3 examples of 'pandas keep only certain columns' in Python

84
def _pdSearch(df, *keep):
85
    keyList = list(df.keys())
86
    # For each keep:
87
    index, indexCol = [], []
88
    for k in keep:
89
        if type(k) is not list:
90
            k = [k]
91
        # For each & condition:
92
        for iN, iS in enumerate(k):
93
            colName, toKeep = iS[0], iS[1]
94
            # Check if the column exist :
95
            if keyList.count(colName):
96
                if type(toKeep) is not list:
97
                    toKeep = [toKeep]
98
                # Then search in df:
99
                indS = []
100
                for jN, jS in enumerate(toKeep):
101
                    indS.extend(list(n.where(df[colName] == jS)[0]))
102
                if iN == 0:
103
                    indSet = set(indS)
104
                else:
105
                    indSet = indSet.intersection(set(indS))
106
            else:
107
                warn('No column "'+colName +
108
                     '" found. This argument has been ignored from your search'
109
                     )
110
                indSet = []
111
112
        index.append(list(indSet))
113
114
    return index

80
def _clean_columns(df, keep_colnames):
81
    new_colnames = []
82
    for i,colname in enumerate(df.columns):
83
        if colname not in keep_colnames:
84
            new_colnames.append(i)
85
        else:
86
            new_colnames.append(colname)
87
    return new_colnames

332
def dropcols(df, start=None, end=None):
333
    """Drop columns that contain NaN within [start, end] inclusive.
334
335
    A wrapper around DataFrame.dropna() that builds an easier *subset*
336
    syntax for tseries-indexed DataFrames.
337
338
    Parameters
339
    ----------
340
    df : DataFrame
341
    start : str or datetime, default None
342
        start cutoff date, inclusive
343
    end : str or datetime, default None
344
        end cutoff date, inclusive
345
346
    Example
347
    -------
348
    df = DataFrame(np.random.randn(10,3),
349
                   index=pd.date_range('2017', periods=10))
350
351
    # Drop in some NaN
352
    df.set_value('2017-01-04', 0, np.nan)
353
    df.set_value('2017-01-02', 2, np.nan)
354
    df.loc['2017-01-05':, 1] = np.nan
355
356
    # only col2 will be kept--its NaN value falls before `start`
357
    print(dropcols(df, start='2017-01-03'))
358
                      2
359
    2017-01-01  0.12939
360
    2017-01-02      NaN
361
    2017-01-03  0.16596
362
    2017-01-04  1.06442
363
    2017-01-05 -1.87040
364
    2017-01-06 -0.17160
365
    2017-01-07  0.94588
366
    2017-01-08  1.49246
367
    2017-01-09  0.02042
368
    2017-01-10  0.75094
369
370
    """
371
372
    if isinstance(df, Series):
373
        raise ValueError("func only applies to `pd.DataFrame`")
374
    if start is None:
375
        start = df.index[0]
376
    if end is None:
377
        end = df.index[-1]
378
    subset = df.index[(df.index >= start) & (df.index <= end)]
379
    return df.dropna(axis=1, subset=subset)