|
| 1 | +""" |
| 2 | +For compatibility with numpy libraries, pandas functions or |
| 3 | +methods have to accept '*args' and '**kwargs' parameters to |
| 4 | +accommodate numpy arguments that are not actually used or |
| 5 | +respected in the pandas implementation. |
| 6 | +
|
| 7 | +To ensure that users do not abuse these parameters, validation |
| 8 | +is performed in 'validators.py' to make sure that any extra |
| 9 | +parameters passed correspond ONLY to those in the numpy signature. |
| 10 | +Part of that validation includes whether or not the user attempted |
| 11 | +to pass in non-default values for these extraneous parameters. As we |
| 12 | +want to discourage users from relying on these parameters when calling |
| 13 | +the pandas implementation, we want them only to pass in the default values |
| 14 | +for these parameters. |
| 15 | +
|
| 16 | +This module provides a set of commonly used default arguments for functions |
| 17 | +and methods that are spread throughout the codebase. This module will make it |
| 18 | +easier to adjust to future upstream changes in the analogous numpy signatures. |
| 19 | +""" |
| 20 | + |
| 21 | +from numpy import ndarray |
| 22 | +from pandas.util.validators import (validate_args, validate_kwargs, |
| 23 | + validate_args_and_kwargs) |
| 24 | +from pandas.core.common import is_integer |
| 25 | +from pandas.compat import OrderedDict |
| 26 | + |
| 27 | + |
| 28 | +class CompatValidator(object): |
| 29 | + def __init__(self, defaults, fname=None, method=None, |
| 30 | + max_fname_arg_count=None): |
| 31 | + self.fname = fname |
| 32 | + self.method = method |
| 33 | + self.defaults = defaults |
| 34 | + self.max_fname_arg_count = max_fname_arg_count |
| 35 | + |
| 36 | + def __call__(self, args, kwargs, fname=None, |
| 37 | + max_fname_arg_count=None, method=None): |
| 38 | + fname = self.fname if fname is None else fname |
| 39 | + max_fname_arg_count = (self.max_fname_arg_count if |
| 40 | + max_fname_arg_count is None |
| 41 | + else max_fname_arg_count) |
| 42 | + method = self.method if method is None else method |
| 43 | + |
| 44 | + if method == 'args': |
| 45 | + validate_args(fname, args, max_fname_arg_count, self.defaults) |
| 46 | + elif method == 'kwargs': |
| 47 | + validate_kwargs(fname, kwargs, self.defaults) |
| 48 | + elif method == 'both': |
| 49 | + validate_args_and_kwargs(fname, args, kwargs, |
| 50 | + max_fname_arg_count, |
| 51 | + self.defaults) |
| 52 | + else: |
| 53 | + raise ValueError("invalid validation method " |
| 54 | + "'{method}'".format(method=method)) |
| 55 | + |
| 56 | +ARGMINMAX_DEFAULTS = dict(out=None) |
| 57 | +validate_argmin = CompatValidator(ARGMINMAX_DEFAULTS, fname='argmin', |
| 58 | + method='both', max_fname_arg_count=1) |
| 59 | +validate_argmax = CompatValidator(ARGMINMAX_DEFAULTS, fname='argmax', |
| 60 | + method='both', max_fname_arg_count=1) |
| 61 | + |
| 62 | + |
| 63 | +def process_skipna(skipna, args): |
| 64 | + if isinstance(skipna, ndarray) or skipna is None: |
| 65 | + args = (skipna,) + args |
| 66 | + skipna = True |
| 67 | + |
| 68 | + return skipna, args |
| 69 | + |
| 70 | + |
| 71 | +def validate_argmin_with_skipna(skipna, args, kwargs): |
| 72 | + """ |
| 73 | + If 'Series.argmin' is called via the 'numpy' library, |
| 74 | + the third parameter in its signature is 'out', which |
| 75 | + takes either an ndarray or 'None', so check if the |
| 76 | + 'skipna' parameter is either an instance of ndarray or |
| 77 | + is None, since 'skipna' itself should be a boolean |
| 78 | + """ |
| 79 | + |
| 80 | + skipna, args = process_skipna(skipna, args) |
| 81 | + validate_argmin(args, kwargs) |
| 82 | + return skipna |
| 83 | + |
| 84 | + |
| 85 | +def validate_argmax_with_skipna(skipna, args, kwargs): |
| 86 | + """ |
| 87 | + If 'Series.argmax' is called via the 'numpy' library, |
| 88 | + the third parameter in its signature is 'out', which |
| 89 | + takes either an ndarray or 'None', so check if the |
| 90 | + 'skipna' parameter is either an instance of ndarray or |
| 91 | + is None, since 'skipna' itself should be a boolean |
| 92 | + """ |
| 93 | + |
| 94 | + skipna, args = process_skipna(skipna, args) |
| 95 | + validate_argmax(args, kwargs) |
| 96 | + return skipna |
| 97 | + |
| 98 | +ARGSORT_DEFAULTS = OrderedDict() |
| 99 | +ARGSORT_DEFAULTS['axis'] = -1 |
| 100 | +ARGSORT_DEFAULTS['kind'] = 'quicksort' |
| 101 | +ARGSORT_DEFAULTS['order'] = None |
| 102 | +validate_argsort = CompatValidator(ARGSORT_DEFAULTS, fname='argsort', |
| 103 | + max_fname_arg_count=0, method='both') |
| 104 | + |
| 105 | + |
| 106 | +def validate_argsort_with_ascending(ascending, args, kwargs): |
| 107 | + """ |
| 108 | + If 'Categorical.argsort' is called via the 'numpy' library, the |
| 109 | + first parameter in its signature is 'axis', which takes either |
| 110 | + an integer or 'None', so check if the 'ascending' parameter has |
| 111 | + either integer type or is None, since 'ascending' itself should |
| 112 | + be a boolean |
| 113 | + """ |
| 114 | + |
| 115 | + if is_integer(ascending) or ascending is None: |
| 116 | + args = (ascending,) + args |
| 117 | + ascending = True |
| 118 | + |
| 119 | + validate_argsort(args, kwargs, max_fname_arg_count=1) |
| 120 | + return ascending |
| 121 | + |
| 122 | +CLIP_DEFAULTS = dict(out=None) |
| 123 | +validate_clip = CompatValidator(CLIP_DEFAULTS, fname='clip', |
| 124 | + method='both', max_fname_arg_count=3) |
| 125 | + |
| 126 | + |
| 127 | +def validate_clip_with_axis(axis, args, kwargs): |
| 128 | + """ |
| 129 | + If 'NDFrame.clip' is called via the numpy library, the third |
| 130 | + parameter in its signature is 'out', which can takes an ndarray, |
| 131 | + so check if the 'axis' parameter is an instance of ndarray, since |
| 132 | + 'axis' itself should either be an integer or None |
| 133 | + """ |
| 134 | + |
| 135 | + if isinstance(axis, ndarray): |
| 136 | + args = (axis,) + args |
| 137 | + axis = None |
| 138 | + |
| 139 | + validate_clip(args, kwargs) |
| 140 | + return axis |
| 141 | + |
| 142 | +COMPRESS_DEFAULTS = OrderedDict() |
| 143 | +COMPRESS_DEFAULTS['axis'] = None |
| 144 | +COMPRESS_DEFAULTS['out'] = None |
| 145 | +validate_compress = CompatValidator(COMPRESS_DEFAULTS, fname='compress', |
| 146 | + method='both', max_fname_arg_count=1) |
| 147 | + |
| 148 | +CUM_FUNC_DEFAULTS = OrderedDict() |
| 149 | +CUM_FUNC_DEFAULTS['dtype'] = None |
| 150 | +CUM_FUNC_DEFAULTS['out'] = None |
| 151 | +validate_cum_func = CompatValidator(CUM_FUNC_DEFAULTS, method='kwargs') |
| 152 | +validate_cumsum = CompatValidator(CUM_FUNC_DEFAULTS, fname='cumsum', |
| 153 | + method='both', max_fname_arg_count=1) |
| 154 | + |
| 155 | +LOGICAL_FUNC_DEFAULTS = dict(out=None) |
| 156 | +validate_logical_func = CompatValidator(LOGICAL_FUNC_DEFAULTS, method='kwargs') |
| 157 | + |
| 158 | +MINMAX_DEFAULTS = dict(out=None) |
| 159 | +validate_min = CompatValidator(MINMAX_DEFAULTS, fname='min', |
| 160 | + method='both', max_fname_arg_count=1) |
| 161 | +validate_max = CompatValidator(MINMAX_DEFAULTS, fname='max', |
| 162 | + method='both', max_fname_arg_count=1) |
| 163 | + |
| 164 | +RESHAPE_DEFAULTS = dict(order='C') |
| 165 | +validate_reshape = CompatValidator(RESHAPE_DEFAULTS, fname='reshape', |
| 166 | + method='both', max_fname_arg_count=1) |
| 167 | + |
| 168 | +REPEAT_DEFAULTS = dict(axis=None) |
| 169 | +validate_repeat = CompatValidator(REPEAT_DEFAULTS, fname='repeat', |
| 170 | + method='both', max_fname_arg_count=1) |
| 171 | + |
| 172 | +ROUND_DEFAULTS = dict(out=None) |
| 173 | +validate_round = CompatValidator(ROUND_DEFAULTS, fname='round', |
| 174 | + method='both', max_fname_arg_count=1) |
| 175 | + |
| 176 | +SORT_DEFAULTS = OrderedDict() |
| 177 | +SORT_DEFAULTS['axis'] = -1 |
| 178 | +SORT_DEFAULTS['kind'] = 'quicksort' |
| 179 | +SORT_DEFAULTS['order'] = None |
| 180 | +validate_sort = CompatValidator(SORT_DEFAULTS, fname='sort', |
| 181 | + method='kwargs') |
| 182 | + |
| 183 | +STAT_FUNC_DEFAULTS = OrderedDict() |
| 184 | +STAT_FUNC_DEFAULTS['dtype'] = None |
| 185 | +STAT_FUNC_DEFAULTS['out'] = None |
| 186 | +validate_stat_func = CompatValidator(STAT_FUNC_DEFAULTS, |
| 187 | + method='kwargs') |
| 188 | +validate_sum = CompatValidator(STAT_FUNC_DEFAULTS, fname='sort', |
| 189 | + method='both', max_fname_arg_count=1) |
| 190 | +validate_mean = CompatValidator(STAT_FUNC_DEFAULTS, fname='mean', |
| 191 | + method='both', max_fname_arg_count=1) |
| 192 | + |
| 193 | +STAT_DDOF_FUNC_DEFAULTS = OrderedDict() |
| 194 | +STAT_DDOF_FUNC_DEFAULTS['dtype'] = None |
| 195 | +STAT_DDOF_FUNC_DEFAULTS['out'] = None |
| 196 | +validate_stat_ddof_func = CompatValidator(STAT_DDOF_FUNC_DEFAULTS, |
| 197 | + method='kwargs') |
| 198 | + |
| 199 | +# Currently, numpy (v1.11) has backwards compatibility checks |
| 200 | +# in place so that this 'kwargs' parameter is technically |
| 201 | +# unnecessary, but in the long-run, this will be needed. |
| 202 | +SQUEEZE_DEFAULTS = dict(axis=None) |
| 203 | +validate_squeeze = CompatValidator(SQUEEZE_DEFAULTS, fname='squeeze', |
| 204 | + method='kwargs') |
| 205 | + |
| 206 | +TAKE_DEFAULTS = OrderedDict() |
| 207 | +TAKE_DEFAULTS['out'] = None |
| 208 | +TAKE_DEFAULTS['mode'] = 'raise' |
| 209 | +validate_take = CompatValidator(TAKE_DEFAULTS, fname='take', |
| 210 | + method='kwargs') |
| 211 | + |
| 212 | + |
| 213 | +def validate_take_with_convert(convert, args, kwargs): |
| 214 | + """ |
| 215 | + If this function is called via the 'numpy' library, the third |
| 216 | + parameter in its signature is 'axis', which takes either an |
| 217 | + ndarray or 'None', so check if the 'convert' parameter is either |
| 218 | + an instance of ndarray or is None |
| 219 | + """ |
| 220 | + |
| 221 | + if isinstance(convert, ndarray) or convert is None: |
| 222 | + args = (convert,) + args |
| 223 | + convert = True |
| 224 | + |
| 225 | + validate_take(args, kwargs, max_fname_arg_count=3, method='both') |
| 226 | + return convert |
| 227 | + |
| 228 | +TRANSPOSE_DEFAULTS = dict(axes=None) |
| 229 | +validate_transpose = CompatValidator(TRANSPOSE_DEFAULTS, fname='transpose', |
| 230 | + method='both', max_fname_arg_count=0) |
| 231 | + |
| 232 | + |
| 233 | +def validate_transpose_for_generic(inst, kwargs): |
| 234 | + try: |
| 235 | + validate_transpose(tuple(), kwargs) |
| 236 | + except ValueError as e: |
| 237 | + klass = type(inst).__name__ |
| 238 | + msg = str(e) |
| 239 | + |
| 240 | + # the Panel class actual relies on the 'axes' parameter if called |
| 241 | + # via the 'numpy' library, so let's make sure the error is specific |
| 242 | + # about saying that the parameter is not supported for particular |
| 243 | + # implementations of 'transpose' |
| 244 | + if "the 'axes' parameter is not supported" in msg: |
| 245 | + msg += " for {klass} instances".format(klass=klass) |
| 246 | + |
| 247 | + raise ValueError(msg) |
0 commit comments