def tonative(trace, format, endian):
if endian == 'little': trace.byteswap(inplace=True)
return native(trace, format=format, copy=False)
def extract(self, trace):
# segyio.tools.native assumes MSB (big endian) and will always byteswap
# on LSB hosts. This effectively means a double byteswap for LSB data
# on LSB hosts.
if self.endian == 'little': trace.byteswap(inplace=True)
return native(trace['samples'], format=self.format, copy=False)
def upload(manifest, fragment_shape, src, filesys):
"""Upload volume to oneseismic
Parameters
----------
manifest : dict
The parsed output of the scan program
fragment_shape : tuple of int
src : io.BaseIO
blob : azure.storage.blob.BlobServiceClient
"""
word1 = manifest['key-words'][0]
word2 = manifest['key-words'][1]
key1s = manifest['dimensions'][0]
key2s = manifest['dimensions'][1]
key3s = manifest['dimensions'][2]
guid = manifest['guid']
# Seek past the textual headers and the binary header
src.seek(int(manifest['byteoffset-first-trace']), io.SEEK_CUR)
# Make a custom dtype that corresponds to a header and a trace. This
# assumes all traces are of same length and sampled similarly, which is
# a safe assumption in practice. This won't be checked though, the check
# belongs in the scan program.
#
# The dtype is quite useful because it means the input can be read into the
# numpy array as a buffer, and then passed on directly as numpy arrays to
# put()
dtype = np.dtype([
('header', 'b', 240),
('samples', 'f4', len(key3s)),
])
trace = np.array(1, dtype=dtype)
fmt = manifest['format']
files = fileset(key1s, key2s, key3s, fragment_shape)
files.setlimits(manifest['key1-last-trace'])
shapeident = '-'.join(map(str, fragment_shape))
prefix = f'src/{shapeident}'
filesys.mkdir(guid)
filesys.cd(guid)
while True:
n = src.readinto(trace)
if n == 0:
break
header = segyio.field.Field(buf=trace['header'], kind='trace')
data = native(data=trace['samples'], format=fmt)
key1 = header[word1]
key2 = header[word2]
files.put(key1, key2, data)
for ident, fragment in files.commit(key1):
ident = '-'.join(map(str, ident))
name = f'{prefix}/{ident}.f32'
print('uploading', name)
with filesys.open(name, mode='wb') as f:
f.write(fragment)
with filesys.open('manifest.json', mode='wb') as f:
f.write(json.dumps(manifest).encode())