v0.16.1.txt

.. _whatsnew_0161:

v0.16.1 (May 11, 2015)
----------------------

This is a minor bug-fix release from 0.16.0 and includes a a large number of
bug fixes along several new features, enhancements, and performance improvements.
We recommend that all users upgrade to this version.

Highlights include:

- Support for a ``CategoricalIndex``, a category based index, see :ref:`here <whatsnew_0161.enhancements.categoricalindex>`
- New section on how-to-contribute to *pandas*, see :ref:`here <contributing>`
- Revised "Merge, join, and concatenate" documentation, including graphical examples to make it easier to understand each operations, see :ref:`here <merging>`
- New method ``sample`` for drawing random samples from Series, DataFrames and Panels. See :ref:`here <whatsnew_0161.enhancements.sample>`
- ``BusinessHour`` date-offset is now supported, see :ref:`here <timeseries.businesshour>`
-  Further enhancement to the ``.str`` accessor to make string operations easier, see :ref:`here <whatsnew_0161.enhancements.string>`

.. contents:: What's new in v0.16.1
    :local:
    :backlinks: none

.. _whatsnew_0161.enhancements:

.. warning::

   In pandas 0.17.0, the sub-package ``pandas.io.data`` will be removed in favor of a separately installable package. See :ref:`here for details <remote_data.pandas_datareader>` (:issue:`8961`)

Enhancements
~~~~~~~~~~~~

- ``BusinessHour`` offset is now supported, which represents business hours starting from 09:00 - 17:00 on ``BusinessDay`` by default. See :ref:`Here <timeseries.businesshour>` for details. (:issue:`7905`)

  .. ipython:: python

     Timestamp('2014-08-01 09:00') + BusinessHour()
     Timestamp('2014-08-01 07:00') + BusinessHour()
     Timestamp('2014-08-01 16:30') + BusinessHour()

- ``DataFrame.diff`` now takes an ``axis`` parameter that determines the direction of differencing (:issue:`9727`)

- Allow ``clip``, ``clip_lower``, and ``clip_upper`` to accept array-like arguments as thresholds (This is a regression from 0.11.0). These methods now have an ``axis`` parameter which determines how the Series or DataFrame will be aligned with the threshold(s). (:issue:`6966`)

- ``DataFrame.mask()`` and ``Series.mask()`` now support same keywords as ``where`` (:issue:`8801`)

- ``drop`` function can now accept ``errors`` keyword to suppress ``ValueError`` raised when any of label does not exist in the target data. (:issue:`6736`)

  .. ipython:: python

    df = DataFrame(np.random.randn(3, 3), columns=['A', 'B', 'C'])
    df.drop(['A', 'X'], axis=1, errors='ignore')

- Allow conversion of values with dtype ``datetime64`` or ``timedelta64`` to strings using ``astype(str)`` (:issue:`9757`)
- ``get_dummies`` function now accepts ``sparse`` keyword.  If set to ``True``, the return ``DataFrame`` is sparse, e.g. ``SparseDataFrame``. (:issue:`8823`)
- ``Period`` now accepts ``datetime64`` as value input. (:issue:`9054`)

- Allow timedelta string conversion when leading zero is missing from time definition, ie `0:00:00` vs `00:00:00`. (:issue:`9570`)
- Allow ``Panel.shift`` with ``axis='items'`` (:issue:`9890`)

- Trying to write an excel file now raises ``NotImplementedError`` if the ``DataFrame`` has a ``MultiIndex`` instead of writing a broken Excel file. (:issue:`9794`)
- Allow ``Categorical.add_categories`` to accept ``Series`` or ``np.array``. (:issue:`9927`)

- Add/delete ``str/dt/cat`` accessors dynamically from ``__dir__``. (:issue:`9910`)
- Add ``normalize`` as a ``dt`` accessor method. (:issue:`10047`)

- ``DataFrame`` and ``Series`` now have ``_constructor_expanddim`` property as overridable constructor for one higher dimensionality data. This should be used only when it is really needed, see :ref:`here <ref-subclassing-pandas>`

- ``pd.lib.infer_dtype`` now returns ``'bytes'`` in Python 3 where appropriate. (:issue:`10032`)

.. _whatsnew_0161.enhancements.categoricalindex:

CategoricalIndex
^^^^^^^^^^^^^^^^

We introduce a ``CategoricalIndex``, a new type of index object that is useful for supporting
indexing with duplicates. This is a container around a ``Categorical`` (introduced in v0.15.0)
and allows efficient indexing and storage of an index with a large number of duplicated elements. Prior to 0.16.1,
setting the index of a ``DataFrame/Series`` with a ``category`` dtype would convert this to regular object-based ``Index``.

.. ipython :: python

   df = DataFrame({'A' : np.arange(6),
                   'B' : Series(list('aabbca')).astype('category',
                                                       categories=list('cab'))
                  })
   df
   df.dtypes
   df.B.cat.categories

setting the index, will create create a ``CategoricalIndex``

.. ipython :: python

   df2 = df.set_index('B')
   df2.index

indexing with ``__getitem__/.iloc/.loc/.ix`` works similarly to an Index with duplicates.
The indexers MUST be in the category or the operation will raise.

.. ipython :: python

   df2.loc['a']

and preserves the ``CategoricalIndex``

.. ipython :: python

   df2.loc['a'].index

sorting will order by the order of the categories

.. ipython :: python

   df2.sort_index()

groupby operations on the index will preserve the index nature as well

.. ipython :: python

   df2.groupby(level=0).sum()
   df2.groupby(level=0).sum().index

reindexing operations, will return a resulting index based on the type of the passed
indexer, meaning that passing a list will return a plain-old-``Index``; indexing with
a ``Categorical`` will return a ``CategoricalIndex``, indexed according to the categories
of the PASSED ``Categorical`` dtype. This allows one to arbitrarly index these even with
values NOT in the categories, similarly to how you can reindex ANY pandas index.

.. ipython :: python

   df2.reindex(['a','e'])
   df2.reindex(['a','e']).index
   df2.reindex(pd.Categorical(['a','e'],categories=list('abcde')))
   df2.reindex(pd.Categorical(['a','e'],categories=list('abcde'))).index

See the :ref:`documentation <advanced.categoricalindex>` for more. (:issue:`7629`, :issue:`10038`, :issue:`10039`)

.. _whatsnew_0161.enhancements.sample:

Sample
^^^^^^

Series, DataFrames, and Panels now have a new method: :meth:`~pandas.DataFrame.sample`.
The method accepts a specific number of rows or columns to return, or a fraction of the
total number or rows or columns. It also has options for sampling with or without replacement,
for passing in a column for weights for non-uniform sampling, and for setting seed values to
facilitate replication. (:issue:`2419`)

.. ipython :: python

   example_series = Series([0,1,2,3,4,5])

   # When no arguments are passed, returns 1
   example_series.sample()

   # One may specify either a number of rows:
   example_series.sample(n=3)

   # Or a fraction of the rows:
   example_series.sample(frac=0.5)

   # weights are accepted.
   example_weights = [0, 0, 0.2, 0.2, 0.2, 0.4]
   example_series.sample(n=3, weights=example_weights)

   # weights will also be normalized if they do not sum to one,
   # and missing values will be treated as zeros.
   example_weights2 = [0.5, 0, 0, 0, None, np.nan]
   example_series.sample(n=1, weights=example_weights2)


When applied to a DataFrame, one may pass the name of a column to specify sampling weights
when sampling from rows.

.. ipython :: python

   df = DataFrame({'col1':[9,8,7,6], 'weight_column':[0.5, 0.4, 0.1, 0]})
   df.sample(n=3, weights='weight_column')


.. _whatsnew_0161.enhancements.string:

String Methods Enhancements
^^^^^^^^^^^^^^^^^^^^^^^^^^^

:ref:`Continuing from v0.16.0 <whatsnew_0160.enhancements.string>`, the following
enhancements make string operations easier and more consistent with standard python string operations.

- The following new methods are accesible via ``.str`` accessor to apply the function to each values. (:issue:`9766`, :issue:`9773`, :issue:`10031`, :issue:`10045`, :issue:`10052`)

  ================  ===============  ===============  ===============  ================
  ..                ..               Methods          ..               ..
  ================  ===============  ===============  ===============  ================
  ``capitalize()``  ``swapcase()``   ``normalize()``  ``partition()``  ``rpartition()``
  ``index()``       ``rindex()``     ``translate()``
  ================  ===============  ===============  ===============  ================



- Added ``StringMethods`` (``.str`` accessor) to ``Index`` (:issue:`9068`)

  The ``.str`` accessor is now available for both ``Series`` and ``Index``.

  .. ipython:: python

     idx = Index([' jack', 'jill ', ' jesse ', 'frank'])
     idx.str.strip()

  One special case for the `.str` accessor on ``Index`` is that if a string method returns ``bool``, the ``.str`` accessor
  will return a ``np.array`` instead of a boolean ``Index`` (:issue:`8875`). This enables the following expression
  to work naturally:

  .. ipython:: python

     idx = Index(['a1', 'a2', 'b1', 'b2'])
     s = Series(range(4), index=idx)
     s
     idx.str.startswith('a')
     s[s.index.str.startswith('a')]


- ``split`` now takes ``expand`` keyword to specify whether to expand dimensionality. ``return_type`` is deprecated. (:issue:`9847`)

  .. ipython:: python

     s = Series(['a,b', 'a,c', 'b,c'])

     # return Series
     s.str.split(',')

     # return DataFrame
     s.str.split(',', expand=True)

     idx = Index(['a,b', 'a,c', 'b,c'])

     # return Index
     idx.str.split(',')

     # return MultiIndex
     idx.str.split(',', expand=True)


- Improved ``extract`` and ``get_dummies`` methods for ``Index.str`` (:issue:`9980`)

.. _whatsnew_0161.api:

API changes
~~~~~~~~~~~





- When passing in an ax to ``df.plot( ..., ax=ax)``, the `sharex` kwarg will now default to `False`.
  The result is that the visibility of xlabels and xticklabels will not anymore be changed. You
  have to do that by yourself for the right axes in your figure or set ``sharex=True`` explicitly
  (but this changes the visible for all axes in the figure, not only the one which is passed in!).
  If pandas creates the subplots itself (e.g. no passed in `ax` kwarg), then the
  default is still ``sharex=True`` and the visibility changes are applied.



- Add support for separating years and quarters using dashes, for
  example 2014-Q1.  (:issue:`9688`)

- :meth:`~pandas.DataFrame.assign` now inserts new columns in alphabetical order. Previously
  the order was arbitrary. (:issue:`9777`)

- By default, ``read_csv`` and ``read_table`` will now try to infer the compression type based on the file extension. Set ``compression=None`` to restore the previous behavior (no decompression). (:issue:`9770`)

.. _whatsnew_0161.deprecations:

Deprecations
^^^^^^^^^^^^

- ``Series.str.split``'s ``return_type`` keyword was removed in favor of ``expand`` (:issue:`9847`)

.. _whatsnew_0161.performance:

Performance Improvements
~~~~~~~~~~~~~~~~~~~~~~~~

- Improved csv write performance with mixed dtypes, including datetimes by up to 5x (:issue:`9940`)
- Improved csv write performance generally by 2x (:issue:`9940`)
- Improved the performance of ``pd.lib.max_len_string_array`` by 5-7x (:issue:`10024`)



.. _whatsnew_0161.bug_fixes:

Bug Fixes
~~~~~~~~~

- Bug where labels did not appear properly in the legend of ``DataFrame.plot()``, passing ``label=`` arguments works, and Series indices are no longer mutated. (:issue:`9542`)
- Bug in json serialization causing a segfault when a frame had zero length. (:issue:`9805`)
- Bug in ``read_csv`` where missing trailing delimiters would cause segfault. (:issue:`5664`)
- Bug in retaining index name on appending (:issue:`9862`)
- Bug in ``scatter_matrix`` draws unexpected axis ticklabels (:issue:`5662`)
- Fixed bug in ``StataWriter`` resulting in changes to input ``DataFrame`` upon save (:issue:`9795`).
- Bug in ``transform`` causing length mismatch when null entries were present and a fast aggregator was being used (:issue:`9697`)
- Bug in ``equals`` causing false negatives when block order differed (:issue:`9330`)
- Bug in grouping with multiple ``pd.Grouper`` where one is non-time based (:issue:`10063`)
- Bug in ``read_sql_table`` error when reading postgres table with timezone (:issue:`7139`)
- Bug in ``DataFrame`` slicing may not retain metadata (:issue:`9776`)
- Bug where ``TimdeltaIndex`` were not properly serialized in fixed ``HDFStore`` (:issue:`9635`)
- Bug with ``TimedeltaIndex`` constructor ignoring ``name`` when given another ``TimedeltaIndex`` as data (:issue:`10025`).
- Bug in ``DataFrameFormatter._get_formatted_index`` with not applying ``max_colwidth`` to the ``DataFrame`` index (:issue:`7856`)
- Bug in ``.loc`` with a read-only ndarray data source (:issue:`10043`)
- Bug in ``groupby.apply()`` that would raise if a passed user defined function either returned only ``None`` (for all input). (:issue:`9685`)
- Always use temporary files in pytables tests (:issue:`9992`)
- Bug in plotting continuously using ``secondary_y`` may not show legend properly. (:issue:`9610`, :issue:`9779`)
- Bug in ``DataFrame.plot(kind="hist")`` results in ``TypeError`` when ``DataFrame`` contains non-numeric columns  (:issue:`9853`)
- Bug where repeated plotting of ``DataFrame`` with a ``DatetimeIndex`` may raise ``TypeError`` (:issue:`9852`)
- Bug in ``setup.py`` that would allow an incompat cython version to build (:issue:`9827`)
- Bug in plotting ``secondary_y`` incorrectly attaches ``right_ax`` property to secondary axes specifying itself recursively. (:issue:`9861`)

- Bug in ``Series.quantile`` on empty Series of type ``Datetime`` or ``Timedelta`` (:issue:`9675`)
- Bug in ``where`` causing incorrect results when upcasting was required (:issue:`9731`)
- Bug in ``FloatArrayFormatter`` where decision boundary for displaying "small" floats in decimal format is off by one order of magnitude for a given display.precision (:issue:`9764`)
- Fixed bug where ``DataFrame.plot()`` raised an error when both ``color`` and ``style`` keywords were passed and there was no color symbol in the style strings (:issue:`9671`)
- Not showing a ``DeprecationWarning`` on combining list-likes with an ``Index`` (:issue:`10083`)
- Bug in ``read_csv`` and ``read_table`` when using ``skip_rows`` parameter if blank lines are present. (:issue:`9832`)
- Bug in ``read_csv()`` interprets ``index_col=True`` as ``1`` (:issue:`9798`)
- Bug in index equality comparisons using ``==`` failing on Index/MultiIndex type incompatibility (:issue:`9785`)
- Bug in which ``SparseDataFrame`` could not take `nan` as a column name (:issue:`8822`)
- Bug in ``to_msgpack`` and ``read_msgpack`` zlib and blosc compression support (:issue:`9783`)

- Bug ``GroupBy.size`` doesn't attach index name properly if grouped by ``TimeGrouper`` (:issue:`9925`)
- Bug causing an exception in slice assignments because ``length_of_indexer`` returns wrong results (:issue:`9995`)
- Bug in csv parser causing lines with initial whitespace plus one non-space character to be skipped. (:issue:`9710`)
- Bug in C csv parser causing spurious NaNs when data started with newline followed by whitespace. (:issue:`10022`)

- Bug causing elements with a null group to spill into the final group when grouping by a ``Categorical`` (:issue:`9603`)
- Bug where .iloc and .loc behavior is not consistent on empty dataframes (:issue:`9964`)

- Bug in invalid attribute access on a ``TimedeltaIndex`` incorrectly raised ``ValueError`` instead of ``AttributeError`` (:issue:`9680`)




- Bug in unequal comparisons between categorical data and a scalar, which was not in the categories (e.g. ``Series(Categorical(list("abc"), ordered=True)) > "d"``. This returned ``False`` for all elements, but now raises a ``TypeError``. Equality comparisons also now return ``False`` for ``==`` and ``True`` for ``!=``. (:issue:`9848`)
- Bug in DataFrame ``__setitem__`` when right hand side is a dictionary (:issue:`9874`)
- Bug in ``where`` when dtype is ``datetime64/timedelta64``, but dtype of other is not (:issue:`9804`)
- Bug in ``MultiIndex.sortlevel()`` results in unicode level name breaks (:issue:`9856`)
- Bug in which ``groupby.transform`` incorrectly enforced output dtypes to match input dtypes. (:issue:`9807`)
- Bug in ``DataFrame`` constructor when ``columns`` parameter is set, and ``data`` is an empty list (:issue:`9939`)
- Bug in bar plot with ``log=True`` raises ``TypeError`` if all values are less than 1 (:issue:`9905`)
- Bug in horizontal bar plot ignores ``log=True`` (:issue:`9905`)



- Bug in PyTables queries that did not return proper results using the index (:issue:`8265`, :issue:`9676`)




- Bug where dividing a dataframe containing values of type ``Decimal`` by another ``Decimal`` would raise. (:issue:`9787`)
- Bug where using DataFrames asfreq would remove the name of the index. (:issue:`9885`)
- Bug causing extra index point when resample BM/BQ (:issue:`9756`)
- Changed caching in ``AbstractHolidayCalendar`` to be at the instance level rather than at the class level as the latter can result in unexpected behaviour. (:issue:`9552`)

- Fixed latex output for multi-indexed dataframes (:issue:`9778`)
- Bug causing an exception when setting an empty range using ``DataFrame.loc`` (:issue:`9596`)




- Bug in hiding ticklabels with subplots and shared axes when adding a new plot to an existing grid of axes (:issue:`9158`)
- Bug in ``transform`` and ``filter`` when grouping on a categorical variable (:issue:`9921`)
- Bug in ``transform`` when groups are equal in number and dtype to the input index (:issue:`9700`)
- Google BigQuery connector now imports dependencies on a per-method basis.(:issue:`9713`)
- Updated BigQuery connector to no longer use deprecated ``oauth2client.tools.run()`` (:issue:`8327`)
- Bug in subclassed ``DataFrame``. It may not return the correct class, when slicing or subsetting it. (:issue:`9632`)
- Bug in ``.median()`` where non-float null values are not handled correctly (:issue:`10040`)
- Bug in Series.fillna() where it raises if a numerically convertible string is given (:issue:`10092`)