matplotlib · NelleV · Jan 13, 2017 · Dec 5, 2016
diff --git a/doc/api/api_changes/2016-12-19-ESDA_sample_data.rst b/doc/api/api_changes/2016-12-19-ESDA_sample_data.rst
@@ -0,0 +1,13 @@
+Cleanup of stock sample data
+````````````````````````````
+
+The sample data of stocks has been cleaned up to remove redundancies and
+increase portability. The ``AAPL.dat.gz``, ``INTC.dat.gz`` and ``aapl.csv``
+files have been removed entirely and will also no longer be available from
+`matplotlib.cbook.get_sample_data`. If a CSV file is required, we suggest using
+the ``msft.csv`` that continues to be shipped in the sample data. If a NumPy
+binary file is acceptable, we suggest using one of the following two new files.
+The ``aapl.npy.gz`` and ``goog.npy`` files have been replaced by ``aapl.npz``
+and ``goog.npz``, wherein the first column's type has changed from
+`datetime.date` to `np.datetime64` for better portability across Python
+versions. Note that matplotlib does not fully support `np.datetime64` as yet.
diff --git a/doc/users/recipes.rst b/doc/users/recipes.rst
@@ -87,28 +87,38 @@ gracefully, and here are some tricks to help you work around them.
 We'll load up some sample date data which contains datetime.date
 objects in a numpy record array::
 
-  In [63]: datafile = cbook.get_sample_data('goog.npy')
+  In [63]: datafile = cbook.get_sample_data('goog.npz')
 
-  In [64]: r = np.load(datafile).view(np.recarray)
+  In [64]: r = np.load(datafile)['price_data'].view(np.recarray)
 
   In [65]: r.dtype
-  Out[65]: dtype([('date', '|O4'), ('', '|V4'), ('open', '<f8'),
+  Out[65]: dtype([('date', '<M8[D]'), ('', '|V4'), ('open', '<f8'),
                   ('high', '<f8'), ('low', '<f8'), ('close', '<f8'),
                   ('volume', '<i8'),  ('adj_close', '<f8')])
 
   In [66]: r.date
   Out[66]:
-  array([2004-08-19, 2004-08-20, 2004-08-23, ..., 2008-10-10, 2008-10-13,
-         2008-10-14], dtype=object)
+  array(['2004-08-19', '2004-08-20', '2004-08-23', ..., '2008-10-10',
+         '2008-10-13', '2008-10-14'], dtype='datetime64[D]')
 
-The dtype of the numpy record array for the field ``date`` is ``|O4``
-which means it is a 4-byte python object pointer; in this case the
-objects are datetime.date instances, which we can see when we print
-some samples in the ipython terminal window.
+The dtype of the NumPy record array for the field ``date`` is ``datetime64[D]``
+which means it is a 64-bit `np.datetime64` in 'day' units. While this format is
+more portable, Matplotlib cannot plot this format natively yet. We can plot
+this data by changing the dates to `datetime.date` instances instead, which can
+be achieved by converting to an object array::
+
+  In [67]: r.date.astype('O')
+  array([datetime.date(2004, 8, 19), datetime.date(2004, 8, 20),
+         datetime.date(2004, 8, 23), ..., datetime.date(2008, 10, 10),
+         datetime.date(2008, 10, 13), datetime.date(2008, 10, 14)],
+        dtype=object)
+
+The dtype of this converted array is now ``object`` and it is filled with
+datetime.date instances instead.
 
 If you plot the data, ::
 
-  In [67]: plot(r.date, r.close)
+  In [67]: plot(r.date.astype('O'), r.close)
   Out[67]: [<matplotlib.lines.Line2D object at 0x92a6b6c>]
 
 you will see that the x tick labels are all squashed together.
@@ -117,18 +127,12 @@ you will see that the x tick labels are all squashed together.
    :context:
 
    import matplotlib.cbook as cbook
-   datafile = cbook.get_sample_data('goog.npy')
-   try:
-       # Python3 cannot load python2 .npy files with datetime(object) arrays
-       # unless the encoding is set to bytes. Hovever this option was
-       # not added until numpy 1.10 so this example will only work with
-       # python 2 or with numpy 1.10 and later.
-       r = np.load(datafile, encoding='bytes').view(np.recarray)
-   except TypeError:
-       # Old Numpy
-       r = np.load(datafile).view(np.recarray)
+   with cbook.get_sample_data('goog.npz') as datafile:
+       r = np.load(datafile)['price_data'].view(np.recarray)
+   # Matplotlib prefers datetime instead of np.datetime64.
+   date = r.date.astype('O')
    plt.figure()
-   plt.plot(r.date, r.close)
+   plt.plot(date, r.close)
    plt.title('Default date handling can cause overlapping labels')
 
 Another annoyance is that if you hover the mouse over the window and
@@ -149,7 +153,7 @@ a number of date formatters built in, so we'll use one of those.
 
    plt.close('all')
    fig, ax = plt.subplots(1)
-   ax.plot(r.date, r.close)
+   ax.plot(date, r.close)
 
    # rotate and align the tick labels so they look better
    fig.autofmt_xdate()
@@ -186,22 +190,17 @@ right.
    import matplotlib.cbook as cbook
 
    # load up some sample financial data
-   datafile = cbook.get_sample_data('goog.npy')
-   try:
-       # Python3 cannot load python2 .npy files with datetime(object) arrays
-       # unless the encoding is set to bytes. Hovever this option was
-       # not added until numpy 1.10 so this example will only work with
-       # python 2 or with numpy 1.10 and later.
-       r = np.load(datafile, encoding='bytes').view(np.recarray)
-   except TypeError:
-       r = np.load(datafile).view(np.recarray)
+   with cbook.get_sample_data('goog.npz') as datafile:
+       r = np.load(datafile)['price_data'].view(np.recarray)
+   # Matplotlib prefers datetime instead of np.datetime64.
+   date = r.date.astype('O')
    # create two subplots with the shared x and y axes
    fig, (ax1, ax2) = plt.subplots(1,2, sharex=True, sharey=True)
 
    pricemin = r.close.min()
 
-   ax1.plot(r.date, r.close, lw=2)
-   ax2.fill_between(r.date, pricemin, r.close, facecolor='blue', alpha=0.5)
+   ax1.plot(date, r.close, lw=2)
+   ax2.fill_between(date, pricemin, r.close, facecolor='blue', alpha=0.5)
 
    for ax in ax1, ax2:
        ax.grid(True)

diff --git a/examples/api/date_demo.py b/examples/api/date_demo.py
@@ -24,31 +24,26 @@
 months = mdates.MonthLocator()  # every month
 yearsFmt = mdates.DateFormatter('%Y')
 
-# load a numpy record array from yahoo csv data with fields date,
-# open, close, volume, adj_close from the mpl-data/example directory.
-# The record array stores python datetime.date as an object array in
-# the date column
-datafile = cbook.get_sample_data('goog.npy')
-try:
-    # Python3 cannot load python2 .npy files with datetime(object) arrays
-    # unless the encoding is set to bytes. However this option was
-    # not added until numpy 1.10 so this example will only work with
-    # python 2 or with numpy 1.10 and later.
-    r = np.load(datafile, encoding='bytes').view(np.recarray)
-except TypeError:
-    r = np.load(datafile).view(np.recarray)
+# Load a numpy record array from yahoo csv data with fields date, open, close,
+# volume, adj_close from the mpl-data/example directory. The record array
+# stores the date as an np.datetime64 with a day unit ('D') in the date column.
+with cbook.get_sample_data('goog.npz') as datafile:
+    r = np.load(datafile)['price_data'].view(np.recarray)
+# Matplotlib works better with datetime.datetime than np.datetime64, but the
+# latter is more portable.
+date = r.date.astype('O')
 
 fig, ax = plt.subplots()
-ax.plot(r.date, r.adj_close)
+ax.plot(date, r.adj_close)
 
 
 # format the ticks
 ax.xaxis.set_major_locator(years)
 ax.xaxis.set_major_formatter(yearsFmt)
 ax.xaxis.set_minor_locator(months)
 
-datemin = datetime.date(r.date.min().year, 1, 1)
-datemax = datetime.date(r.date.max().year + 1, 1, 1)
+datemin = datetime.date(date.min().year, 1, 1)
+datemax = datetime.date(date.max().year + 1, 1, 1)
 ax.set_xlim(datemin, datemax)
 
 

diff --git a/examples/api/date_index_formatter.py b/examples/api/date_index_formatter.py
@@ -10,22 +10,23 @@
 from __future__ import print_function
 import numpy as np
 import matplotlib.pyplot as plt
-import matplotlib.mlab as mlab
 import matplotlib.cbook as cbook
 import matplotlib.ticker as ticker
 
-datafile = cbook.get_sample_data('aapl.csv', asfileobj=False)
-print('loading %s' % datafile)
-r = mlab.csv2rec(datafile)
-
-r.sort()
+# Load a numpy record array from yahoo csv data with fields date, open, close,
+# volume, adj_close from the mpl-data/example directory. The record array
+# stores the date as an np.datetime64 with a day unit ('D') in the date column.
+with cbook.get_sample_data('goog.npz') as datafile:
+    r = np.load(datafile)['price_data'].view(np.recarray)
 r = r[-30:]  # get the last 30 days
-
+# Matplotlib works better with datetime.datetime than np.datetime64, but the
+# latter is more portable.
+date = r.date.astype('O')
 
 # first we'll do it the default way, with gaps on weekends
 fig, axes = plt.subplots(ncols=2, figsize=(8, 4))
 ax = axes[0]
-ax.plot(r.date, r.adj_close, 'o-')
+ax.plot(date, r.adj_close, 'o-')
 ax.set_title("Default")
 fig.autofmt_xdate()
 
@@ -36,7 +37,7 @@
 
 def format_date(x, pos=None):
     thisind = np.clip(int(x + 0.5), 0, N - 1)
-    return r.date[thisind].strftime('%Y-%m-%d')
+    return date[thisind].strftime('%Y-%m-%d')
 
 ax = axes[1]
 ax.plot(ind, r.adj_close, 'o-')

diff --git a/examples/misc/rec_groupby_demo.py b/examples/misc/rec_groupby_demo.py
@@ -3,7 +3,7 @@
 import matplotlib.mlab as mlab
 import matplotlib.cbook as cbook
 
-datafile = cbook.get_sample_data('aapl.csv', asfileobj=False)
+datafile = cbook.get_sample_data('msft.csv', asfileobj=False)
 print('loading', datafile)
 r = mlab.csv2rec(datafile)
 r.sort()

diff --git a/examples/misc/rec_join_demo.py b/examples/misc/rec_join_demo.py
@@ -3,7 +3,7 @@
 import matplotlib.mlab as mlab
 import matplotlib.cbook as cbook
 
-datafile = cbook.get_sample_data('aapl.csv', asfileobj=False)
+datafile = cbook.get_sample_data('msft.csv', asfileobj=False)
 print('loading', datafile)
 r = mlab.csv2rec(datafile)
 

diff --git a/examples/pylab_examples/centered_ticklabels.py b/examples/pylab_examples/centered_ticklabels.py
@@ -19,20 +19,15 @@
 import matplotlib.pyplot as plt
 
 # load some financial data; apple's stock price
-fh = cbook.get_sample_data('aapl.npy.gz')
-try:
-    # Python3 cannot load python2 .npy files with datetime(object) arrays
-    # unless the encoding is set to bytes. However this option was
-    # not added until numpy 1.10 so this example will only work with
-    # python 2 or with numpy 1.10 and later.
-    r = np.load(fh, encoding='bytes')
-except TypeError:
-    r = np.load(fh)
-fh.close()
+with cbook.get_sample_data('aapl.npz') as fh:
+    r = np.load(fh)['price_data'].view(np.recarray)
 r = r[-250:]  # get the last 250 days
+# Matplotlib works better with datetime.datetime than np.datetime64, but the
+# latter is more portable.
+date = r.date.astype('O')
 
 fig, ax = plt.subplots()
-ax.plot(r.date, r.adj_close)
+ax.plot(date, r.adj_close)
 
 ax.xaxis.set_major_locator(dates.MonthLocator())
 ax.xaxis.set_minor_locator(dates.MonthLocator(bymonthday=15))
@@ -46,5 +41,5 @@
     tick.label1.set_horizontalalignment('center')
 
 imid = len(r)//2
-ax.set_xlabel(str(r.date[imid].year))
+ax.set_xlabel(str(date[imid].year))
 plt.show()
diff --git a/examples/pylab_examples/scatter_demo2.py b/examples/pylab_examples/scatter_demo2.py
@@ -5,19 +5,11 @@
 import matplotlib.pyplot as plt
 import matplotlib.cbook as cbook
 
-# Load a numpy record array from yahoo csv data with fields date,
-# open, close, volume, adj_close from the mpl-data/example directory.
-# The record array stores python datetime.date as an object array in
-# the date column
-datafile = cbook.get_sample_data('goog.npy')
-try:
-    # Python3 cannot load python2 .npy files with datetime(object) arrays
-    # unless the encoding is set to bytes. However this option was
-    # not added until numpy 1.10 so this example will only work with
-    # python 2 or with numpy 1.10 and later
-    price_data = np.load(datafile, encoding='bytes').view(np.recarray)
-except TypeError:
-    price_data = np.load(datafile).view(np.recarray)
+# Load a numpy record array from yahoo csv data with fields date, open, close,
+# volume, adj_close from the mpl-data/example directory. The record array
+# stores the date as an np.datetime64 with a day unit ('D') in the date column.
+with cbook.get_sample_data('goog.npz') as datafile:
+    price_data = np.load(datafile)['price_data'].view(np.recarray)
 price_data = price_data[-250:]  # get the most recent 250 trading days
 
 delta1 = np.diff(price_data.adj_close)/price_data.adj_close[:-1]

diff --git a/lib/matplotlib/mpl-data/sample_data/AAPL.dat.gz b/lib/matplotlib/mpl-data/sample_data/AAPL.dat.gz
diff --git a/lib/matplotlib/mpl-data/sample_data/INTC.dat.gz b/lib/matplotlib/mpl-data/sample_data/INTC.dat.gz