comments for understanding process flow + h5py support for importing custom brirs

from typing import Optional, Tuple, Union

from warnings import warn

import h5py

import copy

import numpy as np

from scipy.io import loadmat

        )

    mat_contents = loadmat(filename)

    #mat_contents = h5py.File(filename, 'r')

    try:

        IR = mat_contents["IR"]

        if in_fmt.startswith("MOZART"):

            dataset_suffix = "FULL"

        elif in_fmt in CHANNEL_BASED_AUDIO_FORMATS.keys():

            dataset_suffix = "LS"

        elif in_fmt in CHANNEL_BASED_AUDIO_FORMATS.keys(): ##This is triggered when -if HOA3 is passed from the command line. dataset suffix is being set to LS, and "FULL" mode cannot be used - in_format being used is 7_1_4. 

            dataset_suffix = "FULL"

    elif out_fmt.startswith("BINAURAL"):

        dataset_prefix = "HRIR"

        in_fmt = fromtype(in_fmt)

        tmp_fmt = fromtype("LS")

        IR_tmp = IR.copy()

        #IR_tmp = IR.copy() # Bug: deprecated function, no longer works

        IR_tmp = np.array(IR, copy=True)

        #IR_tmp = copy.deepcopy(IR) # -> error - h5py lists cannot be pickled

        IR = np.zeros([IR_tmp.shape[0], IR_tmp.shape[1], in_fmt.num_channels])

        ir_index = 0

    if trajectory is not None:

        cba.audio = rotate_cba(cba, trajectory)

    IR, _, latency_smp = load_ir(cba.name, bin.name, bin_dataset)

    IR, _, latency_smp = load_ir(cba.name, bin.name, bin_dataset) ##cba.name passed to this function is 7_1_4 -> this caused the renderer to attach suffix LS to the binaural dataset name in the next step

    # render LFE

    if bin_lfe_gain is not None:

            cba.audio, cba.fs, cba.lfe_index, bin_lfe_gain

        )

    ##The following two operations must be stateful for frame-wise, TCP enabled rendering

    # render rest of the signal

    bin.audio = binaural_fftconv(cba.audio, IR, cba.num_channels, cba.lfe_index)

    # compensate delay from binaural dataset

            f"Conversion from {sba.name} to {out.name} is unsupported!"

        )

#Rendered Audio is returned here

    return out

        Name of binaural dataset without prefix or suffix

    """

    if trajectory is not None:

    if trajectory is not None: ##Trajectory option for pre-coded head rotations??

        sba.audio = rotate_sba(sba, trajectory)

    if "ROOM" in bin.name:

    if "ROOM" in bin.name: #this is triggered for BINAURAL_ROOM option 

        ##Does this have be changed based on the new angles mode in SOFA? Or based on input format?

        cba_tmp = audio.fromtype("7_1_4")

        ##Temporary change to 5_1

        ##cba_tmp = audio.fromtype("5_1")

        cba_tmp.fs = sba.fs

        render_sba_to_cba(sba, cba_tmp)

        channelbased.render_cba_to_binaural(cba_tmp, bin, trajectory)

        ##connect cba_tmp to ivas python renderer 

        #Binaural dataset name is not being passeed here

        channelbased.render_cba_to_binaural(cba_tmp, bin, trajectory, bin_dataset)

        #channelbased.render_cba_to_binaural(cba_tmp, bin, **kwargs) - does not work if **kwargs is passed as an argument 

    else:

        IR, _, latency_smp = load_ir(sba.name, bin.name, bin_dataset)

        Channel-based output audio

    """

    ##recreate 

    render_mtx = get_allrad_mtx(sba.ambi_order, cba)

    cba.audio = sba.audio @ render_mtx.T

        ALLRAD matrix

    """

    n_harm = nchan_from_ambi_order(ambi_order)

    n_harm = nchan_from_ambi_order(ambi_order) ## (ambi_order + 1)`** 2`

    if cba.name == "MONO":

        hoa_dec = np.zeros([1, n_harm])