miriad.py

"""Class for reading and writing Miriad files."""
from astropy import constants as const
import os
import shutil
import numpy as np
import copy
import warnings
import aipy
from uvdata import UVData
import telescopes as uvtel
import utils as uvutils
import itertools


class Miriad(UVData):
    """
    Defines a Miriad-specific subclass of UVData for reading and writing Miriad files.
    This class should not be interacted with directly, instead use the read_miriad
    and write_miriad methods on the UVData class.
    """

    def _pol_to_ind(self, pol):
        if self.polarization_array is None:
            raise(ValueError, "Can't index polarization {p} because "
                  "polarization_array is not set".format(p=pol))
        pol_ind = np.argwhere(self.polarization_array == pol)
        if len(pol_ind) != 1:
            raise(ValueError, "multiple matches for pol={pol} in "
                  "polarization_array".format(pol=pol))
        return pol_ind

    def read_miriad(self, filepath, correct_lat_lon=True, run_check=True,
                    check_extra=True, run_check_acceptability=True, phase_type=None,
                    antenna_nums=None, ant_str=None, ant_pairs_nums=None,
                    polarizations=None, time_range=None):
        """
        Read in data from a miriad file.

        Args:
            filepath: The miriad file directory to read from.
            correct_lat_lon: flag -- that only matters if altitude is missing --
                to update the latitude and longitude from the known_telescopes list
            run_check: Option to check for the existence and proper shapes of
                parameters after reading in the file. Default is True.
            check_extra: Option to check optional parameters as well as required
                ones. Default is True.
            run_check_acceptability: Option to check acceptable range of the values of
                parameters after reading in the file. Default is True.
            antenna_nums: The antennas numbers to only read into the object.
            ant_pairs_nums: A list of antenna number tuples (e.g. [(0,1), (3,2)])
                specifying baselines to read into the object. Ordering of the
                numbers within the tuple does not matter. A single antenna iterable
                e.g. (1,) is interpreted as all visibilities with that antenna.
            ant_str: A string containing information about what kinds of visibility data
                to read-in.  Can be 'auto', 'cross', 'all'. Cannot provide ant_str if
                antenna_nums and/or ant_pairs_nums is not None.
            polarizations: List of polarization integers or strings to read-in.
                Ex: ['xx', 'yy', ...]
            time_range: len-2 list containing min and max range of times (Julian Date) to read-in.
                Ex: [2458115.20, 2458115.40]
        """
        if not os.path.exists(filepath):
            raise(IOError, filepath + ' not found')
        uv = aipy.miriad.UV(filepath)

        # list of miriad variables always read
        # NB: this includes variables in try/except (i.e. not all variables are
        # necessarily present in the miriad file)
        default_miriad_variables = ['nchan', 'npol', 'inttime', 'sdf',
                                    'source', 'telescop', 'latitud', 'longitu',
                                    'altitude', 'history', 'visunits',
                                    'instrume', 'dut1', 'gst0', 'rdate',
                                    'timesys', 'xorient', 'cnt', 'ra', 'dec',
                                    'lst', 'pol', 'nants', 'antnames', 'nblts',
                                    'ntimes', 'nbls', 'sfreq', 'epoch',
                                    'antpos', 'antnums', 'degpdy', 'antdiam',
                                    ]
        # list of miriad variables not read, but also not interesting
        # NB: nspect (I think) is number of spectral windows, will want one day
        # NB: xyphase & xyamp are "On-line X Y phase/amplitude measurements" which we may want in
        #     a calibration object some day
        # NB: systemp, xtsys & ytsys are "System temperatures of the antenna/X/Y feed"
        #     which we may want in a calibration object some day
        # NB: freqs, leakage and bandpass may be part of a calibration object some day
        other_miriad_variables = ['nspect', 'obsdec', 'vsource', 'ischan',
                                  'restfreq', 'nschan', 'corr', 'freq',
                                  'freqs', 'leakage', 'bandpass',
                                  'tscale', 'coord', 'veldop', 'time', 'obsra',
                                  'operator', 'version', 'axismax', 'axisrms',
                                  'xyphase', 'xyamp', 'systemp', 'xtsys', 'ytsys',
                                  'baseline']

        extra_miriad_variables = []
        for variable in uv.vars():
            if (variable not in default_miriad_variables
                    and variable not in other_miriad_variables):
                extra_miriad_variables.append(variable)

        miriad_header_data = {'Nfreqs': 'nchan',
                              'Npols': 'npol',
                              'integration_time': 'inttime',
                              'channel_width': 'sdf',  # in Ghz!
                              'object_name': 'source',
                              'telescope_name': 'telescop'
                              }
        for item in miriad_header_data:
            if isinstance(uv[miriad_header_data[item]], str):
                header_value = uv[miriad_header_data[item]].replace('\x00', '')
            else:
                header_value = uv[miriad_header_data[item]]
            setattr(self, item, header_value)

        latitude = uv['latitud']  # in units of radians
        longitude = uv['longitu']
        try:
            altitude = uv['altitude']
            self.telescope_location_lat_lon_alt = (latitude, longitude, altitude)
        except(KeyError):
            # get info from known telescopes. Check to make sure the lat/lon values match reasonably well
            telescope_obj = uvtel.get_telescope(self.telescope_name)
            if telescope_obj is not False:

                tol = 2 * np.pi * 1e-3 / (60.0 * 60.0 * 24.0)  # 1mas in radians
                lat_close = np.isclose(telescope_obj.telescope_location_lat_lon_alt[0],
                                       latitude, rtol=0, atol=tol)
                lon_close = np.isclose(telescope_obj.telescope_location_lat_lon_alt[1],
                                       longitude, rtol=0, atol=tol)
                if correct_lat_lon:
                    self.telescope_location_lat_lon_alt = telescope_obj.telescope_location_lat_lon_alt
                else:
                    self.telescope_location_lat_lon_alt = (latitude, longitude, telescope_obj.telescope_location_lat_lon_alt[2])
                if lat_close and lon_close:
                    if correct_lat_lon:
                        warnings.warn('Altitude is not present in Miriad file, '
                                      'using known location values for '
                                      '{telescope_name}.'.format(telescope_name=telescope_obj.telescope_name))
                    else:
                        warnings.warn('Altitude is not present in Miriad file, '
                                      'using known location altitude value '
                                      'for {telescope_name} and lat/lon from '
                                      'file.'.format(telescope_name=telescope_obj.telescope_name))
                else:
                    warn_string = ('Altitude is not present in file ')
                    if not lat_close and not lon_close:
                        warn_string = warn_string + 'and latitude and longitude values do not match values '
                    else:
                        if not lat_close:
                            warn_string = warn_string + 'and latitude value does not match value '
                        else:
                            warn_string = warn_string + 'and longitude value does not match value '
                    if correct_lat_lon:
                        warn_string = (warn_string + 'for {telescope_name} in known '
                                       'telescopes. Using values from known '
                                       'telescopes.'.format(telescope_name=telescope_obj.telescope_name))
                        warnings.warn(warn_string)
                    else:
                        warn_string = (warn_string + 'for {telescope_name} in known '
                                       'telescopes. Using altitude value from known '
                                       'telescopes and lat/lon from '
                                       'file.'.format(telescope_name=telescope_obj.telescope_name))
                        warnings.warn(warn_string)
            else:
                warnings.warn('Altitude is not present in Miriad file, and '
                              'telescope {telescope_name} is not in '
                              'known_telescopes. Telescope location will be '
                              'set using antenna positions.'
                              .format(telescope_name=self.telescope_name))

        self.history = uv['history']
        if not uvutils.check_history_version(self.history, self.pyuvdata_version_str):
            self.history += self.pyuvdata_version_str
        self.channel_width *= 1e9  # change from GHz to Hz

        # check for pyuvdata variables that are not recognized miriad variables
        if 'visunits' in uv.vartable.keys():
            self.vis_units = uv['visunits'].replace('\x00', '')
        else:
            self.vis_units = 'UNCALIB'  # assume no calibration
        if 'instrume' in uv.vartable.keys():
            self.instrument = uv['instrume'].replace('\x00', '')
        else:
            self.instrument = self.telescope_name  # set instrument = telescope

        if 'dut1' in uv.vartable.keys():
            self.dut1 = uv['dut1']
        if 'degpdy' in uv.vartable.keys():
            self.earth_omega = uv['degpdy']
        if 'gst0' in uv.vartable.keys():
            self.gst0 = uv['gst0']
        if 'rdate' in uv.vartable.keys():
            self.rdate = uv['rdate'].replace('\x00', '')
        if 'timesys' in uv.vartable.keys():
            self.timesys = uv['timesys'].replace('\x00', '')
        if 'xorient' in uv.vartable.keys():
            self.x_orientation = uv['xorient'].replace('\x00', '')

        # read through the file and get the data
        _source = uv['source']  # check source of initial visibility
        # dict of extra variables to see if they change
        check_variables = {}
        for extra_variable in extra_miriad_variables:
            check_variables[extra_variable] = uv[extra_variable]

        history_update_string = '  Downselected to specific '
        n_selects = 0

        # select on ant_str if provided
        if ant_str is not None:
            # type check
            assert isinstance(ant_str, (str, np.str)), "ant_str must be fed as a string"
            assert antenna_nums is None and ant_pairs_nums is None, "ant_str must be None if antenna_nums or ant_pairs_nums is not None"
            aipy.scripting.uv_selector(uv, ant_str)
            if ant_str != 'all':
                history_update_string += 'antenna pairs'
                n_selects += 1
        # select on antenna_nums and/or ant_pairs_nums using aipy.scripting.uv_selector
        if antenna_nums is not None or ant_pairs_nums is not None:
            antpair_str = ''
            if antenna_nums is not None:
                # type check
                err_msg = "antenna_nums must be fed as a list of antenna number integers"
                assert isinstance(antenna_nums, (np.ndarray, list)), err_msg
                assert isinstance(antenna_nums[0], (int, np.int, np.int32)), err_msg
                # get all possible combinations
                antpairs = list(itertools.combinations_with_replacement(antenna_nums, 2))
                # convert antenna numbers to string form required by aipy.scripting.uv_selector
                antpair_str += ','.join(map(lambda ap: '_'.join(map(lambda a: str(a), ap)), antpairs))
                history_update_string += 'antennas'
                n_selects += 1
            if ant_pairs_nums is not None:
                # type check
                err_msg = "ant_pairs_nums must be a list of antnum integer tuples, Ex: [(0, 1), ...]"
                assert isinstance(ant_pairs_nums, list), err_msg
                assert np.array(map(lambda ap: isinstance(ap, tuple), ant_pairs_nums)).all(), err_msg
                assert np.array(map(lambda ap: map(lambda a: isinstance(a, (int, np.int, np.int32)), ap), ant_pairs_nums)).all(), err_msg
                # convert ant-pair tuples to string form required by aipy.scripting.uv_selector
                if len(antpair_str) > 0:
                    antpair_str += ','
                antpair_str += ','.join(map(lambda ap: '_'.join(map(lambda a: str(a), ap)), ant_pairs_nums))
                if n_selects > 0:
                    history_update_string += ', antenna pairs'
                else:
                    history_update_string += 'antenna pairs'
                n_selects += 1
            aipy.scripting.uv_selector(uv, antpair_str)

        # select on time range
        if time_range is not None:
            # type check
            err_msg = "time_range must be a len-2 list of Julian Date floats, Ex: [2458115.2, 2458115.6]"
            assert isinstance(time_range, (list, np.ndarray)), err_msg
            assert len(time_range) == 2, err_msg
            assert np.array(map(lambda t: isinstance(t, (float, np.float, np.float64)), time_range)).all(), err_msg
            uv.select('time', time_range[0], time_range[1], include=True)
            if n_selects > 0:
                history_update_string += ', times'
            else:
                history_update_string += 'times'
            n_selects += 1

        # select on polarizations
        if polarizations is not None:
            # type check
            err_msg = "pols must be a list of polarization strings or ints, Ex: ['xx', ...] or [-5, ...]"
            assert isinstance(polarizations, (list, np.ndarray)), err_msg
            assert np.array(map(lambda p: isinstance(p, (str, np.str, int, np.int, np.int32)), polarizations)).all(), err_msg
            # convert to pol integer if string
            polarizations = [p if isinstance(p, (int, np.int, np.int32)) else uvutils.polstr2num(p) for p in polarizations]
            # iterate through all possible pols and reject if not in pols
            pol_list = []
            for p in np.arange(-8, 5):
                if p not in polarizations:
                    uv.select('polarization', p, p, include=False)
                else:
                    pol_list.append(p)
            # assert not empty
            assert len(pol_list) > 0, "No polarizations in data matched {}".format(polarizations)
            if n_selects > 0:
                history_update_string += ', polarizations'
            else:
                history_update_string += 'polarizations'
            n_selects += 1

        history_update_string += ' using pyuvdata.'
        if n_selects > 0:
            self.history += history_update_string

        data_accumulator = {}
        pol_list = []
        for (uvw, t, (i, j)), d, f in uv.all(raw=True):
            # control for the case of only a single spw not showing up in
            # the dimension
            # Note that the (i, j) tuple is calculated from a baseline number in
            # aipy (see miriad_wrap.h). The i, j values are also adjusted by aipy
            # to start at 0 rather than 1.
            if len(d.shape) == 1:
                d.shape = (1,) + d.shape
                self.Nspws = d.shape[0]
                self.spw_array = np.arange(self.Nspws)
            else:
                raise(ValueError, """Sorry.  Files with more than one spectral
                      window (spw) are not yet supported. A great
                      project for the interested student!""")
            try:
                cnt = uv['cnt']
            except(KeyError):
                cnt = np.ones(d.shape, dtype=np.float)
            ra = uv['ra']
            dec = uv['dec']
            lst = uv['lst']
            source = uv['source']
            if source != _source:
                raise(ValueError, 'This appears to be a multi source file, which is not supported.')
            else:
                _source = source

            # check extra variables for changes compared with initial value
            for extra_variable in check_variables.keys():
                if type(check_variables[extra_variable]) == str:
                    if uv[extra_variable] != check_variables[extra_variable]:
                        check_variables.pop(extra_variable)
                else:
                    if not np.allclose(uv[extra_variable],
                                       check_variables[extra_variable]):
                        check_variables.pop(extra_variable)

            try:
                data_accumulator[uv['pol']].append([uvw, t, i, j, d, f, cnt,
                                                    ra, dec])
            except(KeyError):
                data_accumulator[uv['pol']] = [[uvw, t, i, j, d, f, cnt,
                                                ra, dec]]
                pol_list.append(uv['pol'])
                # NB: flag types in miriad are usually ints

        if len(data_accumulator.keys()) == 0:
            raise ValueError('No data is present, probably as a result of '
                             'select on read that excludes all the data')

        # keep all single valued extra_variables as extra_keywords
        for key in check_variables.keys():
            if type(check_variables[key]) == str:
                value = check_variables[key].replace('\x00', '')
                # check for booleans encoded as strings
                if value == 'True':
                    value = True
                elif value == 'False':
                    value = False
                self.extra_keywords[key] = value
            else:
                self.extra_keywords[key] = check_variables[key]

        # Check for items in itemtable to put into extra_keywords
        # These will end up as variables in written files, but is internally consistent.
        for key in uv.items():
            # A few items that are not needed, we read elsewhere, or is not supported
            # when downselecting, so we don't read here.
            if key not in ['vartable', 'history', 'obstype'] and key not in other_miriad_variables:
                if type(uv[key]) == str:
                    value = uv[key].replace('\x00', '')
                    value = uv[key].replace('\x01', '')
                    if value == 'True':
                        value = True
                    elif value == 'False':
                        value = False
                    self.extra_keywords[key] = value
                else:
                    self.extra_keywords[key] = uv[key]

        for pol, data in data_accumulator.iteritems():
            data_accumulator[pol] = np.array(data)

        self.polarization_array = np.array(pol_list)
        if polarizations is None:
            # A select on read would make the header npols not match the pols in the data
            if len(self.polarization_array) != self.Npols:
                warnings.warn('npols={npols} but found {n} pols in data file'.format(
                              npols=self.Npols, n=len(self.polarization_array)))
        self.Npols = len(pol_list)

        # makes a data_array (and flag_array) of zeroes to be filled in by
        #   data values
        # any missing data will have zeros

        # use set to get the unique list of all times ever listed in the file
        # iterate over polarizations and all spectra (bls and times) in two
        # nested loops, then flatten into a single vector, then set
        # then list again.

        times = list(set(
            np.concatenate([[k[1] for k in d] for d in data_accumulator.values()])))
        times = np.sort(times)

        ant_i_unique = list(set(
            np.concatenate([[k[2] for k in d] for d in data_accumulator.values()])))
        ant_j_unique = list(set(
            np.concatenate([[k[3] for k in d] for d in data_accumulator.values()])))

        sorted_unique_ants = sorted(list(set(ant_i_unique + ant_j_unique)))
        ant_i_unique = np.array(ant_i_unique)
        ant_j_unique = np.array(ant_j_unique)

        # Determine maximum digits needed to distinguish different values
        if sorted_unique_ants[-1] > 0:
            ndig_ant = np.ceil(np.log10(sorted_unique_ants[-1])).astype(int) + 1
        else:
            ndig_ant = 1
        # Be excessive in precision because we use the floating point values as dictionary keys later
        prec_t = - 2 * np.floor(np.log10(self._time_array.tols[-1])).astype(int)
        ndig_t = (np.ceil(np.log10(times[-1])).astype(int) + prec_t + 2)
        blts = []
        for d in data_accumulator.values():
            for k in d:
                blt = ["{1:.{0}f}".format(prec_t, k[1]).zfill(ndig_t),
                       str(k[2]).zfill(ndig_ant), str(k[3]).zfill(ndig_ant)]
                blt = "_".join(blt)
                blts.append(blt)
        unique_blts = np.unique(np.array(blts))

        reverse_inds = dict(zip(unique_blts, range(len(unique_blts))))
        self.Nants_data = len(sorted_unique_ants)

        # Miriad has no way to keep track of antenna numbers, so the antenna
        # numbers are simply the index for each antenna in any array that
        # describes antenna attributes (e.g. antpos for the antenna_postions).
        # Therefore on write, nants (which gives the size of the antpos array)
        # needs to be increased to be the max value of antenna_numbers+1 and the
        # antpos array needs to be inflated with zeros at locations where we
        # don't have antenna information. These inflations need to be undone at
        # read. If the file was written by pyuvdata, then the variable antnums
        # will be present and we can use it, otherwise we need to test for zeros
        # in the antpos array and/or antennas with no visibilities.
        try:
            # The antnums variable will only exist if the file was written with pyuvdata.
            # For some reason Miriad doesn't handle an array of integers properly,
            # so we convert to floats on write and back here
            self.antenna_numbers = uv['antnums'].astype(int)
            self.Nants_telescope = len(self.antenna_numbers)
        except(KeyError):
            self.antenna_numbers = None
            self.Nants_telescope = None

        nants = uv['nants']
        try:
            # Miriad stores antpos values in units of ns, pyuvdata uses meters.
            antpos = uv['antpos'].reshape(3, nants).T * const.c.to('m/ns').value

            # first figure out what are good antenna positions so we can only
            # use the non-zero ones to evaluate position information
            antpos_length = np.sqrt(np.sum(np.abs(antpos)**2, axis=1))
            good_antpos = np.where(antpos_length > 0)[0]
            mean_antpos_length = np.mean(antpos_length[good_antpos])
            if mean_antpos_length > 6.35e6 and mean_antpos_length < 6.39e6:
                absolute_positions = True
            else:
                absolute_positions = False

            # Miriad stores antpos values in a rotated ECEF coordinate system
            # where the x-axis goes through the local meridan. Need to convert
            # these positions back to standard ECEF and if they are absolute positions,
            # subtract off the telescope position to make them relative to the array center.
            ecef_antpos = uvutils.ECEF_from_rotECEF(antpos, longitude)

            if self.telescope_location is not None:
                if absolute_positions:
                    rel_ecef_antpos = ecef_antpos - self.telescope_location
                    # maintain zeros because they mark missing data
                    rel_ecef_antpos[np.where(antpos_length == 0)[0]] = ecef_antpos[np.where(antpos_length == 0)[0]]
                else:
                    rel_ecef_antpos = ecef_antpos
            else:
                self.telescope_location = np.mean(ecef_antpos[good_antpos, :], axis=0)
                valid_location = self._telescope_location.check_acceptability()[0]

                # check to see if this could be a valid telescope_location
                if valid_location:
                    mean_lat, mean_lon, mean_alt = self.telescope_location_lat_lon_alt
                    tol = 2 * np.pi / (60.0 * 60.0 * 24.0)  # 1 arcsecond in radians
                    mean_lat_close = np.isclose(mean_lat, latitude, rtol=0, atol=tol)
                    mean_lon_close = np.isclose(mean_lon, longitude, rtol=0, atol=tol)

                    if mean_lat_close and mean_lon_close:
                        # this looks like a valid telescope_location, and the
                        # mean antenna lat & lon values are close. Set the
                        # telescope_location using the file lat/lons and the mean alt.
                        # Then subtract it off of the antenna positions
                        warnings.warn('Telescope location is not set, but antenna '
                                      'positions are present. Mean antenna latitude and '
                                      'longitude values match file values, so '
                                      'telescope_position will be set using the '
                                      'mean of the antenna altitudes')
                        self.telescope_location_lat_lon_alt = (latitude, longitude, mean_alt)
                        rel_ecef_antpos = ecef_antpos - self.telescope_location

                    else:
                        # this looks like a valid telescope_location, but the
                        # mean antenna lat & lon values are not close. Set the
                        # telescope_location using the file lat/lons at sea level.
                        # Then subtract it off of the antenna positions
                        self.telescope_location_lat_lon_alt = (latitude, longitude, 0)
                        warn_string = ('Telescope location is set at sealevel at '
                                       'the file lat/lon coordinates. Antenna '
                                       'positions are present, but the mean '
                                       'antenna ')
                        rel_ecef_antpos = ecef_antpos - self.telescope_location

                        if not mean_lat_close and not mean_lon_close:
                            warn_string += ('latitude and longitude values do not '
                                            'match file values so they are not used '
                                            'for altiude.')
                        elif not mean_lat_close:
                            warn_string += ('latitude value does not '
                                            'match file values so they are not used '
                                            'for altiude.')
                        else:
                            warn_string += ('longitude value does not '
                                            'match file values so they are not used '
                                            'for altiude.')

                        warnings.warn(warn_string)

                else:
                    # This does not give a valid telescope_location. Instead
                    # calculate it from the file lat/lon and sea level for altiude
                    self.telescope_location_lat_lon_alt = (latitude, longitude, 0)
                    warn_string = ('Telescope location is set at sealevel at '
                                   'the file lat/lon coordinates. Antenna '
                                   'positions are present, but the mean '
                                   'antenna ')

                    warn_string += ('position does not give a '
                                    'telescope_location on the surface of the '
                                    'earth.')
                    if absolute_positions:
                        rel_ecef_antpos = ecef_antpos - self.telescope_location
                    else:
                        warn_string += (' Antenna positions do not appear to be '
                                        'on the surface of the earth and will be treated '
                                        'as relative.')
                        rel_ecef_antpos = ecef_antpos

                    warnings.warn(warn_string)

            if self.Nants_telescope is not None:
                # in this case there is an antnums variable
                # (meaning that the file was written with pyuvdata), so we'll use it
                if nants == self.Nants_telescope:
                    # no inflation, so just use the positions
                    self.antenna_positions = rel_ecef_antpos
                else:
                    # there is some inflation, just use the ones that appear in antnums
                    self.antenna_positions = np.zeros((self.Nants_telescope, 3), dtype=antpos.dtype)
                    for ai, num in enumerate(self.antenna_numbers):
                        self.antenna_positions[ai, :] = rel_ecef_antpos[num, :]
            else:
                # there is no antnums variable (meaning that this file was not
                # written by pyuvdata), so we test for antennas with non-zero
                # positions and/or that appear in the visibility data
                # (meaning that they have entries in ant_1_array or ant_2_array)
                antpos_length = np.sqrt(np.sum(np.abs(antpos)**2, axis=1))
                good_antpos = np.where(antpos_length > 0)[0]
                # take the union of the antennas with good positions (good_antpos)
                # and the antennas that have visisbilities (sorted_unique_ants)
                # if there are antennas with visibilities but zeroed positions we issue a warning below
                ants_use = set(good_antpos).union(sorted_unique_ants)
                # ants_use are the antennas we'll keep track of in the UVData
                # object, so they dictate Nants_telescope
                self.Nants_telescope = len(ants_use)
                self.antenna_numbers = np.array(list(ants_use))
                self.antenna_positions = np.zeros((self.Nants_telescope, 3), dtype=rel_ecef_antpos.dtype)
                for ai, num in enumerate(self.antenna_numbers):
                    if antpos_length[num] == 0:
                        warnings.warn('antenna number {n} has visibilities '
                                      'associated with it, but it has a position'
                                      ' of (0,0,0)'.format(n=num))
                    else:
                        # leave bad locations as zeros to make them obvious
                        self.antenna_positions[ai, :] = rel_ecef_antpos[num, :]

        except(KeyError):
            # there is no antpos variable
            warnings.warn('Antenna positions are not present in the file.')
            self.antenna_positions = None

        if self.antenna_numbers is None:
            # there are no antenna_numbers or antenna_positions, so just use
            # the antennas present in the visibilities
            # (Nants_data will therefore match Nants_telescope)
            self.antenna_numbers = np.array(sorted_unique_ants)
            self.Nants_telescope = len(self.antenna_numbers)

        # antenna names is a foreign concept in miriad but required in other formats.
        try:
            # Here we deal with the way pyuvdata tacks it on to keep the
            # name information if we have it:
            # make it into one long comma-separated string
            ant_name_var = uv['antnames']
            if isinstance(ant_name_var, str):
                ant_name_str = ant_name_var.replace('\x00', '')
                ant_name_list = ant_name_str[1:-1].split(', ')
                self.antenna_names = ant_name_list
            else:
                # Backwards compatibility for old way of storing antenna_names.
                # This is a horrible hack to save & recover antenna_names array.
                # Miriad can't handle arrays of strings and AIPY use to not handle
                # long enough single strings to put them all into one string
                # so we convert them into hex values and then into floats on
                # write and convert back to strings here
                warnings.warn('{file} was written with an old version of '
                              'pyuvdata, which has been deprecated. Rewrite this '
                              'file with write_miriad to ensure future '
                              'compatibility'.format(file=filepath))
                ant_name_flt = uv['antnames']
                ant_name_list = [('%x' % elem.astype(np.int64)).decode('hex') for elem in ant_name_flt]
                self.antenna_names = ant_name_list

        except(KeyError):
            self.antenna_names = self.antenna_numbers.astype(str).tolist()

        # check for antenna diameters
        try:
            self.antenna_diameters = uv['antdiam']
        except(KeyError):
            # backwards compatibility for when keyword was 'diameter'
            try:
                self.antenna_diameters = uv['diameter']
                # if we find it, we need to remove it from extra_keywords to keep from writing it out
                self.extra_keywords.pop('diameter')
            except(KeyError):
                pass
        if self.antenna_diameters is not None:
            self.antenna_diameters = (self.antenna_diameters
                                      * np.ones(self.Nants_telescope, dtype=np.float))

        # form up a grid which indexes time and baselines along the 'long'
        # axis of the visdata array

        tij_grid = np.array(map(lambda x: map(float, x.split("_")), unique_blts))
        t_grid, ant_i_grid, ant_j_grid = tij_grid.T
        # set the data sizes
        if antenna_nums is None and ant_pairs_nums is None and ant_str is None and time_range is None:
            try:
                self.Nblts = uv['nblts']
                if self.Nblts != len(t_grid):
                    warnings.warn('Nblts does not match the number of unique blts in the data')
                    self.Nblts = len(t_grid)
            except(KeyError):
                self.Nblts = len(t_grid)
        else:
            # The select on read will make the header nblts not match the number of unique blts
            self.Nblts = len(t_grid)
        if time_range is None:
            try:
                self.Ntimes = uv['ntimes']
                if self.Ntimes != len(times):
                    warnings.warn('Ntimes does not match the number of unique times in the data')
                    self.Ntimes = len(times)
            except(KeyError):
                self.Ntimes = len(times)
        else:
            # The select on read will make the header ntimes not match the number of unique times
            self.Ntimes = len(times)

        self.time_array = t_grid
        self.ant_1_array = ant_i_grid.astype(int)
        self.ant_2_array = ant_j_grid.astype(int)

        self.baseline_array = self.antnums_to_baseline(ant_i_grid.astype(int),
                                                       ant_j_grid.astype(int))
        if antenna_nums is None and ant_pairs_nums is None and ant_str is None:
            try:
                self.Nbls = uv['nbls']
                if self.Nbls != len(np.unique(self.baseline_array)):
                    warnings.warn('Nbls does not match the number of unique baselines in the data')
                    self.Nbls = len(np.unique(self.baseline_array))
            except(KeyError):
                self.Nbls = len(np.unique(self.baseline_array))
        else:
            # The select on read will make the header nbls not match the number of unique bls
            self.Nbls = len(np.unique(self.baseline_array))

        # slot the data into a grid
        self.data_array = np.zeros((self.Nblts, self.Nspws, self.Nfreqs,
                                    self.Npols), dtype=np.complex64)
        self.flag_array = np.ones(self.data_array.shape, dtype=np.bool)
        self.uvw_array = np.zeros((self.Nblts, 3))
        # NOTE: Using our lst calculator, which uses astropy,
        # instead of aipy values which come from pyephem.
        # The differences are of order 5 seconds.
        if self.telescope_location is not None:
            self.set_lsts_from_time_array()
        self.nsample_array = np.ones(self.data_array.shape, dtype=np.float)
        self.freq_array = (np.arange(self.Nfreqs) * self.channel_width
                           + uv['sfreq'] * 1e9)
        # Tile freq_array to shape (Nspws, Nfreqs).
        # Currently does not actually support Nspws>1!
        self.freq_array = np.tile(self.freq_array, (self.Nspws, 1))

        # Temporary arrays to hold polarization axis, which will be collapsed
        ra_pol_list = np.zeros((self.Nblts, self.Npols))
        dec_pol_list = np.zeros((self.Nblts, self.Npols))
        uvw_pol_list = np.zeros((self.Nblts, 3, self.Npols))
        c_ns = const.c.to('m/ns').value
        for pol, data in data_accumulator.iteritems():
            pol_ind = self._pol_to_ind(pol)
            for ind, d in enumerate(data):
                blt = ["{1:.{0}f}".format(prec_t, d[1]).zfill(ndig_t),
                       str(d[2]).zfill(ndig_ant), str(d[3]).zfill(ndig_ant)]
                blt = "_".join(blt)
                blt_index = reverse_inds[blt]

                self.data_array[blt_index, :, :, pol_ind] = d[4]
                self.flag_array[blt_index, :, :, pol_ind] = d[5]
                self.nsample_array[blt_index, :, :, pol_ind] = d[6]

                # because there are uvws/ra/dec for each pol, and one pol may not
                # have that visibility, we collapse along the polarization
                # axis but avoid any missing visbilities
                uvw = d[0] * c_ns
                uvw.shape = (1, 3)
                uvw_pol_list[blt_index, :, pol_ind] = uvw
                ra_pol_list[blt_index, pol_ind] = d[7]
                dec_pol_list[blt_index, pol_ind] = d[8]

        # Collapse pol axis for ra_list, dec_list, and uvw_list
        ra_list = np.zeros(self.Nblts)
        dec_list = np.zeros(self.Nblts)
        for blt_index in xrange(self.Nblts):
            test = ~np.all(self.flag_array[blt_index, :, :, :], axis=(0, 1))
            good_pol = np.where(test)[0]
            if len(good_pol) == 1:
                # Only one good pol, use it
                self.uvw_array[blt_index, :] = uvw_pol_list[blt_index, :, good_pol]
                ra_list[blt_index] = ra_pol_list[blt_index, good_pol]
                dec_list[blt_index] = dec_pol_list[blt_index, good_pol]
            elif len(good_pol) > 1:
                # Multiple good pols, check for consistency. pyuvdata does not
                # support pol-dependent uvw, ra, or dec.
                if np.any(np.diff(uvw_pol_list[blt_index, :, good_pol], axis=0)):
                    raise ValueError('uvw values are different by polarization.')
                else:
                    self.uvw_array[blt_index, :] = uvw_pol_list[blt_index, :, good_pol[0]]
                if np.any(np.diff(ra_pol_list[blt_index, good_pol])):
                    raise ValueError('ra values are different by polarization.')
                else:
                    ra_list[blt_index] = ra_pol_list[blt_index, good_pol[0]]
                if np.any(np.diff(dec_pol_list[blt_index, good_pol])):
                    raise ValueError('dec values are different by polarization.')
                else:
                    dec_list[blt_index] = dec_pol_list[blt_index, good_pol[0]]
            else:
                # No good pols for this blt. Fill with first one.
                self.uvw_array[blt_index, :] = uvw_pol_list[blt_index, :, 0]
                ra_list[blt_index] = ra_pol_list[blt_index, 0]
                dec_list[blt_index] = dec_pol_list[blt_index, 0]

        # get unflagged blts
        blt_good = np.where(~np.all(self.flag_array, axis=(1, 2, 3)))
        single_ra = np.isclose(np.mean(np.diff(ra_list[blt_good])), 0.)
        single_time = np.isclose(np.mean(np.diff(self.time_array[blt_good])), 0.)

        # first check to see if the phase_type was specified.
        if phase_type is not None:
            if phase_type is 'phased':
                self.set_phased()
            elif phase_type is 'drift':
                self.set_drift()
            else:
                raise ValueError('The phase_type was not recognized. '
                                 'Set the phase_type to "drift" or "phased" to '
                                 'reflect the phasing status of the data')
        else:
            # check if ra is constant throughout file; if it is,
            # file is tracking if not, file is drift scanning
            # check if there's only one unflagged time
            if not single_time:
                if single_ra:
                    self.set_phased()
                else:
                    self.set_drift()
            else:
                # if there's only one time, checking for consistent RAs doesn't work.
                # instead check for the presence of an epoch variable, which isn't
                # really a good option, but at least it prevents crashes.
                if 'epoch' in uv.vartable.keys():
                    self.set_phased()
                else:
                    self.set_drift()

        if self.phase_type == 'phased':
            # check that the RA values do not vary
            if not single_ra:
                raise(ValueError, 'phase_type is "phased" but the RA values are varying.')
            self.phase_center_ra = float(ra_list[0])
            self.phase_center_dec = float(dec_list[0])
            self.phase_center_epoch = uv['epoch']
        else:
            # check that the RA values are not constant (if more than one time present)
            if (single_ra and not single_time):
                raise(ValueError, 'phase_type is "drift" but the RA values are constant.')
            self.zenith_ra = ra_list
            self.zenith_dec = dec_list

        try:
            self.set_telescope_params()
        except ValueError, ve:
            warnings.warn(str(ve))

        # check if object has all required uv_properties set
        if run_check:
            self.check(check_extra=check_extra,
                       run_check_acceptability=run_check_acceptability)

    def write_miriad(self, filepath, run_check=True, check_extra=True,
                     run_check_acceptability=True,
                     clobber=False, no_antnums=False):
        """
        Write the data to a miriad file.

        Args:
            filename: The miriad file directory to write to.
            run_check: Option to check for the existence and proper shapes of
                parameters before writing the file. Default is True.
            check_extra: Option to check optional parameters as well as required
                ones. Default is True.
            run_check_acceptability: Option to check acceptable range of the values of
                parameters before writing the file. Default is True.
            clobber: Option to overwrite the filename if the file already exists.
                Default is False.
            no_antnums: Option to not write the antnums variable to the file.
                Should only be used for testing purposes.
        """
        if run_check:
            self.check(check_extra=check_extra,
                       run_check_acceptability=run_check_acceptability)

        # check for multiple spws
        if self.data_array.shape[1] > 1:
            raise ValueError('write_miriad currently only handles single spw files.')

        if os.path.exists(filepath):
            if clobber:
                print 'File exists: clobbering'
                shutil.rmtree(filepath)
            else:
                raise ValueError('File exists: skipping')

        if self.Nfreqs > 1:
            freq_spacing = self.freq_array[0, 1:] - self.freq_array[0, :-1]
            if not np.isclose(np.min(freq_spacing), np.max(freq_spacing),
                              rtol=self._freq_array.tols[0], atol=self._freq_array.tols[1]):
                raise ValueError('The frequencies are not evenly spaced (probably '
                                 'because of a select operation). The miriad format '
                                 'does not support unevenly spaced frequencies.')
            if not np.isclose(np.max(freq_spacing), self.channel_width,
                              rtol=self._freq_array.tols[0], atol=self._freq_array.tols[1]):
                raise ValueError('The frequencies are separated by more than their '
                                 'channel width (probably because of a select operation). '
                                 'The miriad format does not support frequencies '
                                 'that are spaced by more than their channel width.')

        uv = aipy.miriad.UV(filepath, status='new')

        # initialize header variables
        uv._wrhd('obstype', 'mixed-auto-cross')
        # avoid inserting extra \n.
        if not self.history[-1] == '\n':
            self.history += '\n'
        uv._wrhd('history', self.history)

        # recognized miriad variables
        uv.add_var('nchan', 'i')
        uv['nchan'] = self.Nfreqs
        uv.add_var('npol', 'i')
        uv['npol'] = self.Npols
        uv.add_var('nspect', 'i')
        uv['nspect'] = self.Nspws
        uv.add_var('inttime', 'd')
        uv['inttime'] = self.integration_time
        uv.add_var('sdf', 'd')
        uv['sdf'] = self.channel_width / 1e9  # in GHz
        uv.add_var('source', 'a')
        uv['source'] = self.object_name
        uv.add_var('telescop', 'a')
        uv['telescop'] = self.telescope_name
        uv.add_var('latitud', 'd')
        uv['latitud'] = self.telescope_location_lat_lon_alt[0]
        uv.add_var('longitu', 'd')
        uv['longitu'] = self.telescope_location_lat_lon_alt[1]
        uv.add_var('nants', 'i')
        if self.x_orientation is not None:
            uv.add_var('xorient', 'a')
            uv['xorient'] = self.x_orientation
        if self.antenna_diameters is not None:
            if not np.allclose(self.antenna_diameters, self.antenna_diameters[0]):
                warnings.warn('Antenna diameters are not uniform, but miriad only'
                              'supports a single diameter. Skipping.')
            else:
                uv.add_var('antdiam', 'd')
                uv['antdiam'] = float(self.antenna_diameters[0])

        # These are added to make files written by pyuvdata more "miriad correct", and
        # should be changed when support for more than one spectral window is added.
        # 'nschan' is the number of channels per spectral window, and 'ischan' is the
        # starting channel for each spectral window. Both should be arrays of size Nspws.
        # Also note that indexing in Miriad is 1-based
        uv.add_var('nschan', 'i')
        uv['nschan'] = self.Nfreqs
        uv.add_var('ischan', 'i')
        uv['ischan'] = 1

        # Miriad has no way to keep track of antenna numbers, so the antenna
        # numbers are simply the index for each antenna in any array that
        # describes antenna attributes (e.g. antpos for the antenna_postions).
        # Therefore on write, nants (which gives the size of the antpos array)
        # needs to be increased to be the max value of antenna_numbers+1 and the
        # antpos array needs to be inflated with zeros at locations where we
        # don't have antenna information. These inflations need to be undone at
        # read. If the file was written by pyuvdata, then the variable antnums
        # will be present and we can use it, otherwise we need to test for zeros
        # in the antpos array and/or antennas with no visibilities.
        nants = np.max(self.antenna_numbers) + 1
        uv['nants'] = nants
        if self.antenna_positions is not None:
            # Miriad wants antenna_positions to be in absolute coordinates
            # (not relative to array center) in a rotated ECEF frame where the
            # x-axis goes through the local meridian.
            rel_ecef_antpos = np.zeros((nants, 3), dtype=self.antenna_positions.dtype)
            for ai, num in enumerate(self.antenna_numbers):
                rel_ecef_antpos[num, :] = self.antenna_positions[ai, :]

            # find zeros so antpos can be zeroed there too
            antpos_length = np.sqrt(np.sum(np.abs(rel_ecef_antpos)**2, axis=1))

            ecef_antpos = rel_ecef_antpos + self.telescope_location
            longitude = self.telescope_location_lat_lon_alt[1]
            antpos = uvutils.rotECEF_from_ECEF(ecef_antpos, longitude)

            # zero out bad locations (these are checked on read)
            antpos[np.where(antpos_length == 0), :] = [0, 0, 0]

            uv.add_var('antpos', 'd')
            # Miriad stores antpos values in units of ns, pyuvdata uses meters.
            uv['antpos'] = antpos.T.flatten() / const.c.to('m/ns').value

        uv.add_var('sfreq', 'd')
        uv['sfreq'] = self.freq_array[0, 0] / 1e9  # first spw; in GHz
        if self.phase_type == 'phased':
            uv.add_var('epoch', 'r')
            uv['epoch'] = self.phase_center_epoch

        # required pyuvdata variables that are not recognized miriad variables
        uv.add_var('ntimes', 'i')
        uv['ntimes'] = self.Ntimes
        uv.add_var('nbls', 'i')
        uv['nbls'] = self.Nbls
        uv.add_var('nblts', 'i')
        uv['nblts'] = self.Nblts
        uv.add_var('visunits', 'a')
        uv['visunits'] = self.vis_units
        uv.add_var('instrume', 'a')
        uv['instrume'] = self.instrument
        uv.add_var('altitude', 'd')
        uv['altitude'] = self.telescope_location_lat_lon_alt[2]

        # optional pyuvdata variables that are not recognized miriad variables
        if self.dut1 is not None:
            uv.add_var('dut1', 'd')
            uv['dut1'] = self.dut1
        if self.earth_omega is not None:
            uv.add_var('degpdy', 'd')
            uv['degpdy'] = self.earth_omega
        if self.gst0 is not None:
            uv.add_var('gst0', 'd')
            uv['gst0'] = self.gst0
        if self.rdate is not None:
            uv.add_var('rdate', 'a')
            uv['rdate'] = self.rdate
        if self.timesys is not None:
            uv.add_var('timesys', 'a')
            uv['timesys'] = self.timesys

        # other extra keywords
        # set up dictionaries to map common python types to miriad types
        # NB: arrays/lists/dicts could potentially be written as strings or 1D
        # vectors.  This is not supported at present!
        # NB: complex numbers *should* be supportable, but are not currently
        # supported due to unexplained errors in aipy.miriad and/or its underlying libraries
        numpy_types = {np.int8: int,
                       np.int16: int,
                       np.int32: int,
                       np.int64: int,
                       np.uint8: int,
                       np.uint16: int,
                       np.uint32: int,
                       np.uint64: int,
                       np.float16: float,
                       np.float32: float,
                       np.float64: float,
                       np.float128: float,
                       }
        types = {str: 'a',
                 int: 'i',
                 float: 'd',
                 bool: 'a',  # booleans are stored as strings and changed back on read
                 }
        for key, value in self.extra_keywords.iteritems():
            if type(value) in numpy_types.keys():
                if numpy_types[type(value)] == int:
                    value = int(value)
                elif numpy_types[type(value)] == float:
                    value = float(value)
            elif type(value) == bool:
                value = str(value)
            elif type(value) not in types.keys():
                raise TypeError('Extra keyword {keyword} is of {keytype}. '
                                'Only strings and real numbers are '
                                'supported in miriad.'.format(keyword=key,
                                                              keytype=type(value)))

            if len(str(key)) > 8:
                warnings.warn('key {key} in extra_keywords is longer than 8 '
                              'characters. It will be truncated to 8 as required '
                              'by the miriad file format.'.format(key=key))

            uvkeyname = str(key)[:8]  # name must be string, max 8 letters
            typestring = types[type(value)]
            uv.add_var(uvkeyname, typestring)
            uv[uvkeyname] = value

        if not no_antnums:
            # Add in the antenna_numbers so we have them if we read this file back in.
            # For some reason Miriad doesn't handle an array of integers properly,
            # so convert to floats here and integers on read.
            uv.add_var('antnums', 'd')
            uv['antnums'] = self.antenna_numbers.astype(np.float64)

        # antenna names is a foreign concept in miriad but required in other formats.
        # Miriad can't handle arrays of strings, so we make it into one long
        # comma-separated string and convert back on read.
        ant_name_str = '[' + ', '.join(self.antenna_names) + ']'
        uv.add_var('antnames', 'a')
        uv['antnames'] = ant_name_str

        # variables that can get updated with every visibility
        uv.add_var('pol', 'i')
        uv.add_var('lst', 'd')
        uv.add_var('cnt', 'd')
        uv.add_var('ra', 'd')
        uv.add_var('dec', 'd')

        # write data
        c_ns = const.c.to('m/ns').value
        for viscnt, blt in enumerate(self.data_array):
            uvw = (self.uvw_array[viscnt] / c_ns).astype(np.double)  # NOTE issue 50 on conjugation
            t = self.time_array[viscnt]
            i = self.ant_1_array[viscnt]
            j = self.ant_2_array[viscnt]

            uv['lst'] = self.lst_array[viscnt]
            if self.phase_type == 'phased':
                uv['ra'] = self.phase_center_ra
                uv['dec'] = self.phase_center_dec
            elif self.phase_type == 'drift':
                uv['ra'] = self.zenith_ra[viscnt]
                uv['dec'] = self.zenith_dec[viscnt]
            else:
                raise ValueError('The phasing type of the data is unknown. '
                                 'Set the phase_type to "drift" or "phased" to '
                                 'reflect the phasing status of the data')

            # NOTE only writing spw 0, not supporting multiple spws for write
            for polcnt, pol in enumerate(self.polarization_array):
                uv['pol'] = pol.astype(np.int)
                uv['cnt'] = self.nsample_array[viscnt, 0, :, polcnt].astype(np.double)

                data = self.data_array[viscnt, 0, :, polcnt]
                flags = self.flag_array[viscnt, 0, :, polcnt]
                if i > j:
                    i, j, data = j, i, np.conjugate(data)
                preamble = (uvw, t, (i, j))

                uv.write(preamble, data, flags)