API: disable to_csv and friends on GroupBy objects

[cpcloud]

No description provided.

[jreback]

better than doing a whitelist?

[cpcloud]

my hunch is that there are many more things to whitelist than blacklist

[cpcloud]

i could be wrong....

[jreback]

so many things don't make sense

e.g.

abs,apply,shift,tshift,where,mask,align,groupby,clip....practially every method

allowed of course:

last,first,cum*,mean,sum,min,max,head,tail....

much smaller list

[cpcloud]

@jreback merge?

[jreback]

I would have your test try out some more methods?

[cpcloud]

@jreback Anything else to add here?

[jreback]

can you make this a big more imformative, maybe ' This is a groupby object and cannot perform the operation ....`?, try use apply on tihs object.

[cpcloud]

okay improved the message .. good 2 go?

[jreback]

looks good

[jreback]

closes #4882

[gdraps]

fyi, while testing pandas-master, hit a method not in the new groupby dispatch whitelist: value_counts (on a SeriesGroupBy object). Another possible addition: shift, as used in this SO answer.

Tried to generate a list of blacklisted methods for DataFrame and Series, see below -- needs further filtering, but may reveal useful blocked methods. any thoughts on the remaining methods?

In [18]: sorted([x for x in set(dir(pd.DataFrame)) - pd.core.groupby._apply_whitelist - set(dir(pd.core.groupby.GroupBy)) if not x.startswith(('_', 'T', 'to', 'from', 'as'))])
Out[18]: ['abs', 'add', 'add_prefix', 'add_suffix', 'align', 'all', 'any', 'append', 'applymap', 'at', 'at_time', 'axes', 'between_time', 'bfill', 'blocks', 'bool', 'clip', 'clip_lower', 'clip_upper', 'columns', 'combine', 'combineAdd', 'combineMult', 'combine_first', 'compound', 'consolidate', 'convert_objects', 'copy', 'corr', 'corrwith', 'cov', 'delevel', 'diff', 'div', 'divide', 'dot', 'drop', 'drop_duplicates', 'dropna', 'dtypes', 'duplicated', 'empty', 'eq', 'eval', 'ffill', 'filter', 'first_valid_index', 'floordiv', 'ftypes', 'ge', 'get', 'get_dtype_counts', 'get_ftype_counts', 'get_value', 'get_values', 'groupby', 'gt', 'iat', 'icol', 'idxmax', 'idxmin', 'iget_value', 'iloc', 'index', 'info', 'insert', 'interpolate', 'irow', 'isin', 'isnull', 'iteritems', 'iterkv', 'iterrows', 'itertuples', 'ix', 'join', 'keys', 'kurt', 'kurtosis', 'last_valid_index', 'le', 'load', 'loc', 'lookup', 'lt', 'mad', 'mask', 'merge', 'mod', 'mode', 'mul', 'multiply', 'ndim', 'ne', 'notnull', 'pct_change', 'pivot', 'pivot_table', 'pop', 'pow', 'product', 'query', 'radd', 'rdiv', 'reindex', 'reindex_axis', 'reindex_like', 'rename', 'rename_axis', 'reorder_levels', 'replace', 'reset_index', 'rfloordiv', 'rmod', 'rmul', 'rpow', 'rsub', 'rtruediv', 'save', 'select', 'set_index', 'set_value', 'shape', 'shift', 'skew', 'sort', 'sort_index', 'sortlevel', 'squeeze', 'stack', 'sub', 'subtract', 'swapaxes', 'swaplevel', 'take', 'transpose', 'truediv', 'truncate', 'tshift', 'tz_convert', 'tz_localize', 'unstack', 'update', 'values', 'where', 'xs']

In [19]: sorted([x for x in set(dir(pd.Series)) - pd.core.groupby._apply_whitelist - set(dir(pd.core.groupby.GroupBy)) if not x.startswith(('_', 'T', 'to', 'from', 'as'))]) 
Out[19]: ['abs', 'add', 'add_prefix', 'add_suffix', 'align', 'all', 'any', 'append', 'argmax', 'argmin', 'argsort', 'at', 'at_time', 'autocorr', 'axes', 'base', 'between', 'between_time', 'bfill', 'blocks', 'bool', 'clip', 'clip_lower', 'clip_upper', 'combine', 'combine_first', 'compound', 'consolidate', 'convert_objects', 'copy', 'corr', 'cov', 'data', 'diff', 'div', 'divide', 'dot', 'drop', 'drop_duplicates', 'dropna', 'duplicated', 'empty', 'eq', 'ffill', 'filter', 'first_valid_index', 'flags', 'floordiv', 'ftype', 'ge', 'get', 'get_dtype_counts', 'get_ftype_counts', 'get_value', 'get_values', 'groupby', 'gt', 'iat', 'idxmax', 'idxmin', 'iget', 'iget_value', 'iloc', 'imag', 'index', 'interpolate', 'irow', 'is_time_series', 'isin', 'isnull', 'item', 'iteritems', 'iterkv', 'ix', 'keys', 'kurt', 'kurtosis', 'last_valid_index', 'le', 'load', 'loc', 'lt', 'mad', 'map', 'mask', 'mod', 'mode', 'mul', 'multiply', 'ndim', 'ne', 'nonzero', 'notnull', 'nunique', 'order', 'pct_change', 'pop', 'pow', 'product', 'ptp', 'put', 'radd', 'ravel', 'rdiv', 'real', 'reindex', 'reindex_axis', 'reindex_like', 'rename', 'rename_axis', 'reorder_levels', 'repeat', 'replace', 'reset_index', 'reshape', 'rfloordiv', 'rmod', 'rmul', 'round', 'rpow', 'rsub', 'rtruediv', 'save', 'select', 'set_value', 'shape', 'shift', 'skew', 'sort', 'sort_index', 'sortlevel', 'squeeze', 'str', 'strides', 'sub', 'subtract', 'swapaxes', 'swaplevel', 'take', 'transpose', 'truediv', 'truncate', 'tshift', 'tz_convert', 'tz_localize', 'unique', 'unstack', 'update', 'valid', 'value_counts', 'values', 'view', 'weekday', 'where', 'xs']

[jtratner]

@gdraps can you move this to a separate issue?

[hayd]

Just to throw a spanner in the works here, I wonder if this is going to break existing code which uses this "feature" (i.e. user adds their own method to DataFrame then uses it on a groupby)...