def compute(self, split):
buffers = [list() for i in self.fields]
remain_size = STRIPE_DATA_SIZE
path = os.path.join(self.path, '%04d.dt' % split.index)
indices = dict((i, AdaptiveIndex()) for i in self.indices)
def write_stripe(f, compressed, header, padding=True):
h = compress(marshal.dumps(header))
assert len(h) < STRIPE_HEADER_SIZE
f.write(struct.pack('I', len(h)))
f.write(h)
padding_size = STRIPE_SIZE - len(h) - 4
for c in compressed:
f.write(c)
padding_size -= len(c)
if padding:
f.write('\0' * padding_size)
with atomic_file(path) as f:
stripe_id = 0
for it in chain(self.prev.iterator(sp) for sp in split.splits):
row = it[:len(self.fields)]
size = len(marshal.dumps(tuple(row)))
if size > STRIPE_DATA_SIZE:
raise RuntimeError('Row too big')
if size > remain_size:
compressed = [compress(marshal.dumps(tuple(b))) for b in buffers]
_sizes = tuple(map(len, compressed))
_remain_size = STRIPE_DATA_SIZE - sum(_sizes)
if size > _remain_size:
write_stripe(f, compressed, _sizes)
buffers = [list() for i in self.fields]
remain_size = STRIPE_DATA_SIZE
stripe_id += 1
else:
remain_size = _remain_size
remain_size -= size
for i, value in enumerate(row):
buffers[i].append(value)
field = self.fields[i]
if field in self.indices:
indices[field].add(value, stripe_id)
if any(buffers):
compressed = [compress(marshal.dumps(tuple(b))) for b in buffers]
_sizes = tuple(map(len, compressed))
write_stripe(f, compressed, _sizes, False)
footer_indices = zlib.compress(cPickle.dumps(indices, -1))
footer_fields = compress(marshal.dumps(self.fields))
f.write(footer_indices)
f.write(footer_fields)
f.write(struct.pack('II', len(footer_fields), len(footer_indices)))
yield path
def compress_array(str_list):
"""
Compress an array of strings
By default LZ4 mode is standard in interactive mode,
and high compresion in applications/scripts
"""
if not ENABLE_PARALLEL:
return [lz4.compress(s) for s in str_list]
# Less than 50 chunks its quicker to compress sequentially..
if len(str_list) > LZ4_N_PARALLEL:
return clz4.compressarr(str_list)
else:
return [clz4.compress(s) for s in str_list]
def compress_array(str_list):
"""
Compress an array of strings
By default LZ4 mode is standard in interactive mode,
and high compresion in applications/scripts
"""
if not ENABLE_PARALLEL:
return [lz4.compress(s) for s in str_list]
# Less than 50 chunks its quicker to compress sequentially..
if len(str_list) > LZ4_N_PARALLEL:
return clz4.compressarr(str_list)
else:
return [clz4.compress(s) for s in str_list]
def _put(self, item):
hash_key = item.__hash__().to_bytes(length=20,
byteorder=sys.byteorder,
signed=True)
data = compress(pickle.dumps(
item, protocol=4)) if self.compression else pickle.dumps(
item, protocol=4)
with self._sem:
try:
with self._lmdb.begin(write=True) as txn:
if not txn.replace(hash_key, data, db=self._hashes_db):
key = next(self._idx).to_bytes(length=511,
byteorder='big',
signed=False)
self.logger.debug(
"Queuing new task with SERIAL {sn}".format(
sn=int.from_bytes(
key, byteorder='big', signed=False)))
txn.put(key, hash_key, append=True, db=self._queue_db)
else:
self.logger.debug("Updating already queued task")
except lmdb.MapFullError:
self.logger.critical(
"Database file {path} reached maximum size!".format(
path=self.path))
raise Full()
def scanSubvolume(self, subvolume):
hex = hashlib.sha1(subvolume).hexdigest()
file = self.folder + "/" + hex + ".lz4"
completed = self.folder + "/" + hex + ".complete"
print "scaning: " + subvolume + " (" + hex + ")"
if os.path.exists(completed):
return
p = Popen('btrfs subvolume find-new ' + subvolume + ' 0', shell=True, stdout=PIPE, stderr=STDOUT)
list = []
for line in p.stdout.readlines():
match = self.matchAllocation.match(line)
if match is None:
continue
list.append(match.group(0))
with open(file, "w") as text_file:
text_file.write(lz4.compress("\n".join(list)))
open(completed, 'a').close()
def open_db(self):
self.terms_ldb = leveldb.LevelDB(self.terms_fl)
self.docs_ldb = leveldb.LevelDB(self.docs_fl)
self.doc_buffer_size = 0
self.term_buffer_size = 0
#self.doc_flush_buffer = np.empty(self.max_doc_flush_buffer_size, dtype="S%d" % self.max_doc_size)
self.doc_flush_buffer = [None] * self.max_doc_flush_buffer_size
self.term_flush_buffer = np.empty(self.max_term_flush_buffer_size, dtype="S%d" % self.max_term_size)
self.doc_id_flush_buffer = np.empty(self.max_doc_flush_buffer_size, dtype=np.int64)
self.term_id_flush_buffer = np.empty(self.max_term_flush_buffer_size, dtype=np.int64)
if self.compression == COMPRESSION.NONE:
self.compress = lambda string: string
self.decompress = lambda string: string
elif self.compression == COMPRESSION.ZLIB:
import zlib
self.compress = lambda string: zlib.compress(string, self.compression_level)
self.decompress = lambda string: zlib.decompress(string)
elif self.compression == COMPRESSION.LZMA:
import backports.lzma as lzma
self.compress = lambda string: lzma.compress(bytearray(string), format=lzma.FORMAT_RAW)
self.decompress = lambda data: lzma.decompress(data, format=lzma.FORMAT_RAW)
elif self.compression == COMPRESSION.LZ4R:
import lz4
self.compress = lambda string: lz4.compress(string)
self.decompress = lambda string: lz4.decompress(string)
elif self.compression == COMPRESSION.LZ4H:
import lz4
self.compress = lambda string: lz4.compressHC(string)
self.decompress = lambda string: lz4.decompress(string)
else:
raise Exception("Wrong compression type %r" % self.compression)
def append(self, eventId, data):
try:
with self._sem:
with self.lmdb.begin(write=True) as txn:
eventId = eventId.encode('utf-8')
data = compress(json.dumps(data))
mtime = int(time.time() * 1000).to_bytes(
length=6, byteorder=sys.byteorder, signed=False)
delta_idx = next(self.delta_idx).to_bytes(length=511,
byteorder='big',
signed=False)
index_db = self.lmdb.open_db(key=self.index_name,
txn=txn,
dupsort=True)
mtimes_db = self.lmdb.open_db(key=self.mtimes_name,
txn=txn)
deltas_db = self.lmdb.open_db(key=self.deltas_name,
txn=txn)
txn.put(delta_idx, data, append=True, db=deltas_db)
txn.put(eventId, delta_idx, dupdata=True, db=index_db)
txn.put(eventId, mtime, overwrite=True, db=mtimes_db)
except lmdb.MapFullError:
raise DeltaStoreFull(
"Database file at {path} has reached its maximum size!".format(
path=self.db_path))
def zip_compress(plain, level=9):
if not USE_LZ4:
compressed = zlib.compress(plain, level)
return compressed[2:]
else:
compressed = lz4.compress(plain) if level < 9 else lz4.compressHC(plain)
return compressed[4:]
def compress(file):
data = open(file,"r+b")
if data.name.endswith('.tar'):
compressed_data = open(data.name + ".lz4","w")
compressed_data.write(lz4.compress(data.read()))
else:
print "The file type is not the expected."
def do_test_rountrip_method(compress, i_data, c_data=None):
from lz4 import LZ4_uncompress #@UnresolvedImport
c = compress(i_data)
if c_data is not None:
assert c_data==c, "expected compressed data to look like %s, but got %s" % (hl(c_data), hl(c))
d = LZ4_uncompress(c)
assert d==i_data, "expected decompressed data to look like original %s, but got %s" % (hl(i_data), hl(d))
def run(self, cb_upload, cb_hash):
if not self._blocksize:
raise Exception("Blocksize is not defined")
block_cnt = 0
block_new_cnt = 0
bs = self._block_size * 1024 * 1024
with open(self._device, "rb") as f:
while True:
#Read block and check for EOF
block = f.read(bs)
if not block:
self.log.info("Input EOF reached")
break
block_cnt += 1
#Create hash
h = xxhash.xxh64(block).hexdigest()
#Hash doesnt exist. Upload!
if h not in self.blocks:
block_new_cnt += 1
if self._compression:
block = lz4.compress(block)
cb_upload(h, block)
#Add hash to sync table
cb_hash(h)
self.log.info(
"Dedup::run() completed. Uploaded {} new blocks, {} blocks in total"
.format(block_new_cnt, block_cnt))
def zip_compress(plain, level=9): if not USE_LZ4: compressed = zlib.compress(plain, level) return compressed[2:] else: compressed = lz4.compress(plain) if level < 9 else lz4.compressHC(plain) return compressed[4:]
def send(self, message):
version = (0 & 0xf) << 28
msg_id = message.msg_id
if not msg_id:
msg_id = next(self.msg_ids)
message.msg_ig = msg_id
msg_id = (msg_id & 0xfff) << 16
msg_type = (message._MESSAGE_TYPE & 0xff) << 8
compress = False
compression = 0
if self.compress:
compress = True
elif self.compress is False and message._MESSAGE_TYPE != RESPONSE:
compress = True
msg = message.pack()
if compress and len(msg) >= COMPRESSION_THREASHOLD:
compression = 1
msg = lz4.compress(msg)
header = version + msg_id + msg_type + compression
data = _HEADER.pack(header, len(msg)) + msg
self.sock.sendall(data)
self.last_send = datetime.datetime.now()
def send(self, message):
version = (0 & 0xf) << 28
msg_id = message.msg_id
if not msg_id:
msg_id = next(self.msg_ids)
message.msg_ig = msg_id
msg_id = (msg_id & 0xfff) << 16
msg_type = (message._MESSAGE_TYPE & 0xff) << 8
compress = False
compression = 0
if self.compress:
compress = True
elif self.compress is False and message._MESSAGE_TYPE != RESPONSE:
compress = True
msg = message.pack()
if compress and len(msg) >= COMPRESSION_THREASHOLD:
compression = 1
msg = lz4.compress(msg)
header = version + msg_id + msg_type + compression
data = _HEADER.pack(header, len(msg)) + msg
self.sock.sendall(data)
self.last_send = datetime.datetime.now()
def _compress(
self, data
): # a compression funcion like lrzip in spirit: lz4>lz0>zlib>bz2>lzma
if self.shuffle == True:
try:
print "shuffling..."
data = buff_shuffle(
data) # shuffling will work for a filter < 1GB
print "data shuffled..."
except:
pass
print "Compressing..."
try:
data = lz4.compress(data) # will fail if filter > 1GB
print "lz4 ok"
except:
pass
try:
data = lzo.compress(data) # will fail if filter > 1GB
print "lzo ok"
except:
pass
#data = data.encode('zlib')
#data = zlib.compress(data,1)
#data = zlib.compress(data,9)
#data = bz2.compress(data,9)
data = zlib.compress(data)
data = bz2.compress(data)
data = lzma.compress(data)
return data
def benchmark(data, hcdata=None):
number = 100
size = len(data)
hcdata = hcdata or data
for modname, func in [("pylz4", pylz4.compress), ("rustlz4", rustlz4.compress)]:
timer = timeit.Timer(functools.partial(func, data))
elapsed = timer.timeit(number=number)
perf = size * number / elapsed / 1e6
name = "%s.%s" % (modname, func.__name__)
print("%24s: %8.2f MB/s" % (name, perf))
for modname, func in [("pylz4", pylz4.compressHC), ("rustlz4", rustlz4.compresshc)]:
timer = timeit.Timer(functools.partial(func, hcdata))
elapsed = timer.timeit(number=number)
perf = size * number / elapsed / 1e6
name = "%s.%s" % (modname, func.__name__)
print("%24s: %8.2f MB/s" % (name, perf))
data = pylz4.compress(data)
for modname, func in [("pylz4", pylz4.decompress), ("rustlz4", rustlz4.decompress)]:
timer = timeit.Timer(functools.partial(func, data))
elapsed = timer.timeit(number=number)
perf = size * number / elapsed / 1e6
name = "%s.%s" % (modname, func.__name__)
print("%24s: %8.2f MB/s" % (name, perf))
def fset(self, inst, value):
nprow = getattr(inst, 'NumpyArrayTable__' + self.name)
#~ print 'fset',self.name, nprow, value
if nprow is None:
nprow = self.NumpyArrayTableClass()
setattr(inst, 'NumpyArrayTable__' + self.name, nprow)
if value is None:
if hasattr(inst, self.name + '_array'):
delattr(inst, self.name + '_array')
nprow.shape = None
nprow.dtype = None
nprow.blob = None
nprow.units = None
nprow.compress = None
return
if self.arraytype == np.ndarray:
assert (type(value) == np.ndarray) or (
type(value) == np.memmap
), 'Value is not np.array or np.memmap but {}'.format(type(value))
if self.arraytype == pq.Quantity:
assert type(
value
) == pq.Quantity, '{} {} {} value is not pq.Quantity'.format(
inst.__class__.__name__, self.name, value)
shape = ('{},' * value.ndim)[:-1].format(*value.shape)
if shape.endswith(','): shape = shape[:-1]
nprow.shape = shape
nprow.dtype = value.dtype.str
if self.compress == 'blosc':
blob = blosc.compress(value.tostring(),
typesize=value.dtype.itemsize,
clevel=9)
else:
if not value.flags['C_CONTIGUOUS']:
buf = np.getbuffer(np.array(value, copy=True))
else:
buf = np.getbuffer(value)
if self.compress == 'zlib':
blob = zlib.compress(buf)
elif self.compress == 'lz4':
blob = lz4.compress(buf)
elif self.compress == 'snappy':
blob = snappy.compress(buf)
else:
blob = buf
nprow.compress = self.compress
nprow.blob = blob
if self.arraytype == pq.Quantity:
nprow.units = value.dimensionality.string
setattr(inst, self.name + '_array', value)
def chunks(inp): sr = dill.dumps(inp) inp = compress(sr) bs = 2**14 out = [] for x in range((len(inp) / bs) + 1): out.append(inp[(bs*x):(bs)*(x+1)]) return out
def compute_probabilities(replacement_idx, a, conditional_binary):
"""
Compute the compressed length of a bit string with
number as bit a replacing the number as bit in
position replacement_idx
"""
conditional_binary[replacement_idx] = a
return len(lz4.compress(''.join(conditional_binary))) #54.65
def roundtrip(size=None):
if size is None:
size = struct.unpack(">I", b"\0" + os.urandom(3))[0]
data = os.urandom(size)
assert rustlz4.decompress(pylz4.compress(data)) == data
assert pylz4.decompress(buffer(rustlz4.compress(data))) == data
assert rustlz4.decompress(pylz4.compressHC(data)) == data
assert pylz4.decompress(buffer(rustlz4.compresshc(data))) == data
def process_document(self, session, doc):
data = doc.get_raw(session)
new = lz4.compress(data)
return StringDocument(new,
self.id,
doc.processHistory,
parent=doc.parent,
filename=doc.filename)
def __init__(self, data_dir, obj_to_terms, obj_to_str, str_to_obj):
self.data_dir = data_dir
self.obj_to_terms = obj_to_terms
self.obj_to_str = obj_to_str
self.str_to_obj = str_to_obj
self.id_term_map = None
self.term_id_map = None
self.objnum = 0
try:
import lz4 as compressor
self.compress = compressor.compress
self.compressHC = compressor.compressHC
self.decompress = compressor.decompress
except ImportError:
import zlib as compressor
self.compress = lambda data: compressor.compress(data, 3)
self.compressHC = lambda data: compressor.compress(data, 9)
self.decompress = lambda data: compressor.decompress(data)
def test_val_to_store_info_compress(self):
"""_val_to_store_info() should compress large values.
"""
value = 'foo' * 32
compressed_value = lz4.compress(value)
client = memcache.Client('127.0.0.1', 11211, client_driver=NoopDriver)
result = client._val_to_store_info(value, min_compress_len=1)
self.assertEqual(result, (memcache.Client._FLAG_COMPRESSED,
compressed_value))
def do_test_rountrip_method(compress, i_data, c_data=None):
from lz4 import LZ4_uncompress #@UnresolvedImport
c = compress(i_data)
if c_data is not None:
assert c_data == c, "expected compressed data to look like %s, but got %s" % (
hl(c_data), hl(c))
d = LZ4_uncompress(c)
assert d == i_data, "expected decompressed data to look like original %s, but got %s" % (
hl(i_data), hl(d))
def __init__(self, data_dir, obj_to_terms, obj_to_str, str_to_obj):
self.data_dir = data_dir
self.obj_to_terms = obj_to_terms
self.obj_to_str = obj_to_str
self.str_to_obj = str_to_obj
self.id_term_map = None
self.term_id_map = None
self.objnum = 0
try:
import lz4 as compressor
self.compress = compressor.compress
self.compressHC = compressor.compressHC
self.decompress = compressor.decompress
except ImportError:
import zlib as compressor
self.compress = lambda data: compressor.compress(data, 3)
self.compressHC = lambda data: compressor.compress(data, 9)
self.decompress = lambda data: compressor.decompress(data)
def _dump2stor(store, bucketname, data, compress):
if len(data) == 0:
return ""
key = j.data.hash.md5_string(data)
if not key in objects or not key in new_objects:
if compress:
data = lz4.compress(data)
store.set_object(bucketname, key, data)
new_objects.append(key)
return key
def compress(scheme, data):
hdr = data[:4] + struct.pack(">L", (scheme << 27) + (len(data) & 0x07ffffff))
if scheme == 0 :
return data
elif scheme == 1 and lz4:
res = lz4.compress(hdr + data)
return res
else:
warnings.warn("Table failed to compress by unsupported compression scheme")
return data
def _dump2stor(store, bucketname, data, compress):
if len(data)==0:
return ""
key = j.data.hash.md5_string(data)
if not key in objects or not key in new_objects:
if compress:
data = lz4.compress(data)
store.set_object(bucketname, key, data)
new_objects.append(key)
return key
def serialize_subarray(cls, subarray):
if not subarray.flags['C_CONTIGUOUS']:
subarray = subarray.copy(order='C')
# Buffers larger than 1 GB would overflow
# We could fix this by slicing each slice into smaller pieces...
assert subarray.nbytes <= cls.MAX_LZ4_BUFFER_SIZE, \
"FIXME: This class doesn't support compression of arrays whose slices are each > 1 GB"
return lz4.compress(subarray)
def _put(self, item):
key = next(self._idx).to_bytes(length=511,
byteorder=sys.byteorder,
signed=False)
data = compress(pickle.dumps(
item, protocol=4)) if self.compression else pickle.dumps(
item, protocol=4)
with self._sem:
with self._lmdb.begin(write=True) as txn:
txn.put(key, data, append=True, db=self._queue_db)
def fset(self, inst, value):
nprow = getattr(inst, 'NumpyArrayTable__'+self.name)
#~ print 'fset',self.name, nprow, value
if nprow is None:
nprow = self.NumpyArrayTableClass()
setattr(inst, 'NumpyArrayTable__'+self.name, nprow)
if value is None:
if hasattr(inst, self.name+'_array') :
delattr(inst, self.name+'_array')
nprow.shape = None
nprow.dtype = None
nprow.blob = None
nprow.units = None
nprow.compress = None
return
if self.arraytype == np.ndarray:
assert (type(value) == np.ndarray) or (type(value) == np.memmap) , 'Value is not np.array or np.memmap but {}'.format(type(value))
if self.arraytype == pq.Quantity:
assert type(value) == pq.Quantity , '{} {} {} value is not pq.Quantity'.format(inst.__class__.__name__, self.name, value)
shape = ('{},'*value.ndim)[:-1].format(*value.shape)
if shape.endswith(',') : shape = shape[:-1]
nprow.shape = shape
nprow.dtype = value.dtype.str
if self.compress == 'blosc':
blob = blosc.compress(value.tostring(), typesize = value.dtype.itemsize, clevel= 9)
else:
if not value.flags['C_CONTIGUOUS']:
#~ buf = np.getbuffer(np.array(value, copy = True))
buf = np.array(value, copy=True).data
else:
#~ buf = np.getbuffer(value)
buf = value.data
if self.compress == 'zlib':
blob = zlib.compress(buf)
elif self.compress == 'lz4':
blob = lz4.compress(buf)
elif self.compress == 'snappy':
blob = snappy.compress(buf)
else :
blob = buf
nprow.compress = self.compress
nprow.blob = blob
if self.arraytype == pq.Quantity:
nprow.units = value.dimensionality.string
setattr(inst, self.name+'_array', value)
def test_value_to_store_info_compress_length(self):
""" _val_to_store_info() should not use compressed values if too long.
That is, if the compressed value is longer than the original value, use
the original value instead.
"""
value = '...'
compressed_value = lz4.compress(value)
self.assertGreater(len(compressed_value), len(value))
client = memcache.Client('127.0.0.1', 11211, client_driver=NoopDriver)
result = client._val_to_store_info(value, min_compress_len=1)
self.assertEqual(result, (0, value))
def write_stripe(f, compressed, header, padding=True):
h = compress(marshal.dumps(header))
assert len(h) < STRIPE_HEADER_SIZE
f.write(struct.pack('I', len(h)))
f.write(h)
padding_size = STRIPE_SIZE - len(h) - 4
for c in compressed:
f.write(c)
padding_size -= len(c)
if padding:
f.write('\0' * padding_size)
def write_stripe(f, compressed, header, padding=True):
h = compress(marshal.dumps(header))
assert len(h) < STRIPE_HEADER_SIZE
f.write(struct.pack('I', len(h)))
f.write(h)
padding_size = STRIPE_SIZE - len(h) - 4
for c in compressed:
f.write(c)
padding_size -= len(c)
if padding:
f.write('\0' * padding_size)
def _compress_as_lz4(self):
if self._lz4_items is None:
self._uncompress() # Ensure not currently compressed as draco
compressed = []
flat_vertices = self._vertices_zyx.reshape(-1)
compressed.append( lz4.compress(flat_vertices) ) #@UndefinedVariable
self._vertices_zyx = None
flat_normals = self._normals_zyx.reshape(-1)
compressed.append( lz4.compress(flat_normals) ) #@UndefinedVariable
self._normals_zyx = None
flat_faces = self._faces.reshape(-1)
compressed.append( lz4.compress(flat_faces) ) #@UndefinedVariable
self._faces = None
# Compress twice: still fast, even smaller
self._lz4_items = list(map(lz4.compress, compressed)) #@UndefinedVariable
return sum(map(len, self._lz4_items))
def send(self, data, host, port):
"""Send raw data through a socket"""
if not isinstance(data, dict) and not isinstance(data, list):
raise TypeError("data must be either a list or a dictionary")
# Getting raw data
if self._use_lz4:
data_raw = lz4.compress(self._codec.encode(data))
else:
data_raw = self._codec.encode(data)
# Put the data on the wire as an UTF-8 JSON string
self.sock.sendto(data_raw, (host, port))
def __init__(self, numpy_array):
"""Serializes and compresses the numpy array with LZ4"""
self.raw_buffer = None # only used if we can't compress
self.compressed_label_blocks = None # only used for label arrays of suitable shape
self.compressed_mask_array = None # only used for binary masks
self.serialized_subarrays = []
if numpy_array.flags['F_CONTIGUOUS']:
self.layout = 'F'
else:
self.layout = 'C'
if self.layout == 'F':
numpy_array = numpy_array.transpose()
self.dtype = numpy_array.dtype
self.shape = numpy_array.shape
# TODO: Also support compression of bool arrays via the special DVID binary compression
if self.is_labels(numpy_array):
self.compressed_label_blocks = serialize_uint64_blocks(numpy_array)
elif self.dtype == np.bool and numpy_array.ndim == 3:
# It turns out that encode_mask_array + lz4.compress is better than
# lz4.compression alone (even multiple rounds of lz4 alone)
self.compressed_mask_array = lz4.compress(
encode_mask_array(numpy_array))
else:
if numpy_array.ndim <= 1:
slice_bytes = numpy_array.nbytes
else:
slice_bytes = numpy_array[0].nbytes
if slice_bytes > CompressedNumpyArray.MAX_LZ4_BUFFER_SIZE:
warnings.warn(
"Array is too large to compress -- not compressing.")
if not numpy_array.flags['C_CONTIGUOUS']:
numpy_array = numpy_array.copy(order='C')
self.raw_buffer = bytearray(numpy_array)
else:
# For 2D, 1D or 0D arrays, serialize everything in one buffer.
if numpy_array.ndim <= 2:
self.serialized_subarrays.append(
self.serialize_subarray(numpy_array))
else:
# For ND arrays, serialize each slice independently, to ease RAM usage
for subarray in numpy_array:
self.serialized_subarrays.append(
self.serialize_subarray(subarray))
def __init__(self, data_dir):
# term = str()
# triple = str()
# args(triple) = (int)
self.data_dir = data_dir
# table: id(term) -> term
self.term_id_map = None
# table: id(triple) -> args(triple)
self.triple_id_map = None
# table: id(term) -> args(triple)
self.arg_cache = None
self.rel_id_map = REL_NAME_ID_MAP
self.id_rel_map = REL_ID_NAME_MAP
try:
import lz4 as compressor
self.compress = compressor.compress
self.compressHC = compressor.compressHC
self.decompress = compressor.decompress
except ImportError:
import zlib as compressor
self.compress = lambda data: compressor.compress(data, 3)
self.compressHC = lambda data: compressor.compress(data, 9)
self.decompress = lambda data: compressor.decompress(data)
def __init__(self, data_dir):
# term = str()
# triple = str()
# args(triple) = (int)
self.data_dir = data_dir
# table: id(term) -> term
self.term_id_map = None
# table: id(triple) -> args(triple)
self.triple_id_map = None
# table: id(term) -> args(triple)
self.arg_cache = None
self.rel_id_map = REL_NAME_ID_MAP
self.id_rel_map = REL_ID_NAME_MAP
try:
import lz4 as compressor
self.compress = compressor.compress
self.compressHC = compressor.compressHC
self.decompress = compressor.decompress
except ImportError:
import zlib as compressor
self.compress = lambda data: compressor.compress(data, 3)
self.compressHC = lambda data: compressor.compress(data, 9)
self.decompress = lambda data: compressor.decompress(data)
def save_book_raw_data(book_raw_data_obj):
if use_compression:
f = open(
join("/Volumes/NewVolume/Emotional-Arcs/database/cache",
str(book_raw_data_obj.this_Book.pk) + ".p.lz4"), "wb")
f.write(
lz4.compress(
pickle.dumps(book_raw_data_obj, pickle.HIGHEST_PROTOCOL)))
f.close()
else:
f = open(
join("/Volumes/NewVolume/Emotional-Arcs/database/cache",
str(book_raw_data_obj.this_Book.pk) + ".p"), "wb")
f.write(pickle.dumps(book_raw_data_obj, pickle.HIGHEST_PROTOCOL))
f.close()
def save_book_raw_data(book_raw_data_obj):
if use_compression:
f = open(
join("/Users/andyreagan/projects/2014/09-books/data/cache",
str(book_raw_data_obj.this_Book.pk) + ".p.lz4"), "wb")
f.write(
lz4.compress(
pickle.dumps(book_raw_data_obj, pickle.HIGHEST_PROTOCOL)))
f.close()
else:
f = open(
join("/Users/andyreagan/projects/2014/09-books/data/cache",
str(book_raw_data_obj.this_Book.pk) + ".p"), "wb")
f.write(pickle.dumps(book_raw_data_obj, pickle.HIGHEST_PROTOCOL))
f.close()
def push_to_s3(self, compress=True):
"""Upload object to Amazon s3 bucket"""
s3 = self.connect_and_check_bucket()
# Archive
logging.info('Archiving {}...'.format(self.name))
tar_path = os.path.join(self.PATH, '{}.tar'.format(self.name))
lz4_path = os.path.join(self.PATH, '{}.tar.lz4'.format(self.name))
with tarfile.open(tar_path, 'w') as in_file:
for filename in glob.glob(
os.path.join(self.PATH,
'{stem}.*'.format(stem=self.name))):
in_file.add(filename, arcname=os.path.split(filename)[1])
# Compress
if compress:
logging.info('{} LZ4 compressing...'.format(self.name))
with open(tar_path, 'rb') as in_file:
with open(lz4_path, 'wb') as out_file:
out_file.write(lz4.compress(in_file.read()))
upload_path = lz4_path
else:
upload_path = tar_path
# Get key
key = self.get_key(compress=compress)
# Upload
logging.info('Uploading {} on s3...'.format(self.name))
s3.meta.client.upload_file(upload_path,
self.BUCKET_NAME,
key,
Callback=self.ProgressPercentage(
float(os.path.getsize(upload_path))))
# Clean
os.remove(tar_path)
if compress:
os.remove(lz4_path)
logging.info('Upload {} DONE'.format(self.name))
def compress(data, clibs=[]):
if len(clibs) == 0:
return (False, None, data)
for c in clibs:
try:
if c == "zlib":
import zlib
return (True, "zlib", zlib.compress(data))
elif c == "lz4":
import lz4
return (True, "zlib", lz4.compress(data))
elif c == "snappy":
import pysnappy
return (True, "snappy", pysnappy.compress(data))
else:
import zlib
return (True, "zlib", zlib.compress(data))
except:
return (False, None, data)
return (False, None, data)
def open_db(self):
self.terms_ldb = leveldb.LevelDB(self.terms_fl)
self.docs_ldb = leveldb.LevelDB(self.docs_fl)
self.doc_buffer_size = 0
self.term_buffer_size = 0
#self.doc_flush_buffer = np.empty(self.max_doc_flush_buffer_size, dtype="S%d" % self.max_doc_size)
self.doc_flush_buffer = [None] * self.max_doc_flush_buffer_size
self.term_flush_buffer = np.empty(self.max_term_flush_buffer_size,
dtype="S%d" % self.max_term_size)
self.doc_id_flush_buffer = np.empty(self.max_doc_flush_buffer_size,
dtype=np.int64)
self.term_id_flush_buffer = np.empty(self.max_term_flush_buffer_size,
dtype=np.int64)
if self.compression == COMPRESSION.NONE:
self.compress = lambda string: string
self.decompress = lambda string: string
elif self.compression == COMPRESSION.ZLIB:
import zlib
self.compress = lambda string: zlib.compress(
string, self.compression_level)
self.decompress = lambda string: zlib.decompress(string)
elif self.compression == COMPRESSION.LZMA:
import backports.lzma as lzma
self.compress = lambda string: lzma.compress(
bytearray(string), format=lzma.FORMAT_RAW)
self.decompress = lambda data: lzma.decompress(
data, format=lzma.FORMAT_RAW)
elif self.compression == COMPRESSION.LZ4R:
import lz4
self.compress = lambda string: lz4.compress(string)
self.decompress = lambda string: lz4.decompress(string)
elif self.compression == COMPRESSION.LZ4H:
import lz4
self.compress = lambda string: lz4.compressHC(string)
self.decompress = lambda string: lz4.decompress(string)
else:
raise Exception("Wrong compression type %r" % self.compression)
def compress_fixture(self, fixture_lz4_path, with_header):
"""
Compress json fixture with lz4
fixture_lz4_path:
The path to the compressed fixture
with_header:
Specify if you want to have the mozilla header
at the start of the compressed file
"""
fixture_json_path = 'fixtures/fixture.json'
with open(fixture_json_path, mode='rb') as json_file:
data = json_file.read()
compressed_data = lz4.compress(data)
with open(fixture_lz4_path, mode='wb') as lz4_file:
if with_header:
header = b'mozLz40\0'
lz4_file.write(header)
lz4_file.write(compressed_data)
def _val_to_store_info(self, val, min_compress_len):
"""
Transform val to a storable representation, returning a tuple of the
flags, the length of the new value, and the new value itself.
"""
flags = 0
if isinstance(val, str):
pass
elif isinstance(val, int):
flags |= Client._FLAG_INTEGER
val = "%d" % val
# maxint is pretty tiny. just return
return (flags, val)
elif isinstance(val, long):
flags |= Client._FLAG_LONG
val = "%d" % val
# longs can be huge, so check length and compress if long enough
else:
if self.pickle:
flags |= Client._FLAG_PICKLE
val = pickle.dumps(val, self.pickle_proto)
lv = len(val)
# do not store if value length exceeds maximum
if self.max_value_length and lv > self.max_value_length:
raise MemcacheValueError(
"Value is larger than configured max_value_length. %d > %d" %
(lv, self.max_value_length))
# We should try to compress if min_compress_len > 0 and this
# string is longer than min threshold.
if min_compress_len and lv > min_compress_len:
comp_val = lz4.compress(val)
# Only actually compress if the compressed result is smaller
# than the original.
if len(comp_val) < lv:
flags |= Client._FLAG_COMPRESSED
val = comp_val
return (flags, val)
def compress_fixture(self, fixture_lz4_path, with_header):
"""
Compress json fixture with lz4
fixture_lz4_path:
The path to the compressed fixture
with_header:
Specify if you want to have the mozilla header
at the start of the compressed file
"""
fixture_json_path = 'fixtures/fixture.json'
with open(fixture_json_path, mode='rb') as json_file:
data = json_file.read()
compressed_data = lz4.compress(data)
with open(fixture_lz4_path, mode='wb') as lz4_file:
if with_header:
header = b'mozLz40\0'
lz4_file.write(header)
lz4_file.write(compressed_data)
def run():
fileDir = '/root/thermite/target/release/thermite_test_Neil'
ratioFile_path = sys.argv[1]
ratioList = []
openData = open(fileDir, "rb")
ratioFile = open(ratioFile_path, "w")
vvfd = os.open(fileDir, os.O_RDONLY)
vvData = os.lseek(vvfd, 0, os.SEEK_END)
upper_limit = vvData / 4096
for x in xrange(0, upper_limit, 1):
openData.seek(4096 * x, 0)
data = openData.read(4096)
compOutput = lz4.compress(data)
sys.getsizeof(compOutput)
ratio = sys.getsizeof(compOutput) / float(sys.getsizeof(data))
ratioFile.write(str(ratio) + '\n')
ratioFile.close()
def pack(fname):
global total, compressed
f = open(fname)
data = f.read()
f.close()
checksum = siphashc.siphash('\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0', data)
fsize = len(data)
data = lz4.compress(data)
# size of packet(4), checksum(8), fnamelen(2)+fname, uncompressed size(4), compressed data
l = len(data)
pktlen = 4 + 8 + 2 + len(fname) + 4 + l
total += fsize
compressed += len(data)
sys.stderr.write("%s: %d -> %d\n" %(fname, fsize, len(data)))
sys.stdout.write( struct.pack('<IQH%dsI%ds' % (len(fname), l), pktlen, checksum, len(fname), fname,fsize,data))
sys.stdout.flush()
for line in bz2_fh:
line = line.decode('utf-8')
title_match = title_re.match(line)
if title_match:
if 'lat_d' in old_coords and 'long_d' in old_coords:
lat, lng = normalize_coords(**old_coords)
j += 1
if lat and lng:
abstract, img = extract_abstract(text)
rank = len(text)
#print("\t".join(map(str, (title, lat, lng))))
print(title)
try:
POI(name=title, at=[lng, lat], abstract=lz4.compress(abstract), rank=rank, img=img).save()
#POI(name=title, at=[lng, lat], abstract=abstract).save()
except Exception as e:
print("Insert error:", str(e), title, lat, lng, file=sys.stderr)
# raise
# print("Begin abstract")
# print(abstract)
# print("End abstract")
title = title_match.group(1)
coords, old_coords, lat, lng, text, in_text = None, {}, None, None, '', False
continue
coord_match = coord_re.match(line)
if coord_match:
try:
lat, lng = coord2latlng(line)
# print("\t".join(map(str, (title, lat, lng))))
import uuid
import timeit
import lz4
import snappy
import os
from timeit import Timer
DATA = open("../src/lz4.c", "rb").read()
LZ4_DATA = lz4.compress(DATA)
SNAPPY_DATA = snappy.compress(DATA)
LOOPS = 200000
print("Data Size:")
print(" Input: %d" % len(DATA))
print(" LZ4: %d (%.2f)" % (len(LZ4_DATA), len(LZ4_DATA) / float(len(DATA))))
print(" Snappy: %d (%.2f)" % (len(SNAPPY_DATA), len(SNAPPY_DATA) / float(len(DATA))))
print(" LZ4 / Snappy: %f" % (float(len(LZ4_DATA)) / float(len(SNAPPY_DATA))))
print("Benchmark: %d calls" % LOOPS)
print(" LZ4 Compression: %fs" % Timer("lz4.compress(DATA)", "from __main__ import DATA; import lz4").timeit(number=LOOPS))
print(" Snappy Compression: %fs" % Timer("snappy.compress(DATA)", "from __main__ import DATA; import snappy").timeit(number=LOOPS))
print(" LZ4 Decompression: %fs" % Timer("lz4.uncompress(LZ4_DATA)", "from __main__ import LZ4_DATA; import lz4").timeit(number=LOOPS))
print(" Snappy Decompression : %fs" % Timer("snappy.uncompress(SNAPPY_DATA)", "from __main__ import SNAPPY_DATA; import snappy").timeit(number=LOOPS))
def process_document(self, session, doc):
data = doc.get_raw(session)
new = lz4.compress(data)
return StringDocument(new, self.id, doc.processHistory,
parent=doc.parent, filename=doc.filename)
def test_roundtripLZ4():
_str = "hello world"
cstr = lz4.compress(_str)
assert _str == c.decompress(cstr)
def lz4_pickle_dump(data, filename):
path = pathlib.Path(filename)
with path.open('wb') as f:
f.write(lz4.compress(pickle.dumps(data, protocol=pickle.HIGHEST_PROTOCOL)))
def _test_block(labels, test_name):
# labelarray
with Timer() as timer:
dvid_encoded_list = serialize_uint64_blocks(labels)
dvid_encoded_bytes = sum(map(len, dvid_encoded_list))
dvid_enc_time = timer.seconds
dvid_enc_throughput = (labels.nbytes / dvid_enc_time) / 1e6
with Timer() as timer:
decoded = deserialize_uint64_blocks(dvid_encoded_list, labels.shape)
assert (decoded == labels).all()
dvid_dec_time = timer.seconds
dvid_dec_throughput = (labels.nbytes / dvid_dec_time) / 1e6
# DVID + gzip
with Timer() as timer:
gzipped_dvid_encoded_list = list(map(gzip.compress, dvid_encoded_list))
gzipped_dvid_enc_time = timer.seconds + dvid_enc_time
gzipped_dvid_enc_throughput = (labels.nbytes / gzipped_dvid_enc_time) / 1e6
gzipped_dvid_encoded_bytes = sum(map(len, gzipped_dvid_encoded_list))
print("+ GZIP:", gzipped_dvid_encoded_bytes)
print(f"Compression ratio: {labels.nbytes/gzipped_dvid_encoded_bytes:.1f}x")
print(f"DVID+GZIP encode throughput: {gzipped_dvid_enc_throughput} MB/s")
with Timer() as timer:
unzippped = list(map(gzip.decompress, gzipped_dvid_encoded_list))
assert (decoded == labels).all()
gzipped_dvid_dec_time = timer.seconds + dvid_dec_time
gzipped_dvid_dec_throughput = (labels.nbytes / gzipped_dvid_dec_time) / 1e6
print(f"DVID+GZIP decode throughput: {gzipped_dvid_dec_throughput} MB/s")
# DVID + LZ4
with Timer() as timer:
lz4_dvid_encoded_list = list(map(lz4.compress, dvid_encoded_list))
lz4_dvid_enc_time = timer.seconds + dvid_enc_time
lz4_dvid_enc_throughput = (labels.nbytes / lz4_dvid_enc_time) / 1e6
lz4_dvid_encoded_bytes = sum(map(len, lz4_dvid_encoded_list))
print("+ LZ4:", lz4_dvid_encoded_bytes)
print(f"Compression ratio: {labels.nbytes/lz4_dvid_encoded_bytes:.1f}x")
print(f"DVID+LZ4 encode throughput: {lz4_dvid_enc_throughput} MB/s")
with Timer() as timer:
unzippped = list(map(lz4.decompress, lz4_dvid_encoded_list))
assert (decoded == labels).all()
lz4_dvid_dec_time = timer.seconds + dvid_dec_time
lz4_dvid_dec_throughput = (labels.nbytes / lz4_dvid_dec_time) / 1e6
print(f"DVID+LZ4 decode throughput: {lz4_dvid_dec_throughput} MB/s")
# lz4
with Timer() as timer:
lz4_encoded = lz4.compress(labels)
lz4_encoded_bytes = len(lz4_encoded)
lz4_enc_time = timer.seconds
lz4_enc_throughput = (labels.nbytes / lz4_enc_time) / 1e6
with Timer() as timer:
lz4_decoded = lz4.decompress(lz4_encoded)
decoded_labels = np.frombuffer(lz4_decoded, np.uint64).reshape(labels.shape)
assert (decoded_labels == labels).all()
lz4_dec_time = timer.seconds
lz4_dec_throughput = (labels.nbytes / lz4_dec_time) / 1e6
global HEADER_PRINTED
if not HEADER_PRINTED:
print(f"{'':>20s} {'______ ENCODED BYTES ______ ':^41s} | {'______ ENCODING TIME ______ ':^77s} | {'______ DECODING TIME ______ ':^77s} |")
print(f"{'':>20s} {'LZ4':>10s} {'DVID':>10s} {'D+G':>10s} {'DECREASE':>9s} |"
f"{'------- LZ4 -------':>22s} {'------ DVID ------':>22s} {'---- DVID+GZIP ----':>22s} {'SLOWDOWN':>9s} |"
f"{'------- LZ4 -------':>22s} {'------ DVID ------':>22s} {'---- DVID+GZIP ----':>22s} {'SLOWDOWN':>9s} |")
HEADER_PRINTED = True
print(f"{test_name:>19s}: {lz4_encoded_bytes: 10d} {dvid_encoded_bytes: 10d} {lz4_encoded_bytes/dvid_encoded_bytes:8.1f}x |"
f"{lz4_enc_time:6.2f}s ({lz4_enc_throughput:7.1f} MB/s) {dvid_enc_time:6.2f}s ({dvid_enc_throughput:7.1f} MB/s) {dvid_enc_time/lz4_enc_time:8.1f}x |"
f"{lz4_dec_time:6.2f}s ({lz4_dec_throughput:7.1f} MB/s) {dvid_dec_time:6.2f}s ({dvid_dec_throughput:7.1f} MB/s) {dvid_dec_time/lz4_dec_time:8.1f}x |")
def test_lz4_4_2():
lz4.compress(garbage100k)
def lz4_compress(byts):
# write length in big-endian instead of little-endian
return int32_pack(len(byts)) + lz4.compress(byts)[4:]
from mokujin.resource import StopList
from mokujin.misc import transliterate_ru
from mokujin.resource import ConceptNetList
from mokujin.index import TripleSearchEngine
from mokujin.sourcesearch import TripleStoreExplorer
from mokujin.patternsearch import PatternCollection
try:
import lz4 as comp
comp_format = "lz4"
compress = comp.compressHC
decompress = comp.decompress
except ImportError:
import zlib as comp
comp_format = "zip"
compress = lambda string: comp.compress(string, 9)
decompress = comp.decompress
if __name__ == "__main__":
logging.basicConfig(level=logging.INFO)
parser = argparse.ArgumentParser()
parser.add_argument("-i", "--index", default="data/index", help="Triple store index directory", type=str)
parser.add_argument("-o", "--outputdir", default="output",
help="Directory where script's ouput will be placed", type=str)
parser.add_argument("-q", "--queryterm", default=None, help="Query term", type=str)
parser.add_argument("-qf", "--queryterms_file", default=None, help="File with query terms", type=str)
parser.add_argument("-s", "--stoplist", default="resources/word.freq.ru.csv", help="Stop list file", type=str)
parser.add_argument("-ts", "--t_stop", default=500, help="Stop words frequency threshold", type=float)
parser.add_argument("-tt", "--t_triple", default=5, help="Min frequency treshold for target triples", type=float)
