Module music_df.dedouble

Provides a function to "dedouble" a music_df.

Dedoubling is removing rows that have the same type, onset, release, and pitch.

Functions

def dedouble(df: pandas.DataFrame,
quantize: bool = True,
ticks_per_quarter: int = 16,
match_releases: bool = True,
columns: Iterable[str] = ('type', 'pitch')) ‑> pandas.DataFrame

Expand source code

def dedouble(
    df: pd.DataFrame,
    quantize: bool = True,
    ticks_per_quarter: int = 16,
    match_releases: bool = True,
    # columns are always deduplicated on onset and release
    columns: Iterable[str] = ("type", "pitch"),
) -> pd.DataFrame:
    """
    Dedouble a music_df.

    Args:
        df: The dataframe to dedouble.
        quantize: Whether to quantize the onset and release times before dedoubling.
            Default is True.
        ticks_per_quarter: The number of ticks per quarter note when quantizing.
            Default is 16.
        match_releases: Whether to match releases as well as onsets when dedoubling.
            Default is True.
        columns: The columns, other than onset and release, to deduplicate on.
            Default is ("type", "pitch").

    >>> df = pd.DataFrame(
    ...     {
    ...         "type": ["time_signature", "bar", "note", "note", "note", "bar"],
    ...         "track": [0.0, np.nan, 1.0, 1.0, 2.0, np.nan],
    ...         "channel": [np.nan, np.nan, 0.0, 0.0, 0.0, np.nan],
    ...         "pitch": [np.nan, np.nan, 50.0, 50.0, 50.0, np.nan],
    ...         "onset": [0.0, 0.0, 0.0, 0.0, 0.0, 3.0],
    ...         "release": [np.nan, np.nan, 1.0, 0.99, 1.0, np.nan],
    ...         "velocity": [np.nan, np.nan, 100.0, 100.0, 100.0, np.nan],
    ...         "other": [
    ...             {"numerator": 3, "denominator": 4, "clocks_per_click": None},
    ...             np.nan,
    ...             np.nan,
    ...             np.nan,
    ...             np.nan,
    ...             np.nan,
    ...         ],
    ...     }
    ... )
    >>> pd.set_option(
    ...     "display.width", 200
    ... )  # To avoid issues when the terminal is a different size
    >>> pd.set_option("display.max_columns", None)

    >>> df  # doctest: +NORMALIZE_WHITESPACE
                 type  track  channel  pitch  onset  release  velocity                                              other
    0  time_signature    0.0      NaN    NaN    0.0      NaN       NaN  {'numerator': 3, 'denominator': 4, 'clocks_per...
    1             bar    NaN      NaN    NaN    0.0      NaN       NaN                                                NaN
    2            note    1.0      0.0   50.0    0.0     1.00     100.0                                                NaN
    3            note    1.0      0.0   50.0    0.0     0.99     100.0                                                NaN
    4            note    2.0      0.0   50.0    0.0     1.00     100.0                                                NaN
    5             bar    NaN      NaN    NaN    3.0      NaN       NaN                                                NaN
    >>> dedouble(df)[
    ...     ["original_index", "duplicated_indices"]
    ... ]  # doctest: +NORMALIZE_WHITESPACE
       original_index duplicated_indices
    0               0
    1               1
    2               2              2,3,4
    3               5

    We store the following attributes in .attrs:
    - dedoubled (bool)
    - n_dedoubled_notes (int)
    - n_undedoubled_notes (int)
    >>> dedouble(df).attrs
    {'n_undedoubled_notes': 3, 'n_dedoubled_notes': 1, 'dedoubled': True}

    >>> dedouble(df, quantize=False)[
    ...     ["original_index", "duplicated_indices"]
    ... ]  # doctest: +NORMALIZE_WHITESPACE
       original_index duplicated_indices
    0               0
    1               1
    2               2                2,4
    3               3
    4               5
    >>> dedouble(df, columns=("type", "pitch", "track"))[
    ...     ["original_index", "duplicated_indices"]
    ... ]  # doctest: +NORMALIZE_WHITESPACE
       original_index duplicated_indices
    0               0
    1               1
    2               2                2,3
    3               4
    4               5

    """
    df = df.copy()
    df.attrs["n_undedoubled_notes"] = int((df.type == "note").sum())
    if not match_releases:
        raise NotImplementedError
    if quantize:
        df["quantized_onset"] = (df["onset"] * ticks_per_quarter).round()
        df["quantized_release"] = (df["release"] * ticks_per_quarter).round()
        columns = list(columns) + ["quantized_onset", "quantized_release"]
    else:
        columns = list(columns) + ["onset", "release"]

    # Group by the columns and get the indices of the rows in each group
    groups = df.groupby(columns).indices

    # Filter out the groups with only one row (i.e., no duplicates)
    duplicate_indices = [indices for indices in groups.values() if len(indices) > 1]

    df["duplicated_indices"] = ""
    to_drop = []
    for indices in duplicate_indices:
        df.loc[indices[0], "duplicated_indices"] = ",".join(str(x) for x in indices)
        to_drop.extend(indices[1:])

    deduplicated = df.drop(to_drop, axis=0)

    if quantize:
        deduplicated = deduplicated.drop(
            ["quantized_onset", "quantized_release"], axis=1
        )

    if "original_index" not in deduplicated.columns:
        deduplicated = deduplicated.reset_index(names="original_index")
    else:
        deduplicated = deduplicated.reset_index(drop=True)

    deduplicated.attrs["n_dedoubled_notes"] = int((deduplicated.type == "note").sum())
    deduplicated.attrs["dedoubled"] = True

    return deduplicated

Dedouble a music_df.

Args

df

The dataframe to dedouble.

quantize

Whether to quantize the onset and release times before dedoubling. Default is True.

ticks_per_quarter

The number of ticks per quarter note when quantizing. Default is 16.

match_releases

Whether to match releases as well as onsets when dedoubling. Default is True.

columns

The columns, other than onset and release, to deduplicate on. Default is ("type", "pitch").

>>> df = pd.DataFrame(
...     {
...         "type": ["time_signature", "bar", "note", "note", "note", "bar"],
...         "track": [0.0, np.nan, 1.0, 1.0, 2.0, np.nan],
...         "channel": [np.nan, np.nan, 0.0, 0.0, 0.0, np.nan],
...         "pitch": [np.nan, np.nan, 50.0, 50.0, 50.0, np.nan],
...         "onset": [0.0, 0.0, 0.0, 0.0, 0.0, 3.0],
...         "release": [np.nan, np.nan, 1.0, 0.99, 1.0, np.nan],
...         "velocity": [np.nan, np.nan, 100.0, 100.0, 100.0, np.nan],
...         "other": [
...             {"numerator": 3, "denominator": 4, "clocks_per_click": None},
...             np.nan,
...             np.nan,
...             np.nan,
...             np.nan,
...             np.nan,
...         ],
...     }
... )
>>> pd.set_option(
...     "display.width", 200
... )  # To avoid issues when the terminal is a different size
>>> pd.set_option("display.max_columns", None)

>>> df  # doctest: +NORMALIZE_WHITESPACE
             type  track  channel  pitch  onset  release  velocity                                              other
0  time_signature    0.0      NaN    NaN    0.0      NaN       NaN  {'numerator': 3, 'denominator': 4, 'clocks_per...
1             bar    NaN      NaN    NaN    0.0      NaN       NaN                                                NaN
2            note    1.0      0.0   50.0    0.0     1.00     100.0                                                NaN
3            note    1.0      0.0   50.0    0.0     0.99     100.0                                                NaN
4            note    2.0      0.0   50.0    0.0     1.00     100.0                                                NaN
5             bar    NaN      NaN    NaN    3.0      NaN       NaN                                                NaN
>>> dedouble(df)[
...     ["original_index", "duplicated_indices"]
... ]  # doctest: +NORMALIZE_WHITESPACE
   original_index duplicated_indices
0               0
1               1
2               2              2,3,4
3               5

We store the following attributes in .attrs: - dedoubled (bool) - n_dedoubled_notes (int) - n_undedoubled_notes (int)

>>> dedouble(df).attrs
{'n_undedoubled_notes': 3, 'n_dedoubled_notes': 1, 'dedoubled': True}

>>> dedouble(df, quantize=False)[
...     ["original_index", "duplicated_indices"]
... ]  # doctest: +NORMALIZE_WHITESPACE
   original_index duplicated_indices
0               0
1               1
2               2                2,4
3               3
4               5
>>> dedouble(df, columns=("type", "pitch", "track"))[
...     ["original_index", "duplicated_indices"]
... ]  # doctest: +NORMALIZE_WHITESPACE
   original_index duplicated_indices
0               0
1               1
2               2                2,3
3               4
4               5