def pack_unpack_fp(repeats,
chunk_size=DEFAULT_CHUNK_SIZE,
progress=False,
metadata=None):
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
if progress:
print("Creating test array")
create_array_fp(repeats, in_fp, progress=progress)
in_fp_size = in_fp.tell()
if progress:
print("Compressing")
in_fp.seek(0)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(in_fp_size, chunk_size)
source = PlainFPSource(in_fp)
sink = CompressedFPSink(out_fp)
pack(source, sink, nchunks, chunk_size, last_chunk_size, metadata=metadata)
out_fp.seek(0)
if progress:
print("Decompressing")
source = CompressedFPSource(out_fp)
sink = PlainFPSink(dcmp_fp)
unpack(source, sink)
if progress:
print("Verifying")
cmp_fp(in_fp, dcmp_fp)
return source.metadata
def pack_unpack_fp(repeats, chunk_size=DEFAULT_CHUNK_SIZE,
progress=False, metadata=None):
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
if progress:
print("Creating test array")
create_array_fp(repeats, in_fp, progress=progress)
in_fp_size = in_fp.tell()
if progress:
print("Compressing")
in_fp.seek(0)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(in_fp_size, chunk_size)
source = PlainFPSource(in_fp)
sink = CompressedFPSink(out_fp)
pack(source, sink,
nchunks, chunk_size, last_chunk_size,
metadata=metadata)
out_fp.seek(0)
if progress:
print("Decompressing")
source = CompressedFPSource(out_fp)
sink = PlainFPSink(dcmp_fp)
unpack(source, sink)
if progress:
print("Verifying")
cmp_fp(in_fp, dcmp_fp)
return source.metadata
def test_append_single_chunk():
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking)
orig.seek(0)
new.seek(0)
# append a single chunk
reset_append_fp(orig, new, new_size)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 2)
# append a large content, that amounts to two chunks
new_content = new.read()
new.seek(0)
reset_append_fp(orig, StringIO(new_content * 2), new_size * 2)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 4)
# append half a chunk
reset_append_fp(orig, StringIO(new_content[:len(new_content)]), new_size//2)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 5)
# append a few bytes
reset_append_fp(orig, StringIO(new_content[:1023]), 1024)
# make sure it is squashed into the lat chunk
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 5)
def test_append_into_last_chunk():
# first create an array with a single chunk
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking)
orig.seek(0)
new.seek(0)
# append a few bytes, creating a new, smaller, last_chunk
new_content = new.read()
new.seek(0)
nchunks = reset_append_fp(orig, StringIO(new_content[:1023]), 1023)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(nchunks, 1)
nt.assert_equal(bloscpack_header['last_chunk'], 1023)
# now append into that last chunk
nchunks = reset_append_fp(orig, StringIO(new_content[:1023]), 1023)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(nchunks, 0)
nt.assert_equal(bloscpack_header['last_chunk'], 2046)
# now check by unpacking
source = CompressedFPSource(orig)
sink = PlainFPSink(dcmp)
unpack(source, sink)
dcmp.seek(0)
new.seek(0)
new_str = new.read()
dcmp_str = dcmp.read()
nt.assert_equal(len(dcmp_str), len(new_str) + 2046)
nt.assert_equal(dcmp_str, new_str + new_str[:1023] * 2)
def test_append_single_chunk():
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking)
orig.seek(0)
new.seek(0)
# append a single chunk
reset_append_fp(orig, new, new_size)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 2)
# append a large content, that amounts to two chunks
new_content = new.read()
new.seek(0)
reset_append_fp(orig, StringIO(new_content * 2), new_size * 2)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 4)
# append half a chunk
reset_append_fp(orig, StringIO(new_content[:len(new_content)]),
new_size // 2)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 5)
# append a few bytes
reset_append_fp(orig, StringIO(new_content[:1023]), 1024)
# make sure it is squashed into the lat chunk
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(bloscpack_header['nchunks'], 5)
def pack_unpack_mem(repeats, chunk_size=DEFAULT_CHUNK_SIZE,
progress=False, metadata=None):
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
if progress:
print("Creating test array")
create_array_fp(repeats, in_fp, progress=progress)
in_fp_size = in_fp.tell()
if progress:
print("Compressing")
in_fp.seek(0)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(in_fp_size, chunk_size)
# let us play merry go round
source = PlainFPSource(in_fp)
sink = CompressedMemorySink()
pack(source, sink, nchunks, chunk_size, last_chunk_size, metadata=metadata)
source = CompressedMemorySource(sink)
sink = PlainMemorySink()
unpack(source, sink)
nt.assert_equal(metadata, source.metadata)
source = PlainMemorySource(sink.chunks)
sink = CompressedFPSink(out_fp)
pack(source, sink, nchunks, chunk_size, last_chunk_size, metadata=metadata)
out_fp.seek(0)
source = CompressedFPSource(out_fp)
sink = PlainFPSink(dcmp_fp)
unpack(source, sink)
nt.assert_equal(metadata, source.metadata)
in_fp.seek(0)
dcmp_fp.seek(0)
cmp_fp(in_fp, dcmp_fp)
return source.metadata
def pack_unpack_mem(repeats,
chunk_size=DEFAULT_CHUNK_SIZE,
progress=False,
metadata=None):
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
if progress:
print("Creating test array")
create_array_fp(repeats, in_fp, progress=progress)
in_fp_size = in_fp.tell()
if progress:
print("Compressing")
in_fp.seek(0)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(in_fp_size, chunk_size)
# let us play merry go round
source = PlainFPSource(in_fp)
sink = CompressedMemorySink()
pack(source, sink, nchunks, chunk_size, last_chunk_size, metadata=metadata)
source = CompressedMemorySource(sink)
sink = PlainMemorySink()
unpack(source, sink)
nt.assert_equal(metadata, source.metadata)
source = PlainMemorySource(sink.chunks)
sink = CompressedFPSink(out_fp)
pack(source, sink, nchunks, chunk_size, last_chunk_size, metadata=metadata)
out_fp.seek(0)
source = CompressedFPSource(out_fp)
sink = PlainFPSink(dcmp_fp)
unpack(source, sink)
nt.assert_equal(metadata, source.metadata)
in_fp.seek(0)
dcmp_fp.seek(0)
cmp_fp(in_fp, dcmp_fp)
return source.metadata
def test_append_into_last_chunk():
# first create an array with a single chunk
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking)
orig.seek(0)
new.seek(0)
# append a few bytes, creating a new, smaller, last_chunk
new_content = new.read()
new.seek(0)
nchunks = reset_append_fp(orig, StringIO(new_content[:1023]), 1023)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(nchunks, 1)
nt.assert_equal(bloscpack_header['last_chunk'], 1023)
# now append into that last chunk
nchunks = reset_append_fp(orig, StringIO(new_content[:1023]), 1023)
bloscpack_header = reset_read_beginning(orig)[0]
nt.assert_equal(nchunks, 0)
nt.assert_equal(bloscpack_header['last_chunk'], 2046)
# now check by unpacking
source = CompressedFPSource(orig)
sink = PlainFPSink(dcmp)
unpack(source, sink)
dcmp.seek(0)
new.seek(0)
new_str = new.read()
dcmp_str = dcmp.read()
nt.assert_equal(len(dcmp_str), len(new_str) + 2046)
nt.assert_equal(dcmp_str, new_str + new_str[:1023] * 2)
def test_disable_offsets():
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
create_array_fp(1, in_fp)
in_fp_size = in_fp.tell()
in_fp.seek(0)
bloscpack_args = BloscpackArgs(offsets=False)
source = PlainFPSource(in_fp)
sink = CompressedFPSink(out_fp)
pack(source, sink, *calculate_nchunks(in_fp_size), bloscpack_args=bloscpack_args)
out_fp.seek(0)
bloscpack_header, metadata, metadata_header, offsets = _read_beginning(out_fp)
nt.assert_true(len(offsets) == 0)
def test_offsets():
with create_tmp_files() as (tdir, in_file, out_file, dcmp_file):
create_array(1, in_file)
pack_file(in_file, out_file, chunk_size='2M')
with open(out_file, 'r+b') as input_fp:
bloscpack_header = _read_bloscpack_header(input_fp)
total_entries = bloscpack_header.total_prospective_chunks
offsets = _read_offsets(input_fp, bloscpack_header)
# First chunks should start after header and offsets
first = BLOSCPACK_HEADER_LENGTH + 8 * total_entries
# We assume that the others are correct
nt.assert_equal(offsets[0], first)
nt.assert_equal([
736, 368207, 633319, 902306, 1173771, 1419535, 1666981, 1913995
], offsets)
# try to read the second header
input_fp.seek(offsets[1], 0)
blosc_header_raw = input_fp.read(BLOSC_HEADER_LENGTH)
expected = {
'versionlz': 1,
'blocksize': 262144,
'ctbytes': 265108,
'version': 2,
'flags': 1,
'nbytes': 2097152,
'typesize': 8
}
blosc_header = decode_blosc_header(blosc_header_raw)
nt.assert_equal(expected, blosc_header)
# now check the same thing again, but w/o any max_app_chunks
input_fp, output_fp = StringIO(), StringIO()
create_array_fp(1, input_fp)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(input_fp.tell(), chunk_size='2M')
input_fp.seek(0, 0)
bloscpack_args = BloscpackArgs(max_app_chunks=0)
source = PlainFPSource(input_fp)
sink = CompressedFPSink(output_fp)
pack(source,
sink,
nchunks,
chunk_size,
last_chunk_size,
bloscpack_args=bloscpack_args)
output_fp.seek(0, 0)
bloscpack_header = _read_bloscpack_header(output_fp)
nt.assert_equal(0, bloscpack_header.max_app_chunks)
offsets = _read_offsets(output_fp, bloscpack_header)
nt.assert_equal(
[96, 367567, 632679, 901666, 1173131, 1418895, 1666341, 1913355],
offsets)
def prep_array_for_append(blosc_args=BloscArgs(),
bloscpack_args=BloscpackArgs()):
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking,
blosc_args=blosc_args,
bloscpack_args=bloscpack_args)
orig.seek(0)
new.seek(0)
return orig, new, new_size, dcmp
def prep_array_for_append(blosc_args=BloscArgs(),
bloscpack_args=BloscpackArgs()):
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
chunking = calculate_nchunks(new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking,
blosc_args=blosc_args,
bloscpack_args=bloscpack_args)
orig.seek(0)
new.seek(0)
return orig, new, new_size, dcmp
def test_offsets():
with create_tmp_files() as (tdir, in_file, out_file, dcmp_file):
create_array(1, in_file)
pack_file(in_file, out_file, chunk_size='2M')
with open(out_file, 'r+b') as input_fp:
bloscpack_header = _read_bloscpack_header(input_fp)
total_entries = bloscpack_header.total_prospective_chunks
offsets = _read_offsets(input_fp, bloscpack_header)
# First chunks should start after header and offsets
first = BLOSCPACK_HEADER_LENGTH + 8 * total_entries
# We assume that the others are correct
nt.assert_equal(offsets[0], first)
nt.assert_equal([736, 368207, 633319, 902306, 1173771,
1419535, 1666981, 1913995],
offsets)
# try to read the second header
input_fp.seek(offsets[1], 0)
blosc_header_raw = input_fp.read(BLOSC_HEADER_LENGTH)
expected = {'versionlz': 1,
'blocksize': 262144,
'ctbytes': 265108,
'version': 2,
'flags': 1,
'nbytes': 2097152,
'typesize': 8}
blosc_header = decode_blosc_header(blosc_header_raw)
nt.assert_equal(expected, blosc_header)
# now check the same thing again, but w/o any max_app_chunks
input_fp, output_fp = StringIO(), StringIO()
create_array_fp(1, input_fp)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(input_fp.tell(), chunk_size='2M')
input_fp.seek(0, 0)
bloscpack_args = BloscpackArgs(max_app_chunks=0)
source = PlainFPSource(input_fp)
sink = CompressedFPSink(output_fp)
pack(source, sink,
nchunks, chunk_size, last_chunk_size,
bloscpack_args=bloscpack_args
)
output_fp.seek(0, 0)
bloscpack_header = _read_bloscpack_header(output_fp)
nt.assert_equal(0, bloscpack_header.max_app_chunks)
offsets = _read_offsets(output_fp, bloscpack_header)
nt.assert_equal([96, 367567, 632679, 901666,
1173131, 1418895, 1666341, 1913355],
offsets)
def test_offsets():
with create_tmp_files() as (tdir, in_file, out_file, dcmp_file):
create_array(1, in_file)
pack_file(in_file, out_file, chunk_size='2M')
with open(out_file, 'r+b') as input_fp:
bloscpack_header = _read_bloscpack_header(input_fp)
total_entries = bloscpack_header.total_prospective_chunks
offsets = _read_offsets(input_fp, bloscpack_header)
# First chunks should start after header and offsets
first = BLOSCPACK_HEADER_LENGTH + 8 * total_entries
# We assume that the others are correct
nt.assert_equal(offsets[0], first)
nt.assert_equal([736, 418578, 736870, 1050327,
1363364, 1660766, 1959218, 2257703],
offsets)
# try to read the second header
input_fp.seek(offsets[1], 0)
blosc_header_raw = input_fp.read(BLOSC_HEADER_LENGTH)
expected = {'versionlz': 1,
'blocksize': 131072,
'ctbytes': 318288,
'version': 2,
'flags': 1,
'nbytes': 2097152,
'typesize': 8}
blosc_header = decode_blosc_header(blosc_header_raw)
nt.assert_equal(expected, blosc_header)
# now check the same thing again, but w/o any max_app_chunks
input_fp, output_fp = StringIO(), StringIO()
create_array_fp(1, input_fp)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(input_fp.tell(), chunk_size='2M')
input_fp.seek(0, 0)
bloscpack_args = BloscpackArgs(max_app_chunks=0)
source = PlainFPSource(input_fp)
sink = CompressedFPSink(output_fp)
pack(source, sink,
nchunks, chunk_size, last_chunk_size,
bloscpack_args=bloscpack_args
)
output_fp.seek(0, 0)
bloscpack_header = _read_bloscpack_header(output_fp)
nt.assert_equal(0, bloscpack_header.max_app_chunks)
offsets = _read_offsets(output_fp, bloscpack_header)
nt.assert_equal([96, 417938, 736230, 1049687,
1362724, 1660126, 1958578, 2257063],
offsets)
def test_offsets():
with create_tmp_files() as (tdir, in_file, out_file, dcmp_file):
create_array(1, in_file)
pack_file_to_file(in_file, out_file, chunk_size='2M')
with open(out_file, 'r+b') as input_fp:
bloscpack_header = _read_bloscpack_header(input_fp)
total_entries = bloscpack_header.total_prospective_chunks
offsets = _read_offsets(input_fp, bloscpack_header)
# First chunks should start after header and offsets
first = BLOSCPACK_HEADER_LENGTH + 8 * total_entries
# We assume that the others are correct
nt.assert_equal(offsets[0], first)
nt.assert_equal(736, offsets[0])
# try to read the second header
input_fp.seek(offsets[1], 0)
blosc_header_raw = input_fp.read(BLOSC_HEADER_LENGTH)
expected = {
'versionlz': 1,
'version': 2,
'flags': 1,
'nbytes': 2097152,
'typesize': 8
}
blosc_header = decode_blosc_header(blosc_header_raw)
blosc_header_slice = dict(
(k, blosc_header[k]) for k in expected.keys())
nt.assert_equal(expected, blosc_header_slice)
# now check the same thing again, but w/o any max_app_chunks
input_fp, output_fp = StringIO(), StringIO()
create_array_fp(1, input_fp)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(input_fp.tell(), chunk_size='2M')
input_fp.seek(0, 0)
bloscpack_args = BloscpackArgs(max_app_chunks=0)
source = PlainFPSource(input_fp)
sink = CompressedFPSink(output_fp)
pack(source,
sink,
nchunks,
chunk_size,
last_chunk_size,
bloscpack_args=bloscpack_args)
output_fp.seek(0, 0)
bloscpack_header = _read_bloscpack_header(output_fp)
nt.assert_equal(0, bloscpack_header.max_app_chunks)
offsets = _read_offsets(output_fp, bloscpack_header)
nt.assert_equal(96, offsets[0])
def test_disable_offsets():
in_fp, out_fp, dcmp_fp = StringIO(), StringIO(), StringIO()
create_array_fp(1, in_fp)
in_fp_size = in_fp.tell()
in_fp.seek(0)
bloscpack_args = BloscpackArgs(offsets=False)
source = PlainFPSource(in_fp)
sink = CompressedFPSink(out_fp)
pack(source,
sink,
*calculate_nchunks(in_fp_size),
bloscpack_args=bloscpack_args)
out_fp.seek(0)
bloscpack_header, metadata, metadata_header, offsets = \
_read_beginning(out_fp)
nt.assert_true(len(offsets) == 0)
def test_offsets():
with create_tmp_files() as (tdir, in_file, out_file, dcmp_file):
create_array(1, in_file)
pack_file(in_file, out_file, chunk_size='2M')
with open(out_file, 'r+b') as input_fp:
bloscpack_header = _read_bloscpack_header(input_fp)
total_entries = bloscpack_header.total_prospective_chunks
offsets = _read_offsets(input_fp, bloscpack_header)
# First chunks should start after header and offsets
first = BLOSCPACK_HEADER_LENGTH + 8 * total_entries
# We assume that the others are correct
nt.assert_equal(offsets[0], first)
nt.assert_equal(736, offsets[0])
# try to read the second header
input_fp.seek(offsets[1], 0)
blosc_header_raw = input_fp.read(BLOSC_HEADER_LENGTH)
expected = {'versionlz': 1,
'version': 2,
'flags': 1,
'nbytes': 2097152,
'typesize': 8}
blosc_header = decode_blosc_header(blosc_header_raw)
blosc_header_slice = dict((k, blosc_header[k]) for k in expected.keys())
nt.assert_equal(expected, blosc_header_slice)
# now check the same thing again, but w/o any max_app_chunks
input_fp, output_fp = StringIO(), StringIO()
create_array_fp(1, input_fp)
nchunks, chunk_size, last_chunk_size = \
calculate_nchunks(input_fp.tell(), chunk_size='2M')
input_fp.seek(0, 0)
bloscpack_args = BloscpackArgs(max_app_chunks=0)
source = PlainFPSource(input_fp)
sink = CompressedFPSink(output_fp)
pack(source, sink,
nchunks, chunk_size, last_chunk_size,
bloscpack_args=bloscpack_args
)
output_fp.seek(0, 0)
bloscpack_header = _read_bloscpack_header(output_fp)
nt.assert_equal(0, bloscpack_header.max_app_chunks)
offsets = _read_offsets(output_fp, bloscpack_header)
nt.assert_equal(96, offsets[0])
def test_append_metadata():
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
metadata = {"dtype": "float64", "shape": [1024], "others": []}
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking, metadata=metadata)
orig.seek(0)
new.seek(0)
reset_append_fp(orig, new, new_size)
source = CompressedFPSource(orig)
sink = PlainFPSink(dcmp)
ans = unpack(source, sink)
print(ans)
dcmp.seek(0)
new.seek(0)
new_str = new.read()
dcmp_str = dcmp.read()
nt.assert_equal(len(dcmp_str), len(new_str) * 2)
nt.assert_equal(dcmp_str, new_str * 2)
def test_append_metadata():
orig, new, dcmp = StringIO(), StringIO(), StringIO()
create_array_fp(1, new)
new_size = new.tell()
new.seek(0)
metadata = {"dtype": "float64", "shape": [1024], "others": []}
chunking = calculate_nchunks(new_size, chunk_size=new_size)
source = PlainFPSource(new)
sink = CompressedFPSink(orig)
pack(source, sink, *chunking, metadata=metadata)
orig.seek(0)
new.seek(0)
reset_append_fp(orig, new, new_size)
source = CompressedFPSource(orig)
sink = PlainFPSink(dcmp)
ans = unpack(source, sink)
print(ans)
dcmp.seek(0)
new.seek(0)
new_str = new.read()
dcmp_str = dcmp.read()
nt.assert_equal(len(dcmp_str), len(new_str) * 2)
nt.assert_equal(dcmp_str, new_str * 2)