#!/usr/bin/env python

# Copyright (C) 2014 Andrew Lundgren, Tito Dal Canton
#
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.

"""
Compute the optimal SNRs for every injection in an input HDF file or 
sim_inspiral table and store the result in HDF file datasets or in the
same ligolw table.
"""

import logging
import argparse
import multiprocessing
import numpy as np
import h5py

from ligo.lw import utils as ligolw_utils
from ligo.lw import lsctables

import pycbc
import pycbc.inject
import pycbc.psd
import pycbc.version
from pycbc.filter import sigma, make_frequency_series
from pycbc.types import TimeSeries, FrequencySeries, zeros, float32, \
                        MultiDetOptionAction, load_frequencyseries
from pycbc.io.ligolw import get_table_columns


class TimeIndependentPSD(object):
    def __init__(self, psd_series):
        self.psd_series = psd_series

    def __call__(self, time=None):
        return self.psd_series

class TimeVaryingPSD(object):
    def __init__(self, file_name, length=None, delta_f=None, f_low=None):
        with h5py.File(file_name, 'r') as f:
            self.file_name = file_name
            detector = tuple(f.keys())[0]
            self.start_times = f[detector + '/start_time'][:]
            self.end_times = f[detector + '/end_time'][:]
            self.file_f_low = f.attrs['low_frequency_cutoff']
        self._curr_psd = {}
        self._curr_psd_index = {}
        self.detector = detector
        self.length = length
        self.delta_f = delta_f
        self.f_low = f_low

    def __call__(self, time=None):
        mask = np.logical_and(self.start_times <= time,
                              self.end_times > time)
        if not mask.any():
            return None
        center_times = (self.start_times[mask] + self.end_times[mask]) / 2.
        closest_idx = np.argmin(abs(center_times - time))
        return self.get_psd(np.flatnonzero(mask)[closest_idx])

    def get_psd(self, index):
        curr_pid = multiprocessing.current_process().pid
        if curr_pid not in self._curr_psd_index.keys():
            self._curr_psd_index[curr_pid] = -1
        if not index == self._curr_psd_index[curr_pid]:
            group = self.detector + '/psds/' + str(index)
            psd = load_frequencyseries(self.file_name, group=group)
            if delta_f is not None and psd.delta_f != delta_f:
                psd = pycbc.psd.interpolate(psd, delta_f)
            if self.length is not None and self.length != len(psd):
                psd2 = FrequencySeries(zeros(self.length, dtype=psd.dtype),
                                       delta_f=psd.delta_f)
                if self.length > len(psd):
                    psd2[:] = np.inf
                    psd2[0:len(psd)] = psd
                else:
                    psd2[:] = psd[0:self.length]
                psd = psd2
            if self.f_low is not None and self.f_low < self.file_f_low:
                # avoid using the PSD below the f_low given in the file
                k = int(self.file_f_low / psd.delta_f)
                psd[0:k] = np.inf
            self._curr_psd[curr_pid] = psd
            self._curr_psd_index[curr_pid] = index
        return self._curr_psd[curr_pid]

def parse_injection_range(num_inj, rangestr):
    part = int(rangestr.split('/')[0])
    pieces = int(rangestr.split('/')[1])
    tmin =  num_inj * part // pieces
    tmax =  num_inj * (part + 1) // pieces
    return tmin, tmax

def get_gc_end_time(injection):
    """Return the geocenter end time of an injection. Required for seamless
    compatibility with LIGOLW and HDF injection objects, which use different
    names.
    """
    try:
        # geocent time is robust to potentially incomplete sim tables
        return injection.geocent_end_time
    except AttributeError:
        return injection.tc


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description=__doc__)
    pycbc.add_common_pycbc_options(parser)
    parser.add_argument("--version", action=pycbc.version.Version)
    parser.add_argument('--input-file', '-i', dest='injection_file',
                        required=True,
                        help='Input LIGOLW file defining injections')
    parser.add_argument('--injection-f-ref', type=float,
                        help='Reference frequency in Hz for '
                             'creating CBC injections from an XML '
                             'file.')
    parser.add_argument('--injection-f-final', type=float,
                        help='Override the f_final field of a CBC '
                             'XML injection file.')
    parser.add_argument('--input-sort', type=str, choices=['random', 'time'],
                        default='random',
                        help='Sort injections by the given criterion before '
                             'processing them. Sorting by time may speed up '
                             'the calculation when the approximant is very '
                             'fast and injections are tightly spaced in time')
    parser.add_argument('--output-file', '-o', dest='out_file', required=True,
                        help='Output LIGOLW file')
    parser.add_argument('--f-low', type=float, default=30.,
                        help='Start frequency of matched-filter integration '
                             'in Hz (default %(default)s)')
    parser.add_argument('--seg-length', type=float, default=256,
                        help='Segment duration in seconds (default %(default)s)')
    parser.add_argument('--sample-rate', type=float, default=16384,
                        help='Data sample rate in Hz (default %(default)s)')
    parser.add_argument('--ifos', nargs='+',
                        help='Specify ifos for default HDF dataset output')
    parser.add_argument('--snr-columns', nargs='+', action=MultiDetOptionAction,
                        metavar='DETECTOR:COLUMN',
                        help='For sim_inspiral table output, specify columns'
                        ' to store the optimal SNR for each detector. COLUMN'
                        ' should be an existing sim_inspiral column containing'
                        ' no useful data, alpha1, alpha2 etc. are good'
                        ' candidates. For HDF output the --ifos option should be'
                        ' used, datasets will be named eg "optimal_snr_H1"')
    parser.add_argument('--cores', default=1, type=int,
                        help='Parallelize the computation over the given '
                             'number of cores')
    parser.add_argument('--ignore-waveform-errors', action='store_true',
                        help='Ignore errors in waveform generation and keep '
                             'the corresponding column unchanged')
    parser.add_argument('--progress', action='store_true',
                        help='Show a progress bar (requires tqdm)')
    psd_group = pycbc.psd.insert_psd_option_group_multi_ifo(parser)
    psd_group.add_argument('--time-varying-psds', nargs='*', metavar='FILE',
                           help='Instead of time-independent PSDs, use time-varying '
                           'PSDs from the given HDF5 files and pick the appropriate '
                           'PSD for each injection')
    parser.add_argument('--injection-fraction-range', default='0/1',
                        help='Optional, analyze only a certain range of the '
                             'injections. Format PART/NUM_PARTS')
    opts = parser.parse_args()

    if opts.ifos is not None:
        detectors = opts.ifos
        if opts.snr_columns is not None:
            parser.error("Can't use both --ifos and --snr-columns !")
        opts.snr_columns = {i: 'optimal_snr_' + i for i in opts.ifos}
    else:
        detectors = opts.snr_columns.keys()

    if not opts.time_varying_psds:
        pycbc.psd.verify_psd_options_multi_ifo(opts, parser, detectors)

    pycbc.init_logging(opts.verbose)

    seg_len = opts.seg_length
    sample_rate = opts.sample_rate
    delta_t = 1. / sample_rate
    delta_f = 1. / seg_len
    tlen = int(seg_len * sample_rate)
    flen = tlen // 2 + 1
    f_low = opts.f_low

    logging.info("Loading PSDs")
    if opts.time_varying_psds:
        psds = {}
        for tvpsd_file in opts.time_varying_psds:
            tvpsd = TimeVaryingPSD(tvpsd_file, flen, delta_f, f_low)
            psds[tvpsd.detector] = tvpsd
        if set(detectors) != set(psds.keys()):
            parser.error('Inconsistent detector list in time-varying PSD ' \
                         'specification (%s vs %s)' % (detectors, psds.keys()))
    else:
        psds = pycbc.psd.from_cli_multi_ifos(
            opts,
            length_dict=dict((det, flen) for det in detectors),
            delta_f_dict=dict((det, delta_f) for det in detectors),
            low_frequency_cutoff_dict=dict((det, f_low) for det in detectors),
            ifos=detectors,
            strain_dict=dict((det, None) for det in detectors),
            dyn_range_factor=pycbc.DYN_RANGE_FAC)
        for det in detectors:
            psds[det] = TimeIndependentPSD(psds[det].astype(float32))

    def get_injection(injections, det, injection_time, simulation_id):
        """ Do an injection from the injection XML file, specified by
        IFO and end time"""
        # leave 4 s of padding at the end for possible ringdown
        start_time = int(injection_time + 4. - seg_len)
        strain = TimeSeries(zeros(tlen, dtype=float32), delta_t=delta_t,
                            epoch=start_time)
        injections.apply(strain, det, distance_scale=1./pycbc.DYN_RANGE_FAC,
                         simulation_ids=[simulation_id])
        return make_frequency_series(strain)

    def compute_optimal_snr(inj):
        if not ligolw:
            inj = inj.view(np.recarray)
        for det, column in opts.snr_columns.items():
            injection_time = get_gc_end_time(inj)
            psd = psds[det](injection_time)
            if psd is None:
                continue
            logging.debug('Trying injection %s at %s', inj.simulation_id, det)
            try:
                wave = get_injection(injections, det, injection_time,
                                     simulation_id=inj.simulation_id)
            except Exception as e:
                if opts.ignore_waveform_errors:
                    logging.debug(
                        '%s: waveform generation failed, skipping (%s)',
                        inj.simulation_id,
                        e
                    )
                    continue
                else:
                    logging.error('%s: waveform generation failed with the '
                                  'following exception', inj.simulation_id)
                    raise
            logging.debug(
                'Injection %s at %s completed',
                inj.simulation_id,
                det
            )
            sval = sigma(wave, psd=psd, low_frequency_cutoff=f_low)
            if ligolw:
                setattr(inj, column, sval)
            else:
                inj[column] = sval
        return inj

    logging.info("Loading injections")
    injections = pycbc.inject.InjectionSet.from_cli(opts)
    inj_table = injections.table

    # a bit of ugly special-casing to keep the traditional behavior of
    # pycbc_optimal_snr: if both input and output are LIGOLW documents, we want
    # to preserve the entire content of the document and only modify the
    # particular columns of the sim_inspiral table.  If HDF injections are
    # involved, we do not care.
    ligolw_suffixes = ('.xml', '.xml.gz')
    ligolw = opts.injection_file.endswith(ligolw_suffixes) \
            and opts.out_file.endswith(ligolw_suffixes)

    if not ligolw:
        # create placeholder fields for FieldArray injections
        for det, column in opts.snr_columns.items():
            if column in inj_table:
                continue
            inj_table = inj_table.add_fields(np.zeros(len(inj_table)),
                                             column)

        # make sure we have simulation IDs
        if 'simulation_id' not in inj_table:
            inj_table = inj_table.add_fields(np.arange(len(inj_table)),
                                             'simulation_id')

        inj_dtype = inj_table.dtype

    if opts.input_sort == 'random':
        np.random.seed(100)
        sort_func = lambda x: np.random.random()
    elif opts.input_sort == 'time':
        sort_func = get_gc_end_time
    inj_table = sorted(inj_table, key=sort_func)

    if ligolw:
        new_inj_table = lsctables.New(
            lsctables.SimInspiralTable,
            columns=get_table_columns(injections.table)
        )
    else:
        new_inj_table = []

    # Cut inj_table down to the range defined by opts.injection_fraction_range
    num_injections = len(inj_table)
    imin, imax = parse_injection_range(num_injections,
                                       opts.injection_fraction_range)
    inj_table = inj_table[imin:imax]

    if opts.cores > 1:
        logging.info('Starting workers')
        pool = multiprocessing.Pool(processes=opts.cores)
        iterator = pool.imap_unordered(compute_optimal_snr, inj_table)
    else:
        # do not bother spawning extra processes if running single-core
        iterator = (compute_optimal_snr(inj) for inj in inj_table)

    if opts.progress:
        try:
            from tqdm import tqdm

            iterator = tqdm(iterator, total=len(inj_table))
        except ImportError:
            logging.warning('cannot import tqdm; not showing progress bar')
            pass

    for inj in iterator:
        new_inj_table.append(inj)

    # always store injections sorted by coalescence time
    new_inj_table.sort(key=get_gc_end_time)

    if not ligolw:
        new_inj_table = pycbc.io.FieldArray.from_records(
                new_inj_table, dtype=inj_dtype)

    logging.info('Writing output')

    if ligolw:
        llw_doc = injections.indoc
        llw_root = llw_doc.childNodes[0]
        llw_root.removeChild(injections.table)
        llw_root.appendChild(new_inj_table)
        ligolw_utils.write_filename(llw_doc, opts.out_file, compress='auto')
    else:
        pycbc.inject.InjectionSet.write(opts.out_file, new_inj_table)

    logging.info('Done')