5 examples of 'pandas merge on multiple columns' in Python

813
def merge(old_cols, new_cols):
814
    return old_cols + new_cols

6
def default_merger(x, y):
7
	import pandas as pd
8
	return pd.concat([x, y])

1066
def merge_columns(self, columns=None, sparse=True, sampling_rate='tr'):
1067
    ''' Merge columns into one DF.
1068
    Args:
1069
        columns (list): Optional list of column names to retain; if None,
1070
            all columns are written out.
1071
        sparse (bool): If True, columns will be kept in a sparse format
1072
            provided they are all internally represented as such. If False,
1073
            a dense matrix (i.e., uniform sampling rate for all events)
1074
            will be exported. Will be ignored if at least one column is
1075
            dense.
1076
        sampling_rate (float): If a dense matrix is written out, the
1077
            sampling rate (in Hz) to use for downsampling. Defaults to the
1078
            value currently set in the instance.
1079
    Returns: A pandas DataFrame.
1080
    '''
1081
1082
    if sparse and self._none_dense():
1083
        return super(BIDSEventVariableCollection,
1084
                     self).merge_columns(columns)
1085
1086
    sampling_rate = self._get_sampling_rate(sampling_rate)
1087
1088
    # Make sure all columns have the same sampling rate
1089
    _cols = self.resample(sampling_rate, force_dense=True,
1090
                          in_place=False).values()
1091
1092
    # Retain only specific columns if desired
1093
    if columns is not None:
1094
        _cols = [c for c in _cols if c.name in columns]
1095
1096
    _cols = [c for c in _cols if c.name not in ["event_file_id", "time"]]
1097
1098
    # Merge all data into one DF
1099
    dfs = [pd.Series(c.values.iloc[:, 0], name=c.name) for c in _cols]
1100
    # Convert datetime to seconds and add duration column
1101
    dense_index = self.dense_index.copy()
1102
    onsets = self.dense_index.pop('time').values.astype(float) / 1e+9
1103
    timing = pd.DataFrame({'onset': onsets})
1104
    timing['duration'] = 1. / sampling_rate
1105
    dfs = [timing] + dfs + [dense_index]
1106
    data = pd.concat(dfs, axis=1)
1107
1108
    return data

51
def random_merge(A, B, N=20, on='AnswerId', key='key', n='n'):
52
    """Pair all rows of A with 1 matching row on "on" and N-1 random rows from B
53
    """
54
    assert key not in A and key not in B
55
    X = A.copy()
56
    X[key] = A[on]
57
    Y = B.copy()
58
    Y[key] = B[on]
59
    match = X.merge(Y, on=key).drop(key, axis=1)
60
    match[n] = 0
61
    df_list = [match]
62
    for i in A.index:
63
        X = A.loc[[i]]
64
        Y = B[B[on] != X[on].iloc[0]].sample(N-1)
65
        X[key] = 1
66
        Y[key] = 1
67
        Z = X.merge(Y, how='outer', on=key).drop(key, axis=1)
68
        Z[n] = range(1, N)
69
        df_list.append(Z)
70
    df = pd.concat(df_list, ignore_index=True)
71
    return df

332
def merge_datasets(self, other):
333
    """
334
    This operation combines two dataframes into one new DataFrame.
335
    If the operation is combining two SpatialDataFrames, the
336
    geometry_type must match.
337
    """
338
    if isinstance(other, SpatialDataFrame) and \
339
        other.geometry_type == self.geometry_type:
340
        return pd.concat(objs=[self, other], axis=0)
341
    elif isinstance(other, DataFrame):
342
        return pd.concat(objs=[self, other], axis=0)
343
    elif isinstance(other, Series):
344
        self['merged_datasets'] = other
345
    elif isinstance(other, SpatialDataFrame) and \
346
         other.geometry_type != self.geometry_type:
347
        raise ValueError("Spatial DataFrames must have the same geometry type.")
348
    else:
349
        raise ValueError("Merge datasets cannot merge types %s" % type(other))