def make_index_unique(index: pd.Index, join: str = '-'):
"""Makes the index unique by appending '1', '2', etc.
The first occurance of a non-unique value is ignored.
Parameters
----------
join
The connecting string between name and integer.
Examples
--------
>>> adata1 = sc.AnnData(np.ones((3, 2)), {'obs_names': ['a', 'b', 'c']})
>>> adata2 = sc.AnnData(np.zeros((3, 2)), {'obs_names': ['d', 'b', 'b']})
>>> adata = adata1.concatenate(adata2)
>>> adata.obs_names
Index(['a', 'b', 'c', 'd', 'b', 'b'], dtype='object')
>>> adata.obs_names_make_unique()
>>> adata.obs_names
Index(['a', 'b', 'c', 'd', 'b-1', 'b-2'], dtype='object')
"""
if index.is_unique:
return index
from collections import defaultdict
values = index.values
indices_dup = index.duplicated(keep='first')
values_dup = values[indices_dup]
counter = defaultdict(lambda: 0)
for i, v in enumerate(values_dup):
counter[v] += 1
values_dup[i] += join + str(counter[v])
values[indices_dup] = values_dup
index = pd.Index(values)
return index
def make_index_unique(index: pd.Index, join: str = "-"):
"""
Makes the index unique by appending a number string to each duplicate index element:
'1', '2', etc.
If a tentative name created by the algorithm already exists in the index, it tries
the next integer in the sequence.
The first occurrence of a non-unique value is ignored.
Parameters
----------
join
The connecting string between name and integer.
Examples
--------
>>> from anndata import AnnData
>>> adata = AnnData(np.ones((2, 3)), var=pd.DataFrame(index=["a", "a", "b"]))
>>> adata.var_names
Index(['a', 'a', 'b'], dtype='object')
>>> adata.var_names_make_unique()
>>> adata.var_names
Index(['a', 'a-1', 'b'], dtype='object')
"""
if index.is_unique:
return index
from collections import Counter
values = index.values.copy()
indices_dup = index.duplicated(keep="first")
values_dup = values[indices_dup]
values_set = set(values)
counter = Counter()
issue_interpretation_warning = False
example_colliding_values = []
for i, v in enumerate(values_dup):
while True:
counter[v] += 1
tentative_new_name = v + join + str(counter[v])
if tentative_new_name not in values_set:
values_set.add(tentative_new_name)
values_dup[i] = tentative_new_name
break
issue_interpretation_warning = True
if len(example_colliding_values) < 5:
example_colliding_values.append(tentative_new_name)
if issue_interpretation_warning:
warnings.warn(
f"Suffix used ({join}[0-9]+) to deduplicate index values may make index "
+
"values difficult to interpret. There values with a similar suffixes in "
+ "the index. Consider using a different delimiter by passing " +
"`join={delimiter}`" +
"Example key collisions generated by the make_index_unique algorithm: "
+ str(example_colliding_values))
values[indices_dup] = values_dup
index = pd.Index(values, name=index.name)
return index
def assert_index(obj: pd.Index):
r"""Check if index is conform to audformat."""
if isinstance(obj, pd.MultiIndex) and len(obj.levels) == 2:
if obj.has_duplicates:
max_display = 10
duplicates = obj[obj.duplicated()]
msg_tail = '\n...' if len(duplicates) > max_display else ''
msg_duplicates = '\n'.join([
str(duplicate)
for duplicate in duplicates[:max_display].tolist()
])
raise ValueError('Found duplicates:\n'
f'{msg_duplicates}{msg_tail}')
if not (obj.names[0] == audformat.define.IndexField.START
and obj.names[1] == audformat.define.IndexField.END):
expected_names = [
audformat.define.IndexField.START,
audformat.define.IndexField.END,
]
raise ValueError('Found two levels with names '
f'{obj.names}, '
f'but expected names '
f'{expected_names}.')
if not pd.api.types.is_timedelta64_dtype(obj.levels[0].dtype):
raise ValueError(
"Level 'start' must contain values of type 'timedelta64[ns]'.")
if not pd.api.types.is_timedelta64_dtype(obj.levels[1].dtype):
raise ValueError(
"Level 'end' must contain values of type 'timedelta64[ns]'.")
else:
audformat.assert_index(obj)
def _maybe_check_integrity(self, concat_index: Index):
if self.verify_integrity:
if not concat_index.is_unique:
overlap = concat_index[concat_index.duplicated()].unique()
raise ValueError(
"Indexes have overlapping values: "
"{overlap!s}".format(overlap=overlap)
)
def make_index_unique(index: pd.Index, join: str = "-"):
"""
Makes the index unique by appending a number string to each duplicate index element: '1', '2', etc.
If a tentative name created by the algorithm already exists in the index, it tries the next integer in the sequence.
The first occurrence of a non-unique value is ignored.
Parameters
----------
join
The connecting string between name and integer.
Examples
--------
>>> from anndata import AnnData
>>> adata1 = AnnData(np.ones((3, 2)), dict(obs_names=['a', 'b', 'c']))
>>> adata2 = AnnData(np.zeros((3, 2)), dict(obs_names=['d', 'b', 'b']))
>>> adata = adata1.concatenate(adata2)
>>> adata.obs_names
Index(['a', 'b', 'c', 'd', 'b', 'b'], dtype='object')
>>> adata.obs_names_make_unique()
>>> adata.obs_names
Index(['a', 'b', 'c', 'd', 'b-1', 'b-2'], dtype='object')
"""
if index.is_unique:
return index
from collections import defaultdict
values = index.values
values_set = set(values)
indices_dup = index.duplicated(keep="first")
values_dup = values[indices_dup]
counter = defaultdict(lambda: 0)
for i, v in enumerate(values_dup):
while True:
counter[v] += 1
tentative_new_name = v + join + str(counter[v])
if tentative_new_name not in values_set:
values_set.add(tentative_new_name)
values_dup[i] = tentative_new_name
break
values[indices_dup] = values_dup
index = pd.Index(values)
return index
def check_no_dupes(idx: pd.Index, name: str) -> bool:
dupes = idx.duplicated().any()
if dupes:
warn(f"Duplicated {name}: {idx[idx.duplicated(False)].sort_values()}")
return not dupes
def index_has_duplicates(index: Index) -> bool:
"""Indicates whether a DataFrame's Index contains any duplicates."""
return index.duplicated().any()
def test_index_has_no_duplicates(self, index: Index):
self.assertEqual(first=index.duplicated().any(),
second=False,
msg='The index contains duplicate values.')
