Source code for pyrocko.io.segy

# http://pyrocko.org - GPLv3
#
# The Pyrocko Developers, 21st Century
# ---|P------/S----------~Lg----------
from __future__ import division, absolute_import

import numpy as num
import struct
import calendar

from pyrocko import trace, util
from .io_common import FileLoadError


[docs]def ibm2ieee(ibm): """ Converts an IBM floating point number into IEEE format. :param: ibm - 32 bit unsigned integer: unpack('>L', f.read(4)) """ if ibm == 0: return 0.0 sign = ibm >> 31 & 0x01 exponent = ibm >> 24 & 0x7f mantissa = (ibm & 0x00ffffff) / float(pow(2, 24)) print('x', sign, exponent - 64, mantissa) return (1 - 2 * sign) * mantissa * float(pow(16, exponent - 64))
def unpack_ibm_f4(data): ibm = num.frombuffer(data, dtype='>u4').astype(num.int32) sign = (ibm >> 31) & 0x01 exponent = (ibm >> 24) & 0x7f mantissa = (ibm & 0x00ffffff) / float(pow(2, 24)) xxx = (1 - 2 * sign) * mantissa * (16.0 ** (exponent - 64))\ .astype(float) # for i in range(len(data)/4): # yyy = ibm2ieee(struct.unpack('>L', data[i*4:(i+1)*4])[0]) # print('y', sign[i], exponent[i] - 64, mantissa[i]) # print(xxx[i], yyy) # if xxx[i] != yyy: # sys.exit() # print return xxx
[docs]class SEGYError(Exception): pass
[docs]def iload(filename, load_data, endianness='>'): ''' Read SEGY file. filename -- Name of SEGY file. load_data -- If True, the data is read, otherwise only read headers. ''' endianness = endianness nbth = 3200 nbbh = 400 nbthx = 3200 nbtrh = 240 try: f = open(filename, 'rb') textual_file_header = f.read(nbth) if len(textual_file_header) != nbth: raise SEGYError('incomplete textual file header') binary_file_header = f.read(nbbh) if len(binary_file_header) != nbbh: raise SEGYError('incomplete binary file header') line_number = struct.unpack(endianness+'1I', binary_file_header[4:8]) hvals = struct.unpack(endianness+'24H', binary_file_header[12:12+24*2]) (ntraces, nauxtraces, deltat_us, deltat_us_orig, nsamples, nsamples_orig, format, ensemble_fold) = hvals[0:8] (segy_revision, fixed_length_traces, nextended_headers) = \ struct.unpack(endianness+'3H', binary_file_header[100:100+3*2]) if ntraces == 0 and nauxtraces == 0: ntraces = 1 formats = { 1: (unpack_ibm_f4, 4, "4-byte IBM floating-point"), 2: (endianness+'i4', 4, "4-byte, two's complement integer"), 3: (endianness+'i4', 2, "2-byte, two's complement integer"), 4: (None, 4, "4-byte fixed-point with gain (obolete)"), 5: (endianness+'f4', 4, "4-byte IEEE floating-point"), 6: (None, 0, "not currently used"), 7: (None, 0, "not currently used"), 8: ('i1', 1, "1-byte, two's complement integer")} dtype = formats[format][0] sample_size = formats[format][1] if dtype is None: raise SEGYError('unsupported sample data format %i: %s' % ( format, formats[format][2])) for ihead in range(nextended_headers): f.read(nbthx) atend = False while not atend: for itrace in range((ntraces+nauxtraces)): trace_header = f.read(nbtrh) if len(trace_header) == 0: atend = True break if len(trace_header) != nbtrh: raise SEGYError('incomplete trace header') (scoordx, scoordy, gcoordx, gcoordy) = \ struct.unpack(endianness+'4f', trace_header[72:72+4*4]) (ensemblex, ensembley) = \ struct.unpack(endianness+'2f', trace_header[180:180+2*4]) (ensemble_num,) = \ struct.unpack(endianness+'1I', trace_header[20:24]) (trensemble_num,) = \ struct.unpack(endianness+'1I', trace_header[24:28]) (trace_number,) = \ struct.unpack(endianness+'1I', trace_header[0:4]) (trace_numbersegy,) = \ struct.unpack(endianness+'1I', trace_header[4:8]) (orfield_num,) = \ struct.unpack(endianness+'1I', trace_header[8:12]) (ortrace_num,) = \ struct.unpack(endianness+'1I', trace_header[12:16]) # don't know if this is standard: distance to shot [m] as int (idist,) = struct.unpack(endianness+'1I', trace_header[36:40]) tvals = struct.unpack( endianness+'12H', trace_header[94:94+12*2]) (nsamples_this, deltat_us_this) = tvals[-2:] tscalar = struct.unpack( endianness+'1H', trace_header[214:216])[0] if tscalar == 0: tscalar = 1. elif tscalar < 0: tscalar = 1.0 / tscalar else: tscalar = float(tscalar) tvals = [x * tscalar for x in tvals[:-2]] (year, doy, hour, minute, second) = \ struct.unpack(endianness+'5H', trace_header[156:156+2*5]) # maybe not standard? (msecs, usecs) = struct.unpack( endianness+'2H', trace_header[168:168+4]) try: if year < 100: year += 2000 tmin = util.to_time_float(calendar.timegm( (year, 1, doy, hour, minute, second))) \ + msecs * 1.0e-3 + usecs * 1.0e-6 except Exception: raise SEGYError('invalid start date/time') if fixed_length_traces: if (nsamples_this, deltat_us_this) \ != (nsamples, deltat_us): raise SEGYError( 'trace of incorrect length or sampling ' 'rate (trace=%i)' % itrace+1) if load_data: datablock = f.read(nsamples_this*sample_size) if len(datablock) != nsamples_this*sample_size: raise SEGYError('incomplete trace data') if isinstance(dtype, str): data = num.frombuffer(datablock, dtype=dtype) else: data = dtype(datablock) tmax = None else: f.seek(nsamples_this*sample_size, 1) tmax = tmin + deltat_us_this/1000000.*(nsamples_this-1) data = None if data is not None and isinstance(dtype, str) and ( str(dtype).startswith('<') or str(dtype).startswith('>')): data = data.astype(str(dtype)[1:]) tr = trace.Trace( '', '%05i' % (line_number), '%02i' % (ensemble_num), '%03i' % (ortrace_num), tmin=tmin, tmax=tmax, deltat=deltat_us_this/1000000., ydata=data, meta=dict( orfield_num=orfield_num, distance=float(idist))) yield tr except (OSError, SEGYError) as e: raise FileLoadError(e) finally: f.close()