.prepare_chunk() def write(self, data): '''Write data to the object''' if 'r' in self.mode: raise IOSError(EPERM, "Can't write to stream opened for read") self.obj.send_chunk(data) def close(self): '''Close the object and finish the data transfer''' if 'r' in self.mode: return self.obj.finish_chunk() def read(self, size=65536): '''Read data from the object. We can use just one request because 'seek' is not fully supported. NB: It uses the size passed into the first call for all subsequent calls''' if not self.stream: self.stream = self.obj.stream(size, hdrs=self.headers) logging.debug("read size=%r, total_size=%r, obj.size=%r" % (size, self.total_size, self.obj.size)) try: buff = self.stream.next() self.total_size += len(buff) except StopIteration: return "" else: return buff def seek(self, offset, whence=None): '''Seek in the object. It's supported only for read operations because of the object storage limitations.''' logging.debug("seek offset=%s, whence=%s" % (str(offset), str(whence))) if 'r' in self.mode: if not whence: offs = offset elif whence == 1: offs = self.total_size + offset elif whence == 2: offs = self.obj.size - offset else: raise IOSError(EPERM, "Invalid file offset") if offs < 0 or offs > self.obj.size: raise IOSError(EPERM, "Invalid file offset") # we need to start over after a seek call if self.stream: self.stream = None self.obj = self.container.get_object(self.name) self.headers['Range'] = "bytes=%s-" % offs self.total_size = offs else: raise IOSError(EPERM, "Seek not available for write operations")