Source code for musif.extract.features.texture.handler

from typing import List


import numpy as np

from pandas import DataFrame


from musif.common._constants import VOICE_FAMILY

from musif.config import Configuration

from musif.extract.basic_modules.scoring.constants import FAMILY

from musif.extract.common import _filter_parts_data, _part_matches_filter

from musif.extract.constants import (
    DATA_FAMILY,
    DATA_FAMILY_ABBREVIATION,
    DATA_PART_ABBREVIATION,
    DATA_SOUND_ABBREVIATION,
)

from musif.extract.features.core.handler import DATA_NOTES

from musif.extract.features.prefix import get_part_feature, get_part_prefix

from .constants import *

from musif.extract.features.core.constants import NUM_NOTES


def update_score_objects(
    score_data: dict,
    parts_data: List[dict],
    cfg: Configuration,
    parts_features: List[dict],
    score_features: dict,
):

    parts_data = _filter_parts_data(parts_data, cfg.parts_filter)

    if len(parts_data) == 0:
        return

    features = {}

    for part_data, part_features in zip(parts_data, parts_features):

        part = part_data[DATA_PART_ABBREVIATION]

        features[get_part_feature(part, NUM_NOTES)] = part_features[NUM_NOTES]

    score_features.update(features)

    notes = {}

    for j, part in enumerate(parts_data):

        if part[DATA_PART_ABBREVIATION].startswith("vn"):

            # capitalization to preserve I and II in Violins

            notes[
                part[DATA_PART_ABBREVIATION][0].upper()
                + part[DATA_PART_ABBREVIATION][1:]
            ] = len(part[DATA_NOTES])

        elif part[DATA_FAMILY] == VOICE_FAMILY:

            notes[part[DATA_FAMILY_ABBREVIATION].capitalize()] = int(
                score_features[
                    FAMILY + part[DATA_FAMILY_ABBREVIATION].capitalize() + "_NotesMean"
                ]
            )

        else:

            abbreviation = (
                part[DATA_SOUND_ABBREVIATION][0].upper()
                + part[DATA_SOUND_ABBREVIATION][1:]
            )

            notes[part[DATA_SOUND_ABBREVIATION].capitalize()] = int(
                score_features["Sound" + abbreviation + "_NotesMean"]
            )

    for i, (key, value) in enumerate(notes.items()):

        texture = value / np.asarray(list(notes.values())[i + 1 :])

        for j, t in enumerate(texture):

            part1 = key

            part2 = list(notes.keys())[j + i + 1]

            part1_prefix = get_part_prefix(part1).replace("_", "")

            part2_prefix = get_part_prefix(part2).replace("_", "")

            score_features[f"{part1_prefix}|{part2_prefix}_{TEXTURE}"] = t


def update_part_objects(
    score_data: dict, part_data: dict, cfg: Configuration, part_features: dict
):

    if not _part_matches_filter(part_data[DATA_PART_ABBREVIATION], cfg.parts_filter):

        return {}

    notes = part_data[DATA_NOTES]

    part_features.update({DATA_NOTES: len(notes)})