""" This is the same code as in the corpora repo as copied on September 24, 2020
and then adapted.
"""
import sys, re
from collections import defaultdict
import pandas as pd
import numpy as np
from .utils import abs2rel_key, changes2list, DCML_REGEX, fifths2iv, fifths2name, fifths2pc, fifths2rn, fifths2sd, \
labels2global_tonic, map2elements, name2fifths, rel2abs_key, resolve_relative_keys, roman_numeral2fifths, \
series_is_minor, split_alternatives, split_scale_degree, str_is_minor, transform, sort_tpcs
from .logger import function_logger
################################################################################
# Constants
################################################################################
[docs]class SliceMaker(object):
""" This class serves for storing slice notation such as ``:3`` as a variable or
passing it as function argument.
Examples
--------
.. code-block:: python
SM = SliceMaker()
some_function( slice_this, SM[3:8] )
select_all = SM[:]
df.loc[select_all]
"""
def __getitem__(self, item):
return item
SM = SliceMaker()
[docs]@function_logger
def expand_labels(df, column='label', regex=None, cols={}, dropna=False, propagate=True, volta_structure=None,
relative_to_global=False, chord_tones=True, absolute=False, all_in_c=False):
"""
Split harmony labels complying with the DCML syntax into columns holding their various features
and allows for additional computations and transformations.
Uses: :py:func:`compute_chord_tones`, :py:func:`features2type`, :py:func:`~.utils.labels2global_tonic`, :py:func:`propagate_keys`,
:py:func:`propagate_pedal`, :py:func:`replace_special`, :py:func:`~.utils.roman_numeral2fifths`, :py:func:`~.utils.split_alternatives`, :py:func:`split_labels`,
:py:func:`~.utils.transform`, :py:func:`transpose`
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe where one column contains DCML chord labels.
column : :obj:`str`
Name of the column that holds the harmony labels.
regex : :obj:`re.Pattern`
Compiled regular expression used to split the labels. It needs to have named groups.
The group names are used as column names unless replaced by ``cols``.
cols : :obj:`dict`, optional
Dictionary to map the regex's group names to deviating column names of your choice.
dropna : :obj:`bool`, optional
Pass True if you want to drop rows where ``column`` is NaN/<NA>
propagate: :obj:`bool`, optional
By default, information about global and local keys and about pedal points is spread throughout
the DataFrame. Pass False if you only want to split the labels into their features. This ignores
all following parameters because their expansions depend on information about keys.
volta_structure: :obj:`dict`, optional
{first_mc -> {volta_number -> [mc1, mc2...]} } dictionary as you can get it from
``Score.mscx.volta_structure``. This allows for correct propagation into second and other voltas.
relative_to_global : :obj:`bool`, optional
Pass True if you want all labels expressed with respect to the global key.
This levels and eliminates the features `localkey` and `relativeroot`.
chord_tones : :obj:`bool`, optional
Pass True if you want to add four columns that contain information about each label's
chord, added, root, and bass tones. The pitches are expressed as intervals
relative to the respective chord's local key or, if ``relative_to_global=True``,
to the globalkey. The intervals are represented as integers that represent
stacks of fifths over the tonic, such that 0 = tonic, 1 = dominant, -1 = subdominant,
2 = supertonic etc.
absolute : :obj:`bool`, optional
Pass True if you want to transpose the relative `chord_tones` to the global
key, which makes them absolute so they can be expressed as actual note names.
This implies prior conversion of the chord_tones (but not of the labels) to
the global tonic.
all_in_c : :obj:`bool`, optional
Pass True to transpose `chord_tones` to C major/minor. This performs the same
transposition of chord tones as `relative_to_global` but without transposing
the labels, too. This option clashes with `absolute=True`.
Returns
-------
:obj:`pandas.DataFrame`
Original DataFrame plus additional columns with split features.
"""
assert sum((absolute, all_in_c)) < 2, "Chord tones can be either 'absolute' or 'all_in_c', not both."
assert len(df.index.names) == 1, f"""df has a MultiIndex of {len(df.index.names)} levels, implying that it has information
from several pieces. Apply expand_labels() to one piece at a time."""
df = df.copy()
if regex is None:
regex = DCML_REGEX
### If the index is not unique, it has to be temporarily replaced
tmp_index = not df.index.is_unique
if tmp_index:
ix = df.index
df.reset_index(drop=True, inplace=True)
for col in ['numeral', 'form', 'figbass', 'localkey', 'globalkey', 'phraseend']:
if not col in cols:
cols[col] = col
global_minor = f"{cols['globalkey']}_is_minor"
### Check for too many immediate repetitions
not_nan = df[column].dropna()
immediate_repetitions = not_nan == not_nan.shift()
k = immediate_repetitions.sum()
if k > 0:
if k / len(not_nan.index) > 0.1:
logger.warning(
"DataFrame has many direct repetitions of labels. This function is written for lists of labels only which should have no immediate repetitions.")
else:
logger.debug(f"Immediate repetition of labels:\n{not_nan[immediate_repetitions]}")
### Do the actual expansion
df = split_alternatives(df, column=column, logger=logger)
df = split_labels(df, column=column, regex=regex, cols=cols, dropna=dropna, logger=logger)
df['chord_type'] = transform(df, features2type, [cols[col] for col in ['numeral', 'form', 'figbass']], logger=logger)
df = replace_special(df, regex=regex, merge=True, cols=cols, logger=logger)
### Check phrase annotations
p_col = df[cols['phraseend']]
opening = p_col.fillna('').str.count('{')
closing = p_col.fillna('').str.count('}')
if opening.sum() != closing.sum():
o = df.loc[(opening > 0), ['mc', cols['phraseend']]]
c = df.loc[(closing > 0), ['mc', cols['phraseend']]]
compare = pd.concat([o.reset_index(drop=True), c.reset_index(drop=True)], axis=1).astype({'mc': 'Int64'})
logger.warning(f"Phrase beginning and endings don't match:\n{compare}")
if propagate:
key_cols = {col: cols[col] for col in ['localkey', 'globalkey']}
try:
df = propagate_keys(df, volta_structure=volta_structure, add_bool=True, **key_cols, logger=logger)
except:
logger.error(f"propagate_keys() failed with\n{sys.exc_info()[1]}")
try:
df = propagate_pedal(df, cols=cols, logger=logger)
except:
logger.error(f"propagate_pedal() failed with\n{sys.exc_info()[1]}")
if chord_tones:
ct = compute_chord_tones(df, expand=True, cols=cols, logger=logger)
if relative_to_global or absolute or all_in_c:
transpose_by = transform(df, roman_numeral2fifths, [cols['localkey'], global_minor])
if absolute:
transpose_by += transform(df, name2fifths, [cols['globalkey']])
ct = pd.DataFrame([transpose(tpcs, fifths) for tpcs, fifths in
zip(ct.itertuples(index=False, name=None), transpose_by.values)], index=ct.index,
columns=ct.columns)
df = pd.concat([df, ct], axis=1)
if relative_to_global:
labels2global_tonic(df, inplace=True, cols=cols, logger=logger)
else:
if chord_tones:
logger.info("Chord tones cannot be calculated without propagating keys.")
if relative_to_global:
logger.info("Cannot transpose labels without propagating keys.")
if tmp_index:
df.index = ix
return df
[docs]def transpose(e, n):
""" Add `n` to all elements `e` recursively.
"""
return map2elements(e, lambda x: x + n)
[docs]@function_logger
def split_labels(df, column, regex, cols={}, dropna=False, inplace=False, **kwargs):
""" Split harmony labels complying with the DCML syntax into columns holding their various features.
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe where one column contains DCML chord labels.
column : :obj:`str`
Name of the column that holds the harmony labels.
regex : :obj:`re.Pattern`
Compiled regular expression used to split the labels. It needs to have named groups.
The group names are used as column names unless replaced by `cols`.
cols : :obj:`dict`
Dictionary to map the regex's group names to deviating column names.
dropna : :obj:`bool`, optional
Pass True if you want to drop rows where ``column`` is NaN/<NA>
inplace : :obj:`bool`, optional
Pass True if you want to mutate ``df``.
"""
if regex.__class__ != re.compile('').__class__:
regex = re.compile(regex, re.VERBOSE)
features = regex.groupindex.keys()
if not inplace:
df = df.copy()
if df[column].isna().any():
if dropna:
logger.debug(f"Removing NaN values from label column {column}...")
df = df[df[column].notna()]
else:
logger.debug(f"{column} contains NaN values.")
logger.debug(f"Applying RegEx to column {column}...")
spl = df[column].str.extract(regex, expand=True) # .astype('string') # needs pandas 1.0
for feature in features:
name = cols[feature] if feature in cols else feature
df[name] = spl[feature]
mistakes = spl.isna().apply(all, axis=1) & df[column].notna()
if mistakes.any():
logger.warning(f"The following labels do not match the regEx:\n{df.loc[mistakes, :column].to_string()}")
if not inplace:
return df
[docs]@function_logger
def features2type(numeral, form=None, figbass=None):
""" Turns a combination of the three chord features into a chord type.
Returns
-------
'M': Major triad
'm': Minor triad
'o': Diminished triad
'+': Augmented triad
'mm7': Minor seventh chord
'Mm7': Dominant seventh chord
'MM7': Major seventh chord
'mM7': Minor major seventh chord
'o7': Diminished seventh chord
'%7': Half-diminished seventh chord
'+7': Augmented (minor) seventh chord
'+M7': Augmented major seventh chord
"""
if pd.isnull(numeral) or numeral in ['Fr', 'Ger', 'It']:
return numeral
form, figbass = tuple('' if pd.isnull(val) else val for val in (form, figbass))
# triads
if figbass in ['', '6', '64']:
if form in ['o', '+']:
return form
if form in ['%', 'M', '+M']:
if figbass != '':
logger.error(f"{form} is a seventh chord and cannot have figbass '{figbass}'")
return None
# else: go down, interpret as seventh chord
else:
return 'm' if numeral.islower() else 'M'
# seventh chords
if form in ['o', '%', '+', '+M']:
return f"{form}7"
triad = 'm' if numeral.islower() else 'M'
seventh = 'M' if form == 'M' else 'm'
return f"{triad}{seventh}7"
[docs]@function_logger
def replace_special(df, regex, merge=False, inplace=False, cols={}, special_map={}):
"""
| Move special symbols in the `numeral` column to a separate column and replace them by the explicit chords they stand for.
| In particular, this function replaces the symbols `It`, `Ger`, and `Fr`.
Uses: :py:func:`merge_changes`
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe containing DCML chord labels that have been split by split_labels().
regex : :obj:`re.Pattern`
Compiled regular expression used to split the labels replacing the special symbols.It needs to have named groups.
The group names are used as column names unless replaced by `cols`.
merge : :obj:`bool`, optional
False: By default, existing values, except `figbass`, are overwritten.
True: Merge existing with new values (for `changes` and `relativeroot`).
cols : :obj:`dict`, optional
The special symbols appear in the column `numeral` and are moved to the column `special`.
In case the column names for ``['numeral','form', 'figbass', 'changes', 'relativeroot', 'special']`` deviate, pass a dict, such as
.. code-block:: python
{'numeral': 'numeral_col_name',
'form': 'form_col_name
'figbass': 'figbass_col_name',
'changes': 'changes_col_name',
'relativeroot': 'relativeroot_col_name',
'special': 'special_col_name'}
special_map : :obj:`dict`, optional
In case you want to add or alter special symbols to be replaced, pass a replacement map, e.g.
{'N': 'bII6'}. The column 'figbass' is only altered if it's None to allow for inversions of special chords.
inplace : :obj:`bool`, optional
Pass True if you want to mutate ``df``.
"""
if not inplace:
df = df.copy()
### If the index is not unique, it has to be temporarily replaced
tmp_index = not df.index.is_unique
if tmp_index:
ix = df.index
df.reset_index(drop=True, inplace=True)
special2label = {
'It': 'viio6(b3)/V',
'Ger': 'viio65(b3)/V',
'Fr': 'V7(b5)/V',
}
special2label.update(special_map)
features = ['numeral', 'form', 'figbass', 'changes', 'relativeroot']
for col in features + ['special']:
if not col in cols:
cols[col] = col
feature_cols = list(cols.values())
missing = [cols[f] for f in features if not cols[f] in df.columns]
assert len(
missing) == 0, f"These columns are missing from the DataFrame: {missing}. Either use split_labels() first or give correct `cols` parameter."
select_all_special = df[df[cols['numeral']].isin(special2label.keys())].index
logger.debug(f"Moving special symbols from {cols['numeral']} to {cols['special']}...")
if not cols['special'] in df.columns:
df.insert(df.columns.get_loc(cols['numeral']), cols['special'], np.nan)
df.loc[select_all_special, cols['special']] = df.loc[select_all_special, cols['numeral']]
def repl_spec(frame, special, instead):
"""Check if the selected parts are empty and replace ``special`` by ``instead``."""
new_vals = re.match(regex, instead)
if new_vals is None:
logger.warning(f"{instead} is not a valid label which could replace {special}. Skipped.")
return frame
else:
new_vals = new_vals.groupdict()
for f in features:
if new_vals[f] is not None:
replace_this = SM[:] # by default, replace entire column
if f == 'figbass': # only empty figbass is replaced, with the exception of `Ger6` and `Fr6`
if special in ['Fr', 'Ger']: # For these symbols, a wrong `figbass` == 6 is accepted and replaced
replace_this = (frame[cols['figbass']] == '6') | frame[cols['figbass']].isna()
else:
replace_this = frame[cols['figbass']].isna()
elif f != 'numeral': # numerals always replaced completely
not_empty = frame[cols[f]].notna()
if not_empty.any():
if f in ['changes', 'relativeroot'] and merge:
if f == 'changes':
frame.loc[not_empty, cols[f]] = frame.loc[not_empty, cols[f]].apply(merge_changes,
args=(new_vals[f],))
elif f == 'relativeroot':
frame.loc[not_empty, cols[f]] = frame.loc[not_empty, cols[f]].apply(
lambda x: f"{new_vals[f]}/{x}")
logger.debug(
f"While replacing {special}, the existing '{f}'-values have been merged with '{new_vals[f]}', resulting in :\n{frame.loc[not_empty, cols[f]]}")
replace_this = ~not_empty
else:
logger.warning(
f"While replacing {special}, the following existing '{f}'-values have been overwritten with {new_vals[f]}:\n{frame.loc[not_empty, cols[f]]}")
frame.loc[replace_this, cols[f]] = new_vals[f]
return frame
for special, instead in special2label.items():
select_special = df[cols['special']] == special
df.loc[select_special, feature_cols] = repl_spec(df.loc[select_special, feature_cols].copy(), instead=instead, special=special)
if df[cols['special']].isna().all():
df.drop(columns=cols['special'], inplace=True)
if tmp_index:
df.index = ix
if not inplace:
return df
[docs]def merge_changes(left, right, *args):
"""
Merge two `changes` into one, e.g. `b3` and `+#7` to `+#7b3`.
Uses: :py:func:`changes2list`
"""
all_changes = [changes2list(changes, sort=False) for changes in (left, right, *args)]
res = sum(all_changes, [])
res = sorted(res, key=lambda x: int(x[3]), reverse=True)
return ''.join(e[0] for e in res)
[docs]@function_logger
def propagate_keys(df, volta_structure=None, globalkey='globalkey', localkey='localkey', add_bool=True):
"""
| Propagate information about global keys and local keys throughout the dataframe.
| Pass split harmonies for one piece at a time. For concatenated pieces, use apply().
Uses: :py:func:`series_is_minor`
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe containing DCML chord labels that have been split by split_labels().
volta_structure: :obj:`dict`, optional
{first_mc -> {volta_number -> [mc1, mc2...]} } dictionary as you can get it from
``Score.mscx.volta_structure``. This allows for correct propagation into
second and other voltas.
globalkey, localkey : :obj:`str`, optional
In case you renamed the columns, pass column names.
add_bool : :obj:`bool`, optional
Pass True if you want to add two boolean columns which are true if the respective key is
a minor key.
"""
df = df.copy()
nunique = df[globalkey].nunique()
assert nunique > 0, "No global key specified."
if nunique > 1:
raise NotImplementedError("Several global keys not accepted at the moment.")
logger.debug('Extending global key to all harmonies')
global_key = df[globalkey].iloc[0]
if pd.isnull(global_key):
global_key = df[globalkey].dropna().iloc[0]
logger.warning(
f"Global key is not specified in the first label. Using '{global_key}' from index {df[df[globalkey] == global_key].index[0]}")
df.loc[:, globalkey] = global_key
global_minor = series_is_minor(df[globalkey])
logger.debug('Extending local keys to all harmonies')
if pd.isnull(df[localkey].iloc[0]):
one = 'i' if global_minor.iloc[0] else 'I'
df.iloc[0, df.columns.get_loc(localkey)] = one
if volta_structure is not None and volta_structure != {}:
if 'mc' in df.columns:
volta_mcs = defaultdict(list)
for volta_dict in volta_structure.values():
for volta_no, mcs in volta_dict.items():
volta_mcs[volta_no].extend(mcs)
volta_exclusion = {volta_no: [mc for vn, mcs in volta_mcs.items() for mc in mcs if vn != volta_no] for
volta_no in volta_mcs.keys()}
for volta_no in sorted(volta_exclusion.keys(), reverse=True):
selector = ~df.mc.isin(volta_exclusion[volta_no])
df.loc[selector, localkey] = df.loc[selector, localkey].fillna(method='ffill')
else:
logger.info("Dataframe needs to have a 'mc' column. Ignoring volta_structure.")
df[localkey].fillna(method='ffill', inplace=True)
else:
df[localkey].fillna(method='ffill', inplace=True)
if add_bool:
local_minor = series_is_minor(df[localkey])
gm = f"{globalkey}_is_minor"
lm = f"{localkey}_is_minor"
df[gm] = global_minor
df[lm] = local_minor
return df
[docs]@function_logger
def propagate_pedal(df, relative=True, drop_pedalend=True, cols={}):
"""
Propagate the pedal note for all chords within square brackets.
By default, the note is expressed in relation to each label's localkey.
Uses: :py:func:`rel2abs_key`, :py:func:`abs2rel_key`
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe containing DCML chord labels that have been split by split_labels()
and where the keys have been propagated using propagate_keys().
relative : :obj:`bool`, optional
Pass False if you want the pedal note to stay the same even if the localkey changes.
drop_pedalend : :obj:`bool`, optional
Pass False if you don't want the column with the ending brackets to be dropped.
cols : :obj:`dict`, optional
In case the column names for ``['pedal','pedalend', 'globalkey', 'localkey']`` deviate, pass a dict, such as
.. code-block:: python
{'pedal': 'pedal_col_name',
'pedalend': 'pedalend_col_name',
'globalkey': 'globalkey_col_name',
'localkey': 'localkey_col_name'}
"""
df = df.copy()
### If the index is not unique, it has to be temporarily replaced
tmp_index = not df.index.is_unique
if tmp_index:
ix = df.index
df.reset_index(drop=True, inplace=True)
features = ['pedal', 'pedalend', 'globalkey', 'localkey']
for col in features:
if not col in cols:
cols[col] = col
pedal, pedalend = cols['pedal'], cols['pedalend']
logger.debug('Extending pedal notes to concerned harmonies')
beginnings = df.loc[df[pedal].notna(), ['mc', pedal]]
endings = df.loc[df[pedalend].notna(), ['mc', pedalend]]
n_b, n_e = len(beginnings), len(endings)
def make_comparison():
return pd.concat([beginnings.reset_index(drop=True), endings.reset_index(drop=True)], axis=1).astype({'mc': 'Int64'})
assert n_b == n_e, f"{n_b} organ points started, {n_e} ended:\n{make_comparison()}"
if relative:
assert df[cols[
'localkey']].notna().all(), "Local keys must first be propagated using propagate_keys(), no NaNs allowed."
for (fro, ped), to in zip(beginnings[pedal].items(), endings[pedalend].index):
try:
section = df.loc[fro:to].index
except:
logger.error(
f"Slicing of the DataFrame did not work from {fro} to {to}. Index looks like this:\n{df.head().index}")
localkeys = df.loc[section, cols['localkey']]
if localkeys.nunique() > 1:
first_localkey = localkeys.iloc[0]
globalkeys = df.loc[section, cols['globalkey']].unique()
assert len(globalkeys) == 1, "Several globalkeys appearing within the same organ point."
global_minor = globalkeys[0].islower()
key2pedal = {
key: ped if key == first_localkey else abs2rel_key(rel2abs_key(ped, first_localkey, global_minor), key,
global_minor) for key in localkeys.unique()}
logger.debug(
f"Pedal note {ped} has been transposed relative to other local keys within a global {'minor' if global_minor else 'major'} context: {key2pedal}")
pedals = pd.Series([key2pedal[key] for key in localkeys], index=section)
else:
pedals = pd.Series(ped, index=section)
df.loc[section, pedal] = pedals
if drop_pedalend:
df = df.drop(columns=pedalend)
if tmp_index:
df.index = ix
return df
[docs]@function_logger
def compute_chord_tones(df, bass_only=False, expand=False, cols={}):
"""
Compute the chord tones for DCML harmony labels. They are returned as lists
of tonal pitch classes in close position, starting with the bass note. The
tonal pitch classes represent intervals relative to the local tonic:
-2: Second below tonic
-1: fifth below tonic
0: tonic
1: fifth above tonic
2: second above tonic, etc.
The labels need to have undergone :py:func:`split_labels` and :py:func:`propagate_keys`.
Pedal points are not taken into account.
Uses: :py:func:`features2tpcs`
Parameters
----------
df : :obj:`pandas.DataFrame`
Dataframe containing DCML chord labels that have been split by split_labels()
and where the keys have been propagated using propagate_keys(add_bool=True).
bass_only : :obj:`bool`, optional
Pass True if you need only the bass note.
expand : :obj:`bool`, optional
Pass True if you need chord tones and added tones in separate columns.
cols : :obj:`dict`, optional
In case the column names for ``['mc', 'numeral', 'form', 'figbass', 'changes', 'relativeroot', 'localkey', 'globalkey']`` deviate, pass a dict, such as
.. code-block:: python
{'mc': 'mc',
'numeral': 'numeral_col_name',
'form': 'form_col_name',
'figbass': 'figbass_col_name',
'changes': 'changes_col_name',
'relativeroot': 'relativeroot_col_name',
'localkey': 'localkey_col_name',
'globalkey': 'globalkey_col_name'}
You may also deactivate columns by setting them to None, e.g. {'changes': None}
Returns
-------
:obj:`pandas.Series` or :obj:`pandas.DataFrame`
For every row of `df` one tuple with chord tones, expressed as tonal pitch classes.
If `expand` is True, the function returns a DataFrame with four columns:
Two with tuples for chord tones and added tones, one with the chord root,
and one with the bass note.
"""
df = df.copy()
### If the index is not unique, it has to be temporarily replaced
tmp_index = not df.index.is_unique
if tmp_index:
ix = df.index
df.reset_index(drop=True, inplace=True)
features = ['mc', 'numeral', 'form', 'figbass', 'changes', 'relativeroot', 'localkey', 'globalkey']
for col in features:
if col in df.columns and not col in cols:
cols[col] = col
local_minor, global_minor = f"{cols['localkey']}_is_minor", f"{cols['globalkey']}_is_minor"
if not local_minor in df.columns:
df[local_minor] = series_is_minor(df[cols['localkey']], is_name=False)
logger.debug(f"Boolean column '{local_minor}' created.'")
if not global_minor in df.columns:
df[global_minor] = series_is_minor(df[cols['globalkey']], is_name=True)
logger.debug(f"Boolean column '{global_minor}' created.'")
param_cols = {col: cols[col] for col in ['numeral', 'form', 'figbass', 'changes', 'relativeroot', 'mc'] if cols[col] is not None}
param_cols['minor'] = local_minor
param_tuples = list(df[param_cols.values()].itertuples(index=False, name=None))
#result_dict = {t: features2tpcs(**{a:b for a, b in zip(param_cols.keys(), t)}, bass_only=bass_only, merge_tones=not expand, logger=logger) for t in set(param_tuples)}
result_dict = {}
if bass_only:
default = None
elif not expand:
default = tuple()
else:
default = {
'chord_tones': tuple(),
'added_tones': tuple(),
'root': None,
}
for t in set(param_tuples):
try:
result_dict[t] = features2tpcs(**{a: b for a, b in zip(param_cols.keys(), t)}, bass_only=bass_only,
merge_tones=not expand, logger=logger)
except:
result_dict[t] = default
logger.warning(str(sys.exc_info()[1]))
if expand:
res = pd.DataFrame([result_dict[t] for t in param_tuples], index=df.index)
res['bass_note'] = res.chord_tones.apply(lambda l: np.nan if pd.isnull(l) or len(l) == 0 else l[0])
res[['root', 'bass_note']] = res[['root', 'bass_note']].astype('Int64')
else:
res = pd.Series([result_dict[t] for t in param_tuples], index=df.index)
if tmp_index:
res.index = ix
return res
[docs]@function_logger
def features2tpcs(numeral, form=None, figbass=None, changes=None, relativeroot=None, key='C', minor=None,
merge_tones=True, bass_only=False, mc=None):
"""
Given the features of a chord label, this function returns the chord tones
in the order of the inversion, starting from the bass note. The tones are
expressed as tonal pitch classes, where -1=F, 0=C, 1=G etc.
Uses: :py:func:`~.utils.changes2list`, :py:func:`~.utils.name2fifths`, :py:func:`~.utils.resolve_relative_keys`, :py:func:`~.utils.roman_numeral2fifths`,
:py:func:`~.utils.sort_tpcs`, :py:func:`~.utils.str_is_minor`
Parameters
----------
numeral: :obj:`str`
Roman numeral of the chord's root
form: {None, 'M', 'o', '+' '%'}, optional
Indicates the chord type if not a major or minor triad (for which ``form`` is None).
'%' and 'M' can only occur as tetrads, not as triads.
figbass: {None, '6', '64', '7', '65', '43', '2'}, optional
Indicates chord's inversion. Pass None for triad root position.
changes: :obj:`str`, optional
Added steps such as '+6' or suspensions such as '4' or any combination such as (9+64).
Numbers need to be in descending order.
relativeroot: :obj:`str`, optional
Pass a Roman scale degree if `numeral` is to be applied to a different scale
degree of the local key, as in 'V65/V'
key : :obj:`str` or :obj:`int`, optional
The local key expressed as the root's note name or a tonal pitch class.
If it is a name and `minor` is `None`, uppercase means major and lowercase minor.
If it is a tonal pitch class, `minor` needs to be specified.
minor : :obj:`bool`, optional
Pass True for minor and False for major. Can be omitted if `key` is a note name.
This affects calculation of chords related to III, VI and VII.
merge_tones : :obj:`bool`, optional
Pass False if you want the function to return two tuples, one with (potentially suspended)
chord tones and one with added notes.
bass_only : :obj:`bool`, optional
Return only the bass note instead of all chord tones.
mc : int or str
Pass measure count to display it in warnings.
"""
if pd.isnull(numeral) or numeral == '@none':
if bass_only or merge_tones:
return np.nan
else:
return {
'chord_tones': np.nan,
'added_tones': np.nan,
'root': np.nan,
}
form, figbass, changes, relativeroot = tuple(
'' if pd.isnull(val) else val for val in (form, figbass, changes, relativeroot))
label = f"{numeral}{form}{figbass}{'(' + changes + ')' if changes != '' else ''}{'/' + relativeroot if relativeroot != '' else ''}"
MC = '' if mc is None else f'MC {mc}: '
if minor is None:
try:
minor = str_is_minor(key, is_name=True)
logger.debug(f"Mode inferred from {key}.")
except:
raise ValueError(f"If parameter 'minor' is not specified, 'key' needs to be a string, not {key}")
key = name2fifths(key)
if form in ['%', 'M', '+M']:
assert figbass in ['7', '65', '43',
'2'], f"{MC}{label}: {form} requires figbass (7, 65, 43, or 2) since it specifies a chord's seventh."
if relativeroot != '':
resolved = resolve_relative_keys(relativeroot, minor)
rel_minor = str_is_minor(resolved, is_name=False)
transp = roman_numeral2fifths(resolved, minor)
logger.debug(
f"{MC}Chord applied to {relativeroot}. Therefore transposing it by {transp} fifths.")
return features2tpcs(numeral=numeral, form=form, figbass=figbass, relativeroot=None, changes=changes,
key=key + transp, minor=rel_minor, merge_tones=merge_tones, bass_only=bass_only, mc=mc,
logger=logger)
if numeral.lower() == '#vii' and not minor:
# logger.warning(
# f"{MC}{label} in a major context is most probably an annotation error.")
logger.warning(f"{MC}{numeral} in major context corrected to {numeral[1:]}.")
numeral = numeral[1:]
root_alteration, num_degree = split_scale_degree(numeral, count=True, logger=logger)
# build 2-octave diatonic scale on C major/minor
root = ['I', 'II', 'III', 'IV', 'V', 'VI', 'VII'].index(num_degree.upper())
tpcs = 2 * [i + key for i in (0, 2, -3, -1, 1, -4, -2)] if minor else 2 * [i + key for i in (0, 2, 4, -1, 1, 3, 5)]
# starting the scale from chord root, i.e. root will be tpcs[0], the chord's seventh tpcs[6] etc.
tpcs = tpcs[root:] + tpcs[:root]
root = tpcs[0] + 7 * root_alteration
tpcs[0] = root # octave stays diatonic, is not altered
logger.debug(f"{num_degree}: The {'minor' if minor else 'major'} scale starting from the root: {tpcs}")
def set_iv(chord_interval, interval_size):
""" Add to the interval of a given chord interval in `tpcs` (both viewed from the root note).
Parameters
----------
chord_interval : :obj:`int`
Pass 0 for the root note, 2 for the third, 8 for the ninth etc.
interval_size : :obj:`int`
Stack-of-fifths interval.
"""
nonlocal tpcs, root
iv = root + interval_size
tpcs[chord_interval] = iv
tpcs[chord_interval + 7] = iv
is_triad = figbass in ['', '6', '64']
is_seventh_chord = figbass in ['7', '65', '43', '2']
if not is_triad and not is_seventh_chord:
raise ValueError(f"{MC}{figbass} is not a valid chord inversion.")
if form == 'o':
set_iv(2, -3)
set_iv(4, -6)
if is_seventh_chord:
set_iv(6, -9)
elif form == '%':
set_iv(2, -3)
set_iv(4, -6)
set_iv(6, -2)
elif form == '+':
set_iv(2, 4)
set_iv(4, 8)
if is_seventh_chord:
set_iv(6, -2)
elif form == '+M':
set_iv(2, 4)
set_iv(4, 8)
set_iv(6, 5)
else: # triad with or without major or minor seven
set_iv(4, 1)
if num_degree.isupper():
set_iv(2, 4)
else:
set_iv(2, -3)
if form == 'M':
set_iv(6, 5)
elif is_seventh_chord:
set_iv(6, -2)
tone_functions = (0, 2, 4, 6) if is_seventh_chord else (0, 2, 4)
root_position = {i: [tpcs[i]] for i in tone_functions}
replacements = {i: [] for i in tone_functions}
def replace_chord_tone(which, by):
nonlocal root_position, replacements
if which in root_position:
root_position[which] = []
replacements[which].insert(0, by)
else:
logger.warning(f"Only chord tones [0,2,4,(6)] can be replaced, not {which}")
# apply changes
alts = changes2list(changes, sort=False)
added_notes = []
for full, add_remove, acc, chord_interval in alts:
added = add_remove == '+'
substracted = add_remove == '-'
replacing_upper = add_remove == '^'
replacing_lower = add_remove == 'v'
chord_interval = int(chord_interval) - 1
### From here on, `chord_interval` is decremented, i.e. the root is 0, the seventh is 6 etc. (just like in `tpcs`)
if (chord_interval == 0 and not substracted) or chord_interval > 13:
logger.warning(
f"{MC}Change {full} is meaningless and ignored because it concerns chord tone {chord_interval + 1}.")
continue
next_octave = chord_interval > 7
shift = 7 * (acc.count('#') - acc.count('b'))
new_val = tpcs[chord_interval] + shift
if substracted:
if chord_interval not in tone_functions:
logger.warning(
f"{MC}The change {full} has no effect because it concerns an interval which is not implied by {numeral}{form}{figbass}.")
else:
root_position[chord_interval] = []
elif added:
added_notes.append(new_val)
elif next_octave:
if any((replacing_lower, replacing_upper, substracted)):
logger.warning(f"{MC}{full[0]} has no effect in {full} because the interval is larger than an octave.")
added_notes.append(new_val)
elif chord_interval in [1, 3, 5]: # these are changes to scale degree 2, 4, 6 that replace the lower neighbour unless they have a # or ^
if '#' in acc or replacing_upper:
if '#' in acc and replacing_upper:
logger.warning(f"{MC}^ is redundant in {full}.")
if chord_interval == 5 and is_triad: # leading tone to 7 but not in seventh chord
added_notes.append(new_val)
else:
replace_chord_tone(chord_interval + 1, new_val)
else:
if replacing_lower:
logger.warning(f"{MC}v is redundant in {full}.")
replace_chord_tone(chord_interval - 1, new_val)
else: # chord tone alterations
if replacing_lower:
# TODO: This must be possible, e.g. V(6v5) where 5 is suspension of 4
logger.warning(f"{MC}{full} -> chord tones cannot replace neighbours, use + instead.")
elif chord_interval == 6 and figbass != '7': # 7th are a special case:
if figbass == '': # in root position triads they are added
# TODO: The standard is lacking a distinction, because the root in root pos. can also be replaced from below!
added_notes.append(new_val)
elif figbass in ['6', '64'] or '#' in acc: # in inverted triads they replace the root, as does #7
replace_chord_tone(0, new_val)
else: # otherwise they are unclear
logger.warning(
f"{MC}In seventh chords, such as {label}, it is not clear whether the {full} alters the 7 or replaces the 8 and should not be used.")
elif tpcs[chord_interval] == new_val:
logger.warning(
f"{MC}The change {full} has no effect in {numeral}{form}{figbass}")
else:
root_position[chord_interval] = [new_val]
figbass2bass = {
'': 0,
'7': 0,
'6': 1,
'65': 1,
'64': 2,
'43': 2,
'2': 3
}
bass = figbass2bass[figbass]
chord_tones = []
for tf in tone_functions[bass:] + tone_functions[:bass]:
chord_tone, replacing_tones = root_position[tf], replacements[tf]
if chord_tone == replacing_tones == []:
logger.debug(f"{MC}{label} results in a chord without {tf + 1}.")
if chord_tone != []:
chord_tones.append(chord_tone[0])
if replacing_tones != []:
logger.warning(f"{MC}{label} results in a chord tone {tf + 1} AND its replacement(s) {replacing_tones}.")
chord_tones.extend(replacing_tones)
bass_tpc = chord_tones[0]
if bass_only:
return bass_tpc
elif merge_tones:
return tuple(sort_tpcs(chord_tones + added_notes, start=bass_tpc))
else:
return {
'chord_tones': tuple(chord_tones),
'added_tones': tuple(added_notes),
'root': root,
}
########################################################################################################################
# MOMENTARILY NOT IN USE:
########################################################################################################################
[docs]@function_logger
def chord2tpcs(chord, regex=None, **kwargs):
"""
Split a chord label into its features and apply features2tpcs().
Uses: features2tpcs()
Parameters
----------
chord : :obj:`str`
Chord label that can be split into the features ['numeral', 'form', 'figbass', 'changes', 'relativeroot'].
regex : :obj:`re.Pattern`, optional
Compiled regex with named groups for the five features. By default, the current version of the DCML harmony
annotation standard is used.
**kwargs :
arguments for features2tpcs (pass MC to show it in warnings!)
"""
if regex is None:
regex = DCML_REGEX
chord_features = re.match(regex, chord)
assert chord_features is not None, f"{chord} does not match the regex."
chord_features = chord_features.groupdict()
numeral, form, figbass, changes, relativeroot = tuple(chord_features[f] for f in ('numeral', 'form', 'figbass', 'changes', 'relativeroot'))
return features2tpcs(numeral=numeral, form=form, figbass=figbass, changes=changes, relativeroot=relativeroot,
logger=logger, **kwargs)