def _link_index(self, df_a, df_b):
# Index name conflicts do not occur. They are handled in the
# decorator.
left_on = listify(self.left_on)
right_on = listify(self.right_on)
if self.on:
left_on, right_on = listify(self.on), listify(self.on)
if not left_on or not right_on:
raise ValueError("no column labels given")
if len(left_on) != len(right_on):
raise ValueError(
"length of left and right keys needs to be the same")
blocking_keys = ["blocking_key_%d" % i for i, v in enumerate(left_on)]
# make a dataset for the data on the left
data_left = df_a[left_on].dropna(axis=0, how='any', inplace=False)
data_left.columns = blocking_keys
data_left['index_x'] = data_left.index
# make a dataset for the data on the right
data_right = df_b[right_on].dropna(axis=0, how='any', inplace=False)
data_right.columns = blocking_keys
data_right['index_y'] = data_right.index
# merge the dataframes
pairs = data_left.merge(data_right, how='inner',
on=blocking_keys).set_index(
['index_x', 'index_y'])
return pairs.index.rename([df_a.index.name, df_b.index.name])
def _link_index(self, df_a, df_b):
left_on, right_on = self._get_left_and_right_on()
left_on = listify(left_on)
right_on = listify(right_on)
blocking_keys = ["blocking_key_%d" % i for i, v in enumerate(left_on)]
# make a dataset for the data on the left
# 1. make a dataframe
# 2. rename columns
# 3. add index col
# 4. drop na (last step to presever index)
data_left = pandas.DataFrame(df_a[left_on], copy=False)
data_left.columns = blocking_keys
data_left['index_x'] = numpy.arange(len(df_a))
data_left.dropna(axis=0, how='any', subset=blocking_keys, inplace=True)
# make a dataset for the data on the right
data_right = pandas.DataFrame(df_b[right_on], copy=False)
data_right.columns = blocking_keys
data_right['index_y'] = numpy.arange(len(df_b))
data_right.dropna(axis=0,
how='any',
subset=blocking_keys,
inplace=True)
# merge the dataframes
pairs_df = data_left.merge(data_right, how='inner', on=blocking_keys)
return pandas.MultiIndex(
levels=[df_a.index.values, df_b.index.values],
labels=[pairs_df['index_x'].values, pairs_df['index_y'].values],
verify_integrity=False)
def _compute(self, pairs, x, x_link=None):
# start the timer for the comparing step
start_time = time.time()
sublabels_left = self._get_labels_left(validate=x)
df_a_indexed = frame_indexing(x[sublabels_left], pairs, 0)
if x_link is None:
sublabels_right = self._get_labels_right(validate=x)
df_b_indexed = frame_indexing(x[sublabels_right], pairs, 1)
else:
sublabels_right = self._get_labels_right(validate=x_link)
df_b_indexed = frame_indexing(x_link[sublabels_right], pairs, 1)
# log timing
# index_time = time.time() - start_time
features = []
for feat in self.features:
lbl1 = feat.labels_left
lbl2 = feat.labels_right
data1 = tuple([df_a_indexed[lbl] for lbl in listify(lbl1)])
data2 = tuple([df_b_indexed[lbl] for lbl in listify(lbl2)])
result = feat._compute(data1, data2)
features.append((result, feat.label))
features = self._union(features, pairs)
# log timing
n = pairs.shape[0]
i_max = '?' if self._i_max is None else self._i_max
eta = time.time() - start_time
self._eta.append(eta)
self._n.append(n)
# log
logging.info("comparing [{:d}/{}] - time: {:.2f}s - pairs: {}".format(
self._i, i_max, eta, n))
# log total
if self._output_log_total:
n_total = np.sum(self._n)
eta_total = np.sum(self._eta)
logging.info(
"comparing [{:d}/{}] - time: {:.2f}s - pairs_total: {}".format(
self._i, i_max, eta_total, n_total))
self._i += 1
return features
def compute(self, pairs, x, x_link=None):
"""Compare the records of each record pair.
Calling this method starts the comparing of records.
Parameters
----------
pairs : pandas.MultiIndex
A pandas MultiIndex with the record pairs to compare. The indices
in the MultiIndex are indices of the DataFrame(s) to link.
x : pandas.DataFrame
The DataFrame to link. If `x_link` is given, the comparing is a
linking problem. If `x_link` is not given, the problem is one of
deduplication.
x_link : pandas.DataFrame, optional
The second DataFrame.
Returns
-------
pandas.DataFrame
A pandas DataFrame with feature vectors, i.e. the result of
comparing each record pair.
"""
if not is_pandas_2d_multiindex(pairs):
raise ValueError(
"expected pandas.MultiIndex with record pair indices "
"as first argument")
if not isinstance(x, pandas.DataFrame):
raise ValueError("expected pandas.DataFrame as second argument")
if x_link is not None and not isinstance(x_link, pandas.DataFrame):
raise ValueError("expected pandas.DataFrame as third argument")
labels_left = listify(self.labels_left)
labels_right = listify(self.labels_right)
df_a = frame_indexing(x[labels_left], pairs, 0)
if x_link is None:
df_b = frame_indexing(x[labels_right], pairs, 1)
else:
df_b = frame_indexing(x_link[labels_right], pairs, 1)
data1 = tuple([df_a[lbl] for lbl in listify(self.labels_left)])
data2 = tuple([df_b[lbl] for lbl in listify(self.labels_right)])
results = self._compute(*tuple(data1 + data2))
return results
def __init__(self,
left_on=None,
right_on=None,
max_nulls=0,
max_non_matches=0,
windows=1,
**kwargs):
super(NeighbourhoodBlock, self).__init__(
left_on=left_on, right_on=right_on, **kwargs)
self.max_nulls = max_nulls
self.max_non_matches = max_non_matches
self.windows = listify(windows)
def _link_index(self, df_a, df_b):
if self.on is not None:
if self.left_on is not None or self.right_on is not None:
raise IndexError('Can only pass argument "on" OR "left_on" '
'and "right_on", not a combination of both.')
left_on = right_on = listify(self.on)
else:
if self.left_on is None and self.right_on is None:
raise IndexError('pass argument "on" OR "left_on" and '
'"right_on" at class initalization.')
elif self.left_on is None:
raise IndexError('Argument "left_on" is missing '
'at class initalization.')
elif self.right_on is None:
raise IndexError('Argument "right_on" is missing '
'at class initalization.')
else:
left_on = listify(self.left_on)
right_on = listify(self.right_on)
blocking_keys = ["blocking_key_%d" % i for i, v in enumerate(left_on)]
# make a dataset for the data on the left
data_left = df_a[left_on].dropna(axis=0, how='any', inplace=False)
data_left.columns = blocking_keys
data_left['index_x'] = data_left.index
# make a dataset for the data on the right
data_right = df_b[right_on].dropna(axis=0, how='any', inplace=False)
data_right.columns = blocking_keys
data_right['index_y'] = data_right.index
# merge the dataframes
pairs = data_left.merge(
data_right, how='inner', on=blocking_keys
).set_index(['index_x', 'index_y'])
return pairs.index
def _get_labels_right(self, validate=None):
"""Get all labels of the right dataframe."""
labels = []
for compare_func in self.features:
labels = labels + listify(compare_func.labels_right)
# check requested labels (for better error messages)
if not is_label_dataframe(labels, validate):
error_msg = "label is not found in the dataframe"
raise KeyError(error_msg)
return unique(labels)
def get_normalized_linkage_params():
def default_on_possibilities():
yield self.left_on
yield self.right_on
yield [
c for c in dfs[0].columns
if all(c in df.columns for df in dfs)
]
default_on = next(
iter(
filter(lambda x: x is not None,
default_on_possibilities())))
key_columns = [
listify(side_on or default_on)
for side_on in [self.left_on, self.right_on]
]
key_cols = set(map(len, key_columns))
n_key_cols = next(iter(key_cols))
if (len(key_cols) > 1) or (n_key_cols == 0):
raise IndexError('Invalid blocking keys')
combined_ranks = numpy.vstack([
pandas.concat([df[col] for df, col in zip(dfs, col_grp)]).rank(
method='dense',
na_option='keep').fillna(0).astype(int).values - 1
for col_grp in zip(*key_columns)
]).astype(float).T
combined_ranks[combined_ranks < 0] = numpy.nan
blocks, indices = deduped_blocks_and_indices(
blocks=combined_ranks,
indices=split_to_match(numpy.arange(len(combined_ranks)), dfs))
n_keys = blocks.shape[1]
windows = self.windows + self.windows[-1:] * (
n_keys - len(self.windows))
if (len(windows) > n_keys) or not all(
isinstance(w, int) and (w > 0) and (w % 2 == 1)
for w in windows):
raise ValueError(
'Windows must be positive odd integers and the maximum'
'number allowed is the number of blocking keys'
)
rank_distance_limits = (
numpy.array(windows) // 2).astype(float).reshape((1, -1))
return blocks, indices, rank_distance_limits
def _get_labels(self, frame_i, validate=None):
"""Get all labels.
Parameters
----------
frame_i : str
A string, 'left' or 'right', incidating the dataframe to collect
labels from.
"""
labels = []
for compare_func in self._compare_functions:
labels = labels + listify(compare_func[frame_i])
# check requested labels (for better error messages)
if not is_label_dataframe(labels, validate):
error_msg = "label is not found in the dataframe"
raise KeyError(error_msg)
return unique(labels)
def compute(self, pairs, x, x_link=None):
"""Compare the records of each record pair.
Calling this method starts the comparing of records.
Parameters
----------
pairs : pandas.MultiIndex
A pandas MultiIndex with the record pairs to compare. The indices
in the MultiIndex are indices of the DataFrame(s) to link.
x : pandas.DataFrame
The DataFrame to link. If `x_link` is given, the comparing is a
linking problem. If `x_link` is not given, the problem is one of
deduplication.
x_link : pandas.DataFrame, optional
The second DataFrame.
Returns
-------
pandas.DataFrame
A pandas DataFrame with feature vectors, i.e. the result of
comparing each record pair.
"""
sublabels_left = self._get_labels_left(validate=x)
df_a_indexed = self._loc2(x[sublabels_left], pairs, 0)
if x_link is None:
sublabels_right = self._get_labels_right(validate=x)
df_b_indexed = self._loc2(x[sublabels_right], pairs, 1)
else:
sublabels_right = self._get_labels_right(validate=x_link)
df_b_indexed = self._loc2(x_link[sublabels_right], pairs, 1)
results = pandas.DataFrame(index=pairs)
label_num = 0 # make a label is label is None
for f, lbl1, lbl2, label, args, kwargs in self._compare_functions:
data1 = tuple([df_a_indexed[lbl] for lbl in listify(lbl1)])
data2 = tuple([df_b_indexed[lbl] for lbl in listify(lbl2)])
c = f(*tuple(data1 + data2 + args), **kwargs)
if isinstance(c, (pandas.Series, pandas.DataFrame)):
c = c.values # convert pandas into numpy
if label is not None:
label = listify(label)
n_cols = 1 if len(c.shape) == 1 else c.shape[1]
labels = []
for i in range(0, n_cols):
label_val = label[i] if label is not None else label_num
label_num += 1
labels.append(label_val)
results[label_val] = c
return results
def _link_index(self, df_a, df_b):
if self.on is not None:
if self.left_on is not None or self.right_on is not None:
raise IndexError('Can only pass argument "on" OR "left_on" '
'and "right_on", not a combination of both.')
left_on = right_on = listify(self.on)
else:
if self.left_on is None and self.right_on is None:
raise IndexError('pass argument "on" OR "left_on" and '
'"right_on" at class initalization.')
elif self.left_on is None:
raise IndexError('Argument "left_on" is missing '
'at class initalization.')
elif self.right_on is None:
raise IndexError('Argument "right_on" is missing '
'at class initalization.')
else:
left_on = listify(self.left_on)
right_on = listify(self.right_on)
window = self.window
# Check if window is an odd number
if not isinstance(window, int) or (window < 0) or not bool(window % 2):
raise ValueError(
'window is not a positive and odd integer')
# # sorting key is single column
# if isinstance(self.on, (tuple, list, dict)):
# raise ValueError(
# "sorting key is not a label")
# make blocking keys correct
block_left_on = listify(self.block_left_on)
block_right_on = listify(self.block_right_on)
if self.block_on:
block_left_on = listify(self.block_on)
block_right_on = listify(self.block_on)
# drop missing values and columns without relevant information
data_left = df_a[listify(left_on) + block_left_on].dropna(
axis=0, how='any', inplace=False
)
data_left.columns = ['sorting_key'] + \
["blocking_key_%d" % i for i, v in enumerate(block_left_on)]
data_left['index_x'] = data_left.index
data_right = df_b[listify(right_on) + block_right_on].dropna(
axis=0, how='any', inplace=False
)
data_right.columns = ['sorting_key'] + \
["blocking_key_%d" % i for i, v in enumerate(block_right_on)]
data_right['index_y'] = data_right.index
# sorting_key_values is the terminology in Data Matching [Christen,
# 2012]
if self.sorting_key_values is None:
self.sorting_key_values = self._get_sorting_key_values(
data_left['sorting_key'].values,
data_right['sorting_key'].values
)
sorting_key_factors = pandas.Series(
numpy.arange(len(self.sorting_key_values)),
index=self.sorting_key_values)
data_left['sorting_key'] = data_left[
'sorting_key'].map(sorting_key_factors)
data_right['sorting_key'] = data_right[
'sorting_key'].map(sorting_key_factors)
# Internal window size
_window = int((window - 1) / 2)
def merge_lagged(x, y, w):
"""Merge two dataframes with a lag on in the sorting key."""
y = y.copy()
y['sorting_key'] = y['sorting_key'] + w
return x.merge(y, how='inner')
pairs_concat = [merge_lagged(data_left, data_right, w)
for w in range(-_window, _window + 1)]
pairs = pandas.concat(pairs_concat, axis=0).set_index(
['index_x', 'index_y']
).index
return pairs
def __init__(self, left_on, right_on=None, **kwargs):
super().__init__(**kwargs)
self.left_on = listify(left_on)
self.right_on = listify(right_on) if right_on else self.left_on
self.soundex = fuzzy.Soundex(4)
def _compute(self, pairs, x, x_link=None):
# start the timer for the comparing step
start_time = time.time()
sublabels_left = self._get_labels_left(validate=x)
df_a_indexed = frame_indexing(x[sublabels_left], pairs, 0)
if x_link is None:
sublabels_right = self._get_labels_right(validate=x)
df_b_indexed = frame_indexing(x[sublabels_right], pairs, 1)
else:
sublabels_right = self._get_labels_right(validate=x_link)
df_b_indexed = frame_indexing(x_link[sublabels_right], pairs, 1)
# log timing
# index_time = time.time() - start_time
features = []
for feat in self.features:
# --- DATA1
# None: no data passed to func
if feat.labels_left is None:
data1 = tuple()
# empty array: empty df with index passed to func
elif feat.labels_left == []:
data1 = (df_a_indexed[[]], )
# else: subset columns and pass tuple of series
else:
data1 = tuple(
[df_a_indexed[lbl] for lbl in listify(feat.labels_left)])
# --- DATA2
# None: no data passed to func
if feat.labels_right is None:
data2 = tuple()
# empty array: empty df with index passed to func
elif feat.labels_right == []:
data2 = (df_b_indexed[[]], )
# else: subset columns and pass tuple of series
else:
data2 = tuple(
[df_b_indexed[lbl] for lbl in listify(feat.labels_right)])
result = feat._compute(data1, data2)
features.append((result, feat.label))
features = self._union(features, pairs)
# log timing
n = pairs.shape[0]
i_max = '?' if self._i_max is None else self._i_max
eta = time.time() - start_time
self._eta.append(eta)
self._n.append(n)
# log
logging.info("comparing [{:d}/{}] - time: {:.2f}s - pairs: {}".format(
self._i, i_max, eta, n))
# log total
if self._output_log_total:
n_total = np.sum(self._n)
eta_total = np.sum(self._eta)
logging.info(
"comparing [{:d}/{}] - time: {:.2f}s - pairs_total: {}".format(
self._i, i_max, eta_total, n_total))
self._i += 1
return features
def compare(self, comp_func, labels_a, labels_b, *args, **kwargs):
"""Compare two records.
Core method to compare record pairs. This method takes a function and
data from both records in the record pair. The data is compared with
the compare function. The built-in methods also use this function.
Example
-------
>>> comp = recordlinkage.Compare(PAIRS, DATAFRAME1, DATAFRAME2)
>>> comp.exact('first_name', 'name')
>>> # same as
>>> comp.compare(recordlinkage._compare_exact, 'first_name', 'name')
Parameters
----------
comp_func : function
A comparison function. This function can be a built-in function or
a user defined comparison function.
labels_a : label, pandas.Series, pandas.DataFrame
The labels, Series or DataFrame to compare.
labels_b : label, pandas.Series, pandas.DataFrame
The labels, Series or DataFrame to compare.
name : label
The name of the feature and the name of the column.
store : bool, default True
Store the result in the dataframe.
Returns
-------
pandas.Series
A pandas series with the result of comparing each record pair.
"""
if len(self.pairs) == 0:
raise ValueError(
"need at least one record pair"
)
# the name and store arguments
name = kwargs.pop('name', None)
store = kwargs.pop('store', True)
labels_a = listify(labels_a)
labels_b = listify(labels_b)
data_a = []
for label_a in labels_a:
# the label is a numpy or pandas object
if is_numpy_like(label_a) or is_pandas_like(label_a):
data_a.append(label_a)
# check requested labels (for better error messages)
elif label_a not in self.df_a.columns:
raise KeyError("label '%s' is not found in the first"
"dataframe" % label_a)
else:
if self.low_memory:
df_a_label = self._loc2(self.df_a[label_a], self.pairs, 0)
data_a.append(df_a_label)
# not low memory
else:
if self._df_a_indexed is None:
self._df_a_indexed = self._loc2(
self.df_a, self.pairs, 0)
data_a.append(self._df_a_indexed[label_a])
data_a = tuple(data_a)
data_b = []
for label_b in labels_b:
# the label is a numpy or pandas object
if is_numpy_like(label_b) or is_pandas_like(label_b):
data_b.append(label_b)
# check requested labels (for better error messages)
elif label_b not in self.df_b.columns:
raise KeyError("label '%s' is not found in the second"
"dataframe" % label_b)
else:
if self.low_memory:
df_b_label = self._loc2(self.df_b[label_b], self.pairs, 1)
data_b.append(df_b_label)
# not low memory
else:
if self._df_b_indexed is None:
self._df_b_indexed = self._loc2(
self.df_b, self.pairs, 1)
data_b.append(self._df_b_indexed[label_b])
data_b = tuple(data_b)
if self.njobs > 1:
jobs = []
chunk_size = np.ceil(self.njobs / len(self.pairs))
# each job
for i in range(0, self.njobs):
# The data arguments
args_a = tuple(df_a_indexed.loc[i*chunk_size:(i+1)*chunk_size, da] for da in labels_a)
args_b = tuple(df_b_indexed.loc[i*chunk_size:(i+1)*chunk_size, db] for db in labels_b)
p = mp.Process(target=comp_func, args=args_a + args_b + args, kwargs=kwargs)
jobs.append(p)
for proc in jobs:
# Start the process
p.start()
proc.join()
# merge parts
c = pandas.concat(jobs, axis=0, copy=False)
else:
# # The data arguments
# args_a = tuple(df_a_indexed.loc[:, da] for da in labels_a)
# args_b = tuple(df_b_indexed.loc[:, db] for db in labels_b)
# Compute the comparison
c = comp_func(*tuple(data_a + data_b + args), **kwargs)
# if a pandas series is returned, overwrite the index. The
# returned index can be different than the MultiIndex passed to
# the compare function.
if isinstance(c, pandas.Series):
c.index = self.vectors.index
# append column to Compare.vectors
if store:
name_or_id = name if name else len(self.vectors.columns)
self.vectors[name_or_id] = c
return self.vectors[name_or_id].rename(name)
def _link_index(self, df_a, df_b):
left_on, right_on = self._get_left_and_right_on()
left_on = listify(left_on)
right_on = listify(right_on)
window = self.window
# Check if window is an odd number
if not isinstance(window, int) or (window < 0) or not bool(window % 2):
raise ValueError('window is not a positive and odd integer')
# # sorting key is single column
# if isinstance(self.on, (tuple, list, dict)):
# raise ValueError(
# "sorting key is not a label")
# make blocking keys correct
block_left_on = listify(self.block_left_on)
block_right_on = listify(self.block_right_on)
if self.block_on:
block_left_on = listify(self.block_on)
block_right_on = listify(self.block_on)
blocking_keys = ['sorting_key'] + \
["blocking_key_%d" % i for i, v in enumerate(block_left_on)]
# make a dataset for the data on the left
# 1. make a dataframe
# 2. rename columns
# 3. add index col
# 4. drop na (last step to presever index)
data_left = pandas.DataFrame(df_a[listify(left_on) + block_left_on],
copy=False)
data_left.columns = blocking_keys
data_left['index_x'] = numpy.arange(len(df_a))
data_left.dropna(axis=0, how='any', subset=blocking_keys, inplace=True)
data_right = pandas.DataFrame(df_b[listify(right_on) + block_right_on],
copy=False)
data_right.columns = blocking_keys
data_right['index_y'] = numpy.arange(len(df_b))
data_right.dropna(axis=0,
how='any',
subset=blocking_keys,
inplace=True)
# sorting_key_values is the terminology in Data Matching [Christen,
# 2012]
if self.sorting_key_values is None:
self.sorting_key_values = self._get_sorting_key_values(
data_left['sorting_key'].values,
data_right['sorting_key'].values)
sorting_key_factors = pandas.Series(numpy.arange(
len(self.sorting_key_values)),
index=self.sorting_key_values)
data_left['sorting_key'] = data_left['sorting_key'].map(
sorting_key_factors)
data_right['sorting_key'] = data_right['sorting_key'].map(
sorting_key_factors)
# Internal window size
_window = int((window - 1) / 2)
def merge_lagged(x, y, w):
"""Merge two dataframes with a lag on in the sorting key."""
y = y.copy()
y['sorting_key'] = y['sorting_key'] + w
return x.merge(y, how='inner')
pairs_concat = [
merge_lagged(data_left, data_right, w)
for w in range(-_window, _window + 1)
]
pairs_df = pandas.concat(pairs_concat, axis=0)
return pandas.MultiIndex(
levels=[df_a.index.values, df_b.index.values],
labels=[pairs_df['index_x'].values, pairs_df['index_y'].values],
verify_integrity=False)
def compute(self, pairs, x, x_link=None):
"""Compare the records of each record pair.
Calling this method starts the comparing of records.
Parameters
----------
pairs : pandas.MultiIndex
A pandas MultiIndex with the record pairs to compare. The indices
in the MultiIndex are indices of the DataFrame(s) to link.
x : pandas.DataFrame
The DataFrame to link. If `x_link` is given, the comparing is a
linking problem. If `x_link` is not given, the problem is one of
deduplication.
x_link : pandas.DataFrame, optional
The second DataFrame.
Returns
-------
pandas.DataFrame
A pandas DataFrame with feature vectors, i.e. the result of
comparing each record pair.
"""
if not isinstance(pairs, pandas.MultiIndex):
raise ValueError(
"expected pandas.MultiIndex with record pair indices "
"as first argument")
if not isinstance(x, pandas.DataFrame):
raise ValueError("expected pandas.DataFrame as second argument")
if x_link is not None and not isinstance(x_link, pandas.DataFrame):
raise ValueError("expected pandas.DataFrame as third argument")
logging.info("Comparing - start comparing data")
# start the timer for the comparing step
start_time = time.time()
sublabels_left = self._get_labels_left(validate=x)
df_a_indexed = self._loc2(x[sublabels_left], pairs, 0)
if x_link is None:
sublabels_right = self._get_labels_right(validate=x)
df_b_indexed = self._loc2(x[sublabels_right], pairs, 1)
else:
sublabels_right = self._get_labels_right(validate=x_link)
df_b_indexed = self._loc2(x_link[sublabels_right], pairs, 1)
# log timing
index_time = time.time() - start_time
results = pandas.DataFrame(index=pairs)
label_num = 0 # make a label is label is None
for f, lbl1, lbl2, label, args, kwargs in self._compare_functions:
data1 = tuple([df_a_indexed[lbl] for lbl in listify(lbl1)])
data2 = tuple([df_b_indexed[lbl] for lbl in listify(lbl2)])
c = f(*tuple(data1 + data2 + args), **kwargs)
if isinstance(c, (pandas.Series, pandas.DataFrame)):
c = c.values # convert pandas into numpy
if label is not None:
label = listify(label)
n_cols = 1 if len(c.shape) == 1 else c.shape[1]
labels = []
for i in range(0, n_cols):
label_val = label[i] if label is not None else label_num
label_num += 1
labels.append(label_val)
results[label_val] = c
# log timing
total_time = time.time() - start_time
# log timing
logging.info("Comparing - computation time: ~{:.2f}s (from which "
"indexing: ~{:.2f}s)".format(total_time, index_time))
# log results
logf_result = "Comparing - summary shape={}"
logging.info(logf_result.format(results.shape))
return results
def _link_index(self, df_a, df_b):
# Index name conflicts do not occur. They are handled in the
# decorator.
left_on = listify(self.left_on)
right_on = listify(self.right_on)
if self.on:
left_on = listify(self.on)
right_on = listify(self.on)
if not left_on or not right_on:
raise ValueError("no column labels given")
if len(left_on) != len(right_on):
raise ValueError(
"length of left and right keys needs to be the same")
window = self.window
# Check if window is an odd number
if not isinstance(window, int) or (window < 0) or not bool(window % 2):
raise ValueError('window is not a positive and odd integer')
# # sorting key is single column
# if isinstance(self.on, (tuple, list, dict)):
# raise ValueError(
# "sorting key is not a label")
# make blocking keys correct
block_left_on = listify(self.block_left_on)
block_right_on = listify(self.block_right_on)
if self.block_on:
block_left_on = listify(self.block_on)
block_right_on = listify(self.block_on)
# drop missing values and columns without relevant information
data_left = df_a[listify(left_on) + block_left_on].dropna(
axis=0, how='any', inplace=False)
data_left.columns = ['sorting_key'] + \
["blocking_key_%d" % i for i, v in enumerate(block_left_on)]
data_left['index_x'] = data_left.index
data_right = df_b[listify(right_on) + block_right_on].dropna(
axis=0, how='any', inplace=False)
data_right.columns = ['sorting_key'] + \
["blocking_key_%d" % i for i, v in enumerate(block_right_on)]
data_right['index_y'] = data_right.index
# sorting_key_values is the terminology in Data Matching [Christen,
# 2012]
if self.sorting_key_values is None:
self.sorting_key_values = self._get_sorting_key_values(
data_left['sorting_key'].values,
data_right['sorting_key'].values)
sorting_key_factors = pandas.Series(numpy.arange(
len(self.sorting_key_values)),
index=self.sorting_key_values)
data_left['sorting_key'] = data_left['sorting_key'].map(
sorting_key_factors)
data_right['sorting_key'] = data_right['sorting_key'].map(
sorting_key_factors)
# Internal window size
_window = int((window - 1) / 2)
def merge_lagged(x, y, w):
"""Merge two dataframes with a lag on in the sorting key."""
y = y.copy()
y['sorting_key'] = y['sorting_key'] + w
return x.merge(y, how='inner')
pairs_concat = [
merge_lagged(data_left, data_right, w)
for w in range(-_window, _window + 1)
]
pairs = pandas.concat(pairs_concat, axis=0).set_index(
['index_x',
'index_y']).index.rename([df_a.index.name, df_b.index.name])
return pairs
def compare(self, comp_func, labels_a, labels_b, *args, **kwargs):
"""[DEPRECATED] Compare two records.
Core method to compare record pairs. This method takes a function and
data from both records in the record pair. The data is compared with
the compare function. The built-in methods also use this function.
Example
-------
>>> comp = recordlinkage.Compare(PAIRS, DATAFRAME1, DATAFRAME2)
>>> comp.exact('first_name', 'name')
>>> # same as
>>> comp.compare(recordlinkage._compare_exact, 'first_name', 'name')
Parameters
----------
comp_func : function
A comparison function. This function can be a built-in function or
a user defined comparison function.
labels_a : label, pandas.Series, pandas.DataFrame
The labels, Series or DataFrame to compare.
labels_b : label, pandas.Series, pandas.DataFrame
The labels, Series or DataFrame to compare.
name : label
The name of the feature and the name of the column.
store : bool, default True
Store the result in the dataframe.
Returns
-------
pandas.Series
A pandas series with the result of comparing each record pair.
"""
if isinstance(comp_func, pandas.MultiIndex):
raise ValueError("see new api documentation: "
"use method 'compute' instead of 'compare'")
if len(self.pairs) == 0:
raise ValueError("need at least one record pair")
# the name and store arguments
name = kwargs.pop('name', None)
store = kwargs.pop('store', True)
labels_a = listify(labels_a)
labels_b = listify(labels_b)
data_a = []
for label_a in labels_a:
# the label is a numpy or pandas object
if is_numpy_like(label_a) or is_pandas_like(label_a):
data_a.append(label_a)
# check requested labels (for better error messages)
elif label_a not in self.df_a.columns:
raise KeyError("label '{}' is not found in the first"
"dataframe".format(label_a))
else:
if self.low_memory:
df_a_label = self._loc2(self.df_a[label_a], self.pairs, 0)
data_a.append(df_a_label)
# not low memory
else:
if self._df_a_indexed is None:
self._df_a_indexed = self._loc2(
self.df_a, self.pairs, 0)
data_a.append(self._df_a_indexed[label_a])
data_a = tuple(data_a)
data_b = []
for label_b in labels_b:
# the label is a numpy or pandas object
if is_numpy_like(label_b) or is_pandas_like(label_b):
data_b.append(label_b)
# check requested labels (for better error messages)
elif label_b not in self.df_b.columns:
raise KeyError("label '{}' is not found in the second"
"dataframe".format(label_b))
else:
if self.low_memory:
df_b_label = self._loc2(self.df_b[label_b], self.pairs, 1)
data_b.append(df_b_label)
# not low memory
else:
if self._df_b_indexed is None:
self._df_b_indexed = self._loc2(
self.df_b, self.pairs, 1)
data_b.append(self._df_b_indexed[label_b])
data_b = tuple(data_b)
# Compute the comparison
c = comp_func(*tuple(data_a + data_b + args), **kwargs)
# if a pandas series is returned, overwrite the index. The
# returned index can be different than the MultiIndex passed to
# the compare function.
if isinstance(c, pandas.Series):
c.index = self.vectors.index
# append column to Compare.vectors
if store:
name_or_id = name if name else len(self.vectors.columns)
self.vectors[name_or_id] = c
return self.vectors[name_or_id].rename(name)
def _compute(self, pairs, x, x_link=None):
logging.info("Comparing - start comparing data")
# start the timer for the comparing step
start_time = time.time()
sublabels_left = self._get_labels_left(validate=x)
df_a_indexed = frame_indexing(x[sublabels_left], pairs, 0)
if x_link is None:
sublabels_right = self._get_labels_right(validate=x)
df_b_indexed = frame_indexing(x[sublabels_right], pairs, 1)
else:
sublabels_right = self._get_labels_right(validate=x_link)
df_b_indexed = frame_indexing(x_link[sublabels_right], pairs, 1)
# log timing
index_time = time.time() - start_time
results = pandas.DataFrame(index=pairs)
label_num = 0 # make a label is label is None
for feat, label in self.features:
lbl1 = feat.labels_left
lbl2 = feat.labels_right
data1 = tuple([df_a_indexed[lbl] for lbl in listify(lbl1)])
data2 = tuple([df_b_indexed[lbl] for lbl in listify(lbl2)])
c = feat._compute(*tuple(data1 + data2))
if is_pandas_like(c):
c = c.values # convert pandas into numpy
if label is not None:
label = listify(label)
n_cols = 1 if len(c.shape) == 1 else c.shape[1]
labels = []
for i in range(0, n_cols):
label_val = label[i] if label is not None else label_num
label_num += 1
labels.append(label_val)
results[label_val] = c
# log timing
total_time = time.time() - start_time
# log timing
logging.info("Comparing - computation time: ~{:.2f}s (from which "
"indexing: ~{:.2f}s)".format(total_time, index_time))
# log results
logf_result = "Comparing - summary shape={}"
logging.info(logf_result.format(results.shape))
return results