Module music_df.detremolo

Provides a function for detremoloing (removing rapid repeated notes from) a dataframe.

Functions

def detremolo(df: pandas.DataFrame,
max_tremolo_note_length: float = 0.25,
max_tremolo_note_gap: float = 0.125,
instrument_columns: Iterable[str] = ('instrument', 'midi_instrument', 'track', 'channel')) ‑> pandas.DataFrame

Expand source code

def detremolo(
    df: pd.DataFrame,
    max_tremolo_note_length: float = 0.25,
    max_tremolo_note_gap: float = 0.125,
    instrument_columns: Iterable[str] = (
        "instrument",
        "midi_instrument",
        "track",
        "channel",
    ),
) -> pd.DataFrame:
    """
    Merge rapid repeated notes into single notes.

    Args:
        df: dataframe.
        max_tremolo_note_length: maximum length of a note to be eligible for a tremolo.
        max_tremolo_note_gap: maximum gap between the onset of one note and the repeat
            of the same note to be eligible for a tremolo.
        instrument_columns: columns to group by when detremoloing. This permits us
            to only merge notes that seem to belong to the same instrument.
    """
    tremoli = []
    for _, instr in df[df["type"] == "note"].groupby(
        [c for c in instrument_columns if c in df.columns]
    ):
        for _, group in instr.groupby(["pitch"]):
            current_tremolo = None

            # I think that this will miss the last note in the case where tremolo extends to the
            #   last note

            for (row1_i, note1), (row2_i, note2) in zip(
                group.iloc[:-1].iterrows(), group.iloc[1:].iterrows()
            ):
                if (note1.release - note1.onset <= max_tremolo_note_length) and (
                    note2.onset - note1.release <= max_tremolo_note_gap
                ):
                    if current_tremolo is None:
                        current_tremolo = []
                    current_tremolo.append(row1_i)
                else:
                    if current_tremolo is not None:
                        tremoli.append(current_tremolo)
                        current_tremolo = None
            if current_tremolo is not None:
                tremoli.append(current_tremolo)

    out_df = df.copy()

    to_drop = []

    for tremolo in tremoli:
        out_df.loc[tremolo[0], "release"] = out_df.loc[tremolo[-1], "release"]
        to_drop.extend(tremolo[1:])

    return out_df.drop(to_drop, axis=0)

Merge rapid repeated notes into single notes.

Args

df

dataframe.

max_tremolo_note_length

maximum length of a note to be eligible for a tremolo.

max_tremolo_note_gap

maximum gap between the onset of one note and the repeat of the same note to be eligible for a tremolo.

instrument_columns

columns to group by when detremoloing. This permits us to only merge notes that seem to belong to the same instrument.