generic.py

# pylint: disable=W0231,E1101
import collections
import warnings
import operator
import weakref
import gc
import json

import numpy as np
import pandas as pd

from pandas._libs import tslib, lib
from pandas.core.dtypes.common import (
    _ensure_int64,
    needs_i8_conversion,
    is_scalar,
    is_integer, is_bool,
    is_bool_dtype,
    is_numeric_dtype,
    is_datetime64_dtype,
    is_timedelta64_dtype,
    is_datetime64tz_dtype,
    is_list_like,
    is_dict_like,
    is_re_compilable,
    pandas_dtype)
from pandas.core.dtypes.cast import maybe_promote, maybe_upcast_putmask
from pandas.core.dtypes.missing import isnull, notnull
from pandas.core.dtypes.generic import ABCSeries, ABCPanel

from pandas.core.common import (_values_from_object,
                                _maybe_box_datetimelike,
                                SettingWithCopyError, SettingWithCopyWarning,
                                AbstractMethodError)

from pandas.core.base import PandasObject, SelectionMixin
from pandas.core.index import (Index, MultiIndex, _ensure_index,
                               InvalidIndexError)
import pandas.core.indexing as indexing
from pandas.core.indexes.datetimes import DatetimeIndex
from pandas.core.indexes.period import PeriodIndex, Period
from pandas.core.internals import BlockManager
import pandas.core.algorithms as algos
import pandas.core.common as com
import pandas.core.missing as missing
from pandas.io.formats.printing import pprint_thing
from pandas.io.formats.format import format_percentiles
from pandas.tseries.frequencies import to_offset
from pandas import compat
from pandas.compat.numpy import function as nv
from pandas.compat import (map, zip, lzip, lrange, string_types,
                           isidentifier, set_function_name, cPickle as pkl)
import pandas.core.nanops as nanops
from pandas.util.decorators import Appender, Substitution, deprecate_kwarg
from pandas.util.validators import validate_bool_kwarg
from pandas.core import config

# goal is to be able to define the docs close to function, while still being
# able to share
_shared_docs = dict()
_shared_doc_kwargs = dict(
    axes='keywords for axes', klass='NDFrame',
    axes_single_arg='int or labels for object',
    args_transpose='axes to permute (int or label for object)',
    optional_by="""
        by : str or list of str
            Name or list of names which refer to the axis items.""")


def _single_replace(self, to_replace, method, inplace, limit):
    if self.ndim != 1:
        raise TypeError('cannot replace {0} with method {1} on a {2}'
                        .format(to_replace, method, type(self).__name__))

    orig_dtype = self.dtype
    result = self if inplace else self.copy()
    fill_f = missing.get_fill_func(method)

    mask = missing.mask_missing(result.values, to_replace)
    values = fill_f(result.values, limit=limit, mask=mask)

    if values.dtype == orig_dtype and inplace:
        return

    result = pd.Series(values, index=self.index,
                       dtype=self.dtype).__finalize__(self)

    if inplace:
        self._update_inplace(result._data)
        return

    return result


class NDFrame(PandasObject, SelectionMixin):
    """
    N-dimensional analogue of DataFrame. Store multi-dimensional in a
    size-mutable, labeled data structure

    Parameters
    ----------
    data : BlockManager
    axes : list
    copy : boolean, default False
    """
    _internal_names = ['_data', '_cacher', '_item_cache', '_cache', 'is_copy',
                       '_subtyp', '_name', '_index', '_default_kind',
                       '_default_fill_value', '_metadata', '__array_struct__',
                       '__array_interface__']
    _internal_names_set = set(_internal_names)
    _accessors = frozenset([])
    _metadata = []
    is_copy = None

    def __init__(self, data, axes=None, copy=False, dtype=None,
                 fastpath=False):

        if not fastpath:
            if dtype is not None:
                data = data.astype(dtype)
            elif copy:
                data = data.copy()

            if axes is not None:
                for i, ax in enumerate(axes):
                    data = data.reindex_axis(ax, axis=i)

        object.__setattr__(self, 'is_copy', None)
        object.__setattr__(self, '_data', data)
        object.__setattr__(self, '_item_cache', {})

    def _repr_data_resource_(self):
        """
        Not a real Jupyter special repr method, but we use the same
        naming convention.
        """
        if config.get_option("display.html.table_schema"):
            data = self.head(config.get_option('display.max_rows'))
            payload = json.loads(data.to_json(orient='table'),
                                 object_pairs_hook=collections.OrderedDict)
            return payload

    def _validate_dtype(self, dtype):
        """ validate the passed dtype """

        if dtype is not None:
            dtype = pandas_dtype(dtype)

            # a compound dtype
            if dtype.kind == 'V':
                raise NotImplementedError("compound dtypes are not implemented"
                                          "in the {0} constructor"
                                          .format(self.__class__.__name__))

        return dtype

    def _init_mgr(self, mgr, axes=None, dtype=None, copy=False):
        """ passed a manager and a axes dict """
        for a, axe in axes.items():
            if axe is not None:
                mgr = mgr.reindex_axis(axe,
                                       axis=self._get_block_manager_axis(a),
                                       copy=False)

        # make a copy if explicitly requested
        if copy:
            mgr = mgr.copy()
        if dtype is not None:
            # avoid further copies if we can
            if len(mgr.blocks) > 1 or mgr.blocks[0].values.dtype != dtype:
                mgr = mgr.astype(dtype=dtype)
        return mgr

    # ----------------------------------------------------------------------
    # Construction

    @property
    def _constructor(self):
        """Used when a manipulation result has the same dimensions as the
        original.
        """
        raise AbstractMethodError(self)

    def __unicode__(self):
        # unicode representation based upon iterating over self
        # (since, by definition, `PandasContainers` are iterable)
        prepr = '[%s]' % ','.join(map(pprint_thing, self))
        return '%s(%s)' % (self.__class__.__name__, prepr)

    def _dir_additions(self):
        """ add the string-like attributes from the info_axis """
        return set([c for c in self._info_axis
                    if isinstance(c, string_types) and isidentifier(c)])

    @property
    def _constructor_sliced(self):
        """Used when a manipulation result has one lower dimension(s) as the
        original, such as DataFrame single columns slicing.
        """
        raise AbstractMethodError(self)

    @property
    def _constructor_expanddim(self):
        """Used when a manipulation result has one higher dimension as the
        original, such as Series.to_frame() and DataFrame.to_panel()
        """
        raise NotImplementedError

    # ----------------------------------------------------------------------
    # Axis

    @classmethod
    def _setup_axes(cls, axes, info_axis=None, stat_axis=None, aliases=None,
                    slicers=None, axes_are_reversed=False, build_axes=True,
                    ns=None):
        """Provide axes setup for the major PandasObjects.

        Parameters
        ----------
        axes : the names of the axes in order (lowest to highest)
        info_axis_num : the axis of the selector dimension (int)
        stat_axis_num : the number of axis for the default stats (int)
        aliases : other names for a single axis (dict)
        slicers : how axes slice to others (dict)
        axes_are_reversed : boolean whether to treat passed axes as
            reversed (DataFrame)
        build_axes : setup the axis properties (default True)
        """

        cls._AXIS_ORDERS = axes
        cls._AXIS_NUMBERS = dict((a, i) for i, a in enumerate(axes))
        cls._AXIS_LEN = len(axes)
        cls._AXIS_ALIASES = aliases or dict()
        cls._AXIS_IALIASES = dict((v, k) for k, v in cls._AXIS_ALIASES.items())
        cls._AXIS_NAMES = dict(enumerate(axes))
        cls._AXIS_SLICEMAP = slicers or None
        cls._AXIS_REVERSED = axes_are_reversed

        # typ
        setattr(cls, '_typ', cls.__name__.lower())

        # indexing support
        cls._ix = None

        if info_axis is not None:
            cls._info_axis_number = info_axis
            cls._info_axis_name = axes[info_axis]

        if stat_axis is not None:
            cls._stat_axis_number = stat_axis
            cls._stat_axis_name = axes[stat_axis]

        # setup the actual axis
        if build_axes:

            def set_axis(a, i):
                setattr(cls, a, lib.AxisProperty(i))
                cls._internal_names_set.add(a)

            if axes_are_reversed:
                m = cls._AXIS_LEN - 1
                for i, a in cls._AXIS_NAMES.items():
                    set_axis(a, m - i)
            else:
                for i, a in cls._AXIS_NAMES.items():
                    set_axis(a, i)

        # addtl parms
        if isinstance(ns, dict):
            for k, v in ns.items():
                setattr(cls, k, v)

    def _construct_axes_dict(self, axes=None, **kwargs):
        """Return an axes dictionary for myself."""
        d = dict([(a, self._get_axis(a)) for a in (axes or self._AXIS_ORDERS)])
        d.update(kwargs)
        return d

    @staticmethod
    def _construct_axes_dict_from(self, axes, **kwargs):
        """Return an axes dictionary for the passed axes."""
        d = dict([(a, ax) for a, ax in zip(self._AXIS_ORDERS, axes)])
        d.update(kwargs)
        return d

    def _construct_axes_dict_for_slice(self, axes=None, **kwargs):
        """Return an axes dictionary for myself."""
        d = dict([(self._AXIS_SLICEMAP[a], self._get_axis(a))
                  for a in (axes or self._AXIS_ORDERS)])
        d.update(kwargs)
        return d

    def _construct_axes_from_arguments(self, args, kwargs, require_all=False):
        """Construct and returns axes if supplied in args/kwargs.

        If require_all, raise if all axis arguments are not supplied
        return a tuple of (axes, kwargs).
        """

        # construct the args
        args = list(args)
        for a in self._AXIS_ORDERS:

            # if we have an alias for this axis
            alias = self._AXIS_IALIASES.get(a)
            if alias is not None:
                if a in kwargs:
                    if alias in kwargs:
                        raise TypeError("arguments are mutually exclusive "
                                        "for [%s,%s]" % (a, alias))
                    continue
                if alias in kwargs:
                    kwargs[a] = kwargs.pop(alias)
                    continue

            # look for a argument by position
            if a not in kwargs:
                try:
                    kwargs[a] = args.pop(0)
                except IndexError:
                    if require_all:
                        raise TypeError("not enough/duplicate arguments "
                                        "specified!")

        axes = dict([(a, kwargs.pop(a, None)) for a in self._AXIS_ORDERS])
        return axes, kwargs

    @classmethod
    def _from_axes(cls, data, axes, **kwargs):
        # for construction from BlockManager
        if isinstance(data, BlockManager):
            return cls(data, **kwargs)
        else:
            if cls._AXIS_REVERSED:
                axes = axes[::-1]
            d = cls._construct_axes_dict_from(cls, axes, copy=False)
            d.update(kwargs)
            return cls(data, **d)

    def _get_axis_number(self, axis):
        axis = self._AXIS_ALIASES.get(axis, axis)
        if is_integer(axis):
            if axis in self._AXIS_NAMES:
                return axis
        else:
            try:
                return self._AXIS_NUMBERS[axis]
            except:
                pass
        raise ValueError('No axis named {0} for object type {1}'
                         .format(axis, type(self)))

    def _get_axis_name(self, axis):
        axis = self._AXIS_ALIASES.get(axis, axis)
        if isinstance(axis, string_types):
            if axis in self._AXIS_NUMBERS:
                return axis
        else:
            try:
                return self._AXIS_NAMES[axis]
            except:
                pass
        raise ValueError('No axis named {0} for object type {1}'
                         .format(axis, type(self)))

    def _get_axis(self, axis):
        name = self._get_axis_name(axis)
        return getattr(self, name)

    def _get_block_manager_axis(self, axis):
        """Map the axis to the block_manager axis."""
        axis = self._get_axis_number(axis)
        if self._AXIS_REVERSED:
            m = self._AXIS_LEN - 1
            return m - axis
        return axis

    def _get_axis_resolvers(self, axis):
        # index or columns
        axis_index = getattr(self, axis)
        d = dict()
        prefix = axis[0]

        for i, name in enumerate(axis_index.names):
            if name is not None:
                key = level = name
            else:
                # prefix with 'i' or 'c' depending on the input axis
                # e.g., you must do ilevel_0 for the 0th level of an unnamed
                # multiiindex
                key = '{prefix}level_{i}'.format(prefix=prefix, i=i)
                level = i

            level_values = axis_index.get_level_values(level)
            s = level_values.to_series()
            s.index = axis_index
            d[key] = s

        # put the index/columns itself in the dict
        if isinstance(axis_index, MultiIndex):
            dindex = axis_index
        else:
            dindex = axis_index.to_series()

        d[axis] = dindex
        return d

    def _get_index_resolvers(self):
        d = {}
        for axis_name in self._AXIS_ORDERS:
            d.update(self._get_axis_resolvers(axis_name))
        return d

    @property
    def _info_axis(self):
        return getattr(self, self._info_axis_name)

    @property
    def _stat_axis(self):
        return getattr(self, self._stat_axis_name)

    @property
    def shape(self):
        """Return a tuple of axis dimensions"""
        return tuple(len(self._get_axis(a)) for a in self._AXIS_ORDERS)

    @property
    def axes(self):
        """Return index label(s) of the internal NDFrame"""
        # we do it this way because if we have reversed axes, then
        # the block manager shows then reversed
        return [self._get_axis(a) for a in self._AXIS_ORDERS]

    @property
    def ndim(self):
        """Number of axes / array dimensions"""
        return self._data.ndim

    @property
    def size(self):
        """number of elements in the NDFrame"""
        return np.prod(self.shape)

    @property
    def _selected_obj(self):
        """ internal compat with SelectionMixin """
        return self

    @property
    def _obj_with_exclusions(self):
        """ internal compat with SelectionMixin """
        return self

    def _expand_axes(self, key):
        new_axes = []
        for k, ax in zip(key, self.axes):
            if k not in ax:
                if type(k) != ax.dtype.type:
                    ax = ax.astype('O')
                new_axes.append(ax.insert(len(ax), k))
            else:
                new_axes.append(ax)

        return new_axes

    def set_axis(self, axis, labels):
        """ public verson of axis assignment """
        setattr(self, self._get_axis_name(axis), labels)

    def _set_axis(self, axis, labels):
        self._data.set_axis(axis, labels)
        self._clear_item_cache()

    _shared_docs['transpose'] = """
        Permute the dimensions of the %(klass)s

        Parameters
        ----------
        args : %(args_transpose)s
        copy : boolean, default False
            Make a copy of the underlying data. Mixed-dtype data will
            always result in a copy

        Examples
        --------
        >>> p.transpose(2, 0, 1)
        >>> p.transpose(2, 0, 1, copy=True)

        Returns
        -------
        y : same as input
        """

    @Appender(_shared_docs['transpose'] % _shared_doc_kwargs)
    def transpose(self, *args, **kwargs):

        # construct the args
        axes, kwargs = self._construct_axes_from_arguments(args, kwargs,
                                                           require_all=True)
        axes_names = tuple([self._get_axis_name(axes[a])
                            for a in self._AXIS_ORDERS])
        axes_numbers = tuple([self._get_axis_number(axes[a])
                              for a in self._AXIS_ORDERS])

        # we must have unique axes
        if len(axes) != len(set(axes)):
            raise ValueError('Must specify %s unique axes' % self._AXIS_LEN)

        new_axes = self._construct_axes_dict_from(self, [self._get_axis(x)
                                                         for x in axes_names])
        new_values = self.values.transpose(axes_numbers)
        if kwargs.pop('copy', None) or (len(args) and args[-1]):
            new_values = new_values.copy()

        nv.validate_transpose_for_generic(self, kwargs)
        return self._constructor(new_values, **new_axes).__finalize__(self)

    def swapaxes(self, axis1, axis2, copy=True):
        """
        Interchange axes and swap values axes appropriately

        Returns
        -------
        y : same as input
        """
        i = self._get_axis_number(axis1)
        j = self._get_axis_number(axis2)

        if i == j:
            if copy:
                return self.copy()
            return self

        mapping = {i: j, j: i}

        new_axes = (self._get_axis(mapping.get(k, k))
                    for k in range(self._AXIS_LEN))
        new_values = self.values.swapaxes(i, j)
        if copy:
            new_values = new_values.copy()

        return self._constructor(new_values, *new_axes).__finalize__(self)

    def pop(self, item):
        """
        Return item and drop from frame. Raise KeyError if not found.
        """
        result = self[item]
        del self[item]
        try:
            result._reset_cacher()
        except AttributeError:
            pass

        return result

    def squeeze(self, axis=None):
        """
        Squeeze length 1 dimensions.

        Parameters
        ----------
        axis : None, integer or string axis name, optional
            The axis to squeeze if 1-sized.

            .. versionadded:: 0.20.0

        Returns
        -------
        scalar if 1-sized, else original object
        """
        axis = (self._AXIS_NAMES if axis is None else
                (self._get_axis_number(axis),))
        try:
            return self.iloc[
                tuple([0 if i in axis and len(a) == 1 else slice(None)
                       for i, a in enumerate(self.axes)])]
        except:
            return self

    def swaplevel(self, i=-2, j=-1, axis=0):
        """
        Swap levels i and j in a MultiIndex on a particular axis

        Parameters
        ----------
        i, j : int, string (can be mixed)
            Level of index to be swapped. Can pass level name as string.

        Returns
        -------
        swapped : type of caller (new object)

        .. versionchanged:: 0.18.1

           The indexes ``i`` and ``j`` are now optional, and default to
           the two innermost levels of the index.

        """
        axis = self._get_axis_number(axis)
        result = self.copy()
        labels = result._data.axes[axis]
        result._data.set_axis(axis, labels.swaplevel(i, j))
        return result

    # ----------------------------------------------------------------------
    # Rename

    # TODO: define separate funcs for DataFrame, Series and Panel so you can
    # get completion on keyword arguments.
    _shared_docs['rename'] = """
        Alter axes input function or functions. Function / dict values must be
        unique (1-to-1). Labels not contained in a dict / Series will be left
        as-is. Extra labels listed don't throw an error. Alternatively, change
        ``Series.name`` with a scalar value (Series only).

        Parameters
        ----------
        %(axes)s : scalar, list-like, dict-like or function, optional
            Scalar or list-like will alter the ``Series.name`` attribute,
            and raise on DataFrame or Panel.
            dict-like or functions are transformations to apply to
            that axis' values
        copy : boolean, default True
            Also copy underlying data
        inplace : boolean, default False
            Whether to return a new %(klass)s. If True then value of copy is
            ignored.
        level : int or level name, default None
            In case of a MultiIndex, only rename labels in the specified
            level.

        Returns
        -------
        renamed : %(klass)s (new object)

        See Also
        --------
        pandas.NDFrame.rename_axis

        Examples
        --------
        >>> s = pd.Series([1, 2, 3])
        >>> s
        0    1
        1    2
        2    3
        dtype: int64
        >>> s.rename("my_name") # scalar, changes Series.name
        0    1
        1    2
        2    3
        Name: my_name, dtype: int64
        >>> s.rename(lambda x: x ** 2)  # function, changes labels
        0    1
        1    2
        4    3
        dtype: int64
        >>> s.rename({1: 3, 2: 5})  # mapping, changes labels
        0    1
        3    2
        5    3
        dtype: int64
        >>> df = pd.DataFrame({"A": [1, 2, 3], "B": [4, 5, 6]})
        >>> df.rename(2)
        Traceback (most recent call last):
        ...
        TypeError: 'int' object is not callable
        >>> df.rename(index=str, columns={"A": "a", "B": "c"})
           a  c
        0  1  4
        1  2  5
        2  3  6
        >>> df.rename(index=str, columns={"A": "a", "C": "c"})
           a  B
        0  1  4
        1  2  5
        2  3  6
        """

    @Appender(_shared_docs['rename'] % dict(axes='axes keywords for this'
                                            ' object', klass='NDFrame'))
    def rename(self, *args, **kwargs):

        axes, kwargs = self._construct_axes_from_arguments(args, kwargs)
        copy = kwargs.pop('copy', True)
        inplace = kwargs.pop('inplace', False)
        level = kwargs.pop('level', None)

        if kwargs:
            raise TypeError('rename() got an unexpected keyword '
                            'argument "{0}"'.format(list(kwargs.keys())[0]))

        if com._count_not_none(*axes.values()) == 0:
            raise TypeError('must pass an index to rename')

        # renamer function if passed a dict
        def _get_rename_function(mapper):
            if isinstance(mapper, (dict, ABCSeries)):

                def f(x):
                    if x in mapper:
                        return mapper[x]
                    else:
                        return x
            else:
                f = mapper

            return f

        self._consolidate_inplace()
        result = self if inplace else self.copy(deep=copy)

        # start in the axis order to eliminate too many copies
        for axis in lrange(self._AXIS_LEN):
            v = axes.get(self._AXIS_NAMES[axis])
            if v is None:
                continue
            f = _get_rename_function(v)

            baxis = self._get_block_manager_axis(axis)
            if level is not None:
                level = self.axes[axis]._get_level_number(level)
            result._data = result._data.rename_axis(f, axis=baxis, copy=copy,
                                                    level=level)
            result._clear_item_cache()

        if inplace:
            self._update_inplace(result._data)
        else:
            return result.__finalize__(self)

    rename.__doc__ = _shared_docs['rename']

    def rename_axis(self, mapper, axis=0, copy=True, inplace=False):
        """
        Alter index and / or columns using input function or functions.
        A scalar or list-like for ``mapper`` will alter the ``Index.name``
        or ``MultiIndex.names`` attribute.
        A function or dict for ``mapper`` will alter the labels.
        Function / dict values must be unique (1-to-1). Labels not contained in
        a dict / Series will be left as-is.

        Parameters
        ----------
        mapper : scalar, list-like, dict-like or function, optional
        axis : int or string, default 0
        copy : boolean, default True
            Also copy underlying data
        inplace : boolean, default False

        Returns
        -------
        renamed : type of caller

        See Also
        --------
        pandas.NDFrame.rename
        pandas.Index.rename

        Examples
        --------
        >>> df = pd.DataFrame({"A": [1, 2, 3], "B": [4, 5, 6]})
        >>> df.rename_axis("foo")  # scalar, alters df.index.name
             A  B
        foo
        0    1  4
        1    2  5
        2    3  6
        >>> df.rename_axis(lambda x: 2 * x)  # function: alters labels
           A  B
        0  1  4
        2  2  5
        4  3  6
        >>> df.rename_axis({"A": "ehh", "C": "see"}, axis="columns")  # mapping
           ehh  B
        0    1  4
        1    2  5
        2    3  6
        """
        inplace = validate_bool_kwarg(inplace, 'inplace')
        non_mapper = is_scalar(mapper) or (is_list_like(mapper) and not
                                           is_dict_like(mapper))
        if non_mapper:
            return self._set_axis_name(mapper, axis=axis)
        else:
            axis = self._get_axis_name(axis)
            d = {'copy': copy, 'inplace': inplace}
            d[axis] = mapper
            return self.rename(**d)

    def _set_axis_name(self, name, axis=0):
        """
        Alter the name or names of the axis, returning self.

        Parameters
        ----------
        name : str or list of str
            Name for the Index, or list of names for the MultiIndex
        axis : int or str
           0 or 'index' for the index; 1 or 'columns' for the columns

        Returns
        -------
        renamed : type of caller

        See Also
        --------
        pandas.DataFrame.rename
        pandas.Series.rename
        pandas.Index.rename

        Examples
        --------
        >>> df._set_axis_name("foo")
             A
        foo
        0    1
        1    2
        2    3
        >>> df.index = pd.MultiIndex.from_product([['A'], ['a', 'b', 'c']])
        >>> df._set_axis_name(["bar", "baz"])
                 A
        bar baz
        A   a    1
            b    2
            c    3
        """
        axis = self._get_axis_number(axis)
        idx = self._get_axis(axis).set_names(name)

        renamed = self.copy(deep=True)
        renamed.set_axis(axis, idx)
        return renamed

    # ----------------------------------------------------------------------
    # Comparisons

    def _indexed_same(self, other):
        return all([self._get_axis(a).equals(other._get_axis(a))
                    for a in self._AXIS_ORDERS])

    def __neg__(self):
        values = _values_from_object(self)
        if values.dtype == np.bool_:
            arr = operator.inv(values)
        else:
            arr = operator.neg(values)
        return self.__array_wrap__(arr)

    def __invert__(self):
        try:
            arr = operator.inv(_values_from_object(self))
            return self.__array_wrap__(arr)
        except:

            # inv fails with 0 len
            if not np.prod(self.shape):
                return self

            raise

    def equals(self, other):
        """
        Determines if two NDFrame objects contain the same elements. NaNs in
        the same location are considered equal.
        """
        if not isinstance(other, self._constructor):
            return False
        return self._data.equals(other._data)

    # ----------------------------------------------------------------------
    # Iteration

    def __hash__(self):
        raise TypeError('{0!r} objects are mutable, thus they cannot be'
                        ' hashed'.format(self.__class__.__name__))

    def __iter__(self):
        """Iterate over infor axis"""
        return iter(self._info_axis)

    # can we get a better explanation of this?
    def keys(self):
        """Get the 'info axis' (see Indexing for more)

        This is index for Series, columns for DataFrame and major_axis for
        Panel.
        """
        return self._info_axis

    def iteritems(self):
        """Iterate over (label, values) on info axis

        This is index for Series, columns for DataFrame, major_axis for Panel,
        and so on.
        """
        for h in self._info_axis:
            yield h, self[h]

    def __len__(self):
        """Returns length of info axis"""
        return len(self._info_axis)

    def __contains__(self, key):
        """True if the key is in the info axis"""
        return key in self._info_axis

    @property
    def empty(self):
        """True if NDFrame is entirely empty [no items], meaning any of the
        axes are of length 0.

        Notes
        -----
        If NDFrame contains only NaNs, it is still not considered empty. See
        the example below.

        Examples
        --------
        An example of an actual empty DataFrame. Notice the index is empty:

        >>> df_empty = pd.DataFrame({'A' : []})
        >>> df_empty
        Empty DataFrame
        Columns: [A]
        Index: []
        >>> df_empty.empty
        True

        If we only have NaNs in our DataFrame, it is not considered empty! We
        will need to drop the NaNs to make the DataFrame empty:

        >>> df = pd.DataFrame({'A' : [np.nan]})
        >>> df
            A
        0 NaN
        >>> df.empty
        False
        >>> df.dropna().empty
        True

        See also
        --------
        pandas.Series.dropna
        pandas.DataFrame.dropna
        """
        return any(len(self._get_axis(a)) == 0 for a in self._AXIS_ORDERS)

    def __nonzero__(self):
        raise ValueError("The truth value of a {0} is ambiguous. "
                         "Use a.empty, a.bool(), a.item(), a.any() or a.all()."
                         .format(self.__class__.__name__))

    __bool__ = __nonzero__

    def bool(self):
        """Return the bool of a single element PandasObject.

        This must be a boolean scalar value, either True or False.  Raise a
        ValueError if the PandasObject does not have exactly 1 element, or that
        element is not boolean
        """
        v = self.squeeze()
        if isinstance(v, (bool, np.bool_)):
            return bool(v)
        elif is_scalar(v):
            raise ValueError("bool cannot act on a non-boolean single element "
                             "{0}".format(self.__class__.__name__))

        self.__nonzero__()

    def __abs__(self):
        return self.abs()

    def __round__(self, decimals=0):
        return self.round(decimals)

    # ----------------------------------------------------------------------
    # Array Interface

    def __array__(self, dtype=None):
        return _values_from_object(self)

    def __array_wrap__(self, result, context=None):
        d = self._construct_axes_dict(self._AXIS_ORDERS, copy=False)
        return self._constructor(result, **d).__finalize__(self)

    # ideally we would define this to avoid the getattr checks, but
    # is slower
    # @property
    # def __array_interface__(self):
    #    """ provide numpy array interface method """
    #    values = self.values
    #    return dict(typestr=values.dtype.str,shape=values.shape,data=values)

    def to_dense(self):
        """Return dense representation of NDFrame (as opposed to sparse)"""
        # compat
        return self

    # ----------------------------------------------------------------------
    # Picklability

    def __getstate__(self):
        meta = dict((k, getattr(self, k, None)) for k in self._metadata)
        return dict(_data=self._data, _typ=self._typ, _metadata=self._metadata,
                    **meta)

    def __setstate__(self, state):

        if isinstance(state, BlockManager):
            self._data = state
        elif isinstance(state, dict):
            typ = state.get('_typ')
            if typ is not None:

                # set in the order of internal names
                # to avoid definitional recursion
                # e.g. say fill_value needing _data to be
                # defined
                meta = set(self._internal_names + self._metadata)
                for k in list(meta):
                    if k in state:
                        v = state[k]
                        object.__setattr__(self, k, v)

                for k, v in state.items():
                    if k not in meta:
                        object.__setattr__(self, k, v)

            else:
                self._unpickle_series_compat(state)
        elif isinstance(state[0], dict):
            if len(state) == 5:
                self._unpickle_sparse_frame_compat(state)
            else:
                self._unpickle_frame_compat(state)
        elif len(state) == 4:
            self._unpickle_panel_compat(state)
        elif len(state) == 2:
            self._unpickle_series_compat(state)
        else:  # pragma: no cover
            # old pickling format, for compatibility
            self._unpickle_matrix_compat(state)

        self._item_cache = {}

    # ----------------------------------------------------------------------
    # IO

    # ----------------------------------------------------------------------
    # I/O Methods

    _shared_docs['to_excel'] = """
    Write %(klass)s to an excel sheet
    %(versionadded_to_excel)s

    Parameters
    ----------
    excel_writer : string or ExcelWriter object
        File path or existing ExcelWriter
    sheet_name : string, default 'Sheet1'
        Name of sheet which will contain DataFrame
    na_rep : string, default ''
        Missing data representation
    float_format : string, default None
        Format string for floating point numbers
    columns : sequence, optional
        Columns to write
    header : boolean or list of string, default True
        Write out column names. If a list of string is given it is
        assumed to be aliases for the column names
    index : boolean, default True
        Write row names (index)
    index_label : string or sequence, default None
        Column label for index column(s) if desired. If None is given, and
        `header` and `index` are True, then the index names are used. A
        sequence should be given if the DataFrame uses MultiIndex.
    startrow :
        upper left cell row to dump data frame
    startcol :
        upper left cell column to dump data frame
    engine : string, default None
        write engine to use - you can also set this via the options
        ``io.excel.xlsx.writer``, ``io.excel.xls.writer``, and
        ``io.excel.xlsm.writer``.
    merge_cells : boolean, default True
        Write MultiIndex and Hierarchical Rows as merged cells.
    encoding: string, default None
        encoding of the resulting excel file. Only necessary for xlwt,
        other writers support unicode natively.
    inf_rep : string, default 'inf'
        Representation for infinity (there is no native representation for
        infinity in Excel)
    freeze_panes : tuple of integer (length 2), default None
        Specifies the one-based bottommost row and rightmost column that
        is to be frozen

        .. versionadded:: 0.20.0

    Notes
    -----
    If passing an existing ExcelWriter object, then the sheet will be added
    to the existing workbook.  This can be used to save different
    DataFrames to one workbook:

    >>> writer = pd.ExcelWriter('output.xlsx')
    >>> df1.to_excel(writer,'Sheet1')
    >>> df2.to_excel(writer,'Sheet2')
    >>> writer.save()

    For compatibility with to_csv, to_excel serializes lists and dicts to
    strings before writing.
    """

    def to_json(self, path_or_buf=None, orient=None, date_format=None,
                double_precision=10, force_ascii=True, date_unit='ms',
                default_handler=None, lines=False):
        """
        Convert the object to a JSON string.

        Note NaN's and None will be converted to null and datetime objects
        will be converted to UNIX timestamps.

        Parameters
        ----------
        path_or_buf : the path or buffer to write the result string
            if this is None, return a StringIO of the converted string
        orient : string

            * Series

              - default is 'index'
              - allowed values are: {'split','records','index'}

            * DataFrame

              - default is 'columns'
              - allowed values are:
                {'split','records','index','columns','values'}

            * The format of the JSON string

              - split : dict like
                {index -> [index], columns -> [columns], data -> [values]}
              - records : list like
                [{column -> value}, ... , {column -> value}]
              - index : dict like {index -> {column -> value}}
              - columns : dict like {column -> {index -> value}}
              - values : just the values array
              - table : dict like {'schema': {schema}, 'data': {data}}
                describing the data, and the data component is
                like ``orient='records'``.

                .. versionchanged:: 0.20.0

        date_format : {None, 'epoch', 'iso'}
            Type of date conversion. `epoch` = epoch milliseconds,
            `iso` = ISO8601. The default depends on the `orient`. For
            `orient='table'`, the default is `'iso'`. For all other orients,
            the default is `'epoch'`.
        double_precision : The number of decimal places to use when encoding
            floating point values, default 10.
        force_ascii : force encoded string to be ASCII, default True.
        date_unit : string, default 'ms' (milliseconds)
            The time unit to encode to, governs timestamp and ISO8601
            precision.  One of 's', 'ms', 'us', 'ns' for second, millisecond,
            microsecond, and nanosecond respectively.
        default_handler : callable, default None
            Handler to call if object cannot otherwise be converted to a
            suitable format for JSON. Should receive a single argument which is
            the object to convert and return a serialisable object.
        lines : boolean, default False
            If 'orient' is 'records' write out line delimited json format. Will
            throw ValueError if incorrect 'orient' since others are not list
            like.

            .. versionadded:: 0.19.0

        Returns
        -------
        same type as input object with filtered info axis

        See Also
        --------
        pd.read_json

        Examples
        --------

        >>> df = pd.DataFrame([['a', 'b'], ['c', 'd']],
        ...                   index=['row 1', 'row 2'],
        ...                   columns=['col 1', 'col 2'])
        >>> df.to_json(orient='split')
        '{"columns":["col 1","col 2"],
          "index":["row 1","row 2"],
          "data":[["a","b"],["c","d"]]}'

        Encoding/decoding a Dataframe using ``'index'`` formatted JSON:

        >>> df.to_json(orient='index')
        '{"row 1":{"col 1":"a","col 2":"b"},"row 2":{"col 1":"c","col 2":"d"}}'

        Encoding/decoding a Dataframe using ``'records'`` formatted JSON.
        Note that index labels are not preserved with this encoding.

        >>> df.to_json(orient='records')
        '[{"col 1":"a","col 2":"b"},{"col 1":"c","col 2":"d"}]'

        Encoding with Table Schema

        >>> df.to_json(orient='table')
        '{"schema": {"fields": [{"name": "index", "type": "string"},
                                {"name": "col 1", "type": "string"},
                                {"name": "col 2", "type": "string"}],
                     "primaryKey": "index",
                     "pandas_version": "0.20.0"},
          "data": [{"index": "row 1", "col 1": "a", "col 2": "b"},
                   {"index": "row 2", "col 1": "c", "col 2": "d"}]}'
        """

        from pandas.io import json
        if date_format is None and orient == 'table':
            date_format = 'iso'
        elif date_format is None:
            date_format = 'epoch'
        return json.to_json(path_or_buf=path_or_buf, obj=self, orient=orient,
                            date_format=date_format,
                            double_precision=double_precision,
                            force_ascii=force_ascii, date_unit=date_unit,
                            default_handler=default_handler,
                            lines=lines)

    def to_hdf(self, path_or_buf, key, **kwargs):
        """Write the contained data to an HDF5 file using HDFStore.

        Parameters
        ----------
        path_or_buf : the path (string) or HDFStore object
        key : string
            identifier for the group in the store
        mode : optional, {'a', 'w', 'r+'}, default 'a'

          ``'w'``
              Write; a new file is created (an existing file with the same
              name would be deleted).
          ``'a'``
              Append; an existing file is opened for reading and writing,
              and if the file does not exist it is created.
          ``'r+'``
              It is similar to ``'a'``, but the file must already exist.
        format : 'fixed(f)|table(t)', default is 'fixed'
            fixed(f) : Fixed format
                       Fast writing/reading. Not-appendable, nor searchable
            table(t) : Table format
                       Write as a PyTables Table structure which may perform
                       worse but allow more flexible operations like searching
                       / selecting subsets of the data
        append : boolean, default False
            For Table formats, append the input data to the existing
        data_columns :  list of columns, or True, default None
            List of columns to create as indexed data columns for on-disk
            queries, or True to use all columns. By default only the axes
            of the object are indexed. See `here
            <http://pandas.pydata.org/pandas-docs/stable/io.html#query-via-data-columns>`__.

            Applicable only to format='table'.
        complevel : int, 1-9, default 0
            If a complib is specified compression will be applied
            where possible
        complib : {'zlib', 'bzip2', 'lzo', 'blosc', None}, default None
            If complevel is > 0 apply compression to objects written
            in the store wherever possible
        fletcher32 : bool, default False
            If applying compression use the fletcher32 checksum
        dropna : boolean, default False.
            If true, ALL nan rows will not be written to store.
        """

        from pandas.io import pytables
        return pytables.to_hdf(path_or_buf, key, self, **kwargs)

    def to_msgpack(self, path_or_buf=None, encoding='utf-8', **kwargs):
        """
        msgpack (serialize) object to input file path

        THIS IS AN EXPERIMENTAL LIBRARY and the storage format
        may not be stable until a future release.

        Parameters
        ----------
        path : string File path, buffer-like, or None
            if None, return generated string
        append : boolean whether to append to an existing msgpack
            (default is False)
        compress : type of compressor (zlib or blosc), default to None (no
            compression)
        """

        from pandas.io import packers
        return packers.to_msgpack(path_or_buf, self, encoding=encoding,
                                  **kwargs)

    def to_sql(self, name, con, flavor=None, schema=None, if_exists='fail',
               index=True, index_label=None, chunksize=None, dtype=None):
        """
        Write records stored in a DataFrame to a SQL database.

        Parameters
        ----------
        name : string
            Name of SQL table
        con : SQLAlchemy engine or DBAPI2 connection (legacy mode)
            Using SQLAlchemy makes it possible to use any DB supported by that
            library. If a DBAPI2 object, only sqlite3 is supported.
        flavor : 'sqlite', default None
            DEPRECATED: this parameter will be removed in a future version,
            as 'sqlite' is the only supported option if SQLAlchemy is not
            installed.
        schema : string, default None
            Specify the schema (if database flavor supports this). If None, use
            default schema.
        if_exists : {'fail', 'replace', 'append'}, default 'fail'
            - fail: If table exists, do nothing.
            - replace: If table exists, drop it, recreate it, and insert data.
            - append: If table exists, insert data. Create if does not exist.
        index : boolean, default True
            Write DataFrame index as a column.
        index_label : string or sequence, default None
            Column label for index column(s). If None is given (default) and
            `index` is True, then the index names are used.
            A sequence should be given if the DataFrame uses MultiIndex.
        chunksize : int, default None
            If not None, then rows will be written in batches of this size at a
            time.  If None, all rows will be written at once.
        dtype : dict of column name to SQL type, default None
            Optional specifying the datatype for columns. The SQL type should
            be a SQLAlchemy type, or a string for sqlite3 fallback connection.

        """
        from pandas.io import sql
        sql.to_sql(self, name, con, flavor=flavor, schema=schema,
                   if_exists=if_exists, index=index, index_label=index_label,
                   chunksize=chunksize, dtype=dtype)

    def to_pickle(self, path, compression='infer',
                  protocol=pkl.HIGHEST_PROTOCOL):
        """
        Pickle (serialize) object to input file path.

        Parameters
        ----------
        path : string
            File path
        compression : {'infer', 'gzip', 'bz2', 'xz', None}, default 'infer'
            a string representing the compression to use in the output file

            .. versionadded:: 0.20.0
        protocol : int
            Int which indicates which protocol should be used by the pickler,
            default HIGHEST_PROTOCOL (see [1], paragraph 12.1.2). The possible
            values for this parameter depend on the version of Python. For
            Python 2.x, possible values are 0, 1, 2. For Python>=3.0, 3 is a
            valid value. For Python >= 3.4, 4 is a valid value.A negative value
            for the protocol parameter is equivalent to setting its value to
            HIGHEST_PROTOCOL.

            .. [1] https://docs.python.org/3/library/pickle.html
            .. versionadded:: 0.21.0

        """
        from pandas.io.pickle import to_pickle
        return to_pickle(self, path, compression=compression,
                         protocol=protocol)

    def to_clipboard(self, excel=None, sep=None, **kwargs):
        """
        Attempt to write text representation of object to the system clipboard
        This can be pasted into Excel, for example.

        Parameters
        ----------
        excel : boolean, defaults to True
                if True, use the provided separator, writing in a csv
                format for allowing easy pasting into excel.
                if False, write a string representation of the object
                to the clipboard
        sep : optional, defaults to tab
        other keywords are passed to to_csv

        Notes
        -----
        Requirements for your platform
          - Linux: xclip, or xsel (with gtk or PyQt4 modules)
          - Windows: none
          - OS X: none
        """
        from pandas.io import clipboard
        clipboard.to_clipboard(self, excel=excel, sep=sep, **kwargs)

    def to_xarray(self):
        """
        Return an xarray object from the pandas object.

        Returns
        -------
        a DataArray for a Series
        a Dataset for a DataFrame
        a DataArray for higher dims

        Examples
        --------
        >>> df = pd.DataFrame({'A' : [1, 1, 2],
                               'B' : ['foo', 'bar', 'foo'],
                               'C' : np.arange(4.,7)})
        >>> df
           A    B    C
        0  1  foo  4.0
        1  1  bar  5.0
        2  2  foo  6.0

        >>> df.to_xarray()
        <xarray.Dataset>
        Dimensions:  (index: 3)
        Coordinates:
          * index    (index) int64 0 1 2
        Data variables:
            A        (index) int64 1 1 2
            B        (index) object 'foo' 'bar' 'foo'
            C        (index) float64 4.0 5.0 6.0

        >>> df = pd.DataFrame({'A' : [1, 1, 2],
                               'B' : ['foo', 'bar', 'foo'],
                               'C' : np.arange(4.,7)}
                             ).set_index(['B','A'])
        >>> df
                 C
        B   A
        foo 1  4.0
        bar 1  5.0
        foo 2  6.0

        >>> df.to_xarray()
        <xarray.Dataset>
        Dimensions:  (A: 2, B: 2)
        Coordinates:
          * B        (B) object 'bar' 'foo'
          * A        (A) int64 1 2
        Data variables:
            C        (B, A) float64 5.0 nan 4.0 6.0

        >>> p = pd.Panel(np.arange(24).reshape(4,3,2),
                         items=list('ABCD'),
                         major_axis=pd.date_range('20130101', periods=3),
                         minor_axis=['first', 'second'])
        >>> p
        <class 'pandas.core.panel.Panel'>
        Dimensions: 4 (items) x 3 (major_axis) x 2 (minor_axis)
        Items axis: A to D
        Major_axis axis: 2013-01-01 00:00:00 to 2013-01-03 00:00:00
        Minor_axis axis: first to second

        >>> p.to_xarray()
        <xarray.DataArray (items: 4, major_axis: 3, minor_axis: 2)>
        array([[[ 0,  1],
                [ 2,  3],
                [ 4,  5]],
               [[ 6,  7],
                [ 8,  9],
                [10, 11]],
               [[12, 13],
                [14, 15],
                [16, 17]],
               [[18, 19],
                [20, 21],
                [22, 23]]])
        Coordinates:
          * items       (items) object 'A' 'B' 'C' 'D'
          * major_axis  (major_axis) datetime64[ns] 2013-01-01 2013-01-02 2013-01-03  # noqa
          * minor_axis  (minor_axis) object 'first' 'second'

        Notes
        -----
        See the `xarray docs <http://xarray.pydata.org/en/stable/>`__
        """
        import xarray
        if self.ndim == 1:
            return xarray.DataArray.from_series(self)
        elif self.ndim == 2:
            return xarray.Dataset.from_dataframe(self)

        # > 2 dims
        coords = [(a, self._get_axis(a)) for a in self._AXIS_ORDERS]
        return xarray.DataArray(self,
                                coords=coords,
                                )

    # ----------------------------------------------------------------------
    # Fancy Indexing

    @classmethod
    def _create_indexer(cls, name, indexer):
        """Create an indexer like _name in the class."""

        if getattr(cls, name, None) is None:
            iname = '_%s' % name
            setattr(cls, iname, None)

            def _indexer(self):
                i = getattr(self, iname)
                if i is None:
                    i = indexer(self, name)
                    setattr(self, iname, i)
                return i

            setattr(cls, name, property(_indexer, doc=indexer.__doc__))

            # add to our internal names set
            cls._internal_names_set.add(iname)

    def get(self, key, default=None):
        """
        Get item from object for given key (DataFrame column, Panel slice,
        etc.). Returns default value if not found.

        Parameters
        ----------
        key : object

        Returns
        -------
        value : type of items contained in object
        """
        try:
            return self[key]
        except (KeyError, ValueError, IndexError):
            return default

    def __getitem__(self, item):
        return self._get_item_cache(item)

    def _get_item_cache(self, item):
        """Return the cached item, item represents a label indexer."""
        cache = self._item_cache
        res = cache.get(item)
        if res is None:
            values = self._data.get(item)
            res = self._box_item_values(item, values)
            cache[item] = res
            res._set_as_cached(item, self)

            # for a chain
            res.is_copy = self.is_copy
        return res

    def _set_as_cached(self, item, cacher):
        """Set the _cacher attribute on the calling object with a weakref to
        cacher.
        """
        self._cacher = (item, weakref.ref(cacher))

    def _reset_cacher(self):
        """Reset the cacher."""
        if hasattr(self, '_cacher'):
            del self._cacher

    def _iget_item_cache(self, item):
        """Return the cached item, item represents a positional indexer."""
        ax = self._info_axis
        if ax.is_unique:
            lower = self._get_item_cache(ax[item])
        else:
            lower = self.take(item, axis=self._info_axis_number, convert=True)
        return lower

    def _box_item_values(self, key, values):
        raise AbstractMethodError(self)

    def _maybe_cache_changed(self, item, value):
        """The object has called back to us saying maybe it has changed.

        numpy < 1.8 has an issue with object arrays and aliasing
        GH6026
        """
        self._data.set(item, value, check=pd._np_version_under1p8)

    @property
    def _is_cached(self):
        """Return boolean indicating if self is cached or not."""
        return getattr(self, '_cacher', None) is not None

    def _get_cacher(self):
        """return my cacher or None"""
        cacher = getattr(self, '_cacher', None)
        if cacher is not None:
            cacher = cacher[1]()
        return cacher

    @property
    def _is_view(self):
        """Return boolean indicating if self is view of another array """
        return self._data.is_view

    def _maybe_update_cacher(self, clear=False, verify_is_copy=True):
        """
        See if we need to update our parent cacher if clear, then clear our
        cache.

        Parameters
        ----------
        clear : boolean, default False
            clear the item cache
        verify_is_copy : boolean, default True
            provide is_copy checks

        """

        cacher = getattr(self, '_cacher', None)
        if cacher is not None:
            ref = cacher[1]()

            # we are trying to reference a dead referant, hence
            # a copy
            if ref is None:
                del self._cacher
            else:
                try:
                    ref._maybe_cache_changed(cacher[0], self)
                except:
                    pass

        if verify_is_copy:
            self._check_setitem_copy(stacklevel=5, t='referant')

        if clear:
            self._clear_item_cache()

    def _clear_item_cache(self, i=None):
        if i is not None:
            self._item_cache.pop(i, None)
        else:
            self._item_cache.clear()

    def _slice(self, slobj, axis=0, kind=None):
        """
        Construct a slice of this container.

        kind parameter is maintained for compatibility with Series slicing.
        """
        axis = self._get_block_manager_axis(axis)
        result = self._constructor(self._data.get_slice(slobj, axis=axis))
        result = result.__finalize__(self)

        # this could be a view
        # but only in a single-dtyped view slicable case
        is_copy = axis != 0 or result._is_view
        result._set_is_copy(self, copy=is_copy)
        return result

    def _set_item(self, key, value):
        self._data.set(key, value)
        self._clear_item_cache()

    def _set_is_copy(self, ref=None, copy=True):
        if not copy:
            self.is_copy = None
        else:
            if ref is not None:
                self.is_copy = weakref.ref(ref)
            else:
                self.is_copy = None

    def _check_is_chained_assignment_possible(self):
        """
        Check if we are a view, have a cacher, and are of mixed type.
        If so, then force a setitem_copy check.

        Should be called just near setting a value

        Will return a boolean if it we are a view and are cached, but a
        single-dtype meaning that the cacher should be updated following
        setting.
        """
        if self._is_view and self._is_cached:
            ref = self._get_cacher()
            if ref is not None and ref._is_mixed_type:
                self._check_setitem_copy(stacklevel=4, t='referant',
                                         force=True)
            return True
        elif self.is_copy:
            self._check_setitem_copy(stacklevel=4, t='referant')
        return False

    def _check_setitem_copy(self, stacklevel=4, t='setting', force=False):
        """

        Parameters
        ----------
        stacklevel : integer, default 4
           the level to show of the stack when the error is output
        t : string, the type of setting error
        force : boolean, default False
           if True, then force showing an error

        validate if we are doing a settitem on a chained copy.

        If you call this function, be sure to set the stacklevel such that the
        user will see the error *at the level of setting*

        It is technically possible to figure out that we are setting on
        a copy even WITH a multi-dtyped pandas object. In other words, some
        blocks may be views while other are not. Currently _is_view will ALWAYS
        return False for multi-blocks to avoid having to handle this case.

        df = DataFrame(np.arange(0,9), columns=['count'])
        df['group'] = 'b'

        # This technically need not raise SettingWithCopy if both are view
        # (which is not # generally guaranteed but is usually True.  However,
        # this is in general not a good practice and we recommend using .loc.
        df.iloc[0:5]['group'] = 'a'

        """

        if force or self.is_copy:

            value = config.get_option('mode.chained_assignment')
            if value is None:
                return

            # see if the copy is not actually refererd; if so, then disolve
            # the copy weakref
            try:
                gc.collect(2)
                if not gc.get_referents(self.is_copy()):
                    self.is_copy = None
                    return
            except:
                pass

            # we might be a false positive
            try:
                if self.is_copy().shape == self.shape:
                    self.is_copy = None
                    return
            except:
                pass

            # a custom message
            if isinstance(self.is_copy, string_types):
                t = self.is_copy

            elif t == 'referant':
                t = ("\n"
                     "A value is trying to be set on a copy of a slice from a "
                     "DataFrame\n\n"
                     "See the caveats in the documentation: "
                     "http://pandas.pydata.org/pandas-docs/stable/"
                     "indexing.html#indexing-view-versus-copy"
                     )

            else:
                t = ("\n"
                     "A value is trying to be set on a copy of a slice from a "
                     "DataFrame.\n"
                     "Try using .loc[row_indexer,col_indexer] = value "
                     "instead\n\nSee the caveats in the documentation: "
                     "http://pandas.pydata.org/pandas-docs/stable/"
                     "indexing.html#indexing-view-versus-copy"
                     )

            if value == 'raise':
                raise SettingWithCopyError(t)
            elif value == 'warn':
                warnings.warn(t, SettingWithCopyWarning, stacklevel=stacklevel)

    def __delitem__(self, key):
        """
        Delete item
        """
        deleted = False

        maybe_shortcut = False
        if hasattr(self, 'columns') and isinstance(self.columns, MultiIndex):
            try:
                maybe_shortcut = key not in self.columns._engine
            except TypeError:
                pass

        if maybe_shortcut:
            # Allow shorthand to delete all columns whose first len(key)
            # elements match key:
            if not isinstance(key, tuple):
                key = (key, )
            for col in self.columns:
                if isinstance(col, tuple) and col[:len(key)] == key:
                    del self[col]
                    deleted = True
        if not deleted:
            # If the above loop ran and didn't delete anything because
            # there was no match, this call should raise the appropriate
            # exception:
            self._data.delete(key)

        # delete from the caches
        try:
            del self._item_cache[key]
        except KeyError:
            pass

    def take(self, indices, axis=0, convert=True, is_copy=True, **kwargs):
        """
        Analogous to ndarray.take

        Parameters
        ----------
        indices : list / array of ints
        axis : int, default 0
        convert : translate neg to pos indices (default)
        is_copy : mark the returned frame as a copy

        Returns
        -------
        taken : type of caller
        """
        nv.validate_take(tuple(), kwargs)
        self._consolidate_inplace()
        new_data = self._data.take(indices,
                                   axis=self._get_block_manager_axis(axis),
                                   convert=True, verify=True)
        result = self._constructor(new_data).__finalize__(self)

        # maybe set copy if we didn't actually change the index
        if is_copy:
            if not result._get_axis(axis).equals(self._get_axis(axis)):
                result._set_is_copy(self)

        return result

    def xs(self, key, axis=0, level=None, drop_level=True):
        """
        Returns a cross-section (row(s) or column(s)) from the
        Series/DataFrame. Defaults to cross-section on the rows (axis=0).

        Parameters
        ----------
        key : object
            Some label contained in the index, or partially in a MultiIndex
        axis : int, default 0
            Axis to retrieve cross-section on
        level : object, defaults to first n levels (n=1 or len(key))
            In case of a key partially contained in a MultiIndex, indicate
            which levels are used. Levels can be referred by label or position.
        drop_level : boolean, default True
            If False, returns object with same levels as self.

        Examples
        --------
        >>> df
           A  B  C
        a  4  5  2
        b  4  0  9
        c  9  7  3
        >>> df.xs('a')
        A    4
        B    5
        C    2
        Name: a
        >>> df.xs('C', axis=1)
        a    2
        b    9
        c    3
        Name: C

        >>> df
                            A  B  C  D
        first second third
        bar   one    1      4  1  8  9
              two    1      7  5  5  0
        baz   one    1      6  6  8  0
              three  2      5  3  5  3
        >>> df.xs(('baz', 'three'))
               A  B  C  D
        third
        2      5  3  5  3
        >>> df.xs('one', level=1)
                     A  B  C  D
        first third
        bar   1      4  1  8  9
        baz   1      6  6  8  0
        >>> df.xs(('baz', 2), level=[0, 'third'])
                A  B  C  D
        second
        three   5  3  5  3

        Returns
        -------
        xs : Series or DataFrame

        Notes
        -----
        xs is only for getting, not setting values.

        MultiIndex Slicers is a generic way to get/set values on any level or
        levels.  It is a superset of xs functionality, see
        :ref:`MultiIndex Slicers <advanced.mi_slicers>`

        """
        axis = self._get_axis_number(axis)
        labels = self._get_axis(axis)
        if level is not None:
            loc, new_ax = labels.get_loc_level(key, level=level,
                                               drop_level=drop_level)

            # create the tuple of the indexer
            indexer = [slice(None)] * self.ndim
            indexer[axis] = loc
            indexer = tuple(indexer)

            result = self.iloc[indexer]
            setattr(result, result._get_axis_name(axis), new_ax)
            return result

        if axis == 1:
            return self[key]

        self._consolidate_inplace()

        index = self.index
        if isinstance(index, MultiIndex):
            loc, new_index = self.index.get_loc_level(key,
                                                      drop_level=drop_level)
        else:
            loc = self.index.get_loc(key)

            if isinstance(loc, np.ndarray):
                if loc.dtype == np.bool_:
                    inds, = loc.nonzero()
                    return self.take(inds, axis=axis, convert=False)
                else:
                    return self.take(loc, axis=axis, convert=True)

            if not is_scalar(loc):
                new_index = self.index[loc]

        if is_scalar(loc):
            new_values = self._data.fast_xs(loc)

            # may need to box a datelike-scalar
            #
            # if we encounter an array-like and we only have 1 dim
            # that means that their are list/ndarrays inside the Series!
            # so just return them (GH 6394)
            if not is_list_like(new_values) or self.ndim == 1:
                return _maybe_box_datetimelike(new_values)

            result = self._constructor_sliced(
                new_values, index=self.columns,
                name=self.index[loc], dtype=new_values.dtype)

        else:
            result = self.iloc[loc]
            result.index = new_index

        # this could be a view
        # but only in a single-dtyped view slicable case
        result._set_is_copy(self, copy=not result._is_view)
        return result

    _xs = xs

    # TODO: Check if this was clearer in 0.12
    def select(self, crit, axis=0):
        """
        Return data corresponding to axis labels matching criteria

        Parameters
        ----------
        crit : function
            To be called on each index (label). Should return True or False
        axis : int

        Returns
        -------
        selection : type of caller
        """
        axis = self._get_axis_number(axis)
        axis_name = self._get_axis_name(axis)
        axis_values = self._get_axis(axis)

        if len(axis_values) > 0:
            new_axis = axis_values[
                np.asarray([bool(crit(label)) for label in axis_values])]
        else:
            new_axis = axis_values

        return self.reindex(**{axis_name: new_axis})

    def reindex_like(self, other, method=None, copy=True, limit=None,
                     tolerance=None):
        """Return an object with matching indices to myself.

        Parameters
        ----------
        other : Object
        method : string or None
        copy : boolean, default True
        limit : int, default None
            Maximum number of consecutive labels to fill for inexact matches.
        tolerance : optional
            Maximum distance between labels of the other object and this
            object for inexact matches.

            .. versionadded:: 0.17.0

        Notes
        -----
        Like calling s.reindex(index=other.index, columns=other.columns,
                               method=...)

        Returns
        -------
        reindexed : same as input
        """
        d = other._construct_axes_dict(axes=self._AXIS_ORDERS, method=method,
                                       copy=copy, limit=limit,
                                       tolerance=tolerance)

        return self.reindex(**d)

    def drop(self, labels, axis=0, level=None, inplace=False, errors='raise'):
        """
        Return new object with labels in requested axis removed.

        Parameters
        ----------
        labels : single label or list-like
        axis : int or axis name
        level : int or level name, default None
            For MultiIndex
        inplace : bool, default False
            If True, do operation inplace and return None.
        errors : {'ignore', 'raise'}, default 'raise'
            If 'ignore', suppress error and existing labels are dropped.

            .. versionadded:: 0.16.1

        Returns
        -------
        dropped : type of caller
        """
        inplace = validate_bool_kwarg(inplace, 'inplace')
        axis = self._get_axis_number(axis)
        axis_name = self._get_axis_name(axis)
        axis, axis_ = self._get_axis(axis), axis

        if axis.is_unique:
            if level is not None:
                if not isinstance(axis, MultiIndex):
                    raise AssertionError('axis must be a MultiIndex')
                new_axis = axis.drop(labels, level=level, errors=errors)
            else:
                new_axis = axis.drop(labels, errors=errors)
            dropped = self.reindex(**{axis_name: new_axis})
            try:
                dropped.axes[axis_].set_names(axis.names, inplace=True)
            except AttributeError:
                pass
            result = dropped

        else:
            labels = com._index_labels_to_array(labels)
            if level is not None:
                if not isinstance(axis, MultiIndex):
                    raise AssertionError('axis must be a MultiIndex')
                indexer = ~axis.get_level_values(level).isin(labels)
            else:
                indexer = ~axis.isin(labels)

            slicer = [slice(None)] * self.ndim
            slicer[self._get_axis_number(axis_name)] = indexer

            result = self.loc[tuple(slicer)]

        if inplace:
            self._update_inplace(result)
        else:
            return result

    def _update_inplace(self, result, verify_is_copy=True):
        """
        Replace self internals with result.

        Parameters
        ----------
        verify_is_copy : boolean, default True
            provide is_copy checks

        """
        # NOTE: This does *not* call __finalize__ and that's an explicit
        # decision that we may revisit in the future.

        self._reset_cache()
        self._clear_item_cache()
        self._data = getattr(result, '_data', result)
        self._maybe_update_cacher(verify_is_copy=verify_is_copy)

    def add_prefix(self, prefix):
        """
        Concatenate prefix string with panel items names.

        Parameters
        ----------
        prefix : string

        Returns
        -------
        with_prefix : type of caller
        """
        new_data = self._data.add_prefix(prefix)
        return self._constructor(new_data).__finalize__(self)

    def add_suffix(self, suffix):
        """
        Concatenate suffix string with panel items names.

        Parameters
        ----------
        suffix : string

        Returns
        -------
        with_suffix : type of caller
        """
        new_data = self._data.add_suffix(suffix)
        return self._constructor(new_data).__finalize__(self)

    _shared_docs['sort_values'] = """
        Sort by the values along either axis

        .. versionadded:: 0.17.0

        Parameters
        ----------%(optional_by)s
        axis : %(axes_single_arg)s, default 0
            Axis to direct sorting
        ascending : bool or list of bool, default True
             Sort ascending vs. descending. Specify list for multiple sort
             orders.  If this is a list of bools, must match the length of
             the by.
        inplace : bool, default False
             if True, perform operation in-place
        kind : {'quicksort', 'mergesort', 'heapsort'}, default 'quicksort'
             Choice of sorting algorithm. See also ndarray.np.sort for more
             information.  `mergesort` is the only stable algorithm. For
             DataFrames, this option is only applied when sorting on a single
             column or label.
        na_position : {'first', 'last'}, default 'last'
             `first` puts NaNs at the beginning, `last` puts NaNs at the end

        Returns
        -------
        sorted_obj : %(klass)s
        """

    def sort_values(self, by, axis=0, ascending=True, inplace=False,
                    kind='quicksort', na_position='last'):
        raise AbstractMethodError(self)

    _shared_docs['sort_index'] = """
        Sort object by labels (along an axis)

        Parameters
        ----------
        axis : %(axes)s to direct sorting
        level : int or level name or list of ints or list of level names
            if not None, sort on values in specified index level(s)
        ascending : boolean, default True
            Sort ascending vs. descending
        inplace : bool, default False
            if True, perform operation in-place
        kind : {'quicksort', 'mergesort', 'heapsort'}, default 'quicksort'
             Choice of sorting algorithm. See also ndarray.np.sort for more
             information.  `mergesort` is the only stable algorithm. For
             DataFrames, this option is only applied when sorting on a single
             column or label.
        na_position : {'first', 'last'}, default 'last'
             `first` puts NaNs at the beginning, `last` puts NaNs at the end.
             Not implemented for MultiIndex.
        sort_remaining : bool, default True
            if true and sorting by level and index is multilevel, sort by other
            levels too (in order) after sorting by specified level

        Returns
        -------
        sorted_obj : %(klass)s
        """

    @Appender(_shared_docs['sort_index'] % dict(axes="axes", klass="NDFrame"))
    def sort_index(self, axis=0, level=None, ascending=True, inplace=False,
                   kind='quicksort', na_position='last', sort_remaining=True):
        inplace = validate_bool_kwarg(inplace, 'inplace')
        axis = self._get_axis_number(axis)
        axis_name = self._get_axis_name(axis)
        labels = self._get_axis(axis)

        if level is not None:
            raise NotImplementedError("level is not implemented")
        if inplace:
            raise NotImplementedError("inplace is not implemented")

        sort_index = labels.argsort()
        if not ascending:
            sort_index = sort_index[::-1]

        new_axis = labels.take(sort_index)
        return self.reindex(**{axis_name: new_axis})

    _shared_docs['reindex'] = """
        Conform %(klass)s to new index with optional filling logic, placing
        NA/NaN in locations having no value in the previous index. A new object
        is produced unless the new index is equivalent to the current one and
        copy=False

        Parameters
        ----------
        %(axes)s : array-like, optional (can be specified in order, or as
            keywords)
            New labels / index to conform to. Preferably an Index object to
            avoid duplicating data
        method : {None, 'backfill'/'bfill', 'pad'/'ffill', 'nearest'}, optional
            method to use for filling holes in reindexed DataFrame.
            Please note: this is only  applicable to DataFrames/Series with a
            monotonically increasing/decreasing index.

            * default: don't fill gaps
            * pad / ffill: propagate last valid observation forward to next
              valid
            * backfill / bfill: use next valid observation to fill gap
            * nearest: use nearest valid observations to fill gap

        copy : boolean, default True
            Return a new object, even if the passed indexes are the same
        level : int or name
            Broadcast across a level, matching Index values on the
            passed MultiIndex level
        fill_value : scalar, default np.NaN
            Value to use for missing values. Defaults to NaN, but can be any
            "compatible" value
        limit : int, default None
            Maximum number of consecutive elements to forward or backward fill
        tolerance : optional
            Maximum distance between original and new labels for inexact
            matches. The values of the index at the matching locations most
            satisfy the equation ``abs(index[indexer] - target) <= tolerance``.

            .. versionadded:: 0.17.0

        Examples
        --------

        Create a dataframe with some fictional data.

        >>> index = ['Firefox', 'Chrome', 'Safari', 'IE10', 'Konqueror']
        >>> df = pd.DataFrame({
        ...      'http_status': [200,200,404,404,301],
        ...      'response_time': [0.04, 0.02, 0.07, 0.08, 1.0]},
        ...       index=index)
        >>> df
                   http_status  response_time
        Firefox            200           0.04
        Chrome             200           0.02
        Safari             404           0.07
        IE10               404           0.08
        Konqueror          301           1.00

        Create a new index and reindex the dataframe. By default
        values in the new index that do not have corresponding
        records in the dataframe are assigned ``NaN``.

        >>> new_index= ['Safari', 'Iceweasel', 'Comodo Dragon', 'IE10',
        ...             'Chrome']
        >>> df.reindex(new_index)
                       http_status  response_time
        Safari               404.0           0.07
        Iceweasel              NaN            NaN
        Comodo Dragon          NaN            NaN
        IE10                 404.0           0.08
        Chrome               200.0           0.02

        We can fill in the missing values by passing a value to
        the keyword ``fill_value``. Because the index is not monotonically
        increasing or decreasing, we cannot use arguments to the keyword
        ``method`` to fill the ``NaN`` values.

        >>> df.reindex(new_index, fill_value=0)
                       http_status  response_time
        Safari                 404           0.07
        Iceweasel                0           0.00
        Comodo Dragon            0           0.00
        IE10                   404           0.08
        Chrome                 200           0.02

        >>> df.reindex(new_index, fill_value='missing')
                      http_status response_time
        Safari                404          0.07
        Iceweasel         missing       missing
        Comodo Dragon     missing       missing
        IE10                  404          0.08
        Chrome                200          0.02

        To further illustrate the filling functionality in
        ``reindex``, we will create a dataframe with a
        monotonically increasing index (for example, a sequence
        of dates).

        >>> date_index = pd.date_range('1/1/2010', periods=6, freq='D')
        >>> df2 = pd.DataFrame({"prices": [100, 101, np.nan, 100, 89, 88]},
        ...                    index=date_index)
        >>> df2
                    prices
        2010-01-01     100
        2010-01-02     101
        2010-01-03     NaN
        2010-01-04     100
        2010-01-05      89
        2010-01-06      88

        Suppose we decide to expand the dataframe to cover a wider
        date range.

        >>> date_index2 = pd.date_range('12/29/2009', periods=10, freq='D')
        >>> df2.reindex(date_index2)
                    prices
        2009-12-29     NaN
        2009-12-30     NaN
        2009-12-31     NaN
        2010-01-01     100
        2010-01-02     101
        2010-01-03     NaN
        2010-01-04     100
        2010-01-05      89
        2010-01-06      88
        2010-01-07     NaN

        The index entries that did not have a value in the original data frame
        (for example, '2009-12-29') are by default filled with ``NaN``.
        If desired, we can fill in the missing values using one of several
        options.

        For example, to backpropagate the last valid value to fill the ``NaN``
        values, pass ``bfill`` as an argument to the ``method`` keyword.

        >>> df2.reindex(date_index2, method='bfill')
                    prices
        2009-12-29     100
        2009-12-30     100
        2009-12-31     100
        2010-01-01     100
        2010-01-02     101
        2010-01-03     NaN
        2010-01-04     100
        2010-01-05      89
        2010-01-06      88
        2010-01-07     NaN

        Please note that the ``NaN`` value present in the original dataframe
        (at index value 2010-01-03) will not be filled by any of the
        value propagation schemes. This is because filling while reindexing
        does not look at dataframe values, but only compares the original and
        desired indexes. If you do want to fill in the ``NaN`` values present
        in the original dataframe, use the ``fillna()`` method.

        Returns
        -------
        reindexed : %(klass)s
        """

    # TODO: Decide if we care about having different examples for different
    #       kinds

    @Appender(_shared_docs['reindex'] % dict(axes="axes", klass="NDFrame"))
    def reindex(self, *args, **kwargs):

        # construct the args
        axes, kwargs = self._construct_axes_from_arguments(args, kwargs)
        method = missing.clean_reindex_fill_method(kwargs.pop('method', None))
        level = kwargs.pop('level', None)
        copy = kwargs.pop('copy', True)
        limit = kwargs.pop('limit', None)
        tolerance = kwargs.pop('tolerance', None)
        fill_value = kwargs.pop('fill_value', np.nan)

        if kwargs:
            raise TypeError('reindex() got an unexpected keyword '
                            'argument "{0}"'.format(list(kwargs.keys())[0]))

        self._consolidate_inplace()

        # if all axes that are requested to reindex are equal, then only copy
        # if indicated must have index names equal here as well as values
        if all([self._get_axis(axis).identical(ax)
                for axis, ax in axes.items() if ax is not None]):
            if copy:
                return self.copy()
            return self

        # check if we are a multi reindex
        if self._needs_reindex_multi(axes, method, level):
            try:
                return self._reindex_multi(axes, copy, fill_value)
            except:
                pass

        # perform the reindex on the axes
        return self._reindex_axes(axes, level, limit, tolerance, method,
                                  fill_value, copy).__finalize__(self)

    def _reindex_axes(self, axes, level, limit, tolerance, method, fill_value,
                      copy):
        """Perform the reindex for all the axes."""
        obj = self
        for a in self._AXIS_ORDERS:
            labels = axes[a]
            if labels is None:
                continue

            ax = self._get_axis(a)
            new_index, indexer = ax.reindex(labels, level=level, limit=limit,
                                            tolerance=tolerance, method=method)

            axis = self._get_axis_number(a)
            obj = obj._reindex_with_indexers({axis: [new_index, indexer]},
                                             fill_value=fill_value,
                                             copy=copy, allow_dups=False)

        return obj

    def _needs_reindex_multi(self, axes, method, level):
        """Check if we do need a multi reindex."""
        return ((com._count_not_none(*axes.values()) == self._AXIS_LEN) and
                method is None and level is None and not self._is_mixed_type)

    def _reindex_multi(self, axes, copy, fill_value):
        return NotImplemented

    _shared_docs[
        'reindex_axis'] = ("""Conform input object to new index with optional
        filling logic, placing NA/NaN in locations having no value in the
        previous index. A new object is produced unless the new index is
        equivalent to the current one and copy=False

        Parameters
        ----------
        labels : array-like
            New labels / index to conform to. Preferably an Index object to
            avoid duplicating data
        axis : %(axes_single_arg)s
        method : {None, 'backfill'/'bfill', 'pad'/'ffill', 'nearest'}, optional
            Method to use for filling holes in reindexed DataFrame:

            * default: don't fill gaps
            * pad / ffill: propagate last valid observation forward to next
              valid
            * backfill / bfill: use next valid observation to fill gap
            * nearest: use nearest valid observations to fill gap

        copy : boolean, default True
            Return a new object, even if the passed indexes are the same
        level : int or name
            Broadcast across a level, matching Index values on the
            passed MultiIndex level
        limit : int, default None
            Maximum number of consecutive elements to forward or backward fill
        tolerance : optional
            Maximum distance between original and new labels for inexact
            matches. The values of the index at the matching locations most
            satisfy the equation ``abs(index[indexer] - target) <= tolerance``.

            .. versionadded:: 0.17.0

        Examples
        --------
        >>> df.reindex_axis(['A', 'B', 'C'], axis=1)

        See Also
        --------
        reindex, reindex_like

        Returns
        -------
        reindexed : %(klass)s
        """)

    @Appender(_shared_docs['reindex_axis'] % _shared_doc_kwargs)
    def reindex_axis(self, labels, axis=0, method=None, level=None, copy=True,
                     limit=None, fill_value=np.nan):
        self._consolidate_inplace()

        axis_name = self._get_axis_name(axis)
        axis_values = self._get_axis(axis_name)
        method = missing.clean_reindex_fill_method(method)
        new_index, indexer = axis_values.reindex(labels, method, level,
                                                 limit=limit)
        return self._reindex_with_indexers({axis: [new_index, indexer]},
                                           fill_value=fill_value, copy=copy)

    def _reindex_with_indexers(self, reindexers, fill_value=np.nan, copy=False,
                               allow_dups=False):
        """allow_dups indicates an internal call here """

        # reindex doing multiple operations on different axes if indicated
        new_data = self._data
        for axis in sorted(reindexers.keys()):
            index, indexer = reindexers[axis]
            baxis = self._get_block_manager_axis(axis)

            if index is None:
                continue

            index = _ensure_index(index)
            if indexer is not None:
                indexer = _ensure_int64(indexer)

            # TODO: speed up on homogeneous DataFrame objects
            new_data = new_data.reindex_indexer(index, indexer, axis=baxis,
                                                fill_value=fill_value,
                                                allow_dups=allow_dups,
                                                copy=copy)

        if copy and new_data is self._data:
            new_data = new_data.copy()

        return self._constructor(new_data).__finalize__(self)

    def _reindex_axis(self, new_index, fill_method, axis, copy):
        new_data = self._data.reindex_axis(new_index, axis=axis,
                                           method=fill_method, copy=copy)

        if new_data is self._data and not copy:
            return self
        else:
            return self._constructor(new_data).__finalize__(self)

    def filter(self, items=None, like=None, regex=None, axis=None):
        """
        Subset rows or columns of dataframe according to labels in
        the specified index.

        Note that this routine does not filter a dataframe on its
        contents. The filter is applied to the labels of the index.

        Parameters
        ----------
        items : list-like
            List of info axis to restrict to (must not all be present)
        like : string
            Keep info axis where "arg in col == True"
        regex : string (regular expression)
            Keep info axis with re.search(regex, col) == True
        axis : int or string axis name
            The axis to filter on.  By default this is the info axis,
            'index' for Series, 'columns' for DataFrame

        Returns
        -------
        same type as input object

        Examples
        --------
        >>> df
        one  two  three
        mouse     1    2      3
        rabbit    4    5      6

        >>> # select columns by name
        >>> df.filter(items=['one', 'three'])
        one  three
        mouse     1      3
        rabbit    4      6

        >>> # select columns by regular expression
        >>> df.filter(regex='e$', axis=1)
        one  three
        mouse     1      3
        rabbit    4      6

        >>> # select rows containing 'bbi'
        >>> df.filter(like='bbi', axis=0)
        one  two  three
        rabbit    4    5      6

        See Also
        --------
        pandas.DataFrame.select

        Notes
        -----
        The ``items``, ``like``, and ``regex`` parameters are
        enforced to be mutually exclusive.

        ``axis`` defaults to the info axis that is used when indexing
        with ``[]``.
        """
        import re

        nkw = sum([x is not None for x in [items, like, regex]])
        if nkw > 1:
            raise TypeError('Keyword arguments `items`, `like`, or `regex` '
                            'are mutually exclusive')

        if axis is None:
            axis = self._info_axis_name
        axis_name = self._get_axis_name(axis)
        axis_values = self._get_axis(axis_name)

        if items is not None:
            return self.reindex(**{axis_name:
                                   [r for r in items if r in axis_values]})
        elif like:
            matchf = lambda x: (like in x if isinstance(x, string_types) else
                                like in str(x))
            return self.select(matchf, axis=axis_name)
        elif regex:
            matcher = re.compile(regex)
            return self.select(lambda x: matcher.search(str(x)) is not None,
                               axis=axis_name)
        else:
            raise TypeError('Must pass either `items`, `like`, or `regex`')

    def head(self, n=5):
        """
        Returns first n rows
        """
        return self.iloc[:n]

    def tail(self, n=5):
        """
        Returns last n rows
        """
        if n == 0:
            return self.iloc[0:0]
        return self.iloc[-n:]

    def sample(self, n=None, frac=None, replace=False, weights=None,
               random_state=None, axis=None):
        """
        Returns a random sample of items from an axis of object.

        .. versionadded:: 0.16.1

        Parameters
        ----------
        n : int, optional
            Number of items from axis to return. Cannot be used with `frac`.
            Default = 1 if `frac` = None.
        frac : float, optional
            Fraction of axis items to return. Cannot be used with `n`.
        replace : boolean, optional
            Sample with or without replacement. Default = False.
        weights : str or ndarray-like, optional
            Default 'None' results in equal probability weighting.
            If passed a Series, will align with target object on index. Index
            values in weights not found in sampled object will be ignored and
            index values in sampled object not in weights will be assigned
            weights of zero.
            If called on a DataFrame, will accept the name of a column
            when axis = 0.
            Unless weights are a Series, weights must be same length as axis
            being sampled.
            If weights do not sum to 1, they will be normalized to sum to 1.
            Missing values in the weights column will be treated as zero.
            inf and -inf values not allowed.
        random_state : int or numpy.random.RandomState, optional
            Seed for the random number generator (if int), or numpy RandomState
            object.
        axis : int or string, optional
            Axis to sample. Accepts axis number or name. Default is stat axis
            for given data type (0 for Series and DataFrames, 1 for Panels).

        Returns
        -------
        A new object of same type as caller.

        Examples
        --------

        Generate an example ``Series`` and ``DataFrame``:

        >>> s = pd.Series(np.random.randn(50))
        >>> s.head()
        0   -0.038497
        1    1.820773
        2   -0.972766
        3   -1.598270
        4   -1.095526
        dtype: float64
        >>> df = pd.DataFrame(np.random.randn(50, 4), columns=list('ABCD'))
        >>> df.head()
                  A         B         C         D
        0  0.016443 -2.318952 -0.566372 -1.028078
        1 -1.051921  0.438836  0.658280 -0.175797
        2 -1.243569 -0.364626 -0.215065  0.057736
        3  1.768216  0.404512 -0.385604 -1.457834
        4  1.072446 -1.137172  0.314194 -0.046661

        Next extract a random sample from both of these objects...

        3 random elements from the ``Series``:

        >>> s.sample(n=3)
        27   -0.994689
        55   -1.049016
        67   -0.224565
        dtype: float64

        And a random 10% of the ``DataFrame`` with replacement:

        >>> df.sample(frac=0.1, replace=True)
                   A         B         C         D
        35  1.981780  0.142106  1.817165 -0.290805
        49 -1.336199 -0.448634 -0.789640  0.217116
        40  0.823173 -0.078816  1.009536  1.015108
        15  1.421154 -0.055301 -1.922594 -0.019696
        6  -0.148339  0.832938  1.787600 -1.383767
        """

        if axis is None:
            axis = self._stat_axis_number

        axis = self._get_axis_number(axis)
        axis_length = self.shape[axis]

        # Process random_state argument
        rs = com._random_state(random_state)

        # Check weights for compliance
        if weights is not None:

            # If a series, align with frame
            if isinstance(weights, pd.Series):
                weights = weights.reindex(self.axes[axis])

            # Strings acceptable if a dataframe and axis = 0
            if isinstance(weights, string_types):
                if isinstance(self, pd.DataFrame):
                    if axis == 0:
                        try:
                            weights = self[weights]
                        except KeyError:
                            raise KeyError("String passed to weights not a "
                                           "valid column")
                    else:
                        raise ValueError("Strings can only be passed to "
                                         "weights when sampling from rows on "
                                         "a DataFrame")
                else:
                    raise ValueError("Strings cannot be passed as weights "
                                     "when sampling from a Series or Panel.")

            weights = pd.Series(weights, dtype='float64')

            if len(weights) != axis_length:
                raise ValueError("Weights and axis to be sampled must be of "
                                 "same length")

            if (weights == np.inf).any() or (weights == -np.inf).any():
                raise ValueError("weight vector may not include `inf` values")

            if (weights < 0).any():
                raise ValueError("weight vector many not include negative "
                                 "values")

            # If has nan, set to zero.
            weights = weights.fillna(0)

            # Renormalize if don't sum to 1
            if weights.sum() != 1:
                if weights.sum() != 0:
                    weights = weights / weights.sum()
                else:
                    raise ValueError("Invalid weights: weights sum to zero")

            weights = weights.values

        # If no frac or n, default to n=1.
        if n is None and frac is None:
            n = 1
        elif n is not None and frac is None and n % 1 != 0:
            raise ValueError("Only integers accepted as `n` values")
        elif n is None and frac is not None:
            n = int(round(frac * axis_length))
        elif n is not None and frac is not None:
            raise ValueError('Please enter a value for `frac` OR `n`, not '
                             'both')

        # Check for negative sizes
        if n < 0:
            raise ValueError("A negative number of rows requested. Please "
                             "provide positive value.")

        locs = rs.choice(axis_length, size=n, replace=replace, p=weights)
        return self.take(locs, axis=axis, is_copy=False)

    _shared_docs['pipe'] = ("""
        Apply func(self, \*args, \*\*kwargs)

        .. versionadded:: 0.16.2

        Parameters
        ----------
        func : function
            function to apply to the %(klass)s.
            ``args``, and ``kwargs`` are passed into ``func``.
            Alternatively a ``(callable, data_keyword)`` tuple where
            ``data_keyword`` is a string indicating the keyword of
            ``callable`` that expects the %(klass)s.
        args : positional arguments passed into ``func``.
        kwargs : a dictionary of keyword arguments passed into ``func``.

        Returns
        -------
        object : the return type of ``func``.

        Notes
        -----

        Use ``.pipe`` when chaining together functions that expect
        on Series or DataFrames. Instead of writing

        >>> f(g(h(df), arg1=a), arg2=b, arg3=c)

        You can write

        >>> (df.pipe(h)
        ...    .pipe(g, arg1=a)
        ...    .pipe(f, arg2=b, arg3=c)
        ... )

        If you have a function that takes the data as (say) the second
        argument, pass a tuple indicating which keyword expects the
        data. For example, suppose ``f`` takes its data as ``arg2``:

        >>> (df.pipe(h)
        ...    .pipe(g, arg1=a)
        ...    .pipe((f, 'arg2'), arg1=a, arg3=c)
        ...  )

        See Also
        --------
        pandas.DataFrame.apply
        pandas.DataFrame.applymap
        pandas.Series.map
    """)

    @Appender(_shared_docs['pipe'] % _shared_doc_kwargs)
    def pipe(self, func, *args, **kwargs):
        if isinstance(func, tuple):
            func, target = func
            if target in kwargs:
                raise ValueError('%s is both the pipe target and a keyword '
                                 'argument' % target)
            kwargs[target] = self
            return func(*args, **kwargs)
        else:
            return func(self, *args, **kwargs)

    _shared_docs['aggregate'] = ("""
    Aggregate using callable, string, dict, or list of string/callables

    %(versionadded)s

    Parameters
    ----------
    func : callable, string, dictionary, or list of string/callables
        Function to use for aggregating the data. If a function, must either
        work when passed a %(klass)s or when passed to %(klass)s.apply. For
        a DataFrame, can pass a dict, if the keys are DataFrame column names.

        Accepted Combinations are:

        - string function name
        - function
        - list of functions
        - dict of column names -> functions (or list of functions)

    Notes
    -----
    Numpy functions mean/median/prod/sum/std/var are special cased so the
    default behavior is applying the function along axis=0
    (e.g., np.mean(arr_2d, axis=0)) as opposed to
    mimicking the default Numpy behavior (e.g., np.mean(arr_2d)).

    agg is an alias for aggregate. Use it.

    Returns
    -------
    aggregated : %(klass)s
    """)

    _shared_docs['transform'] = ("""
    Call function producing a like-indexed %(klass)s
    and return a %(klass)s with the transformed values`

    .. versionadded:: 0.20.0

    Parameters
    ----------
    func : callable, string, dictionary, or list of string/callables
        To apply to column

        Accepted Combinations are:

        - string function name
        - function
        - list of functions
        - dict of column names -> functions (or list of functions)

    Returns
    -------
    transformed : %(klass)s

    Examples
    --------
    >>> df = pd.DataFrame(np.random.randn(10, 3), columns=['A', 'B', 'C'],
    ...                   index=pd.date_range('1/1/2000', periods=10))
    df.iloc[3:7] = np.nan

    >>> df.transform(lambda x: (x - x.mean()) / x.std())
                       A         B         C
    2000-01-01  0.579457  1.236184  0.123424
    2000-01-02  0.370357 -0.605875 -1.231325
    2000-01-03  1.455756 -0.277446  0.288967
    2000-01-04       NaN       NaN       NaN
    2000-01-05       NaN       NaN       NaN
    2000-01-06       NaN       NaN       NaN
    2000-01-07       NaN       NaN       NaN
    2000-01-08 -0.498658  1.274522  1.642524
    2000-01-09 -0.540524 -1.012676 -0.828968
    2000-01-10 -1.366388 -0.614710  0.005378

    See also
    --------
    pandas.%(klass)s.aggregate
    pandas.%(klass)s.apply

    """)

    # ----------------------------------------------------------------------
    # Attribute access

    def __finalize__(self, other, method=None, **kwargs):
        """
        Propagate metadata from other to self.

        Parameters
        ----------
        other : the object from which to get the attributes that we are going
            to propagate
        method : optional, a passed method name ; possibly to take different
            types of propagation actions based on this

        """
        if isinstance(other, NDFrame):
            for name in self._metadata:
                object.__setattr__(self, name, getattr(other, name, None))
        return self

    def __getattr__(self, name):
        """After regular attribute access, try looking up the name
        This allows simpler access to columns for interactive use.
        """

        # Note: obj.x will always call obj.__getattribute__('x') prior to
        # calling obj.__getattr__('x').

        if (name in self._internal_names_set or name in self._metadata or
                name in self._accessors):
            return object.__getattribute__(self, name)
        else:
            if name in self._info_axis:
                return self[name]
            return object.__getattribute__(self, name)

    def __setattr__(self, name, value):
        """After regular attribute access, try setting the name
        This allows simpler access to columns for interactive use.
        """

        # first try regular attribute access via __getattribute__, so that
        # e.g. ``obj.x`` and ``obj.x = 4`` will always reference/modify
        # the same attribute.

        try:
            object.__getattribute__(self, name)
            return object.__setattr__(self, name, value)
        except AttributeError:
            pass

        # if this fails, go on to more involved attribute setting
        # (note that this matches __getattr__, above).
        if name in self._internal_names_set:
            object.__setattr__(self, name, value)
        elif name in self._metadata:
            object.__setattr__(self, name, value)
        else:
            try:
                existing = getattr(self, name)
                if isinstance(existing, Index):
                    object.__setattr__(self, name, value)
                elif name in self._info_axis:
                    self[name] = value
                else:
                    object.__setattr__(self, name, value)
            except (AttributeError, TypeError):
                object.__setattr__(self, name, value)

    # ----------------------------------------------------------------------
    # Getting and setting elements

    # ----------------------------------------------------------------------
    # Consolidation of internals

    def _protect_consolidate(self, f):
        """Consolidate _data -- if the blocks have changed, then clear the
        cache
        """
        blocks_before = len(self._data.blocks)
        result = f()
        if len(self._data.blocks) != blocks_before:
            self._clear_item_cache()
        return result

    def _consolidate_inplace(self):
        """Consolidate data in place and return None"""

        def f():
            self._data = self._data.consolidate()

        self._protect_consolidate(f)

    def _consolidate(self, inplace=False):
        """
        Compute NDFrame with "consolidated" internals (data of each dtype
        grouped together in a single ndarray).

        Parameters
        ----------
        inplace : boolean, default False
            If False return new object, otherwise modify existing object

        Returns
        -------
        consolidated : type of caller
        """
        inplace = validate_bool_kwarg(inplace, 'inplace')
        if inplace:
            self._consolidate_inplace()
        else:
            f = lambda: self._data.consolidate()
            cons_data = self._protect_consolidate(f)
            return self._constructor(cons_data).__finalize__(self)

    def consolidate(self, inplace=False):
        """
        DEPRECATED: consolidate will be an internal implementation only.
        """
        # 15483
        warnings.warn("consolidate is deprecated and will be removed in a "
                      "future release.", FutureWarning, stacklevel=2)
        return self._consolidate(inplace)

    @property
    def _is_mixed_type(self):
        f = lambda: self._data.is_mixed_type
        return self._protect_consolidate(f)

    @property
    def _is_numeric_mixed_type(self):
        f = lambda: self._data.is_numeric_mixed_type
        return self._protect_consolidate(f)

    @property
    def _is_datelike_mixed_type(self):
        f = lambda: self._data.is_datelike_mixed_type
        return self._protect_consolidate(f)

    def _check_inplace_setting(self, value):
        """ check whether we allow in-place setting with this type of value """

        if self._is_mixed_type:
            if not self._is_numeric_mixed_type:

                # allow an actual np.nan thru
                try:
                    if np.isnan(value):
                        return True
                except:
                    pass

                raise TypeError('Cannot do inplace boolean setting on '
                                'mixed-types with a non np.nan value')

        return True

    def _get_numeric_data(self):
        return self._constructor(
            self._data.get_numeric_data()).__finalize__(self)

    def _get_bool_data(self):
        return self._constructor(self._data.get_bool_data()).__finalize__(self)

    # ----------------------------------------------------------------------
    # Internal Interface Methods

    def as_matrix(self, columns=None):
        """
        Convert the frame to its Numpy-array representation.

        Parameters
        ----------
        columns: list, optional, default:None
            If None, return all columns, otherwise, returns specified columns.

        Returns
        -------
        values : ndarray
            If the caller is heterogeneous and contains booleans or objects,
            the result will be of dtype=object. See Notes.


        Notes
        -----
        Return is NOT a Numpy-matrix, rather, a Numpy-array.

        The dtype will be a lower-common-denominator dtype (implicit
        upcasting); that is to say if the dtypes (even of numeric types)
        are mixed, the one that accommodates all will be chosen. Use this
        with care if you are not dealing with the blocks.

        e.g. If the dtypes are float16 and float32, dtype will be upcast to
        float32.  If dtypes are int32 and uint8, dtype will be upcase to
        int32. By numpy.find_common_type convention, mixing int64 and uint64
        will result in a flot64 dtype.

        This method is provided for backwards compatibility. Generally,
        it is recommended to use '.values'.

        See Also
        --------
        pandas.DataFrame.values
        """
        self._consolidate_inplace()
        if self._AXIS_REVERSED:
            return self._data.as_matrix(columns).T
        return self._data.as_matrix(columns)

    @property
    def values(self):
        """Numpy representation of NDFrame

        Notes
        -----
        The dtype will be a lower-common-denominator dtype (implicit
        upcasting); that is to say if the dtypes (even of numeric types)
        are mixed, the one that accommodates all will be chosen. Use this
        with care if you are not dealing with the blocks.

        e.g. If the dtypes are float16 and float32, dtype will be upcast to
        float32.  If dtypes are int32 and uint8, dtype will be upcast to
        int32. By numpy.find_common_type convention, mixing int64 and uint64
        will result in a flot64 dtype.
        """
        return self.as_matrix()

    @property
    def _values(self):
        """internal implementation"""
        return self.values

    @property
    def _get_values(self):
        # compat
        return self.as_matrix()

    def get_values(self):
        """same as values (but handles sparseness conversions)"""
        return self.as_matrix()

    def get_dtype_counts(self):
        """Return the counts of dtypes in this object."""
        from pandas import Series
        return Series(self._data.get_dtype_counts())

    def get_ftype_counts(self):
        """Return the counts of ftypes in this object."""
        from pandas import Series
        return Series(self._data.get_ftype_counts())

    @property
    def dtypes(self):
        """Return the dtypes in this object."""
        from pandas import Series
        return Series(self._data.get_dtypes(), index=self._info_axis,
                      dtype=np.object_)

    @property
    def ftypes(self):
        """
        Return the ftypes (indication of sparse/dense and dtype)
        in this object.
        """
        from pandas import Series
        return Series(self._data.get_ftypes(), index=self._info_axis,
                      dtype=np.object_)

    def as_blocks(self, copy=True):
        """
        Convert the frame to a dict of dtype -> Constructor Types that each has
        a homogeneous dtype.

        NOTE: the dtypes of the blocks WILL BE PRESERVED HERE (unlike in
              as_matrix)

        Parameters
        ----------
        copy : boolean, default True

               .. versionadded: 0.16.1

        Returns
        -------
        values : a dict of dtype -> Constructor Types
        """
        self._consolidate_inplace()

        bd = {}
        for b in self._data.blocks:
            bd.setdefault(str(b.dtype), []).append(b)

        result = {}
        for dtype, blocks in bd.items():
            # Must combine even after consolidation, because there may be
            # sparse items which are never consolidated into one block.
            combined = self._data.combine(blocks, copy=copy)
            result[dtype] = self._constructor(combined).__finalize__(self)

        return result

    @property
    def blocks(self):
        """Internal property, property synonym for as_blocks()"""
        return self.as_blocks()

    @deprecate_kwarg(old_arg_name='raise_on_error', new_arg_name='errors',
                     mapping={True: 'raise', False: 'ignore'})
    def astype(self, dtype, copy=True, errors='raise', **kwargs):
        """
        Cast object to input numpy.dtype
        Return a copy when copy = True (be really careful with this!)

        Parameters
        ----------
        dtype : data type, or dict of column name -> data type
            Use a numpy.dtype or Python type to cast entire pandas object to
            the same type. Alternatively, use {col: dtype, ...}, where col is a
            column label and dtype is a numpy.dtype or Python type to cast one
            or more of the DataFrame's columns to column-specific types.
        errors : {'raise', 'ignore'}, default 'raise'.
            Control raising of exceptions on invalid data for provided dtype.

            - ``raise`` : allow exceptions to be raised
            - ``ignore`` : suppress exceptions. On error return original object

            .. versionadded:: 0.20.0

        raise_on_error : DEPRECATED use ``errors`` instead
        kwargs : keyword arguments to pass on to the constructor

        Returns
        -------
        casted : type of caller
        """
        if isinstance(dtype, collections.Mapping):
            if self.ndim == 1:  # i.e. Series
                if len(dtype) > 1 or list(dtype.keys())[0] != self.name:
                    raise KeyError('Only the Series name can be used for '
                                   'the key in Series dtype mappings.')
                new_type = list(dtype.values())[0]
                return self.astype(new_type, copy, errors, **kwargs)
            elif self.ndim > 2:
                raise NotImplementedError(
                    'astype() only accepts a dtype arg of type dict when '
                    'invoked on Series and DataFrames. A single dtype must be '
                    'specified when invoked on a Panel.'
                )
            for col_name in dtype.keys():
                if col_name not in self:
                    raise KeyError('Only a column name can be used for the '
                                   'key in a dtype mappings argument.')
            from pandas import concat
            results = []
            for col_name, col in self.iteritems():
                if col_name in dtype:
                    results.append(col.astype(dtype[col_name], copy=copy))
                else:
                    results.append(results.append(col.copy() if copy else col))
            return concat(results, axis=1, copy=False)

        # else, only a single dtype is given
        new_data = self._data.astype(dtype=dtype, copy=copy, errors=errors,
                                     **kwargs)
        return self._constructor(new_data).__finalize__(self)

    def copy(self, deep=True):
        """
        Make a copy of this objects data.

        Parameters
        ----------
        deep : boolean or string, default True
            Make a deep copy, including a copy of the data and the indices.
            With ``deep=False`` neither the indices or the data are copied.

            Note that when ``deep=True`` data is copied, actual python objects
            will not be copied recursively, only the reference to the object.
            This is in contrast to ``copy.deepcopy`` in the Standard Library,
            which recursively copies object data.

        Returns
        -------
        copy : type of caller
        """
        data = self._data.copy(deep=deep)
        return self._constructor(data).__finalize__(self)

    def __copy__(self, deep=True):
        return self.copy(deep=deep)

    def __deepcopy__(self, memo=None):
        if memo is None:
            memo = {}
        return self.copy(deep=True)

    def _convert(self, datetime=False, numeric=False, timedelta=False,
                 coerce=False, copy=True):
        """
        Attempt to infer better dtype for object columns

        Parameters
        ----------
        datetime : boolean, default False
            If True, convert to date where possible.
        numeric : boolean, default False
            If True, attempt to convert to numbers (including strings), with
            unconvertible values becoming NaN.
        timedelta : boolean, default False
            If True, convert to timedelta where possible.
        coerce : boolean, default False
            If True, force conversion with unconvertible values converted to
            nulls (NaN or NaT)
        copy : boolean, default True
            If True, return a copy even if no copy is necessary (e.g. no
            conversion was done). Note: This is meant for internal use, and
            should not be confused with inplace.

        Returns
        -------
        converted : same as input object
        """
        return self._constructor(
            self._data.convert(datetime=datetime, numeric=numeric,
                               timedelta=timedelta, coerce=coerce,
                               copy=copy)).__finalize__(self)

    # TODO: Remove in 0.18 or 2017, which ever is sooner
    def convert_objects(self, convert_dates=True, convert_numeric=False,
                        convert_timedeltas=True, copy=True):
        """
        Deprecated.

        Attempt to infer better dtype for object columns

        Parameters
        ----------
        convert_dates : boolean, default True
            If True, convert to date where possible. If 'coerce', force
            conversion, with unconvertible values becoming NaT.
        convert_numeric : boolean, default False
            If True, attempt to coerce to numbers (including strings), with
            unconvertible values becoming NaN.
        convert_timedeltas : boolean, default True
            If True, convert to timedelta where possible. If 'coerce', force
            conversion, with unconvertible values becoming NaT.
        copy : boolean, default True
            If True, return a copy even if no copy is necessary (e.g. no
            conversion was done). Note: This is meant for internal use, and
            should not be confused with inplace.

        See Also
        --------
        pandas.to_datetime : Convert argument to datetime.
        pandas.to_timedelta : Convert argument to timedelta.
        pandas.to_numeric : Return a fixed frequency timedelta index,
            with day as the default.

        Returns
        -------
        converted : same as input object
        """
        from warnings import warn
        warn("convert_objects is deprecated.  Use the data-type specific "
             "converters pd.to_datetime, pd.to_timedelta and pd.to_numeric.",
             FutureWarning, stacklevel=2)

        return self._constructor(
            self._data.convert(convert_dates=convert_dates,
                               convert_numeric=convert_numeric,
                               convert_timedeltas=convert_timedeltas,
                               copy=copy)).__finalize__(self)

    # ----------------------------------------------------------------------
    # Filling NA's

    _shared_docs['fillna'] = ("""
        Fill NA/NaN values using the specified method

        Parameters
        ----------
        value : scalar, dict, Series, or DataFrame
            Value to use to fill holes (e.g. 0), alternately a
            dict/Series/DataFrame of values specifying which value to use for
            each index (for a Series) or column (for a DataFrame). (values not
            in the dict/Series/DataFrame will not be filled). This value cannot
            be a list.
        method : {'backfill', 'bfill', 'pad', 'ffill', None}, default None
            Method to use for filling holes in reindexed Series
            pad / ffill: propagate last valid observation forward to next valid
            backfill / bfill: use NEXT valid observation to fill gap
        axis : %(axes_single_arg)s
        inplace : boolean, default False
            If True, fill in place. Note: this will modify any
            other views on this object, (e.g. a no-copy slice for a column in a
            DataFrame).
        limit : int, default None
            If method is specified, this is the maximum number of consecutive
            NaN values to forward/backward fill. In other words, if there is
            a gap with more than this number of consecutive NaNs, it will only
            be partially filled. If method is not specified, this is the
            maximum number of entries along the entire axis where NaNs will be
            filled. Must be greater than 0 if not None.
        downcast : dict, default is None
            a dict of item->dtype of what to downcast if possible,
            or the string 'infer' which will try to downcast to an appropriate
            equal type (e.g. float64 to int64 if possible)

        See Also
        --------
        reindex, asfreq

        Returns
        -------
        filled : %(klass)s
        """)

    @Appender(_shared_docs['fillna'] % _shared_doc_kwargs)
    def fillna(self, value=None, method=None, axis=None, inplace=False,
               limit=None, downcast=None):
        inplace = validate_bool_kwarg(inplace, 'inplace')

        if isinstance(value, (list, tuple)):
            raise TypeError('"value" parameter must be a scalar or dict, but '
                            'you passed a "{0}"'.format(type(value).__name__))
        self._consolidate_inplace()

        # set the default here, so functions examining the signaure
        # can detect if something was set (e.g. in groupby) (GH9221)
        if axis is None:
            axis = 0
        axis = self._get_axis_number(axis)
        method = missing.clean_fill_method(method)
        from pandas import DataFrame
        if value is None:
            if method is None:
                raise ValueError('must specify a fill method or value')
            if self._is_mixed_type and axis == 1:
                if inplace:
                    raise NotImplementedError()
                result = self.T.fillna(method=method, limit=limit).T

                # need to downcast here because of all of the transposes
                result._data = result._data.downcast()

                return result

            # > 3d
            if self.ndim > 3:
                raise NotImplementedError('Cannot fillna with a method for > '
                                          '3dims')

            # 3d
            elif self.ndim == 3:

                # fill in 2d chunks
                result = dict([(col, s.fillna(method=method, value=value))
                               for col, s in self.iteritems()])
                new_obj = self._constructor.\
                    from_dict(result).__finalize__(self)
                new_data = new_obj._data

            else:
                # 2d or less
                method = missing.clean_fill_method(method)
                new_data = self._data.interpolate(method=method, axis=axis,
                                                  limit=limit, inplace=inplace,
                                                  coerce=True,
                                                  downcast=downcast)
        else:
            if method is not None:
                raise ValueError('cannot specify both a fill method and value')

            if len(self._get_axis(axis)) == 0:
                return self

            if self.ndim == 1:
                if isinstance(value, (dict, ABCSeries)):
                    from pandas import Series
                    value = Series(value)
                elif not is_list_like(value):
                    pass
                else:
                    raise ValueError("invalid fill value with a %s" %
                                     type(value))

                new_data = self._data.fillna(value=value, limit=limit,
                                             inplace=inplace,
                                             downcast=downcast)

            elif isinstance(value, (dict, ABCSeries)):
                if axis == 1:
                    raise NotImplementedError('Currently only can fill '
                                              'with dict/Series column '
                                              'by column')

                result = self if inplace else self.copy()
                for k, v in compat.iteritems(value):
                    if k not in result:
                        continue
                    obj = result[k]
                    obj.fillna(v, limit=limit, inplace=True, downcast=downcast)
                return result
            elif not is_list_like(value):
                new_data = self._data.fillna(value=value, limit=limit,
                                             inplace=inplace,
                                             downcast=downcast)
            elif isinstance(value, DataFrame) and self.ndim == 2:
                new_data = self.where(self.notnull(), value)
            else:
                raise ValueError("invalid fill value with a %s" % type(value))

        if inplace:
            self._update_inplace(new_data)
        else:
            return self._constructor(new_data).__finalize__(self)

    def ffill(self, axis=None, inplace=False, limit=None, downcast=None):
        """
        Synonym for :meth:`DataFrame.fillna(method='ffill') <DataFrame.fillna>`
        """
        return self.fillna(method='ffill', axis=axis, inplace=inplace,
                           limit=limit, downcast=downcast)

    def bfill(self, axis=None, inplace=False, limit=None, downcast=None):
        """
        Synonym for :meth:`DataFrame.fillna(method='bfill') <DataFrame.fillna>`
        """
        return self.fillna(method='bfill', axis=axis, inplace=inplace,
                           limit=limit, downcast=downcast)

    def replace(self, to_replace=None, value=None, inplace=False, limit=None,
                regex=False, method='pad', axis=None):
        """
        Replace values given in 'to_replace' with 'value'.

        Parameters
        ----------
        to_replace : str, regex, list, dict, Series, numeric, or None

            * str or regex:

                - str: string exactly matching `to_replace` will be replaced
                  with `value`
                - regex: regexs matching `to_replace` will be replaced with
                  `value`

            * list of str, regex, or numeric:

                - First, if `to_replace` and `value` are both lists, they
                  **must** be the same length.
                - Second, if ``regex=True`` then all of the strings in **both**
                  lists will be interpreted as regexs otherwise they will match
                  directly. This doesn't matter much for `value` since there
                  are only a few possible substitution regexes you can use.
                - str and regex rules apply as above.

            * dict:

                - Nested dictionaries, e.g., {'a': {'b': nan}}, are read as
                  follows: look in column 'a' for the value 'b' and replace it
                  with nan. You can nest regular expressions as well. Note that
                  column names (the top-level dictionary keys in a nested
                  dictionary) **cannot** be regular expressions.
                - Keys map to column names and values map to substitution
                  values. You can treat this as a special case of passing two
                  lists except that you are specifying the column to search in.

            * None:

                - This means that the ``regex`` argument must be a string,
                  compiled regular expression, or list, dict, ndarray or Series
                  of such elements. If `value` is also ``None`` then this
                  **must** be a nested dictionary or ``Series``.

            See the examples section for examples of each of these.
        value : scalar, dict, list, str, regex, default None
            Value to use to fill holes (e.g. 0), alternately a dict of values
            specifying which value to use for each column (columns not in the
            dict will not be filled). Regular expressions, strings and lists or
            dicts of such objects are also allowed.
        inplace : boolean, default False
            If True, in place. Note: this will modify any
            other views on this object (e.g. a column form a DataFrame).
            Returns the caller if this is True.
        limit : int, default None
            Maximum size gap to forward or backward fill
        regex : bool or same types as `to_replace`, default False
            Whether to interpret `to_replace` and/or `value` as regular
            expressions. If this is ``True`` then `to_replace` *must* be a
            string. Otherwise, `to_replace` must be ``None`` because this
            parameter will be interpreted as a regular expression or a list,
            dict, or array of regular expressions.
        method : string, optional, {'pad', 'ffill', 'bfill'}
            The method to use when for replacement, when ``to_replace`` is a
            ``list``.

        See Also
        --------
        NDFrame.reindex
        NDFrame.asfreq
        NDFrame.fillna

        Returns
        -------
        filled : NDFrame

        Raises
        ------
        AssertionError
            * If `regex` is not a ``bool`` and `to_replace` is not ``None``.
        TypeError
            * If `to_replace` is a ``dict`` and `value` is not a ``list``,
              ``dict``, ``ndarray``, or ``Series``
            * If `to_replace` is ``None`` and `regex` is not compilable into a
              regular expression or is a list, dict, ndarray, or Series.
        ValueError
            * If `to_replace` and `value` are ``list`` s or ``ndarray`` s, but
              they are not the same length.

        Notes
        -----
        * Regex substitution is performed under the hood with ``re.sub``. The
          rules for substitution for ``re.sub`` are the same.
        * Regular expressions will only substitute on strings, meaning you
          cannot provide, for example, a regular expression matching floating
          point numbers and expect the columns in your frame that have a
          numeric dtype to be matched. However, if those floating point numbers
          *are* strings, then you can do this.
        * This method has *a lot* of options. You are encouraged to experiment
          and play with this method to gain intuition about how it works.

        """
        inplace = validate_bool_kwarg(inplace, 'inplace')
        if not is_bool(regex) and to_replace is not None:
            raise AssertionError("'to_replace' must be 'None' if 'regex' is "
                                 "not a bool")
        if axis is not None:
            from warnings import warn
            warn('the "axis" argument is deprecated and will be removed in'
                 'v0.13; this argument has no effect')

        self._consolidate_inplace()

        if value is None:
            # passing a single value that is scalar like
            # when value is None (GH5319), for compat
            if not is_dict_like(to_replace) and not is_dict_like(regex):
                to_replace = [to_replace]

            if isinstance(to_replace, (tuple, list)):
                return _single_replace(self, to_replace, method, inplace,
                                       limit)

            if not is_dict_like(to_replace):
                if not is_dict_like(regex):
                    raise TypeError('If "to_replace" and "value" are both None'
                                    ' and "to_replace" is not a list, then '
                                    'regex must be a mapping')
                to_replace = regex
                regex = True

            items = list(compat.iteritems(to_replace))
            keys, values = lzip(*items) or ([], [])

            are_mappings = [is_dict_like(v) for v in values]

            if any(are_mappings):
                if not all(are_mappings):
                    raise TypeError("If a nested mapping is passed, all values"
                                    " of the top level mapping must be "
                                    "mappings")
                # passed a nested dict/Series
                to_rep_dict = {}
                value_dict = {}

                for k, v in items:
                    keys, values = lzip(*v.items()) or ([], [])
                    if set(keys) & set(values):
                        raise ValueError("Replacement not allowed with "
                                         "overlapping keys and values")
                    to_rep_dict[k] = list(keys)
                    value_dict[k] = list(values)

                to_replace, value = to_rep_dict, value_dict
            else:
                to_replace, value = keys, values

            return self.replace(to_replace, value, inplace=inplace,
                                limit=limit, regex=regex)
        else:

            # need a non-zero len on all axes
            for a in self._AXIS_ORDERS:
                if not len(self._get_axis(a)):
                    return self

            new_data = self._data
            if is_dict_like(to_replace):
                if is_dict_like(value):  # {'A' : NA} -> {'A' : 0}
                    res = self if inplace else self.copy()
                    for c, src in compat.iteritems(to_replace):
                        if c in value and c in self:
                            # object conversion is handled in
                            # series.replace which is called recursivelly
                            res[c] = res[c].replace(to_replace=src,
                                                    value=value[c],
                                                    inplace=False,
                                                    regex=regex)
                    return None if inplace else res

                # {'A': NA} -> 0
                elif not is_list_like(value):
                    keys = [(k, src) for k, src in compat.iteritems(to_replace)
                            if k in self]
                    keys_len = len(keys) - 1
                    for i, (k, src) in enumerate(keys):
                        convert = i == keys_len
                        new_data = new_data.replace(to_replace=src,
                                                    value=value,
                                                    filter=[k],
                                                    inplace=inplace,
                                                    regex=regex,
                                                    convert=convert)
                else:
                    raise TypeError('value argument must be scalar, dict, or '
                                    'Series')

            elif is_list_like(to_replace):  # [NA, ''] -> [0, 'missing']
                if is_list_like(value):
                    if len(to_replace) != len(value):
                        raise ValueError('Replacement lists must match '
                                         'in length. Expecting %d got %d ' %
                                         (len(to_replace), len(value)))

                    new_data = self._data.replace_list(src_list=to_replace,
                                                       dest_list=value,
                                                       inplace=inplace,
                                                       regex=regex)

                else:  # [NA, ''] -> 0
                    new_data = self._data.replace(to_replace=to_replace,
                                                  value=value, inplace=inplace,
                                                  regex=regex)
            elif to_replace is None:
                if not (is_re_compilable(regex) or
                        is_list_like(regex) or is_dict_like(regex)):
                    raise TypeError("'regex' must be a string or a compiled "
                                    "regular expression or a list or dict of "
                                    "strings or regular expressions, you "
                                    "passed a"
                                    " {0!r}".format(type(regex).__name__))
                return self.replace(regex, value, inplace=inplace, limit=limit,
                                    regex=True)
            else:

                # dest iterable dict-like
                if is_dict_like(value):  # NA -> {'A' : 0, 'B' : -1}
                    new_data = self._data

                    for k, v in compat.iteritems(value):
                        if k in self:
                            new_data = new_data.replace(to_replace=to_replace,
                                                        value=v, filter=[k],
                                                        inplace=inplace,
                                                        regex=regex)

                elif not is_list_like(value):  # NA -> 0
                    new_data = self._data.replace(to_replace=to_replace,
                                                  value=value, inplace=inplace,
                                                  regex=regex)
                else:
                    msg = ('Invalid "to_replace" type: '
                           '{0!r}').format(type(to_replace).__name__)
                    raise TypeError(msg)  # pragma: no cover

        if inplace:
            self._update_inplace(new_data)
        else:
            return self._constructor(new_data).__finalize__(self)

    _shared_docs['interpolate'] = """
        Please note that only ``method='linear'`` is supported for
        DataFrames/Series with a MultiIndex.

        Parameters
        ----------
        method : {'linear', 'time', 'index', 'values', 'nearest', 'zero',
                  'slinear', 'quadratic', 'cubic', 'barycentric', 'krogh',
                  'polynomial', 'spline', 'piecewise_polynomial',
                  'from_derivatives', 'pchip', 'akima'}

            * 'linear': ignore the index and treat the values as equally
              spaced. This is the only method supported on MultiIndexes.
              default
            * 'time': interpolation works on daily and higher resolution
              data to interpolate given length of interval
            * 'index', 'values': use the actual numerical values of the index
            * 'nearest', 'zero', 'slinear', 'quadratic', 'cubic',
              'barycentric', 'polynomial' is passed to
              ``scipy.interpolate.interp1d``. Both 'polynomial' and 'spline'
              require that you also specify an `order` (int),
              e.g. df.interpolate(method='polynomial', order=4).
              These use the actual numerical values of the index.
            * 'krogh', 'piecewise_polynomial', 'spline', 'pchip' and 'akima'
              are all wrappers around the scipy interpolation methods of
              similar names. These use the actual numerical values of the
              index. For more information on their behavior, see the
              `scipy documentation
              <http://docs.scipy.org/doc/scipy/reference/interpolate.html#univariate-interpolation>`__
              and `tutorial documentation
              <http://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html>`__
            * 'from_derivatives' refers to BPoly.from_derivatives which
              replaces 'piecewise_polynomial' interpolation method in
              scipy 0.18

            .. versionadded:: 0.18.1

               Added support for the 'akima' method
               Added interpolate method 'from_derivatives' which replaces
               'piecewise_polynomial' in scipy 0.18; backwards-compatible with
               scipy < 0.18

        axis : {0, 1}, default 0
            * 0: fill column-by-column
            * 1: fill row-by-row
        limit : int, default None.
            Maximum number of consecutive NaNs to fill. Must be greater than 0.
        limit_direction : {'forward', 'backward', 'both'}, default 'forward'
            If limit is specified, consecutive NaNs will be filled in this
            direction.

            .. versionadded:: 0.17.0

        inplace : bool, default False
            Update the NDFrame in place if possible.
        downcast : optional, 'infer' or None, defaults to None
            Downcast dtypes if possible.
        kwargs : keyword arguments to pass on to the interpolating function.

        Returns
        -------
        Series or DataFrame of same shape interpolated at the NaNs

        See Also
        --------
        reindex, replace, fillna

        Examples
        --------

        Filling in NaNs

        >>> s = pd.Series([0, 1, np.nan, 3])
        >>> s.interpolate()
        0    0
        1    1
        2    2
        3    3
        dtype: float64

        """

    @Appender(_shared_docs['interpolate'] % _shared_doc_kwargs)
    def interpolate(self, method='linear', axis=0, limit=None, inplace=False,
                    limit_direction='forward', downcast=None, **kwargs):
        """
        Interpolate values according to different methods.
        """
        inplace = validate_bool_kwarg(inplace, 'inplace')

        if self.ndim > 2:
            raise NotImplementedError("Interpolate has not been implemented "
                                      "on Panel and Panel 4D objects.")

        if axis == 0:
            ax = self._info_axis_name
            _maybe_transposed_self = self
        elif axis == 1:
            _maybe_transposed_self = self.T
            ax = 1
        else:
            _maybe_transposed_self = self
        ax = _maybe_transposed_self._get_axis_number(ax)

        if _maybe_transposed_self.ndim == 2:
            alt_ax = 1 - ax
        else:
            alt_ax = ax

        if (isinstance(_maybe_transposed_self.index, MultiIndex) and
                method != 'linear'):
            raise ValueError("Only `method=linear` interpolation is supported "
                             "on MultiIndexes.")

        if _maybe_transposed_self._data.get_dtype_counts().get(
                'object') == len(_maybe_transposed_self.T):
            raise TypeError("Cannot interpolate with all NaNs.")

        # create/use the index
        if method == 'linear':
            # prior default
            index = np.arange(len(_maybe_transposed_self._get_axis(alt_ax)))
        else:
            index = _maybe_transposed_self._get_axis(alt_ax)

        if pd.isnull(index).any():
            raise NotImplementedError("Interpolation with NaNs in the index "
                                      "has not been implemented. Try filling "
                                      "those NaNs before interpolating.")
        data = _maybe_transposed_self._data
        new_data = data.interpolate(method=method, axis=ax, index=index,
                                    values=_maybe_transposed_self, limit=limit,
                                    limit_direction=limit_direction,
                                    inplace=inplace, downcast=downcast,
                                    **kwargs)

        if inplace:
            if axis == 1:
                new_data = self._constructor(new_data).T._data
            self._update_inplace(new_data)
        else:
            res = self._constructor(new_data).__finalize__(self)
            if axis == 1:
                res = res.T
            return res

    # ----------------------------------------------------------------------
    # Timeseries methods Methods

    def asof(self, where, subset=None):
        """
        The last row without any NaN is taken (or the last row without
        NaN considering only the subset of columns in the case of a DataFrame)

        .. versionadded:: 0.19.0 For DataFrame

        If there is no good value, NaN is returned for a Series
        a Series of NaN values for a DataFrame

        Parameters
        ----------
        where : date or array of dates
        subset : string or list of strings, default None
           if not None use these columns for NaN propagation

        Notes
        -----
        Dates are assumed to be sorted
        Raises if this is not the case

        Returns
        -------
        where is scalar

          - value or NaN if input is Series
          - Series if input is DataFrame

        where is Index: same shape object as input

        See Also
        --------
        merge_asof

        """

        if isinstance(where, compat.string_types):
            from pandas import to_datetime
            where = to_datetime(where)

        if not self.index.is_monotonic:
            raise ValueError("asof requires a sorted index")

        is_series = isinstance(self, ABCSeries)
        if is_series:
            if subset is not None:
                raise ValueError("subset is not valid for Series")
        elif self.ndim > 2:
            raise NotImplementedError("asof is not implemented "
                                      "for {type}".format(type(self)))
        else:
            if subset is None:
                subset = self.columns
            if not is_list_like(subset):
                subset = [subset]

        is_list = is_list_like(where)
        if not is_list:
            start = self.index[0]
            if isinstance(self.index, PeriodIndex):
                where = Period(where, freq=self.index.freq).ordinal
                start = start.ordinal

            if where < start:
                if not is_series:
                    from pandas import Series
                    return Series(index=self.columns, name=where)
                return np.nan

            # It's always much faster to use a *while* loop here for
            # Series than pre-computing all the NAs. However a
            # *while* loop is extremely expensive for DataFrame
            # so we later pre-compute all the NAs and use the same
            # code path whether *where* is a scalar or list.
            # See PR: https://github.com/pandas-dev/pandas/pull/14476
            if is_series:
                loc = self.index.searchsorted(where, side='right')
                if loc > 0:
                    loc -= 1

                values = self._values
                while loc > 0 and isnull(values[loc]):
                    loc -= 1
                return values[loc]

        if not isinstance(where, Index):
            where = Index(where) if is_list else Index([where])

        nulls = self.isnull() if is_series else self[subset].isnull().any(1)
        if nulls.all():
            if is_series:
                return self._constructor(np.nan, index=where, name=self.name)
            elif is_list:
                from pandas import DataFrame
                return DataFrame(np.nan, index=where, columns=self.columns)
            else:
                from pandas import Series
                return Series(np.nan, index=self.columns, name=where[0])

        locs = self.index.asof_locs(where, ~(nulls.values))

        # mask the missing
        missing = locs == -1
        data = self.take(locs, is_copy=False)
        data.index = where
        data.loc[missing] = np.nan
        return data if is_list else data.iloc[-1]

    # ----------------------------------------------------------------------
    # Action Methods

    _shared_docs['isnull'] = """
        Return a boolean same-sized object indicating if the values are null.

        See Also
        --------
        notnull : boolean inverse of isnull
        """

    @Appender(_shared_docs['isnull'])
    def isnull(self):
        return isnull(self).__finalize__(self)

    _shared_docs['isnotnull'] = """
        Return a boolean same-sized object indicating if the values are
        not null.

        See Also
        --------
        isnull : boolean inverse of notnull
        """

    @Appender(_shared_docs['isnotnull'])
    def notnull(self):
        return notnull(self).__finalize__(self)

    def clip(self, lower=None, upper=None, axis=None, *args, **kwargs):
        """
        Trim values at input threshold(s).

        Parameters
        ----------
        lower : float or array_like, default None
        upper : float or array_like, default None
        axis : int or string axis name, optional
            Align object with lower and upper along the given axis.

        Returns
        -------
        clipped : Series

        Examples
        --------
        >>> df
          0         1
        0  0.335232 -1.256177
        1 -1.367855  0.746646
        2  0.027753 -1.176076
        3  0.230930 -0.679613
        4  1.261967  0.570967
        >>> df.clip(-1.0, 0.5)
                  0         1
        0  0.335232 -1.000000
        1 -1.000000  0.500000
        2  0.027753 -1.000000
        3  0.230930 -0.679613
        4  0.500000  0.500000
        >>> t
        0   -0.3
        1   -0.2
        2   -0.1
        3    0.0
        4    0.1
        dtype: float64
        >>> df.clip(t, t + 1, axis=0)
                  0         1
        0  0.335232 -0.300000
        1 -0.200000  0.746646
        2  0.027753 -0.100000
        3  0.230930  0.000000
        4  1.100000  0.570967
        """
        if isinstance(self, ABCPanel):
            raise NotImplementedError("clip is not supported yet for panels")

        axis = nv.validate_clip_with_axis(axis, args, kwargs)

        # GH 2747 (arguments were reversed)
        if lower is not None and upper is not None:
            if is_scalar(lower) and is_scalar(upper):
                lower, upper = min(lower, upper), max(lower, upper)

        result = self
        if lower is not None:
            result = result.clip_lower(lower, axis)
        if upper is not None:
            result = result.clip_upper(upper, axis)

        return result

    def clip_upper(self, threshold, axis=None):
        """
        Return copy of input with values above given value(s) truncated.

        Parameters
        ----------
        threshold : float or array_like
        axis : int or string axis name, optional
            Align object with threshold along the given axis.

        See Also
        --------
        clip

        Returns
        -------
        clipped : same type as input
        """
        if np.any(isnull(threshold)):
            raise ValueError("Cannot use an NA value as a clip threshold")

        subset = self.le(threshold, axis=axis) | isnull(self)
        return self.where(subset, threshold, axis=axis)

    def clip_lower(self, threshold, axis=None):
        """
        Return copy of the input with values below given value(s) truncated.

        Parameters
        ----------
        threshold : float or array_like
        axis : int or string axis name, optional
            Align object with threshold along the given axis.

        See Also
        --------
        clip

        Returns
        -------
        clipped : same type as input
        """
        if np.any(isnull(threshold)):
            raise ValueError("Cannot use an NA value as a clip threshold")

        subset = self.ge(threshold, axis=axis) | isnull(self)
        return self.where(subset, threshold, axis=axis)

    def groupby(self, by=None, axis=0, level=None, as_index=True, sort=True,
                group_keys=True, squeeze=False, **kwargs):
        """
        Group series using mapper (dict or key function, apply given function
        to group, return result as series) or by a series of columns.

        Parameters
        ----------
        by : mapping, function, str, or iterable
            Used to determine the groups for the groupby.
            If ``by`` is a function, it's called on each value of the object's
            index. If a dict or Series is passed, the Series or dict VALUES
            will be used to determine the groups (the Series' values are first
            aligned; see ``.align()`` method). If an ndarray is passed, the
            values are used as-is determine the groups. A str or list of strs
            may be passed to group by the columns in ``self``
        axis : int, default 0
        level : int, level name, or sequence of such, default None
            If the axis is a MultiIndex (hierarchical), group by a particular
            level or levels
        as_index : boolean, default True
            For aggregated output, return object with group labels as the
            index. Only relevant for DataFrame input. as_index=False is
            effectively "SQL-style" grouped output
        sort : boolean, default True
            Sort group keys. Get better performance by turning this off.
            Note this does not influence the order of observations within each
            group.  groupby preserves the order of rows within each group.
        group_keys : boolean, default True
            When calling apply, add group keys to index to identify pieces
        squeeze : boolean, default False
            reduce the dimensionality of the return type if possible,
            otherwise return a consistent type

        Examples
        --------
        DataFrame results

        >>> data.groupby(func, axis=0).mean()
        >>> data.groupby(['col1', 'col2'])['col3'].mean()

        DataFrame with hierarchical index

        >>> data.groupby(['col1', 'col2']).mean()

        Returns
        -------
        GroupBy object

        """
        from pandas.core.groupby import groupby

        if level is None and by is None:
            raise TypeError("You have to supply one of 'by' and 'level'")
        axis = self._get_axis_number(axis)
        return groupby(self, by=by, axis=axis, level=level, as_index=as_index,
                       sort=sort, group_keys=group_keys, squeeze=squeeze,
                       **kwargs)

    def asfreq(self, freq, method=None, how=None, normalize=False,
               fill_value=None):
        """
        Convert TimeSeries to specified frequency.

        Optionally provide filling method to pad/backfill missing values.

        Returns the original data conformed to a new index with the specified
        frequency. ``resample`` is more appropriate if an operation, such as
        summarization, is necessary to represent the data at the new frequency.

        Parameters
        ----------
        freq : DateOffset object, or string
        method : {'backfill'/'bfill', 'pad'/'ffill'}, default None
            Method to use for filling holes in reindexed Series (note this
            does not fill NaNs that already were present):

            * 'pad' / 'ffill': propagate last valid observation forward to next
              valid
            * 'backfill' / 'bfill': use NEXT valid observation to fill
        how : {'start', 'end'}, default end
            For PeriodIndex only, see PeriodIndex.asfreq
        normalize : bool, default False
            Whether to reset output index to midnight
        fill_value: scalar, optional
            Value to use for missing values, applied during upsampling (note
            this does not fill NaNs that already were present).

            .. versionadded:: 0.20.0

        Returns
        -------
        converted : type of caller

        Examples
        --------

        Start by creating a series with 4 one minute timestamps.

        >>> index = pd.date_range('1/1/2000', periods=4, freq='T')
        >>> series = pd.Series([0.0, None, 2.0, 3.0], index=index)
        >>> df = pd.DataFrame({'s':series})
        >>> df
                               s
        2000-01-01 00:00:00    0.0
        2000-01-01 00:01:00    NaN
        2000-01-01 00:02:00    2.0
        2000-01-01 00:03:00    3.0

        Upsample the series into 30 second bins.

        >>> df.asfreq(freq='30S')
                               s
        2000-01-01 00:00:00    0.0
        2000-01-01 00:00:30    NaN
        2000-01-01 00:01:00    NaN
        2000-01-01 00:01:30    NaN
        2000-01-01 00:02:00    2.0
        2000-01-01 00:02:30    NaN
        2000-01-01 00:03:00    3.0

        Upsample again, providing a ``fill value``.

        >>> df.asfreq(freq='30S', fill_value=9.0)
                               s
        2000-01-01 00:00:00    0.0
        2000-01-01 00:00:30    9.0
        2000-01-01 00:01:00    NaN
        2000-01-01 00:01:30    9.0
        2000-01-01 00:02:00    2.0
        2000-01-01 00:02:30    9.0
        2000-01-01 00:03:00    3.0

        Upsample again, providing a ``method``.

        >>> df.asfreq(freq='30S', method='bfill')
                               s
        2000-01-01 00:00:00    0.0
        2000-01-01 00:00:30    NaN
        2000-01-01 00:01:00    NaN
        2000-01-01 00:01:30    2.0
        2000-01-01 00:02:00    2.0
        2000-01-01 00:02:30    3.0
        2000-01-01 00:03:00    3.0

        See Also
        --------
        reindex

        Notes
        -----
        To learn more about the frequency strings, please see `this link
        <http://pandas.pydata.org/pandas-docs/stable/timeseries.html#offset-aliases>`__.
        """
        from pandas.core.resample import asfreq
        return asfreq(self, freq, method=method, how=how, normalize=normalize,
                      fill_value=fill_value)

    def at_time(self, time, asof=False):
        """
        Select values at particular time of day (e.g. 9:30AM).

        Parameters
        ----------
        time : datetime.time or string

        Returns
        -------
        values_at_time : type of caller
        """
        try:
            indexer = self.index.indexer_at_time(time, asof=asof)
            return self.take(indexer, convert=False)
        except AttributeError:
            raise TypeError('Index must be DatetimeIndex')

    def between_time(self, start_time, end_time, include_start=True,
                     include_end=True):
        """
        Select values between particular times of the day (e.g., 9:00-9:30 AM).

        Parameters
        ----------
        start_time : datetime.time or string
        end_time : datetime.time or string
        include_start : boolean, default True
        include_end : boolean, default True

        Returns
        -------
        values_between_time : type of caller
        """
        try:
            indexer = self.index.indexer_between_time(
                start_time, end_time, include_start=include_start,
                include_end=include_end)
            return self.take(indexer, convert=False)
        except AttributeError:
            raise TypeError('Index must be DatetimeIndex')

    def resample(self, rule, how=None, axis=0, fill_method=None, closed=None,
                 label=None, convention='start', kind=None, loffset=None,
                 limit=None, base=0, on=None, level=None):
        """
        Convenience method for frequency conversion and resampling of time
        series.  Object must have a datetime-like index (DatetimeIndex,
        PeriodIndex, or TimedeltaIndex), or pass datetime-like values
        to the on or level keyword.

        Parameters
        ----------
        rule : string
            the offset string or object representing target conversion
        axis : int, optional, default 0
        closed : {'right', 'left'}
            Which side of bin interval is closed
        label : {'right', 'left'}
            Which bin edge label to label bucket with
        convention : {'start', 'end', 's', 'e'}
        loffset : timedelta
            Adjust the resampled time labels
        base : int, default 0
            For frequencies that evenly subdivide 1 day, the "origin" of the
            aggregated intervals. For example, for '5min' frequency, base could
            range from 0 through 4. Defaults to 0
        on : string, optional
            For a DataFrame, column to use instead of index for resampling.
            Column must be datetime-like.

            .. versionadded:: 0.19.0

        level : string or int, optional
            For a MultiIndex, level (name or number) to use for
            resampling.  Level must be datetime-like.

            .. versionadded:: 0.19.0

        Notes
        -----
        To learn more about the offset strings, please see `this link
        <http://pandas.pydata.org/pandas-docs/stable/timeseries.html#offset-aliases>`__.

        Examples
        --------

        Start by creating a series with 9 one minute timestamps.

        >>> index = pd.date_range('1/1/2000', periods=9, freq='T')
        >>> series = pd.Series(range(9), index=index)
        >>> series
        2000-01-01 00:00:00    0
        2000-01-01 00:01:00    1
        2000-01-01 00:02:00    2
        2000-01-01 00:03:00    3
        2000-01-01 00:04:00    4
        2000-01-01 00:05:00    5
        2000-01-01 00:06:00    6
        2000-01-01 00:07:00    7
        2000-01-01 00:08:00    8
        Freq: T, dtype: int64

        Downsample the series into 3 minute bins and sum the values
        of the timestamps falling into a bin.

        >>> series.resample('3T').sum()
        2000-01-01 00:00:00     3
        2000-01-01 00:03:00    12
        2000-01-01 00:06:00    21
        Freq: 3T, dtype: int64

        Downsample the series into 3 minute bins as above, but label each
        bin using the right edge instead of the left. Please note that the
        value in the bucket used as the label is not included in the bucket,
        which it labels. For example, in the original series the
        bucket ``2000-01-01 00:03:00`` contains the value 3, but the summed
        value in the resampled bucket with the label``2000-01-01 00:03:00``
        does not include 3 (if it did, the summed value would be 6, not 3).
        To include this value close the right side of the bin interval as
        illustrated in the example below this one.

        >>> series.resample('3T', label='right').sum()
        2000-01-01 00:03:00     3
        2000-01-01 00:06:00    12
        2000-01-01 00:09:00    21
        Freq: 3T, dtype: int64

        Downsample the series into 3 minute bins as above, but close the right
        side of the bin interval.

        >>> series.resample('3T', label='right', closed='right').sum()
        2000-01-01 00:00:00     0
        2000-01-01 00:03:00     6
        2000-01-01 00:06:00    15
        2000-01-01 00:09:00    15
        Freq: 3T, dtype: int64

        Upsample the series into 30 second bins.

        >>> series.resample('30S').asfreq()[0:5] #select first 5 rows
        2000-01-01 00:00:00   0.0
        2000-01-01 00:00:30   NaN
        2000-01-01 00:01:00   1.0
        2000-01-01 00:01:30   NaN
        2000-01-01 00:02:00   2.0
        Freq: 30S, dtype: float64

        Upsample the series into 30 second bins and fill the ``NaN``
        values using the ``pad`` method.

        >>> series.resample('30S').pad()[0:5]
        2000-01-01 00:00:00    0
        2000-01-01 00:00:30    0
        2000-01-01 00:01:00    1
        2000-01-01 00:01:30    1
        2000-01-01 00:02:00    2
        Freq: 30S, dtype: int64

        Upsample the series into 30 second bins and fill the
        ``NaN`` values using the ``bfill`` method.

        >>> series.resample('30S').bfill()[0:5]
        2000-01-01 00:00:00    0
        2000-01-01 00:00:30    1
        2000-01-01 00:01:00    1
        2000-01-01 00:01:30    2
        2000-01-01 00:02:00    2
        Freq: 30S, dtype: int64

        Pass a custom function via ``apply``

        >>> def custom_resampler(array_like):
        ...     return np.sum(array_like)+5

        >>> series.resample('3T').apply(custom_resampler)
        2000-01-01 00:00:00     8
        2000-01-01 00:03:00    17
        2000-01-01 00:06:00    26
        Freq: 3T, dtype: int64

        For DataFrame objects, the keyword ``on`` can be used to specify the
        column instead of the index for resampling.

        >>> df = pd.DataFrame(data=9*[range(4)], columns=['a', 'b', 'c', 'd'])
        >>> df['time'] = pd.date_range('1/1/2000', periods=9, freq='T')
        >>> df.resample('3T', on='time').sum()
                             a  b  c  d
        time
        2000-01-01 00:00:00  0  3  6  9
        2000-01-01 00:03:00  0  3  6  9
        2000-01-01 00:06:00  0  3  6  9

        For a DataFrame with MultiIndex, the keyword ``level`` can be used to
        specify on level the resampling needs to take place.

        >>> time = pd.date_range('1/1/2000', periods=5, freq='T')
        >>> df2 = pd.DataFrame(data=10*[range(4)],
                               columns=['a', 'b', 'c', 'd'],
                               index=pd.MultiIndex.from_product([time, [1, 2]])
                               )
        >>> df2.resample('3T', level=0).sum()
                             a  b   c   d
        2000-01-01 00:00:00  0  6  12  18
        2000-01-01 00:03:00  0  4   8  12
        """
        from pandas.core.resample import (resample,
                                          _maybe_process_deprecations)
        axis = self._get_axis_number(axis)
        r = resample(self, freq=rule, label=label, closed=closed,
                     axis=axis, kind=kind, loffset=loffset,
                     convention=convention,
                     base=base, key=on, level=level)
        return _maybe_process_deprecations(r,
                                           how=how,
                                           fill_method=fill_method,
                                           limit=limit)

    def first(self, offset):
        """
        Convenience method for subsetting initial periods of time series data
        based on a date offset.

        Parameters
        ----------
        offset : string, DateOffset, dateutil.relativedelta

        Examples
        --------
        ts.first('10D') -> First 10 days

        Returns
        -------
        subset : type of caller
        """
        from pandas.tseries.frequencies import to_offset
        if not isinstance(self.index, DatetimeIndex):
            raise NotImplementedError("'first' only supports a DatetimeIndex "
                                      "index")

        if len(self.index) == 0:
            return self

        offset = to_offset(offset)
        end_date = end = self.index[0] + offset

        # Tick-like, e.g. 3 weeks
        if not offset.isAnchored() and hasattr(offset, '_inc'):
            if end_date in self.index:
                end = self.index.searchsorted(end_date, side='left')
                return self.iloc[:end]

        return self.loc[:end]

    def last(self, offset):
        """
        Convenience method for subsetting final periods of time series data
        based on a date offset.

        Parameters
        ----------
        offset : string, DateOffset, dateutil.relativedelta

        Examples
        --------
        ts.last('5M') -> Last 5 months

        Returns
        -------
        subset : type of caller
        """
        from pandas.tseries.frequencies import to_offset
        if not isinstance(self.index, DatetimeIndex):
            raise NotImplementedError("'last' only supports a DatetimeIndex "
                                      "index")

        if len(self.index) == 0:
            return self

        offset = to_offset(offset)

        start_date = start = self.index[-1] - offset
        start = self.index.searchsorted(start_date, side='right')
        return self.iloc[start:]

    def rank(self, axis=0, method='average', numeric_only=None,
             na_option='keep', ascending=True, pct=False):
        """
        Compute numerical data ranks (1 through n) along axis. Equal values are
        assigned a rank that is the average of the ranks of those values

        Parameters
        ----------
        axis : {0 or 'index', 1 or 'columns'}, default 0
            index to direct ranking
        method : {'average', 'min', 'max', 'first', 'dense'}
            * average: average rank of group
            * min: lowest rank in group
            * max: highest rank in group
            * first: ranks assigned in order they appear in the array
            * dense: like 'min', but rank always increases by 1 between groups
        numeric_only : boolean, default None
            Include only float, int, boolean data. Valid only for DataFrame or
            Panel objects
        na_option : {'keep', 'top', 'bottom'}
            * keep: leave NA values where they are
            * top: smallest rank if ascending
            * bottom: smallest rank if descending
        ascending : boolean, default True
            False for ranks by high (1) to low (N)
        pct : boolean, default False
            Computes percentage rank of data

        Returns
        -------
        ranks : same type as caller
        """
        axis = self._get_axis_number(axis)

        if self.ndim > 2:
            msg = "rank does not make sense when ndim > 2"
            raise NotImplementedError(msg)

        def ranker(data):
            ranks = algos.rank(data.values, axis=axis, method=method,
                               ascending=ascending, na_option=na_option,
                               pct=pct)
            ranks = self._constructor(ranks, **data._construct_axes_dict())
            return ranks.__finalize__(self)

        # if numeric_only is None, and we can't get anything, we try with
        # numeric_only=True
        if numeric_only is None:
            try:
                return ranker(self)
            except TypeError:
                numeric_only = True

        if numeric_only:
            data = self._get_numeric_data()
        else:
            data = self

        return ranker(data)

    _shared_docs['align'] = ("""
        Align two object on their axes with the
        specified join method for each axis Index

        Parameters
        ----------
        other : DataFrame or Series
        join : {'outer', 'inner', 'left', 'right'}, default 'outer'
        axis : allowed axis of the other object, default None
            Align on index (0), columns (1), or both (None)
        level : int or level name, default None
            Broadcast across a level, matching Index values on the
            passed MultiIndex level
        copy : boolean, default True
            Always returns new objects. If copy=False and no reindexing is
            required then original objects are returned.
        fill_value : scalar, default np.NaN
            Value to use for missing values. Defaults to NaN, but can be any
            "compatible" value
        method : str, default None
        limit : int, default None
        fill_axis : %(axes_single_arg)s, default 0
            Filling axis, method and limit
        broadcast_axis : %(axes_single_arg)s, default None
            Broadcast values along this axis, if aligning two objects of
            different dimensions

            .. versionadded:: 0.17.0

        Returns
        -------
        (left, right) : (%(klass)s, type of other)
            Aligned objects
        """)

    @Appender(_shared_docs['align'] % _shared_doc_kwargs)
    def align(self, other, join='outer', axis=None, level=None, copy=True,
              fill_value=None, method=None, limit=None, fill_axis=0,
              broadcast_axis=None):
        from pandas import DataFrame, Series
        method = missing.clean_fill_method(method)

        if broadcast_axis == 1 and self.ndim != other.ndim:
            if isinstance(self, Series):
                # this means other is a DataFrame, and we need to broadcast
                # self
                cons = self._constructor_expanddim
                df = cons(dict((c, self) for c in other.columns),
                          **other._construct_axes_dict())
                return df._align_frame(other, join=join, axis=axis,
                                       level=level, copy=copy,
                                       fill_value=fill_value, method=method,
                                       limit=limit, fill_axis=fill_axis)
            elif isinstance(other, Series):
                # this means self is a DataFrame, and we need to broadcast
                # other
                cons = other._constructor_expanddim
                df = cons(dict((c, other) for c in self.columns),
                          **self._construct_axes_dict())
                return self._align_frame(df, join=join, axis=axis, level=level,
                                         copy=copy, fill_value=fill_value,
                                         method=method, limit=limit,
                                         fill_axis=fill_axis)

        if axis is not None:
            axis = self._get_axis_number(axis)
        if isinstance(other, DataFrame):
            return self._align_frame(other, join=join, axis=axis, level=level,
                                     copy=copy, fill_value=fill_value,
                                     method=method, limit=limit,
                                     fill_axis=fill_axis)
        elif isinstance(other, Series):
            return self._align_series(other, join=join, axis=axis, level=level,
                                      copy=copy, fill_value=fill_value,
                                      method=method, limit=limit,
                                      fill_axis=fill_axis)
        else:  # pragma: no cover
            raise TypeError('unsupported type: %s' % type(other))

    def _align_frame(self, other, join='outer', axis=None, level=None,
                     copy=True, fill_value=np.nan, method=None, limit=None,
                     fill_axis=0):
        # defaults
        join_index, join_columns = None, None
        ilidx, iridx = None, None
        clidx, cridx = None, None

        is_series = isinstance(self, ABCSeries)

        if axis is None or axis == 0:
            if not self.index.equals(other.index):
                join_index, ilidx, iridx = self.index.join(
                    other.index, how=join, level=level, return_indexers=True)

        if axis is None or axis == 1:
            if not is_series and not self.columns.equals(other.columns):
                join_columns, clidx, cridx = self.columns.join(
                    other.columns, how=join, level=level, return_indexers=True)

        if is_series:
            reindexers = {0: [join_index, ilidx]}
        else:
            reindexers = {0: [join_index, ilidx], 1: [join_columns, clidx]}

        left = self._reindex_with_indexers(reindexers, copy=copy,
                                           fill_value=fill_value,
                                           allow_dups=True)
        # other must be always DataFrame
        right = other._reindex_with_indexers({0: [join_index, iridx],
                                              1: [join_columns, cridx]},
                                             copy=copy, fill_value=fill_value,
                                             allow_dups=True)

        if method is not None:
            left = left.fillna(axis=fill_axis, method=method, limit=limit)
            right = right.fillna(axis=fill_axis, method=method, limit=limit)

        # if DatetimeIndex have different tz, convert to UTC
        if is_datetime64tz_dtype(left.index):
            if left.index.tz != right.index.tz:
                if join_index is not None:
                    left.index = join_index
                    right.index = join_index

        return left.__finalize__(self), right.__finalize__(other)

    def _align_series(self, other, join='outer', axis=None, level=None,
                      copy=True, fill_value=None, method=None, limit=None,
                      fill_axis=0):

        is_series = isinstance(self, ABCSeries)

        # series/series compat, other must always be a Series
        if is_series:
            if axis:
                raise ValueError('cannot align series to a series other than '
                                 'axis 0')

            # equal
            if self.index.equals(other.index):
                join_index, lidx, ridx = None, None, None
            else:
                join_index, lidx, ridx = self.index.join(other.index, how=join,
                                                         level=level,
                                                         return_indexers=True)

            left = self._reindex_indexer(join_index, lidx, copy)
            right = other._reindex_indexer(join_index, ridx, copy)

        else:
            # one has > 1 ndim
            fdata = self._data
            if axis == 0:
                join_index = self.index
                lidx, ridx = None, None
                if not self.index.equals(other.index):
                    join_index, lidx, ridx = self.index.join(
                        other.index, how=join, level=level,
                        return_indexers=True)

                if lidx is not None:
                    fdata = fdata.reindex_indexer(join_index, lidx, axis=1)

            elif axis == 1:
                join_index = self.columns
                lidx, ridx = None, None
                if not self.columns.equals(other.index):
                    join_index, lidx, ridx = self.columns.join(
                        other.index, how=join, level=level,
                        return_indexers=True)

                if lidx is not None:
                    fdata = fdata.reindex_indexer(join_index, lidx, axis=0)
            else:
                raise ValueError('Must specify axis=0 or 1')

            if copy and fdata is self._data:
                fdata = fdata.copy()

            left = self._constructor(fdata)

            if ridx is None:
                right = other
            else:
                right = other.reindex(join_index, level=level)

        # fill
        fill_na = notnull(fill_value) or (method is not None)
        if fill_na:
            left = left.fillna(fill_value, method=method, limit=limit,
                               axis=fill_axis)
            right = right.fillna(fill_value, method=method, limit=limit)

        # if DatetimeIndex have different tz, convert to UTC
        if is_series or (not is_series and axis == 0):
            if is_datetime64tz_dtype(left.index):
                if left.index.tz != right.index.tz:
                    if join_index is not None:
                        left.index = join_index
                        right.index = join_index

        return left.__finalize__(self), right.__finalize__(other)

    def _where(self, cond, other=np.nan, inplace=False, axis=None, level=None,
               try_cast=False, raise_on_error=True):
        """
        Equivalent to public method `where`, except that `other` is not
        applied as a function even if callable. Used in __setitem__.
        """
        inplace = validate_bool_kwarg(inplace, 'inplace')

        # align the cond to same shape as myself
        cond = com._apply_if_callable(cond, self)
        if isinstance(cond, NDFrame):
            cond, _ = cond.align(self, join='right', broadcast_axis=1)
        else:
            if not hasattr(cond, 'shape'):
                cond = np.asanyarray(cond)
            if cond.shape != self.shape:
                raise ValueError('Array conditional must be same shape as '
                                 'self')
            cond = self._constructor(cond, **self._construct_axes_dict())

        # make sure we are boolean
        fill_value = True if inplace else False
        cond = cond.fillna(fill_value)

        msg = "Boolean array expected for the condition, not {dtype}"

        if not isinstance(cond, pd.DataFrame):
            # This is a single-dimensional object.
            if not is_bool_dtype(cond):
                raise ValueError(msg.format(dtype=cond.dtype))
        else:
            for dt in cond.dtypes:
                if not is_bool_dtype(dt):
                    raise ValueError(msg.format(dtype=dt))

        cond = cond.astype(bool, copy=False)
        cond = -cond if inplace else cond

        # try to align with other
        try_quick = True
        if hasattr(other, 'align'):

            # align with me
            if other.ndim <= self.ndim:

                _, other = self.align(other, join='left', axis=axis,
                                      level=level, fill_value=np.nan)

                # if we are NOT aligned, raise as we cannot where index
                if (axis is None and
                        not all([other._get_axis(i).equals(ax)
                                 for i, ax in enumerate(self.axes)])):
                    raise InvalidIndexError

            # slice me out of the other
            else:
                raise NotImplemented("cannot align with a higher dimensional "
                                     "NDFrame")

        elif is_list_like(other):

            if self.ndim == 1:

                # try to set the same dtype as ourselves
                try:
                    new_other = np.array(other, dtype=self.dtype)
                except ValueError:
                    new_other = np.array(other)
                except TypeError:
                    new_other = other

                # we can end up comparing integers and m8[ns]
                # which is a numpy no no
                is_i8 = needs_i8_conversion(self.dtype)
                if is_i8:
                    matches = False
                else:
                    matches = (new_other == np.array(other))

                if matches is False or not matches.all():

                    # coerce other to a common dtype if we can
                    if needs_i8_conversion(self.dtype):
                        try:
                            other = np.array(other, dtype=self.dtype)
                        except:
                            other = np.array(other)
                    else:
                        other = np.asarray(other)
                        other = np.asarray(other,
                                           dtype=np.common_type(other,
                                                                new_other))

                    # we need to use the new dtype
                    try_quick = False
                else:
                    other = new_other
            else:

                other = np.array(other)

        if isinstance(other, np.ndarray):

            if other.shape != self.shape:

                if self.ndim == 1:

                    icond = cond.values

                    # GH 2745 / GH 4192
                    # treat like a scalar
                    if len(other) == 1:
                        other = np.array(other[0])

                    # GH 3235
                    # match True cond to other
                    elif len(cond[icond]) == len(other):

                        # try to not change dtype at first (if try_quick)
                        if try_quick:

                            try:
                                new_other = _values_from_object(self).copy()
                                new_other[icond] = other
                                other = new_other
                            except:
                                try_quick = False

                        # let's create a new (if we failed at the above
                        # or not try_quick
                        if not try_quick:

                            dtype, fill_value = maybe_promote(other.dtype)
                            new_other = np.empty(len(icond), dtype=dtype)
                            new_other.fill(fill_value)
                            maybe_upcast_putmask(new_other, icond, other)
                            other = new_other

                    else:
                        raise ValueError('Length of replacements must equal '
                                         'series length')

                else:
                    raise ValueError('other must be the same shape as self '
                                     'when an ndarray')

            # we are the same shape, so create an actual object for alignment
            else:
                other = self._constructor(other, **self._construct_axes_dict())

        if axis is None:
            axis = 0

        if self.ndim == getattr(other, 'ndim', 0):
            align = True
        else:
            align = (self._get_axis_number(axis) == 1)

        block_axis = self._get_block_manager_axis(axis)

        if inplace:
            # we may have different type blocks come out of putmask, so
            # reconstruct the block manager

            self._check_inplace_setting(other)
            new_data = self._data.putmask(mask=cond, new=other, align=align,
                                          inplace=True, axis=block_axis,
                                          transpose=self._AXIS_REVERSED)
            self._update_inplace(new_data)

        else:
            new_data = self._data.where(other=other, cond=cond, align=align,
                                        raise_on_error=raise_on_error,
                                        try_cast=try_cast, axis=block_axis,
                                        transpose=self._AXIS_REVERSED)

            return self._constructor(new_data).__finalize__(self)

    _shared_docs['where'] = ("""
        Return an object of same shape as self and whose corresponding
        entries are from self where cond is %(cond)s and otherwise are from
        other.

        Parameters
        ----------
        cond : boolean %(klass)s, array-like, or callable
            If cond is callable, it is computed on the %(klass)s and
            should return boolean %(klass)s or array. The callable must
            not change input %(klass)s (though pandas doesn't check it).

            .. versionadded:: 0.18.1
                A callable can be used as cond.

        other : scalar, %(klass)s, or callable
            If other is callable, it is computed on the %(klass)s and
            should return scalar or %(klass)s. The callable must not
            change input %(klass)s (though pandas doesn't check it).

            .. versionadded:: 0.18.1
                A callable can be used as other.

        inplace : boolean, default False
            Whether to perform the operation in place on the data
        axis : alignment axis if needed, default None
        level : alignment level if needed, default None
        try_cast : boolean, default False
            try to cast the result back to the input type (if possible),
        raise_on_error : boolean, default True
            Whether to raise on invalid data types (e.g. trying to where on
            strings)

        Returns
        -------
        wh : same type as caller

        Notes
        -----
        The %(name)s method is an application of the if-then idiom. For each
        element in the calling DataFrame, if ``cond`` is ``%(cond)s`` the
        element is used; otherwise the corresponding element from the DataFrame
        ``other`` is used.

        The signature for :func:`DataFrame.where` differs from
        :func:`numpy.where`. Roughly ``df1.where(m, df2)`` is equivalent to
        ``np.where(m, df1, df2)``.

        For further details and examples see the ``%(name)s`` documentation in
        :ref:`indexing <indexing.where_mask>`.

        Examples
        --------
        >>> s = pd.Series(range(5))
        >>> s.where(s > 0)
        0    NaN
        1    1.0
        2    2.0
        3    3.0
        4    4.0

        >>> df = pd.DataFrame(np.arange(10).reshape(-1, 2), columns=['A', 'B'])
        >>> m = df %% 3 == 0
        >>> df.where(m, -df)
           A  B
        0  0 -1
        1 -2  3
        2 -4 -5
        3  6 -7
        4 -8  9
        >>> df.where(m, -df) == np.where(m, df, -df)
              A     B
        0  True  True
        1  True  True
        2  True  True
        3  True  True
        4  True  True
        >>> df.where(m, -df) == df.mask(~m, -df)
              A     B
        0  True  True
        1  True  True
        2  True  True
        3  True  True
        4  True  True

        See Also
        --------
        :func:`DataFrame.%(name_other)s`
        """)

    @Appender(_shared_docs['where'] % dict(_shared_doc_kwargs, cond="True",
                                           name='where', name_other='mask'))
    def where(self, cond, other=np.nan, inplace=False, axis=None, level=None,
              try_cast=False, raise_on_error=True):

        other = com._apply_if_callable(other, self)
        return self._where(cond, other, inplace, axis, level, try_cast,
                           raise_on_error)

    @Appender(_shared_docs['where'] % dict(_shared_doc_kwargs, cond="False",
                                           name='mask', name_other='where'))
    def mask(self, cond, other=np.nan, inplace=False, axis=None, level=None,
             try_cast=False, raise_on_error=True):

        inplace = validate_bool_kwarg(inplace, 'inplace')
        cond = com._apply_if_callable(cond, self)

        return self.where(~cond, other=other, inplace=inplace, axis=axis,
                          level=level, try_cast=try_cast,
                          raise_on_error=raise_on_error)

    _shared_docs['shift'] = ("""
        Shift index by desired number of periods with an optional time freq

        Parameters
        ----------
        periods : int
            Number of periods to move, can be positive or negative
        freq : DateOffset, timedelta, or time rule string, optional
            Increment to use from the tseries module or time rule (e.g. 'EOM').
            See Notes.
        axis : %(axes_single_arg)s

        Notes
        -----
        If freq is specified then the index values are shifted but the data
        is not realigned. That is, use freq if you would like to extend the
        index when shifting and preserve the original data.

        Returns
        -------
        shifted : %(klass)s
    """)

    @Appender(_shared_docs['shift'] % _shared_doc_kwargs)
    def shift(self, periods=1, freq=None, axis=0):
        if periods == 0:
            return self

        block_axis = self._get_block_manager_axis(axis)
        if freq is None:
            new_data = self._data.shift(periods=periods, axis=block_axis)
        else:
            return self.tshift(periods, freq)

        return self._constructor(new_data).__finalize__(self)

    def slice_shift(self, periods=1, axis=0):
        """
        Equivalent to `shift` without copying data. The shifted data will
        not include the dropped periods and the shifted axis will be smaller
        than the original.

        Parameters
        ----------
        periods : int
            Number of periods to move, can be positive or negative

        Notes
        -----
        While the `slice_shift` is faster than `shift`, you may pay for it
        later during alignment.

        Returns
        -------
        shifted : same type as caller
        """
        if periods == 0:
            return self

        if periods > 0:
            vslicer = slice(None, -periods)
            islicer = slice(periods, None)
        else:
            vslicer = slice(-periods, None)
            islicer = slice(None, periods)

        new_obj = self._slice(vslicer, axis=axis)
        shifted_axis = self._get_axis(axis)[islicer]
        new_obj.set_axis(axis, shifted_axis)

        return new_obj.__finalize__(self)

    def tshift(self, periods=1, freq=None, axis=0):
        """
        Shift the time index, using the index's frequency if available.

        Parameters
        ----------
        periods : int
            Number of periods to move, can be positive or negative
        freq : DateOffset, timedelta, or time rule string, default None
            Increment to use from the tseries module or time rule (e.g. 'EOM')
        axis : int or basestring
            Corresponds to the axis that contains the Index

        Notes
        -----
        If freq is not specified then tries to use the freq or inferred_freq
        attributes of the index. If neither of those attributes exist, a
        ValueError is thrown

        Returns
        -------
        shifted : NDFrame
        """

        index = self._get_axis(axis)
        if freq is None:
            freq = getattr(index, 'freq', None)

        if freq is None:
            freq = getattr(index, 'inferred_freq', None)

        if freq is None:
            msg = 'Freq was not given and was not set in the index'
            raise ValueError(msg)

        if periods == 0:
            return self

        if isinstance(freq, string_types):
            freq = to_offset(freq)

        block_axis = self._get_block_manager_axis(axis)
        if isinstance(index, PeriodIndex):
            orig_freq = to_offset(index.freq)
            if freq == orig_freq:
                new_data = self._data.copy()
                new_data.axes[block_axis] = index.shift(periods)
            else:
                msg = ('Given freq %s does not match PeriodIndex freq %s' %
                       (freq.rule_code, orig_freq.rule_code))
                raise ValueError(msg)
        else:
            new_data = self._data.copy()
            new_data.axes[block_axis] = index.shift(periods, freq)

        return self._constructor(new_data).__finalize__(self)

    def truncate(self, before=None, after=None, axis=None, copy=True):
        """Truncates a sorted NDFrame before and/or after some particular
        index value. If the axis contains only datetime values, before/after
        parameters are converted to datetime values.

        Parameters
        ----------
        before : date
            Truncate before index value
        after : date
            Truncate after index value
        axis : the truncation axis, defaults to the stat axis
        copy : boolean, default is True,
            return a copy of the truncated section

        Returns
        -------
        truncated : type of caller
        """

        if axis is None:
            axis = self._stat_axis_number
        axis = self._get_axis_number(axis)
        ax = self._get_axis(axis)

        # if we have a date index, convert to dates, otherwise
        # treat like a slice
        if ax.is_all_dates:
            from pandas.core.tools.datetimes import to_datetime
            before = to_datetime(before)
            after = to_datetime(after)

        if before is not None and after is not None:
            if before > after:
                raise ValueError('Truncate: %s must be after %s' %
                                 (after, before))

        slicer = [slice(None, None)] * self._AXIS_LEN
        slicer[axis] = slice(before, after)
        result = self.loc[tuple(slicer)]

        if isinstance(ax, MultiIndex):
            setattr(result, self._get_axis_name(axis),
                    ax.truncate(before, after))

        if copy:
            result = result.copy()

        return result

    def tz_convert(self, tz, axis=0, level=None, copy=True):
        """
        Convert tz-aware axis to target time zone.

        Parameters
        ----------
        tz : string or pytz.timezone object
        axis : the axis to convert
        level : int, str, default None
            If axis ia a MultiIndex, convert a specific level. Otherwise
            must be None
        copy : boolean, default True
            Also make a copy of the underlying data

        Returns
        -------

        Raises
        ------
        TypeError
            If the axis is tz-naive.
        """
        axis = self._get_axis_number(axis)
        ax = self._get_axis(axis)

        def _tz_convert(ax, tz):
            if not hasattr(ax, 'tz_convert'):
                if len(ax) > 0:
                    ax_name = self._get_axis_name(axis)
                    raise TypeError('%s is not a valid DatetimeIndex or '
                                    'PeriodIndex' % ax_name)
                else:
                    ax = DatetimeIndex([], tz=tz)
            else:
                ax = ax.tz_convert(tz)
            return ax

        # if a level is given it must be a MultiIndex level or
        # equivalent to the axis name
        if isinstance(ax, MultiIndex):
            level = ax._get_level_number(level)
            new_level = _tz_convert(ax.levels[level], tz)
            ax = ax.set_levels(new_level, level=level)
        else:
            if level not in (None, 0, ax.name):
                raise ValueError("The level {0} is not valid".format(level))
            ax = _tz_convert(ax, tz)

        result = self._constructor(self._data, copy=copy)
        result.set_axis(axis, ax)
        return result.__finalize__(self)

    @deprecate_kwarg(old_arg_name='infer_dst', new_arg_name='ambiguous',
                     mapping={True: 'infer',
                              False: 'raise'})
    def tz_localize(self, tz, axis=0, level=None, copy=True,
                    ambiguous='raise'):
        """
        Localize tz-naive TimeSeries to target time zone.

        Parameters
        ----------
        tz : string or pytz.timezone object
        axis : the axis to localize
        level : int, str, default None
            If axis ia a MultiIndex, localize a specific level. Otherwise
            must be None
        copy : boolean, default True
            Also make a copy of the underlying data
        ambiguous : 'infer', bool-ndarray, 'NaT', default 'raise'
            - 'infer' will attempt to infer fall dst-transition hours based on
              order
            - bool-ndarray where True signifies a DST time, False designates
              a non-DST time (note that this flag is only applicable for
              ambiguous times)
            - 'NaT' will return NaT where there are ambiguous times
            - 'raise' will raise an AmbiguousTimeError if there are ambiguous
              times
        infer_dst : boolean, default False (DEPRECATED)
            Attempt to infer fall dst-transition hours based on order

        Returns
        -------

        Raises
        ------
        TypeError
            If the TimeSeries is tz-aware and tz is not None.
        """
        axis = self._get_axis_number(axis)
        ax = self._get_axis(axis)

        def _tz_localize(ax, tz, ambiguous):
            if not hasattr(ax, 'tz_localize'):
                if len(ax) > 0:
                    ax_name = self._get_axis_name(axis)
                    raise TypeError('%s is not a valid DatetimeIndex or '
                                    'PeriodIndex' % ax_name)
                else:
                    ax = DatetimeIndex([], tz=tz)
            else:
                ax = ax.tz_localize(tz, ambiguous=ambiguous)
            return ax

        # if a level is given it must be a MultiIndex level or
        # equivalent to the axis name
        if isinstance(ax, MultiIndex):
            level = ax._get_level_number(level)
            new_level = _tz_localize(ax.levels[level], tz, ambiguous)
            ax = ax.set_levels(new_level, level=level)
        else:
            if level not in (None, 0, ax.name):
                raise ValueError("The level {0} is not valid".format(level))
            ax = _tz_localize(ax, tz, ambiguous)

        result = self._constructor(self._data, copy=copy)
        result.set_axis(axis, ax)
        return result.__finalize__(self)

    # ----------------------------------------------------------------------
    # Numeric Methods
    def abs(self):
        """
        Return an object with absolute value taken--only applicable to objects
        that are all numeric.

        Returns
        -------
        abs: type of caller
        """
        return np.abs(self)

    def describe(self, percentiles=None, include=None, exclude=None):
        """
        Generates descriptive statistics that summarize the central tendency,
        dispersion and shape of a dataset's distribution, excluding
        ``NaN`` values.

        Analyzes both numeric and object series, as well
        as ``DataFrame`` column sets of mixed data types. The output
        will vary depending on what is provided. Refer to the notes
        below for more detail.

        Parameters
        ----------
        percentiles : list-like of numbers, optional
            The percentiles to include in the output. All should
            fall between 0 and 1. The default is
            ``[.25, .5, .75]``, which returns the 25th, 50th, and
            75th percentiles.
        include : 'all', list-like of dtypes or None (default), optional
            A white list of data types to include in the result. Ignored
            for ``Series``. Here are the options:

            - 'all' : All columns of the input will be included in the output.
            - A list-like of dtypes : Limits the results to the
              provided data types.
              To limit the result to numeric types submit
              ``numpy.number``. To limit it instead to categorical
              objects submit the ``numpy.object`` data type. Strings
              can also be used in the style of
              ``select_dtypes`` (e.g. ``df.describe(include=['O'])``)
            - None (default) : The result will include all numeric columns.
        exclude : list-like of dtypes or None (default), optional,
            A black list of data types to omit from the result. Ignored
            for ``Series``. Here are the options:

            - A list-like of dtypes : Excludes the provided data types
              from the result. To select numeric types submit
              ``numpy.number``. To select categorical objects submit the data
              type ``numpy.object``. Strings can also be used in the style of
              ``select_dtypes`` (e.g. ``df.describe(include=['O'])``)
            - None (default) : The result will exclude nothing.

        Returns
        -------
        summary:  Series/DataFrame of summary statistics

        Notes
        -----
        For numeric data, the result's index will include ``count``,
        ``mean``, ``std``, ``min``, ``max`` as well as lower, ``50`` and
        upper percentiles. By default the lower percentile is ``25`` and the
        upper percentile is ``75``. The ``50`` percentile is the
        same as the median.

        For object data (e.g. strings or timestamps), the result's index
        will include ``count``, ``unique``, ``top``, and ``freq``. The ``top``
        is the most common value. The ``freq`` is the most common value's
        frequency. Timestamps also include the ``first`` and ``last`` items.

        If multiple object values have the highest count, then the
        ``count`` and ``top`` results will be arbitrarily chosen from
        among those with the highest count.

        For mixed data types provided via a ``DataFrame``, the default is to
        return only an analysis of numeric columns. If ``include='all'``
        is provided as an option, the result will include a union of
        attributes of each type.

        The `include` and `exclude` parameters can be used to limit
        which columns in a ``DataFrame`` are analyzed for the output.
        The parameters are ignored when analyzing a ``Series``.

        Examples
        --------
        Describing a numeric ``Series``.

        >>> s = pd.Series([1, 2, 3])
        >>> s.describe()
        count    3.0
        mean     2.0
        std      1.0
        min      1.0
        25%      1.5
        50%      2.0
        75%      2.5
        max      3.0

        Describing a categorical ``Series``.

        >>> s = pd.Series(['a', 'a', 'b', 'c'])
        >>> s.describe()
        count     4
        unique    3
        top       a
        freq      2
        dtype: object

        Describing a timestamp ``Series``.

        >>> s = pd.Series([
        ...   np.datetime64("2000-01-01"),
        ...   np.datetime64("2010-01-01"),
        ...   np.datetime64("2010-01-01")
        ... ])
        >>> s.describe()
        count                       3
        unique                      2
        top       2010-01-01 00:00:00
        freq                        2
        first     2000-01-01 00:00:00
        last      2010-01-01 00:00:00
        dtype: object

        Describing a ``DataFrame``. By default only numeric fields
        are returned.

        >>> df = pd.DataFrame([[1, 'a'], [2, 'b'], [3, 'c']],
        ...                   columns=['numeric', 'object'])
        >>> df.describe()
               numeric
        count      3.0
        mean       2.0
        std        1.0
        min        1.0
        25%        1.5
        50%        2.0
        75%        2.5
        max        3.0

        Describing all columns of a ``DataFrame`` regardless of data type.

        >>> df.describe(include='all')
                numeric object
        count       3.0      3
        unique      NaN      3
        top         NaN      b
        freq        NaN      1
        mean        2.0    NaN
        std         1.0    NaN
        min         1.0    NaN
        25%         1.5    NaN
        50%         2.0    NaN
        75%         2.5    NaN
        max         3.0    NaN

        Describing a column from a ``DataFrame`` by accessing it as
        an attribute.

        >>> df.numeric.describe()
        count    3.0
        mean     2.0
        std      1.0
        min      1.0
        25%      1.5
        50%      2.0
        75%      2.5
        max      3.0
        Name: numeric, dtype: float64

        Including only numeric columns in a ``DataFrame`` description.

        >>> df.describe(include=[np.number])
               numeric
        count      3.0
        mean       2.0
        std        1.0
        min        1.0
        25%        1.5
        50%        2.0
        75%        2.5
        max        3.0

        Including only string columns in a ``DataFrame`` description.

        >>> df.describe(include=[np.object])
               object
        count       3
        unique      3
        top         b
        freq        1

        Excluding numeric columns from a ``DataFrame`` description.

        >>> df.describe(exclude=[np.number])
               object
        count       3
        unique      3
        top         b
        freq        1

        Excluding object columns from a ``DataFrame`` description.

        >>> df.describe(exclude=[np.object])
               numeric
        count      3.0
        mean       2.0
        std        1.0
        min        1.0
        25%        1.5
        50%        2.0
        75%        2.5
        max        3.0

        See Also
        --------
        DataFrame.count
        DataFrame.max
        DataFrame.min
        DataFrame.mean
        DataFrame.std
        DataFrame.select_dtypes
        """
        if self.ndim >= 3:
            msg = "describe is not implemented on Panel or PanelND objects."
            raise NotImplementedError(msg)
        elif self.ndim == 2 and self.columns.size == 0:
            raise ValueError("Cannot describe a DataFrame without columns")

        if percentiles is not None:
            # explicit conversion of `percentiles` to list
            percentiles = list(percentiles)

            # get them all to be in [0, 1]
            self._check_percentile(percentiles)

            # median should always be included
            if 0.5 not in percentiles:
                percentiles.append(0.5)
            percentiles = np.asarray(percentiles)
        else:
            percentiles = np.array([0.25, 0.5, 0.75])

        # sort and check for duplicates
        unique_pcts = np.unique(percentiles)
        if len(unique_pcts) < len(percentiles):
            raise ValueError("percentiles cannot contain duplicates")
        percentiles = unique_pcts

        formatted_percentiles = format_percentiles(percentiles)

        def describe_numeric_1d(series):
            stat_index = (['count', 'mean', 'std', 'min'] +
                          formatted_percentiles + ['max'])
            d = ([series.count(), series.mean(), series.std(), series.min()] +
                 [series.quantile(x) for x in percentiles] + [series.max()])
            return pd.Series(d, index=stat_index, name=series.name)

        def describe_categorical_1d(data):
            names = ['count', 'unique']
            objcounts = data.value_counts()
            count_unique = len(objcounts[objcounts != 0])
            result = [data.count(), count_unique]
            if result[1] > 0:
                top, freq = objcounts.index[0], objcounts.iloc[0]

                if is_datetime64_dtype(data):
                    asint = data.dropna().values.view('i8')
                    names += ['top', 'freq', 'first', 'last']
                    result += [lib.Timestamp(top), freq,
                               lib.Timestamp(asint.min()),
                               lib.Timestamp(asint.max())]
                else:
                    names += ['top', 'freq']
                    result += [top, freq]

            return pd.Series(result, index=names, name=data.name)

        def describe_1d(data):
            if is_bool_dtype(data):
                return describe_categorical_1d(data)
            elif is_numeric_dtype(data):
                return describe_numeric_1d(data)
            elif is_timedelta64_dtype(data):
                return describe_numeric_1d(data)
            else:
                return describe_categorical_1d(data)

        if self.ndim == 1:
            return describe_1d(self)
        elif (include is None) and (exclude is None):
            # when some numerics are found, keep only numerics
            data = self.select_dtypes(include=[np.number])
            if len(data.columns) == 0:
                data = self
        elif include == 'all':
            if exclude is not None:
                msg = "exclude must be None when include is 'all'"
                raise ValueError(msg)
            data = self
        else:
            data = self.select_dtypes(include=include, exclude=exclude)

        ldesc = [describe_1d(s) for _, s in data.iteritems()]
        # set a convenient order for rows
        names = []
        ldesc_indexes = sorted([x.index for x in ldesc], key=len)
        for idxnames in ldesc_indexes:
            for name in idxnames:
                if name not in names:
                    names.append(name)

        d = pd.concat(ldesc, join_axes=pd.Index([names]), axis=1)
        d.columns = data.columns.copy()
        return d

    def _check_percentile(self, q):
        """Validate percentiles (used by describe and quantile)."""

        msg = ("percentiles should all be in the interval [0, 1]. "
               "Try {0} instead.")
        q = np.asarray(q)
        if q.ndim == 0:
            if not 0 <= q <= 1:
                raise ValueError(msg.format(q / 100.0))
        else:
            if not all(0 <= qs <= 1 for qs in q):
                raise ValueError(msg.format(q / 100.0))
        return q

    _shared_docs['pct_change'] = """
        Percent change over given number of periods.

        Parameters
        ----------
        periods : int, default 1
            Periods to shift for forming percent change
        fill_method : str, default 'pad'
            How to handle NAs before computing percent changes
        limit : int, default None
            The number of consecutive NAs to fill before stopping
        freq : DateOffset, timedelta, or offset alias string, optional
            Increment to use from time series API (e.g. 'M' or BDay())

        Returns
        -------
        chg : %(klass)s

        Notes
        -----

        By default, the percentage change is calculated along the stat
        axis: 0, or ``Index``, for ``DataFrame`` and 1, or ``minor`` for
        ``Panel``. You can change this with the ``axis`` keyword argument.
        """

    @Appender(_shared_docs['pct_change'] % _shared_doc_kwargs)
    def pct_change(self, periods=1, fill_method='pad', limit=None, freq=None,
                   **kwargs):
        # TODO: Not sure if above is correct - need someone to confirm.
        axis = self._get_axis_number(kwargs.pop('axis', self._stat_axis_name))
        if fill_method is None:
            data = self
        else:
            data = self.fillna(method=fill_method, limit=limit, axis=axis)

        rs = (data.div(data.shift(periods=periods, freq=freq, axis=axis,
                                  **kwargs)) - 1)
        if freq is None:
            mask = isnull(_values_from_object(self))
            np.putmask(rs.values, mask, np.nan)
        return rs

    def _agg_by_level(self, name, axis=0, level=0, skipna=True, **kwargs):
        grouped = self.groupby(level=level, axis=axis)
        if hasattr(grouped, name) and skipna:
            return getattr(grouped, name)(**kwargs)
        axis = self._get_axis_number(axis)
        method = getattr(type(self), name)
        applyf = lambda x: method(x, axis=axis, skipna=skipna, **kwargs)
        return grouped.aggregate(applyf)

    @classmethod
    def _add_numeric_operations(cls):
        """Add the operations to the cls; evaluate the doc strings again"""

        axis_descr, name, name2 = _doc_parms(cls)

        cls.any = _make_logical_function(
            cls, 'any', name, name2, axis_descr,
            'Return whether any element is True over requested axis',
            nanops.nanany)
        cls.all = _make_logical_function(
            cls, 'all', name, name2, axis_descr,
            'Return whether all elements are True over requested axis',
            nanops.nanall)

        @Substitution(outname='mad',
                      desc="Return the mean absolute deviation of the values "
                           "for the requested axis",
                      name1=name, name2=name2, axis_descr=axis_descr)
        @Appender(_num_doc)
        def mad(self, axis=None, skipna=None, level=None):
            if skipna is None:
                skipna = True
            if axis is None:
                axis = self._stat_axis_number
            if level is not None:
                return self._agg_by_level('mad', axis=axis, level=level,
                                          skipna=skipna)

            data = self._get_numeric_data()
            if axis == 0:
                demeaned = data - data.mean(axis=0)
            else:
                demeaned = data.sub(data.mean(axis=1), axis=0)
            return np.abs(demeaned).mean(axis=axis, skipna=skipna)

        cls.mad = mad

        cls.sem = _make_stat_function_ddof(
            cls, 'sem', name, name2, axis_descr,
            "Return unbiased standard error of the mean over requested "
            "axis.\n\nNormalized by N-1 by default. This can be changed "
            "using the ddof argument",
            nanops.nansem)
        cls.var = _make_stat_function_ddof(
            cls, 'var', name, name2, axis_descr,
            "Return unbiased variance over requested axis.\n\nNormalized by "
            "N-1 by default. This can be changed using the ddof argument",
            nanops.nanvar)
        cls.std = _make_stat_function_ddof(
            cls, 'std', name, name2, axis_descr,
            "Return sample standard deviation over requested axis."
            "\n\nNormalized by N-1 by default. This can be changed using the "
            "ddof argument",
            nanops.nanstd)

        @Substitution(outname='compounded',
                      desc="Return the compound percentage of the values for "
                      "the requested axis", name1=name, name2=name2,
                      axis_descr=axis_descr)
        @Appender(_num_doc)
        def compound(self, axis=None, skipna=None, level=None):
            if skipna is None:
                skipna = True
            return (1 + self).prod(axis=axis, skipna=skipna, level=level) - 1

        cls.compound = compound

        cls.cummin = _make_cum_function(
            cls, 'cummin', name, name2, axis_descr, "cumulative minimum",
            lambda y, axis: np.minimum.accumulate(y, axis), "min",
            np.inf, np.nan)
        cls.cumsum = _make_cum_function(
            cls, 'cumsum', name, name2, axis_descr, "cumulative sum",
            lambda y, axis: y.cumsum(axis), "sum", 0., np.nan)
        cls.cumprod = _make_cum_function(
            cls, 'cumprod', name, name2, axis_descr, "cumulative product",
            lambda y, axis: y.cumprod(axis), "prod", 1., np.nan)
        cls.cummax = _make_cum_function(
            cls, 'cummax', name, name2, axis_descr, "cumulative max",
            lambda y, axis: np.maximum.accumulate(y, axis), "max",
            -np.inf, np.nan)

        cls.sum = _make_stat_function(
            cls, 'sum', name, name2, axis_descr,
            'Return the sum of the values for the requested axis',
            nanops.nansum)
        cls.mean = _make_stat_function(
            cls, 'mean', name, name2, axis_descr,
            'Return the mean of the values for the requested axis',
            nanops.nanmean)
        cls.skew = _make_stat_function(
            cls, 'skew', name, name2, axis_descr,
            'Return unbiased skew over requested axis\nNormalized by N-1',
            nanops.nanskew)
        cls.kurt = _make_stat_function(
            cls, 'kurt', name, name2, axis_descr,
            "Return unbiased kurtosis over requested axis using Fisher's "
            "definition of\nkurtosis (kurtosis of normal == 0.0). Normalized "
            "by N-1\n",
            nanops.nankurt)
        cls.kurtosis = cls.kurt
        cls.prod = _make_stat_function(
            cls, 'prod', name, name2, axis_descr,
            'Return the product of the values for the requested axis',
            nanops.nanprod)
        cls.product = cls.prod
        cls.median = _make_stat_function(
            cls, 'median', name, name2, axis_descr,
            'Return the median of the values for the requested axis',
            nanops.nanmedian)
        cls.max = _make_stat_function(
            cls, 'max', name, name2, axis_descr,
            """This method returns the maximum of the values in the object.
            If you want the *index* of the maximum, use ``idxmax``. This is
            the equivalent of the ``numpy.ndarray`` method ``argmax``.""",
            nanops.nanmax)
        cls.min = _make_stat_function(
            cls, 'min', name, name2, axis_descr,
            """This method returns the minimum of the values in the object.
            If you want the *index* of the minimum, use ``idxmin``. This is
            the equivalent of the ``numpy.ndarray`` method ``argmin``.""",
            nanops.nanmin)

    @classmethod
    def _add_series_only_operations(cls):
        """Add the series only operations to the cls; evaluate the doc
        strings again.
        """

        axis_descr, name, name2 = _doc_parms(cls)

        def nanptp(values, axis=0, skipna=True):
            nmax = nanops.nanmax(values, axis, skipna)
            nmin = nanops.nanmin(values, axis, skipna)
            return nmax - nmin

        cls.ptp = _make_stat_function(
            cls, 'ptp', name, name2, axis_descr,
            """Returns the difference between the maximum value and the
            minimum value in the object. This is the equivalent of the
            ``numpy.ndarray`` method ``ptp``.""",
            nanptp)

    @classmethod
    def _add_series_or_dataframe_operations(cls):
        """Add the series or dataframe only operations to the cls; evaluate
        the doc strings again.
        """

        from pandas.core import window as rwindow

        @Appender(rwindow.rolling.__doc__)
        def rolling(self, window, min_periods=None, freq=None, center=False,
                    win_type=None, on=None, axis=0, closed=None):
            axis = self._get_axis_number(axis)
            return rwindow.rolling(self, window=window,
                                   min_periods=min_periods, freq=freq,
                                   center=center, win_type=win_type,
                                   on=on, axis=axis, closed=closed)

        cls.rolling = rolling

        @Appender(rwindow.expanding.__doc__)
        def expanding(self, min_periods=1, freq=None, center=False, axis=0):
            axis = self._get_axis_number(axis)
            return rwindow.expanding(self, min_periods=min_periods, freq=freq,
                                     center=center, axis=axis)

        cls.expanding = expanding

        @Appender(rwindow.ewm.__doc__)
        def ewm(self, com=None, span=None, halflife=None, alpha=None,
                min_periods=0, freq=None, adjust=True, ignore_na=False,
                axis=0):
            axis = self._get_axis_number(axis)
            return rwindow.ewm(self, com=com, span=span, halflife=halflife,
                               alpha=alpha, min_periods=min_periods, freq=freq,
                               adjust=adjust, ignore_na=ignore_na, axis=axis)

        cls.ewm = ewm

        @Appender(_shared_docs['transform'] % _shared_doc_kwargs)
        def transform(self, func, *args, **kwargs):
            result = self.agg(func, *args, **kwargs)
            if is_scalar(result) or len(result) != len(self):
                raise ValueError("transforms cannot produce "
                                 "aggregated results")

            return result

        cls.transform = transform


def _doc_parms(cls):
    """Return a tuple of the doc parms."""
    axis_descr = "{%s}" % ', '.join(["{0} ({1})".format(a, i)
                                     for i, a in enumerate(cls._AXIS_ORDERS)])
    name = (cls._constructor_sliced.__name__
            if cls._AXIS_LEN > 1 else 'scalar')
    name2 = cls.__name__
    return axis_descr, name, name2


_num_doc = """

%(desc)s

Parameters
----------
axis : %(axis_descr)s
skipna : boolean, default True
    Exclude NA/null values. If an entire row/column is NA, the result
    will be NA
level : int or level name, default None
    If the axis is a MultiIndex (hierarchical), count along a
    particular level, collapsing into a %(name1)s
numeric_only : boolean, default None
    Include only float, int, boolean columns. If None, will attempt to use
    everything, then use only numeric data. Not implemented for Series.

Returns
-------
%(outname)s : %(name1)s or %(name2)s (if level specified)\n"""

_num_ddof_doc = """

%(desc)s

Parameters
----------
axis : %(axis_descr)s
skipna : boolean, default True
    Exclude NA/null values. If an entire row/column is NA, the result
    will be NA
level : int or level name, default None
    If the axis is a MultiIndex (hierarchical), count along a
    particular level, collapsing into a %(name1)s
ddof : int, default 1
    degrees of freedom
numeric_only : boolean, default None
    Include only float, int, boolean columns. If None, will attempt to use
    everything, then use only numeric data. Not implemented for Series.

Returns
-------
%(outname)s : %(name1)s or %(name2)s (if level specified)\n"""

_bool_doc = """

%(desc)s

Parameters
----------
axis : %(axis_descr)s
skipna : boolean, default True
    Exclude NA/null values. If an entire row/column is NA, the result
    will be NA
level : int or level name, default None
    If the axis is a MultiIndex (hierarchical), count along a
    particular level, collapsing into a %(name1)s
bool_only : boolean, default None
    Include only boolean columns. If None, will attempt to use everything,
    then use only boolean data. Not implemented for Series.

Returns
-------
%(outname)s : %(name1)s or %(name2)s (if level specified)\n"""

_cnum_doc = """

Parameters
----------
axis : %(axis_descr)s
skipna : boolean, default True
    Exclude NA/null values. If an entire row/column is NA, the result
    will be NA

Returns
-------
%(outname)s : %(name1)s\n


See also
--------
pandas.core.window.Expanding.%(accum_func_name)s : Similar functionality
    but ignores ``NaN`` values.

"""


def _make_stat_function(cls, name, name1, name2, axis_descr, desc, f):
    @Substitution(outname=name, desc=desc, name1=name1, name2=name2,
                  axis_descr=axis_descr)
    @Appender(_num_doc)
    def stat_func(self, axis=None, skipna=None, level=None, numeric_only=None,
                  **kwargs):
        nv.validate_stat_func(tuple(), kwargs, fname=name)
        if skipna is None:
            skipna = True
        if axis is None:
            axis = self._stat_axis_number
        if level is not None:
            return self._agg_by_level(name, axis=axis, level=level,
                                      skipna=skipna)
        return self._reduce(f, name, axis=axis, skipna=skipna,
                            numeric_only=numeric_only)

    return set_function_name(stat_func, name, cls)


def _make_stat_function_ddof(cls, name, name1, name2, axis_descr, desc, f):
    @Substitution(outname=name, desc=desc, name1=name1, name2=name2,
                  axis_descr=axis_descr)
    @Appender(_num_ddof_doc)
    def stat_func(self, axis=None, skipna=None, level=None, ddof=1,
                  numeric_only=None, **kwargs):
        nv.validate_stat_ddof_func(tuple(), kwargs, fname=name)
        if skipna is None:
            skipna = True
        if axis is None:
            axis = self._stat_axis_number
        if level is not None:
            return self._agg_by_level(name, axis=axis, level=level,
                                      skipna=skipna, ddof=ddof)
        return self._reduce(f, name, axis=axis, numeric_only=numeric_only,
                            skipna=skipna, ddof=ddof)

    return set_function_name(stat_func, name, cls)


def _make_cum_function(cls, name, name1, name2, axis_descr, desc,
                       accum_func, accum_func_name, mask_a, mask_b):
    @Substitution(outname=name, desc=desc, name1=name1, name2=name2,
                  axis_descr=axis_descr, accum_func_name=accum_func_name)
    @Appender("Return {0} over requested axis.".format(desc) +
              _cnum_doc)
    def cum_func(self, axis=None, skipna=True, *args, **kwargs):
        skipna = nv.validate_cum_func_with_skipna(skipna, args, kwargs, name)
        if axis is None:
            axis = self._stat_axis_number
        else:
            axis = self._get_axis_number(axis)

        y = _values_from_object(self).copy()

        if (skipna and
                issubclass(y.dtype.type, (np.datetime64, np.timedelta64))):
            result = accum_func(y, axis)
            mask = isnull(self)
            np.putmask(result, mask, tslib.iNaT)
        elif skipna and not issubclass(y.dtype.type, (np.integer, np.bool_)):
            mask = isnull(self)
            np.putmask(y, mask, mask_a)
            result = accum_func(y, axis)
            np.putmask(result, mask, mask_b)
        else:
            result = accum_func(y, axis)

        d = self._construct_axes_dict()
        d['copy'] = False
        return self._constructor(result, **d).__finalize__(self)

    return set_function_name(cum_func, name, cls)


def _make_logical_function(cls, name, name1, name2, axis_descr, desc, f):
    @Substitution(outname=name, desc=desc, name1=name1, name2=name2,
                  axis_descr=axis_descr)
    @Appender(_bool_doc)
    def logical_func(self, axis=None, bool_only=None, skipna=None, level=None,
                     **kwargs):
        nv.validate_logical_func(tuple(), kwargs, fname=name)
        if skipna is None:
            skipna = True
        if axis is None:
            axis = self._stat_axis_number
        if level is not None:
            if bool_only is not None:
                raise NotImplementedError("Option bool_only is not "
                                          "implemented with option level.")
            return self._agg_by_level(name, axis=axis, level=level,
                                      skipna=skipna)
        return self._reduce(f, axis=axis, skipna=skipna,
                            numeric_only=bool_only, filter_type='bool',
                            name=name)

    return set_function_name(logical_func, name, cls)


# install the indexes
for _name, _indexer in indexing.get_indexers_list():
    NDFrame._create_indexer(_name, _indexer)