core.py

#! /usr/bin/env python
# -*- coding: utf-8 -*-
"""
GCF bindings to python
"""
import ctypes
from io import IOBase

import math
import os
import warnings

import numpy as np

from obspy import UTCDateTime
from obspy.core import Stream, Trace, AttribDict
from obspy.core.util.libnames import _load_cdll

# supported sampling rates outside range 1-250 Hz
# key:value = samplingrate:fractional_time_denominator
_SPS_MAP = {
    0.1: 1,
    0.125: 1,
    0.2: 1,
    0.25: 1,
    0.5: 1,
    400: 8,
    500: 2,
    625: 5,
    800: 16,
    1000: 4,
    1250: 5,
    2000: 8,
    2500: 10,
    4000: 16,
    5000: 20
}
# Unsupported sampling rates in range 1-250 Hz
_SPS_RESERVED = [157, 161, 162, 164, 167, 171, 174, 175, 176, 179, 181, 182,
                 191, 193, 194]
# Valid differential gains
_VALID_GAIN = [-1, 0, 1, 2, 4, 8, 16, 32, 64]
# Min/max data values
_GCF_MIN = -2147483648
_GCF_MAX = 2147483647

# Mapping between error codes set by reader and messages
_ERRORMSG_READ = {
    -1: "data block %d is not a data block",
    1: "stream ID in header of data block %d use extended system ID format",
    2: "stream ID in header of data block %d use double-extended "
       "system ID format",
    3: "unknown compression in data block %d",
    4: "to few/many data samples indicated in block header of data block %d",
    5: "start time of first data sample in data block %d is negative",
    9: "failure to decode header values of data block %d",
    10: "failure to decode data block %d (last data != RIC)",
    11: "1'st first difference != 0 in data block %d",
    21: "failure to decode data block %d (last data != RIC, 1'st first "
        "difference != 0)"
}
_ERROR_WRITE = {
    -2: IOError("failed to write to disc"),
    -1: IOError("failed to open file to write to"),
    1: ValueError("no data or inconsistent headers"),
    2: ValueError("unsupported sampling rate"),
    3: ValueError("bad fractional start time"),
    4: ValueError("unsupported gain"),
    5: ValueError("unsupported instrument type"),
    6: ValueError("too long system_id"),
    }


class _GcfSeg(ctypes.Structure):
    _fields_ = [
      ('streamID', ctypes.c_char * 7),
      ('systemID', ctypes.c_char * 7),
      ('start', ctypes.c_int32),
      ('t_numerator', ctypes.c_int),
      ('t_denominator', ctypes.c_int),
      ('t_leap', ctypes.c_int),
      ('gain', ctypes.c_int),
      ('sysType', ctypes.c_int),
      ('digi', ctypes.c_int),
      ('ttl', ctypes.c_int),
      ('blk', ctypes.c_int),
      ('err', ctypes.c_int),
      ('sps', ctypes.c_int),
      ('sps_denom', ctypes.c_int),
      ('compr', ctypes.c_int),
      ('FIC', ctypes.c_int32),
      ('RIC', ctypes.c_int32),
      ('n_data', ctypes.c_int32),
      ('n_alloc', ctypes.c_int32),
      ('data', ctypes.POINTER(ctypes.c_int32))
    ]


class _GcfFile(ctypes.Structure):
    _fields_ = [
      ('n_blk', ctypes.c_int),
      ('n_seg', ctypes.c_int),
      ('n_alloc', ctypes.c_int),
      ('n_errHead', ctypes.c_int),
      ('n_errData', ctypes.c_int),
      ('seg', ctypes.POINTER(_GcfSeg))
    ]


def compatible_sps(sps):
    """
    Checks if a sampling rate is compatible with the GCF format

    :type sps: float, int
    :param sps: sampling rate to test
    :rtype: bool
    :returns: True if sampling rate is compatible else False
    """
    if not (isinstance(sps, int) or isinstance(sps, float)):
        return False
    if 1 <= sps <= 250:
        if int(sps)-sps != 0 or sps in _SPS_RESERVED:
            return False
    elif sps not in _SPS_MAP:
        return False
    return True


def get_time_denominator(sps):
    """
    Returns the time fractional offset denominator, d, associated with an
    input sampling rate.

    Any data written to a gcf file must have its first sample sampled at a
    fractional time:
        n/d
    where 0 <= n < d

    :type sps: float, int
    :param sps: sampling rate (samples per second)
    :rtype: int
    :returns: fractional offset denominator, if 0 input sampling rate is not
        compatible with implemented GCF format, if 1 first data sample must be
        sampled in integer time
    """
    denom = 0
    if compatible_sps(sps):
        if sps <= 250:
            denom = 1
        else:
            denom = _SPS_MAP[sps]
    return denom


def merge_gcf_stream(st):
    """
    Merges GCF stream (replacing Stream.merge(-1) for headonly=True)

    :type st: :class:`~obspy.core.stream.Stream`
    :param st: GCF Stream object with no data
    :rtype: :class:`~obspy.core.stream.Stream`
    :returns: Stream object containing header and data.
    """
    traces = []
    for tr in st:
        delta = tr.stats.delta
        starttime = tr.stats.starttime
        endtime = tr.stats.endtime
        for trace in traces:
            if tr.id == trace.id and delta == trace.stats.delta \
              and not starttime == trace.stats.starttime:
                if 0 < starttime - trace.stats.endtime <= delta:
                    trace.stats.npts += tr.stats.npts
                    break
                elif 0 < trace.stats.starttime - endtime <= delta:
                    trace.stats.starttime = UTCDateTime(starttime)
                    trace.stats.npts += tr.stats.npts
                    break
        else:
            traces.append(tr)
    return Stream(traces=traces)


def _is_gcf(file):
    """
    Checks whether a file is GCF or not.

    :type file: str or file-like object
    :param file: path to GCF file to be checked
    :rtype: bool
    :return: True if a object pointed to by path is a GCF file.
    """
    input_is_gcf = True

    # Load shared library
    gcf_io = _load_cdll("gcf")

    # declare function argument and return types
    gcf_io.free_GcfFile.argtypes = [ctypes.POINTER(_GcfFile)]
    gcf_io.free_GcfFile.restype = None
    # avoid releasing uninitialized memory in free_GcfFile later:
    gcf_io.init_GcfFile.argtypes = [ctypes.POINTER(_GcfFile)]
    gcf_io.init_GcfFile.restype = None

    obj = None
    try:
        # Decode first block:
        obj = _GcfFile()
        # make sure pointers in obj are properly initialized to NULL:
        gcf_io.init_GcfFile(obj)
        ret = _fill_gcf_from_file(file, obj, 3)

        # b_filename = filename.encode('utf-8')
        # ret = gcf_io.read_gcf(fid, obj, 3)
        if ret or (obj.n_errHead and obj.seg[0].err not in (10, 11, 21)) \
                or obj.n_errData:
            input_is_gcf = False
    except Exception:
        input_is_gcf = False
    finally:
        # release allocated memory
        if obj is not None:
            gcf_io.free_GcfFile(obj)

    return input_is_gcf


def _fill_gcf_from_file(file, obj, mode):
    """Read data from `file` into the given `_GcfFile` object `obj`.
    This function is the Python implementation of the C function
    `gcf_io.read_gcf` to allow reading from both file paths
    and file-like objects, and should not be called directly:
    see `_read_gcf` and `_is_gcf` for details.

    :param file: str, Path or file-like object
    :param obj: an empty instance of a _GcfFile` object
    :param mode: the read mode. See `_read_gcf` for details
    """
    # Load shared library
    gcf_io = _load_cdll("gcf")

    # declare function argument and return types
    gcf_io.read_gcf.argtypes = [ctypes.c_char_p,
                                ctypes.POINTER(_GcfFile),
                                ctypes.c_int]
    gcf_io.read_gcf.restype = ctypes.c_int
    gcf_io.init_GcfSeg_for_read.argtypes = [ctypes.POINTER(_GcfSeg),
                                            ctypes.c_int]
    gcf_io.init_GcfSeg_for_read.restype = None
    gcf_io.is_LittleEndian_gcf.argtypes = []
    gcf_io.is_LittleEndian_gcf.restype = ctypes.c_int
    gcf_io.read_gcf_block.argtypes = [ctypes.POINTER(_GcfFile),
                                      ctypes.c_char_p,
                                      ctypes.POINTER(_GcfSeg),
                                      ctypes.c_int,
                                      ctypes.c_int,
                                      ctypes.c_double]
    gcf_io.read_gcf_block.restype = ctypes.c_int
    gcf_io.free_GcfSeg.argtypes = [ctypes.POINTER(_GcfSeg)]
    gcf_io.free_GcfSeg.restype = None
    gcf_io.merge_GcfFile.argtypes = [ctypes.POINTER(_GcfFile),
                                     ctypes.c_int,
                                     ctypes.c_double]
    gcf_io.merge_GcfFile.restype = None

    d = 0  # number of empty blocks
    cur_pos = None
    ret = 0
    fpt = None  # file-like object
    try:
        if isinstance(file, IOBase):
            cur_pos = file.tell()
            fpt = file
        else:
            fpt = open(file, "rb")
    except Exception:
        ret = -1

    # get endianess of current machine
    endian = gcf_io.is_LittleEndian_gcf()

    if not ret:
        b1 = 0
        tol = 1.E-3
        seg = _GcfSeg()
        # initiate and allocate segment
        gcf_io.init_GcfSeg_for_read(seg, mode)

        # adjust mode if nedded
        if mode > 2:
            mode = 2
            b1 = 1

        while True:
            buffer = fpt.read(1024)
            if not buffer:
                break
            d += gcf_io.read_gcf_block(obj, buffer, seg, mode, endian, tol)
            if b1:
                break

        # free segment
        gcf_io.free_GcfSeg(seg)

        # merge segments if asked for
        if abs(mode) < 2:
            gcf_io.merge_GcfFile(obj, mode, tol)

    if obj.n_blk == d:
        ret = 1 + endian  # no data blocks in file

    # close file or restore position:
    if fpt is not None:
        if cur_pos is not None:
            fpt.seek(cur_pos, 0)
        else:
            fpt.close()

    return ret


def _read_gcf(file, headonly=False, network='', station='',
              location='', channel_prefix='HH', blockmerge=True,
              cleanoverlap=True, errorret=False, **kwargs):
    """
    Reads a GCF file and returns a :class:`~obspy.core.stream.Stream`

    Only GCF data records are supported. Function supports format as described
    by GCF Reference
    `SWA-RFC-GCFR Issue F, December 2021
    <https://www.guralp.com/apps/ok?doc=GCF_format>`_

    .. warning::
            This function should NOT be called directly, it registers via the
            ObsPy :func:`~obspy.core.stream.read` function, call this instead.

    .. rubric:: Exceptions

        Function will raise ``IOError`` on problems to read file

    .. rubric:: GCF header data

        GCF specific meta data will be set for each
            :class:`~obspy.core.trace.Trace` object in the returned
            :class:`~obspy.core.stream.Stream` object under attribute
            ``stats.gcf``, more precisely the following attributes will be set:

            ``stats.gcf.system_id`` : str
                system_id set in block header
            ``stats.gcf.stream_id`` : str
                6-character stream_id set in block, typically consists of:
                 (4-character unitID) + (1-character orientation code) + (int,
                 tap)
            ``stats.gcf.sys_type`` : int
                system_id type:
                 ``0`` - regular
                 ``1`` - extended
                 ``2`` - double extended
            ``stats.gcf.t_leap`` : bool
                1 if block start at leap second, else 0
            ``stats.gcf.gain`` : int
                variable gain setting (if ``-1`` not used)
            ``stats.gcf.digi`` : int
                digitizer type, combine with ``stats.gcf.sys_type`` to find the
                digitizer according to:
                ==================  ==============  ===========================
                stats.gcf.sys_type  stats.gcf.digi  digitizer
                ==================  ==============  ===========================
                0                   0               unknown (probably DM24 Mk2)
                1                   0               DM24
                1                   1               CD24
                2                   0               Affinity
                2                   1               Minimus
                ==================  ==============  ===========================
            ``stats.gcf.ttl`` : int
                tap-table-lookup to retrieve sequence of decimation filters
                used (see GCF reference linked above for further info)
            ``stats.gcf.blk`` : int
                Lowest block number in gcf file of blocks included in trace
                (count start at 0)
            ``stats.gcf.FIC`` : int
                Forward Integration Constant (i.e. first data value)
            ``stats.gcf.RIC`` : int
                Reverse Integration Constant (ideally last data value)
            ``stats.gcf.stat`` : int
                Status code, available codes are:
                ==== ==========================================================
                code description
                ==== ==========================================================
                -1   not a data block
                0    no detected block issues
                1    stream ID use extended system ID format
                2    stream ID use double-extended system ID format
                3    unknown compression
                4    to few/many data samples indicated in block header
                5    start of first data sample is negative
                9    failure to decode header values
                10   failure to decode data values (last data != RIC, may be
                     due to bad RIC!!!)
                11   first first difference != 0 (may not be critical!!!)
                21   status codes 10 + 11
                ==== ==========================================================

    .. rubric:: Example

        >>> from obspy import read
        >>> st = read("/path/to/20160603_1955n.gcf", format="GCF")

    :type file: str or :class:`~pathlib.Path` or
        file-like object( e.g. `ByesIO`)
    :param file: path to GCF file to read
    :type headonly: bool, optional
    :param headonly: if True only read block headers
    :type network: str, optional
    :param network: network code to use
    :type station: str, optional
    :param station: station code to use, if not specified unitID in gcf header
       will be used (first 4 characters in stream_id)
    :type location: str, optional
    :param location: location code to use
    :type channel_prefix: str, optional
    :param channel_prefix: 2-character channel prefix to use (e.g. ``"HH"`` for
        a high-gain seismometer sampled at 200 Hz)
    :type blockmerge: bool, optional
    :param blockmerge: if True merge blocks if aligned
    :type cleanoverlap: bool, optional
    :param cleanoverlap: ignored if blockmerge is False, if True, remove
        overlapping identical data prior to merging blocks, if False
        overlapping blocks will not be merged even if data in overlap is
        identical
    :type errorret: bool, optional
    :param errorret: if ``True`` block and header issues will be set in
        trace.stats.gcf.stat for each :class:`~obspy.core.trace.Trace`
        object in returned :class:`~obspy.core.stream.Stream` object, else
        function will raise raise an ``IOError`` with an appropriate
        error message
    :rtype: :class:`~obspy.core.stream.Stream`
    :returns: Stream object
    """
    # Load shared library
    gcf_io = _load_cdll("gcf")
    # declare function argument and return types
    gcf_io.free_GcfFile.argtypes = [ctypes.POINTER(_GcfFile)]
    gcf_io.free_GcfFile.restype = None

    # set reader mode
    if headonly:
        if not blockmerge:
            mode = -2
        else:
            mode = -1
    else:
        if not blockmerge:
            mode = 2
        elif not cleanoverlap:
            mode = 1
        else:
            mode = 0

    # read file
    obj = _GcfFile()
    # b_filename = filename.encode('utf-8')
    # ret = gcf_io.read_gcf(fid, obj, mode)
    ret = _fill_gcf_from_file(file, obj, mode)

    if isinstance(file, IOBase):
        if ret:
            raise IOError("failed to read GCF data (error code %d)" % ret)
    else:
        filename = str(file)
        if ret == -1:
            if os.path.isfile(filename):
                raise IOError("cannot open file %s in read mode (missing "
                              "read permissions for user?)" % filename)
            else:
                raise IOError("%s is not a file" % filename)
        elif ret == 1:
            raise IOError("file %s is not a GCF data file" % filename)
        elif ret:
            raise IOError("failed to read file %s (unknown error code %d "
                          "returned)" % (filename, ret))

    err_msg = ""
    # So far so good, set up trace objects
    traces = []
    for i in range(int(obj.n_seg)):
        if obj.seg[i].err and not errorret:
            err_msg = _ERRORMSG_READ.get(
                obj.seg[i].err,
                "unknown error code (%d) set for data block %%d" %
                (obj.seg[i].err)
                ) % (i+1)
            break
        chn = obj.seg[i].streamID.decode('utf-8')[-2]
        channel = channel_prefix + chn.upper()
        if headonly or obj.seg[i].n_data <= 0:
            data = None
        else:
            data = np.array(obj.seg[i].data[:obj.seg[i].n_data],
                            dtype=np.int32)
        sps = obj.seg[i].sps*1./obj.seg[i].sps_denom
        dt = obj.seg[i].t_leap
        start = obj.seg[i].start+obj.seg[i].t_numerator * \
            1./obj.seg[i].t_denominator-dt
        def_stat = obj.seg[i].streamID[:4].decode('utf-8')
        stats = {
            "network": network,
            "station": station if station else def_stat,
            "location": location,
            "channel": channel,
            "sampling_rate": sps,
            "starttime": UTCDateTime(start),
            "npts": obj.seg[i].n_data,
            "gcf": AttribDict({
                "system_id": obj.seg[i].systemID.decode('utf-8'),
                "stream_id": obj.seg[i].streamID.decode('utf-8'),
                "sys_type": obj.seg[i].sysType,
                "t_leap": True if obj.seg[i].t_leap else False,
                "gain": obj.seg[i].gain,
                "digi": obj.seg[i].digi,
                "ttl": obj.seg[i].ttl,
                "blk": obj.seg[i].blk,
                "FIC": obj.seg[i].FIC,
                "RIC": obj.seg[i].RIC,
                "stat": obj.seg[i].err
            })
        }
        if data is not None:
            traces.append(Trace(data=data, header=stats))
        else:
            traces.append(Trace(header=stats))

    # Set up the stream object
    if not err_msg:
        stream = Stream(traces=traces)

    # Free memory allocated by the C-function
    gcf_io.free_GcfFile(obj)

    if err_msg:
        raise IOError(err_msg)
    return stream


def _write_gcf(stream, filename, stream_id=None, system_id=None, is_leap=False,
               gain=None, ttl=None, digi=None, sys_type=None, misalign=0.1,
               **kwargs):
    """
    Writes a :class:`~obspy.core.stream.Stream` or a
    :class:`~obspy.core.trace.Trace` to a GCF file

    Only GCF data records are supported. Function supports format as described
    by `GCF Reference <https://www.guralp.com/apps/ok?doc=GCF_format>`_
    SWA-RFC-GCFR Issue F, December 2021.

    .. warning::

            This function should NOT be called directly, it registers via the
            ObsPy :func:`~obspy.core.stream.write` function, call this
            instead.

    .. note::

        Sampling rate is restricted, most sampling rates between 1-250 samples
        per second (minus a few exceptions)are supported, for greater or lower
        sampling rates format support can be checked with function
        :func:`compatible_sps`.

        First data sample in each trace may only be sampled at non-integer
        second if sampling rate > 250. For sampling rates > 250 first data
        sample in each trace must start at an integer nominator of the
        denominator associated with the ampling rate. e.g. for a sampling rate
        of 1000 samples per second the associated denominator is 4 hence first
        data sample must be sampled at either ss.00 (0/4), ss.25 (1/4), ss.50
        (2/4), or ss.75 (3/4). Use function :func:`get_time_denominator` to get
        the associated denominator.

        The GCF format is only guaranteed to support 32-bit signed integer
        values.  While data with values out of range may be properly stored in
        the GCF format (if first and last data sample can be represented as a
        32-bit signed integer as well as all first difference values of the
        data vector) the current implementaion only permitts input data to be
        representable as a 32-bit signed integer. If input waveforms cannot be
        representable as 32-bit signed integers they will be clipped at
        -2,147,483,648 and 2,147,483,647

    .. rubric:: Exceptions

        ``TypeError`` will be raised upon unsupported input, ``IOError`` will
        be raised upon failure to output file

    .. rubric:: Example

        >>> st.write('GCF-filename.gcf', format='GCF')  # doctest: +SKIP

    :type stream: :class:`~obspy.core.stream.Stream` or
                  :class:`~obspy.core.trace.Trace`
    :param stream: waveform to write to file,
    :type filename: str
    :param filename: path to GCF file to read
    :type stream_id: str
    :param stream_id: 6-character stream ID, first 4 characters (the unit ID)
        should be in the range 0-9, A-Z fifth character should be the
        orientation code (e.g. Z, N, E) and the last character an integer
        yielding the tap the data were output from in the digitizer. If not
        specified stream_id will be taken from ``stats.gcf.stream_id`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` object if present else stream_id
        will be constructed as:
        ``stats.station.ljust(4, x)[:4]+stats.channel[-1]+'0'``
    :type system_id: str
    :param system_id: 4- to 6-character SysID (4 if ``sys_type`` is 2, 5 if
        ``sys_type`` is 1 and 6 if ``sys_type`` is 0). If not specified
       system_id  will be taken from ``stats.gcf.system_id`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` object and if present else system_id
        will be extracted from ``stream_id`` starting from first character.
    :type is_leap: int
    :param is_leap: should be set to ``True`` if first sample is sampled at an
        integer second that is a leap second. If not specified value will be
        taken from ``stats.gcf.t_leap`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` if present else the default value
        is ``0``
    :type gain: int
    :param gain: digitizer gain (permitted values are 2^n; n <= 6, and -1 if
        combined with ``sys_type`` = 0), if not specified gain will be taken
        from ``stats.gcf.gain`` for each :class:`~obspy.core.trace.Trace`
        object in the input  :class:`~obspy.core.stream.Stream` object if
        present else the default value is ``0``.
    :type ttl: int
    :param ttl: tap-table-lookup reference. if not specified ttl will be
        taken from t``stats.gcf.gain`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` object if present else the default
        value is ``0``
    :type digi: int
    :param digi: digitizer type (permitted values are 0 or 1), if not
        specified type will be taken from ``stats.gcf.digi`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` object if present else the default
        value is ``0``
    :type sys_type: int
    :param sys_type: system_id type (permitted values are 0, 1, or 2), if not
        specified sys_type will be taken from ``stats.gcf.sys_type`` for each
        :class:`~obspy.core.trace.Trace` object in the input
        :class:`~obspy.core.stream.Stream` object if present else the default
        value is ``0``
    :type misalign: float
    :param misalign: fraction of a sampling interval (permitted range is
        0-0.5) of tolerated misalignment of starttime. If not specified
        default value is ``0.1``
    """
    # avoid altering input object
    stream = stream.copy()

    # Make sure we have a Stream or Trace object
    if not isinstance(stream, Stream):
        if not isinstance(stream, Trace):
            raise TypeError("write requires either a Stream or a Trace object")
        else:
            stream = Stream(traces=stream)

    # Check output path
    if os.path.exists(filename) and not os.path.isfile(filename):
        IOError("path %s exists but is not a file object" % (filename))
    d, f = os.path.split(filename)
    if d and not os.path.exists(d):
        try:
            os.makedirs(d)
        except Exception:
            raise IOError("failed to create output directorie(s) %s" % (d))

    # check misalign parameter for supported range
    if misalign < 0 or misalign > 0.5:
        raise TypeError("argument misalign out of range (permitted range "
                        "0 - 0.5)")

    ret = 0
    # Load shared library
    gcf_io = _load_cdll("gcf")

    # declare function argument and return types
    gcf_io.write_gcf.argtypes = [ctypes.c_char_p, ctypes.POINTER(_GcfFile)]
    gcf_io.write_gcf.restype = ctypes.c_int

    # prepare the GcfFile object
    obj = _GcfFile()
    obj.n_seg = len(stream)
    obj.seg = (_GcfSeg * obj.n_seg)()
    obj.n_alloc = obj.n_seg
    ctypes.cast(obj.seg, ctypes.POINTER(_GcfSeg))

    for i, trace in enumerate(stream):
        gcf_stats = trace.stats.gcf if hasattr(trace.stats, 'gcf') \
            else None

        # Input checks, all of these are checked in the C-writer function
        #  but still duplicated here for clarity and ease of formatting
        #  proper exception messages

        # Check data type
        if not isinstance(trace.data[0], np.int32):
            if np.any(trace.data > _GCF_MAX) or \
                    np.any(trace.data < _GCF_MIN):
                msg = ('some samples were exceeding the data range that '
                       'GCF format can store (signed 32-bit integer), so '
                       'the data written to file will be clipped')
                warnings.warn(msg)
            trace.data = trace.data.clip(_GCF_MIN, _GCF_MAX)
            data = np.require(trace.data, dtype=np.int32,
                              requirements=['C_CONTIGUOUS'])
            if not np.all(data == trace.data):
                msg = ('data was downcast to int32 and data was altered '
                       'by this operation')
                warnings.warn(msg)
            trace.data = data
        else:
            trace.data = np.ascontiguousarray(trace.data)

        # Check sampling rate
        sps_denom = 1
        if not compatible_sps(trace.stats.sampling_rate):
            sps = ".3f" % (trace.stats.sampling_rate) if \
                           trace.stats.sampling_rate < 1 else \
                           "%d" % (trace.stats.sampling_rate)
            raise ValueError("trace sampling rate, %s, in trace %d not "
                             "supported in GCF format" % (sps, i+1))
        else:
            sps = trace.stats.sampling_rate
            if sps < 1:
                if sps < 0.12:
                    sps_denom = 10
                elif sps < 0.19:
                    sps_denom = 8
                elif sps < 0.24:
                    sps_denom = 5
                elif sps < 0.49:
                    sps_denom = 4
                else:
                    sps_denom = 2
                sps = 1
            sps = int(sps)

        # check start time
        sec = trace.stats.starttime.timestamp
        t_numerator = 0
        if sps <= 250:
            # fractional start time not supported, allow for rounding of
            #  up 10% of the sampling interval
            # hmm is this enough for sampling rates < 1 Hz
            start = int(round(sec))
            if abs(start-sec) > 1./sps*misalign:
                raise ValueError("fractional start time not supported for "
                                 "sampling rates <= 250 Hz (in trace %d)" %
                                 (i+1))
        else:
            # fractional start time supported but restricted, allow for
            # 10% misalignment
            start = int(math.floor(sec))
            dt = sec-start
            t_denom = _SPS_MAP[sps]
            numerator = dt*t_denom
            t_numerator = int(round(numerator))
            if abs(numerator-t_numerator)/t_denom > misalign/sps:
                raise ValueError("start time in trace %d not aligned with "
                                 "supported (fractional) start time" %
                                 (i+1))

        # Check if is_leap is set else set
        use_is_leap = 0
        if is_leap is not None:
            use_is_leap = 1 if is_leap else 0
        elif hasattr(gcf_stats, "is_leap"):
            use_is_leap = 1 if gcf_stats.is_leap else 0

        # Check if gain is set else set
        use_gain = 0
        gain_in_stats = False
        if gain is not None:
            use_gain = gain
        elif hasattr(gcf_stats, "gain"):
            gain_in_stats = True
            use_gain = gcf_stats.gain
        if use_gain not in _VALID_GAIN:
            if gain is None:
                gains = ', '.join(["%d" % g for g in _VALID_GAIN])
                raise ValueError("bad value on stats.gcf.gain, %s, in "
                                 "stats.gcf in trace %d (permitted values:"
                                 " %s)" % (use_gain, i+1, gains))
            else:
                gains = ', '.join(["%d" % g for g in _VALID_GAIN])
                raise ValueError("bad value on argument gain, %s "
                                 "(permitted values: %s)" %
                                 (use_gain, gains))

        # Check if ttl is set else set
        use_ttl = 0
        if ttl is not None:
            use_ttl = ttl
        elif hasattr(gcf_stats, "ttl"):
            use_ttl = gcf_stats.ttl

        # Check if type is set else set
        use_digi = 0
        if digi is not None:
            use_digi = digi
        elif hasattr(gcf_stats, "digi"):
            use_digi = gcf_stats.digi
        if use_digi not in [0, 1]:
            if digi is None:
                raise ValueError("bad value on stats.gcf.digi, %s, in "
                                 "trace %d (permitted values: 0, 1)" %
                                 (use_digi, i+1))
            else:
                raise ValueError("bad value on argument digi, %s "
                                 "(permitted values: 0, 1)" % (use_digi))

        # Check if sys_type is set else set
        use_sys_type = 0
        st_in_stats = False
        if sys_type is not None:
            use_sys_type = sys_type
        elif hasattr(gcf_stats, "sys_type"):
            st_in_stats = True
            use_sys_type = gcf_stats.sys_type
        if use_sys_type not in [0, 1, 2]:
            if digi is None:
                raise ValueError("bad value on stats.gcf.digi, %s, in "
                                 "trace %d (permitted values: 0, 1, 2)"
                                 % (use_sys_type, i+1))
            else:
                raise ValueError("bad value on argument digi, %s, "
                                 "in trace %d (permitted values: 0, 1, 2)"
                                 % (use_sys_type, i+1))
        elif use_sys_type != 0 and use_gain == -1:
            if gain_in_stats and st_in_stats:
                raise ValueError(
                    "value -1 on stats.gcf.gain may only be "
                    "combined with value 0 on "
                    "stats.gcf.sys_type (in trace %s)" % (i+1))
            elif gain_in_stats:
                raise ValueError("value -1 on stats.gcf.gain may only be "
                                 "combined with value 0 on argument "
                                 "sys_type (in trace %s)" % (i+1))
            elif st_in_stats:
                raise ValueError(
                    "value -1 on argument gain may only be "
                    "combined with value 0 on "
                    "stats.gcf.sys_type (in trace %s)" % (i+1))
            else:
                raise ValueError("value -1 on argument gain may only be "
                                 "combined with value 0 on argument "
                                 "sys_type")

        # Check if stream_id is set else build
        if stream_id is not None:
            use_stream_id = stream_id
        elif gcf_stats is not None and hasattr(gcf_stats, "stream_id"):
            use_stream_id = gcf_stats.stream_id.upper()
        else:
            use_stream_id = (trace.stats.station.ljust(4, "X")[:4] +
                             (trace.stats.channel[-1]
                              if trace.stats.channel
                              else 'X')+'0').upper()
        if len(use_stream_id) != 6:
            if stream_id is None:
                raise ValueError("bad value on stats.gcf.stream_id, %s, "
                                 "in trace %d (must be 6-character long)" %
                                 (use_stream_id, i+1))
            else:
                raise ValueError(
                    "bad value on argument stream_id, %s (must"
                    " be 6-character long)" % (use_stream_id))

        # Check if system_id is set else set
        len_system_id = 6 if use_sys_type == 0 \
            else (5 if use_sys_type == 1 else 4)
        use_system_id = use_stream_id[:len_system_id]
        si_in_stats = False
        if system_id is not None:
            use_system_id = system_id.upper()
        elif gcf_stats is not None and hasattr(gcf_stats, "system_id"):
            si_in_stats = True
            use_system_id = gcf_stats.system_id.upper()
        if len(use_system_id) > len_system_id:
            if st_in_stats and si_in_stats:
                raise ValueError("stats.gcf.system_id not compatible with "
                                 "stats.gcf.sys_type in trace %d (max "
                                 "length %d characters )" %
                                 (i+1, len_system_id))
            elif st_in_stats:
                raise ValueError("argument system_id not compatible with "
                                 "stats.gcf.sys_type in trace %d (max "
                                 "length %d characters )" %
                                 (i+1, len_system_id))
            elif si_in_stats:
                raise ValueError("stats.gcf.system_id not compatible with "
                                 "argument sys_type in trace %d (max length"
                                 " %d characters )" % (i+1, len_system_id))
            else:
                raise ValueError("argument system_id not compatible with "
                                 "argument sys_type (max length %d "
                                 "characters )" % (len_system_id))

        # Populate segment
        tr = _GcfSeg()
        tr.streamID = use_stream_id.encode('utf-8')
        tr.systemID = use_system_id.encode('utf-8')
        tr.start = start
        tr.t_numerator = t_numerator
        tr.t_leap = use_is_leap
        tr.gain = use_gain
        tr.sysType = use_sys_type
        tr.digi = use_digi
        tr.ttl = use_ttl
        tr.sps = sps
        tr.sps_denom = sps_denom
        tr.n_data = trace.data.size
        tr.n_alloc = trace.data.size
        tr.data = trace.data.ctypes.data_as(ctypes.POINTER(ctypes.c_int32))

        # Add segment to file object
        obj.seg[i] = tr

    # Write to file
    ret = gcf_io.write_gcf(filename.encode('utf-8'), obj)
    if ret == 0:
        pass
    elif ret in _ERROR_WRITE:
        raise _ERROR_WRITE['ret']
    else:
        raise IOError("unknown error code %d" % (ret))